Browse the corpus

Summary box What is already known about this subject? ▸  Factors affecting adenomagenesis and carcinogenesis are thought to be different. Previous work has shown that the relative risks of each vary between populations. Patients with a history of adenoma are at greater risk of metachronous adenoma, which may support the existence of colorectal fields. To date, little is known about the molecular basis for such fields. ▸  Keratins are a type of intermediate filament proteins, which as part of cellular cytoskeleton have important regulatory functions on the colonic mucosa. K8 null mice develop colitis and K8 is shown to modulate tumour necrosis factor (TNF) action. ▸  Butyrate is widely hypothesised as being a key chemoprotective molecule, through effects on cellular programming and cell fate determination and more recently through regulation of metabolic phenotype of the cell. What are the new findings? ▸  Validated proteomic analysis of the macroscopically normal mucosa in the colon of participants with an adenoma was compared to tissue from the adenoma and from participants free of pathology. The analysis indicated that several groups of proteins were altered, but particularly proteins of the keratin family. ▸  These data suggest that there may be field-wide changes in the colon associating with the presence of a lesion, and contribute to the molecular evidence for existence of fields. ▸  These changes may be restored to a normal phenotype (for keratins) by resection of lesions and by dietary fibre interventions.

What are the new findings? ▸  Validated proteomic analysis of the macroscopically normal mucosa in the colon of participants with an adenoma was compared to tissue from the adenoma and from participants free of pathology. The analysis indicated that several groups of proteins were altered, but particularly proteins of the keratin family. ▸  These data suggest that there may be field-wide changes in the colon associating with the presence of a lesion, and contribute to the molecular evidence for existence of fields. ▸  These changes may be restored to a normal phenotype (for keratins) by resection of lesions and by dietary fibre interventions. How might it impact on clinical practice in the foreseeable future? ▸  The data provide evidence for the existence of colorectal fields. As such this introduces the possibility of using markers (eg, the levels, distributions and form of keratin) as biomarkers of fields. Critically this work suggests that active interventions may modulate fields, implying they are regressible. Both resection and increased intake of fibre rich food both restored normality to keratin expression. As such secondary chemopreventive strategies have been proven in principle and offer the opportunity to give lifestyle advice to modulate risk of metachronous disease.

ay modulate fields, implying they are regressible. Both resection and increased intake of fibre rich food both restored normality to keratin expression. As such secondary chemopreventive strategies have been proven in principle and offer the opportunity to give lifestyle advice to modulate risk of metachronous disease. Introduction Metachronous adenoma (AD) is a greater risk than incident AD,1 an observation that has been used in support of a theory of colonic field effects. The existence of field effects around lesions or giving rise to lesions was proposed over 50 years ago.2 In the case of the colon, there is debate on whether the immediate region around the lesion or the entire colon represents the field. We have recently summarised the evidence and proposed a series of qualitative models for metachronous AD.3 Direct molecular evidence of field effects in the colon has remained elusive but a gel-based proteomic analysis by Polley et al4 identified proteins altered in the vicinity of ADs, including keratin 8. This study supported fields being local to the lesion, but did not examine sites distant to the lesion. Additional indirect evidence for crypt expansion to fields comes from measuring extent of hyperploidy and crypt expansion in patients with colitis.5 Nonetheless, direct evidence for molecular changes in colorectal fields remains limited.

fields being local to the lesion, but did not examine sites distant to the lesion. Additional indirect evidence for crypt expansion to fields comes from measuring extent of hyperploidy and crypt expansion in patients with colitis.5 Nonetheless, direct evidence for molecular changes in colorectal fields remains limited. Keratins are key components of intermediate filaments, a cytoskeletal structure responsible for the structural integrity of the epithelium through cell-cell contacts, cell shear stress and through regulation of signal transduction and cell polarisation. Keratin expression occurs in specific combinations, dependent on the type of epithelium, and differentiation status related to the epithelial type and stage of cellular differentiation.6 K8, K18 and K19 are the predominantly expressed in colonic epithelium.7 Keratin 8 (K8) forms a heterodimer with keratin 18 (K18) in the colonocyte. Several lines of evidence suggest a role for K8 in maintenance of a functional colorectal epithelium. Alteration in K8 patterns occur around colorectal neoplasia4; polymerisation-inhibiting mutations in K8 were observed in a subset of inflammatory bowel disease (IBD) patients8; transgenic mice lacking K8 exhibit hyperplasia in the colon and impaired barrier and absorptive function in addition to developing colorectal neoplasia.9 The literature on associations between keratin expression, function and colorectal disease has recently been reviewed.7

of inflammatory bowel disease (IBD) patients8; transgenic mice lacking K8 exhibit hyperplasia in the colon and impaired barrier and absorptive function in addition to developing colorectal neoplasia.9 The literature on associations between keratin expression, function and colorectal disease has recently been reviewed.7 Keratins (and other proteins in cancer pathways) are subject to modification by acetylation.10 In the case of keratin we have shown that acetylation may drive transition between different protein states (soluble/insoluble).11 12 There is substantial evidence from epidemiological studies to suggest that high dietary fibre intakes may protect against colorectal cancer.13 14 Mechanisms may include reduction in transit time, increase in stool volume, both of which would reduce exposure to carcinogens, and butyrate production. Butyrate, produced by bacteria-mediated fermentation regulates cell cycle and apoptosis in vitro,15–17 and is chemopreventive in rat models of colorectal carcinogenesis.18 These properties are attributed to butyrate's molecular action as an inhibitor of histone deacetylation which results in increased protein acetylation.19 Several microarray-based studies20 21 reveal a substantial portion of the transcriptome is altered in response to butyrate but that protein acetylation is not limited to histones and acetylation is as important in the regulation of protein function as phosphorylation.22

which results in increased protein acetylation.19 Several microarray-based studies20 21 reveal a substantial portion of the transcriptome is altered in response to butyrate but that protein acetylation is not limited to histones and acetylation is as important in the regulation of protein function as phosphorylation.22 Proteomic approaches to the analysis of clinical specimens allow objective and impartial evaluation of the changes occurring between compared samples. Proteomics is a suite of protein separation and characterisation approaches, employing gel and gel-free approaches for protein separation. Two dimensional gel electrophoresis (2DGE) separates proteins by charge and mass, gel-free approaches such as isobaric tags for relative and absolute quantification (iTRAQ) enable quantitative and global profiling in a multiplex format,23 24 with identification of larger numbers of proteins, and the relative quantification of all identifiable proteins, but with the limitation of loss of separation of subspecies of a protein, for example distinction of a phosphorylated and non-phosphorylated species. iTRAQ is an emerging standard for assessment of changes in a global proteome, while 2DGE has utility for assessment of post-translational modifications of smaller numbers of species.10 We applied multiple proteomic workflows to protein extracts from biopsies taken at, near and distant to colorectal neoplasia, stratified by levels of butyrate present in a stool sample to explore the effects of both colorectal fields and of butyrate on the status of epithelial keratins.

Proteomic approaches to the analysis of clinical specimens allow objective and impartial evaluation of the changes occurring between compared samples. Proteomics is a suite of protein separation and characterisation approaches, employing gel and gel-free approaches for protein separation. Two dimensional gel electrophoresis (2DGE) separates proteins by charge and mass, gel-free approaches such as isobaric tags for relative and absolute quantification (iTRAQ) enable quantitative and global profiling in a multiplex format,23 24 with identification of larger numbers of proteins, and the relative quantification of all identifiable proteins, but with the limitation of loss of separation of subspecies of a protein, for example distinction of a phosphorylated and non-phosphorylated species. iTRAQ is an emerging standard for assessment of changes in a global proteome, while 2DGE has utility for assessment of post-translational modifications of smaller numbers of species.10 We applied multiple proteomic workflows to protein extracts from biopsies taken at, near and distant to colorectal neoplasia, stratified by levels of butyrate present in a stool sample to explore the effects of both colorectal fields and of butyrate on the status of epithelial keratins. Methods and materials Participants and recruitment This study includes two arms: a cross-sectional study and a fibre-intervention. The study design has been reported elsewhere.25 Ethics committee approval was obtained from the North Sheffield Research Ethics Committee prior to recruiting (Reference number: 06/Q2308/93).

Methods and materials Participants and recruitment This study includes two arms: a cross-sectional study and a fibre-intervention. The study design has been reported elsewhere.25 Ethics committee approval was obtained from the North Sheffield Research Ethics Committee prior to recruiting (Reference number: 06/Q2308/93). Cross-sectional study (FACT OBS) Participants included in this study were recruited from colonoscopy lists at Sheffield Teaching Hospitals between October 2007 and June 2008. In total 62 participants were included in this analysis, of whom 34 were found to be free from disease during endoscopy (the normal group), while 28 participants were found to have histologically confirmed ADs (the AD group). The normal group had a younger mean age (62.1±11.4 vs 68.1±0.1, p=0.047). There were no significant differences between groups for BMI or weight. In patients with adenomatous polyps biopsies were taken from the AD itself, and from macroscopically normal mucosa on the contralateral (CL) wall to the adenoma CL and from the mid-sigmoid (MS) colon (see online supplementary information (SOI); section 1). Endoscopists judged on macroscopic appearance which polyps were adenomatous, all judgements were independently confirmed by a histopathologist and non-adenomatous samples were removed from the analysis. In lesion-free participants biopsies were taken from the MS. Three biopsies were taken at each site—two for proteomic analysis and one for immunohistochemistry. Participants additionally provided a stool sample (while bowel habit was normal) for assessment of faecal SCFA levels.

samples were removed from the analysis. In lesion-free participants biopsies were taken from the MS. Three biopsies were taken at each site—two for proteomic analysis and one for immunohistochemistry. Participants additionally provided a stool sample (while bowel habit was normal) for assessment of faecal SCFA levels. Intervention study (FACT INT) An initial screening tool for fibre intake was developed (DL and EAW, unpublished) to identify participants with low dietary fibre intake (<12 g/day). The screening tool was in the form of a brief food frequency questionnaire that asked about the consumption and portion size of high fibre breads, (cereals, pasta, rice, vegetables and fruit). Nine participants were included in the intervention trial. Biopsies were taken during scheduled endoscopy from MS only and either flash frozen for protein analysis or formalin fixed for IHC. Diet at baseline was assessed using a 4-day food diary (FD) which confirmed low baseline fibre intake. A food replacement strategy was used to increase fibre intake. The high fibre intervention was 8 weeks in duration and included an initial 2-week period when intake was increased incrementally. Increased fibre consumption was encouraged via supply and weekly delivery of fibre-rich foods (cereals, breads, fruit, vegetables) directly to the participants’ homes. Participants were invited to choose their own foods from a list of high-fibre foods and were offered advice as to how to incorporate more fibre into their diets. Participants were provided with a simple fibre reckoner to assess their own daily fibre intake and were asked progressively to increase fibre intake, to achieve an equivalent of over 20 g/day by the end week of the intervention. FD, endoscopy, biopsy and stool sampling were repeated at the end of the intervention.

ets. Participants were provided with a simple fibre reckoner to assess their own daily fibre intake and were asked progressively to increase fibre intake, to achieve an equivalent of over 20 g/day by the end week of the intervention. FD, endoscopy, biopsy and stool sampling were repeated at the end of the intervention. Biopsy lysis, pooling and fractionation Colorectal pinch biopsies (∼5 mg) were suspended in kinase buffer (50 mM Tris-HCl pH7.5; 10 mM MgCl2; 0.1 mM EDTA; 2 mM DTT) followed by homogenisation. The lysate was centrifuged to provide crude separation between the insoluble and soluble fractions. Resulting fractions were grouped by diagnosis and region and were ranked by faecal butyrate level and then pooled: eight lysates (two each from four patients) were used in each pool. Acetyl proteins were separated from other soluble proteins using antibody immobilised to a Pierce Seize matrix. The insoluble fraction, principally intermediate filaments, was prepared using our integrated workflow for iTRAQ-compatible analysis26 which modified the high-salt extraction technique of Achstaetter27 and Herrmann.28 The insoluble and soluble fractions were analysed using iTRAQ workflows while acetyl proteins were analysed by 2D gel electrophoresis (vide infra). The workflow is summarised in figure 1.

tegrated workflow for iTRAQ-compatible analysis26 which modified the high-salt extraction technique of Achstaetter27 and Herrmann.28 The insoluble and soluble fractions were analysed using iTRAQ workflows while acetyl proteins were analysed by 2D gel electrophoresis (vide infra). The workflow is summarised in figure 1. Figure 1 Experimental workflow. The figure summarises the workflow used in this study. Participants recruited had biopsies taken and provided stool (for faecal butyrate) and diet information. The frozen biopsies were lysed to yield soluble and insoluble fractions. The former was pooled according to butyrate and subject to immunoprecipitation for acetyl proteins, to yield an acetyl-enriched and a soluble fraction. The former was analysed by 2DGE whereas the latter was analysed by iTRAQ. The insoluble fraction was processed to yield IF and then analysed by iTRAQ. iTRAQ, isobaric tags for relative and absolute quantification; 2DGE, two dimensional gel electrophoresis; IF, intermediate filament.

tyl-enriched and a soluble fraction. The former was analysed by 2DGE whereas the latter was analysed by iTRAQ. The insoluble fraction was processed to yield IF and then analysed by iTRAQ. iTRAQ, isobaric tags for relative and absolute quantification; 2DGE, two dimensional gel electrophoresis; IF, intermediate filament. Proteomic methods iTRAQ workflow Proteomes were analysed using an iTRAQ workflow as previously described.29 Briefly proteins were reduced, alkylated and subject to proteolytic digestion using trypsin prior to labelling using the iTRAQ 8 plex kit (ABSciex, Warrington, UK), according to manufacturer's instructions. Peptide fractionation was undertaken using Strong Cation Exchange on a BioLC HPLC system (Dionex, Surrey, UK). Fractions were collected, dried and stored prior for mass spectrometric analysis. Mass spectrometry (MS) was performed using ESI-qQ-TOF-MS/MS platforms (Bruker, Bremen, GmBH) coupled with an online nano-flow liquid chromatography system (U3000, Dionex, Camberley, UK). Protein identifications were obtained using the the Phenyx software platform (GeneBio, Geneva) to perform database searching against the human UniProt (SwissProt, Trembl downloaded 11 May 2010) database. A concatenated target-decoy database search strategy was also employed to estimate the rate of false discovery rate,30 calculated to be 1% and as such were well within the 5% recommended for reporting proteomic data.31

orm database searching against the human UniProt (SwissProt, Trembl downloaded 11 May 2010) database. A concatenated target-decoy database search strategy was also employed to estimate the rate of false discovery rate,30 calculated to be 1% and as such were well within the 5% recommended for reporting proteomic data.31 Mathematical, bioinformatic and statistical analysis Hierarchical clustering and principal component analysis were performed to group the data based on the degree of similarity between samples. Agglomerative clustering using the squared Euclidean distance between log10 iTRAQ ratios and smallest inter-cluster dissimilarity linkage procedure was performed (Mathematica 7.0.0 for Mac). Statistically significant changes in protein level were identified using our t test algorithm.32 Pathway analysis was undertaken using the Instance Browser in Reactome.33 Protein interaction networks were analysed using STRING V. 9.0.34 Statistical analysis of immunoblot densitometry data and immunohistochemical data was carried out in Microsoft Excel and SPSS.

n level were identified using our t test algorithm.32 Pathway analysis was undertaken using the Instance Browser in Reactome.33 Protein interaction networks were analysed using STRING V. 9.0.34 Statistical analysis of immunoblot densitometry data and immunohistochemical data was carried out in Microsoft Excel and SPSS. Immunoblotting Proteins were separated by SDS-PAGE and transferred to PVDF. Membranes were incubated with primary antibody solutions in blocking buffer. Primary antibodies used include: keratin 19 (mAb3238, Millipore, UK); ApoA1(Ab48647, Abcam, UK), M2PK (Ab38327, Abcam, UK); GAPDH (AM4300, Ambion); and α-tubulin (Ab7792, Abcam). Cross-reactions were visualised using HRP-conjugated secondary antibodies (Dako), Immobilon Western HRP substrate (Millipore, UK). A Chemigenius Bioimaging system was employed for band visualisation and densitometric analysis. Immunohistochemistry for keratin 8 Our previously established protocols for keratin 8 immunohistochemistry and scoring were used.35 Antigen retrieval was performed on 4-micron formalin-fixed, paraffin-embedded sections with EDTA (1 mM, pH8) in microwave at high power for 8 min. Sections were incubated with primary antibody (K8 mouse monoclonal, ab9023) and biotinylated secondary antibody (anti mouse IgG, RTU Vectastain Universal). Detection was performed using DAB kit (sk-4100Vector lab, UK).

ormalin-fixed, paraffin-embedded sections with EDTA (1 mM, pH8) in microwave at high power for 8 min. Sections were incubated with primary antibody (K8 mouse monoclonal, ab9023) and biotinylated secondary antibody (anti mouse IgG, RTU Vectastain Universal). Detection was performed using DAB kit (sk-4100Vector lab, UK). Results Recruitment and participant demographics The inclusion, exclusion criteria, justification and study design have previously been reported.36 Of 81 male participants attending for diagnostic colonoscopy were recruited, 19 were excluded from pooling by butyrate owing to a diagnosis of cancer (n=11), failure to provide a stool sample (n=3), or bowel preparation other than Kleanprep (n=5). Biopsies from participants prepared for endoscopy with Picolax (n=5) were pooled separately as bowel preparation influences proliferation of colon epithelial cells.37 Colonoscopy was completed to the caecal pole in all participants. Adenomatous polyps (confirmed histologically) were present in 28 participants and macroscopically normal colonic mucosa in the remaining 34. The morphometric and AD positional data are shown in the SOI, section 4.

oliferation of colon epithelial cells.37 Colonoscopy was completed to the caecal pole in all participants. Adenomatous polyps (confirmed histologically) were present in 28 participants and macroscopically normal colonic mucosa in the remaining 34. The morphometric and AD positional data are shown in the SOI, section 4. Pooling for proteomics was undertaken following stratification by faecal butyrate concentration. Stool samples were extracted and butyrate determined as previously described.35 38 There was no significant difference in the normal and AD subject groups’ highest (15.5 mM vs 13.0 mM) and lowest (0.9 vs 0.8 mM) mean butyrate levels (SOI, section 4). Individuals were stratified by faecal butyrate, and samples grouped by diagnosis/biopsy location (normal MS, AD MS, CL and lesion AD). Workflow The workflow for this study is set out in figure 1. The insoluble fraction was processed to yield intermediate filament (IF)-enriched material.26 The soluble fractions were pooled in fours (empirically determined to yield analysable acetyl-proteins by IP, data not shown). Samples from the highest and lowest butyrate pools from normal participants, and AD participants at each of the three biopsy sites (a total of eight pools) were immunoprecipitated to yield a pooled IP eluate (acetyl proteins), and a pooled flow-through (other soluble proteins).

e acetyl-proteins by IP, data not shown). Samples from the highest and lowest butyrate pools from normal participants, and AD participants at each of the three biopsy sites (a total of eight pools) were immunoprecipitated to yield a pooled IP eluate (acetyl proteins), and a pooled flow-through (other soluble proteins). For proteomic analysis, the insoluble and soluble pools were analysed by 8-plex iTRAQ workflow; the IP pools were analysed by 2DGE. IP pools were indicative of a profound effect of bowel preparation on protein modification (SOI, section 15) and this analysis was not pursued further. Orthogonal validation was provided at two tiers: (1) immunoblot analysis of the pooled protein extracts analysed by iTRAQ and (2) IHC on archived FFPE biopsy material for individual participants from the study.

ect of bowel preparation on protein modification (SOI, section 15) and this analysis was not pursued further. Orthogonal validation was provided at two tiers: (1) immunoblot analysis of the pooled protein extracts analysed by iTRAQ and (2) IHC on archived FFPE biopsy material for individual participants from the study. The insoluble proteome shows lesion-associated changes in the keratin profile A small number of studies have converged on keratins as hallmarks of colorectal mucosal health7 we undertook a proteomic analysis of the insoluble fraction which is enriched for intermediate filament proteins, including keratins.26 Samples grouped by diagnosis/biopsy location (normal MS, AD MS, CL and lesion AD) and were stratified by low (∼1mM) and high (∼14 mM) butyrate (see above and SOI for concentrations). The iTRAQ workflow identified 55 proteins were represented by ≥2 peptides (complete list: SOI, section 5). Intermediate filament proteins, including keratins were present. Principal component analysis (PCA) (figure 2Ai) of the proteins and associated iTRAQ ratios across the data set, indicated that the samples clustered clearly according to butyrate status, with a single axis (component 2) distinguishing low from high butyrate. The data were further analysed using a Euclidian agglomerative clustering approach, the data suggested that the high-butyrate group were more alike than the low-butyrate group. Interacting protein networks were identified with STRINGS DB (figure 2B and SOI, section 6). Intriguingly, when all lines of evidence are considered the entire intermediate filament proteome forms a single interaction network, with clustering around the keratin-vimentin and collagen centres in additional to a metabolism cluster. A pathways analysis using Reactome (SOI, section 7), indicated that collagen formation and extracellular matrix organisation were well represented (p=2.8×10−8 and 1.8×10−8, respectively). The pathway for NCAM signalling in neurite outgrowth was very highly represented (p=1.2×10−10). The data are consistent with the proposal that IF networks form scaffolding on which associated proteins organise and are regulated to control metabolic activities.39

well represented (p=2.8×10−8 and 1.8×10−8, respectively). The pathway for NCAM signalling in neurite outgrowth was very highly represented (p=1.2×10−10). The data are consistent with the proposal that IF networks form scaffolding on which associated proteins organise and are regulated to control metabolic activities.39 Figure 2 iTRAQ analysis of the insoluble proteome shows effects of lesional proximity and butyrate level. Samples were separated and relatively quantified by 8-plex iTRAQ. Global analysis of the data was undertaken by principal component analysis (PCA) (panel A) and hierarchical clustering analysis (HCA) (panel B). PCA showed clustering by butyrate level with low butyrate samples in the blue oval and high butyrate in the orange oval. HCA showed the high-butyrate samples were more alike than the other samples. A protein interaction network was generated from the whole dataset (orphan nodes not shown) indicating proteins interlinked with clusters around extracellular matrix (solid line), keratins (dashed line) and metabolism (dotted line). Significant differences between samples according to lesional proximity and controlling for butyrate were computed and are shown in panel C. Proteins listed in red are significantly downregulated, while those in green are upregulated. iTRAQ, isobaric tags for relative and absolute quantification.

abolism (dotted line). Significant differences between samples according to lesional proximity and controlling for butyrate were computed and are shown in panel C. Proteins listed in red are significantly downregulated, while those in green are upregulated. iTRAQ, isobaric tags for relative and absolute quantification. We hypothesised that the alterations in insoluble keratins observed associated with lesions or butyrate level could either reflect a reduction in total keratin in the tissue, or could represent a shift in protein space into the soluble phase. We therefore undertook undirected and directed analyses of the soluble fractions to establish whether changes in keratin were reflected across the fractions (vide infra). Parallel analysis of the soluble fraction reveals observed changes in keratin are at protein level, not a change in protein solubility.

We hypothesised that the alterations in insoluble keratins observed associated with lesions or butyrate level could either reflect a reduction in total keratin in the tissue, or could represent a shift in protein space into the soluble phase. We therefore undertook undirected and directed analyses of the soluble fractions to establish whether changes in keratin were reflected across the fractions (vide infra). Parallel analysis of the soluble fraction reveals observed changes in keratin are at protein level, not a change in protein solubility. SDS PAGE analysis of the soluble fraction indicated that the proteins remained intact without degradation or smearing and that no part of the mass range was distorted (SOI section 8). The soluble proteome was analysed by iTRAQ (SOI section 9). Analysis using Instance Browser showed that no pathways were particularly enriched (SOI, section 10). PCA (figure 3Ai) shows that the macroscopically normal tissues are much more similar than the lesional tissue. These samples may be further separated along linear axes. The dotted axis suggests that one component alone can distinguish the effects of butyrate on tissue. The solid axis demonstrates separability of macroscopically normal tissue according to the presence or absence of a lesion. Hierarchical clustering analysis (HCA) suggested again that lesional tissue is most divergent, but that within macroscopically normal tissue, the effect of butyrate is greater than the effect of lesional proximity (figure 3Aii). Protein interaction networks were built as described above. A large network accounted for many of the proteins, with several smaller clusters (SOI section 11). Significant changes were analysed either by controlling for the butyrate status and assessing the impact of lesion on the proteome (figure 3C) or by controlling for the lesional status and assessing the impact of butyrate on the proteome (figure 3D). For clarity on cytoskeletal and related proteins are shown, versions with complete listings appear in the SOI (SOI, section 11). Venns show proteins significantly altered at each site by high or low butyrate.

3C) or by controlling for the lesional status and assessing the impact of butyrate on the proteome (figure 3D). For clarity on cytoskeletal and related proteins are shown, versions with complete listings appear in the SOI (SOI, section 11). Venns show proteins significantly altered at each site by high or low butyrate. Figure 3 iTRAQ analysis of the global soluble proteome shows effects of butyrate and presence of an adenoma. Samples were separated and relatively quantified by 8-plex iTRAQ. Global analysis of the data was undertaken by principal component analysis (PCA) (panel Ai) and hierarchical clustering analysis (HCA) (panel Aii). PCA showed macroscopically normal samples could be separated by a single factor to distinguish high from low butyrate (dashed axis) and a straight line (function of two factors) could separate normal from lesion-associated samples. Panel B hierarchical clustering was used to group the data based on the degree of similarity between the samples analysed using the complete iTRAQ data set. Quantitative data were used to identify significant differences between the soluble proteome samples according to lesional proximity and controlling for butyrate. Panel C shows comparisons of lesion proximity in high and low butyrate samples and panel D shows changes associated with butyrate level, controlling for lesional proximity. Proteins listed in red are significantly down-regulated, while those in green are up-regulated. iTRAQ, isobaric tags for relative and absolute quantification.

ws comparisons of lesion proximity in high and low butyrate samples and panel D shows changes associated with butyrate level, controlling for lesional proximity. Proteins listed in red are significantly down-regulated, while those in green are up-regulated. iTRAQ, isobaric tags for relative and absolute quantification. Seven proteins were subject to orthogonal validation by western immunoblot: keratin 8, keratin 19, α-tubulin, ApoA1, M2PK, GAPDH and cofilin (SOI section 12). The data show different responses to butyrate with progressive lesional proximity. Assessment of the effect of lesional proximity on ApoA1, tubulin and keratins 8 and 19 was undertaken (SOI, section 13). ApoA1 and α–tubulin show a consistent trend of increased expression with transition from normal to field to AD tissue (SOI, section 13). In contrast keratins 8 and 19 show a consistent trend towards downregulation across the same series.

roximity on ApoA1, tubulin and keratins 8 and 19 was undertaken (SOI, section 13). ApoA1 and α–tubulin show a consistent trend of increased expression with transition from normal to field to AD tissue (SOI, section 13). In contrast keratins 8 and 19 show a consistent trend towards downregulation across the same series. Data for keratin were extracted each iTRAQ to allow comparison of the changes observed between the soluble and insoluble fractions in the same the same subject pools were undertaken. Analysis revealed consistent trends in data (figure 4A), supporting a model whereby changes in keratin reflected the change in the total level of protein in association with a lesion or levels of butyrate. Further orthogonal validation by antibody-based methods (western immunoblotting, immunohistochemistry) was undertaken (figure 4B–F). Soluble fraction pools were analysed by western immunoblot (figure 4Bi) for keratin 8 and keratin 19. Quantification revealed that their level is constant between the MS and CL sampling positions, but that in lesions levels are significantly reduced (figure 4Bii). In the insoluble fraction, analysis of the proteomic data (figure 4C) indicated that alteration in keratin 8 and 18 was pronounced around lesions and that this was also a function of butyrate status. Immunoblot of keratin 8 and 18 in insoluble samples (figure 4C) shows that the keratin 8 immunoreactivity profile was markedly different to the soluble form: in the insoluble material the keratin 8 pool was represented by multiple bands in the 50 kDa region. Samples extracted from lesional tissue had lost some of the higher molecular weight forms, and lower bands appeared, suggestive of proteolysis. The immunoreactivity profile of low butyrate samples from macroscopically normal tissue resembled the lesional tissue, with loss of the higher molecular weight forms of keratin. The immunoblots suggest reduction is greater in low butyrate conditions.

ar weight forms, and lower bands appeared, suggestive of proteolysis. The immunoreactivity profile of low butyrate samples from macroscopically normal tissue resembled the lesional tissue, with loss of the higher molecular weight forms of keratin. The immunoblots suggest reduction is greater in low butyrate conditions. Figure 4 Integrative analysis of changes in keratin. Data on relative expression of keratins 8, 18 and 19 are extracted from the isobaric tags for relative and absolute quantification (iTRAQ) analyses and presented in panel A to allow comparison of trends across samples. Panel Bi shows immunoblot for keratin 8 and keratin 19 of different soluble fraction pools with varying mean butyrate at the mid-sigmoid (MS), contralateral (CL) and adenoma (AD) biopsy sites. Bands were quantified by densitometry and are represented in Bii, MS is used as the reference sample, levels at the CL wall and lesion AD are shown in the white and grey bars respectively. Panel C shows immunoblot analysis of keratin 8 and 18 immunoreactivity in the insoluble fractions at varying butyrate level (high and low). Comparative analysis of trend in change in the iTRAQ data for soluble and insoluble fractions is shown in Panel D. FFPE sections were stained and scored for keratin 8. Panels Ei and Eii show representative sections for high and how scores. Three aspects of keratin organisation were scored: surface intensity, crypt intensity and crypt depth. Box-and-whiskers plots show distributions of each data for each end point between three different sample sets—mid-sigmoid (normal and adenoma) and contralateral to adenoma. All end points the data showed significant differences (Jonkheere-Terpstra). When the normal group were separated into new cases (sporadic—spor) or surveillance cases free from pathology but with a history of adenoma (surveillance) there were no significant differences between end points.

and contralateral to adenoma. All end points the data showed significant differences (Jonkheere-Terpstra). When the normal group were separated into new cases (sporadic—spor) or surveillance cases free from pathology but with a history of adenoma (surveillance) there were no significant differences between end points. Since the trend is the same in the soluble and insoluble fractions (using iTRAQ and western immunoblot analysis) the hypothesis of change in cellular level of each of K8, K19 is a better fit to the data. To assess whether changes were reflected at a histological level and to allow analysis at individual subject level in unpooled samples, FFPE sections were assessed using our protocol.35 Figure 4Ei and ii show examples of crypts with strong and deep staining and weak and shallow staining, respectively, reflecting the scoring criteria developed: crypt depth (KCD), crypt intensity (KCI) and surface intensity (KSI) (ibid.). Analysis in figure 4Eiii, iv and v shows each end point against lesional proximity in samples from MS (normal) MS (AD present) and CL to AD. All outcomes were tested statistically and revealed significant differences between each measure and lesional proximity (p=0.019, p=0.014, p=0.045, respectively, Jonkheere-Terpstra test). As the normal group was a mixture of individuals from the AD surveillance and index colonoscopies, retrospective data on history of AD was obtained (where available) and a subgroup analysis undertaken. There were no significant differences between any of the keratin end points between subgroups (figure 5Fi-iii and SOI, section 14). Taken together these data reveal significant differences in expression of multiple keratins and their products as a function of butyrate status and lesional proximity. The similarity between the sporadic and surveillance subgroups suggests that the changes observed associated with presence of a lesion may be reversible by lesional resection.

ta reveal significant differences in expression of multiple keratins and their products as a function of butyrate status and lesional proximity. The similarity between the sporadic and surveillance subgroups suggests that the changes observed associated with presence of a lesion may be reversible by lesional resection. Figure 5 A high-fibre intervention elevates keratin level. Panel A graphical protocol: participants were recruited to an 8-week fibre intervention. Biopsies, food diary and faecal samples were collected at baseline and at exit from the intervention. Non-starch polysaccharide intake was significantly increased (panel B), however, there were no significant effects on faecal acetate or butyrate (Ci, Ciii) whereas propionate (Cii) approached a significant reduction. Levels of keratins 8 and 19 were measured by immunoblot, in samples preintervention and postintervention. Owing to effects of SCFAs and diet on ‘housekeeping’ markers (not shown), Coomassie staining of duplicate gels was used to show controlled loading of gels.

ereas propionate (Cii) approached a significant reduction. Levels of keratins 8 and 19 were measured by immunoblot, in samples preintervention and postintervention. Owing to effects of SCFAs and diet on ‘housekeeping’ markers (not shown), Coomassie staining of duplicate gels was used to show controlled loading of gels. A fibre intervention elevates keratin expression As dietary fibre intake is associated with a reduced risk of cancer,13 we assessed the effect of increasing fibre intake on keratin expression. A visual protocol is shown in figure 5A, and details are in the methods and reference.24 Mean fibre intake was very significantly increased (figure 5B), although surprisingly little effect was seen on levels of faecal acetate or butyrate (figure 5Ci and iii), and a trend towards reduced propionate was noted (p=0.086, figure 5Cii). Soluble protein fractions from these biopsies were extracted and immunoprobed for K8 and K19 (figure 5D). Expression of both proteins was higher following fibre intervention.

t was seen on levels of faecal acetate or butyrate (figure 5Ci and iii), and a trend towards reduced propionate was noted (p=0.086, figure 5Cii). Soluble protein fractions from these biopsies were extracted and immunoprobed for K8 and K19 (figure 5D). Expression of both proteins was higher following fibre intervention. Discussion In this study we have collected biopsies from participants with and without a neoplasia at positions with various proximities to the lesion. These were subject to a three-tier proteomic analysis, with orthogonal validation by immunoblot and immunohistochemistry and finally an independent intervention trial. Our experimental design reveals a very consistent pattern of decreased global level of keratins in lesional tissue and with butyrate concentration. Polymerised keratins are a key component of IF, and occupy the insoluble fraction of lysed material, whereas depolymerised forms may appear in the soluble fraction. The consistent pattern of change between these fractions is indicative of reduced global level of keratin with lesional proximity. The protein acetylation environment of epithelia will be influenced by microenvironmental levels of butyrate.40 41 We and others have also shown that keratins are themselves acetylated10 42 and that keratin acetylation is associated with depolymerisation.11 12 Our data also indicate that keratin forms may be altered as a consequence of the lesion and butyrate status: in lesional tissue the insoluble keratins appeared to be degraded, an observation paralleled in low-butyrate conditions in lesion-free participants. This suggests a contributing chemoprevention action of butyrate. Critically, keratin changes may remain modifiable risk factors: the fibre intervention increased keratin level and retrospective subgroup analysis implied restoration of keratin postpolypectomy.

eled in low-butyrate conditions in lesion-free participants. This suggests a contributing chemoprevention action of butyrate. Critically, keratin changes may remain modifiable risk factors: the fibre intervention increased keratin level and retrospective subgroup analysis implied restoration of keratin postpolypectomy. Changes in metabolic proteins were also noted. The colonocyte utilises butyrate as a fuel source,43 proximity to a lesion impacts on the expression profile of metabolic enzymes, possibly indicating local areas of Warburg metabolism in a primarily β–oxidative tissue. Alterations in pyruvate handling may change the functional threshold for butyrate's histone deacetylase (HDAC) inhibitory activity.41

as a fuel source,43 proximity to a lesion impacts on the expression profile of metabolic enzymes, possibly indicating local areas of Warburg metabolism in a primarily β–oxidative tissue. Alterations in pyruvate handling may change the functional threshold for butyrate's histone deacetylase (HDAC) inhibitory activity.41 An integrative model for pre-neoplastic changes Figure 6 shows an integrative model accounting for parameters we report in the context of the wider literature. Keratins are responsive to butyrate resulting in changed expression and polymerisation, however, we have also recently reported upregulation of keratin 8 in response to inflammation in active colitis (Corfe et al submitted to Journal of Pathology). Subclinical inflammation is a risk factor for carcinogenesis and has recently been shown to impact on fatty acid oxidation.44 The extracellular matrix, contributes to cell viability signalling, and alters flexibility in the neoplastic colon, interactions with and may signal via keratins. As such the metabolic, inflammatory and extracellular matrix environments may all impact on keratin at the levels of expression, polymerisation and degradation. The inflammatory and metabolic environments are partly a function of diet and the microbiome: diet may directly influence inflammation (eg, through imbalance of ω-3 and ω-6 inter alia), and carbohydrate and fibre and the microbiome together mediate the levels of SCFA in the gut. We propose a model wherein modifiable risk factors may exert subclinical effects on the colonocyte, exhibited or integrated through changes in keratin and intermediate filament function. In turn such changes may form a positive feedback loop, accelerating dysbiosis or inflammation.

ediate the levels of SCFA in the gut. We propose a model wherein modifiable risk factors may exert subclinical effects on the colonocyte, exhibited or integrated through changes in keratin and intermediate filament function. In turn such changes may form a positive feedback loop, accelerating dysbiosis or inflammation. Figure 6 Integrative model of impact of microenvironment on keratin and functional consequences. Our data provide direct molecular evidence of field-changes. In the preneoplastic colon these events remain, in principle, modifiable and offer a therapeutic window. Future research should address whether such changes are cause or effect of neoplasm. Supplementary Material Supplementary Materials The authors thank LJ Croucher and L Shaw for extraction of biopsies, Saw Yen Ow for iTRAQ sample processing. This work was funded by the Food Standards Agency (Ref N12017) the EPSRC (EP/E036252/1) and the University of Sheffield.

Our data provide direct molecular evidence of field-changes. In the preneoplastic colon these events remain, in principle, modifiable and offer a therapeutic window. Future research should address whether such changes are cause or effect of neoplasm. Supplementary Material Supplementary Materials The authors thank LJ Croucher and L Shaw for extraction of biopsies, Saw Yen Ow for iTRAQ sample processing. This work was funded by the Food Standards Agency (Ref N12017) the EPSRC (EP/E036252/1) and the University of Sheffield. Contributors: CAE designed the proteomic strategy, undertook sample preparation, undertook first and second pass analysis of the proteomic data and prepared relevant sections for the first draft of the manuscript and approved the final version. RR undertook the extraction and proteomic analysis of the insoluble fraction, data analysis, all related orthogonal work, and prepared figures for the manuscript and approved the final version. JSW designed, undertook and interpreted immunoblot analyses, contributed to the first draft of the paper and approved the final version. JN developed and designed mathematical strategies for clustering analysis and interpretation of the proteomic data, prepared relevant sections for the first draft of the manuscript and approved the final version. DL undertook the majority of consenting and recruiting participants and ran the intervention trial, maintained databases, undertook extraction of SCFA and approved the final manuscript. PCW oversaw the processing of proteomic sample, reviewed data analytic strategy and approved the final version of the manuscript. EAW co-conceived the project, continuously contributed to refinement of design, supervised the intervention arm, supervised the analysis of nutritional data prepared relevant sections for the first draft of the manuscript and approved the final version. SAR co-conceived the project and took clinical responsibility for the work, continuously contributed to refinement of design, prepared relevant sections for the first draft of the manuscript and approved the final version. JPB co-conceived the project, continuously contributed to refinement of design, oversaw relevant statistical analysis, prepared relevant sections for the first draft of the manuscript and approved the final version. BMC conceived the project, assumes overall responsibility for the project, is its director and is guarantor, undertook bioinformatic analyses, wrote the first draft of the manuscript, finalised each draft of the manuscript and produced the final version.

the first draft of the manuscript and approved the final version. BMC conceived the project, assumes overall responsibility for the project, is its director and is guarantor, undertook bioinformatic analyses, wrote the first draft of the manuscript, finalised each draft of the manuscript and produced the final version. Funding: This work was funded by the Food Standards Agency (Ref N12017) the EPSRC (EP/E036252/1) and the University of Sheffield. Competing interests: None declared. Ethics approval: North Sheffield REC. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: Proteomics datasets will be made publicly available through PRIDE on acceptance of the manuscript.

Summary box What is already known about this subject? ▸  Ulcerative colitis (UC) is associated with an increased risk of colitis-associated cancers; factors such as severity of colonic inflammation (endoscopic and histological), duration of disease and concomitant primary sclerosing cholangitis are associated with increased risk. ▸  Keratins are a type of intermediate filament (IF) proteins, which, as part of cellular cytoskeleton have important regulatory functions on the colonic mucosa. Keratin 8 (K8) null mice develop colitis and K8 is shown to modulate tumour necrosis factor action. Missense mutations in the keratin 8 gene have been noted in a subset of patients with inflammatory bowel diseases. ▸  Vimentin, a type III IF protein is often considered as a canonical marker of epithelial, mesenchymal transformation. Increased levels have been noted in aggressive colorectal cancers. What are the new findings? Proteomics demonstrate that acute inflammation results in reduction in keratins 8, 18, 19 (K8, K18, K19) and vimentin (VIM) levels. Phosphorylated K8 (K8pS23) levels relative to total K8, is increased in controls and long-standing pancolitis, while a reduced ratio is seen in dysplasia. There is also a presence of low molecular weight K8 forms and reduced K8 phosphorylation in inflamed mucosa compared to proximal uninflamed colon. Further proteomic analysis has shown potential differences between these proteoforms, suggesting potential value as biomarkers.

Phosphorylated K8 (K8pS23) levels relative to total K8, is increased in controls and long-standing pancolitis, while a reduced ratio is seen in dysplasia. There is also a presence of low molecular weight K8 forms and reduced K8 phosphorylation in inflamed mucosa compared to proximal uninflamed colon. Further proteomic analysis has shown potential differences between these proteoforms, suggesting potential value as biomarkers. How might it impact on clinical practice in the foreseeable future? Keratin (K) levels and phosphorylation are reduced in acute inflammation, but restore or increase following clinical and endoscopic remission. This process may be impaired in patients who have an elevated risk of cancer suggesting that keratin regulation in remission may be a pivotal factor influencing subsequent risk of development of colitis-associated cancer. Monitoring of relative ratio of vimentin to K8, phosphorylated K8, or appearance of novel proteoforms might be useful as markers of aggressive disease phenotype in ulcerative colitis.

ng that keratin regulation in remission may be a pivotal factor influencing subsequent risk of development of colitis-associated cancer. Monitoring of relative ratio of vimentin to K8, phosphorylated K8, or appearance of novel proteoforms might be useful as markers of aggressive disease phenotype in ulcerative colitis. Introduction Intermediate filaments (IF) are an important component of the cellular cytoskeleton, with keratins constituting the major IF proteins in epithelial cells and accounting for approximately 5% of total cellular protein.1 2 In the intestinal epithelium, principally expressed keratins are keratins 8, 18 and 19 (K8, K18, K19).2 Vimentin (VIM), a type III IF protein is primarily found in cells of mesenchymal origin3 and replaces keratins early in epithelial–mesenchymal transition. IF proteins share a similar structure.2 The structure, solubility and functions of IFs are regulated, in part, by a range of post-translational modifications (PTMs), including phosphorylation, glycosylation, acetylation and cleavage.2 4 Specific sites of phosphorylation, predominantly in the head and tail domains, have been noted in serine (S) residues of K8 at S23, S73 and S431. Phosphorylation of these sites is under control of protein kinases including p38,5 mitogen-activated protein kinases (MAPK)6 and c-jun kinase.7 The curated PTM deposition database, phosphosite (http://www.phosphosite.org), indicates additional PTM types including acetylation and glycosylation. Sumoylation, which exhibits reciprocity with acetylation at some sites, has also been demonstrated to modify K8 function in liver.8 IFs are dynamic and undergo reorganisation in response to a variety of stimuli including, mitosis, apoptosis and other cellular stresses often with redistribution between an insoluble (filamentous) and a soluble pool.9

ciprocity with acetylation at some sites, has also been demonstrated to modify K8 function in liver.8 IFs are dynamic and undergo reorganisation in response to a variety of stimuli including, mitosis, apoptosis and other cellular stresses often with redistribution between an insoluble (filamentous) and a soluble pool.9 Keratins are associated with the pathogenesis of various colorectal diseases, including cancer and inflammatory bowel diseases (IBD). They have been shown to be critical in maintaining the epithelial integrity and to protect against mechanical and non-mechanical stresses,10 and regulate effects of signalling pathways. Keratins also function in cell-death signalling pathways, in particular apoptosis mediated by Fas and tumour necrosis factor (TNF) α.11 12 Stromal VIM expression in colorectal cancers correlates with malignant potential of the tumours.13

anical and non-mechanical stresses,10 and regulate effects of signalling pathways. Keratins also function in cell-death signalling pathways, in particular apoptosis mediated by Fas and tumour necrosis factor (TNF) α.11 12 Stromal VIM expression in colorectal cancers correlates with malignant potential of the tumours.13 The pathogenesis of ulcerative colitis (UC) is still not fully understood, and factors such as altered immune response to luminal microbiota14 and alterations in mucosal barrier function15 may be contributory. A potential role of impaired epithelial barrier function due to alterations in keratin levels has been hypothesised.16 Notably, the keratin gene superfamily is located within the IBD2 locus on chromosome 12.16 Heterozygous missense mutations in the K8 gene have been reported in patients with IBD with in vitro experiments demonstrating inefficient filament assembly by the mutant K8 in comparison with wild type.17 K8-deficient mice (mK8−/−FVB/N) develop colonic inflammation and mucosal hyperplasia;18 a mucosal chronic T-helper type 2 inflammatory response has been noted in K8−/− mice that develop colitis.19 The latter change reverses with antibiotic therapy suggesting a contributory role of luminal bacterial microbiota.20

ype.17 K8-deficient mice (mK8−/−FVB/N) develop colonic inflammation and mucosal hyperplasia;18 a mucosal chronic T-helper type 2 inflammatory response has been noted in K8−/− mice that develop colitis.19 The latter change reverses with antibiotic therapy suggesting a contributory role of luminal bacterial microbiota.20 UC elevates risk of developing colitis-associated cancer (CAC). Duration of colitis is an important factor in increasing colon cancer risk, with cumulative colorectal cancer (CRC) incidence noted at 2%, 8% and 18%, at 10, 20 and 30 years, respectively.21 Extent of colonic involvement,22 concomitant primary sclerosing cholangitis (PSC)23 as well as early age of onset of UC are also risk factors for CAC.22 Severity of inflammation increases the risk of dysplasia and CRC,24 and both microscopic as well as inflammation evident endoscopically are considered important.25 26 Quantitative proteomic techniques enable not only identification of proteins in biological samples, but also their alteration in level between samples.27 This permits determination of changes in response to disease-related factors. Few studies have investigated mucosal proteomic changes in IBD, and only one has shown changes in cytoskeletal proteins, showing elevated levels of actin cytoskeletal and associated proteins in patients with a higher risk of UC progression.28 Given associations between K8 and colorectal cancer progression and its interactions with inflammatory pathways, we investigated the mucosal changes in IFs in patients with UC with differing cancer risk and inflammatory status.

tin cytoskeletal and associated proteins in patients with a higher risk of UC progression.28 Given associations between K8 and colorectal cancer progression and its interactions with inflammatory pathways, we investigated the mucosal changes in IFs in patients with UC with differing cancer risk and inflammatory status. Materials and methods Patients Patients (18–75 years old) with histologically proven UC and healthy controls (CON) were recruited prospectively from outpatient clinics and inpatient wards at the Royal Hallamshire Hospital, Sheffield, UK. The study was approved by South Yorkshire Research Ethics Committee (10/H1310/21). Written informed consent was obtained from all patients. Clinical activity in patients with UC was defined by the Baron UC disease activity index score.29 Patients were categorised into groups of differing CRC risk and inflammatory status. Recent onset (<5 years duration) UC (ROUC). Long-standing pancolitis (LSPC)—with disease duration between 20 and 40 years. Patients in the LSPC and ROUC groups were in established deep remission with no evidence of endoscopic activity (Baron endoscopic scores 0 and 1, respectively) and no microscopic activity (histological activity scores 0 in both groups). UC with concomitant primary sclerosing cholangitis (PSC); Dysplasia in patients with pancolitis (dysplastic tissue, DT); Rectal mucosa distant to areas of dysplasia (DR). Biopsies were obtained in patients with active distal colitis from The inflamed rectal mucosa (ACTively inflamed mucosa; ACT); The proximal, endoscopically uninflamed mucosa from the same patients as 6 (INACT);

Patients in the LSPC and ROUC groups were in established deep remission with no evidence of endoscopic activity (Baron endoscopic scores 0 and 1, respectively) and no microscopic activity (histological activity scores 0 in both groups). UC with concomitant primary sclerosing cholangitis (PSC); Dysplasia in patients with pancolitis (dysplastic tissue, DT); Rectal mucosa distant to areas of dysplasia (DR). Biopsies were obtained in patients with active distal colitis from The inflamed rectal mucosa (ACTively inflamed mucosa; ACT); The proximal, endoscopically uninflamed mucosa from the same patients as 6 (INACT); CONs were patients undergoing lower gastrointestinal (GI) endoscopic examination which was endoscopically and histologically normal, with no prior history of IBD, nor from hospital records did they develop IBD in the 2 years after biopsies were obtained. Biopsies were obtained at the rectum from all the patients with UC in remission as well as normal controls. Lower GI endoscopy and colonic biopsy All patients underwent flexible sigmoidoscopy or colonoscopy following bowel preparation using either Kleanprep (Norgine Limited, UK) or Picolax (Ferring Pharmaceuticals, UK). Endoscopic biopsies were collected from the colon using Radial Jaw 4, biopsy forceps (Boston Scientific Corporation, Massachusetts, USA). Biopsies were snap-frozen in liquid nitrogen and stored at −80°C (for proteomic analyses) or formalin fixed for histological assessment. Exclusion criteria and patient data are in the online supplementary data.

Lower GI endoscopy and colonic biopsy All patients underwent flexible sigmoidoscopy or colonoscopy following bowel preparation using either Kleanprep (Norgine Limited, UK) or Picolax (Ferring Pharmaceuticals, UK). Endoscopic biopsies were collected from the colon using Radial Jaw 4, biopsy forceps (Boston Scientific Corporation, Massachusetts, USA). Biopsies were snap-frozen in liquid nitrogen and stored at −80°C (for proteomic analyses) or formalin fixed for histological assessment. Exclusion criteria and patient data are in the online supplementary data. Preparation of IF-enriched fraction We followed our technique for isolation and solubilisation of IFs for mass spectrometry (MS).30 The IF fraction from cultured MCF-7 cells used as controls in the western immunoblot and as internal standards for densitometry.30 Experimental details are provided in full (see online supplementary data, sections 3 and 5). Proteomic methods iTRAQ (isobaric tags for relative and absolute quantitation) protein profiling: Briefly, pooled samples were tryptically digested, labelled with 8-plex iTRAQ reagents, separated by strong cation exchange offline before tandem MS. Peptides were identified with Phenyx and Uniprot, before relative quantification and statistical analysis.31 Details are in the online supplementary data, section 4A. Label-free MS: Gel bands were excised and digested in-gel prior to LC/MS-MS analysis coupled with ion mobility (HDMSE). Data processing used UniProt and Progenesis QI software.32 Details are in the online supplementary data, section 4B.

Proteomic methods iTRAQ (isobaric tags for relative and absolute quantitation) protein profiling: Briefly, pooled samples were tryptically digested, labelled with 8-plex iTRAQ reagents, separated by strong cation exchange offline before tandem MS. Peptides were identified with Phenyx and Uniprot, before relative quantification and statistical analysis.31 Details are in the online supplementary data, section 4A. Label-free MS: Gel bands were excised and digested in-gel prior to LC/MS-MS analysis coupled with ion mobility (HDMSE). Data processing used UniProt and Progenesis QI software.32 Details are in the online supplementary data, section 4B. Sodium dodecyl sulfate polyacrylamide gel electrophoresis and western immunoblotting Buffer exchange, western transfer and immunoblotting were as previously described.4 Primary antibodies were mouse monoclonal: anti-K8 (ab9023); K18 (ab668), K19 (ab7754); K8 phospho-specific rabbit antiphospho S73 antibody (ab32579) and phospho S431 (ab59434) (supplied by Abcam, Cambridge, UK). In-house antibodies raised against acetylated lysine10 residue of K8 (rabbit).33 Densitometric analysis is detailed in online supplementary data, section 7. Immunoblotting was performed for (n=56) patients.

specific rabbit antiphospho S73 antibody (ab32579) and phospho S431 (ab59434) (supplied by Abcam, Cambridge, UK). In-house antibodies raised against acetylated lysine10 residue of K8 (rabbit).33 Densitometric analysis is detailed in online supplementary data, section 7. Immunoblotting was performed for (n=56) patients. Histological assessments of inflammation and keratin expression Biopsies were formalin fixed, and tissue sections (4 µm thick) were cut. Sections were H&E stained, and reviewed by a single histopathologist (SSC) blinded to the diagnosis. Dysplasia and histological severity of inflammation were defined according to standard criteria.26 34 Immununohistochemistry of K8, K18 and K19 was essentially as described using antibodies to K8, K19 ab9023, ab7754, respectively (Abcam, Cambridge, UK) and K18 (in-house monoclonal antibody). Scoring of keratin for crypt intensity, crypt depth and surface intensity were as previously described.35 Immunohistochemistry (n=48, was performed where formalin-fixed paraffin embedded (FFPE) tissue was adequately well oriented for scoring). Statistical analysis Data have been described as median and range/interquartile range, as indicated. Differences between groups were evaluated with the non-parametric Mann-Whitney U test. Analysis of iTRAQ data and relative quantification of fold changes in protein levels was performed as described previously31 where the criteria for alterations on protein level were p value <0.05, with Bonnferroni multiple test correction applied to reduce false positives.

d with the non-parametric Mann-Whitney U test. Analysis of iTRAQ data and relative quantification of fold changes in protein levels was performed as described previously31 where the criteria for alterations on protein level were p value <0.05, with Bonnferroni multiple test correction applied to reduce false positives. Results Patients and workflow Patients: 62 adult patients with histologically proven UC healthy CONs were recruited prospectively from outpatient clinics or inpatient wards at Royal Hallamshire Hospital, Sheffield, UK. CONs were patients undergoing lower GI endoscopic examination who had normal endoscopic and histological appearance of the colonic mucosa. Patients were stratified into groups based on cancer risk. Inflammatory status measured by the Baron endoscopic appearance and the Histological Activity Score is detailed in the Methods section and in the online supplementary table S1.

ination who had normal endoscopic and histological appearance of the colonic mucosa. Patients were stratified into groups based on cancer risk. Inflammatory status measured by the Baron endoscopic appearance and the Histological Activity Score is detailed in the Methods section and in the online supplementary table S1. Acute inflammation results in reduced levels of keratins and VIM The insoluble proteome was extracted from flash-frozen biopsies and analysed using an iTRAQ workflow (figure 1 and online supplementary data, section 2). The demographic summary of samples used in the pooling is in online supplementary data, section 6. The list of proteins identified and relatively quantified in the analysis is detailed in the online supplementary data, section 6. In total, 53 proteins were identified and relatively quantified, of which 31 had information from two or more peptide sequences. A qualitative analysis of the whole dataset was undertaken using STRINGS, a database of known and predicted protein interactions to determine protein interacting networks (PINs; figure 2A) represented by the data set. Two principle PINs emerged in this analysis, with clusters around fibronectin (FN1) and the major keratin components K8, K18, K19 indicating linkage to extracellular matrix, consistent with IF function. The proteomic data for ACTively inflamed mucosa (ACT) and unINvolved mucosa in subjects with ACTive inflammation (INACT) were compared to identify alterations in protein levels between groups. Keratins were significantly decreased, with fold change in levels of K8 (0.4-fold), K18 (0.7-fold), K19 (0.4-fold) and VIM (0.6-fold) (all p<0.05) in inflamed mucosa in comparison with CONs as well as to the uninvolved proximal colon (figure 2B). The uninflamed mucosa did not show any significant difference in protein levels by MS from the mucosa of controls. Immunoblotting was undertaken to validate the proteomic data and, using a bank of antibodies to K8 and key post-translational modifications (figure 2Ci), to explore further the nature of changes in K8 associated with inflammation. The positions of the pSer23, pSer73 and pSer431 are shown relative to AcLys10 (subject to western blot, shown in the context of previously identified4 lysine acetylation sites; figure 2Ci). The western blot data is shown in figure 1Cii. Total K8 levels were reduced in ACT, both in comparison with CON and INACT; in ACT, K8 was predominantly concentrated in the lower molecular weight forms (37 kDa).

ubject to western blot, shown in the context of previously identified4 lysine acetylation sites; figure 2Ci). The western blot data is shown in figure 1Cii. Total K8 levels were reduced in ACT, both in comparison with CON and INACT; in ACT, K8 was predominantly concentrated in the lower molecular weight forms (37 kDa). A reduced ratio of Vim:K8 was also noted in ACT and not paralleled in changed ratios between K8 and K18 (figure 2D). Phosphorylation of K8 is reduced in tumour progression.36 The relative phosphorylation at sites pSer431, pSer23 and pSer73 in INACT, ACT and control tissues was therefore evaluated by western immunoblot and relative densitometry to an internal standard of MCF-7 IF extract (figure 2Cii, E). The data suggest that there is much less relative phosphorylation at all three sites (ie, when differential levels of K8 are controlled for) in the samples from inflamed colon, and that this is more pronounced in the inflamed versus the non-inflamed region, suggesting that a progressive dephosphorylation of K8 is associated with inflammatory status. Figure 1 Graphical summary of the workflow. Protein profiles of samples from the eight-patient group were compared using iTRAQ (isobaric tags for relative and absolute quantitation)-based quantitative proteomics. Downstream validation was undertaken using both immunoblotting (n=56) and immunohistochemistry (n=48, as not all formalin-fixed paraffin embedded (FFPE) tissue was adequately well oriented for scoring).

group were compared using iTRAQ (isobaric tags for relative and absolute quantitation)-based quantitative proteomics. Downstream validation was undertaken using both immunoblotting (n=56) and immunohistochemistry (n=48, as not all formalin-fixed paraffin embedded (FFPE) tissue was adequately well oriented for scoring). Figure 2 Effect on active inflammation on intermediate filament (IF) proteome. IF extracts were subject to proteomic analysis (see online supplementary data, sections 4 and 6). (A) Strings analysis of the global proteome showing key protein interacting networks (PINs). (B) Proteins identified included K8, K18, K19 and VIM (vimentin), as well as IF-associated proteins spectrin β chain brain 4, Xin actin-binding protein 1 and KRT1B as well as histone proteins and collagen. White bars show change in ACT relative to INACT, and grey bars showing ACT relative to control (CON). (Ci) Position of key antibodies used in this study and (Cii) immunoblots of ACT and INACT pooled samples entered into the proteomic analysis. Densitometry was undertaken and (D) ratios of Krt8 and either Krt18 or VIM; (E) alteration in relative phosphorylation at each site. As the pooled samples used in iTRAQ and immunoblot may mask heterogeneity. We undertook two tiers of orthogonal validation of unpooled samples (see online supplementary data, section 8). Antibody-based methods of dot blotting and immunohistochemistry (IHC) were employed.

Figure 2 Effect on active inflammation on intermediate filament (IF) proteome. IF extracts were subject to proteomic analysis (see online supplementary data, sections 4 and 6). (A) Strings analysis of the global proteome showing key protein interacting networks (PINs). (B) Proteins identified included K8, K18, K19 and VIM (vimentin), as well as IF-associated proteins spectrin β chain brain 4, Xin actin-binding protein 1 and KRT1B as well as histone proteins and collagen. White bars show change in ACT relative to INACT, and grey bars showing ACT relative to control (CON). (Ci) Position of key antibodies used in this study and (Cii) immunoblots of ACT and INACT pooled samples entered into the proteomic analysis. Densitometry was undertaken and (D) ratios of Krt8 and either Krt18 or VIM; (E) alteration in relative phosphorylation at each site. As the pooled samples used in iTRAQ and immunoblot may mask heterogeneity. We undertook two tiers of orthogonal validation of unpooled samples (see online supplementary data, section 8). Antibody-based methods of dot blotting and immunohistochemistry (IHC) were employed. Figure 3A summarises by-patient changes in keratins assessed by dot-blotting for all three major keratins. Medians and distribution patterns are shown and agree with the pooled data (for raw data, see online supplementary data, section 8). Immunohistochemistry of keratin levels in crypts adds spatial information on the distribution within tissue, with the caveat that it is less sensitive, and the need to select scorable crypts potentially biases results. Analysis of distribution of keratins expression (figure 3B and see online supplementary figures S8.1, S8.2) showed significant changes in the intensity of staining in the crypt, as opposed to total intensity or depth of staining, agreed with the proteomic and immunoblot data.

scorable crypts potentially biases results. Analysis of distribution of keratins expression (figure 3B and see online supplementary figures S8.1, S8.2) showed significant changes in the intensity of staining in the crypt, as opposed to total intensity or depth of staining, agreed with the proteomic and immunoblot data. Figure 3 Analysis of individual variation in unpooled samples in response to inflammation. (A) Relative signal intensity differences for keratins 8, 18 and19 in colonic biopsies from patients with active colitis (ACT), and reciprocal proximal inactive colonic mucosa (INACT) from the same individuals in comparison with MCF-7 as an external control. Relative signal intensity differences for keratins 8, 18 and 19 from rectal colonic biopsy for five normal controls (CON) relative to MCF-7 are also shown. Solid horizontal lines represent the median with vertical bars as IQR in each group. Differences are tested using a Mann-Whitney U test. Raw data are shown in the supplementary online information. (B) Corresponding formalin-fixed paraffin embedded (FFPE) sections are stained and scored for keratins identified in this study. Colonic epithelial average crypt intensity for K8, K18 and K19 using immunohistochemical staining comparing patients with active colitis (ACT) with reciprocal proximal inactive colonic mucosa (INACT) from the same individuals; FFPE rectal tissues from normal individuals are IHC stained for K18 as CON; solid horizontal lines represent the median with vertical bars as IQR.

8 and K19 using immunohistochemical staining comparing patients with active colitis (ACT) with reciprocal proximal inactive colonic mucosa (INACT) from the same individuals; FFPE rectal tissues from normal individuals are IHC stained for K18 as CON; solid horizontal lines represent the median with vertical bars as IQR. Long-standing quiescent disease associates with a distinct keratin expression and phosphorylation profile compared with either recent onset or higher cancer risk groups The expression levels of key IF proteins (K8, K18, K19 and VIM) in the iTRAQ data set were compared across the patient groups in this study (figure 4A). Pooled samples entered into the proteomic analysis were subject to orthogonal validation by western immunoblot (figure 4B). When the control group was the normalising reference point (figure 4Ai), the analysis revealed partitioning of profiles: LSPC exhibited generally higher levels of IF proteins relative to control, whereas ROUC, PSC, DR and DT samples all showed reduced levels of most IF proteins relative to control, a marked opposition of direction of change. The increases noted in LSPC were significant: K8 and K19 (1.7-fold each, p<0.05) and VIM levels (2.2-fold) (figure 4Ai, B). By contrast, in ROUC, despite having macroscopically and microscopically quiescent disease, exhibited significantly reduced keratin levels relative to controls (K8, K18, K19 and VIM 0.3, 0.5, 0.4 and 0.5-fold respectively, all p<0.05). Coexisting PSC results in a fourfold increase in the colon cancer risk in colitis patients;23 likewise, we assessed tissue from patients with dysplasia both at the DT and uninvolved rectal mucosa (DR). In PSC, there was a significant reduction in keratin and VIM levels (0.8, 0.7, 0.8 and 0.9-fold for K8, K18 and K19 and VIM, respectively; p<0.05, figure 4Ai) in comparison with CON (figure 4Ai, B). Analysis of DT and rectal mucosa (DR) also showed a similar pattern of changes in DR and DT relative to controls (figure 4Ai, B). These changes suggest a possible pan-colonic field change in IF proteome alterations in patients with dysplasia.

IM, respectively; p<0.05, figure 4Ai) in comparison with CON (figure 4Ai, B). Analysis of DT and rectal mucosa (DR) also showed a similar pattern of changes in DR and DT relative to controls (figure 4Ai, B). These changes suggest a possible pan-colonic field change in IF proteome alterations in patients with dysplasia. Figure 4 Proteomic analysis revealling intermediate filament (IF) proteins are differentially altered in ulcerative colitis (UC) subgroups. (A) iTRAQ (isobaric tags for relative and absolute quantitation) data of the changes in IF in patients with long-standing pancolitis (LSPC), recent onset UC (ROUC) and primary sclerosing cholangitis (PSC), and those with dysplasia (DR and DR, respectively). By comparison with controls (Ai), LSPC showing a significant increase in levels of K8, K19 and VIM (vimentin); in ROUC, PSC, DR and DT the levels of K8, K18 and K19 are significantly decreased. The oppositional direction of change between LSPC and other groups when set against controls reinforces evidence that this is a biological not a technical event. A comparison of major proteins in quiescent and LSPC disease was undertaken (Aii): LSPC samples exhibited a marked increase in all IF proteins and IF-associated proteins (Spectrin, Xin) compared with ROUC. Comparison of samples from higher risk subjects (PSC, DR, DT) against LSPC revealed marked changes in K8, K19 and VIM, with the VIM directionality distinguishing PSC from DT/DT. Finally in (Aiv) DT samples showed reduced levels of K8, K18, Xin and Spectrin relative to DR. (B) Results of orthogonal validation undertaken by western immunoblot using specific antibodies to K8, K18, K19 and VIM. Multiple isoforms of K8 are noted in all the patient groups including healthy controls with lower levels and low molecular weight forms predominating in ROUC and DT. Immunoblotting using specific antibodies to three major phosphorylation sites of K8 (pS23, pS73 and pS431) and one acetylation site of K8 (AcLys10) was undertaken, along with immunoblotting for K8 in the same samples. A control MCF-7 sample was included in all immunoblots to allow normalisation between experiments. Bands were quantified by densitometry, normalised within-blot to the MCF-7 control, and then across-blot to the K8 sample. Relative phosphorylation levels and VIM:K8 thereby determined are shown in (C and D).

ame samples. A control MCF-7 sample was included in all immunoblots to allow normalisation between experiments. Bands were quantified by densitometry, normalised within-blot to the MCF-7 control, and then across-blot to the K8 sample. Relative phosphorylation levels and VIM:K8 thereby determined are shown in (C and D). In order to investigate effects of disease duration on IF proteins, LSPC data were compared with ROUC data (ROUC as reference set, figure 4Aii). In comparison with ROUC, a significant increase in levels of IF proteins (4.3, 1.2, 4 and 3.5-fold for K8, K18, K19 and VIM, respectively; p<0.05), and IF-associated proteins Xin and Spectrin was noted in LSPC (figure 4Aii), suggesting an effect of duration of disease on levels of principal keratins. The LSPC group exhibited oppositionality of changes for most keratins by comparison with the control group through a re-plot of the data comparing ROUC, PSC, DR and DT with LSPC as reference group (figure 4Aiii, B). In PSC, there was a significant reduction in keratin and VIM levels (0.8, 0.7, 0.8 and 0.9-fold for K8, K18 and K19 and VIM, respectively; p<0.05, figure 4Ai; online supplementary table S2) in comparison with the CON group (figure 4Ai, 2B). Analysis of DT relative to LSPC revealed a significant reduction in levels of IFs (0.3, 0.2 and 0.7-fold for K8, K19 and VIM, p<0.05) (figure 4Aiii), as well as KRT1B, Xin and Spectrin (figure 4Aiii;), 0.6-fold for both, p<0.05). Rectal mucosa (DR) also showed reduced levels (figure 3Aiv) relative to LSPC. Finally, in order to establish whether there were changes associated with dysplasia which may discriminate DR and DT tissue, a direct comparison was undertaken (figure 4Aiv) which revealed reduction of K8, K19, Xin and Spectrin in dysplastic tissue relative to matched controls.

ed levels (figure 3Aiv) relative to LSPC. Finally, in order to establish whether there were changes associated with dysplasia which may discriminate DR and DT tissue, a direct comparison was undertaken (figure 4Aiv) which revealed reduction of K8, K19, Xin and Spectrin in dysplastic tissue relative to matched controls. As keratin phosphorylation in inflamed tissue was significantly reduced (vide supra), we sought to establish whether any such changes associate with particular subgroups of UC. Pooled samples, as entered into the proteomic analysis, were immunoprobed for K8, for phosphorylation at pSer23, pSer73, pSer431 and lysine acetylation at AcLys10 (figure 4B). Multiple immunoreactive bands of K8 corresponding to as many as six discrete bands between 37 and 50 kDa on western blot were noted in control patients, LSPC, PSC and DR (figure 3B). By contrast with controls, the isoforms in LSPC were evident at higher molecular weights, whereas in PSC and DR they tended to predominate at lower molecular weights. Similar pattern of K8 alteration is noted in quiescent ROUC mucosa as well as in DT. In order to assess whether these K8 forms arose through differential reversible post-translational modifications, we undertook western blot to identify common PTMs in K8 in the form of phosphorylation at pSer23, pSer73, pSer431 and lysine acetylation at AcLys10. In healthy controls, intense phosphorylation at Ser73 and Ser 431 was noted predominantly at a single band around 37 kDa, while phosphoSer23 antibody cross-reacted with a more diverse set of molecular weights. The pattern (if not the intensity) of cross-reaction was broadly conserved in LSPC and DR. DT and ROUC samples exhibited lower levels of K8 cross-reaction, a less diverse array of molecular weights, and so had correspondingly fewer cross-reactions with anti-PTM antibodies. LSPC, by contrast, exhibited cross-reactions for PTMs across a wider range of molecular weights and was the only in vivo sample to exhibit cross-reaction for lysine10. Densitometry allowed the quantitative assessment of changes in relatively phosphorylation, corrected for K8 level (see figure 3C and online supplementary data, section 3). The analysis reveals that relative phosphorylation in all sites is consistent between LSPC and CON, but lower in other conditions, markedly so at pSer23.

tometry allowed the quantitative assessment of changes in relatively phosphorylation, corrected for K8 level (see figure 3C and online supplementary data, section 3). The analysis reveals that relative phosphorylation in all sites is consistent between LSPC and CON, but lower in other conditions, markedly so at pSer23. As VIM levels were reduced in acute inflammation and in ROUC, and were increased in LSPC, PSC, DR and DT, a densitometry analysis (see online supplementary data, section 7 for regions quantified) of VIM and K8 immunoblots was undertaken (figure 4D). We undertook a novel analysis of the ratio of VIM to K8 which showed reduced levels in ROUC, in comparison with CONs. There was a progressive increase in VIM/K8 ratios from LSPC to PSC, with markedly increased levels noted in DT and DR (other comparisons were undertaken; see online supplementary data).

ken (figure 4D). We undertook a novel analysis of the ratio of VIM to K8 which showed reduced levels in ROUC, in comparison with CONs. There was a progressive increase in VIM/K8 ratios from LSPC to PSC, with markedly increased levels noted in DT and DR (other comparisons were undertaken; see online supplementary data). Orthogonal validation of changes in levels of the major keratins in individual patients was undertaken as for ACT/INACT samples by IHC. This complementary methodology is considerably less sensitive than iTRAQ or western blot, but provides an assessment of total keratin as well as information changes in keratin level across the tissue architecture. Fold-changes observed relative to control samples were modest at the iTRAQ level (<2-fold for keratins in LSPC; <0.5-fold for keratins in other groups) and were not found to be significant (figure 5A), however, the general trends observed matched those identified by protein chemistry. It is likely that this also reflects that relatively fewer samples were scorable by IHC, and numbers in some pools were low.

for keratins in LSPC; <0.5-fold for keratins in other groups) and were not found to be significant (figure 5A), however, the general trends observed matched those identified by protein chemistry. It is likely that this also reflects that relatively fewer samples were scorable by IHC, and numbers in some pools were low. Figure 5 Analysis of individual variation in unpooled samples in response to inflammation. Corresponding formalin-fixed paraffin embedded (FFPE) sections were stained and scored for keratins identified in this study. Colonic epithelial average surface and crypt intensity for K8 (top panel), K18 (centre panel) and K19 (bottom panel) were scored comparing patients with: active colitis (ACT), reciprocal proximal inactive colonic mucosa (INACT) from the same individuals, long-standing pancolitis (LSPC), recent-onset ulcerative colitis (UC) (ROUC), UC with concomitant primary sclerosing cholangitis (PSC), UC with dysplasia, and rectal sample from UC with dysplasia (DR); FFPE rectal tissues from normal individuals were immunohistochemical staining for each keratin as control (CON). Each group was compared with the CON and the statistically significant difference marked with a (*) sign at the top. *p<0.05, Mann-Whitney U test.

th dysplasia, and rectal sample from UC with dysplasia (DR); FFPE rectal tissues from normal individuals were immunohistochemical staining for each keratin as control (CON). Each group was compared with the CON and the statistically significant difference marked with a (*) sign at the top. *p<0.05, Mann-Whitney U test. Mass spectrometric analysis of proteoforms of keratins identifies candidate biomarkers Immunoblot analysis using K8-specific antibodies revealed a series of immunoreactive bands: potentially representing different proteoforms (representing different splice variants and/or PTM status) of K8 (figure 2Cii, 4B). In order to investigate this further, paired IF fractions from distal, active, inflamed rectal mucosa, and from proximal, uninflamed mucosa in the same patient were studied. Patient samples were first resolved by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE; figure 5A). Individual bands (R1, S1-S5) were excised from the gel, proteolytically digested with trypsin, and analysed using label-free MS37 (see online supplementary data, section 4B). While all bands excised contained K8, this was not the predominant species for all, and relative recovery of K8 is shown in figure 6B. PTMs were identified: lysine acetylations at sites were detected at AcLys107, AcLys122, AcLys158, AcLys325 and phoshorylation of pSer23, pSer34 and pSer431 which represent the major phosphorylations sites of K8. The deconvoluted peptide maps of each species are shown in figure 6C and sequence coverage mapped in figure 6D. The data indicate a substantially reduced complexity of K8 proteoforms in active colitis.

158, AcLys325 and phoshorylation of pSer23, pSer34 and pSer431 which represent the major phosphorylations sites of K8. The deconvoluted peptide maps of each species are shown in figure 6C and sequence coverage mapped in figure 6D. The data indicate a substantially reduced complexity of K8 proteoforms in active colitis. Figure 6 Analysis of keratin proteoforms using a label-free proteomic approach. (A) Proteins from intermediate filament (IF) fractions derived from biopsy materials from two patients with active ulcerative colitis (UC) and paired inactive biopsy material were resolved by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE; 10%). Bands in the region corresponding to K8 were excised from active (R1) and inactive (S1–S5) gel lanes. (B) The amount of K8 protein in each band was calculated. All samples were run in duplicate and the results are the mean values. (C) All peptides corresponding to K8, are shown in yellow for each sample with modified residues shown in green (5Bi), and (ii) phosphorylated and (iii) acetylated residues were identified and indicated in green. (D) The sequence coverage is shown for K8 in the gel bands analysed, peptides shown in yellow with modified residues oxidised methione, acetylated lysine or phosphoserine highlighted in green.

ed residues shown in green (5Bi), and (ii) phosphorylated and (iii) acetylated residues were identified and indicated in green. (D) The sequence coverage is shown for K8 in the gel bands analysed, peptides shown in yellow with modified residues oxidised methione, acetylated lysine or phosphoserine highlighted in green. Discussion Development of colon cancer in UC is multifactorial, and inflammation is now considered to play a significant role in its pathogenesis.25 The natural history of UC is variable: the frequency of disease flares is unpredictable and cancer develops in a small proportion of such patients. Ongoing active histological inflammation has been shown to predispose to development of CAC.26 In view of the evidence implicating inflammation in cancer and colitis, and the potential association of keratins in the colonic inflammation and the pathogenesis of IBD,17 18 we investigated proteomic changes in IFs in well-characterised groups of patients with UC at varying levels of cancer risk. We have shown that acute inflammation is associated with a reduction in levels of keratins in the inflamed colonic mucosa, whereas the levels in the uninflamed proximal mucosa parallel those of healthy colonic mucosa in control patients. Keratins are dynamic and are involved in a range of inflammatory pathways, in particular protection from apoptosis mediated by TNF-α and Fas, with epithelial cells lacking K8/K18 being prone to apoptosis, and K8−/− mice exhibiting hyperplasia, fragile mucosa and impaired ion transport.11 12 38 Our data show that keratin levels are at, or slightly higher than, normal level in patients with persistent quiescent disease, but that in active inflammation and in quiescent subjects with recent disease, there is a reduced level of keratin. Taken together, these data suggest a delay in the restoration of keratins in the mucosa following acute inflammation despite macroscopic and microscopic mucosal healing. Previous studies have also shown persistent cellular and molecular damage with activated kinase and transcription factor signalling pathways in the colon despite apparent microscopic healing.39

in the restoration of keratins in the mucosa following acute inflammation despite macroscopic and microscopic mucosal healing. Previous studies have also shown persistent cellular and molecular damage with activated kinase and transcription factor signalling pathways in the colon despite apparent microscopic healing.39 Phosphorylation of K8 affects its solubility and function.2 Reduced phosphorylation of K8 due to increased phosphatase of regenerating liver-3 expression is associated with colorectal cancer progression.36 Our data show reduced or a loss of phosphorylation in acute inflammation, with similar changes persisting in ROUC as well as DT. This suggests a protective effect of physiological phosphorylation of K8 at pSer23 would be lost in ROUC and DR/DT. Increased levels and phosphorylations of keratins in LSPC may be significant: LSPC is believed to be an important factor contributing to the pathogenesis of CAC, but levels of keratins in quiescent LSPC points towards a protective mechanism in this situation. The environmental and regulatory factors governing K8/K18 are not well characterised and merit analysis in future studies. There are limited data on the role of acetylation in regulation of keratin function, but we have previously shown association between Lys483 acetylation and depolymerisation. We noted that the K8 recovered from inflamed tissue appeared more extensively acetylated (by number of sites, not necessarily stoichiometry) which may suggest progression to entropy of IF and consequent plasticity of cells.

ction, but we have previously shown association between Lys483 acetylation and depolymerisation. We noted that the K8 recovered from inflamed tissue appeared more extensively acetylated (by number of sites, not necessarily stoichiometry) which may suggest progression to entropy of IF and consequent plasticity of cells. Taken together with the results of acute inflammation on keratins, we can suggest that inadequate restoration of keratin expression following acute inflammation may contribute to the pathogenesis of CAC by affecting the repair process in the mucosa. This is supported by the changes seen in patients with PSC and dysplasia, but the number of patients in these groups is small. The findings, therefore, need confirmation in prospective studies in these subgroups. This study remains cross-sectional, and future prospective analyses will reveal whether repeated inflammation or impaired keratin expression is the more causal factor in CAC risk phenotypes. In this study, we show an acute reduction in VIM levels in the acutely inflamed mucosa as well as in ROUC. Increased expression of VIM, generally expressed in cells with a mesenchymal phenotype in LSPC may reflect morphological colonic tissue remodelling and architectural alterations, reflecting the chronic relapsing/remitting course of the disease as a consequence of accumulated damage during each active phase.40 41 The VIM:K8 ratio progressively increases with development of a more aggressive phenotype, potentially due to epithelial denudation and crypt shortening associated with disease. These results suggest that relative overexpression of VIM in the colonic mucosa may herald the development of CAC. The data are integrated into a qualitative model for CAC progression in figure 7.

ment of a more aggressive phenotype, potentially due to epithelial denudation and crypt shortening associated with disease. These results suggest that relative overexpression of VIM in the colonic mucosa may herald the development of CAC. The data are integrated into a qualitative model for CAC progression in figure 7. Figure 7 Proposed model of involvement of intermediate filament (IF) proteins in pathogenesis of dysplastic changes in ulcerative colitis (UC). A small number of patients despite an initial acute episode have long-standing quiescent disease process; these patients are at a lower risk of colorectal cancer (CRC) in spite of having a long-standing disease process. On the contrary, recurrent bouts of acute inflammation and inadequate recovery/restoration of keratins following the acute inflammation may contribute to pathogenesis of CRC. Some patients may exhibit inflammation even after restoration of keratin, and so re-enter the disease path (red dashed line).

ng disease process. On the contrary, recurrent bouts of acute inflammation and inadequate recovery/restoration of keratins following the acute inflammation may contribute to pathogenesis of CRC. Some patients may exhibit inflammation even after restoration of keratin, and so re-enter the disease path (red dashed line). This is the first study investigating changes in insoluble IF levels in the mucosa in patients with UC with differing disease phenotype. Unlike previous studies which have focused on the soluble fraction of proteins, we have undertaken subcellular fractionation to investigate changes in the relatively insoluble fraction which, under physiological states, constitutes the bulk of the keratin pool. Our data suggest a model for the pathogenesis of CAC whereby acute inflammation reduces keratin levels and affects mucosal IF protein integrity which lags behind apparent clinical, microscopic and endoscopic recovery. Persistent failure of such recovery may be the cornerstone of pathogenesis of CAC. Additionally, the role of altered VIM:K8 ratio is significant and it might have a potential role as a mucosal marker of progressive disease. Nevertheless, prospective studies are needed, as are development of targeted strategies to modulate IF expression in the colonic mucosa. Supplementary Material Supplementary Materials This study was funded by a grant from the Bardhan Research and Educational Trust, and by EPSRC (EP/E036252/1).

This is the first study investigating changes in insoluble IF levels in the mucosa in patients with UC with differing disease phenotype. Unlike previous studies which have focused on the soluble fraction of proteins, we have undertaken subcellular fractionation to investigate changes in the relatively insoluble fraction which, under physiological states, constitutes the bulk of the keratin pool. Our data suggest a model for the pathogenesis of CAC whereby acute inflammation reduces keratin levels and affects mucosal IF protein integrity which lags behind apparent clinical, microscopic and endoscopic recovery. Persistent failure of such recovery may be the cornerstone of pathogenesis of CAC. Additionally, the role of altered VIM:K8 ratio is significant and it might have a potential role as a mucosal marker of progressive disease. Nevertheless, prospective studies are needed, as are development of targeted strategies to modulate IF expression in the colonic mucosa. Supplementary Material Supplementary Materials This study was funded by a grant from the Bardhan Research and Educational Trust, and by EPSRC (EP/E036252/1). Contributors: AJL conceived the study, wrote the funding bid, directed the research including taking clinical responsibility. BMC co-conceived the study, co-wrote the funding bid, directed and supervised the laboratory component, and undertook some of the bioinformatics analysis. CAE conceived the proteomic strategies, undertook all the iTRAQ work, including first-pass data analysis. DM co-wrote the funding bid, recruited and consented patients and undertook sample preparation, western blotting and some of the secondary data analysis. SJC reviewed IHC methods and strategy and took clinical responsibility for the histopathology. JC undertook all label-free proteomics, analysed the primary data and undertook data visualisation. AMRM undertook all immunohistochemistry and primary scoring, and analysed the data. AA undertook all dot-blotting, quantifications and analysed the data as well as sample preparation for label-free proteomics. DM, AA, AMRM, CAE and JC wrote sections of the paper in their areas of contribution. DM, additionally, drafted other sections. BMC organised and wrote a substantive draft of the paper. All authors contributed to the review process and have had sight of the final version.

as sample preparation for label-free proteomics. DM, AA, AMRM, CAE and JC wrote sections of the paper in their areas of contribution. DM, additionally, drafted other sections. BMC organised and wrote a substantive draft of the paper. All authors contributed to the review process and have had sight of the final version. Funding: This study was funded by a grant from the Bardhan Research and Educational Trust and by EPSRC (EP/E036252/1). Competing interests: None declared. Ethics approval: South Yorkshire Research Ethics Committee. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: Proteomic data will be made available through online repositories at point of acceptance and publication. Other data sets will be made available on request to the lead author.

Summary box What is already known about this subject? Oesophageal squamous cell carcinoma (OSCC) is the most common type of oesophageal cancer worldwide and is extremely prevalent in certain delineated locations such as across the African Rift Valley. Malawi has the highest incidence of OSCC in the world. Despite this, there is a relative lack of data in the African and particularly Malawian setting including characterising endoscopic findings and used treatment. What are the new findings? This is the first study of its type to characterise endoscopic data for one of the principal endoscopy sites in Malawi. This confirms the large predominance of OSCC as a diagnostic finding at endoscopy, validating the emerging importance that is being placed on this disease. Our population cohort had a predominance of OSCC located in the upper oesophagus compared with Africa-wide data and a female predominance in this upper OSCC cohort. Likely due to resource availability, bougie dilatation of OSCC predominated as a palliative measure over stent placement and was safe in this setting. How might it impact on clinical practice in the foreseeable future? This validates the increasing importance that OSCC is having on Malawian and international health agendas. This is the first study of its type to characterise this cohort including their demographics and disease characteristics. Bougie dilatation is common and safe in the Malawian setting for OSCC.

How might it impact on clinical practice in the foreseeable future? This validates the increasing importance that OSCC is having on Malawian and international health agendas. This is the first study of its type to characterise this cohort including their demographics and disease characteristics. Bougie dilatation is common and safe in the Malawian setting for OSCC. Introduction There is wide geographical variation in the incidence of many cancers which may reflect aetiological risk factors; in particular in the case of oesophageal squamous cell carcinoma (OSCC), there is a strikingly high incidence in the African Rift valley, the so-called African oesophageal SCC corridor1 extending from Sudan to the Eastern Cape Province of South Africa and including Malawi. Malawi ranks number one in the world for the incidence of oesophageal cancer with a weighted age standardised rate (ASR) of 24.2 (world ASR=5.9).2 Malawi also ranks as one of the poorest countries in the world, ranked second lowest in income (after Burundi) in the world with a per capita gross national income of US$320 per annum in 20173—which will likely be affected in 2018 by the anticipated erratic rainfall and armyworm infestation on agriculture production.

2 Malawi also ranks as one of the poorest countries in the world, ranked second lowest in income (after Burundi) in the world with a per capita gross national income of US$320 per annum in 20173—which will likely be affected in 2018 by the anticipated erratic rainfall and armyworm infestation on agriculture production. Unfortunately, OSCC has particularly poor outcomes in low-income settings, where the disease often presents too late for therapeutic intervention, with only 13% of patients in sub-Saharan Africa being surgical candidates and far fewer being able to have access to or afford such a treatment.4 As such a diagnosis is generally terminal, OSCC in such settings has been neglected from research agendas and it is widely accepted more work is needed in this area to further understand the risks and disease development in this context.5 6 The endoscopy unit at the Queen Elizabeth Central Hospital in Blantyre has a single procedure room and is a World Gastroenterology Organisation International Endoscopy Training Centre forming the centre of a hub-and-spoke endoscopy training programme for the three other central hospitals in the country. It performs around 1200 upper gastrointestinal endoscopies per annum, of which 300 are therapeutic—mainly dilatation or stenting of SCC and banding of oesophageal varices. All cases are recorded in a bespoke electronic recording system. In view of the enormous burden of oesophageal SCC that we see here, we audited the data from the endoscopy unit to analyse clinical and endoscopic features of SCC presenting over 5 years.

y dilatation or stenting of SCC and banding of oesophageal varices. All cases are recorded in a bespoke electronic recording system. In view of the enormous burden of oesophageal SCC that we see here, we audited the data from the endoscopy unit to analyse clinical and endoscopic features of SCC presenting over 5 years. Methods The endoscopy computer database is a Microsoft Access non-relational database recording patient demographics, endoscopy findings, therapeutic intervention and recommendations for treatment. All endoscopy records held in the departmental computer database for a 5-year period (2013–2017) were examined and any associated histology reports sought. Repeat endoscopies on the same individual were eliminated. Surgical and Oncology databases were also interrogated for supplementary routine clinical details of patients with SCC during the same time period, but direct linkage was impossible in the absence of hospital numbers, consistent spelling of names or date of birth (hospital numbers are not assigned, many patients do not know their date of birth, or even age, and spelling of names is variable). The data were analysed using Statistical Package for Statistical Sciences (SPSS/PC+ V.3.1) to undertake one-way analysis of variance and regression analysis between the age and gender of patients stratified according to the distance of the tumour from the teeth/gums (upper <25 cm, mid 25–29.9 cm, lower 30–34.9 cm and GOJ (gastro-oesophageal junction) ≥35 cm) using Student-Newman-Keuls (SNK) multiple range testing for significant differences in sorted means depending on the number of steps between the two means being tested, and χ2 test for gender differences stratified by distance, with Kendall’s Tau C as a measure of association for ordinal variables.

al junction) ≥35 cm) using Student-Newman-Keuls (SNK) multiple range testing for significant differences in sorted means depending on the number of steps between the two means being tested, and χ2 test for gender differences stratified by distance, with Kendall’s Tau C as a measure of association for ordinal variables. Results Of a total of 5882 endoscopy records, 1586 patients (27%) with an endoscopic diagnosis of oesophageal cancer were identified, 68% of procedures carried out by four of the authors (PJF, AK, JM, LM). The youngest patient was 17 years old and the oldest 100 years, median age 55 years. The presenting symptom was dysphagia in 84% of cases. Results are summarised in table 1. Table 1 Oesophageal cancer data stratified by site of tumour (upper <25 cm, mid 25–29.9 cm, lower 30–34.9 cm and GOJ ≥35 cm) Site Upper Mid Lower GOJ All number at each site 349 449 488 300 1586 percent distribution at each site 22% 28% 31% 19% 100% Age at each site Median 49 52 52 55 55 IQR 24 22 23 21 23 Minimum age 22 18 17 24 17 Maximum age 87 95 94 100 100 Female 186 205 202 123 716 % Female 53.3% 45.7% 41.4% 41.0% 45.1% Histology at each site Squamous carcinoma 39 51 68 33 191 Undifferentiated carcinoma 1 2 3 Squamous dysplasia 4 12 6 5 27 Adenocarcinoma 6 6 Non-malignant 4 8 16 8 36 All 47 72 92 52 263 Statistically distinct identities indicated by the box outlines. GOJ, gastro-oesophageal junction.

Site Upper Mid Lower GOJ All number at each site 349 449 488 300 1586 percent distribution at each site 22% 28% 31% 19% 100% Age at each site Median 49 52 52 55 55 IQR 24 22 23 21 23 Minimum age 22 18 17 24 17 Maximum age 87 95 94 100 100 Female 186 205 202 123 716 % Female 53.3% 45.7% 41.4% 41.0% 45.1% Histology at each site Squamous carcinoma 39 51 68 33 191 Undifferentiated carcinoma 1 2 3 Squamous dysplasia 4 12 6 5 27 Adenocarcinoma 6 6 Non-malignant 4 8 16 8 36 All 47 72 92 52 263 Statistically distinct identities indicated by the box outlines. GOJ, gastro-oesophageal junction. Analysis of variance showed highly significant differences in the age of patients stratified according to the site of the tumour (F3,1582=6.7, p<0.001), and SNK range testing revealed this was due to the ‘Upper’ oesophagus group being significantly (p<0.05) different from all the other groups which were statistically indistinguishable—the identity difference is indicated in table 1 by the boxes around the median ages for each site group. Likewise, there were highly significant differences in the gender of patients stratified according to the site of the tumour (F3,1582=4.8, p<0.01) and SNK range testing revealed this was due to the ‘Upper’ oesophagus having more women than all the other groups with the same identity differences as for age. Multiple regression analysis revealed a highly significant association between age and gender, and site: F2,1583=14.6, p<0.001 with a positive association between age and distance of the cancer down the oesophagus and more men with increasing distance down the oesophagus. Analysis of variance showed no gender differences when stratified by age: F1,1584=0.28, p=0.6.

significant association between age and gender, and site: F2,1583=14.6, p<0.001 with a positive association between age and distance of the cancer down the oesophagus and more men with increasing distance down the oesophagus. Analysis of variance showed no gender differences when stratified by age: F1,1584=0.28, p=0.6. Chi-squared analysis confirmed significant gender differences stratified by site of the tumour (χ2=14.3, df 3, p<0.01), and Kendall’s Tau C confirmed the female gender was associated with higher tumours (τc=−0.10, p<0.001), 53.3% women in the ‘Upper’ oesophagus group and 41.0% in the ‘GOJ’ group. A total of 263 histology reports were found. Most patients underwent biopsy at the time of endoscopy when it was technically feasible, but unfortunately, histology was only available for a limited proportion of the 5-year period in the Ministry of Health due to staff shortages and lack of reagents for staining and so on. When a histologist was available for reporting, patients were asked to make a contribution (~US$14) to fund the reagents and many chose not to. Likewise, at times a histology report was only available in the private sector at a cost of around US$41. The vast majority (83%, n=218) of tumours were shown to be due to squamous cancer/dysplasia and only 2% (n=6) due to adenocarcinoma. A number of biopsies (14%, n=36) showed non-malignant pathology (ulceration, slough, missed tumour etc), although the endoscopic diagnosis was clearly that of a tumour.

t of around US$41. The vast majority (83%, n=218) of tumours were shown to be due to squamous cancer/dysplasia and only 2% (n=6) due to adenocarcinoma. A number of biopsies (14%, n=36) showed non-malignant pathology (ulceration, slough, missed tumour etc), although the endoscopic diagnosis was clearly that of a tumour. Thirty-nine per cent (620 patients) underwent bougie dilatation of their tumour for symptom relief, 11% (179 patients) had placement of a self-expanding metal stent (only sporadically available in our hospital), and one patient had alcohol injection of the tumour for debulking. Two perforations were identified after bougie dilatation and were managed conservatively. One per cent (17 patients) underwent an Ivor Lewis oesophagectomy with end-to-end anastomosis and 1% (22 patients) had palliative gastrostomy tubes inserted. Seventeen per cent (274 patients) received chemotherapy. Radiotherapy is not available at present anywhere in the country. We have no outcome data for any of these patients, in the absence of a Hospital Records System.

hagectomy with end-to-end anastomosis and 1% (22 patients) had palliative gastrostomy tubes inserted. Seventeen per cent (274 patients) received chemotherapy. Radiotherapy is not available at present anywhere in the country. We have no outcome data for any of these patients, in the absence of a Hospital Records System. Discussion Lamentably, data on OSCC in sub-Saharan Africa are scarce and this is particularly true in Malawi which has the highest incidence of OSCC in the world.2 This study validates the observation that OSCC predominates in sub-Saharan Africa in Malawi over other forms of oesophageal carcinoma, the mechanism for which is still debated. One previous study has prospectively evaluated oesophageal carcinoma in this context,7 but it recruited much smaller numbers over a smaller period (143 patients over 9 months vs 1586 patients over 60 months) and was primarily aimed at following up expandable oesophageal stent placement rather than a descriptive assessment of all oesophageal cancers. Compared with Africa-wide data, our population was similar though had a predominance of tumours anatomically located in the upper oesophagus (22% vs <20% for oesophageal cancer in a previous estimated prevalence study across Africa in 2012).8 Otherwise, the anatomical distribution seemed broadly similar (28% in the mid-oesophagus for ours vs 30%–70% in Africa in 2012,8 50% in the lower oesophagus/GOJ vs 20%–50% in Africa in 2012).8

d in the upper oesophagus (22% vs <20% for oesophageal cancer in a previous estimated prevalence study across Africa in 2012).8 Otherwise, the anatomical distribution seemed broadly similar (28% in the mid-oesophagus for ours vs 30%–70% in Africa in 2012,8 50% in the lower oesophagus/GOJ vs 20%–50% in Africa in 2012).8 We noted a female preponderance in the upper third of the oesophagus, but a male preponderance overall which accords with previous Africa-wide data in 20128 (though there ares heterogeneous demographic data across the continent). The relatively young age of patients with OSCC at diagnosis compared with other high-incidence locations is consistent with other Africa-wide data; for example, 8% of OSCC cases in the Bomet district of West Kenya were under 30 years of age at diagnosis.9 Although smoking and alcohol play a prominent role in a higher-income context, these are unlikely to be causative for the observed prevalence in the high-incidence (low-resource) areas.10 11 First exposure to these risk factors does not reflect the observed OSCC disease prevalence; there are often similar rates of OSCC in men and women despite significantly different exposures to smoking and alcohol.12 Furthermore, exposure to these risk factors are not prominent practices in other high-prevalence areas such as China and repeatedly been shown not to be a major risk factor in OSCC development.10

; there are often similar rates of OSCC in men and women despite significantly different exposures to smoking and alcohol.12 Furthermore, exposure to these risk factors are not prominent practices in other high-prevalence areas such as China and repeatedly been shown not to be a major risk factor in OSCC development.10 There are many putative mechanisms for the observed high OSCC prevalence in this context, and a compelling potential cause includes fumonisin exposure—a mutagenic mycotoxin found on maize and associated with high OSCC rates.13–16 In the Malawian context, this may be driven by the cultural and financial reliance of maize as the predominant dietary constituent, as well as changes in traditional methods of storage of maize—from nkhokwe (well-ventilated grain silos on stilts) to indoor storage because of security from theft driven by extreme poverty—which facilitates fungal growth. Over-reliance on maize on a stable may also cause micronutrient deficiencies particularly due to the cooking methods in this context (though this is controversial)8 and due to silica contamination.17

s on stilts) to indoor storage because of security from theft driven by extreme poverty—which facilitates fungal growth. Over-reliance on maize on a stable may also cause micronutrient deficiencies particularly due to the cooking methods in this context (though this is controversial)8 and due to silica contamination.17 However, other potential reasons include thermal damage to oesophageal mucosa (by hot drinks such as tea, but also the hot maize porridge ‘phala’; the young age of consumption and exposure may explain the young presentation of OSCC) as well as micronutrient including selenium deficiencies in an estimated 80% of the Malawian population due to poor crop uptake of selenium in the low pH soils,18 as well as exposure to polycyclic aromatic hydrocarbons from cook fire smoke in the home (particularly when lower socioeconomic groups are forced to use poorly combustible fuels due to financial constraints).19 20 Ultimately, the causative mechanism has not been proven and ultimately remains unclear. As risk factors may disproportionately affect OSCC risk in different parts of the oesophageal mucosa,21 the female and younger individual preponderance in the ‘Upper’ oesophagus group in our setting may be attributable to different behaviours and therefore exposures (such as thermal damage) based on gender and age. However, this is speculative as it may also reflect a different underlying population, and therefore remains to be confirmed.

nger individual preponderance in the ‘Upper’ oesophagus group in our setting may be attributable to different behaviours and therefore exposures (such as thermal damage) based on gender and age. However, this is speculative as it may also reflect a different underlying population, and therefore remains to be confirmed. Though palliative stenting has good efficacy in our setting (<4% complication rate and a median survival of 210 days),7 the cost is unfortunately prohibitive for the health service and for patients, and as such, bougie dilation is often undertaken. There is a paucity of evidence of long-term outcomes for this procedure though itseems effective in other settings22 and carries a risk of malignant perforation.23 The strengths of these data include the large patient numbers and the long period the audit was undertaken, which enabled the detection of a distinct group of patients who have a higher risk of suffering from tumours higher in the oesophageal tract (ie, younger age and female). However, though this institution serves a large area, this study was undertaken at a single site and was unable to provide histological confirmation in the majority of cases given local funding constraints.

ts who have a higher risk of suffering from tumours higher in the oesophageal tract (ie, younger age and female). However, though this institution serves a large area, this study was undertaken at a single site and was unable to provide histological confirmation in the majority of cases given local funding constraints. Conclusion We present retrospective endoscopic data of oesophageal cancers in the Malawian context, which validate the predominance of OSCC in this context. We assert this study demonstrates the need for prioritisation of OSCC on the national health agenda, including consideration of characterising the causative mechanisms, promoting preventative strategies and a cost-effectiveness assessment for a radiotherapy facility. Preventative strategies could take the form of joint public health strategies, for example if cook fire is validated as a risk factor, there already exist strategies to reduce domestic exposure due to its pulmonary sequelae,24 and micronutrient deficiency correction would have benefits beyond OSCC.25 We look forward to further data in this context which further characterise the risk factors. We acknowledge the contribution of Dr Paul O’Toole in setting up the Endoscopy Access database and Professor Eric Borgstein in obtaining the historical surgical data. MAG and LM contributed equally as senior author.

Conclusion We present retrospective endoscopic data of oesophageal cancers in the Malawian context, which validate the predominance of OSCC in this context. We assert this study demonstrates the need for prioritisation of OSCC on the national health agenda, including consideration of characterising the causative mechanisms, promoting preventative strategies and a cost-effectiveness assessment for a radiotherapy facility. Preventative strategies could take the form of joint public health strategies, for example if cook fire is validated as a risk factor, there already exist strategies to reduce domestic exposure due to its pulmonary sequelae,24 and micronutrient deficiency correction would have benefits beyond OSCC.25 We look forward to further data in this context which further characterise the risk factors. We acknowledge the contribution of Dr Paul O’Toole in setting up the Endoscopy Access database and Professor Eric Borgstein in obtaining the historical surgical data. MAG and LM contributed equally as senior author. Contributors: PJF conceived the study, acquired the data and carried out statistical analysis. JDC wrote the manuscript and made amendments. PJF, AK, JM and LM performed the majority of the endoscopies. MAG leads the endoscopy training centre, which provided training, equipment and quality assurance, and which also developed the database audit. Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Competing interests: None declared. Patient consent: Not required.

Contributors: PJF conceived the study, acquired the data and carried out statistical analysis. JDC wrote the manuscript and made amendments. PJF, AK, JM and LM performed the majority of the endoscopies. MAG leads the endoscopy training centre, which provided training, equipment and quality assurance, and which also developed the database audit. Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Competing interests: None declared. Patient consent: Not required. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Summary box What is already known about this subject? Laparotomy is one of the most common emergency surgical interventions with higher postoperative morbidity and mortality than elective procedures. In elective surgery, these outcomes can be supplemented by patient-reported outcome measures (PROMs), but they have not been used routinely for emergency admissions. While the feasibility of asking emergency laparotomy patients to recall their pre-admission health status has been demonstrated, their likelihood of responding to a mailed postdischarge questionnaire is unknown. What are the new findings? PROMs can be successfully collected in patients 3 months after emergency laparotomy with a response rate of 74% using mailed follow-up. Most patients have regained their prior level of gastrointestinal health and their general health also improved. How might it impact on clinical practice in the foreseeable future? PROMs offer the opportunity for routinely assessing the impact of treatment from the patient’s perspective. Meaningful comparisons of surgeons and hospitals based on PROMs could be undertaken to supplement clinical measures such as mortality, morbidity and complications. Introduction In England, 40% of National Health Service (NHS) hospital admissions and 18% of procedures are emergencies.1 2 In 2016/2017, there were about 1.96 million procedures conducted by general surgeons for digestive tract conditions (excluding appendicectomy) of which 116 000 (6%) were emergency operations.3 Emergency laparotomy has a higher postoperative morbidity and mortality than elective procedures.4

ocedures are emergencies.1 2 In 2016/2017, there were about 1.96 million procedures conducted by general surgeons for digestive tract conditions (excluding appendicectomy) of which 116 000 (6%) were emergency operations.3 Emergency laparotomy has a higher postoperative morbidity and mortality than elective procedures.4 If the aim of healthcare is to restore a patient to his or her full potential, we need to be able to compare patients’ outcomes with their health status before the sudden and unexpected event that led to their emergency admission. Patient-reported outcomes measures (PROMs) are one of the ways to measure effectiveness and to determine the benefit of resources spent.5 6 PROMs are self-reported questionnaires designed to be completed by patients to capture their health at specific points in time to detect a health change over a period. They are multidimensional measures which may cover symptoms, functional status or health-related quality of life.6 It is known that short-term clinical outcomes, such as morbidity and mortality, following emergency surgical care vary significantly between hospitals.7 8 In contrast, little is known about the longer-term health status of those who survive who make up the vast majority of patients. Capturing PROMs would provide an additional means of routinely assessing the effectiveness of emergency surgical care. Currently, we know little about whether PROMs for emergency surgery vary between hospitals and whether there is any unwarranted variation.

f those who survive who make up the vast majority of patients. Capturing PROMs would provide an additional means of routinely assessing the effectiveness of emergency surgical care. Currently, we know little about whether PROMs for emergency surgery vary between hospitals and whether there is any unwarranted variation. There is minimal existing research about the feasibility of collecting routine follow-up PROMs from patients who have completed a PROM during their inpatient episode. The relevance of the available evidence is unclear as studies either involved only a few centres or were restricted to protocol-driven intervention trials instead of routine use.9 10 In addition, studies were mostly conducted in other countries, so the results may not be applicable in England.11–18 Response rates ranged between 51% and 71% for mailed questionnaires, and between 51% and 84% for interviewer-administered questionnaires. The only attempt in England to collect PROMs in multiple sites involved 28 major trauma centres and achieved about a 50% response rate using mailed or online follow-up at 6 months (personal communication: Antoinette Edwards).

or mailed questionnaires, and between 51% and 84% for interviewer-administered questionnaires. The only attempt in England to collect PROMs in multiple sites involved 28 major trauma centres and achieved about a 50% response rate using mailed or online follow-up at 6 months (personal communication: Antoinette Edwards). The reliability of patients’ recalling their prior health status via the use of a retrospective PROM has been demonstrated in six studies mostly in the USA.19 20 To determine the feasibility of employing PROMs in emergency admissions, we undertook two exploratory studies, one on a medical condition and the other in surgery (emergency laparotomy). Patients’ recollected state of health prior to their admission was collected shortly after their laparotomy but before discharge from hospital to provide a baseline assessment. The feasibility of recruiting representative samples of patients has already been reported.21 This paper reports on the follow-up response rate for patients, identifies any response biases and explores the feasibility of comparing patients’ outcome at 3 months with their retrospectively collected PROMs at baseline.

The reliability of patients’ recalling their prior health status via the use of a retrospective PROM has been demonstrated in six studies mostly in the USA.19 20 To determine the feasibility of employing PROMs in emergency admissions, we undertook two exploratory studies, one on a medical condition and the other in surgery (emergency laparotomy). Patients’ recollected state of health prior to their admission was collected shortly after their laparotomy but before discharge from hospital to provide a baseline assessment. The feasibility of recruiting representative samples of patients has already been reported.21 This paper reports on the follow-up response rate for patients, identifies any response biases and explores the feasibility of comparing patients’ outcome at 3 months with their retrospectively collected PROMs at baseline. Methods Site and patient recruitment A multisite study was carried out to ensure there would be variation in the administration of patient recruitment and data collection. This would allow us to gain insights into the relative merits of recruiting in different settings and with different personnel involved. Fourteen hospitals were selected, on the basis of their high case ascertainment rates in the National Emergency Laparotomy Audit (NELA), of which 13 agreed to participate and 11 successfully recruited patients for the 15-week duration of the study.

recruiting in different settings and with different personnel involved. Fourteen hospitals were selected, on the basis of their high case ascertainment rates in the National Emergency Laparotomy Audit (NELA), of which 13 agreed to participate and 11 successfully recruited patients for the 15-week duration of the study. Patients who met the NELA inclusion criteria and were alive at discharge were eligible for inclusion in this study unless they were not literate in English, deemed not to have sufficient cognitive ability or were not residents in the UK. For NELA, all patients over the age of 18 years, having a general surgical emergency laparotomy in all NHS hospitals in England and Wales, are eligible for inclusion and are enrolled on a prospective basis into the audit. The inclusion criteria for the audit aims to include all emergency gastrointestinal procedures on the stomach, large and small bowel, for conditions such as perforation, bleeding, abdominal abscess or obstruction, via open or laparoscopic approaches. Emergency laparotomies following elective surgical complications are also included. Patients requiring vascular surgery, gynaecological surgery, surgery on the renal tract, appendicectomy for appendicitis and laparotomy following trauma are excluded from the audit.22

or obstruction, via open or laparoscopic approaches. Emergency laparotomies following elective surgical complications are also included. Patients requiring vascular surgery, gynaecological surgery, surgery on the renal tract, appendicectomy for appendicitis and laparotomy following trauma are excluded from the audit.22 Patients were invited to participate after surgery, before discharge and as close to the discharge date as possible to ensure the immediate effects of the intervention (such as a general anaesthetic and immediate postoperative complications including ileus, respiratory depression and side effects of opioids) were minimised to ensure that the patients were medically able to complete the questionnaire.21 Clinical staff explained the study to patients, provided written information and obtained written consent. Questionnaires recalling their pre-admission baseline health status were completed by patients without assistance from staff or family except when they were impeded by physical disability or sensory impairment. Three-month follow-up Patients were mailed a follow-up questionnaire (QF) from the London School of Hygiene and Tropical Medicine 12 weeks (84 days) after their date of admission to hospital. Patient vital status was first checked against the Personal Demographics Service at NHS Digital prior to sending a follow-up questionnaire. After 2 weeks, non-responders were sent a reminder questionnaire.

m the London School of Hygiene and Tropical Medicine 12 weeks (84 days) after their date of admission to hospital. Patient vital status was first checked against the Personal Demographics Service at NHS Digital prior to sending a follow-up questionnaire. After 2 weeks, non-responders were sent a reminder questionnaire. Questionnaires The questionnaires completed during the admission included demographic information, self-reported comorbidities, a disease-specific PROM and a generic PROM. Patients were asked to recall how they were a month before their current admission. A systematic review identified suitable PROMs with adequate psychometric properties. Clinicians were then consulted in an unstructured meeting (a formal consensus development method was not used) to determine the final choice. This included consideration of the length and likely burden on patients of instruments.

sion. A systematic review identified suitable PROMs with adequate psychometric properties. Clinicians were then consulted in an unstructured meeting (a formal consensus development method was not used) to determine the final choice. This included consideration of the length and likely burden on patients of instruments. The disease-specific PROM was the Gastrointestinal Quality of Life Index (GIQLI), developed by Eypasch and colleagues.23 It consists of 36 questions relating to the gastrointestinal system and the impact of symptoms and treatment on individuals’ physical, emotional and social status. It takes 5–10 min to complete and has good test–retest reliability (intraclass correlation coefficient 0.92) and internal consistency (Cronbach’s alpha >0.90). The GIQLI is the most commonly used validated GI system–specific PROM for studies investigating outcomes in emergency abdominal surgery. The GIQLI score provides a global index score from 0 (poor health) to 144 (excellent health). The index score comprises four subscales: GIQLI symptoms (0–76), GIQLI physical score (0–28), social score (0–16) and emotion score (0–20) (online supplementary appendix 1). One item, on sex life, may not be applicable for some patients, but the option of such a response is not available. Despite this, some patients wrote ‘not applicable’ on their questionnaire. They were coded as ‘not at all’. 10.1136/bmjgast-2018-000238.supp1Supplementary data

The disease-specific PROM was the Gastrointestinal Quality of Life Index (GIQLI), developed by Eypasch and colleagues.23 It consists of 36 questions relating to the gastrointestinal system and the impact of symptoms and treatment on individuals’ physical, emotional and social status. It takes 5–10 min to complete and has good test–retest reliability (intraclass correlation coefficient 0.92) and internal consistency (Cronbach’s alpha >0.90). The GIQLI is the most commonly used validated GI system–specific PROM for studies investigating outcomes in emergency abdominal surgery. The GIQLI score provides a global index score from 0 (poor health) to 144 (excellent health). The index score comprises four subscales: GIQLI symptoms (0–76), GIQLI physical score (0–28), social score (0–16) and emotion score (0–20) (online supplementary appendix 1). One item, on sex life, may not be applicable for some patients, but the option of such a response is not available. Despite this, some patients wrote ‘not applicable’ on their questionnaire. They were coded as ‘not at all’. 10.1136/bmjgast-2018-000238.supp1Supplementary data A generic PROM was also included as it allows for comparisons across conditions and treatment groups and the potential to derive quality-adjusted life years for economic evaluations. The generic PROM used was the EQ-5D-3L, which has five items: mobility, usual activities, personal care, pain/discomfort and anxiety/depression. It takes up to 5 min to complete.24 For each of these questions, the respondent chooses from three responses indicating the level of their function. A multi-attribute utility score where death and perfect health are represented by 0 and 1 are calculated.25 Scores less than 0 are considered worse than death and one is the maximum score possible. The EQ-5D-3L was used rather than the EQ-5D-5L as the former is still the version used in the National PROMs Programme in England.

attribute utility score where death and perfect health are represented by 0 and 1 are calculated.25 Scores less than 0 are considered worse than death and one is the maximum score possible. The EQ-5D-3L was used rather than the EQ-5D-5L as the former is still the version used in the National PROMs Programme in England. Analysis Participating patients’ characteristics were summarised using means and SDs for continuous variables or percentages for binary variables. Response rates were calculated and reported for patients grouped by age, sex, living arrangements, socioeconomic status (SES), baseline GIQLI scores and baseline EQ-5D scores. SES was measured using the English Index of Multiple Deprivation (IMD) based on patients’ residential postcodes26 with patients assigned to quintiles of the national ranking of IMD scores. We conducted χ2 and paired t-test for differences to compare characteristics of participants who responded to the 3-month QF with those who did not. Patients’ outcomes at 3 months were compared with their baseline using paired t-test to assess evidence of change in health status. Change scores, with the 95% CI, were also used to describe reasonable limits on the extent of any change in order to assess whether the results were consistent with recovery to baseline (no change or an improvement in scores).

mpared with their baseline using paired t-test to assess evidence of change in health status. Change scores, with the 95% CI, were also used to describe reasonable limits on the extent of any change in order to assess whether the results were consistent with recovery to baseline (no change or an improvement in scores). Using the characteristics of those who were more likely to respond, we compared baseline and follow-up scores in these groups and quantified the potential impact this would have had on the mean health change. We then calculated what the health change would have been should the response rate be the same as recruitment proportions for the patient characteristics shown have a statistically significant non-response association. These calculations are estimations based on the assumption that non-responders would have reported similar PROM changes as responders, to help quantify the extent of influence non-response has on the change in PROM scores as an illustration of the potential impact. Results Response rates A total of 268 patients were recruited and completed baseline questionnaires (online supplementary appendix 2). Of these, 13 (4.9%) patients who were discharged from hospital then died during the postdischarge period before the follow-up contact. Of the 255 survivors, 190 patients (74.1%) responded to the follow-up PROM questionnaire: 146 responded to the first request and 44 after one reminder. The mean time between completing the baseline (Q1) and the QF was 85 (SD 19) days, and between admission and QF, 94 days.

Results Response rates A total of 268 patients were recruited and completed baseline questionnaires (online supplementary appendix 2). Of these, 13 (4.9%) patients who were discharged from hospital then died during the postdischarge period before the follow-up contact. Of the 255 survivors, 190 patients (74.1%) responded to the follow-up PROM questionnaire: 146 responded to the first request and 44 after one reminder. The mean time between completing the baseline (Q1) and the QF was 85 (SD 19) days, and between admission and QF, 94 days. Response bias Responders and non-responders were similar as regards comorbidities, living arrangements and health status (EQ-5D and GIQLI) (table 1). Responders differed from non-responders in three ways: they were older (mean age 65.0 (SD 16; range 18–91) vs 53.4 (SD 18; range 19–88); p<0.0001) (figure 1), more likely to be women and more likely to come from more affluent SES. Figure 1 Age distribution of responders and non-responders. PROM, patient-reported outcome measure. Table 1 Characteristics of responders compared with non-responders

Response bias Responders and non-responders were similar as regards comorbidities, living arrangements and health status (EQ-5D and GIQLI) (table 1). Responders differed from non-responders in three ways: they were older (mean age 65.0 (SD 16; range 18–91) vs 53.4 (SD 18; range 19–88); p<0.0001) (figure 1), more likely to be women and more likely to come from more affluent SES. Figure 1 Age distribution of responders and non-responders. PROM, patient-reported outcome measure. Table 1 Characteristics of responders compared with non-responders Patient characteristic Overall (n=255) Responders (n=189) Non-responders (n=66) P values* Sex Male 118 (46.0) 80 (42.3) 38 (57.6) 0.03 Female 137 (54.0) 109 (57.7) 28 (42.4) SES 1 (least deprived) 34 (14.8) 29 (17.1) 5 (8.3) 0.03 2 47 (20.4) 37 (21.0) 10 (16.7) 3 49 (23.3) 38 (22.3) 11 (18.3) 4 49 (21.3) 37 (21.8) 12 (20.0) 5 (most deprived) 51 (22.2) 29 (17.6) 22 (36.7) Missing 25 19 6 Comorbidities 0 58 (24.2) 37 (21.2) 21 (33.3) 0.186 1 78 (32.6) 64 (36.4) 14 (22.2) 2 44 (18.4) 33 (18.7) 11 (17.5) 3 32 (13.4) 22 (12.5) 10 (15.9) 4 or more 27 (11.3) 20 (11.4) 7 (11.1) Missing 16 13 3 Living arrangements With family 203 (79.6) 149 (82.3) 54 (84.3) 0.685 Alone 40 (15.7) 30 (16.5) 10 (15.6) Other 2 (0.8) 2 (1.1) 0 Missing 10 8 2 Mean EQ-5D (SD) 0.57 (0.40) 0.58 (0.39) 0.54 (0.42) 0.494 Missing 12 10 2 Mean GIQLI (SD), g 94.1 (31.3) 94.7 (31.4) 92.3 (31.04) 0.619 Missing 25 18 7 *From χ2. GIQLI, Gastrointestinal Quality of Life Index; SES, socioeconomic status.

Patient characteristic Overall (n=255) Responders (n=189) Non-responders (n=66) P values* Sex Male 118 (46.0) 80 (42.3) 38 (57.6) 0.03 Female 137 (54.0) 109 (57.7) 28 (42.4) SES 1 (least deprived) 34 (14.8) 29 (17.1) 5 (8.3) 0.03 2 47 (20.4) 37 (21.0) 10 (16.7) 3 49 (23.3) 38 (22.3) 11 (18.3) 4 49 (21.3) 37 (21.8) 12 (20.0) 5 (most deprived) 51 (22.2) 29 (17.6) 22 (36.7) Missing 25 19 6 Comorbidities 0 58 (24.2) 37 (21.2) 21 (33.3) 0.186 1 78 (32.6) 64 (36.4) 14 (22.2) 2 44 (18.4) 33 (18.7) 11 (17.5) 3 32 (13.4) 22 (12.5) 10 (15.9) 4 or more 27 (11.3) 20 (11.4) 7 (11.1) Missing 16 13 3 Living arrangements With family 203 (79.6) 149 (82.3) 54 (84.3) 0.685 Alone 40 (15.7) 30 (16.5) 10 (15.6) Other 2 (0.8) 2 (1.1) 0 Missing 10 8 2 Mean EQ-5D (SD) 0.57 (0.40) 0.58 (0.39) 0.54 (0.42) 0.494 Missing 12 10 2 Mean GIQLI (SD), g 94.1 (31.3) 94.7 (31.4) 92.3 (31.04) 0.619 Missing 25 18 7 *From χ2. GIQLI, Gastrointestinal Quality of Life Index; SES, socioeconomic status. Comparing change in PROM scores The distribution of the EQ-5D at baseline was bimodal, with the majority of patients above 0.5 and a smaller peak between −0.5 and 0.5 (figure 2). The distribution of the GIQLI score was broadly normal with a left skew. Figure 2 Baseline Gastrointestinal Quality of Life Index (GIQLI) score and baseline EQ-5D score distributions. Three months after surgery, patients’ GIQLI emotion score and GIQLI physical score had returned to the baseline score (table 2). The GIQLI symptom score had improved (51.9 vs 59.6; p<0.001), whereas the GIQLI social score had deteriorated (11.2 vs 9.8; p<0.001).

Figure 2 Baseline Gastrointestinal Quality of Life Index (GIQLI) score and baseline EQ-5D score distributions. Three months after surgery, patients’ GIQLI emotion score and GIQLI physical score had returned to the baseline score (table 2). The GIQLI symptom score had improved (51.9 vs 59.6; p<0.001), whereas the GIQLI social score had deteriorated (11.2 vs 9.8; p<0.001). Table 2 Comparison of baseline and follow-up PROM scores PROM Number with complete data Baseline (Q1) Mean (SE, 95% CI) Follow-up (QF) Mean (SE, 95% CI) Change (95% CI, p values) GIQLI 158 93.3 (2.55, 88.3 to 98.4) 97.9 (1.77, 94.4 to 101.4) +4.6 (0.37 to 8.83, 0.048) GIQLI symptom 168 52.0 (1.18, 49.6 to 54.2) 59.5 (0.76, 58.0 to 61.0) +7.5 (5.68 to 9.32, <0.0001) GIQLI emotion 177 12.0 (0.45, 11.12 to 12.9) 12.3 (0.35, 11.6 to 13.0) +0.3 (–0.43 to 1.04, 0.37) GIQLI physical 176 14.0 (0.61, 12.8 to 15.2) 13.3 (0.46, 12.4 to 14.2) −0.7 (–1.68 to 0.28, 0.18) GIQLI social 174 11.0 (0.34, 10.4 to 11.7) 9.8 (0.29, 9.27 to 10.4) −1.2 (–1.82 to –0.58, 0.0006) EQ-5D index 175 0.58 (0.03, 0.52 to 0.64) 0.64 (0.03, 0.59 to 0.69) +0.06 (0.00 to 0.12, 0.06) GIQLI, Gastrointestinal Quality of Life Index; PROM, patient-reported outcome measure; QF, follow-up questionnaire. The GIQLI score had improved (93.3 vs 97.9, p=0.048) and EQ-5D score had improved considerably (0.58 vs 0.64), although this difference was not quite statistically significant (p=0.06).

PROM Number with complete data Baseline (Q1) Mean (SE, 95% CI) Follow-up (QF) Mean (SE, 95% CI) Change (95% CI, p values) GIQLI 158 93.3 (2.55, 88.3 to 98.4) 97.9 (1.77, 94.4 to 101.4) +4.6 (0.37 to 8.83, 0.048) GIQLI symptom 168 52.0 (1.18, 49.6 to 54.2) 59.5 (0.76, 58.0 to 61.0) +7.5 (5.68 to 9.32, <0.0001) GIQLI emotion 177 12.0 (0.45, 11.12 to 12.9) 12.3 (0.35, 11.6 to 13.0) +0.3 (–0.43 to 1.04, 0.37) GIQLI physical 176 14.0 (0.61, 12.8 to 15.2) 13.3 (0.46, 12.4 to 14.2) −0.7 (–1.68 to 0.28, 0.18) GIQLI social 174 11.0 (0.34, 10.4 to 11.7) 9.8 (0.29, 9.27 to 10.4) −1.2 (–1.82 to –0.58, 0.0006) EQ-5D index 175 0.58 (0.03, 0.52 to 0.64) 0.64 (0.03, 0.59 to 0.69) +0.06 (0.00 to 0.12, 0.06) GIQLI, Gastrointestinal Quality of Life Index; PROM, patient-reported outcome measure; QF, follow-up questionnaire. The GIQLI score had improved (93.3 vs 97.9, p=0.048) and EQ-5D score had improved considerably (0.58 vs 0.64), although this difference was not quite statistically significant (p=0.06). Influence of non-response on change in health status Change in the GIQLI score and in the EQ-5D score was not associated with patients’ SES (table 3). However, change was greater in younger (under 70 years) and female patients though the differences did not reach statistical significance except for EQ-5D in women. Table 3 Change in PROMs scores by age, sex and SES

Influence of non-response on change in health status Change in the GIQLI score and in the EQ-5D score was not associated with patients’ SES (table 3). However, change was greater in younger (under 70 years) and female patients though the differences did not reach statistical significance except for EQ-5D in women. Table 3 Change in PROMs scores by age, sex and SES Patient characteristic Change in GIQLI (SD) (n=158) P values* Change in EQ-5D (SD) (n=160) P values* Age (years) >70 1.5 (25.0) 0.46 0.03 (0.36) 0.70 50–70 6.9 (25.8) 0.07 (0.37) <50 7.8 (39.3) 0.09 (0.50) Sex Male 2.46 (28.7) 0.43 −0.01 (0.40) 0.047 Female 6.14 (29.3) 0.11 (0.39) SES 1 (least deprived) 2.39 (23.7) 0.69 0.13 (0.32) 0.61 2 –0.75 (24.4) 0.47 (0.32) 3 3.94 (28.1) 0.04 (0.39) 4 4.52 (26.4) 0.11 (0.42) 5 (most deprived) 9.96 (27.5) –0.01 (0.49) *From ANOVA. GIQLI, Gastrointestinal Quality of Life Index; PROM, patient-reported outcome measure; SES, socioeconomic status.

Patient characteristic Change in GIQLI (SD) (n=158) P values* Change in EQ-5D (SD) (n=160) P values* Age (years) >70 1.5 (25.0) 0.46 0.03 (0.36) 0.70 50–70 6.9 (25.8) 0.07 (0.37) <50 7.8 (39.3) 0.09 (0.50) Sex Male 2.46 (28.7) 0.43 −0.01 (0.40) 0.047 Female 6.14 (29.3) 0.11 (0.39) SES 1 (least deprived) 2.39 (23.7) 0.69 0.13 (0.32) 0.61 2 –0.75 (24.4) 0.47 (0.32) 3 3.94 (28.1) 0.04 (0.39) 4 4.52 (26.4) 0.11 (0.42) 5 (most deprived) 9.96 (27.5) –0.01 (0.49) *From ANOVA. GIQLI, Gastrointestinal Quality of Life Index; PROM, patient-reported outcome measure; SES, socioeconomic status. Assessment of non-response bias Assessment of potential biases that might have been introduced by some patients not responding was based on the assumption that patients with similar baseline characteristics (sex and age) would have had similar follow-up EQ-5D or GIQLI scores. To illustrate the impact on non-response linked to sex and age, we estimated the mean change in GIQLI and EQ-5D scores had there been 100% follow-up response rate, compared with the observed mean changes. With this assumption, if responses were as per recruitment proportions by gender, the change in GIQLI would have been 4.55 (for all participants including non-responders) compared with 4.60 (observed in responders) and the mean change in EQ-5D would have been 0.055 compared with the observed mean change of 0.060. If responses were as per recruitment proportions by age, the change in GIQLI would have been 5.10 instead of 4.60, and the mean change in EQ-5D would have been 0.061 instead of 0.060.

Assessment of non-response bias Assessment of potential biases that might have been introduced by some patients not responding was based on the assumption that patients with similar baseline characteristics (sex and age) would have had similar follow-up EQ-5D or GIQLI scores. To illustrate the impact on non-response linked to sex and age, we estimated the mean change in GIQLI and EQ-5D scores had there been 100% follow-up response rate, compared with the observed mean changes. With this assumption, if responses were as per recruitment proportions by gender, the change in GIQLI would have been 4.55 (for all participants including non-responders) compared with 4.60 (observed in responders) and the mean change in EQ-5D would have been 0.055 compared with the observed mean change of 0.060. If responses were as per recruitment proportions by age, the change in GIQLI would have been 5.10 instead of 4.60, and the mean change in EQ-5D would have been 0.061 instead of 0.060. Discussion Main findings Retrospective and 3-month follow-up PROMs can be successfully collected in representative samples of patients undergoing emergency laparotomy surgery across the country with a response rate of 74% using mailed follow-up. Although responders and non-responders were similar with regards to their living arrangements, number of comorbidities and baseline health status, responders were more likely to be older, women and of a higher SES. The impact of any response bias appears to be slight. Response bias due to sex could overestimate the improvement in health status by 1% (0.05/4.45) on the GIQLI score and by 9% (0.005/0.060) on the EQ-5D index. In contrast, age bias may underestimate the improvement by 10% (0.5/4.6) on the GIQLI score and by 2% (0.001/0.060) on the EQ-5D.

ponse bias appears to be slight. Response bias due to sex could overestimate the improvement in health status by 1% (0.05/4.45) on the GIQLI score and by 9% (0.005/0.060) on the EQ-5D index. In contrast, age bias may underestimate the improvement by 10% (0.5/4.6) on the GIQLI score and by 2% (0.001/0.060) on the EQ-5D. The mean GIQLI had improved by 3 months from 93.3 to 97.9. This suggests that patients regain their prior level of GI health after major emergency surgery and there is an improvement compared with a month before their emergency admission. GIQLI symptom also improve, by eight when compared with baseline, though GIQLI social decreased by 1.3. Patients’ overall health status measured by the EQ-5D showed a considerable increase (0.58 vs 0.64), although this was not quite statistically significant.

ment compared with a month before their emergency admission. GIQLI symptom also improve, by eight when compared with baseline, though GIQLI social decreased by 1.3. Patients’ overall health status measured by the EQ-5D showed a considerable increase (0.58 vs 0.64), although this was not quite statistically significant. What this study adds This study has demonstrated the feasibility of collecting PROMs 3 months after emergency surgery among patients who, during their admission, had supplied retrospective accounts of the pre-event health status. It was shown that with high response rates, any responder bias is slight and will not undermine comparisons of providers. The higher response rates achieved in our study compared with a prior study in England9 may reflect the severity of emergency laparotomy as a subset of emergency surgical admissions. In elective surgery, higher response rates are observed with major procedures such as hip replacement than minor procedures such as inguinal hernia repair.27 In addition, mailing the questionnaires from a university rather than the hospital created the perception of independent assessment which may have encouraged participation. The observation that the GIQLI social score worsens despite the symptom score improving was unexpected. It may be that the use of retrospective reporting of preoperative symptoms exaggerates their severity though such a bias was not detected in studies of elective surgery.19 20 It could be that the GIQLI social score items require a longer recovery trajectory than GIQLI symptom items.

ptom score improving was unexpected. It may be that the use of retrospective reporting of preoperative symptoms exaggerates their severity though such a bias was not detected in studies of elective surgery.19 20 It could be that the GIQLI social score items require a longer recovery trajectory than GIQLI symptom items. The improvement of generic health status, as seen by the increase in EQ-5D, may reflect that emergency laparotomies are primarily performed in lifesaving situations; the improved health outcomes would imply that these procedures are lifesaving and restorative and also goes further and improves the quality of life of patients. This is not unsurprising, as a proportion of emergency laparotomies will be performed for conditions that may be associated with chronic symptoms prior to acute presentation (such as acute colonic perforation in diverticular disease). As such, recall of symptoms in the month preceding surgery may also encompass the impact of chronic disease. Strengths and limitations This is the first study of using retrospective PROMs to collect patients’ baseline health status and a 3-month follow-up for those admitted for emergency surgical operations in England. It was also conducted in multiple sites (11 hospital trusts) in different regions in England. This confirmed the feasibility of recruiting patients from diverse different geographical populations, as well as assessing PROMs use in different hospital organisational cultures and environments.

gical operations in England. It was also conducted in multiple sites (11 hospital trusts) in different regions in England. This confirmed the feasibility of recruiting patients from diverse different geographical populations, as well as assessing PROMs use in different hospital organisational cultures and environments. There are several limitations. First, some patients did not respond to the GIQLI item on their sexual life as there was no option to report ‘not applicable’. This introduced some response bias. Second, some patients (defined by their cognitive or literacy ability) were not eligible for inclusion in NELA so could not be included in this study. Third, PROMs have not been widely used in emergency surgery and their psychometric properties (eg, inter-rater reliability) in such patients have not been demonstrated. Among the disease-specific PROMs with adequate measurement properties, the GIQLI is the most commonly used. There is a need for further research in the systematic development of PROMs for use in emergency admissions, including psychometric testing for use in emergency laparotomy. Fourth, although not the subject of this paper, determining the outcome of emergency laparotomy using PROMs will inevitably depend on the validity of patients’ recall of their pre-event health status. This has been addressed in a previous paper.21 Finally, only one follow-up was conducted so there is little insight into the recovery trajectory of emergency laparotomy patients. Additional follow-ups should be considered in future studies.

Fourth, although not the subject of this paper, determining the outcome of emergency laparotomy using PROMs will inevitably depend on the validity of patients’ recall of their pre-event health status. This has been addressed in a previous paper.21 Finally, only one follow-up was conducted so there is little insight into the recovery trajectory of emergency laparotomy patients. Additional follow-ups should be considered in future studies. Conclusion This approach assesses from the patient’s perspective the impact of emergency laparotomy treatment. It also offers an insight into the opportunity for assessing other hospital admissions that are emergencies. The generalisability of these findings needs to be investigated with research on other causes of emergency admissions. Further research is needed in a larger sample of patients to explore longer-term outcomes enabling mapping of recovery trajectories. In addition, by capturing clinical data on patients (eg, P-POSSUM scores), such as by linkage to national clinical audit data, it would be possible to determine any association with indications for surgery, diagnosis and severity. This would be essential to be able to make meaningful comparisons of hospitals’ outcomes and to ensure the PROMs data could support clinical decisions.

SUM scores), such as by linkage to national clinical audit data, it would be possible to determine any association with indications for surgery, diagnosis and severity. This would be essential to be able to make meaningful comparisons of hospitals’ outcomes and to ensure the PROMs data could support clinical decisions. Routine collection of PROMs in emergency admissions could be feasible by their inclusion in national clinical audits. Such data would enhance quality improvement by including, alongside clinical outcomes, information on patients’ views of their symptoms, functional status and quality of life. For patients undergoing emergency laparotomy, there is a paucity of information available on the longer-term functional outcomes. Evidence obtained from PROMs can help inform shared decision-making before undertaking potentially high-risk surgery. We thank the patients and staff of 11 hospitals who participated in the study. Collaborators: NELA PROMs project team: Dave Murray, Mike Grocott, David Saunders, Jose Lourtie, Esther Kwong, Nick Black NELA PROMs Site and study leads: Michael Lewis, Gill Pout, Patricia Dickens, James Kirkby-Bott, Pauline Bartlett, Guy Titley, Emma Willett, Nina Barratt, Tanuja Shah, Kathleen Holding, Lianne Hufton, Veeranna Shatkar, Ruwan Weerakkody, Caron Baldwin, Sarah Hare, Annette Woods, Ewen Griffiths, Arlo Whitehouse, Jugdeep Dhesi, Jane Okello, Philip Braude, Karen Wilson, Kirsty Gibson, Abdul Quddus, Davina Ross-Anderson, Katherine MacGloin, Hasan Mukhtar, Kathryn Simpson, Kayleigh Gilbert

Kathleen Holding, Lianne Hufton, Veeranna Shatkar, Ruwan Weerakkody, Caron Baldwin, Sarah Hare, Annette Woods, Ewen Griffiths, Arlo Whitehouse, Jugdeep Dhesi, Jane Okello, Philip Braude, Karen Wilson, Kirsty Gibson, Abdul Quddus, Davina Ross-Anderson, Katherine MacGloin, Hasan Mukhtar, Kathryn Simpson, Kayleigh Gilbert Contributors: EK was the Principal Investigator of the study. NB is the Doctoral Supervisor. EK and NB wrote the paper with input from JN and DM. Funding: EK is funded by Economic and Social Research Council doctoral fellowship (grant reference: ES/J500021/1). DM has been funded by NELA for his time as National Clinical Lead (2012–2017) and Chair of the Project Team (since 2017). The research was supported by the National Institute for Health Research (NIHR) Collaboration for Leadership in Applied Health Research and Care North Thames at Barts Health NHS Trust. Disclaimer: The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. Competing interests: None declared. Patient consent: Obtained. Ethics approval: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and national research committee (NHS Health Research Authority) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. NHS ethical approval was obtained from South East Coast–Brighton and Sussex Research Ethics Committee (REC reference: 16/LO/2053) and was incorporated in the NIHR Research Network Portfolio.

(NHS Health Research Authority) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. NHS ethical approval was obtained from South East Coast–Brighton and Sussex Research Ethics Committee (REC reference: 16/LO/2053) and was incorporated in the NIHR Research Network Portfolio. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Introduction Acute liver decompensation is the main cause of hospitalisation in patients with cirrhosis, and it has been defined as the rapid development of at least one clinical complication between ascites, hepatic encephalopathy, gastrointestinal haemorrhage and bacterial infection.1 One of the keys to adequate management of hepatic decompensation is the prompt identification of its precipitating event, if any. This can be a direct liver injury (ie, a binge causing alcoholic hepatitis, drug-induced liver toxicity, superimposed viral hepatitis, portal vein thrombosis, ischaemia) or the consequence of systemic insults such as surgery, variceal bleeding or infection.2 In a significant proportion of patients (up to 43% of more severe cases1), the precipitant factor remains undetected.

lcoholic hepatitis, drug-induced liver toxicity, superimposed viral hepatitis, portal vein thrombosis, ischaemia) or the consequence of systemic insults such as surgery, variceal bleeding or infection.2 In a significant proportion of patients (up to 43% of more severe cases1), the precipitant factor remains undetected. Recent data from the European Association for the Study of the Liver-Chronic Liver Failure (EASL-CLIF) Consortium show that 3-month mortality of patients hospitalised for acute liver decompensation is 22%, while 9% undergo liver transplantation. In the subgroup of patients with more severe illness, that is those patients who present with or develop acute-on-chronic liver failure (ACLF), 3-month mortality reaches 51%.1 Unfortunately, there is no universally accepted definition of ACLF, thus precluding the acquisition of clear-cut epidemiological data. Recently, Jalan et al3 proposed the following working definition of ACLF: a syndrome occurring in patients with chronic liver disease, characterised by acute hepatic decompensation resulting in liver failure (jaundice and prolongation of the international normalised ratio (INR)) and failure of one or more extra-hepatic organs.

, Jalan et al3 proposed the following working definition of ACLF: a syndrome occurring in patients with chronic liver disease, characterised by acute hepatic decompensation resulting in liver failure (jaundice and prolongation of the international normalised ratio (INR)) and failure of one or more extra-hepatic organs. The pathophysiology of ACLF remains largely unknown. An important mechanism in the development of this syndrome seems to be altered host response to injury, with deregulated inflammation, and further predisposition to infection.2 In particular, systemic inflammatory response syndrome (SIRS), once established, is associated with increased inflammatory cytokine response, ACLF development and perpetuation, with establishment of a vicious cycle of inflammation-decompensation.4 For these reasons, prompt identification and adequate management of any precipitating events impinge on prognosis. We report the case of two brothers, one with alcohol-related and one with mixed aetiology cirrhosis, who developed a first episode of acute liver decompensation and ACLF in the absence of obvious precipitants. Cases In December 2013, a 44-year-old man presented to our department with a history of fever, vomiting, diarrhoea (up to 8 watery stools per day) and abdominal pain over a period of 10 days. He also reported cough, dysuria and rapid weight gain. His history was significant for smoking, previous hepatitis C virus (HCV) infection with spontaneous resolution and alcohol misuse, but he reported abstinence over the previous 15 days.

ea (up to 8 watery stools per day) and abdominal pain over a period of 10 days. He also reported cough, dysuria and rapid weight gain. His history was significant for smoking, previous hepatitis C virus (HCV) infection with spontaneous resolution and alcohol misuse, but he reported abstinence over the previous 15 days. Physical examination revealed fever (38°C), jaundice, rhonchi and wheezes over the middle right lung field, an enlarged liver, abdominal discomfort, mild ascites and flapping tremor. Blood pressure was 120/80 mm Hg and heart rate 130 bpm. Routine laboratory examination showed high white cell count with a relative increase in monocytes (1800 el/μL on a total of 13 040 el/μL leucocytes), elevated C reactive protein (54.7 mg/L) and abnormal liver function tests (aspartate aminotransferase (AST) 185 U/L, alanine aminotransferase (ALT) 54 U/L, bilirubin 162.1 μmol/L, γ-glutamyl transferase (GGT) 365 U/L, INR 1.53, ammonia 58 μmol/L). The chronic liver failure-sequential organ failure assessment (CLIF-SOFA) score was 7. Abdominal ultrasound confirmed cirrhosis (Model for End-Stage Liver Disease (MELD) score 20) and ascites, but explorative paracentesis was not feasible because of the small volume of fluid.

GGT) 365 U/L, INR 1.53, ammonia 58 μmol/L). The chronic liver failure-sequential organ failure assessment (CLIF-SOFA) score was 7. Abdominal ultrasound confirmed cirrhosis (Model for End-Stage Liver Disease (MELD) score 20) and ascites, but explorative paracentesis was not feasible because of the small volume of fluid. A diagnosis of ACLF and SIRS was made, and acute alcoholic hepatitis suspected; however, steroids were not started. Chest X-ray documented parahilar opacity in the right upper lobe, and urine dipstick was compatible with urinary tract infection. The patient was started on broad-spectrum antibiotics, with rapid remission of the abdominal pain. On day 11 after admission, diarrhoea and fever (up to 38.2°C) continued (figure 1, top panel). Faecal culture and assessment for Clostridium difficile toxins were negative. Figure 1 Time course of relevant laboratory and clinical variables in patient 1 (top panel) and patient 2 (bottom panel) during their hospital stay. Thick grey bars indicate episodes of diarrhoea, and the grey-shaded area indicates the period of treatment with ganciclovir in patient 1 (WCC, white cell count). Contrast-enhanced thoracic and abdominal computerised tomography was performed and showed a 2.5 cm nodule in the right upper lobe of the lung, 1 cm enlarged lymph nodes on mediastinal, axillary, abdominal and pelvic sites, and irregular liver margins. Echocardiography showed no evidence of endocarditis, and tuberculosis screening was negative.

c and abdominal computerised tomography was performed and showed a 2.5 cm nodule in the right upper lobe of the lung, 1 cm enlarged lymph nodes on mediastinal, axillary, abdominal and pelvic sites, and irregular liver margins. Echocardiography showed no evidence of endocarditis, and tuberculosis screening was negative. Extensive microbiological screening, to include HAV, HBV, HCV, HEV, Borrelia burgdorferi, Bartonella, Leptospira, Rubella, Chlamydia, Treponema pallidum, Mycoplasma, Plasmodium, HIV and cytomegalovirus (CMV), was performed. Of note, HAV-IgM was negative, HBV serology showed no signs of previous infection and HCV antibodies were positive with negative HCV-RNA. CMV-IgM was positive, and CMV-DNA viral load was 1473 copies/mL. Ganciclovir was started on day 20 at a dose of 5 mg/kg/die. Fever slowly lowered and disappeared on day 28. The patient was discharged with a MELD score of 14, no ascites and no symptoms. A positron emission tomography scan confirmed pulmonary malignancy, which was removed by thoracoscopic lobectomy; histology was compatible with adenocarcinoma and no radio/chemotherapy was needed. On 8-month follow-up, the patient remains well and substantially compensated from a liver standpoint (no ascites, MELD 15).

A positron emission tomography scan confirmed pulmonary malignancy, which was removed by thoracoscopic lobectomy; histology was compatible with adenocarcinoma and no radio/chemotherapy was needed. On 8-month follow-up, the patient remains well and substantially compensated from a liver standpoint (no ascites, MELD 15). In April 2014, this patient's 49-year-old brother, suffering from HCV and alcohol-related cirrhosis, was referred to our outpatient clinic for rapid development of tense ascites and ankle swelling. The patient reported progressive weight gain over the previous week (from 65 to 79 kg), accompanied by abdominal discomfort, diarrhoea and low-grade fever (up to 37.3°C). He was a smoker and had a history of chronic alcohol misuse, but he had been abstinent for 2 months, which was confirmed by relatives. He had precarious living and social conditions. He had been diagnosed with cirrhosis (MELD score 18) 1 month earlier, and abdominal computerised tomography had also shown considerable fatty infiltration and a minor amount of peri-hepatic ascites. Upper gastrointestinal endoscopy revealed duodenitis with duodenal erosions; no oesophageal varices were documented.

ons. He had been diagnosed with cirrhosis (MELD score 18) 1 month earlier, and abdominal computerised tomography had also shown considerable fatty infiltration and a minor amount of peri-hepatic ascites. Upper gastrointestinal endoscopy revealed duodenitis with duodenal erosions; no oesophageal varices were documented. Physical examination showed jaundice, moderate ascites and marked ankle swelling. Blood pressure was 125/70 mm Hg and heart rate 85 bpm. Laboratory examination showed high white cell count (15 090 el/µL, 43% lymphocytes), mild thrombocytopaenia (132.000/mL) and worsening of liver function (AST 110 U/L, ALT 65 U/L, INR 2.02 after vitamin K supplementation, bilirubin 147 μmol/L; MELD score 22). Renal function was normal. HCV load was 67 UI/mL. HAV, HBV and HEV serology was negative, with no signs of previous infection. The patient was admitted with a diagnosis of acute liver decompensation; an exploratory paracentesis showed no evidence of bacterial peritonitis, urine dipstick and blood cultures were negative. Abdominal ultrasound showed no signs of portal thrombosis; an echocardiogram showed normal cardiac function with minor increase in pulmonary pressure. The patient was treated with empiric antibiotic therapy (piperacillin and tazobactam) and parenteral diuretics (up to 700 mg/day of potassium canrenoate and 40 mg/day of furosemide, based on electrolytes). Nevertheless, his weight continued to increase and a 6 L paracentesis was performed on day 7 (figure 1, bottom panel).

patient was treated with empiric antibiotic therapy (piperacillin and tazobactam) and parenteral diuretics (up to 700 mg/day of potassium canrenoate and 40 mg/day of furosemide, based on electrolytes). Nevertheless, his weight continued to increase and a 6 L paracentesis was performed on day 7 (figure 1, bottom panel). Considering his recent family history, viral screening was performed soon after his hospital admission: CMV-DNA quantification was negative but CMV-IgM was positive, suggesting recent viral infection. Antiviral treatment was not started because of rapid viral clearance and unfavourable risk–benefit ratio, considering the tendency towards spontaneous resolution of symptoms: fever had resolved on day 5 and diarrhoea had progressively improved with no treatment. The patient was discharged on day 25 with no fever, no abdominal symptoms, a MELD score of 19 and a weight of 63 kg. He remains well on 90-day follow-up, with an MELD score of 18. Discussion CMV is a ubiquitous double-stranded DNA virus belonging to the Herpesviridae family. CMV infection can occur via blood or tissue exposure, or through close contact. Infection is diagnosed as the presence of at least one of the following: detection of CMV-DNA via culture or molecular techniques, seroconversion with the appearance of anti-CMV IgM antibodies or a fourfold increase in anti-CMV IgG titres. Active infection can be either the first infection in a naïve patient, a reactivation of an endogenous latent virus, or a reinfection by a different strain in an already infected patient.5

molecular techniques, seroconversion with the appearance of anti-CMV IgM antibodies or a fourfold increase in anti-CMV IgG titres. Active infection can be either the first infection in a naïve patient, a reactivation of an endogenous latent virus, or a reinfection by a different strain in an already infected patient.5 CMV disease is defined as a combination of infection, and clinical symptoms and signs. The spectrum of illness is strictly dependent on the host immune system: in the immunocompetent adult, primary CMV infection is most commonly subclinical and can sometimes result in a heterophile-negative mononucleosis syndrome, but it is, rarely, able to lead to severe organ-specific complications. Treatment is usually not warranted in immunocompetent patients with self-limiting illness. In the immunocompromised host, reactivation is more frequent and, as well as primary CMV infection and re-infection, is associated with significant morbidity and mortality. The spectrum of illness in this population can more easily vary from organ-specific impairment such as colitis, hepatitis and pneumonia, to a multisystem disorder. Treatment of these patients is usually initiated with intravenous ganciclovir until resolution of symptoms, often followed by a course of oral valganciclovir. Over the past few years, a grey zone has been defined between patients considered to be more at risk of infection, notably transplant recipients and HIV-infected patients, and the immunocompetent population: critically ill patients, such as those admitted to intensive care units, have shown a prevalence of active infection of up to 36%, with a negative influence on their prognosis.6 The extent of this grey zone has not been clearly defined. It has been shown that molecular signs of recent CMV reactivation are also relatively common in cirrhotic patients7 8 before transplantation, but infection has been suggested to be mild or asymptomatic.

up to 36%, with a negative influence on their prognosis.6 The extent of this grey zone has not been clearly defined. It has been shown that molecular signs of recent CMV reactivation are also relatively common in cirrhotic patients7 8 before transplantation, but infection has been suggested to be mild or asymptomatic. When considering the clinical cases presented here, the first doubt is whether CMV acted as a simple opportunistic ‘spectator’ or as an active pathogen with a direct influence on prognosis. In other words, whether our patients had CMV infection or CMV disease. Of note, biopsies were not performed, thus the diagnosis rested on clinical and laboratory findings. Both our patients showed symptoms (diarrhoea, fever and abdominal pain) that are typical of CMV colitis, along with decompensation of a previously stable cirrhosis. We can think of two different roles for the virus in our patients: (1) a primary precipitating illness; (2) a key component of the well-described vicious cycle in which an inflammatory response leads to immune deregulation and to a consequent infection, worsening inflammation.4 As for the first point, it is known that in a significant proportion of patients presenting with liver decompensation or ACLF, clinicians are unable to define a precipitating factor.1 Routine investigations in such patients are not usually as comprehensive as those performed in major immunocompromised states, such as the post-transplant setting. Our patients were not severely immunocompromised, but liver disease and alcohol misuse may have acted synergistically on reactivation and loss of control of the replicating virus. Cirrhosis is a condition in which depression and overstimulation of the immune system coexist,9 and alcohol misuse contributes to a further impairment in the immune response.

unocompromised, but liver disease and alcohol misuse may have acted synergistically on reactivation and loss of control of the replicating virus. Cirrhosis is a condition in which depression and overstimulation of the immune system coexist,9 and alcohol misuse contributes to a further impairment in the immune response. This is interesting when considering the potential for CMV infection or reactivation, for two main reasons: first, latent CMV control is strictly dependent on CD4+ and CD8+ T-cell function, which is deranged in cirrhosis as a result of high antigen load due to bacterial translocation,10 and natural killer cell function, which is impaired in relation to chronic alcohol consumption.11 Second, CMV reactivation is stimulated by three main pathways: (1) release of tumour necrosis factor α (TNFα), which binds to TNFα receptor on latently infected cells, with the activation of nuclear factor κB and initiation of viral replication; (2) release of inflammatory prostaglandins, with cyclic AMP (cAMP) activation and (3) catecholamine-mediated production of cAMP.5 TNFα production is enhanced both in liver failure and in chronic alcohol consumption, because of enhanced gut permeability and consequent lipopolysaccharide-mediated Kupffer cell activation, resulting in increased serum levels of TNFα and other pro-inflammatory cytochines.12 This latter mechanism is amplified in the setting of our second hypothesis, that is, viral reactivation as a result of immune deregulation caused by SIRS, which, in turn, leads to further worsening of the inflammatory state. This mechanism might have been in action in the first patient, in whom bacterial infection could have increased the predisposition to CMV reactivation, until viral infection became prevalent.

ivation as a result of immune deregulation caused by SIRS, which, in turn, leads to further worsening of the inflammatory state. This mechanism might have been in action in the first patient, in whom bacterial infection could have increased the predisposition to CMV reactivation, until viral infection became prevalent. The two patients differed in clinical severity (acute decompensation vs ACLF), concomitant precipitant events (none vs bacterial infection), and the type of management required for the CMV infection. The second patient was not treated with antivirals, as he had shown greater control of infection with rapid clearance of CMV-DNA and progressive amelioration of symptoms. The first patient was treated with antivirals, as he showed much poorer response to the infection and worse general status. This somehow parallels the intermediate position of our patients between severely immunocompromised patients, and the immunocompetent population. As far as treatment is concerned, it is also worth noting that in our first patient, clinical history could also suggest alcoholic hepatitis, potentially leading to deleterious initiation of steroids if the viral aetiology had not been considered.

severely immunocompromised patients, and the immunocompetent population. As far as treatment is concerned, it is also worth noting that in our first patient, clinical history could also suggest alcoholic hepatitis, potentially leading to deleterious initiation of steroids if the viral aetiology had not been considered. In conclusion, we have presented the cases of two brothers in whom CMV infection/reactivation may have played a crucial role in precipitating acute hepatic decompensation and ACLF, respectively. As identification and early treatment of precipitants impinge on the prognosis of both such conditions, it is our impression that screening for CMV may be worthy, especially when additional predisposing factors such as alcohol consumption are present. Twitter: Follow Sara Montagnese at @MontagneseSara Contributors: SR and VP contributed to drafting of the manuscript along with technical and material support. MC and AS contributed to patient management along with technical and material support. PAn and PAm contributed to critical revision of the manuscript for important intellectual content. SM contributed to patient management and drafting of the manuscript. Funding: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. Competing interests: None declared. Patient consent: Obtained. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Introduction Hepatitis C virus (HCV) genotype 4 (HCV-4) infection is highly prevalent in Africa and the Middle East, accounting for over 80% of patients with chronic HCV infection in many areas of this region.1 Globally, HCV-4 is responsible for 20% of the estimated 130–150 million patients with HCV infection.2 Treatment success using pegylated interferon and ribavirin (P/R), defined by achieving sustained virological response (SVR), is estimated at 53% for treatment-naïve patients with HCV-4 infection.3 This means that approximately half of the patients who undergo such regimen experience failure due to null-response or relapse. While newer direct-acting antiviral (DAA) medications have demonstrated remarkable efficacy in patients with HCV-1 infection who had experienced prior failure with P/R,4 their cost and accessibility pose barriers to rapid worldwide adoption.5 Therefore, in areas where newer therapies are limited, another course of P/R may be the only option at the present.

AA) medications have demonstrated remarkable efficacy in patients with HCV-1 infection who had experienced prior failure with P/R,4 their cost and accessibility pose barriers to rapid worldwide adoption.5 Therefore, in areas where newer therapies are limited, another course of P/R may be the only option at the present. To this effect, we conducted a meta-analysis and review of available literature to determine the SVR after re-treatment with P/R in treatment-experienced patients with HCV-4 infection. All published primary papers and abstracts regardless of study design were considered, and those fulfilling predetermined eligibility criteria were included. To the best of our knowledge, no such effort on this topic has been undertaken before. This meta-analysis adds to existing literature by providing a comprehensive evaluation of re-treatment efficacy using P/R in patients who have chronic HCV-4 infection and experienced prior failures with interferon (IFN)-based therapy. Methods We adhered to the PRISMA guidelines in the design, implementation, analysis and reporting of this meta-analysis.6 A review protocol was established during the design stage by the consensus of all authors involved, and was followed during the identification of eligible studies and data extraction (see online supplementary data).

d to the PRISMA guidelines in the design, implementation, analysis and reporting of this meta-analysis.6 A review protocol was established during the design stage by the consensus of all authors involved, and was followed during the identification of eligible studies and data extraction (see online supplementary data). Data identification We conducted a comprehensive search in the literature databases MEDLINE and EMBASE, last accessed 18 January 2014, using the search term: (‘genotype 4’). We excluded articles that were not in English; no other search restrictions were imposed. We also completed a manual search of abstracts from annual scientific meetings by the American Association for the Study of Liver Diseases (AASLD), the European Association for the Study of the Liver (EASL), the Digestive Diseases Weeks (DDW), and the Asian Pacific Association for the Study of Liver (APASL) between the years 2012 and 2013 using the terms: (‘experienced’, ‘re’, ‘prior’, ‘before’, ‘previously’, and ‘failed’). We manually reviewed the bibliographies of published studies for additional studies eligible for inclusion. Two authors (BZ and BEY) independently conducted all searches. Search sources and search terms were predetermined by the agreement of all authors prior to execution. Only published data extracted from eligible studies were included in this meta-analysis.

published studies for additional studies eligible for inclusion. Two authors (BZ and BEY) independently conducted all searches. Search sources and search terms were predetermined by the agreement of all authors prior to execution. Only published data extracted from eligible studies were included in this meta-analysis. Study eligibility Studies were considered for inclusion in our analysis if they were – original studies (i) featuring at least 10 adult (ii) treatment-experienced patients (iii) with chronic HCV-4 infection (iv) who have failed prior interferon-based therapy (either due to non-response or relapse) (v) and were subsequently retreated with another course of P/R combination therapy (vi). Studies were excluded if they included children or adolescents (i) or otherwise failed to satisfy the inclusion criteria (ii). Two authors (BZ and BEY) independently reviewed the titles and abstracts of relevant studies for eligibility. Duplicate publications were identified independently by cross-referencing the first and the last authors, collaboration groups, and the number of participants featured in each study. Any discrepancy was reviewed by a third review author (NHN) and resolved by consensus. Unless specifically stated by the included studies’ authors, patients were assumed to not be co-infected with hepatitis B virus (HBV)/HIV, or have any other liver-related diseases.

d the number of participants featured in each study. Any discrepancy was reviewed by a third review author (NHN) and resolved by consensus. Unless specifically stated by the included studies’ authors, patients were assumed to not be co-infected with hepatitis B virus (HBV)/HIV, or have any other liver-related diseases. Data extraction and statistical analysis We designed a data extraction form to record both the total number of patients and those who achieved SVR. We also recorded each study's design (prospective vs retrospective), study type (randomised controlled trial vs observational), intention-to-treat (ITT) analysis (yes or no), and country of origin. Baseline patient characteristics, treatment protocol, and treatment responses of each study were recorded, including age, sex, prior relapser versus non-responder, treatment duration, rapid virological response (RVR), early virological response (EVR), and SVR. The extraction form initially contained a number of possible prognosticators of treatment success, as discussed by Yee et al,3 but were removed after their omission from most articles we reviewed in-depth. Data extractions were performed by two authors (BZ and BEY).

nse (RVR), early virological response (EVR), and SVR. The extraction form initially contained a number of possible prognosticators of treatment success, as discussed by Yee et al,3 but were removed after their omission from most articles we reviewed in-depth. Data extractions were performed by two authors (BZ and BEY). Statistical analysis Our prespecified primary outcome is the probability of successfully re-treating chronic HCV-4 infection with a second course of P/R for the patients who had experienced failure with prior IFN-based therapy. Secondary outcomes included estimates of achieving SVR in prior relapsers versus non-responders, and the probability of achieving SVR in patients who achieve RVR or EVR. We obtained pooled event rates with corresponding 95% CIs using random-effects model (DerSimonian and Laird method), and inverse variance method.7 Study heterogeneity was determined with χ2-based Cochrane Q-statistic (p value set at ≤0.05) and I2 set at ≥50%, in accordance to the standard of quality for reporting meta-analysis from the Cochrane handbook.7 To assess for bias introduced by varying study characteristics in the primary analysis, we attempted univariate and multivariate random-effects meta-regression to identify heterogeneity in primary outcome. All statistical tests were two-sided. All analyses were performed using Comprehensive Meta-Analysis, V.2 (Biostat, Englewood, New Jersey, USA), and STATA 11 (StataCorp, College Station, Texas, USA).8–11

we attempted univariate and multivariate random-effects meta-regression to identify heterogeneity in primary outcome. All statistical tests were two-sided. All analyses were performed using Comprehensive Meta-Analysis, V.2 (Biostat, Englewood, New Jersey, USA), and STATA 11 (StataCorp, College Station, Texas, USA).8–11 Results Literature search results Our literature search identified 1798 studies (729 from MEDLINE and 1069 from EMBASE), and 14 648 abstracts from major liver society meetings (4327 from AASLD, 4204 from DDW, 3260 from APASL and 2857 from EASL) held between 2012 and 2013. One hundred and forty-eight studies (see online supplementary data) were selected for closer evaluation of eligibility and extractable data. Of these studies, 145 were excluded—121 for naïve or unknown prior treatment status, 15 for redundancy, 4 for non-extractable data, 3 for number of total patients less than 10, and 2 for including children and adolescents. After exclusion of aforementioned studies, two full-length articles and one abstract were included in the meta-analysis12–14 (figure 1). Figure 1 Flow chart of studies from initial search, and those meeting or failing to meet inclusion criteria for meta-analysis (AASLD, American Association for the Study of Liver Diseases; APASL, Asian Pacific Association for the Study of Liver; DDW, Digestive Diseases Weeks; EASL, European Association for the Study of the Liver; HCV-4, hepatitis C virus genotype 4).

ch, and those meeting or failing to meet inclusion criteria for meta-analysis (AASLD, American Association for the Study of Liver Diseases; APASL, Asian Pacific Association for the Study of Liver; DDW, Digestive Diseases Weeks; EASL, European Association for the Study of the Liver; HCV-4, hepatitis C virus genotype 4). Study characteristics Three studies with a total of 126 individuals were included in our pooled analysis (figure 1). Two studies originated from Saudi Arabia12 13 and another from Portugal.14 All three studies were retrospective in design. One study13 provided analysis based on ITT, with 37 of 56 (66%) patients actually completing treatment. Of the other two studies, 51 of 59 patients (86.4%) completed treatment in one,12 while the other14 did not disclose the rate of treatment completion. One study14 reported that all re-treated patients had previously failed regimens of standard IFN with or without ribavirin (RBV). Only one study12 provided baseline characteristics for treatment-experienced patients, including age (52.2±10.7; mean±SD), gender (male=36/59, or 61%), HBV/HIV co-infection (10/59, or 16.9%), and previous organ transplant (7/59, or 11.9%). Intended length of treatment was a minimum of 48 weeks in all three studies. Sustained viral response On basis of the three studies and a total of 126 patients with HCV-4 infection who received re-treatment with P/R, 41 achieved SVR, amounting to a pooled treatment success rate of 32.7% (CI 25.0% to 41.4%) (figure 2). Heterogeneity was not detected among these studies (Q-value=1.13, p=0.57, I2=0.00).

ned viral response On basis of the three studies and a total of 126 patients with HCV-4 infection who received re-treatment with P/R, 41 achieved SVR, amounting to a pooled treatment success rate of 32.7% (CI 25.0% to 41.4%) (figure 2). Heterogeneity was not detected among these studies (Q-value=1.13, p=0.57, I2=0.00). Figure 2 Pooled rate of sustained virological response. One study13 noted that SVR was higher in patients who experienced previous treatment failure due to relapse (50%) versus the non-responders (23.1%). Discussion As expected, the rate of achieving SVR with another course of P/R in treatment-experienced patients with HCV-4 infection who had failed prior IFN-based therapy was greatly diminished compared with treatment-naïve individuals,3 despite being slightly better than that previously reported for mixed genotypes.15 16 Prior relapsers were significantly more likely to achieve SVR, consistent with published findings, and this is even demonstrated in the re-treatment of patients with newer DAA therapeutics.4 15 Such high probability of re-treatment failure with P/R in patients with HCV-4 infection supports the use of DAA for re-treatment even in resource-limited areas, and especially for individuals who were non-responders to the initial course of IFN-based therapy. Our finding supports current practice guidelines issued by the EASL and the AASLD, which recommend the incorporation of DAA in the re-treatment of patients with HCV-4 infection who have failed the previous P/R therapy.17 18

d especially for individuals who were non-responders to the initial course of IFN-based therapy. Our finding supports current practice guidelines issued by the EASL and the AASLD, which recommend the incorporation of DAA in the re-treatment of patients with HCV-4 infection who have failed the previous P/R therapy.17 18 This study has several strengths. We performed an extensive literature search using two study databases, abstracts from four major liver conferences spanning 2 years, and reviewed articles’ bibliographies. No such comprehensive review on this topic has been published before. Despite the small number of studies included in analysis, the outcome lends evidence to the current EASL and AASLD recommendations even in resource-limited countries and regions.

er conferences spanning 2 years, and reviewed articles’ bibliographies. No such comprehensive review on this topic has been published before. Despite the small number of studies included in analysis, the outcome lends evidence to the current EASL and AASLD recommendations even in resource-limited countries and regions. A number of weaknesses also merit careful consideration, mostly stemming from paucity of available data. Despite our thorough review of literature, only three studies were identified and incorporated into the analysis. Only one of these studies reported baseline characteristics for the targeted patient population, preventing in-depth analysis of possible prognostic factors for SVR.12 The pooled estimate for the rate of re-treatment success with P/R between prior relapsers versus non-responders, which may vary significantly,13 could not be determined due to lack of primary data. Study patients were also heterogeneous with regard to types of IFN or use of RBV during prior therapies; nevertheless, pooled SVR was low even with inclusion of patients who may have had only standard IFN and without RBV. As with any comprehensive review, differences in the quality of studies and reporting exist, and patient populations differ across the included studies. Newer publications since the implementation of this meta-analysis are not considered.

en with inclusion of patients who may have had only standard IFN and without RBV. As with any comprehensive review, differences in the quality of studies and reporting exist, and patient populations differ across the included studies. Newer publications since the implementation of this meta-analysis are not considered. In conclusion, the result of our meta-analysis identified a low probability of success with repeated P/R combination therapy in patients with HCV-4 infection with prior IFN-based therapy failure, supporting the use of DAAs when re-treatment is considered even in resource-limited regions. Supplementary Material Supplementary Materials Contributors: BZ contributed in the study design, data collection, data analysis and interpretation, and in the drafting of the manuscript. NHN contributed in the study design, data analysis and interpretation, and in the drafting of the manuscript. BEY contributed in the study design, data collection and interpretation, and in the drafting of the manuscript. BY participated in the data collection. MHN contributed in the study concept and design, data analysis and interpretation, and in the critical revision of the manuscript. All the authors have critically reviewed the manuscript and approved its final version. Funding: This study was funded in part by the NIH National Center for Research Resources, TL1 training grants, 1TL1RR03197, to NHN and BZ. Competing interests: MHN has served as an advisory board member and has received research support from Gilead Sciences, Janssen Pharmaceuticals, and Roche Laboratories. Patient consent: Obtained.

Funding: This study was funded in part by the NIH National Center for Research Resources, TL1 training grants, 1TL1RR03197, to NHN and BZ. Competing interests: MHN has served as an advisory board member and has received research support from Gilead Sciences, Janssen Pharmaceuticals, and Roche Laboratories. Patient consent: Obtained. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Summary box What is already known about this subject? ▸  The change of hospitalisation of liver cirrhosis in time has been studied generally with a small number of patients in China. ▸  Evidence of the change of hospitalisation of liver cirrhosis as an outcome of the change of aetiological factors is relatively weak. ▸  Sex ratio of hospitalisation has not been well studied as different types of liver cirrhosis. What are the new findings? ▸  The pattern of hospitalisation of liver cirrhosis is changing accordingly with the change of aetiological factors. ▸  Hospitalisation rates of three different types of liver cirrhosis (viral hepatitis cirrhosis (VHC), non-viral hepatitis cirrhosis (NVHC) and alcoholic cirrhosis) are all higher for males than for females. ▸  Sex ratio of hospitalisation shrinks with time for VHC, but it widens with time for NVHC. How might this impact on clinical practice in the foreseeable future? ▸  The changing pattern of hospitalisation of liver cirrhosis in China warrants clinical attention. ▸  Increased knowledge of admission trends may allow early intervention and treatment for liver cirrhosis.

▸  Sex ratio of hospitalisation shrinks with time for VHC, but it widens with time for NVHC. How might this impact on clinical practice in the foreseeable future? ▸  The changing pattern of hospitalisation of liver cirrhosis in China warrants clinical attention. ▸  Increased knowledge of admission trends may allow early intervention and treatment for liver cirrhosis. Introduction Liver cirrhosis is defined as a diffuse process characterised by fibrosis and the conversion of normal liver architecture into structurally abnormal nodules,1 representing an advanced stage of chronic liver diseases.2 Of aetiological causes, chronic viral hepatitis and alcohol consumption are the two most common causes,3 4 not only accounting for the majority of cirrhosis, but also leading to an epidemic of the disease worldwide. Other causes of liver cirrhosis include autoimmune diseases,5 fatty liver diseases6 and several inherited metabolic disorders.6

chronic viral hepatitis and alcohol consumption are the two most common causes,3 4 not only accounting for the majority of cirrhosis, but also leading to an epidemic of the disease worldwide. Other causes of liver cirrhosis include autoimmune diseases,5 fatty liver diseases6 and several inherited metabolic disorders.6 With the rapidly growing economy, the aetiological factors of liver cirrhosis in China have also significantly changed over the past few decades. As reported, alcohol production rose from 7.11 million tons in 1984 to 31 million tons in 2001,7 and the average prevalence of weekly regular alcohol drinking in the Chinese population was as high as over 33% during 2004–2008.8 On the other hand, the prevalence of HBsAg for the population aged 1–59 years decreased from 9.8% in 1992 to 7.2% in 2006.9 As a result, the changes of the two factors in population may have influenced the hospitalisation trends of liver cirrhosis. Although some features of hospitalisation from cirrhosis in China had been previously studied, the results of those studies were generally based on a small sample size, and specific demography such as sex was not emphasised.10–12 In this study, we used the 2006–2010 data from 31 top-ranked hospitals in Beijing to evaluate the changes of hospitalisation of liver cirrhosis, involving more than 2.3 million hospitalised patients. This study, with its large sample, provides an opportunity to reliably analyse the trends of hospitalisation according to major types of liver cirrhosis and demographic groups.

nked hospitals in Beijing to evaluate the changes of hospitalisation of liver cirrhosis, involving more than 2.3 million hospitalised patients. This study, with its large sample, provides an opportunity to reliably analyse the trends of hospitalisation according to major types of liver cirrhosis and demographic groups. Methods Data source Data were obtained from the 2006 to 2010 hospitalisation summary reports (HSRs) in the 31 Grade 3A hospitals in Beijing. Hospitals in China are divided into 3 grades and 10 classes. The ranking for hospitals is according to infrastructure, level of services, and quality and safety of care. Grade 3A hospitals are those with the highest rank, and generally receive patients referred from smaller hospitals. Eligible hospitals in the study have at least 500 or more beds. To better describe the results, we excluded some special hospitals (infectious disease, gynaecology and chest). The HSR data in the hospitals were electronically submitted to the Beijing Municipal Health Bureau, through a centralised health information system, according to the administrative requirement of the Ministry of Health. The medical information on HSRs includes basic demographics, dates of admission and discharge, eight discharge diagnoses in Chinese and corresponding ICD-10-CM codes (one principal and seven supplementary diagnoses), treatments (mainly surgical information including date, coding, anaesthetist and surgeon), outcome of hospitalisations (survival status, drug allergy and hospitalisation infection) and financial costs.

ht discharge diagnoses in Chinese and corresponding ICD-10-CM codes (one principal and seven supplementary diagnoses), treatments (mainly surgical information including date, coding, anaesthetist and surgeon), outcome of hospitalisations (survival status, drug allergy and hospitalisation infection) and financial costs. Study patients We obtained 3 821 987 hospitalisation records for the years 2006–2010. Because the unit of analysis is an individual patient rather than a hospitalisation record when hospitalisation rate is calculated, we selected the first hospitalisation record of each patient as study record. After excluding readmission, a total of 2 517 628 hospitalised patients were included in the study. Furthermore, we also excluded inpatients who were younger than 20 years of age (N=227 484), and then had 2 290 144 hospitalised patients (988 178 men and 1 301 966 women). On the basis of the hospitalised patient population, we used ICD-10 K70.2, K70.3 and K74 codes to identify hospitalisation records on liver cirrhosis occurring in any of the eight listed diagnostic codes. A total of 26 691 records with liver cirrhosis were identified. To better conduct analyses, we excluded records with missing data on Chinese diagnostic name (N=53) and records with a diagnosis of schistosomiasis (N=90). After these exclusions, 26 548 hospitalised patients with liver cirrhosis remained (18 371 men and 8177 women).

26 691 records with liver cirrhosis were identified. To better conduct analyses, we excluded records with missing data on Chinese diagnostic name (N=53) and records with a diagnosis of schistosomiasis (N=90). After these exclusions, 26 548 hospitalised patients with liver cirrhosis remained (18 371 men and 8177 women). To explore the cause of trend, we distinguished viral hepatitis cirrhosis from non-viral hepatitis cirrhosis by examining whether patients had a diagnosis of viral hepatitis (ICD-10 codes: B15—B19, Z22.5) in any of the eight diagnoses. Lastly, we acquired 15 404 patients with viral hepatitis cirrhosis and 7621 with non-viral hepatitis cirrhosis. In addition, we also acquired 3523 hospitalised patients with non-viral hepatitis alcoholic cirrhosis (K70.2 and K70.3).

gnosis of viral hepatitis (ICD-10 codes: B15—B19, Z22.5) in any of the eight diagnoses. Lastly, we acquired 15 404 patients with viral hepatitis cirrhosis and 7621 with non-viral hepatitis cirrhosis. In addition, we also acquired 3523 hospitalised patients with non-viral hepatitis alcoholic cirrhosis (K70.2 and K70.3). Statistical analysis We used STATA software V.12.0 (StataCorp LP, College Station, Texas, USA) to analyse the changes of hospitalisation rates by time, age and gender. To address these changes, we used a Poisson regression model to estimate hospitalisation rate ratios (RR) and 95% CIs. In the Poisson model, the hospitalisation rate for the jth observed population is assumed to be given by. To fit models, we established a dataset of dependent variables (including the numbers of patients with viral hepatitis cirrhosis, non-viral hepatitis cirrhosis and non-viral hepatitis alcoholic cirrhosis (numerators), the numbers of hospitalised patients (denominator) and independent variables (including year, age and gender). Using the hospitalisation rate of year 2006, age group <40 years and female as reference, we estimated RRs and 95% CIs for the other years, age groups and gender.

al hepatitis alcoholic cirrhosis (numerators), the numbers of hospitalised patients (denominator) and independent variables (including year, age and gender). Using the hospitalisation rate of year 2006, age group <40 years and female as reference, we estimated RRs and 95% CIs for the other years, age groups and gender. In the Poisson models (), Cj is the jth dependent variable (the jth number of patient with liver cirrhosis), Ej is the jth number of hospitalisation and βj is the jth regression coefficient of dependent variables χk such as year (year 2007, 2008, 2009, or 2010 vs year 2006), age group (20–39, 40–49, 50–59, 60–69, or ≥70 years vs <40 years), and gender (male vs female). In STATA, the Option IRR of Poisson model was used to obtain adjusted RRs and 95% CIs, and the statistical significance of the RRs was determined by the Z test. A RR >1 with p value <0.05 indicates a statistically significant change of rate of hospitalisation for liver cirrhosis.

40 years), and gender (male vs female). In STATA, the Option IRR of Poisson model was used to obtain adjusted RRs and 95% CIs, and the statistical significance of the RRs was determined by the Z test. A RR >1 with p value <0.05 indicates a statistically significant change of rate of hospitalisation for liver cirrhosis. Results Approximately 2.3 million hospitalised patients were included in the analysis (table 1), of whom 15 404 were associated with viral hepatitis cirrhosis (6.73‰), 7621 with non-viral hepatitis cirrhosis (3.33‰) and 3523 with alcoholic cirrhosis (1.54‰). About half of the hospitalised patients with liver cirrhosis was due to viral hepatitis cirrhosis (58%), and the other half was due to non-viral hepatitis cirrhosis (28.7%) and alcoholic cirrhosis (13.3%). The most notable findings were that the rates of hospitalisation for viral hepatitis cirrhosis decrease with time and the rates for non-viral hepatitis cirrhosis and alcoholic cirrhosis increase with time. In addition, the rates of hospitalisation were significantly higher for males than for females in all three types of liver cirrhosis. Table 1 Hospitalisation rates for liver cirrhosis by year, age and gender among hospitalised patients in 31 top hospitals in Beijing

Results Approximately 2.3 million hospitalised patients were included in the analysis (table 1), of whom 15 404 were associated with viral hepatitis cirrhosis (6.73‰), 7621 with non-viral hepatitis cirrhosis (3.33‰) and 3523 with alcoholic cirrhosis (1.54‰). About half of the hospitalised patients with liver cirrhosis was due to viral hepatitis cirrhosis (58%), and the other half was due to non-viral hepatitis cirrhosis (28.7%) and alcoholic cirrhosis (13.3%). The most notable findings were that the rates of hospitalisation for viral hepatitis cirrhosis decrease with time and the rates for non-viral hepatitis cirrhosis and alcoholic cirrhosis increase with time. In addition, the rates of hospitalisation were significantly higher for males than for females in all three types of liver cirrhosis. Table 1 Hospitalisation rates for liver cirrhosis by year, age and gender among hospitalised patients in 31 top hospitals in Beijing Viral hepatitis cirrhosis Non-viral hepatitis cirrhosis Alcoholic cirrhosis Hospitalised patients (n) Patients (n) Rate (‰) Patients (n) Rate (‰) Patients (n) Rate (‰) Total 2 290 144 15 404 6.73 7621 3.33 3523 1.54 Year 2006 396 601 2846 7.18 1129 2.85 536 1.35 2007 432 335 2920 6.75 1208 2.79 584 1.35 2008 455 350 3106 6.82 1528 3.36 648 1.42 2009 489 527 3100 6.33 1713 3.50 806 1.65 2010 516 331 3432 6.65 2043 3.96 949 1.84 Age (years) 20–39 779 792 1821 2.34 726 0.93 343 0.44 40–49 372 264 3875 10.41 1378 3.70 1223 3.29 50–59 435 855 5072 11.64 2103 4.82 1234 2.83 60–69 325 334 2980 9.16 1596 4.91 460 1.41 ≥70 376 899 1656 4.39 1818 4.82 263 0.70 Gender Female 1 301 966 4475 3.44 3636 2.79 66 0.05 Male 988 178 10 929 11.06 3985 4.03 3457 3.50 Table 2 shows the comparisons of year, age and gender in the rate of hospitalisation according to three types of liver cirrhosis. Between 2006 and 2010, hospitalisation rates for viral hepatitis cirrhosis decreased by 10% (adjusted RR=0.90, 95% CI 0.86 to 0.95, p for trend <0.001), while rates for non-viral hepatitis cirrhosis and alcoholic cirrhosis inversely increased by 35% (adjusted RR=1.35, 1.26 to 1.46, p for trend <0.001) and by 33% (adjusted RR=1.33, 1.19 to 1.47, p for trend <0.001), respectively. Compared with the age group of 20–39 years, adjusted rate ratios significantly decreased with the increases in age (40–49 years to ≥70 years), from 3.58 (3.38 to 3.78) to 1.44 (1.35 to 1.54) in viral hepatitis cirrhosis (p for trend <0.001), and from 4.31 (3.82 to 4.86) to 0.83 (0.71 to 0.97) in alcoholic cirrhosis (p for trend <0.001). However, the age-associated decreasing trend was not seen in non-hepatitis cirrhosis, in which adjusted rate ratio reached a peak in the age group 50–59 years, and subsequently remained stable until 70 years. Additionally, adjusted sex ratio was 2.71 (2.62 to 2.81, p<0.001) for viral hepatitis cirrhosis, 1.14 (1.09 to 1.19, p<0.001) for non-viral hepatitis cirrhosis and 59.9 (46.9 to 76.5, p<0.001) for alcoholic cirrhosis.

te ratio reached a peak in the age group 50–59 years, and subsequently remained stable until 70 years. Additionally, adjusted sex ratio was 2.71 (2.62 to 2.81, p<0.001) for viral hepatitis cirrhosis, 1.14 (1.09 to 1.19, p<0.001) for non-viral hepatitis cirrhosis and 59.9 (46.9 to 76.5, p<0.001) for alcoholic cirrhosis. Table 2 Hospitalisation rate ratios and 95% CIs * for year, age and gender by types of liver cirrhosis Viral hepatitis cirrhosis Non-viral hepatitis cirrhosis Alcoholic cirrhosis Year 2006 Reference Reference Reference 2007 0.95 (0.90 to 1.00) 0.99 (0.91 to 1.08) 1.02 (0.90 to 1.14) 2008 0.95 (0.91 to 1.00) 1.18 (1.09 to 1.27) 1.07 (0.95 to 1.20) 2009 0.87 (0.83 to 0.92) 1.22 (1.13 to 1.31) 1.21 (1.08 to 1.35) 2010 0.90 (0.86 to 0.95) 1.35 (1.26 to 1.46) 1.33 (1.19 to 1.47) P trend <0.001 P trend <0.001 P trend <0.001 Age (years) 20–39 Reference Reference Reference 40–49 3.58 (3.38 to 3.78) 3.86 (3.52 to 4.22) 4.31 (3.82 to 4.86) 50–59 3.92 (3.72 to 4.14) 4.98 (4.58 to 5.43) 3.53 (3.13 to 3.98) 60–69 3.06 (2.88 to 3.24) 5.08 (4.65 to 5.55) 1.73 (1.51 to 1.99) ≥70 1.44 (1.35 to 1.54) 4.98 (4.57 to 5.44) 0.83 (0.71 to 0.97) P trend <0.001 P trend <0.001 P trend <0.001 Gender Female Reference Reference Reference Male 2.71 (2.62 to 2.81) 1.14 (1.09 to 1.19) 59.9 (46.9 to 76.5) p<0.001 p<0.001 p<0.001 *Hospitalisation rate ratios (RR) and 95% CIs in the table were estimated by Poisson regression models, and were mutually adjusted for the effects of year, age and gender. p Values for linear trend were obtained from modelling the continuous form of year or age variable.

.19) 59.9 (46.9 to 76.5) p<0.001 p<0.001 p<0.001 *Hospitalisation rate ratios (RR) and 95% CIs in the table were estimated by Poisson regression models, and were mutually adjusted for the effects of year, age and gender. p Values for linear trend were obtained from modelling the continuous form of year or age variable. Table 3 shows the results of sex ratio trend analysis for viral hepatitis cirrhosis and non-viral hepatitis cirrhosis by age group. Sex ratio trends were found mainly in the age groups under 60 years. The sex ratio for viral hepatitis cirrhosis was smaller in time compared with that for non-viral hepatitis cirrhosis. The sex ratio generally decreased with time in viral hepatitis cirrhosis, but increased in non-viral hepatitis cirrhosis. Between 2006 and 2010, the sex ratio for non-viral hepatitis cirrhosis rose by 37%, 17.7% and 48.8% in the age groups 20–39, 40–49 and 50–59 years, respectively. The age-associated sex ratio trend analysis was not conducted for alcoholic cirrhosis due to the small sample size of female patients (N=66). Table 3 Sex ratio trend analysis for viral hepatitis and non-viral hepatitis cirrhosis by age group

Table 3 shows the results of sex ratio trend analysis for viral hepatitis cirrhosis and non-viral hepatitis cirrhosis by age group. Sex ratio trends were found mainly in the age groups under 60 years. The sex ratio for viral hepatitis cirrhosis was smaller in time compared with that for non-viral hepatitis cirrhosis. The sex ratio generally decreased with time in viral hepatitis cirrhosis, but increased in non-viral hepatitis cirrhosis. Between 2006 and 2010, the sex ratio for non-viral hepatitis cirrhosis rose by 37%, 17.7% and 48.8% in the age groups 20–39, 40–49 and 50–59 years, respectively. The age-associated sex ratio trend analysis was not conducted for alcoholic cirrhosis due to the small sample size of female patients (N=66). Table 3 Sex ratio trend analysis for viral hepatitis and non-viral hepatitis cirrhosis by age group Viral hepatitis cirrhosis Non-viral hepatitis cirrhosis Male* Female* Sex ratio (95% CI)† Male Female Sex ratio (95% CI) 20–39 years 2006 8.35 0.58 14.4 (10.8 to 19.0) 1.59 0.42 3.81 (2.57 to 5.65) 2007 7.51 0.63 11.9 (9.2 to 15.4) 1.16 0.48 2.43 (1.64 to 3.58) 2008 7.69 0.46 16.6 (12.4 to 22.2) 2.33 0.40 5.80 (4.09 to 8.23) 2009 6.38 0.48 13.4 (10.1 to 17.8) 2.27 0.65 3.47 (2.59 to 4.66) 2010 5.97 0.49 12.1 (9.2 to 16.0) 2.83 0.54 5.22 (3.88 to 7.02) 40–49 years 2006 19.21 4.72 4.07 (3.43 to 4.84) 3.99 2.29 1.75 (1.32 to 2.31) 2007 17.99 4.77 3.77 (3.18 to 4.48) 4.22 2.54 1.66 (1.27 to 2.16) 2008 16.47 4.41 3.73 (3.14 to 4.44) 4.42 2.43 1.82 (1.41 to 2.36) 2009 14.89 3.60 4.13 (3.45 to 4.96) 4.88 2.74 1.78 (1.41 to 2.25) 2010 16.33 4.13 3.95 (3.35 to 4.67) 6.31 3.06 2.06 (1.67 to 2.54) 50–59 years 2006 17.54 7.31 2.40 (2.07 to 2.77) 3.88 4.74 0.82 (0.65 to 1.03) 2007 16.27 7.00 2.32 (2.02 to 2.68) 4.33 3.42 1.27 (1.01 to 1.59) 2008 16.02 7.56 2.12 (1.86 to 2.42) 5.64 4.95 1.14 (0.95 to 1.37) 2009 15.31 7.19 2.13 (1.87 to 2.42) 5.43 4.23 1.28 (1.07 to 1.54) 2010 15.03 6.99 2.15 (1.90 to 2.44) 6.01 4.92 1.22 (1.03 to 1.44) 60–69 years 2006 11.21 7.19 1.56 (1.31 to 1.86) 3.79 4.78 0.79 (0.62 to 1.02) 2007 10.99 7.32 1.50 (1.26 to 1.78) 3.63 4.53 0.80 (0.62 to 1.03) 2008 11.27 7.29 1.55 (1.31 to 1.83) 4.74 4.86 0.98 (0.78 to 1.22) 2009 10.46 7.05 1.48 (1.26 to 1.75) 4.98 5.50 0.91 (0.74 to 1.11) 2010 10.87 7.35 1.48 (1.27 to 1.72) 5.22 6.52 0.80 (0.66 to 0.97) ≥70 years 2006 4.53 3.55 1.28 (1.00 to 1.63) 3.21 5.33 0.60 (0.47 to 0.77) 2007 3.94 4.08 0.96 (0.76 to 1.22) 3.75 5.59 0.67 (0.54 to 0.84) 2008 5.61 3.97 1.41 (1.15 to 1.74) 3.81 6.21 0.61 (0.50 to 0.75) 2009 4.41 3.96 1.11 (0.90 to 1.38) 4.08 6.17 0.66 (0.54 to 0.80) 2010 5.42 3.93 1.38 (1.13 to 1.68) 3.76 6.85 0.55 (0.45 to 0.67) *The numbers listed in the table columns are the hospitalisation rates of liver cirrhosis (‰).

3.75 5.59 0.67 (0.54 to 0.84) 2008 5.61 3.97 1.41 (1.15 to 1.74) 3.81 6.21 0.61 (0.50 to 0.75) 2009 4.41 3.96 1.11 (0.90 to 1.38) 4.08 6.17 0.66 (0.54 to 0.80) 2010 5.42 3.93 1.38 (1.13 to 1.68) 3.76 6.85 0.55 (0.45 to 0.67) *The numbers listed in the table columns are the hospitalisation rates of liver cirrhosis (‰). †Male-to-female sex ratios and 95% CIs of hospitalisation rate were estimated by Poisson regression models. Discussion Using the HSRs from 31 top-ranking hospitals in Beijing, we found that between 2006 and 2010, hospitalisation rates for viral hepatitis cirrhosis slightly declined by 10%, but rates for non-viral hepatitis cirrhosis and alcoholic cirrhosis significantly grew by more than 30%. Male to female ratio was significantly higher for each type of cirrhosis, and the changes in the sex ratios in time were associated with the type of cirrhosis and age. While our observation time is very short, this large analysis provides reliable estimates for the trend of hospitalisation rates and sex ratio of liver cirrhosis among the Chinese population.

higher for each type of cirrhosis, and the changes in the sex ratios in time were associated with the type of cirrhosis and age. While our observation time is very short, this large analysis provides reliable estimates for the trend of hospitalisation rates and sex ratio of liver cirrhosis among the Chinese population. We found that chronic viral hepatitis remained the major cause of cirrhosis and accounted for 60% of all cirrhosis admissions. Because of the extremely high prevalence of hepatitis B in the Chinese population, the high hospitalisation rate of viral hepatitis cirrhosis may be associated mainly with hepatitis B rather than with the other types of viral hepatitis. A recent hospital study in Beijing reported that hepatitis B was responsible for the largest proportion (46%) of all hepatitis admissions, while hepatitis E was responsible for the second largest (32%) and the other types of hepatitis, A, C and D, were together responsible for the smallest proportion (2.1%).13 The study also reported that the numbers of admission from hepatitis A, B and E decreased, respectively, by 99%, 45% and 62% between 2002 and 2011.13 Except for the impacts of decrease of viral hepatitis, the decreased hospitalisation trend of viral hepatitis cirrhosis was also possibly affected by the effectiveness of antiviral therapy for hepatitis.14

mbers of admission from hepatitis A, B and E decreased, respectively, by 99%, 45% and 62% between 2002 and 2011.13 Except for the impacts of decrease of viral hepatitis, the decreased hospitalisation trend of viral hepatitis cirrhosis was also possibly affected by the effectiveness of antiviral therapy for hepatitis.14 A significant increase in hospitalisation for alcoholic cirrhosis during a short period of 5 years substantially reflects the status of alcohol-induced liver damage in China. The proportion of alcoholic cirrhosis in all liver cirrhosis was 14.8% in 2010, an increase of 24.4% as compared to that in 2006. The result in proportion was close to that in Japan 15 and Korea,16 but much lower than that in some western countries, where the proportion of alcohol cirrhosis ranged from 53% to 84%.17–19 Because the proportion of cirrhosis is only a relative statistic, it may be affected by the variation of proportion of other cirrhosis. For example, the proportion of alcoholic cirrhosis would increase relative to the decline of numbers of viral hepatitis cirrhosis.

ion of alcohol cirrhosis ranged from 53% to 84%.17–19 Because the proportion of cirrhosis is only a relative statistic, it may be affected by the variation of proportion of other cirrhosis. For example, the proportion of alcoholic cirrhosis would increase relative to the decline of numbers of viral hepatitis cirrhosis. We defined non-viral hepatitis cirrhosis in the study as non-viral and non-alcoholic hepatitis, to aetiologically distinguish the causes of the two other cirrhosis. However, our definition is ambiguous and, therefore, the trend of the disease may be affected by various causes. We noted that the temporal increased trend of non-viral hepatitis cirrhosis paralleled the non-alcoholic fatty liver disease (NAFLD) trend,20 21 as well as the drug-induced liver damage trend22 in the Chinese population. NAFLD is a common cause for cryptogenic cirrhosis,23 and its prevalence was found to be as high as 36% among adults in Beijing.24 On the other hand, it is well known that many drugs, including herbs, have hepatic toxicity and thus whether the use of these drugs is increasing and influences the increase of cirrhosis is worthy to speculate. A recent report showed that about 20–30% of patients with liver diseases had ever used herbal medicine, an increase of three to five times over the past decades.25

ncluding herbs, have hepatic toxicity and thus whether the use of these drugs is increasing and influences the increase of cirrhosis is worthy to speculate. A recent report showed that about 20–30% of patients with liver diseases had ever used herbal medicine, an increase of three to five times over the past decades.25 The hospitalisation for non-viral hepatitis cirrhosis seemed to increase more quickly for males than for females in time, especially for males aged 50–59 years. While the reasons for this are clearly uncertain, the authors hypothesise that some particular environmental agents may have influenced the trend of time. A suspicious agent is smoking. A few studies reported that smoking was associated with more advanced hepatic fibrosis in primary liver cirrhosis.26 27 In China, in 2010, the rate of smoking was 53% among men and 2.5% among women, but smoking habits among Chinese men have remained stable over time.28

the trend of time. A suspicious agent is smoking. A few studies reported that smoking was associated with more advanced hepatic fibrosis in primary liver cirrhosis.26 27 In China, in 2010, the rate of smoking was 53% among men and 2.5% among women, but smoking habits among Chinese men have remained stable over time.28 Patients with liver cirrhosis admitted were predominantly men. Overall, the male proportion in admission from cirrhosis was 98% for alcoholic, 71% for viral hepatitis and 52% of non-viral and non-alcoholic hepatitis. The sex-specific preponderance can be associated with the high frequencies of alcohol drinking and viral hepatitis infection among the male population. In our study, however, patients with non-viral hepatitis cirrhosis mainly consist of those with cryptogenic cirrhosis, including primary biliary cirrhosis. In western countries, primary biliary cirrhosis was recently found to have a high female preponderance and to affect middle-aged women.29 Female to male ratio for the disease was as high as 9:1–20:1.30 The female preponderance of primary biliary cirrhosis is uncertain but was recently found to be associated with smoking.26 Since smoking occurs predominantly among Chinese men, the factor may play a role in making gender preponderance different between China and western countries.

the disease was as high as 9:1–20:1.30 The female preponderance of primary biliary cirrhosis is uncertain but was recently found to be associated with smoking.26 Since smoking occurs predominantly among Chinese men, the factor may play a role in making gender preponderance different between China and western countries. Because of the large sample size and comprehensive representation of the entire population in Beijing, HSRs are particularly useful for studying the time trend of disease. However, despite quality control, the HSR data are still subject to measurement errors, including incomplete or inaccurate information recorded, as well as processing errors. As a result, misclassification of diseases may have occurred, especially of subtypes of liver cirrhosis. To avoid potential misclassification, we used three main types rather than subtypes of liver cirrhosis, for analysis. The main types of liver cirrhosis are specific aetiologically and, therefore, can help to identify trends and associated causes. Because HSRs are restricted only to hospitalised patients, cirrhosis patients who were not hospitalised could not be analysed. However, we infer that the defect, if it exists, would occur randomly and would not depend on year, age or gender. In this study, we did not describe any changes in the reasons for hospitalisation of liver cirrhosis, in the duration of hospitalisation stay and in the financial implications of hospitalisation.

analysed. However, we infer that the defect, if it exists, would occur randomly and would not depend on year, age or gender. In this study, we did not describe any changes in the reasons for hospitalisation of liver cirrhosis, in the duration of hospitalisation stay and in the financial implications of hospitalisation. Conclusion We conducted a large data-descriptive analysis for hospitalisation trend and sex ratio of liver cirrhosis in Beijing. Hospitalisation rate for liver cirrhosis has significantly changed in recent years. The patterns of change are consistent to the changing patterns of major risk factors such as viral hepatitis infection and alcoholic consumption in China. Increased knowledge of admission trends may allow early intervention for liver cirrhosis. Contributors: G-PY and X-YB designed the research, analysed the data and drafted the manuscript. All the authors interpreted the results. B-BX and Y-HH made critical revision of the manuscript. YL and Y-HH took the responsibility of administration and provided important suggestions. All the authors have seen and approved the final manuscript. Funding: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. Competing interests: None declared. Ethics approval: This study was considered exempt from ethical review by the IRB of Peking University since data were collected mainly for an administrative purpose and were used for a secondary data analysis. Provenance and peer review: Not commissioned; externally peer reviewed.

Funding: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. Competing interests: None declared. Ethics approval: This study was considered exempt from ethical review by the IRB of Peking University since data were collected mainly for an administrative purpose and were used for a secondary data analysis. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Summary box What is already known about this subject? ▸  Colorectal cancer (CRC) surgery is hampered by a 30-day mortality of approximately 5%. ▸  Hypoalbuminaemia (<35 g/L) in patients undergoing CRC surgery varies from 10% to 57%. ▸  Human serum albumin (HSA) below 35 g/L is associated with overall poor survival in patients with CRC, but the impact on short-term survival after CRC surgery is unknown. What are the new findings? ▸  Decrement of preoperative HSA was associated with a concentration-dependent increased risk of 30-day mortality following CRC surgery. ▸  Thirty-day crude mortality following CRC surgery increased from 2.0% among patients with HSA above 40 g/L to 26.9% among patients with HSA equal to or below 25 g/L. ▸  Compared with patients with HSA above 40 g/L, the 30-day mortality HR increased from 1.75 (95% CI 1.25 to 2.45) among patients with HSA 36–40 g/L to 7.59 (95% CI 4.95 to 11.64) among patients with HSA equal to or below 25 g/L. ▸  The negative prognostic impact associated with decrement of HSA persisted independently from other preoperative conditions, including presence of systemic inflammation. How might it impact on clinical practice in the foreseeable future? ▸  The study highlights the importance of optimise preoperative serum albumin concentration in patients undergoing CRC surgery and aids to evaluate patient's postoperative mortality risk.

▸  The negative prognostic impact associated with decrement of HSA persisted independently from other preoperative conditions, including presence of systemic inflammation. How might it impact on clinical practice in the foreseeable future? ▸  The study highlights the importance of optimise preoperative serum albumin concentration in patients undergoing CRC surgery and aids to evaluate patient's postoperative mortality risk. Introduction Surgery is the only potentially curable treatment for colorectal cancer (CRC), but it is hampered by an overall 30-day mortality of approximately 5%.1 2 Advanced cancer stage, presence of comorbidity, and old age at time of CRC surgery are conditions known to impair postoperative mortality.1 3 A common feature associated with these conditions is a decrease in human serum albumin (HSA) concentration.4 5 HSA is the main determinant of plasma oncotic pressure and is a multifunctional protein with a wide range of properties including antioxidant, immunomodulatory, and detoxification functions.6 A previous study showed that a decrease in HSA concentrations from greater than 46 g/L to less than 21 g/L was associated with an increase in 30-day mortality from 1% to 29% among American veterans undergoing major non-cardiac surgery.7 Although it is estimated that hypoalbuminaemia is present in 10–57% of patients undergoing CRC surgery,8–18 the prognostic impact of HSA among these patients has been sparsely examined.8–12 Particularly, no studies have examined the concentration-dependent association between HSA concentrations and mortality among patients with CRC, the prognostic impact among subgroups of patients with CRC, or have been able to properly control for potential confounding.8–12 Moreover, decrement of HSA is often associated with systemic inflammation4 and patients with CRC with hypoalbuminaemia (<35 g/L) and C reactive protein (CRP) concentration above 10 mg/L are at high risk of mortality.19 However, it is not clear if prognostic impact of HSA is independent from the degree of inflammation. Understanding the prognostic impact of HSA will aid with appropriate risk stratification, allocation of patients to intensive perioperative and postoperative care, and better knowledge to inform patients about postoperative risks. We therefore examined, in a population-based setting, the impact of preoperative HSA on 30-day mortality following CRC surgery overall and among patients with different preoperative conditions and CRP levels.

ensive perioperative and postoperative care, and better knowledge to inform patients about postoperative risks. We therefore examined, in a population-based setting, the impact of preoperative HSA on 30-day mortality following CRC surgery overall and among patients with different preoperative conditions and CRP levels. Materials and methods Setting We conducted this cohort study using prospectively collected data from medical registries in North and Central Denmark (with approximately 2.15 million inhabitants) from 1 January 1997 to 31 December 2011. Since 1968, a unique civil personal registration (CPR) number has been assigned to every Danish resident at birth or on immigration and allows accurate record linkage at the individual level among all Danish registries.20

ral Denmark (with approximately 2.15 million inhabitants) from 1 January 1997 to 31 December 2011. Since 1968, a unique civil personal registration (CPR) number has been assigned to every Danish resident at birth or on immigration and allows accurate record linkage at the individual level among all Danish registries.20 Patients with CRC Our study cohort included patients undergoing first-time CRC surgery in North and Central Denmark during the period 1997–2011. Patients with CRC were identified using the Danish Cancer Registry (DCR), which contains data on date of diagnosis, stage, and other information of incident cases of malignant neoplasms in Denmark since 1943.21 Tumours registered after 1 January 1978 have been reclassified according to the International Classification of Diseases (ICD), 10th revision (ICD-10). To obtain information on surgery, we linked these patients with CRC to the Danish National Patient Registry (DNPR). The DNPR records information from all hospitalisations since 1977 and from outpatient contacts since 1995.22 Each record includes the dates of hospital admission and discharge, up to 20 discharge diagnoses recorded according to the ICD-8 until 1993 and according to the ICD-10 thereafter, type of admission (non-elective or elective), and type and date of surgical procedures. Since 1996, surgical procedures have been coded according to the Nordic Medico-Statistical Committee (NOMESCO) Classification of Surgical Procedures.23 First-time CRC surgery was defined as the first procedure involving colorectal surgery performed during a hospitalisation where CRC was listed as a diagnosis in the DNPR. CRC surgery was categorised according to the intention of eradicating the primary tumour as radical resection and non-eradicative procedures.24 Radical resection included surgeries such as partial and total resections of the colon and/or rectum while non-eradicative procedures included colostomy, stent placement, or excision of a very small part of the colon. Radical resection was further divided into laparoscopic and open surgery. CRC stage was reported as localised (Dukes’ stage A or B), regionally spread (Dukes’ stage C), and metastasised. Surgeries were also categorised as elective and non-elective according to the type of admission.25 We restricted the study population to patients living in the study region at the time of the CRC surgery using data from the Civil Registration System (CRS).

A or B), regionally spread (Dukes’ stage C), and metastasised. Surgeries were also categorised as elective and non-elective according to the type of admission.25 We restricted the study population to patients living in the study region at the time of the CRC surgery using data from the Civil Registration System (CRS). This system is updated on a daily basis and tracks the vital status, marital status, and residence of all Danish residents.26 The study population was further restricted to patients undergoing CRC surgery in North Jutland after 1 January 1997, in Aarhus after 1 January 2000, in Viborg after 1 January 2005, and in Ringkjøbing after 1 January 2006, reflecting the availability of laboratory data (see online supplementary table S1).27 The laboratory database contains laboratory tests from inpatient stays, outpatient clinic visits, and visits to general practitioners.27 The National Health Service provides tax-funded medical care covering surgery for all Danish residents and all types of CRC surgery that were provided in the study region during the period investigated.

base contains laboratory tests from inpatient stays, outpatient clinic visits, and visits to general practitioners.27 The National Health Service provides tax-funded medical care covering surgery for all Danish residents and all types of CRC surgery that were provided in the study region during the period investigated. Preoperative albumin serum concentration For each patient undergoing CRC surgery, we searched the laboratory database for preoperative measurements of HSA (rounded to the nearest integer). Only measurements 1–30 days prior to surgery date were used. In the event of several preoperative HSA measurements within the same individual, we used the most recent one for analysis. Normal HSA interval was defined as 36–45 g/L for persons aged 40–70 years and 34–45 g/L for those older than 70 years. We categorised HSA levels as follows: ≤25 g/L (severe hypoalbuminaemia), >25 g/L and ≤30 g/L (moderate hypoalbuminaemia), >30 g/L and ≤35 g/L (mild hypoalbuminaemia), >35 g/L and ≤40 g/L (low normal albuminaemia), and >40 g/L (high normal albuminaemia).

sons aged 40–70 years and 34–45 g/L for those older than 70 years. We categorised HSA levels as follows: ≤25 g/L (severe hypoalbuminaemia), >25 g/L and ≤30 g/L (moderate hypoalbuminaemia), >30 g/L and ≤35 g/L (mild hypoalbuminaemia), >35 g/L and ≤40 g/L (low normal albuminaemia), and >40 g/L (high normal albuminaemia). Covariates We quantified patients’ burden of comorbidity using the Charlson Comorbidity Index (CCI) that includes 19 diseases, each assigned a score between one and six.28 Using diagnoses registered in the DNPR, we identified the diseases in the CCI at any time prior to or during the CRC surgery admission,29 excluding CRC and CRC metastases. We classified patients as having low (score=0), moderate (score=1–2), or high comorbidity level (score ≥3). We also identified patients with alcohol-related disease defined as alcohol abuse or alcohol-related diseases disregarding alcoholic liver disease.30 Information about marital status was obtained using the CRS.31 Using the laboratory database, we also collected data on other preoperative blood tests from 1 to 30 days prior to surgery including haemoglobin, sodium, potassium, leucocytes, creatinine, CRP, platelets, international normalised ratio (INR), bilirubin, and alanine aminotransferase. Finally, we computed the preoperative Model for End-stage Liver Disease (MELD) scores using values of creatinine, total bilirubin and INR.32 33 Diagnostic and surgical codes are provided in online supplementary appendix.

Using the laboratory database, we also collected data on other preoperative blood tests from 1 to 30 days prior to surgery including haemoglobin, sodium, potassium, leucocytes, creatinine, CRP, platelets, international normalised ratio (INR), bilirubin, and alanine aminotransferase. Finally, we computed the preoperative Model for End-stage Liver Disease (MELD) scores using values of creatinine, total bilirubin and INR.32 33 Diagnostic and surgical codes are provided in online supplementary appendix. Thirty-day mortality We followed patients from date of CRC surgery until death, day 30 postoperatively, or end of study, whichever came first. Information about date of death was obtained from the CRS. Statistical analysis We used the Kaplan-Meier method to compute 30-day postoperative mortality of patients with CRC in each HSA category.34 To quantify the excess mortality associated with the decrement of HSA, we computed absolute mortality difference with 95% CIs between patients with HSA above 40 g/L and patients with HSA equal to or below 25 g/L.35 We used Cox regression analysis to compute HRs with 95% CIs as a measure of the relative risk of postoperative mortality using patients with preoperative HSA above 40 g/L as reference and controlling for potential confounding factors.

ween patients with HSA above 40 g/L and patients with HSA equal to or below 25 g/L.35 We used Cox regression analysis to compute HRs with 95% CIs as a measure of the relative risk of postoperative mortality using patients with preoperative HSA above 40 g/L as reference and controlling for potential confounding factors. We stratified patients by elective versus non-elective hospital admissions, age category (0–59, 60–69, 70–79, and 80+ years), gender, cancer site (colon or rectum), cancer stage, type of surgery, comorbidity level, marital status, MELD score (<10 and ≥10), year of surgery (1997–2005, 2006–2011), and by preoperative concentration of CRP (≤10, >10 and ≤20, >20 and ≤50, and >50 mg/L). The regression analysis using categorisation of HSA assumes that the impact of HSA on mortality is equal within the same HSA category and that it is discontinuous as interval boundaries are crossed.36 In order to bridge such limitation, we used fractional polynomial Cox regression analysis to graphically describe adjusted HRs for 30-day mortality associated with preoperative HSA (as continuous variable) overall and stratified by CRP levels, assuming HRs equal to 1 for HSA equal to 40 g/L.37

boundaries are crossed.36 In order to bridge such limitation, we used fractional polynomial Cox regression analysis to graphically describe adjusted HRs for 30-day mortality associated with preoperative HSA (as continuous variable) overall and stratified by CRP levels, assuming HRs equal to 1 for HSA equal to 40 g/L.37 Finally, we conducted a sensitivity analysis where missing data for type of admission, CRC stage (ie,stage unknown), preoperative HSA, and other laboratory measurements were imputed deterministically with 20 cycles of regression switching assuming data were ‘missing at random’.38 Adjusted HRs and 95% CIs for 30-day mortality for each HSA category compared with HSA above 40 g/L were assessed using the imputed data sets. Analyses were performed using STATA V.12.0 (StataCorp LP, College Station, Texas, USA). The study was approved by the Danish Data Protection Agency (record number 2009-41-3866). Data obtained from Danish registries are generally available to researchers and their use does not require informed consent. Results Descriptive data We identified 10 347 patients (median age 71, IQR 62–78) undergoing first-time CRC surgery during the study period (table 1 and figure 1). Table 1 Characteristics of patients undergoing surgery for colorectal cancer

Analyses were performed using STATA V.12.0 (StataCorp LP, College Station, Texas, USA). The study was approved by the Danish Data Protection Agency (record number 2009-41-3866). Data obtained from Danish registries are generally available to researchers and their use does not require informed consent. Results Descriptive data We identified 10 347 patients (median age 71, IQR 62–78) undergoing first-time CRC surgery during the study period (table 1 and figure 1). Table 1 Characteristics of patients undergoing surgery for colorectal cancer Serum albumin concentration Hypoalbuminaemia Normal albuminaemia Severe Moderate Mild Low High Missing albumin ≤25 g/L n=401 26–30 g/L n=784 31–35 g/L n=1742 36–40 g/L n=3065 >40 g/L n=3347 n=1008 Gender (%) Male 143 (35.7) 371 (47.3) 844 (48.5) 1660 (54.2) 1910 (57.1) 514 (51.0) Female 258 (64.3) 413 (52.7) 898 (51.6) 1405 (45.8) 1437 (42.9) 494 (49.0) Median age (IQR) 76 (68–82) 76 (68–82) 75 (65–81) 71 (63–78) 67 (59–74) 71 (61–79) Age, years, (%) <60 48 (12.0) 64 (8.2) 234 (13.4) 549 (17.9) 857 (25.6) 210 (20.8) 60–69 60 (15.0) 170 (21.7) 380 (21.8) 824 (26.9) 1122 (33.5) 237 (23.5) 70–79 154 (38.4) 280 (35.7) 605 (34.7) 1064 (34.7) 990 (29.6) 317 (31.5) ≥80 139 (34.7) 270 (34.4) 523 (30.0) 628 (20.5) 378 (11.3) 244 (24.1) Type of admission (%) Elective 122 (30.4) 348 (44.4) 1128 (64.8) 2544 (83.1) 3118 (93.2) 534 (53.5) Non-elective 279 (69.6) 435 (55.6) 612 (35.2) 516 (16.9) 227 (6.8) 465 (46.2) Missing 0 1 (0.1) 2 (0.1) 5 (0.16) 2 (0.1) 9 (0.9) Cancer site (%) Colon 302 (75.3) 599 (76.4) 1212 (69.6) 1856 (60.6) 1689 (50.5) 779 (77.3) Rectum 99 (24.7) 185 (23.6) 530 (30.4) 1209 (39.5) 1658 (49.5) 229 (22.7) Cancer stage (%) Localised 124 (30.9) 251 (32.0) 671 (38.5) 1287 (42.0) 1517 (45.3) 385 (38.2) Regional 98 (24.4) 205 (26.2) 460 (26.4) 841 (27.4) 1051 (31.4) 299 (29.7) Metastasised 119 (29.7) 215 (27.4) 365 (21.0) 527 (17.2) 340 (10.2) 206 (20.4) Unknown 60 (15.0) 113 (14.4) 246 (14.1) 410 (13.4) 439 (13.1) 118 (11.7) Type of surgery (%) Open radical resection 266 (66.3) 579 (73.9) 1369 (78.6) 2470 (80.6) 2546 (76.1) 721 (71.5) Laparoscopic radical resection 4 (1.0) 7 (0.9) 65 (3.7) 271 (8.8) 541 (16.2) 121 (12.0) Non-eradicative procedures 131 (32.7) 198 (25.3) 308 (17.7) 324 (10.6) 260 (7.8) 166 (16.5) Comorbidity (%) Low 205 (51.1) 375 (47.8) 928 (53.3) 1752 (57.2) 2153 (64.3) 651 (64.6) Moderate 134 (33.4) 279 (35.6) 585 (33.6) 969 (31.6) 916 (27.4) 271 (26.9) High 62 (15.5) 130 (16.6) 229 (13.2) 344 (11.2) 278 (8.3) 86 (8.5) Alcohol-related disease (%) No 386 (96.3) 764 (97.5) 1707 (98.0) 3017 (98.4) 3290 (98.3) 988 (98.0) Yes 15 (3.7) 20 (2.6) 35 (2.0) 48 (1.6) 57 (1.7) 20 (2.0) Marital status (%) Married 172 (42.9) 359 (45.8) 855 (49.1) 1731 (5

31.6) 916 (27.4) 271 (26.9) High 62 (15.5) 130 (16.6) 229 (13.2) 344 (11.2) 278 (8.3) 86 (8.5) Alcohol-related disease (%) No 386 (96.3) 764 (97.5) 1707 (98.0) 3017 (98.4) 3290 (98.3) 988 (98.0) Yes 15 (3.7) 20 (2.6) 35 (2.0) 48 (1.6) 57 (1.7) 20 (2.0) Marital status (%) Married 172 (42.9) 359 (45.8) 855 (49.1) 1731 (5 6.5) 2139 (63.9) 516 (51.2) Never married 40 (10.0) 55 (7.0) 137 (7.9) 218 (7.1) 214 (6.4) 82 (8.1) Other 189 (47.1) 370 (47.2) 750 (43.1) 1116 (36.4) 994 (29.7) 410 (40.7) Figure 1 Flow chart showing the selection of patients with colorectal cancer (CRC).

31.6) 916 (27.4) 271 (26.9) High 62 (15.5) 130 (16.6) 229 (13.2) 344 (11.2) 278 (8.3) 86 (8.5) Alcohol-related disease (%) No 386 (96.3) 764 (97.5) 1707 (98.0) 3017 (98.4) 3290 (98.3) 988 (98.0) Yes 15 (3.7) 20 (2.6) 35 (2.0) 48 (1.6) 57 (1.7) 20 (2.0) Marital status (%) Married 172 (42.9) 359 (45.8) 855 (49.1) 1731 (5 6.5) 2139 (63.9) 516 (51.2) Never married 40 (10.0) 55 (7.0) 137 (7.9) 218 (7.1) 214 (6.4) 82 (8.1) Other 189 (47.1) 370 (47.2) 750 (43.1) 1116 (36.4) 994 (29.7) 410 (40.7) Figure 1 Flow chart showing the selection of patients with colorectal cancer (CRC). Of those, 9339 (90.3%) had at least one measurement of HSA in the 30 days before surgery and 8381 (81.0%) in the week before surgery. HSA below 35 g/L was present in 26.4% (n=2464) of patients with preoperative HSA measurement. Patients with HSA equal to or below 40 g/L were more likely to be old, female, and to have comorbid conditions including alcohol-related disease as compared with those with HSA above 40 g/L (table 1). Moreover, compared with patients with HSA above 40 g/L, patients with lower HSA were more likely to have metastasised cancer and primary localisation in the colon. The prevalence of non-elective admission increased with the decrease of HSA. The distribution of individual diseases included in the CCI for each HSA category is reported in online supplementary table S2. Information on other preoperative blood tests and MELD score for each HSA category is reported in table 2. Preoperative CRP was available for 66.1% (n=6841) of patients and it was markedly increased among patients with low HSA (table 2). Information about patients with missing preoperative HSA (n=1008) is reported in tables 1 and 2, and online supplementary table S3.

MELD score for each HSA category is reported in table 2. Preoperative CRP was available for 66.1% (n=6841) of patients and it was markedly increased among patients with low HSA (table 2). Information about patients with missing preoperative HSA (n=1008) is reported in tables 1 and 2, and online supplementary table S3. Table 2 Preoperative blood measurements in patients undergoing surgery for colorectal cancer

MELD score for each HSA category is reported in table 2. Preoperative CRP was available for 66.1% (n=6841) of patients and it was markedly increased among patients with low HSA (table 2). Information about patients with missing preoperative HSA (n=1008) is reported in tables 1 and 2, and online supplementary table S3. Table 2 Preoperative blood measurements in patients undergoing surgery for colorectal cancer Serum albumin concentration Hypoalbuminaemia Normal albuminaemia Severe Moderate Mild Low High Missing albumin ≤25 g/L n=401 26–30 g/L n=784 31–35 g/L n=1742 36–40 g/L n=3065 >40 g/L n=3347 n=1008 Albumin, g/L Mean (SD) 22.0 (3.0) 28.4 (1.4) 33.3 (1.4) 38.1 (1.4) 43.4 (2.1) – Median (IQR) 23 (21–24) 29 (27–30) 34 (32–35) 38 (37–39) 43 (42–45) – Haemoglobin, mmol/L Median (IQR) 6.6 (6.1–7.6) 6.9 (6.3–7.6) 7.3 (6.5–8) 7.9 (7.1–8.7) 8.5 (7.8–9.1) 8.0 (7.1–8.8) Missing, n (%) 0 2 (0.3) 3 (0.2) 10 (0.3) 8 (0.2) 553 (54.9) Sodium, mmol/L Median (IQR) 137 (134–139) 137 (135–140) 139 (137–141) 140 (138–142) 140 (139–142) 140 (137–141) Missing, n (%) 0 2 (0.3) 2 (0.1) 3 (0.1) 2 (0.1) 572 (56.8) Potassium, mmol/L Median (IQR) 3.8 (3.4–4.2) 3.9 (3.5–4.3) 4 (3.7–4.3) 4.1 (3.8–4.3) 4.1 (3.8–4.3) 4 (3.7–4.3) Missing, n (%) 0 3 (0.4) 2 (0.1) 3 (0.1) 1 (<0.1) 569 (56.5) Leucocytes, 109/L Median (IQR) 11.2 (8.5–15.1) 10.0 (7.8–13.2) 8.8 (6.9–11.2) 8.0 (6.5–9.9) 7.4 (6.1–9.0) 9.1 (6.9–11.8) Missing, n (%) 28 (7.0) 82 (10.5) 359 (20.6) 846 (27.6) 945 (28.2) 752 (74.6) Creatinine, µmol/L Median (IQR) 68 (55–85) 74 (61–93) 78 (66–93) 79 (68–95) 77 (67–90) 78 (67–91) Missing, n (%) 0 1 (0.1) 0 5 (0.2) 1 (<0.1) 556 (55.2) C reactive protein, mg/L Median (IQR) 85 (41–146) 51 (24–102) 26 (10–62) 10 (10–26) 10 (8–10) 19 (10–55) Missing, n (%) 40 (10.0) 121 (15.4) 460 (26.4) 1019 (33.3) 1054 (31.5) 812 (80.6) Platelet, 109/L Median (IQR) 411 (309–537) 392 (300–514) 353 (278–455) 316 (250–398) 290 (242–353) 324 (253–417) Missing, n (%) 80 (20.0) 197 (25.1) 537 (30.8) 1027 (33.5) 1065 (31.8) 818 (81.2) INR Median (IQR) 1.2 (1.1–1.3) 1.1 (1.0–1.2) 1.0 (1.0–1.1) 1.0 (1.0–1.1) 1.0 (0.9–1.1) 1.1 (1.0–1.2) Missing, n (%) 121 (30.2) 276 (35.2) 682 (39.2) 1162 (37.9) 1123 (33.6) 845 (83.8) Bilirubin, µmol/L Median (IQR) 9 (6–13) 8 (6–13) 8 (5–11) 8 (6–11) 8 (6–11) 10 (7–12) Missing, n (%) 72 (18.0) 214 (27.3) 625 (35.9) 1130 (36.9) 1036 (31.0) 821 (81.5) ALAT, U/L Median (IQR) 17 (11–28) 17 (12–28) 17 (12–25) 17 (13–25) 20 (15–27) 18 (13–28) Missing, n (%) 135 (33.7) 324 (41.33) 812 (46.6) 1410 (46.0) 1204 (36.0) 815 (80.9) MELD <10, n (%) 177 (44.1) 331 (42.2) 720 (41.3) 1308 (42.7) 1742 (52.1) 1

n (%) 72 (18.0) 214 (27.3) 625 (35.9) 1130 (36.9) 1036 (31.0) 821 (81.5) ALAT, U/L Median (IQR) 17 (11–28) 17 (12–28) 17 (12–25) 17 (13–25) 20 (15–27) 18 (13–28) Missing, n (%) 135 (33.7) 324 (41.33) 812 (46.6) 1410 (46.0) 1204 (36.0) 815 (80.9) MELD <10, n (%) 177 (44.1) 331 (42.2) 720 (41.3) 1308 (42.7) 1742 (52.1) 1 00 (9.9) ≥10, n (%) 76 (19.0) 97 (12.4) 120 (6.9) 200 (6.5) 127 (3.8) 20 (2.0) Missing, n (%) 148 (36.9) 356 (45.4) 902 (51.8) 1557 (50.8) 1478 (44.16) 888 (88.1) ALAT, alanine aminotransferase; INR, international normalised ratio; MELD, Model for End-Stage Liver Disease. Postoperative mortality Overall 30-day mortality increased from 2.0% in patients with HSA above 40 g/L to 26.9% in patients with HSA equal to or below 25 g/L (table 3 and figure 2), corresponding to an absolute mortality difference of 25.0% (95% CI 20.6% to 29.3%). Table 3 Thirty-day mortality and corresponding HRs in patients with different preoperative serum albumin concentration undergoing surgery because of colorectal cancer, overall and stratified by type of admission, cancer type, and period of surgery

Postoperative mortality Overall 30-day mortality increased from 2.0% in patients with HSA above 40 g/L to 26.9% in patients with HSA equal to or below 25 g/L (table 3 and figure 2), corresponding to an absolute mortality difference of 25.0% (95% CI 20.6% to 29.3%). Table 3 Thirty-day mortality and corresponding HRs in patients with different preoperative serum albumin concentration undergoing surgery because of colorectal cancer, overall and stratified by type of admission, cancer type, and period of surgery HR (95% CI) Number of patients N Number of deaths N 30-Day mortality %* (95% CI) Crude Adjusted† Colorectal cancer surgery, albumin g/L ≤25 401 108 26.9 (22.9 to 31.6) 15.89 (11.69 to 21.59) 7.59 (4.95 to 11.64) 26–30 784 154 19.6 (17.0 to 22.6) 10.91 (8.18 to 14.56) 5.19 (3.53 to 7.63) 31–35 1742 163 9.4 (8.1 to 10.8) 4.92 (3.70 to 6.56) 2.58 (1.80 to 3.69) 36–40 3065 149 4.9 (4.2 to 5.7) 2.50 (1.87 to 3.34) 1.75 (1.25 to 2.45) >40 3347 66 2.0 (1.6 to 2.5) 1.00 1.00 Elective admission‡, albumin g/L ≤25 122 30 24.6 (17.9 to 33.2) 15.32 (9.84 to 23.83) 8.08 (4.45 to 14.67) 26–30 348 62 17.8 (14.2 to 22.3) 10.57 (7.38 to 15.15) 5.31 (3.27 to 8.62) 31–35 1128 90 8.0 (6.5 to 9.7) 4.50 (3.23 to 6.26) 2.56 (1.68 to 3.90) 36–40 2544 99 3.9 (3.2 to 4.7) 2.15 (1.55 to 2.98) 1.52 (1.04 to 2.21) >40 3118 57 1.8 (1.4 to 2.4) 1.00 1.00 Non-elective admission‡, albumin g/L ≤25 279 78 28.0 (23.1 to 33.6) 8.19 (4.11 to 16.33) 7.78 (3.17 to 19.12) 26–30 435 92 21.2 (17.6 to 25.3) 5.83 (2.94 to 11.57) 5.36 (2.23 to 12.88) 31–35 612 73 11.9 (9.6 to 14.8) 3.14 (1.57 to 6.28) 2.68 (1.12 to 6.44) 36–40 516 50 9.7 (7.4 to 12.6) 2.51 (1.23 to 5.10) 2.32 (0.97 to 5.55) >40 227 9 4.0 (2.1 to 7.5) 1.00 1.00 Age 0–59 years, albumin g/L ≤25 48 6 12.5 (5.8 to 25.7) 23.03 (7.03 to 75.46) 7.39 (1.32 to 41.22) 26–30 64 9 14.1 (7.6 to 25.3) 25.55 (8.56 to 76.24) 11.84 (2.65 to 52.97) 31–35 234 9 3.9 (2.0 to 7.3) 6.68 (2.24 to 19.94) 2.18 (0.51 to 9.35) 36–40 549 10 1.8 (1.0 to 3.4) 3.15 (1.08 to 9.20) 2.29 (0.69 to 7.61) >40 857 5 0.6 (0.2 to 1.4) 1.00 1.00 Age 60–69 years, albumin g/L ≤25 60 9 15.0 (8.1 to 26.8) 15.11 (6.37 to 35.87) 6.63 (2.13 to 20.61) 26–30 170 26 15.3 (10.7 to 21.6) 15.39 (7.77 to 30.50) 4.75 (1.77 to 12.76) 31–35 380 21 5.5 (3.6 to 8.4) 5.26 (2.59 to 10.69) 2.55 (1.03 to 6.31) 36–40 824 17 2.1 (1.3 to 3.3) 1.94 (0.92 to 4.05) 1.34 (0.57 to 3.15) >40 1122 12 1.1 (0.6 to 1.9) 1.00 1.00 Age 70–79 years, albumin g/L ≤25 154 43 27.9 (21.5 to 35.7) 11.89 (7.35 to 19.24) 8.76 (4.40 to 17.44) 26–30 280 54 19.3 (15.1 to 24.4) 7.71 (4.86 to 12.24) 5.23 (2.79 to 9.77) 31–35 605 46 7.6 (5.8 to 10.0) 2.87 (1.78 to 4.61) 2.12 (1.19 to 3.80) 36–40 1064 49 4.6 (3.5 to 6.1) 1.70 (1.06 to 2

.1 (0.6 to 1.9) 1.00 1.00 Age 70–79 years, albumin g/L ≤25 154 43 27.9 (21.5 to 35.7) 11.89 (7.35 to 19.24) 8.76 (4.40 to 17.44) 26–30 280 54 19.3 (15.1 to 24.4) 7.71 (4.86 to 12.24) 5.23 (2.79 to 9.77) 31–35 605 46 7.6 (5.8 to 10.0) 2.87 (1.78 to 4.61) 2.12 (1.19 to 3.80) 36–40 1064 49 4.6 (3.5 to 6.1) 1.70 (1.06 to 2 .72) 1.41 (0.83 to 2.40) >40 990 27 2.7 (1.9 to 4.0) 1.00 1.00 Age 80+ years, albumin g/L ≤25 139 50 36.0 (28.6 to 44.5) 7.54 (4.56 to 12.45) 7.09 (3.50 to 14.36) 26–30 270 65 24.1 (19.4 to 29.6) 4.57 (2.82 to 7.40) 4.57 (2.40 to 8.68) 31–35 523 87 16.6 (13.7 to 20.1) 3.03 (1.90 to 4.83) 2.66 (1.46 to 4.85) 36–40 628 73 11.6 (9.4 to 14.4) 2.05 (1.28 to 3.31) 2.17 (1.22 to 3.86) >40 378 22 5.8 (3.9 to 8.7) 1.00 1.00 Age and admission type were largely responsible for the change in estimates by adjustment. *Calculated using the Kaplan-Meier method. †Mutually adjusted for gender, age (both as a continuous and a categorical variable), type of admission, operation year (calendar year), county, cancer site, cancer stage (excluded patients with stage ‘unknown’), comorbidity level, alcohol-related disease, liver disease, marital status, haemoglobin, sodium, potassium and creatinine (number of observations with complete data=8033). ‡Information on type of admission is missing for some patients. Therefore, the sum of patients with non-elective and elective admissions is not equal to the number of all patients included in the study. Figure 2 Crude 30-day mortality curves for patients undergoing surgery for colorectal cancer according to preoperative serum albumin concentration.

‡Information on type of admission is missing for some patients. Therefore, the sum of patients with non-elective and elective admissions is not equal to the number of all patients included in the study. Figure 2 Crude 30-day mortality curves for patients undergoing surgery for colorectal cancer according to preoperative serum albumin concentration. The adjusted HRs increased from 1.75 (95% CI 1.25 to 2.45) among patients with HSA 35–40 g/L to 7.59 (95% CI 4.95 to 11.64) among patients with HSA equal to or below 25 g/L, compared with patients with HSA above 40 g/L. As expected, in each HSA category 30-day mortality was higher among patients admitted non-electively than among patients with an elective admission. Similarly, elderly patients had higher postoperative mortality than younger patients (table 3). However, adjusted HRs showed that a decrease in HSA was associated with a gradually increased risk of mortality regardless of types of admission or age group (table 3). Table 4 shows 30-day mortality and corresponding adjusted HRs for subgroups of patients undergoing CRC surgery. Table 4 Thirty-day mortality and corresponding adjusted HRs with 95% CIs in subgroups of patients with different preoperative serum albumin concentrations undergoing surgery because of colorectal cancer Serum albumin concentration ≤25 g/L 26–30 g/L 31–35 g/L 36–40 g/L >40 g/L 30-Day mortality* % (n) Adjusted HR† (95% CI) 30-DAy mortality* % (n) Adjusted HR† (95% CI) 30-Day mortality* % (n) Adjusted HR† (95% CI) 30-Day mortality* % (n) Adjusted HR† (95% CI) 30-Day mortality* % (n) Ref.

Table 4 Thirty-day mortality and corresponding adjusted HRs with 95% CIs in subgroups of patients with different preoperative serum albumin concentrations undergoing surgery because of colorectal cancer Serum albumin concentration ≤25 g/L 26–30 g/L 31–35 g/L 36–40 g/L >40 g/L 30-Day mortality* % (n) Adjusted HR† (95% CI) 30-DAy mortality* % (n) Adjusted HR† (95% CI) 30-Day mortality* % (n) Adjusted HR† (95% CI) 30-Day mortality* % (n) Adjusted HR† (95% CI) 30-Day mortality* % (n) Ref. group Gender Male 30.1 (43) 8.23 (4.49 to 15.09) 24.5 (91) 5.79 (3.45 to 9.71) 10.3 (87) 2.68 (1.65 to 4.35) 5.2 (86) 1.73 (1.11 to 2.71) 1.9 (37) 1.00 Female 25.2 (65) 6.79 (3.65 to 12.63) 15.3 (63) 4.28 (2.39 to 7.65) 8.5 (76) 2.29 (1.39 to 3.92) 4.5 (63) 1.64 (0.99 to 2.72) 2.0 (29) 1.00 Cancer site Colon 29.8 (90) 6.68 (4.01 to 11.14) 18.4 (110) 3.96 (2.47 to 6.34) 9.2 (112) 2.05 (1.32 to 3.20) 5.2 (96) 1.43 (0.94 to 2.19) 2.4 (41) 1.00 Rectum 18.2 (18) 8.90 (3.77 to 21.05) 23.8 (44) 10.76 (5.40 to 21.44) 9.6 (51) 3.83 (2.06 to 7.12) 4.4 (53) 2.52 (1.44 to 4.41) 1.5 (25) 1.00 Cancer stage Localised 17.7 (22) 4.22 (2.14 to 8.36) 16.7 (42) 4.02 (2.25 to 7.19) 7.2 (48) 2.08 (1.24 to 3.46) 4.8 (62) 1.74 (1.10 to 2.73) 2.1 (32) 1.00 Regional 25.5 (25) 11.01 (4.44 to 27.33) 15.1 (31) 5.88 (2.65 to 13.07) 7.8 (36) 3.90 (1.88 to 8.11) 3.7 (31) 2.11 (1.05 to 4.25) 1.2 (13) 1.00 Metastasised 38.7 (46) 10.61 (4.72 to 23.88) 27.9 (60) 6.58 (3.05 to 14.18) 11.8 (43) 2.64 (1.25 to 5.58) 5.7 (30) 1.53 (0.74 to 3.19) 2.9 (10) 1.00 Year of surgery 1997–2005 29.8 (78) 4.40 (2.52 to 7.67) 20.4 (95) 3.03 (1.82 to 5.05) 8.2 (82) 1.39 (0.85 to 2.28) 4.5 (68) 1.05 (0.66 to 1.68) 3.1 (28) 1.00 2006–2011 21.6 (30) 10.11 (5.17 to 19.79) 18.5 (59) 8.49 (4.84 to 14.92) 11.0 (81) 4.90 (2.94 to 8.15) 5.2 (81) 2.81 (1.76 to 4.49) 1.6 (38) 1.00 Type of surgery Open radical resection 24.8 (66) 5.32 (3.26 to 8.68) 16.2 (94) 3.43 (2.21 to 5.30) 8.6 (118) 2.03 (1.37 to 3.01) 4.9 (120) 1.52 (1.06 to 2.18) 2.2 (57) 1.00 Laparoscopic radical resection 0 –- 14.3 (1) – 7.7 (5) 9.13 (1.32 to 63.34) 3.3 (9) 3.19 (0.74 to 13.78) 0.7 (4) 1,00 Non-eradicative resection 32.1 (42) 44.40 (9.48 to 207.97) 29.8 (59) 30.81 (6.91 to 41.59) 13.0 (40) 9.35 (2.10 to 41.59) 6.2 (20) 4.43 (0.98 to 20.05) 1.9 (5) 1,00 Comorbidity Low 23.7 (49) 7.87 (4.05 to 15.26) 15.9 (60) 5.39 (2.96 to 9.82) 6.6 (61) 1.91 (1.08 to 3.37) 2.8 (49) 1.22 (0.72 to 2.06) 1.3 (28) 1.00 Moderate 29.9 (40) 9.36 (4.65 to 18.83) 23.0 (64) 6.68 (3.59 to 12.45) 10.9 (64) 3.31 (1.86 to 5.86) 6.6 (64) 2.10 (1.24 to 3.58) 2.9 (27) 1.00 High 31.7 (19) 6.91 (2.45 to 19.53) 23.3 (30) 4.77 (1.91 to 11.94) 17.0 (38) 3.55 (1.51 to 8.34) 10.7 (36) 2.84

1.91 (1.08 to 3.37) 2.8 (49) 1.22 (0.72 to 2.06) 1.3 (28) 1.00 Moderate 29.9 (40) 9.36 (4.65 to 18.83) 23.0 (64) 6.68 (3.59 to 12.45) 10.9 (64) 3.31 (1.86 to 5.86) 6.6 (64) 2.10 (1.24 to 3.58) 2.9 (27) 1.00 High 31.7 (19) 6.91 (2.45 to 19.53) 23.3 (30) 4.77 (1.91 to 11.94) 17.0 (38) 3.55 (1.51 to 8.34) 10.7 (36) 2.84 (1.25 to 6.43) 4.1 (11) 1.00 Marital status Married 27.3 (47) 11.09 (5.95 to 20.66) 20.1 (72) 6.70 (3.83 to 11.74) 8.3 (71) 3.52 (2.11 to 5.89) 3.7 (64) 1.72 (1.06 to 2.79) 1.5 (31) 1.00 Never married 25.0 (10) 10.50 (1.72 to 64.07) 27.3 (15) 5.58 (1.05 to 29.76) 8.0 (11) 1.26 (0.27 to 5.87) 6.0 (13) 2.29 (0.59 to 8.79) 2.3 (5) 1.00 Other 27.0 (51) 5.75 (3.03 to 10.92) 18.1 (67) 4.15 (2.32 to 7.41) 10.8 (81) 2.23 (1.30 to 3.81) 6.5 (72) 1.76 (1.06 to 2.91) 3.0 (30) 1.00 Creactive protein‡, mg/L ≤10.0 14.3 (2) 5.33 (1.06 to 26.83) 9.5 (7) 1.81 (0.57 to 5.76) 6.2 (20) 1.88 (0.88 to 4.00) 4.1 (41) 1.48 (0.82 to 2.67) 1.6 (29) 1.00 10.1–20.0 13.8 (4) 4.31 (0.87 to 21.32) 9.7 (7) 1.91 (0.53 to 6.85) 12.2 (27) 3.09 (1.21 to 7.86) 6.7 (27) 1.93 (0.83 to 4.50) 3.6 (11) 1.00 20.1–50.0 23.6 (17) 7.47 (2.22 to 25.08) 21.0 (38) 7.26 (2.40 to 22.00) 9.6 (33) 3.38 (1.18 to 9.70) 5.6 (22) 1.88 (0.67 to 5.31) 2.7 (5) 1.00 >50.0 29.3 (72) 4.19 (0.97 to 18.12) 26.5 (89) 3.63 (0.86 to 15.35) 13.4 (53) 1.36 (0.32 to 5.85) 10.6 (26) 1.44 (0.33 to 6.25) 8.1 (3) 1.00 MELD§ <10 23.2 (41) 7.06 (3.61 to 13.80) 18.1 (60) 5.20 (2.87 to 9.42) 9.2 (66) 2.74 (1.60 to 4.71) 4.5 (59) 1.63 (0.99 to 2.69) 1.9 (33) 1.00 ≥10 43.4 (33) 9.22 (3.27 to 25.97) 43.4 (42) 8.62 (3.17 to 23.42) 19.2 (23) 2.75 (1.03 to 7.32) 15.5 (31) 2.36 (0.94 to 5.92) 6.3 (8) 1.00 *Calculated using the Kaplan-Meier method.

§ <10 23.2 (41) 7.06 (3.61 to 13.80) 18.1 (60) 5.20 (2.87 to 9.42) 9.2 (66) 2.74 (1.60 to 4.71) 4.5 (59) 1.63 (0.99 to 2.69) 1.9 (33) 1.00 ≥10 43.4 (33) 9.22 (3.27 to 25.97) 43.4 (42) 8.62 (3.17 to 23.42) 19.2 (23) 2.75 (1.03 to 7.32) 15.5 (31) 2.36 (0.94 to 5.92) 6.3 (8) 1.00 *Calculated using the Kaplan-Meier method. †Mutually adjusted for gender, age (both as a continuous and a categorical variable), type of admission, operation year (calendar year), county, cancer site, cancer stage (excluded patients with stage ‘unknown’), comorbidity level, alcohol-related disease, liver disease, marital status, haemoglobin, sodium, potassium and creatinine (number of observations with complete data=8033). ‡Patients with preoperative C reactive protein measurement=6841. §Patients with preoperative MELD=5018. MELD, Model for End-stage Liver Disease; Ref., reference. For all subgroups, postoperative mortality was lowest among patients with HSA above 40 g/L and gradually increased with decreasing HSA levels. Notably, 30-day mortality was affected even by changes within what is clinically considered the normal HSA range (35–40 g/L as compared with >40 g/L) (tables 3 and 4). Adjusted HRs similarly showed a markedly high mortality associated with a decrease in HSA in most of the subgroups, especially among patients with HSA below 35 g/L. However, 95% CIs reflected imprecise estimates. The strong relation between 30-day mortality and preoperative HSA was confirmed from the fractional polynomial analysis (figure 3).

For all subgroups, postoperative mortality was lowest among patients with HSA above 40 g/L and gradually increased with decreasing HSA levels. Notably, 30-day mortality was affected even by changes within what is clinically considered the normal HSA range (35–40 g/L as compared with >40 g/L) (tables 3 and 4). Adjusted HRs similarly showed a markedly high mortality associated with a decrease in HSA in most of the subgroups, especially among patients with HSA below 35 g/L. However, 95% CIs reflected imprecise estimates. The strong relation between 30-day mortality and preoperative HSA was confirmed from the fractional polynomial analysis (figure 3). Figure 3 Adjusted HRs for 30-day mortality, overall (A) and stratified by C reactive protein (CRP) levels (B), associated with preoperative human serum albumin concentration. Adjusted HRs in (A) are provided with 95% CI (dash lines). Sensitivity analysis The analysis using data with imputation of missing information on preoperative HSA and other variables revealed results similar to those reported above: the adjusted 30-day HRs were 7.50 (95% CI 5.10 to 11.03) for HSA below 25 g/L, 4.99 (95% CI 3.51 to 7.10) for HSA 26–30 g/L, 2.76 (95% CI 2.00 to 3.79) for HSA 31–35 g/L, and 1.78 (95% CI 1.32 to 2.40) for HSA 36–40 g/L, compared with HSA above 40 g/L.

d other variables revealed results similar to those reported above: the adjusted 30-day HRs were 7.50 (95% CI 5.10 to 11.03) for HSA below 25 g/L, 4.99 (95% CI 3.51 to 7.10) for HSA 26–30 g/L, 2.76 (95% CI 2.00 to 3.79) for HSA 31–35 g/L, and 1.78 (95% CI 1.32 to 2.40) for HSA 36–40 g/L, compared with HSA above 40 g/L. Discussion In this cohort study conducted within a population-based hospital setting, we found that 30-day mortality after CRC surgery was inversely associated with preoperative HSA concentration. Even within HSA levels that are clinically considered normal, we found that a decrease in HSA is associated with increased mortality. Moreover, the prognostic impact of decrement of HSA was independent of other preoperative conditions and of different CRP levels.

nversely associated with preoperative HSA concentration. Even within HSA levels that are clinically considered normal, we found that a decrease in HSA is associated with increased mortality. Moreover, the prognostic impact of decrement of HSA was independent of other preoperative conditions and of different CRP levels. Our study extends current knowledge by examining the impact of preoperative HSA concentration on 30-day mortality following CRC surgery overall and in subgroups of patients within a population-based setting. Previous studies investigating short-term prognosis in patients undergoing CRC surgery reported increased risk of postoperative complications among patients with HSA below 35 g/L.8–12 14 15 39 40 Lai et al8 found a twofold increased 30-day mortality comparing patients with HSA below 35 g/L to those with HSA above 35 g/L. Our findings showed that the impact of HSA on 30-day mortality had a concentration–response pattern that was not limited to the cut-off of 35 g/L. Moreover, our results showed that the prognostic impact of HSA persisted even in subgroups of patients with different preoperative conditions that are reported to be associated with a decrement of HSA.

impact of HSA on 30-day mortality had a concentration–response pattern that was not limited to the cut-off of 35 g/L. Moreover, our results showed that the prognostic impact of HSA persisted even in subgroups of patients with different preoperative conditions that are reported to be associated with a decrement of HSA. Decrement in HSA may have a direct impact on prognosis affecting organ vascularisation, hamper distribution of antibiotics, perpetuate inflammation, and promote intravascular coagulation. Indeed, during the past decade a better understanding of HSA structure and function has led to the concept that HSA has multifunctional properties ranging from provision of oncotic pressure, immune regulation, and endothelial stabilisation to being a molecule that works in the intracellular compartment modifying several key pathophysiological mechanisms.41 A criticism that is often raised regarding causality between decreased HSA concentration and prognosis is that clinical trials provided contradictory results about prognosis following HSA administration, especially among critically ill patients.42–46 However, HSA replacement has been shown beneficial in specific clinical conditions such as spontaneous bacterial peritonitis, patients with ascites receiving paracentesis, hepatic encephalopathy, and hepatorenal syndrome.41 47 Moreover, previous studies showing no benefit from HSA administration were performed in acute patients when their conditions were already worsened and this may have prevented the effect of HSA on prognosis. Finally, the quality of administrated HSA should also be questioned based on previous studies investigating HSA in commercial solutions that reported high prevalence of oxidised forms 48 and in vitro immunosuppressive activity.49

nditions were already worsened and this may have prevented the effect of HSA on prognosis. Finally, the quality of administrated HSA should also be questioned based on previous studies investigating HSA in commercial solutions that reported high prevalence of oxidised forms 48 and in vitro immunosuppressive activity.49 The main strengths of our study include its large size, its population-based design with uniform access to healthcare in Denmark, comprehensive laboratory data, accurate history of preadmission comorbidity,29 and complete follow-up data. However, additional issues should be considered when interpreting our results. Exclusion of patients with CRC with missing data from the main analysis may have introduced selection bias.50 However, the sensitivity analysis using imputed data provided estimates that were similar to those obtained from the complete case analysis. Therefore, it seems unlikely that missing data could have biased our findings. Unmeasured or only partially measured conditions (eg, malnutrition, infection, alcohol consumption, and smoking) known to affect HSA concentration may increase the risk of postoperative mortality also through a non-HSA-related pathway and therefore might have biased our estimates. However, the strength of association and the concentration–response pattern of HRs are unlikely explained only by residual or unmeasured confounding. Finally, we do not have information on cause of death and this prevents us from investigating possible differences in the specific complication leading to death among patients with different preoperative HSA concentrations.

ation–response pattern of HRs are unlikely explained only by residual or unmeasured confounding. Finally, we do not have information on cause of death and this prevents us from investigating possible differences in the specific complication leading to death among patients with different preoperative HSA concentrations. In conclusion, our results showed that a decrement of preoperative HSA was associated with a concentration-dependent increased risk of mortality in the 30 days following CRC surgery even within concentration levels clinically considered normal. Furthermore, we showed that other preoperative conditions and patient characteristics including presence of systemic inflammation did not markedly affect the prognostic impact of HSA on postoperative mortality. Supplementary Material Supplementary Materials Contributors: JM contributed in the study concept and design, statistical analysis, data interpretation, and manuscript preparation. SA contributed in the statistical analysis supervision and manuscript reviewing. RE, TN and HTS contributed in the study concept and design, critical analysis of the data, manuscript reviewing, and study supervision. All the authors approved the final draft submitted for publication. Funding: The study was supported by a grant from the Danish Cancer Society (R73-A4284-13-S17) and from Aarhus University Research Foundation. JM has received a scholarship from Aarhus University. Competing interests: None declared. Ethics approval: The study was approved by the Danish Data Protection Agency (record number 2009-41-3866).

Funding: The study was supported by a grant from the Danish Cancer Society (R73-A4284-13-S17) and from Aarhus University Research Foundation. JM has received a scholarship from Aarhus University. Competing interests: None declared. Ethics approval: The study was approved by the Danish Data Protection Agency (record number 2009-41-3866). Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Summary box What is already known about this subject? ▸  Irritable bowel syndrome (IBS) is a chronic functional gastrointestinal (GI) disorder that affects 10–15% of western populations, yet its pathogenesis and management are poorly characterised. ▸  Vitamin D supplementation has been associated with positive outcomes in other lower-GI conditions, including cancer prevention and colitis. ▸  Few studies to date have reported any relationship between vitamin D and IBS, although we recently reported a case study in this area. What are the new findings? ▸  The IBS population exhibits low concentrations of serum 25OHD; these levels respond to supplementation. ▸  There is a significant positive association between quality of life and circulating 25OHD concentrations. ▸  We provide data on which to inform power calculations for a larger clinical trial of vitamin D supplementation in IBS patients. How might it impact on clinical practice in the foreseeable future? ▸  The data provide evidence for widespread vitamin D insufficiency in people with IBS. There may be benefit to testing IBS patients’ vitamin D status and providing supplementation. There is a positive association between vitamin D status and quality of life measures, supporting a potential interaction between vitamin D and mental health and well-being. Clinicians may consider trialling vitamin D with patients with IBS as part of a nutritional management strategy. The results suggest a larger, adequately powered trial may be justified.

vitamin D status and quality of life measures, supporting a potential interaction between vitamin D and mental health and well-being. Clinicians may consider trialling vitamin D with patients with IBS as part of a nutritional management strategy. The results suggest a larger, adequately powered trial may be justified. Introduction Irritable bowel syndrome (IBS) is a chronic disorder which profoundly affects quality of life with a prevalence of 10–15% in the industrialised world.1 IBS is a relapsing condition that has a large social impact and is associated with significant direct and indirect healthcare costs.1 It is a heterogeneous disorder that can be subtyped depending on the bowel habits of patients; IBS with constipation (IBS-C), IBS with diarrhoea (IBS-D) or IBS with mixed bowel habits (IBS-M).2 IBS pathogenesis is poorly understood and it is generally regarded as a multifactorial disorder involving host and environmental factors, including diet. Most hypotheses involve altered intraluminal milieu, immune activation, enteric neuromuscular dysfunction and brain–gut axis dysregulation. It is thought that the intestinal microbiota might play an important role as bacterial infection, antibiotic use and chronic low-grade inflammation are associated with IBS onset.3 Several studies have implicated changes in the colon microbiota may be associated with IBS symptoms, including perturbation in bile acid metabolism and electrolyte absorption.4 This theory is further supported by generally beneficial effects of bacteriotherapeutic interventions (vide infra).

associated with IBS onset.3 Several studies have implicated changes in the colon microbiota may be associated with IBS symptoms, including perturbation in bile acid metabolism and electrolyte absorption.4 This theory is further supported by generally beneficial effects of bacteriotherapeutic interventions (vide infra). Therapy for IBS is primarily targeted at treating the symptoms experienced, and include loperamide for diarrhoea, methylcellulose for constipation and smooth-muscle relaxants for abdominal pain.2 Nonetheless, symptom treatment meets with limited success and may not be effective for long-term management of IBS. Mounting data suggest that probiotics may be beneficial in the management of IBS. A recent systematic review of 35 trials investigating probiotic use in IBS showed that 25 of these studies reported a beneficial effect on the primary outcomes including global symptom severity.5 6 The effect on secondary outcomes varied between studies with improvements seen in flatulence and abdominal bloating following treatment with VSL#3.7 8 Other studies demonstrated a decrease in abdominal pain after 4 weeks of treatment with Lactobacillus plantarum 299V when compared to control.9 Our previous work has demonstrated an improvement in global symptom severity in people with IBS following 8 weeks of supplementation with LAB4 probiotic.10

tment with VSL#3.7 8 Other studies demonstrated a decrease in abdominal pain after 4 weeks of treatment with Lactobacillus plantarum 299V when compared to control.9 Our previous work has demonstrated an improvement in global symptom severity in people with IBS following 8 weeks of supplementation with LAB4 probiotic.10 We recently reported a case study of an IBS patient taking high dose (3000 IU daily) vitamin D. The participant reported remission of IBS symptoms following supplementation, with a recurrence of the symptoms on supplementation cessation. Additionally, analysis of social media (blogs/forums) reports from 37 IBS patients found that 70% described improvements in their symptoms with supplementation and the majority of these individuals reported being vitamin D deficient before supplementation.11 A role for vitamin D supplementation in gastrointestinal health is also supported by studies showing associations between vitamin D deficiency and inflammatory bowel disease (IBD).12 A recent systematic review suggested there may be benefits of vitamin D supplementation in IBD.13 On the basis of our previous case report,11 we hypothesised that patients with IBS who are vitamin D insufficient, would report improvement in symptoms following vitamin D supplementation. We further sought to test whether vitamin D supplementation in combination with a probiotic preparation would act synergistically.

We recently reported a case study of an IBS patient taking high dose (3000 IU daily) vitamin D. The participant reported remission of IBS symptoms following supplementation, with a recurrence of the symptoms on supplementation cessation. Additionally, analysis of social media (blogs/forums) reports from 37 IBS patients found that 70% described improvements in their symptoms with supplementation and the majority of these individuals reported being vitamin D deficient before supplementation.11 A role for vitamin D supplementation in gastrointestinal health is also supported by studies showing associations between vitamin D deficiency and inflammatory bowel disease (IBD).12 A recent systematic review suggested there may be benefits of vitamin D supplementation in IBD.13 On the basis of our previous case report,11 we hypothesised that patients with IBS who are vitamin D insufficient, would report improvement in symptoms following vitamin D supplementation. We further sought to test whether vitamin D supplementation in combination with a probiotic preparation would act synergistically. There are no published randomised controlled trial of vitamin D supplementation in IBS and therefore no data on which to base a power calculation. A formal aim of this trial is to provide data to allow such calculations to be undertaken, enabling properly scaled clinical trials to be designed.

On the basis of our previous case report,11 we hypothesised that patients with IBS who are vitamin D insufficient, would report improvement in symptoms following vitamin D supplementation. We further sought to test whether vitamin D supplementation in combination with a probiotic preparation would act synergistically. There are no published randomised controlled trial of vitamin D supplementation in IBS and therefore no data on which to base a power calculation. A formal aim of this trial is to provide data to allow such calculations to be undertaken, enabling properly scaled clinical trials to be designed. Materials and methods Participants Ethical approval for this study was granted by The University of Sheffield Research Ethics Committee (Ref: SMBRER278). Participants were recruited via poster advertisements at the University of Sheffield. All participants had a previous clinical diagnosis of IBS and met the Rome III criteria at baseline.14 Participants who reported any antibiotic use in the past 4 weeks prior to recruitment, recent changes in IBS medication, pregnancy, current use of vitamins or probiotic supplements, history of gastrointestinal surgery, diabetes or current use of antidepressants or antipsychotics were excluded. The study population was stratified by vitamin D status hypothesising that deficient or insufficient individuals would respond if deficiency were causal in IBS.

rent use of vitamins or probiotic supplements, history of gastrointestinal surgery, diabetes or current use of antidepressants or antipsychotics were excluded. The study population was stratified by vitamin D status hypothesising that deficient or insufficient individuals would respond if deficiency were causal in IBS. Patient measures Throughout the study participants were assessed using a number of outcomes. To assess vitamin D status participants provided a blood sample from which baseline and exit 25OHD was measured in serum. Participants were given a food frequency questionnaire (FFQ) from which dietary intake was derived using FETA opensource software.i Baseline IBS symptom questionnaires were completed.15 The questionnaire assessed abdominal pain (pain severity and number of days with pain), bloating, bowel habits (minimum and maximum bowel movement per day and satisfaction with bowel habit) and quality of life over each 2-week period. Study design This was a 12-week double-blind, placebo-controlled, stratified study. Participants attended the Clinical Research Facility, Royal Hallamshire Hospital, Sheffield for three visits (figure 1). At the first visit, baseline 25OHD was measured in serum, participants completed FFQ and baseline IBS symptom questionnaires were completed.15 Figure 1 Flow diagram of study protocol. IBS, irritable bowel syndrome; SSQ, Symptom Severity Questionnaires.

Participants attended the Clinical Research Facility, Royal Hallamshire Hospital, Sheffield for three visits (figure 1). At the first visit, baseline 25OHD was measured in serum, participants completed FFQ and baseline IBS symptom questionnaires were completed.15 Figure 1 Flow diagram of study protocol. IBS, irritable bowel syndrome; SSQ, Symptom Severity Questionnaires. The second visit took place approximately 2 weeks following visit 1. Participants were issued with a 12-week supply of supplement and were given IBS symptom questionnaires to be completed biweekly from the date of visit 2. At the final visit participants provided a second blood sample from which serum 25OHD was measured. Of the 51 patients issued with supplements 9 failed to return questionnaires for the full period (4 in the placebo group, 3 vitamin D alone and 2 combined intervention).

The second visit took place approximately 2 weeks following visit 1. Participants were issued with a 12-week supply of supplement and were given IBS symptom questionnaires to be completed biweekly from the date of visit 2. At the final visit participants provided a second blood sample from which serum 25OHD was measured. Of the 51 patients issued with supplements 9 failed to return questionnaires for the full period (4 in the placebo group, 3 vitamin D alone and 2 combined intervention). Sample size and randomisation Based on our previous study10 we aimed to recruit approximately 150 participants. By the end of the study 51 participants were recruited. Recruitment was between January 2014 and April 2014. Participants were stratified by baseline vitamin D statusii (‘Vitamin D deficient’ (25OHD <20 ng/mL) and ‘Vitamin D replete’ (25OHD >20 ng/mL)) and randomised. Participants were randomised to receive either double placebo (n=17), vitamin D supplementation and probiotic placebo (n=16) or probiotic and vitamin D supplementation (n=18). Participants were allocated in a 1:1:1 ratio to the three arms of the study according to a computer-generated random sequence using block randomisation with a block-size of six. The randomisation was performed by an independent statistician. Participants were enrolled and assigned sequentially to placebo or active products. The allocation sequence was not available to any member of the research team until databases had been completed and locked.

ng block randomisation with a block-size of six. The randomisation was performed by an independent statistician. Participants were enrolled and assigned sequentially to placebo or active products. The allocation sequence was not available to any member of the research team until databases had been completed and locked. Intervention The probiotic, vitamin D3 and corresponding placebos were provided by Cultech Ltd, Port Talbot, UK. Vitamin D3 and the corresponding placebo were provided as 15 mL liquid sublingual sprays. Both contained identical buffers with placebo lacking the active vitamin D3. Each spray of vitamin D3 gave a dose containing 3000 IU vitamin D3. The probiotic preparation contained two strains of Lactobacillus acidophilus, CUL60 (NCIMB 30157), CUL21 (NCIMB 30156), Bifidobacterium bifidum CUL20 (NCIMB 30153) and Bifidobacterium animalis subsp. lactis CUL34 (NCIMB 30172) at a total of 2.5×1010 colony forming units (cfu) per capsule. Volunteers were instructed to ingest one capsule per day with water and one spray orally per day for 12 weeks. Compliance was assessed by counting the number of capsules remaining and weighing sprays at the end of the intervention—98% compliance was achieved. Adverse events There were no adverse events. One participant on the placebo arm reported feeling unwell after starting the trial with acute indigestion, heartburn and pain in the back which subsided after 2 weeks. The participant halted supplements for 4 weeks and then returned to the study under the initial time frame.

Adverse events There were no adverse events. One participant on the placebo arm reported feeling unwell after starting the trial with acute indigestion, heartburn and pain in the back which subsided after 2 weeks. The participant halted supplements for 4 weeks and then returned to the study under the initial time frame. Biochemical assay Serum 25OHD was measured using the Cobas e411 automated immunoassay from Roche Diagnostics (Germany). The inter assay coefficient of variation (CV) was 4.6%.

Adverse events There were no adverse events. One participant on the placebo arm reported feeling unwell after starting the trial with acute indigestion, heartburn and pain in the back which subsided after 2 weeks. The participant halted supplements for 4 weeks and then returned to the study under the initial time frame. Biochemical assay Serum 25OHD was measured using the Cobas e411 automated immunoassay from Roche Diagnostics (Germany). The inter assay coefficient of variation (CV) was 4.6%. Statistical analysis Number of days with pain were expressed as a percentage of each 14-day period. All other variables were assessed using a visual analogue scale and scored of 100. Individual scores were then combined to give a total symptom severity score of 500.15 χ2 Test was used to assess the distribution of participants by vitamin D stratification. Associations between baseline 25OHD stratification (deficient: <12.5 ng/mL; insufficient: 12.5–20 ng/mL; replete 20–50 ng/mL; toxic: >50 ng/mL, no recordings were made at or near the toxic range) and baseline symptom scores, follow-up 25OHD and change in symptom score were determined by analysis of variance(ANOVA) with Bonferroni correction. Pearson's correlations for vitamin D intake and baseline serum 25OHD were undertaken. Change in symptom score over 7 time points were analysed by repeated measures ANOVA, missing data were not imputed and were assumed to be missing at random in the ANOVA model. Comparisons of deficient and replete symptom scores were assessed by independent t tests. All tests were two sided with a significance value of <0.05. Analysis was carried out using SPSS V.22 (IBM, Armonk, New York, USA).

ANOVA, missing data were not imputed and were assumed to be missing at random in the ANOVA model. Comparisons of deficient and replete symptom scores were assessed by independent t tests. All tests were two sided with a significance value of <0.05. Analysis was carried out using SPSS V.22 (IBM, Armonk, New York, USA). Results Baseline vitamin D status shows a significant association with quality of life Demographics of the population are shown in table 1. The three arms were similar in terms of number, gender, age, IBS subtype and serum 25OHD. Overall IBS symptom severity at baseline (week -2) was similar between groups (243±67 placebo, 244±92 vitamin D and 237±67 vitamin D with probiotic, p=0.241) and between IBS subtype (figure 2A) as was the baseline serum 25OHD (15±8.4, 16±8.0, 14±8.3 ng/mL, p=0.295). Table 1 Baseline characteristics of participants Placebo Vitamin D Vitamin D+Probiotic Participants n 18 17 16 Female n 17 15 15 Mean age (±SD) 36 (±15) 34 (±12) 34 (±14) IBS subtype n Mixed 10 7 11 Constipation 2 5 2 Diarrhoea 6 5 3 Vitamin D status Deficient 14 14 12 Replete 4 3 4 Mean serum 25OHD ng/mL (±SD) 15 (±8.4) 14 (±8.3) 16 (±8.0) Mean IBS severity score (±SD) 243 (67) 244 (92) 237 (67) 25OHD, 25-hydroxyvitamin D; IBS, Irritable bowel syndrome.

(±SD) 36 (±15) 34 (±12) 34 (±14) IBS subtype n Mixed 10 7 11 Constipation 2 5 2 Diarrhoea 6 5 3 Vitamin D status Deficient 14 14 12 Replete 4 3 4 Mean serum 25OHD ng/mL (±SD) 15 (±8.4) 14 (±8.3) 16 (±8.0) Mean IBS severity score (±SD) 243 (67) 244 (92) 237 (67) 25OHD, 25-hydroxyvitamin D; IBS, Irritable bowel syndrome. Figure 2 Baseline characteristics of participants (A). Distribution of IBS-subtype across serum 25OHD levels at baseline shows no significant association of IBS subtype with vitamin D status (B). Mean score for symptom severity at baseline in IBS-C, D and M participants stratified by vitamin D status at baseline (C). Correlation of baseline 25OHD with vitamin D intake (D). Table summarising IBS symptom score with vitamin D intake and baseline serum 25OHD. No correlations were detected with in response to vitamin D intake, however, serum 25OHD showed a negative correlation between quality of life and serum 25OHD. IBS, irritable bowel syndrome; FFQ, food frequency questionnaire.

ntake (D). Table summarising IBS symptom score with vitamin D intake and baseline serum 25OHD. No correlations were detected with in response to vitamin D intake, however, serum 25OHD showed a negative correlation between quality of life and serum 25OHD. IBS, irritable bowel syndrome; FFQ, food frequency questionnaire. The majority of participants had baseline 25OHD levels that are considered insufficient/severely deficient16 with an overall sample mean 25OHD of 15.3±7.9 ng/mL and 81.8% of IBS-C (n=7), 70% of IBS-D (n=9) and 81.6% of IBS-M (n=24) with <20 ng/mL circulating 25OHD levels. This level of deficiency was similar to that seen in a BMI-matched cohort in Sheffield at this time of year (S. Bowles personal communication, 2014). There was no association between IBS-subtype and 25OHD status at baseline (figure 2B). However, participants with low 25OHD at baseline reported a greater impact on quality of life (on the basis of the IBS questionnaire) than their replete counterparts (p=0.034) (table 2). There was no difference in other symptom scores between 25OHD deficient and replete individuals (table 2). Data exploration suggested, a trend for lower affected quality of life scores in each IBS subtype in vitamin D replete participants against the non-replete cohort (see online supplementary, figure S1c). Table 2 Baseline symptoms by vitamin D status* Deficient Replete Symptom severity 247±70 233±77 Pain severity 44±20 47±24 Pain frequency 37±25 36±24 Distention severity 42±29 40±30 Bowel satisfaction 60±20 58±23 Affected life 62±16 49±21* Data are means±SD.

The majority of participants had baseline 25OHD levels that are considered insufficient/severely deficient16 with an overall sample mean 25OHD of 15.3±7.9 ng/mL and 81.8% of IBS-C (n=7), 70% of IBS-D (n=9) and 81.6% of IBS-M (n=24) with <20 ng/mL circulating 25OHD levels. This level of deficiency was similar to that seen in a BMI-matched cohort in Sheffield at this time of year (S. Bowles personal communication, 2014). There was no association between IBS-subtype and 25OHD status at baseline (figure 2B). However, participants with low 25OHD at baseline reported a greater impact on quality of life (on the basis of the IBS questionnaire) than their replete counterparts (p=0.034) (table 2). There was no difference in other symptom scores between 25OHD deficient and replete individuals (table 2). Data exploration suggested, a trend for lower affected quality of life scores in each IBS subtype in vitamin D replete participants against the non-replete cohort (see online supplementary, figure S1c). Table 2 Baseline symptoms by vitamin D status* Deficient Replete Symptom severity 247±70 233±77 Pain severity 44±20 47±24 Pain frequency 37±25 36±24 Distention severity 42±29 40±30 Bowel satisfaction 60±20 58±23 Affected life 62±16 49±21* Data are means±SD. *Denotes p<0.05 w.r.t. corresponding deficient symptom, independent t test.

Table 2 Baseline symptoms by vitamin D status* Deficient Replete Symptom severity 247±70 233±77 Pain severity 44±20 47±24 Pain frequency 37±25 36±24 Distention severity 42±29 40±30 Bowel satisfaction 60±20 58±23 Affected life 62±16 49±21* Data are means±SD. *Denotes p<0.05 w.r.t. corresponding deficient symptom, independent t test. We sought to determine whether potential benefits of replete vitamin D status may be attributed to associations between systemic effects of 25OHD or the potential mucosal action of ingested dietary vitamin D. There was no association between vitamin D intake and symptom severity, suggesting the effect may be systemically mediated (figure 2C, D). Effect of vitamin D intervention The distribution of participants by vitamin D status in each treatment arm was compared at baseline and exit. An equal distribution was seen at baseline (table 3) and the distribution was altered by 12 weeks of vitamin D supplementation. Of participants receiving supplementation in combination or alone the percentage of participants who were replete improved from 22.2% and 25.0% to 87.5% and 92.3%, respectively. Participants on placebo also had higher 25OHD levels with 60% of participants replete at exit compared to only 18.5% at baseline (figure 3A). Table 3 Vitamin D status at baseline and exit

ith TAC (TAC group) and the remaining 7 patients were treated with IFX (IFX group). The choice of TAC or IFX treatment for the induction therapy was conducted according to the primary physician's decision at that time. There were no significant differences in the clinical characteristics between the TAC and IFX groups. Short-term outcomes The clinical course of short-term outcomes of the 29 patients with severe UC is shown in figure 1. Of 22 patients in the TAC group, 14 achieved a CR at 8 weeks after initiating TAC and 6 of the remaining 8 patients could not achieve a CR, although they responded to TAC. The remaining two patients did not respond to TAC and required sequential therapy (switched to IFX) within 8 weeks after initiating TAC. Of seven patients in the IFX group, five achieved a CR at 8 weeks after initiating IFX, while one required sequential therapy (switched to TAC). One patient who did not respond to IFX resulted in a colectomy without sequential therapy. The CR rate at 8 weeks from initiating TAC or IFX including sequential therapy in the 29 patients with severe UC was 65.5% (19/29). The colectomy rate at 8 weeks was 3.4% (1/29).

Effect of vitamin D intervention The distribution of participants by vitamin D status in each treatment arm was compared at baseline and exit. An equal distribution was seen at baseline (table 3) and the distribution was altered by 12 weeks of vitamin D supplementation. Of participants receiving supplementation in combination or alone the percentage of participants who were replete improved from 22.2% and 25.0% to 87.5% and 92.3%, respectively. Participants on placebo also had higher 25OHD levels with 60% of participants replete at exit compared to only 18.5% at baseline (figure 3A). Table 3 Vitamin D status at baseline and exit Intervention arm Baseline (% sufficient) Exit (% sufficient) Placebo 18.5 60.0 Vitamin D3 22.2 92.3 Vitamin D3+Probiotic 25.0 87.5 Figure 3 Effect of intervention on participant serum 25OHD concentration and total symptom severity (A). Box plot serum 25OHD concentrations (ng/mL) before and after intervention with placebo, vitamin D3 or vitamin D3+Probiotic (all participants), a significantly greater increase with vitamin D supplementation alone or with probiotic was seen over placebo group (B). Repeated measure of total symptom severity against time point (all participants) (C). Box plot serum 25OHD concentrations (ng/mL) before and after intervention with placebo, vitamin D3 or vitamin D3+Probiotic. (deplete participants only), a significant increase in serum 25OHD was seen in all groups with a greater response associated with vitamin D supplementation alone or with probiotic when compared to placebo group (D). Repeated measure of total symptom severity against time point (deplete participants only).

D3+Probiotic. (deplete participants only), a significant increase in serum 25OHD was seen in all groups with a greater response associated with vitamin D supplementation alone or with probiotic when compared to placebo group (D). Repeated measure of total symptom severity against time point (deplete participants only). Participants who received vitamin D alone or in combination with probiotic had significantly higher 25OHD at follow-up compared to baseline (15.8±8.0 to 37.2±9.3 ng/mL and 14.9±8.4 to 37.1±11.7 ng/mL, respectively, p<0.001). These two groups also had higher levels of 25OHD compared to the placebo group at exit with mean 25OHD levels of 37 ng/mL while the placebo group mean was 25.3±8.0 ng/mL (p=0.008 vitamin D, p=0.005 vitamin D+Probiotic, respectively) (figure 3A). While it appeared that there were improved scores in all reported symptoms from baseline to exit irrespective of the intervention arm, this was not statistically significant for any symptom tested. Additionally, treatment arm did not interact with any of the symptom scores at any of the time point (figure 3B and online supplementary figure S2). The analysis was repeated limited to participants who were deficient at baseline. A significant increase was seen between baseline and exit 25OHD in all arms (figure 3C) with no significant relationship of intervention arm on total symptom severity (figure 3D) or individual IBS scores (data not shown).

tary figure S2). The analysis was repeated limited to participants who were deficient at baseline. A significant increase was seen between baseline and exit 25OHD in all arms (figure 3C) with no significant relationship of intervention arm on total symptom severity (figure 3D) or individual IBS scores (data not shown). Discussion We have presented evidence of high-vitamin D insufficiency in a population with IBS, We have previously reported anecdotal evidence suggesting a benefit of vitamin D in management of IBS.11 Baseline data from our population would support a need for monitoring vitamin D status and suggests that the impact of vitamin D deficiency is most acute on perceived quality of life.

iency in a population with IBS, We have previously reported anecdotal evidence suggesting a benefit of vitamin D in management of IBS.11 Baseline data from our population would support a need for monitoring vitamin D status and suggests that the impact of vitamin D deficiency is most acute on perceived quality of life. A correlation indicated a potential relationship between 25OHD and quality of life scores; however, this did not reach significance. Reported quality of life will be influenced by the depression status of participants, a common comorbidity in IBS.17 Numerous studies have reported that vitamin D levels are correlated with depression. In one such study vitamin D deficiency in young healthy individuals predicted clinical depression across a 4-week period, noting that as vitamin D levels increased due to seasonal difference in sunlight exposure volunteers showed lower levels of depression.18 A similar pattern was seen in patients suffering post-stroke depression: lower serum vitamin D correlated with depression.19 The exact mechanism through which vitamin D improves depression is not fully understood. However, vitamin D status has been implicated in neurological development,14 the vitamin D receptor (VDR) is expressed throughout the nervous system where its activation is linked to neurotransmitter levels and serotonin synthesis.20 Additionally binding of 25OHD-VDR complex results in expression of 1-α-hydroxylase which converts 25OHD to 1,25-dihydroxyvitamin D.21 This 25OHD metabolite has been shown to upregulate neurotrophins which in turn promote survival and differentiation of nerve cells.22 VDR is also expressed in the gut and regulates epithelial barrier function and bowel inflammation23 suggesting that a vitamin D deficient diet may directly impact bowel function and hence IBS symptomology. Our attempt to distinguish the effects of circulating and ingested vitamin D did not support either model, likely due to underpowering.

in the gut and regulates epithelial barrier function and bowel inflammation23 suggesting that a vitamin D deficient diet may directly impact bowel function and hence IBS symptomology. Our attempt to distinguish the effects of circulating and ingested vitamin D did not support either model, likely due to underpowering. Supplementation with vitamin D increased serum 25OHD concentrations from baseline with a significantly greater improvement seen compared to participants receiving placebo. Nonetheless the group receiving placebo showed an improvement in 25OHD levels. This may have prevented us from detecting a significant difference in symptom scores between the placebo and supplemented groups at exit. This general increase in serum 25OHD across intervention arms was likely due to seasonal differences in sun exposure. It has been found in previous studies that the percentage of individuals in the UK deficient for vitamin D ranges from 30–40% (January–March) to 2–13% (July–September).16 Study participants were recruited in the early months of 2014 (January–April) with most follow-up visits taking place from May to August, most likely explaining the rise in 25OHD in all groups.

he percentage of individuals in the UK deficient for vitamin D ranges from 30–40% (January–March) to 2–13% (July–September).16 Study participants were recruited in the early months of 2014 (January–April) with most follow-up visits taking place from May to August, most likely explaining the rise in 25OHD in all groups. An exploratory analysis of participants by IBS subtype showed that there was a marked improvement in response to vitamin D supplementation across nearly all IBS symptoms in participants with IBS-C (see online supplementary figure S3). Interestingly the proportion of these participants that were considered severely deficient was low in comparison to the other IBS subtypes (figure 2B). A clear limitation is the very small numbers in groups when the data is split but the observation suggests a future study into vitamin D supplementation in this subgroup may be justified.

roportion of these participants that were considered severely deficient was low in comparison to the other IBS subtypes (figure 2B). A clear limitation is the very small numbers in groups when the data is split but the observation suggests a future study into vitamin D supplementation in this subgroup may be justified. Throughout this study we failed to observe a significant interaction with intervention with either vitamin D alone or in combination with probiotic. This is likely explained by a number of limitations in the study. First, we observed an effect of the placebo on symptom scores: IBS symptoms have a highly variable course (compare week-2 and 0 in figures 3B, D for example). It is recommended that active intervention in IBS last a minimum of 4 weeks and potentially up to 6 months for long-term analysis.24 Furthermore, as a pilot study the current trial is inherently underpowered. In order to detect a statistically significant response in total symptom severity scores we would need larger sample sizes. With the data generated in this study it was calculated that a sample size of 74 per arm would be needed. This was calculated based on change in total symptom severity at exit with a placebo mean of −16, a mean of −54 for vitamin D intervention, a combined SD of 82 and assuming a desired power of 80% with an α error of 0.05.

he data generated in this study it was calculated that a sample size of 74 per arm would be needed. This was calculated based on change in total symptom severity at exit with a placebo mean of −16, a mean of −54 for vitamin D intervention, a combined SD of 82 and assuming a desired power of 80% with an α error of 0.05. It is possible that seasonal variations in vitamin D levels may impact IBS in the current study. Our data show a general improvement in symptom score and improved 25OHD over the duration of this study is tempting to speculate that increased sunlight exposure increases vitamin D levels which in turn improved IBS symptoms. Such effects have been previously shown for IBS and IBD symptoms.25 26 However, there was a significantly greater effect of vitamin D intervention than placebo on circulating vitamin D. Taken together, these data suggestion that seasonal variation should also be taken into account in the design of trials to assess the potential application of vitamin D in IBS. In summary we have shown that IBS sufferers deficient in vitamin D experience poorer quality of lives than their replete counterparts. Our data demonstrate the need for additional studies of more than 8 weeks in duration with recruitment of a larger cohort of IBS patients. The authors thank the staff at Sheffield Clinical Research Facility for support phlebotomy and use of their research environment; Ms Fatma Gossiel undertook clinical biochemical assays for vitamin D; Ms Dolappo Owoeye and Ms Farhana ab Hadi entered diet data.

In summary we have shown that IBS sufferers deficient in vitamin D experience poorer quality of lives than their replete counterparts. Our data demonstrate the need for additional studies of more than 8 weeks in duration with recruitment of a larger cohort of IBS patients. The authors thank the staff at Sheffield Clinical Research Facility for support phlebotomy and use of their research environment; Ms Fatma Gossiel undertook clinical biochemical assays for vitamin D; Ms Dolappo Owoeye and Ms Farhana ab Hadi entered diet data. Contributors: ST completed exit analyses, entered all data, undertook all statistical analyses, wrote the first draft of the paper. NR and ART recruited and consented participants and undertook day-to-day running of the trial in its inception and initial phases and reviewed drafts of the paper. ALE contributed to exit analyses, undertook first-pass data analysis and reviewed drafts of the paper. VAG co-conceived the study and undertook qualitative exit interviews with participants and reviewed drafts of the paper. IG; SFP co-designed the study, advised on dosing and delivery of probiotics and vitamin D, contributed to the analysis and interpretation of the data and reviewed drafts of the paper. EAW co-designed the study and contributed to its direction, supervised and interpreted the nutritional analyses and reviewed drafts of the paper. BMC conceived the study, assumes overall responsibility for the project, is its chief investigator, finalised each draft of the manuscript and produced the final version. Funding: This work was funded by Cultech Ltd.

one required sequential therapy (switched to TAC). One patient who did not respond to IFX resulted in a colectomy without sequential therapy. The CR rate at 8 weeks from initiating TAC or IFX including sequential therapy in the 29 patients with severe UC was 65.5% (19/29). The colectomy rate at 8 weeks was 3.4% (1/29). Figure 1 Short-term and long-term outcomes of the 29 patients with severe ulcerative colitis (UC). Short-term outcomes: 14 of 22 patients in the tacrolimus (TAC) group and 5 of 7 patients in the infliximab (IFX) group achieved a clinical remission (CR) at 8 weeks after initiating TAC or IFX. Two patients in the TAC group were refractory to TAC and required sequential therapy. Two patients in the IFX group were refractory to IFX and one could achieve a CR after switching to TAC. The remaining one resulted in colectomy without switching to TAC. The CR rate at 8 weeks in the TAC and IFX groups was 63.6% and 71.4%, respectively. Long-term outcomes: In the TAC group, of 14 patients who achieved a CR at 8 weeks after initiating TAC, 11 patients could maintain a CR and 3 flared-up. A total of nine patients could not achieve and maintain a CR with TAC. Six of the nine patients could achieve a CR by sequential therapy, and the remaining three resulted in colectomy (one patient within 6 months). Two patients who required sequential therapy within 8 weeks after initiating TAC resulted in colectomy within 6 months after initiating TAC. In the IFX group, of four patients who achieved a CR at 8 weeks after initiating IFX, three could maintain a CR and one flared-up. A total of two patients could not achieve and maintain a CR with IFX and they could achieve a CR by sequential therapy. One patient who achieved a CR by sequential therapy within 8 weeks after initiating IFX could maintain a CR. *Reveals the patient who resulted in colectomy within 6 months after initiating TAC or IFX.

Contributors: ST completed exit analyses, entered all data, undertook all statistical analyses, wrote the first draft of the paper. NR and ART recruited and consented participants and undertook day-to-day running of the trial in its inception and initial phases and reviewed drafts of the paper. ALE contributed to exit analyses, undertook first-pass data analysis and reviewed drafts of the paper. VAG co-conceived the study and undertook qualitative exit interviews with participants and reviewed drafts of the paper. IG; SFP co-designed the study, advised on dosing and delivery of probiotics and vitamin D, contributed to the analysis and interpretation of the data and reviewed drafts of the paper. EAW co-designed the study and contributed to its direction, supervised and interpreted the nutritional analyses and reviewed drafts of the paper. BMC conceived the study, assumes overall responsibility for the project, is its chief investigator, finalised each draft of the manuscript and produced the final version. Funding: This work was funded by Cultech Ltd. Competing interests: IG and SFP are employees of Cultech Ltd, a manufacturer of supplements and probiotics. Patient consent: Obtained. Ethics approval: University of Sheffield. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available. i http://www.srl.cam.ac.uk/epic/epicffq/ ii https://ods.od.nih.gov/factsheets/VitaminD-HealthProfessional/

Summary box What is already known about this subject? ▸  According to randomised trials, Colorectal Cancer (CRC) screening with the guaiac-based faecal occult blood test (FOBT) reduces CRC-related mortality. ▸  Based on the available evidence, the EU recommends CRC screening. ▸  Evidence on the impact of guaiac-based FOBT screening on resource use is missing. What are the new findings? ▸  Our study showed a slight increase in the use of hospital services (outpatient visits, inpatient episodes and colonoscopies) in the screening group compared to the control group. ▸  People who were invited to screening but who did not attend were more likely to use hospital resources for gastrointestinal reasons than people with a negative FOBT result. How might it impact on clinical practice in the foreseeable future? ▸  This study only evaluates the resource use on a hospital level and not the impact of screening on primary healthcare services. Introduction Colorectal cancer is the third most common cancer, following breast and prostate cancers, and the second most common cause of cancer deaths after lung cancer, in Europe, including in Finland. Within the 27 Member states of the European Union, 340 000 people were diagnosed with colorectal cancer in 2012 and about 150 000 people died because of it.1 Among 5.4 million Finns, the corresponding figures are approximately 2500 people diagnosed with colorectal cancer and more than 1000 deaths every year. The incidence of colorectal cancer increased for decades while mortality was stable among men and decreased among women since the 1970s.2 3

people died because of it.1 Among 5.4 million Finns, the corresponding figures are approximately 2500 people diagnosed with colorectal cancer and more than 1000 deaths every year. The incidence of colorectal cancer increased for decades while mortality was stable among men and decreased among women since the 1970s.2 3 Colorectal cancer screening aims to reduce the mortality of the disease by discovering the cancer at an early, asymptomatic stage, to enable more effective and less invasive treatment. Screening with a faecal occult blood test (FOBT) has been shown to reduce mortality from colorectal cancer in a number of randomised screening studies.4–7 In a Cochrane review,8 the combined results of these studies showed a 16% reduction in the relative risk of colorectal cancer mortality. Three of these studies4 6 7 used biennial screening and showed an average of 15% relative risk reduction in colorectal cancer mortality. The reduction in mortality was seen in all three trials even though there was large variation in the design, such as the age of the target population, screening intervals, attendance rate and length of follow-up. Also, the cost-effectiveness of colorectal cancer screening has been modelled in a number of studies, in various countries, using different screening tests and strategies.9–15 Based on the available evidence, the Council of the European Union already, in 2003, recommended the implementation of a population-based screening programme for colorectal cancer to the EU Member States.16

s been modelled in a number of studies, in various countries, using different screening tests and strategies.9–15 Based on the available evidence, the Council of the European Union already, in 2003, recommended the implementation of a population-based screening programme for colorectal cancer to the EU Member States.16 In Finland, colorectal cancer screening started in 2004 as a population-based randomised study. The idea was to launch a new screening programme in such a way that it can be reliably evaluated. The target population of screening is men and women aged 60–69 years, based on the fact that the incidence of and mortality from colorectal cancer increase after 60 years of age.3 The municipalities responsible for providing health services in Finland can voluntarily join the programme, and the number of piloting municipalities has increased from 22 (of 444) in 2004 to 153 (of 336) in 2012. The target population in these municipalities was randomised at an individual level within municipality, birth year and in sex specific blocks, to screening and control groups. Owing to the gradual expansion of the programme as a randomised health services study,17 the effects of the ultimate national screening programme on the allocation of health care resources can be evaluated as a part of routine health service. The aim of this article was to estimate the difference in the use of hospital resources between those invited and the controls in the Finnish colorectal cancer screening programme within the first screening round.

In Finland, colorectal cancer screening started in 2004 as a population-based randomised study. The idea was to launch a new screening programme in such a way that it can be reliably evaluated. The target population of screening is men and women aged 60–69 years, based on the fact that the incidence of and mortality from colorectal cancer increase after 60 years of age.3 The municipalities responsible for providing health services in Finland can voluntarily join the programme, and the number of piloting municipalities has increased from 22 (of 444) in 2004 to 153 (of 336) in 2012. The target population in these municipalities was randomised at an individual level within municipality, birth year and in sex specific blocks, to screening and control groups. Owing to the gradual expansion of the programme as a randomised health services study,17 the effects of the ultimate national screening programme on the allocation of health care resources can be evaluated as a part of routine health service. The aim of this article was to estimate the difference in the use of hospital resources between those invited and the controls in the Finnish colorectal cancer screening programme within the first screening round. Materials and methods When the Finnish colorectal cancer screening study was launched, a national screening centre was established in the city of Tampere, at the local cancer society, to deal with the invitations, responses and recommendations for referrals, and to analyse the faecal samples. FOBT kits (Hemoccult; Beckman Coulter, USA) are posted to those offered screening along with a letter of invitation to participate in the programme and advice on how to take the sample. The kits are also returned by post. If any traces of blood are found in the sample, the person is referred to a contact nurse in their home municipality, for further examinations. Diagnostic confirmation, treatment and follow-up of people who screened positive are conducted within the normal health services at the local level and are paid for mostly by the home municipality of the participant being screened. Those in the control arm are identified but not contacted. The control population receives routine health services available in Finland.18

f people who screened positive are conducted within the normal health services at the local level and are paid for mostly by the home municipality of the participant being screened. Those in the control arm are identified but not contacted. The control population receives routine health services available in Finland.18 All individuals (n=246 079) who were randomised to the colorectal cancer screening (n=123 149) or control (n=122 930) groups during years 2004–2009 were identified and their data on healthcare resource use were gathered from the Finnish Hospital Discharge Register (FHDR). The follow-up of each individual was from the beginning of the year of randomisation up to the end of the next calendar year (two full years). The study period was 2004–2010. For those randomised during the years 2004–2009, healthcare utilisation data on all gastrointestinal-related hospital-based outpatient visits and inpatient episodes of care were gathered on the year of randomisation and the following year. The data were restricted by ICD-10 codes that were related to gastrointestinal symptoms: C15-C26; D12 and K00-K93. Colonoscopies performed during those visits and episodes in hospitals were also examined, including procedure codes UJF32, UJF34 and UJF45. Colonoscopies performed in primary healthcare (including the screening colonoscopies) are not included in the analysis for either the screening or the control group. See table 1 for the explanation of the codes used in this article. Table 1 Explanation of the diagnose and procedure codes used in the article

For those randomised during the years 2004–2009, healthcare utilisation data on all gastrointestinal-related hospital-based outpatient visits and inpatient episodes of care were gathered on the year of randomisation and the following year. The data were restricted by ICD-10 codes that were related to gastrointestinal symptoms: C15-C26; D12 and K00-K93. Colonoscopies performed during those visits and episodes in hospitals were also examined, including procedure codes UJF32, UJF34 and UJF45. Colonoscopies performed in primary healthcare (including the screening colonoscopies) are not included in the analysis for either the screening or the control group. See table 1 for the explanation of the codes used in this article. Table 1 Explanation of the diagnose and procedure codes used in the article ICD-10 codes C15-C26 Malignant neoplasms of digestive organs C15 Malignant neoplasm of oesophagus C16 Malignant neoplasm of stomach C17 Malignant neoplasm of small intestine C18 Malignant neoplasm of colon C19 Malignant neoplasm of rectosigmoid junction C20 Malignant neoplasm of rectum C21 Malignant neoplasm of anus and anal canal C22 Malignant neoplasm of liver and intrahepatic bile ducts C23 Malignant neoplasm of gall bladder C24 Malignant neoplasm of other and unspecified parts of biliary tract C25 Malignant neoplasm of pancreas C26 Malignant neoplasm of other ill-defined digestive organs D12 Benign neoplasm of colon, rectum, anus and anal canal K00-K93 Diseases of the digestive system K00-K14 Diseases of oral cavity, salivary glands and jaws K20-K31 Diseases of oesophagus, stomach and duodenum K35-K38 Diseases of appendix K40-K46 Hernia K50-K52 Non-infective enteritis and colitis K55-K63 Other diseases of intestines K65-K67 Diseases of peritoneum K70-K77 Diseases of liver K80-K87 Disorders of gallbladder, biliary tract and pancreas K90-K93 Other diseases of the digestive system National procedure codes* UJF32 Colonoscopy UJF34 Left colonoscopy UJF45 Flexible sigmoidoscopy with biopsy *Based on the Nordic Classification of Surgical Procedures.

ases of peritoneum K70-K77 Diseases of liver K80-K87 Disorders of gallbladder, biliary tract and pancreas K90-K93 Other diseases of the digestive system National procedure codes* UJF32 Colonoscopy UJF34 Left colonoscopy UJF45 Flexible sigmoidoscopy with biopsy *Based on the Nordic Classification of Surgical Procedures. The FHDR is one of the oldest individual level hospital discharge registers covering an entire country. The FHDR has total (legislative) coverage of all inpatient care provided at university, general and psychiatric hospitals or primary care health centres, as well as treatment in military and prison wards and private hospitals, since 1969. Each record contains data on several variables, such as personal identity number, age at admission, gender, hospital identifier code, admission and discharge dates, reason for admission (ICD-10 code since 1996), and diagnostic and treatment activities. The register includes data on clinical procedures from 1986 onwards and the current coding system for procedures has been in use since 2004. The data on hospital outpatient visits have been collected since 1998 with some limitations and they are considered to be comparable from 2006 onwards. The completeness and accuracy of the FHDR has been reviewed to be good. Data on outpatient services at the primary care level (general practitioner visits, etc) are available from 1994 onwards at aggregate level.19 20

A total of two patients could not achieve and maintain a CR with IFX and they could achieve a CR by sequential therapy. One patient who achieved a CR by sequential therapy within 8 weeks after initiating IFX could maintain a CR. *Reveals the patient who resulted in colectomy within 6 months after initiating TAC or IFX. Long-term outcomes Of the 29 patients with severe UC, 25 were followed up for more than 6 months after initiating treatment with TAC or IFX. The median follow-up period of the 25 patients (TAC group: n=19, IFX group: n=6) was 29 months (range 7–118 months). The clinical course of long-term outcomes of the 25 patients with severe UC is shown in figure 1. Of 14 patients who achieved a CR at 8 weeks after initiating TAC, 11 could maintain a CR continuing TAC. The remaining three had a flare-up of UC, of whom one resulted in colectomy without sequential therapy within 6 months after initiating TAC. The remaining two patients who had a flare-up of UC despite achieving a CR and six patients who could not achieve a CR at 8 weeks after initiating TAC required sequential therapy (switched to IFX). After induction of sequential therapy, six of the eight patients could achieve and maintain a CR, and the remaining two required colectomy. Moreover, two patients who required the sequential therapy (switched to IFX) within 8 weeks after initiating TAC could not achieve a CR and required colectomy. However, in the IFX group, three of the four patients who achieved a CR at 8 weeks after initiating IFX could maintain a CR. The remaining one patient who had a flare-up of UC and the one patient who could not achieve a CR at 8 weeks after initiating IFX required sequential therapy (switched to TAC), and they could achieve and maintain a CR. The patient who required sequential therapy (switched to TAC) within 8 weeks after initiating IFX could achieve and maintain a CR. Overall, total 23 patients (TAC group: n=17, IFX group: n=6) could maintain a CR at the end of follow-up. The medications in these patients are summarised in table 2. Regarding their maintenance therapy at the end of follow-up, five patients in the TAC group and one patient in the IFX group were treated with IFX in conjunction with IM and 5-ASA. One patient in the TAC group and two patients in the IFX group were treated with IFX in conjunction with 5-ASA. Three patients in the TAC group were treated with TAC in conjunction with IM and 5-ASA.

collected since 1998 with some limitations and they are considered to be comparable from 2006 onwards. The completeness and accuracy of the FHDR has been reviewed to be good. Data on outpatient services at the primary care level (general practitioner visits, etc) are available from 1994 onwards at aggregate level.19 20 The percentage of people in the target population who used hospital services and the cumulative number of outpatient visits, hospital inpatient episodes and hospital colonoscopies, were calculated. We compared those randomised to screening to those in the control group. The screening group was also divided according to uptake of screening and test result. The differences between the arms in distributions of the individual variables were tested for statistical significance by the χ2 test. Results The study population included 246 079 people, 123 149 (50%) in the screening group and 122 930 in the control group. Based on the randomised setting of the study, the groups were similar in demographic characteristics. By 2009, the programme covered about 40% of the ultimate target population and 20% were invited to screening. The attendance rate within the first screening round was 68% during the years 2004–2009, with some variation by calendar year. During these years, 2944 individuals (2.4%) had a positive FOBT (table 2). Table 2 Number and percentage of individuals in the Finnish Colorectal Cancer Screening programme by year of invitation, arm and screening test result

By 2009, the programme covered about 40% of the ultimate target population and 20% were invited to screening. The attendance rate within the first screening round was 68% during the years 2004–2009, with some variation by calendar year. During these years, 2944 individuals (2.4%) had a positive FOBT (table 2). Table 2 Number and percentage of individuals in the Finnish Colorectal Cancer Screening programme by year of invitation, arm and screening test result Year of invitation Screening Screening total Control Positive test result Negative test result Non-attender n Per cent n Per cent n Per cent n n 2004 109 2.4 3283 72.3 1149 25.3 4539 4539 2005 542 2.3 16 709 70.9 6311 26.8 23 562 23 565 2006 585 2.3 16 625 66.8 7687 30.9 24 897 24 898 2007 513 2.4 14 303 66.2 6778 31.4 21 594 21 375 2008 576 2.4 14 756 60.9 8899 36.7 24 231 24 229 2009 621 2.6 15 299 62.9 8406 34.6 24 326 24 324 Total 2944 2.4 80 975 65.8 39 230 31.9 123 149 122 930 The number and proportion of people who had at least one gastrointestinal-related hospital outpatient visit, inpatient episode or colonoscopy during the year of randomisation or the next year, is presented in table 3. Most of the study population in both screening and control groups had not had any gastrointestinal-related hospital outpatient visits or inpatient episodes within the 2 years following randomisation. The proportion of people with at least one outpatient visit, inpatient episode or colonoscopy, was higher in the screening group than in the control group, for all the types of resource use evaluated. The difference between the groups was the largest in outpatient visits: there were 523 people more in the screening group than in the control group with at least one gastrointestinal-related outpatient visit. The corresponding figures for people with at least one inpatient episode or hospital-based colonoscopy were 132 and 315, respectively. The people in the screening group were more likely to undergo more colonoscopies per person than those in the control group (figure 1).

st one gastrointestinal-related outpatient visit. The corresponding figures for people with at least one inpatient episode or hospital-based colonoscopy were 132 and 315, respectively. The people in the screening group were more likely to undergo more colonoscopies per person than those in the control group (figure 1). Table 3 Number and proportion of individuals with at least one event (outpatient visit, inpatient episode and colonoscopy) during the first 2 years following randomisation by arm, screening result and type of resource use, in the Finnish colorectal cancer screening programme Event Screening Total Control p Value Positive test result Negative test result Non-attender Number of individuals 2944 80 975 39 230 123 149 122 930 Outpatient visits* N 764 5705 3190 9659 9136 <0.001 Per cent 26.0 7.0 8.1 7.8 7.4 Inpatient episodes* N 243 2690 1911 4844 4712 <0.001 Per cent 8.3 3.3 4.9 3.9 3.8 Colonoscopies N 339 942 593 1874 1559 <0.001 Per cent 11.5 1.2 1.5 1.5 1.3 The data includes the calendar years 2004–2010. *ICD-10 codes C15-C26, D12 and K00-K93. Figure 1 Number of colonoscopies per individual in the Finnish colorectal cancer screening trial in 2004–2010 by arm.

Event Screening Total Control p Value Positive test result Negative test result Non-attender Number of individuals 2944 80 975 39 230 123 149 122 930 Outpatient visits* N 764 5705 3190 9659 9136 <0.001 Per cent 26.0 7.0 8.1 7.8 7.4 Inpatient episodes* N 243 2690 1911 4844 4712 <0.001 Per cent 8.3 3.3 4.9 3.9 3.8 Colonoscopies N 339 942 593 1874 1559 <0.001 Per cent 11.5 1.2 1.5 1.5 1.3 The data includes the calendar years 2004–2010. *ICD-10 codes C15-C26, D12 and K00-K93. Figure 1 Number of colonoscopies per individual in the Finnish colorectal cancer screening trial in 2004–2010 by arm. A positive FOBT result increased the probability of using hospital resources. After a positive test result, a quarter of the people had at least one outpatient visit, while 93% of those with a negative test result had no outpatient visits at all. Also, the probability of undergoing colonoscopy while in hospital was 9.6 times higher in those with a positive test result compared to those with a negative test result. Furthermore, those who were invited to screening but did not attend were more likely to use hospital resources for gastrointestinal reasons than those with a negative FOBT result (table 3). On average, 2.4% of the screening population had only one outpatient visit and 5.5% had two visits or more. The corresponding numbers for the control arm were 3.9% and 3.5%, respectively. Maximum number of visits per person was 86 in the screening group and 141 in the control group.

alised patients is twofold higher than that of ambulatory outpatients.4 The implications of a poor prep include a higher chance of missed pathology, aborted or delayed procedures and the need for a repeat procedure; all of which contribute to patient inconvenience as well as increased costs to the healthcare system.5–8 Colon cleansing using a split-dose administration of 4 L polyethylene glyocol (PEG) is associated with higher compliance rates, better quality bowel preparations and higher colonoscopy completion rates.9–11 Despite evidence of improved outcomes with the use of a split-dose PEG solution, the majority of patients who are hospitalised receive the standard full dose (NON-SPLIT) 4 L PEG solution prior to inpatient colonoscopy.1 The reasons for this discrepancy remain unclear. Recognising this gap between evidence and clinical practice, we performed this multiphase study to evaluate the feasibility of implementing a split-dose bowel prep regimen for inpatient colonoscopy at our institution. Methods Study design This was a single-centre, prospective observational study conducted at the Malcom Randall VA Medical Center.

A positive FOBT result increased the probability of using hospital resources. After a positive test result, a quarter of the people had at least one outpatient visit, while 93% of those with a negative test result had no outpatient visits at all. Also, the probability of undergoing colonoscopy while in hospital was 9.6 times higher in those with a positive test result compared to those with a negative test result. Furthermore, those who were invited to screening but did not attend were more likely to use hospital resources for gastrointestinal reasons than those with a negative FOBT result (table 3). On average, 2.4% of the screening population had only one outpatient visit and 5.5% had two visits or more. The corresponding numbers for the control arm were 3.9% and 3.5%, respectively. Maximum number of visits per person was 86 in the screening group and 141 in the control group. The cumulative number of outpatient visits, inpatient episodes and hospital colonoscopies over the whole study period is consistently higher in the screening arm than in the control arm (table 4). The difference between the arms was largest in the number of hospital colonoscopies: there were 53% more colonoscopies performed per individual in the screening group than in the control group. For the outpatient visits and inpatient episodes, the number of events was 18% and 17% higher, respectively, in the screening group compared to the control group.

largest in the number of hospital colonoscopies: there were 53% more colonoscopies performed per individual in the screening group than in the control group. For the outpatient visits and inpatient episodes, the number of events was 18% and 17% higher, respectively, in the screening group compared to the control group. Table 4 Number of individuals, number of first screening round events (outpatient visits, inpatient episodes and colonoscopies), number of events per individual by arm and type of resource use, during the first 2 years following randomisation in the Finnish colorectal cancer screening programme Event Number of individuals Number of events Number of events per individual p Value Screening Control Screening Control Screening Control Outpatient visit* 123 149 122 930 31 975 27 061 0.26 0.22 <0.001 Inpatient episode* 123 149 122 930 9260 7903 0.075 0.064 <0.001 Colonoscopy 123 149 122 930 2686 1756 0.022 0.014 <0.001 The data includes calendar years 2004–2010. *ICD-10 codes C15-C26, D12 and K00-K93.

Event Number of individuals Number of events Number of events per individual p Value Screening Control Screening Control Screening Control Outpatient visit* 123 149 122 930 31 975 27 061 0.26 0.22 <0.001 Inpatient episode* 123 149 122 930 9260 7903 0.075 0.064 <0.001 Colonoscopy 123 149 122 930 2686 1756 0.022 0.014 <0.001 The data includes calendar years 2004–2010. *ICD-10 codes C15-C26, D12 and K00-K93. Discussion According to this study, the vast majority of the study population, that is, 60–69-year-old Finnish men and women, had not used hospital resources at all for any gastrointestinal-related reasons. Slightly more people in the screening group than in the control group had at least one hospital-based outpatient visit (7.8% vs 7.4%) or inpatient episode (3.9% vs 3.8%). In total, the screening group had 4914 more cumulative outpatient visits and 1357 more inpatient episodes than the control group. The proportions of test-positive people with at least one outpatient visit (26.0% vs 7.0%) or with at least one inpatient episode (8.3% vs 3.3%) were clearly higher than those of people with a negative FOBT result. The difference in use of hospital resources between the screening and control group is reliably comparable. The setting of the Finnish colorectal cancer screening programme makes it possible to evaluate the feasibility and effectiveness of screening in a randomised, yet real-world healthcare setting. Any possible systematic errors within the Finnish Hospital Discharge registry will be equally present in both groups.

parable. The setting of the Finnish colorectal cancer screening programme makes it possible to evaluate the feasibility and effectiveness of screening in a randomised, yet real-world healthcare setting. Any possible systematic errors within the Finnish Hospital Discharge registry will be equally present in both groups. Limitations of the study One limitation of the study is the lack of primary care data. Resources used at the primary healthcare level, either by the screening or the control groups, were not registered at FHDR during the study years. Small municipalities with inadequate resources refer patients to hospital for the primary colonoscopy. Other municipalities have their own endoscopy units and their primary colonoscopies are not included in our data. In both cases, if any further examination or treatment (eg, removal of large polyps, suspicious cancer) is necessary, the patient is referred to the secondary healthcare. From earlier work,2 we know that 90% of the test-positives complied to colonoscopy, which means that approximately 2650 primary colonoscopies should be added to the figures of the screening group reported here. However, the lack of data on primary healthcare concerns both, and no comprehensive data are available on possible colonoscopies or other diagnostics at the primary healthcare level in the control group or non-attenders. According to national care guidelines,21 the primary investigation for symptomatic patients is endoscopy (not FOBT) and thus controls also have access to primary care endoscopy. The difference in use of primary healthcare services between the two groups can only be speculated, but it is likely that the proportional differences based on FHDR data should indicate any overall differences.

or symptomatic patients is endoscopy (not FOBT) and thus controls also have access to primary care endoscopy. The difference in use of primary healthcare services between the two groups can only be speculated, but it is likely that the proportional differences based on FHDR data should indicate any overall differences. A major concern in FOBT screening and resource use has been the possible increase in the number of diagnostic procedures, mainly colonoscopies, and our study confirms this. In the UK, Price et al22 reported an increase of 21–31% in the number of hospital colonoscopies due to FOBT screening, but a simultaneous increase was also seen in the demand for symptom-based colonoscopies. In Australia, it has been estimated that a biennial FOBT would result in five times more colonoscopies than the option of no screening.23 In our study, a total of 930 more hospital colonoscopies were conducted for the screening group than for the control group, resulting in a 53% increase in hospital colonoscopies. The proportion of people with at least one colonoscopy was slightly higher in the screening group than in the control group (1.5% vs 1.3%). In particular, patients with a positive FOBT result seemed to increase the probability of hospital resource use, including colonoscopies, compared to those with a negative FOBT or to those who did not attend the screening. This implies that an abnormal finding in the screening results in a number of colonoscopies while without screening, the control group will receive only one diagnostic colonoscopy based on the symptoms, without FOBT.

ing colonoscopies, compared to those with a negative FOBT or to those who did not attend the screening. This implies that an abnormal finding in the screening results in a number of colonoscopies while without screening, the control group will receive only one diagnostic colonoscopy based on the symptoms, without FOBT. Based on the present study, it is, however, impossible to say whether the increase is only temporary. For example, Thiis-Evensen et al24 showed that, with a 9-year observation period after a screening colonoscopy, the screening group had 50% fewer usual-care distal endoscopic examinations than the control group. Furthermore, Seeff et al25 have argued that if all of the potential endoscopic capacity existing in the USA was used for organised colorectal cancer screening, the unscreened population could be screened within 1 year using the FOBT and colonoscopy for those with a positive test result. For example, follow-up after polypectomy is carried out at more frequent intervals than recommended. In comparison, many published estimates of colonoscopies in cost-effectiveness analyses are reported only for different screening strategies, while the alternative of ‘no systematic screening’ is often left out or assumed to be zero.11 12 26

after polypectomy is carried out at more frequent intervals than recommended. In comparison, many published estimates of colonoscopies in cost-effectiveness analyses are reported only for different screening strategies, while the alternative of ‘no systematic screening’ is often left out or assumed to be zero.11 12 26 Our screening interval was 2 years. After that, the confounding effect of the second screen is introduced, presuming that the effects of the first and second round are different. In the Finnish CRC screening programme, everyone in the screening group will receive an invitation to screening also during the second round, regardless of the screening result in the first round. The long term evaluation is thus not that straightforward and may include a source of bias. In conclusion, colorectal cancer screening using the FOBT seems to slightly increase the volume of hospital outpatient visits, inpatient episodes and hospital colonoscopies in Finland. The authors would like to acknowledge Jouni Rasilainen (the National Institute for Health and Welfare) for his effort in defining and collecting data. Contributors: NM and MH were involved in study concept and design of the screening programme. All the authors were involved in the study concept and design of this resource allocation study. NM and SM were involved in collecting the data. SM, NM, MH and PR were involved in the data and statistical analyses. All the authors read and approved the final version of the manuscript and were involved in the manuscript drafting.

thors were involved in the study concept and design of this resource allocation study. NM and SM were involved in collecting the data. SM, NM, MH and PR were involved in the data and statistical analyses. All the authors read and approved the final version of the manuscript and were involved in the manuscript drafting. Funding: Dr. Malila reports grants from Cancer Society of Finland during the conduct of the study. Competing interests: None. Ethics approval: National Institute for Health and Welfare, Helsinki, Finland (THL/619/5.05.00/2010). Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Summary box What is already known about this subject? ▸  Corticosteroid (CS) is a first-line therapy for patients with severe ulcerative colitis (UC). ▸  Second-line rescue therapy using tacrolimus (TAC) or infliximab (IFX) is considered for severe UC cases refractory to CS. ▸  Data on the short-term and long-term outcomes of patients with severe UC treated with TAC and IFX are limited. ▸  The efficacy and safety levels of sequential therapies (TAC→IFX or IFX→TAC) remain unclear. What are the new findings? ▸  TAC and IFX had similar effects on remission induction in patients with severe UC. ▸  Sequential therapies could avoid colectomy in patients with severe UC refractory to initial treatment with TAC or IFX. ▸  In our study, there were no serious adverse events in the treatment with sequential therapies. How might it impact on clinical practice in the foreseeable future? ▸  Our data demonstrated that sequential therapies could be considered for patients with severe UC who failed to respond to initial treatment with TAC or IFX. Introduction Ulcerative colitis (UC) is a chronic inflammatory disorder of the colon characterised by diarrhoea, bloody stool, abdominal pain, fever, anaemia and weight loss.1

How might it impact on clinical practice in the foreseeable future? ▸  Our data demonstrated that sequential therapies could be considered for patients with severe UC who failed to respond to initial treatment with TAC or IFX. Introduction Ulcerative colitis (UC) is a chronic inflammatory disorder of the colon characterised by diarrhoea, bloody stool, abdominal pain, fever, anaemia and weight loss.1 A recent study showed that 25% of patients with UC experienced episodes of severe colitis.2 Although introduction of intravenous corticosteroids (CS) has modified the natural history of severe acute relapse, approximately 30–40% of patients with UC are refractory to steroid therapy, and without further medical management these patients will require emergency colectomy.3 Therefore, the development of a second-line rescue therapy for steroid-resistant cases remains an important challenge. The calcineurin-inhibitor cyclosporine (CsA) is an immunosuppressant that inhibits T-cell-mediated production of interleukin-2 (IL-2). CsA was first shown to be an effective rescue therapy for acute severe steroid-refractory UC in 1994.4 Recent studies showed response rates to CsA ranging from 64% to 82%.4 5 Despite the benefits of CsA treatment, it is associated with significant toxicity and adverse reactions, and thus alternative agents are required.

sA was first shown to be an effective rescue therapy for acute severe steroid-refractory UC in 1994.4 Recent studies showed response rates to CsA ranging from 64% to 82%.4 5 Despite the benefits of CsA treatment, it is associated with significant toxicity and adverse reactions, and thus alternative agents are required. Tacrolimus (TAC) is an immunosuppressive macrolide isolated from Streptomyces tsukubaensis.6 TAC inhibits the complexation of calcineurin with its respective cytoplasmic receptors, cyclophilin and FK-binding protein 12, both of which regulate calmodulin-dependent phosphatase activity, thereby interrupting the signal transduction pathway in T cells.7 Although TAC and CsA have similar modes of action, TAC has a 30-fold to 100-fold greater immunosuppressive effect in vitro and a 10-fold to 20-fold greater effect in vivo than CsA, as well as more reliable intestinal absorption, even in the presence of gastrointestinal disease.8 A randomised controlled trial and several case studies have demonstrated that TAC is effective for inducing and maintaining remission in patients with refractory UC.8–11 TAC is thus considered to be a promising therapeutic option for refractory UC. In the past 10 years, infliximab (IFX), a monoclonal antibody that binds free and membrane-bound tumour necrosis factor α, has demonstrated efficacy for UC. Its superior efficacy compared to placebo in moderate-to-severe non-acute UC is well established based on UC trials (ACT1 and ACT2), with a response rate between 61% and 69%.12

0 years, infliximab (IFX), a monoclonal antibody that binds free and membrane-bound tumour necrosis factor α, has demonstrated efficacy for UC. Its superior efficacy compared to placebo in moderate-to-severe non-acute UC is well established based on UC trials (ACT1 and ACT2), with a response rate between 61% and 69%.12 While the use of both TAC and IFX has been shown to be effective and safe as rescue therapy, data directly comparing the effects of these two agents on severe UC are limited.13 14 In the present study, we reviewed a series of patients with severe UC admitted consecutively to our hospital from 2001 up to the present, who had been treated with TAC or IFX. Methods Patients This was a retrospective single centre study in Kyoto University Hospital. We reviewed the medical records of all patients who received at least one IFX infusion or TAC treatment for treatment of severe active UC. The diagnosis of UC was confirmed according to standardised criteria by prior clinical assessment, radiology, endoscopy and histology. We used the same definitions of steroid dependency and refractoriness as the European Crohn's and Colitis Organization consensus.15 According to the Montreal classification, a disease event was categorised as proctitis if present up to 15 cm from the anal verge, as left-sided colitis if present up to but not beyond the splenic flexure, or as extensive colitis if present beyond the splenic flexure.16

ean Crohn's and Colitis Organization consensus.15 According to the Montreal classification, a disease event was categorised as proctitis if present up to 15 cm from the anal verge, as left-sided colitis if present up to but not beyond the splenic flexure, or as extensive colitis if present beyond the splenic flexure.16 From October 2001 to February 2014, all patients with severe UC were treated with TAC or IFX at Kyoto University Hospital. All patients with UC in this study had severe disease activity with a Mayo score greater than 10 and moderate-to-severe inflammation on sigmoidoscopy, despite concurrent treatment with CS and/or immunomodulators (IM).

February 2014, all patients with severe UC were treated with TAC or IFX at Kyoto University Hospital. All patients with UC in this study had severe disease activity with a Mayo score greater than 10 and moderate-to-severe inflammation on sigmoidoscopy, despite concurrent treatment with CS and/or immunomodulators (IM). Treatment regimens were as follows: TAC was administered orally at an initial dose of 0.1 mg/kg/day or was administered intravenously at an initial dose of 0.01 mg/kg/day, and the dose was adjusted to produce whole blood trough levels of 10–15 ng/mL to induce remission. IFX episodic therapy was defined as a single infusion of 5 mg/kg of IFX at induction, followed by further infusion if necessary. Scheduled therapy was defined as intravenous infusions of IFX at a dose of 5 mg/kg at weeks 0, 2 and 6, followed by regular infusions every 8 weeks thereafter. For patients who did not respond to initial treatments either with TAC or IFX, and those who had relapsed despite their response to initial treatment, sequential therapies (TAC→IFX or IFX→TAC) were performed under the global assessment of physicians.17 We collected data on the demographics, baseline characteristics of severe UC, previous and concomitant medications, clinical outcome and adverse events of both treatments. Informed consent for participation in this study, which was approved by the institutional ethics committee, was provided by all patients.

Treatment regimens were as follows: TAC was administered orally at an initial dose of 0.1 mg/kg/day or was administered intravenously at an initial dose of 0.01 mg/kg/day, and the dose was adjusted to produce whole blood trough levels of 10–15 ng/mL to induce remission. IFX episodic therapy was defined as a single infusion of 5 mg/kg of IFX at induction, followed by further infusion if necessary. Scheduled therapy was defined as intravenous infusions of IFX at a dose of 5 mg/kg at weeks 0, 2 and 6, followed by regular infusions every 8 weeks thereafter. For patients who did not respond to initial treatments either with TAC or IFX, and those who had relapsed despite their response to initial treatment, sequential therapies (TAC→IFX or IFX→TAC) were performed under the global assessment of physicians.17 We collected data on the demographics, baseline characteristics of severe UC, previous and concomitant medications, clinical outcome and adverse events of both treatments. Informed consent for participation in this study, which was approved by the institutional ethics committee, was provided by all patients. Clinical outcomes To evaluate disease activity and response to both treatments, a partial Mayo score18–20 was retrospectively determined both before treatment and after 8 weeks of treatment, as well as at the last follow-up visit. In Kyoto University Hospital, we evaluate disease activity of all patients with UC, at least on every 4-week visit after starting their induction therapy, by the partial Mayo score. Clinical remission (CR) was defined as a partial Mayo score of less than 3. Relapse was defined as an increase in the partial Mayo score to 3 or higher with additional therapies required. Mucosal healing was defined as endoscopic Mayo score of 0. Short-term outcomes were evaluated at 8 weeks after initiating TAC or IFX treatment. Long-term outcomes were evaluated using data from patients who were followed up for more than 6 months after TAC or IFX treatment. Moreover, colectomy-free survival was evaluated in all 29 patients with severe UC during the period from the time of their initial treatments to the last follow-up visit.

or IFX treatment. Long-term outcomes were evaluated using data from patients who were followed up for more than 6 months after TAC or IFX treatment. Moreover, colectomy-free survival was evaluated in all 29 patients with severe UC during the period from the time of their initial treatments to the last follow-up visit. Definition of cytomegalovirus reactivation All patients were tested for colonic cytomegalovirus (CMV) reactivation by immunohistochemistry and quantitative real-time PCR using colonic biopsy specimens (mucosal PCR). Colonic biopsy specimens were obtained from inflamed colonic mucosa, fixed in formalin, embedded in paraffin and stained with H&E, and immunohistochemistry was performed using anti-CMV monoclonal antibodies (Dako Cytomation, Kyoto, Japan).21 Cases in which the CMV-DNA copy number was over 10 copies/μg DNA were defined as positive for CMV reactivation. Cases in which CMV was detected by either one of these methods were diagnosed with positive colonic CMV reactivation.22

ry was performed using anti-CMV monoclonal antibodies (Dako Cytomation, Kyoto, Japan).21 Cases in which the CMV-DNA copy number was over 10 copies/μg DNA were defined as positive for CMV reactivation. Cases in which CMV was detected by either one of these methods were diagnosed with positive colonic CMV reactivation.22 Assessment and statistics The primary end point of this study was short-term outcome at 8 weeks after induction of TAC or IFX. Secondary end points included long-term outcome and colectomy-free survival at 2014 after the initiation of treatment. Moreover, we assessed the efficacy and safety of the sequential therapies (TAC→IFX or IFX→TAC) in patients with UC refractory to the initial treatment. Cumulative colectomy-free survival was assessed using the Kaplan-Meir method. Comparisons of differences within groups were performed using the Mann-Whitney U test or Student t test, depending on the distribution. Categorical data were compared by χ2 analysis. A p value of less than 0.05 was considered to be statistically significant.

tomy-free survival was assessed using the Kaplan-Meir method. Comparisons of differences within groups were performed using the Mann-Whitney U test or Student t test, depending on the distribution. Categorical data were compared by χ2 analysis. A p value of less than 0.05 was considered to be statistically significant. Results Patient characteristics Clinical characteristics of all patients with severe UC are shown in table 1. Median age at diagnosis of severe UC was 34 years (range 19–69 years). Median disease duration before the induction therapy was 43 months (range 2–348 months). Median follow-up period was 27 months (range 0.5–118 months). Of the 29 patients, 18 were men and 11 were women. Of the 29 patients, 24 were hospitalised at the induction therapy. Median partial Mayo score was 9 (range 8–9) and the endoscopic Mayo score was 3 (range 2–3). Of the 29 patients, 23 had extensive disease type and the other 6 patients had left-sided disease. C reactive protein level was 7.0 (0.1–27.1) mg/dL, haemoglobin level was 10.3 (7.1–13.4) g/dL and serum albumin level was 2.9 (1.7–4.9) g/dL. Regarding previous response to CS, 14 of the 29 patients (48.3%) were refractory to CS, 11 (37.9%) were dependent on CS and the remaining 4 patients (13.8%) were naïve to CS. As concomitant therapy with the induction therapy, except for one patient who was allergic to 5-aminosalicylate (5-ASA), the patients were all treated with 5-ASA. Prior to initiating treatment with TAC or IFX, 16 of the 29 patients were treated with CS (9 with intravenous CS and 7 with oral CS). Median duration of treatment with systemic CS before initiating TAC or IFX was 36 days (range 14–4200 days). At initiating treatment with TAC or IFX, CS had already tapered off in 9 of the 29 patients. CS was not used for these patients’ induction therapies because they had previous histories of long-term use of CS. Despite being naïve to CS in the remaining four patients, they were not treated with CS because they refused it. Nine of the 29 patients were treated with concomitant thiopurine and 15 were treated with concomitant cytapheresis.

e patients’ induction therapies because they had previous histories of long-term use of CS. Despite being naïve to CS in the remaining four patients, they were not treated with CS because they refused it. Nine of the 29 patients were treated with concomitant thiopurine and 15 were treated with concomitant cytapheresis. Table 1 Clinical characteristics of 29 patients with severe UC

e patients’ induction therapies because they had previous histories of long-term use of CS. Despite being naïve to CS in the remaining four patients, they were not treated with CS because they refused it. Nine of the 29 patients were treated with concomitant thiopurine and 15 were treated with concomitant cytapheresis. Table 1 Clinical characteristics of 29 patients with severe UC All patients Tacrolimus group Infliximab group p Value* (n=29) (n=22) (n=7) Age at diagnosis of severe UC (median (range)) (years) 34 (19–69) 34.5 (19–69) 33 (22–58) 0.799 Disease duration prior to the induction therapy (median (range)) (months) 43 (2–348) 45 (2–348) 32 (6–157) 0.760 Follow-up period (median (range)) (months) 27 (0.5–118) 21 (2–118) 30 (0.5–79) 0.541 Gender 0.758 Male 18 14 4 Female 11 8 3 Hospitalised patients at the induction therapy 24 19 5 0.569 Partial Mayo score (median (range)) 9 (8–9) 8.5 (8–9) 9 (8–9) 0.779 Endoscopic Mayo score (median (range)) 3 (2–3) 3 (3–3) 3 (2–3) 0.476 Disease extent 0.555 Extensive 23 18 5 Left sided 6 4 2 Laboratory examination (mean±SD (range)) (mg/dL) C reactive protein 7.0±7.1 (0.1–27.1) 5.7±5.5 (0.1–19) 11.2±10.3 (1.3–27.1) 0.270 Haemoglobin 10.3±1.8 (7.1–13.4) 10.4±2.0 (7.1–13.4) 10.0±1.1 (9.3–12.5) 0.858 Albumin 2.9±0.6 (1.7–4.0) 2.9±0.6 (1.8–4.0) 2.8±0.7 (1.7–3.7) 0.939 Previous response to corticosteroid Corticosteroid refractory 14 12 2 0.436 Corticosteroid dependent 11 6 5 0.096 Corticosteroid naïve 4 4 0 0.544 Concomitant therapy at the induction therapy† 5-aminosalitirate 28 21 7 1.000 Corticosteroid 16 12 4 1.000 Thiopurine 9 6 3 0.746 Cytapheresis 15 13 2 0.323 Positive rate of CMV reactivation (percentage (positive/all)) (%) 44.8 (13/29) 40.9 (9/22) 57.1 (4/7) 0.452 Numbers of patients are shown unless specified.

0 0.544 Concomitant therapy at the induction therapy† 5-aminosalitirate 28 21 7 1.000 Corticosteroid 16 12 4 1.000 Thiopurine 9 6 3 0.746 Cytapheresis 15 13 2 0.323 Positive rate of CMV reactivation (percentage (positive/all)) (%) 44.8 (13/29) 40.9 (9/22) 57.1 (4/7) 0.452 Numbers of patients are shown unless specified. *Comparison of differences between tacrolimus group and infliximab group was evaluated. †Some patients were treated in conjunction with other therapies. CMV, cytomegalovirus; UC, ulcerative colitis. For the induction therapy, 22 of the 29 patients were treated with TAC (TAC group) and the remaining 7 patients were treated with IFX (IFX group). The choice of TAC or IFX treatment for the induction therapy was conducted according to the primary physician's decision at that time. There were no significant differences in the clinical characteristics between the TAC and IFX groups.

five patients in the TAC group and one patient in the IFX group were treated with IFX in conjunction with IM and 5-ASA. One patient in the TAC group and two patients in the IFX group were treated with IFX in conjunction with 5-ASA. Three patients in the TAC group were treated with TAC in conjunction with IM and 5-ASA. Four patients in the TAC group and two patients in the IFX group were treated with IM and 5-ASA. One patient in the TAC group was treated with IM alone due to intolerance of 5-ASA. Three patients in the TAC group and one patient in the IFX group were treated with 5-ASA. Regarding cumulative dosage of each drug during the observational period, the mean cumulative dosage of IM in the TAC group was more than twice that in the IFX group. Table 2 Medications in patients who achieved a CR at the end of follow-up TAC group IFX group (n=17) (n=6) Observational period (median (range)) (months) 29 (14–118) 32 (22–79) Maintenance therapy at the end of follow-up IFX+IM+5-ASA 5 1 IFX+5-ASA 1 2 TAC+IM+5-ASA 3 0 IM+5-ASA 4 2 IM 1 0 5-ASA 3 1 Cumulative dosage of each drug during observational period TAC (mean (range)) (mg) 2852 (389–5937) 1016 (0–2861) IFX (mean (range)) (mg) 3866 (1500–19 260) 5843 (225–20 520) IM (mean (range)) (mg) 38 910 (7800–110 550) 16 122 (0–42 700) 5-ASA (mean (range)) (g) 4537 (0–10 138) 3609 (828–6994) Numbers of patients are shown unless specified. 5-ASA, 5-aminosalicylate; CR, clinical remission; IFX, infliximab; IM, immunomodulators; TAC, tacrolimus.

TAC group IFX group (n=17) (n=6) Observational period (median (range)) (months) 29 (14–118) 32 (22–79) Maintenance therapy at the end of follow-up IFX+IM+5-ASA 5 1 IFX+5-ASA 1 2 TAC+IM+5-ASA 3 0 IM+5-ASA 4 2 IM 1 0 5-ASA 3 1 Cumulative dosage of each drug during observational period TAC (mean (range)) (mg) 2852 (389–5937) 1016 (0–2861) IFX (mean (range)) (mg) 3866 (1500–19 260) 5843 (225–20 520) IM (mean (range)) (mg) 38 910 (7800–110 550) 16 122 (0–42 700) 5-ASA (mean (range)) (g) 4537 (0–10 138) 3609 (828–6994) Numbers of patients are shown unless specified. 5-ASA, 5-aminosalicylate; CR, clinical remission; IFX, infliximab; IM, immunomodulators; TAC, tacrolimus. Comparison of maintenance therapy between the patients who achieved a CR with TAC and those who relapsed with TAC Of the 29 patients with severe UC, 25 (22 in the TAC group and 3 in the IFX group) were treated with TAC in their clinical course. We analysed the differences in the maintenance therapy between those who achieved a CR with TAC (CR group, n=14) and those who relapsed with TAC (Relapse group, n=11), as shown in table 3. In the 18 patients (10 in the CR group and 8 in the Relapse group), long-term use of TAC (more than 6 months) was performed. As a result, 10 of 18 patients (55.6%) could achieve a CR. Regarding the maintenance therapy with TAC, the median treatment duration with TAC was 471 days (range 325–995 days) in the 10 patients who were finally switched to IM or treated with the combination of TAC and IM.

use of TAC (more than 6 months) was performed. As a result, 10 of 18 patients (55.6%) could achieve a CR. Regarding the maintenance therapy with TAC, the median treatment duration with TAC was 471 days (range 325–995 days) in the 10 patients who were finally switched to IM or treated with the combination of TAC and IM. Table 3 Comparison of maintenance therapy between the patients who achieved a CR with TAC and those who relapsed with TAC CR Relapse (n=14) (n=11) Duration of TAC use (median (range)) (days) 411 (121–995) 405 (63–962) Cessation of TAC within 6 months 4 3 Maintenance therapy with TAC IM+5-ASA 13 7 IM 1 0 5-ASA 0 4 Numbers of patients are shown unless specified. 5-ASA, 5-aminosalicylate; CR, clinical remission; IM, immunomodulators; TAC, tacrolimus. Intensification of IFX treatment Of the 29 patients with severe UC, 17 patients (10 in the TAC group and 7 in the IFX group) were treated with at least more than one infusion of IFX. Nine (6 in the TAC group and 3 in the IFX group) of the 17 patients could achieve a CR after the treatment with IFX. Six (4 in the TAC group and 2 in the IFX group) of the nine patients needed the intensification of IFX treatment such as dose escalation and shortening of interval in order to maintain a CR. All the eight patients (4 in the TAC group and 4 in the IFX group) who could not achieve a CR from the treatment with IFX resulted in colectomy or were switched to TAC without intensifying IFX treatment.

d the intensification of IFX treatment such as dose escalation and shortening of interval in order to maintain a CR. All the eight patients (4 in the TAC group and 4 in the IFX group) who could not achieve a CR from the treatment with IFX resulted in colectomy or were switched to TAC without intensifying IFX treatment. Mucosal healing During our observational periods, 27 of the 29 patients were evaluated by colonoscopy after their treatments, although the timing and frequency of colonoscopy varied among the patients. At the first endoscopic analysis after the initial treatment (median: 31 days, range: 9–1782 days), four patients in the TAC group could achieve mucosal healing. Two of the four patients had been continued with TAC treatment at the first endoscopic examination and the remaining two patients had not. At the latest endoscopic analysis after the initial treatment (median: 851 days, range: 248–2091 days), seven patients could achieve mucosal healing. One of the seven patients could achieve a CR after switching IFX to TAC and the remaining six patients could achieve a CR after treatment with TAC.

two patients had not. At the latest endoscopic analysis after the initial treatment (median: 851 days, range: 248–2091 days), seven patients could achieve mucosal healing. One of the seven patients could achieve a CR after switching IFX to TAC and the remaining six patients could achieve a CR after treatment with TAC. Colectomy-free survival Of the 29 patients with severe UC, 6 patients required colectomy. One resulted in colectomy within 8 weeks after initiating IFX and 3 resulted in colectomy within 6 months after initiating TAC. The remaining two resulted in colectomy later than 6 months after initiating TAC. Based on Kaplan-Meier survival analysis, the overall cumulative colectomy-free survival was estimated to be 79.3% at 118 months (figure 2A). Moreover, the cumulative colectomy-free survival was 77.3% at 118 months (range 2–118 months) in the TAC group and 85.7% at 79 months (range 0.5–79 months) in the IFX group (p=0.704; figure 2B). Regarding the effect of the sequential therapies, 9 of 13 patients who were refractory to the initial treatments (TAC group: n=10, IFX group: n=3) responded to the sequential therapies and avoided colectomy.

118 months) in the TAC group and 85.7% at 79 months (range 0.5–79 months) in the IFX group (p=0.704; figure 2B). Regarding the effect of the sequential therapies, 9 of 13 patients who were refractory to the initial treatments (TAC group: n=10, IFX group: n=3) responded to the sequential therapies and avoided colectomy. Figure 2 (A) Proportion of patients who avoided colectomy (n=23). The overall cumulative colectomy-free survival was estimated to be 79.3% at 118 months, based on Kaplan-Meier survival analysis. (B) Proportion of patients who avoided colectomy (n=17 in the tacrolimus (TAC) group and n=6 in the infliximab (IFX) group). The cumulative colectomy-free survival was estimated to be 77.3% at 118 months in the TAC group (solid line) and 85.7% at 79 months in the IFX group (dashed line) (p=0.704).

er survival analysis. (B) Proportion of patients who avoided colectomy (n=17 in the tacrolimus (TAC) group and n=6 in the infliximab (IFX) group). The cumulative colectomy-free survival was estimated to be 77.3% at 118 months in the TAC group (solid line) and 85.7% at 79 months in the IFX group (dashed line) (p=0.704). Predictors of colectomy To evaluate the predictive factors for colectomy in patients with severe UC, we analysed the differences in the patient characteristics between patients with severe UC who required a colectomy (colectomy group) and those who did not (non-colectomy group; table 4). The positive rate of CMV reactivation in the colectomy group was significantly higher than that in the non-colectomy group (p=0.033). There were no significant differences between the colectomy group and non-colectomy group with regard to age, disease duration, sex, partial Mayo score, endoscopic Mayo score, disease extent, laboratory examination, daily dosage of CS and copy number of CMV-DNA. Therefore, our data suggest that positivity of colonic CMV reactivation is a predictive factor for colectomy, regardless of the copy number of CMV-DNA in the colonic mucosa. Table 4 Predictors in colectomy

Predictors of colectomy To evaluate the predictive factors for colectomy in patients with severe UC, we analysed the differences in the patient characteristics between patients with severe UC who required a colectomy (colectomy group) and those who did not (non-colectomy group; table 4). The positive rate of CMV reactivation in the colectomy group was significantly higher than that in the non-colectomy group (p=0.033). There were no significant differences between the colectomy group and non-colectomy group with regard to age, disease duration, sex, partial Mayo score, endoscopic Mayo score, disease extent, laboratory examination, daily dosage of CS and copy number of CMV-DNA. Therefore, our data suggest that positivity of colonic CMV reactivation is a predictive factor for colectomy, regardless of the copy number of CMV-DNA in the colonic mucosa. Table 4 Predictors in colectomy Colectomy Non-colectomy p Value (n=6) (n=23) Age at diagnosis of severe UC (median (range)) (years) 49.5 (20–64) 33 (19–69) 0.153 Disease duration prior to the induction therapy (median (range)) (months) 50 (4–348) 34 (2–157) 0.590 Gender 0.494 Male 3 15 Female 3 8 Partial Mayo score (median (range)) 8.5 (8–9) 9 (8–9) 0.936 Endoscopic Mayo score (median (range)) 3 (2–3) 3 (3–3) 0.893 Disease extent 0.160 Extensive 6 17 Left sided 0 6 Laboratory examination (mean±SD (range)) (mg/dL) C reactive protein 8.1±9.8 (1.2–27.1) 6.8±6.5 (0.1–18.7) 0.893 Haemoglobin 9.7±1.5 (7.5–11.8) 10.4±1.8 (7.9–13.1) 0.518 Albumin 2.7±0.6 (1.7±3.4) 3.0±0.6 (1.8–4.0) 0.318 Daily dosage of corticosteroid prior to TAC or IFX (median (range)) (mg/day) 17.5 (0–60) 7.5 (0–60) 0.647 Positive rate of CMV reactivation (percentage (positive/all)) (%) 83.3 (5/6) 34.8 (8/23) 0.033 Copy number of CMV-DNA in patients with CMV reactivation (median (range)) (copies/μg) 975 (320–1400) 1300 (390–12 000) 0.445 Numbers of patients are shown unless specified.

edian (range)) (mg/day) 17.5 (0–60) 7.5 (0–60) 0.647 Positive rate of CMV reactivation (percentage (positive/all)) (%) 83.3 (5/6) 34.8 (8/23) 0.033 Copy number of CMV-DNA in patients with CMV reactivation (median (range)) (copies/μg) 975 (320–1400) 1300 (390–12 000) 0.445 Numbers of patients are shown unless specified. CMV, cytomegalovirus; IFX, infliximab; TAC, tacrolimus; UC, ulcerative colitis. Adverse events During our observational periods, no patients died and no life-threatening events occurred. In addition, opportunistic infections and reactivation of tuberculosis or hepatitis B virus were not observed in any of the 29 patients with severe UC. Discussion The purpose of this retrospective study was to evaluate the short-term and long-term clinical outcome of Japanese patients with severe UC treated with TAC, IFX and sequential therapies. The rate of induction of remission at 8 weeks in the TAC and IFX groups was similar. In addition, 69.2% (9/13) of patients required alternative induction regimens (so-called sequential therapies) to avoid colectomy. During this observational study, no serious complications related to these medical therapies were observed. The present data first suggest that sequential therapies could avoid colectomy in cases of severe UC failing to respond to TAC or IFX.

ative induction regimens (so-called sequential therapies) to avoid colectomy. During this observational study, no serious complications related to these medical therapies were observed. The present data first suggest that sequential therapies could avoid colectomy in cases of severe UC failing to respond to TAC or IFX. Patients with severe UC require a pro-active approach with either effective medical treatment or a timely surgical approach. To avoid the increasing mortality of these patients, physicians should carefully determine the therapeutic strategy depending on the patient's condition. Intravenous CS remains the first-line treatment.23 24 High-dose and prolonged exposure to CS, however, predisposes patients to increased infections and peri-operative complications such as thrombosis and anastomotic leakage.25 Thus, in patients with severe UC who did not respond to CS therapies, rescue medical therapies with either a calcineurin inhibitor or antitumour necrosis factor α antibodies are necessary.

owever, predisposes patients to increased infections and peri-operative complications such as thrombosis and anastomotic leakage.25 Thus, in patients with severe UC who did not respond to CS therapies, rescue medical therapies with either a calcineurin inhibitor or antitumour necrosis factor α antibodies are necessary. First, we assessed the rate of induction of remission in patients with severe UC treated with TAC or IFX. Regarding the effect of IFX in hospitalised patients with severely active UC, two randomised control trials (RCTs) evaluated 56 patients and compared biological therapy with placebo. According to these RCTs, IFX tended to be superior to placebo.26 27 As for the efficacy of TAC in severely active UC, previous reports showed that CR rates in patients with severe UC at 1 and 3 months after administering TAC were 45% and 72%, respectively.9 28 Our study demonstrated that there were no differences in the CR rates in patients with severe UC at 8 weeks between the TAC and IFX groups, suggesting that the effects of TAC, a calcineurin inhibitor, are potentially similar to those of IFX on severe UC. Actually, recent RCTs revealed that the efficacies of CsA, another calcineurin inhibitor, were equivalent to those of IFX for acute patients with severe UC refractory to intravenous CS.29

C and IFX groups, suggesting that the effects of TAC, a calcineurin inhibitor, are potentially similar to those of IFX on severe UC. Actually, recent RCTs revealed that the efficacies of CsA, another calcineurin inhibitor, were equivalent to those of IFX for acute patients with severe UC refractory to intravenous CS.29 Second, we assessed the efficacy and safety of sequential therapies in this study. Several studies have examined the effect of sequential therapy with alternative drugs in patients with refractory UC.13 14 30–32 Maser et al reported the efficacy of the sequential therapies in patients with UC refractory to CsA or IFX. The CR rate of the IFX-salvage group (CsA-patients with refractory UC treated with IFX) and the CsA-salvage group (IFX-refractory those with CsA) was 40% and 33%, respectively.32 In the present study, the CR rate of the IFX-salvage group (TAC-refractory, receiving IFX) and the TAC-salvage group (IFX-refractory, receiving TAC) was 60% (6/10) and 100% (3/3), respectively. Of note, all patients requiring sequential therapies had no serious adverse events. Maser's data,32 however, showed a high incidence (16%, 3/19) of serious adverse events, including one death related to infection. The different frequency of serious side effects between our study and Maser's might be due to different doses of CS with which patients in each study had been treated before starting calcineurin inhibitor or biologicals.33 However, gastroenterologists must deliberately follow-up patients with UC treated with sequential therapies to avoid adverse events as much as possible. Several studies have reported the long-term outcomes of patients with severe CS-refractory UC treated with CsA or IFX. Croft et al34 reported that the colectomy-free rate at 12 months was 42% in the CsA group and 65% in the IFX group. The 3-year colectomy-free rates ranged from 43% to 55% in the CsA group, and from 70% to 73% in the IFX group.35–37 In our study, the cumulative colectomy-free survival was 77.3% at 118 months in the TAC group and 85.7% at 79 months in the IFX group (figure 2B). Thus, sequential therapies appeared to contribute to extending the colectomy-free survival term. Moreover, the maintenance therapy, particularly after the induction therapy with TAC in severe UC, is an important issue to prevent colectomy. According to our study, long-term treatment with TAC could induce a clinically better outcome in patients with severe UC.

ntribute to extending the colectomy-free survival term. Moreover, the maintenance therapy, particularly after the induction therapy with TAC in severe UC, is an important issue to prevent colectomy. According to our study, long-term treatment with TAC could induce a clinically better outcome in patients with severe UC. However, further studies will be required to ascertain the efficacy and safety levels of long-term treatment with TAC in patients with refractory UC.

ntribute to extending the colectomy-free survival term. Moreover, the maintenance therapy, particularly after the induction therapy with TAC in severe UC, is an important issue to prevent colectomy. According to our study, long-term treatment with TAC could induce a clinically better outcome in patients with severe UC. However, further studies will be required to ascertain the efficacy and safety levels of long-term treatment with TAC in patients with refractory UC. Next, we analysed the predictors of colectomy in this study. Our data demonstrated that positivity of CMV-DNA in the colonic mucosa was a significant predictor of colectomy. In general, CMV can lead to worsening colitis in patients with moderate-to-severe UC.38 39 Therefore, we always consider the involvement of CMV infection in patients with UC refractory to immunosuppressive therapies. We assessed the CMV-DNA in the colonic mucosa of all enrolled patients before starting TAC or IFX therapy. In the present study, however, there was no significant difference in the copy number of CMV-DNA between the two groups. Heterogeneity of patients’ characteristics might affect these discrepancies. Therefore, the relevance of the copy number of CMV-DNA is an issue to be addressed in the future. Of 13 patients who were positive for colonic CMV-DNA, 7 patients were treated with ganciclovir on the basis of physician's decision. The use of ganciclovir did not affect colectomy rate in this retrospective observational study (p=0.12, data by χ2 analysis). Because there was no difference in the CR ratio at 8 weeks between the two groups, our data suggest that optimal control of colonic inflammation could lead to UC remission despite the lack of antiviral therapy.

clovir did not affect colectomy rate in this retrospective observational study (p=0.12, data by χ2 analysis). Because there was no difference in the CR ratio at 8 weeks between the two groups, our data suggest that optimal control of colonic inflammation could lead to UC remission despite the lack of antiviral therapy. There are some limitations to our study, including the small sample size, heterogeneity of patients’ prior treatment history and therapeutic protocol. Despite the small number of patients, our data suggested the efficacy of sequential therapies for avoiding colectomy in patients with severe UC. However, the difference in the history of patients’ prior treatments might have affected subsequent treatment outcome. Moreover, because this study was a retrospective uncontrolled study, our current data should be assessed carefully. Conclusions In conclusion, our data indicate that TAC, IFX and their sequential therapies could allow some patients with severe UC to avoid colectomy, although further investigation with larger cases is necessary. Contributors: NM, TY, MM and HN designed the study. NM wrote the manuscript and performed a major role in collecting patients’ data. YK, SY, TT, AM, YH, TY, MM and HN collected patients’ data and were involved in editing the manuscript. All authors read and approved the final manuscript.

Conclusions In conclusion, our data indicate that TAC, IFX and their sequential therapies could allow some patients with severe UC to avoid colectomy, although further investigation with larger cases is necessary. Contributors: NM, TY, MM and HN designed the study. NM wrote the manuscript and performed a major role in collecting patients’ data. YK, SY, TT, AM, YH, TY, MM and HN collected patients’ data and were involved in editing the manuscript. All authors read and approved the final manuscript. Funding: This work was supported by the Japan Society for the Promotion of Science “KAKENHI” Grants-in-aid for Scientific Research (24590941, 25860532, 26460967 and 26893122), and Health and Labour Sciences Research Grants for Research on Rare and Intractable Disease from the Ministry of Health, Labour and Welfare, Japan. Competing interests: None. Patient consent: Obtained. Ethics approval: Ethics committee of Kyoto University Hospital. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Summary box What is already known about this subject? ▸  There is a paucity of data on the prevalence of faecal incontinence in irritable bowel syndrome (IBS). ▸  It is generally assumed that, if it occurs, faecal incontinence is largely confined to patients with the diarrhoea predominant variety of the condition. ▸  It is not known to whom IBS patients with faecal incontinence disclose their problem. What are the new findings? ▸  This study shows that in IBS patients referred to secondary and tertiary care, the prevalence of faecal incontinence equals that observed in care homes for the elderly. ▸  There is little difference in the prevalence of faecal incontinence between patients with diarrhoea, constipation and mixed varieties of IBS. ▸  Unless specifically asked, many patients with IBS-related faecal incontinence may not admit to this problem. How might it impact on clinical practice in the foreseeable future? ▸  Future guidelines on the management of IBS need to highlight the fact that faecal incontinence is a major problem in this condition.

▸  There is little difference in the prevalence of faecal incontinence between patients with diarrhoea, constipation and mixed varieties of IBS. ▸  Unless specifically asked, many patients with IBS-related faecal incontinence may not admit to this problem. How might it impact on clinical practice in the foreseeable future? ▸  Future guidelines on the management of IBS need to highlight the fact that faecal incontinence is a major problem in this condition. Introduction Faecal incontinence (FI) is a distressing and disabling condition with many negative, social and psychological consequences such as embarrassment, anxiety, social isolation and loss of employment.1 2 It presents in two main forms which are either the involuntary but recognised passage of gas or liquid or solid stool (urge incontinence), or anal leakage of mucus, fluid or stool (passive incontinence).3 There is a considerable variation in the reported prevalence of FI in the general population, which is due to a variety of reasons including the reluctance of patients to report this symptom, different working definitions and the inadequacy of data collection methods.4 Accordingly, the prevalence ranges from 0.4% to 18% in the general population,1 4–7 although higher prevalences have been reported in at-risk populations. For example, it ranges from 5% to 24% in community-dwelling women in the USA,8 33–46% in patients with inflammatory bowel disease9 and 47–54% in residential nursing homes.10 11 Some diseases are well known to be associated with faecal impaction or FI of which diabetes mellitus, multiple sclerosis and inflammatory bowel disease are examples.10 However, considering the high prevalence of irritable bowel syndrome (IBS), there is surprisingly little information on the rate of incontinence in this condition.12

are well known to be associated with faecal impaction or FI of which diabetes mellitus, multiple sclerosis and inflammatory bowel disease are examples.10 However, considering the high prevalence of irritable bowel syndrome (IBS), there is surprisingly little information on the rate of incontinence in this condition.12 IBS is a functional gastrointestinal disorder with a community prevalence which appears to vary in different geographical areas such as 3% in Iran, 12% in the UK and 30% in Nigeria,13 although this is probably due to different methods of identification. However, the prevalence in the UK appears to be consistently between 10% and 15%. Apart from gastrointestinal symptoms such as abdominal pain, bloating and bowel dysfunction, there are some associated non-gastrointestinal symptoms including lethargy, back pain, urinary symptoms and dyspareunia.12 14 This complex of different symptoms results in a poor quality of life (QOL) and consequently a high rate of psychological problems such as anxiety and depression.14 15 It has also been shown that the emotional burden of disease can lead to the development of hopelessness and suicidal ideation.15 This already distressing condition can be made even worse if a patient has to deal with an embarrassing problem which is difficult to disclose such as FI. Drossman and colleagues surveyed college students and young hospital employees for urgency and faecal soiling. They categorised the participants into three groups consisting of patients with IBS who did and did not visit physicians and patients without IBS. They showed that 20% (13/65) of the patients with IBS who visited a physician and 6.2% (10/162) of the patients who did not visit a doctor reported faecal soiling.16 However, this study was confined to a young population not representative of patients with IBS in general and did not take into consideration the different bowel habit subtypes of IBS. In another study, Longstreth and Wolde-Tsadik screened members of a health maintenance organisation for IBS symptoms based on Manning criteria and subdivided the patients with IBS into less severe and more severe groups according to the number of abdominal pain episodes. Their results showed that 18.5% of the participants in the less severe IBS symptom group and 22.7% in the more severe group suffered from FI,17 but again the patients were not divided into the various bowel habit subtypes.

h IBS into less severe and more severe groups according to the number of abdominal pain episodes. Their results showed that 18.5% of the participants in the less severe IBS symptom group and 22.7% in the more severe group suffered from FI,17 but again the patients were not divided into the various bowel habit subtypes. Although it is generally accepted that IBS can lead to FI especially when associated with diarrhoea, to the best of our knowledge, this issue has not been systematically investigated in an unselected group of patients particularly in relation to bowel habit subtype. Therefore, the aim of this study was to assess the prevalence of FI and its consequences in a consecutive group of patients with IBS attending a clinic specialising in the care of this condition.

has not been systematically investigated in an unselected group of patients particularly in relation to bowel habit subtype. Therefore, the aim of this study was to assess the prevalence of FI and its consequences in a consecutive group of patients with IBS attending a clinic specialising in the care of this condition. Methods In this study, consecutive new and follow-up patients with IBS without any significant concomitant disease attending the clinic fulfilling Rome III criteria for IBS were asked to complete a series of questionnaires over a period of 1 year starting January 2012. The following questionnaires were administered: the IBS Symptom Severity Score (IBS-SSS), the non-colonic symptom score, a QOL measure and the Hospital Anxiety Depression (HAD) scale. A further questionnaire was completed detailing bowel habits (diarrhoea, constipation and alternating), FI, urgency, incomplete evacuation, history of previous surgery (appendicectomy, cholecystectomy, hysterectomy, sterilisation and caesarean section), number of pregnancies and urinary incontinence. The FI questionnaire recorded FI as follows: (1) Never, (2) Less than once a year, (3) Once a year, (4) More than once a year, (5) Once a month, (6) Once a week, (7) More than once a week, (8) Nocturnal and (9) In response to a laxative. The severity of FI was subdivided into mild, moderate and severe categories based on its frequency being ‘once a year or less’, ‘once a month or less but more than once a year’ and ‘once a week or more or nocturnal’, respectively. For constipation patients, particular attention was paid to the relationship between any incontinence resulting from the use of laxatives. In addition, coping strategies such as the use of medication, carrying a change of clothes and the use of incontinence pads were recorded. The IBS-SSS is a validated questionnaire consisting of five subscales: pain severity, pain frequency, abdominal bloating, bowel habit dissatisfaction and life interference with a maximum score of 500.18 Patients are classified as mild, moderate or severe based on a score of 75–175, 175–300 or greater than 300, respectively. The measure of non-colonic symptoms19 includes nausea/vomiting, early satiety, headache, backache, lethargy, excess wind, heartburn, urinary symptoms, thigh pain and bodily aches with a maximum score of 500. The QOL measure19 assesses psychological well-being, physical well-being, mood, locus of control and social/relationships with a maximum score of 500.

includes nausea/vomiting, early satiety, headache, backache, lethargy, excess wind, heartburn, urinary symptoms, thigh pain and bodily aches with a maximum score of 500. The QOL measure19 assesses psychological well-being, physical well-being, mood, locus of control and social/relationships with a maximum score of 500. The HAD scale20 is a validated measure consisting of seven anxiety and seven depression items with a maximum score of 21 for each one. For the purposes of this study, a score of 10 or more was regarded as indicative of significant anxiety or depression. Ethical statement All the questionnaires used in this study have been routinely used in our department for many years to monitor the progress of patients and their response to treatment. All new patients are routinely asked about FI and a more detailed response to this question was recorded for the purposes of this study. The study was submitted to the NHS Health Research Authority and Medical Research Council decision tool, which concluded that a formal ethical review was not necessary.

to treatment. All new patients are routinely asked about FI and a more detailed response to this question was recorded for the purposes of this study. The study was submitted to the NHS Health Research Authority and Medical Research Council decision tool, which concluded that a formal ethical review was not necessary. Statistical analysis Power calculation: for an overall sample of 500 patients, the study would be able to estimate the prevalence of FI with the accuracy of ±5% (ie, the 95% CI for the observed prevalence value would extend 5% to either side of this estimate). For particular subgroups of patients (such as IBS severity) of a size greater than 100, the CIs would not extend more than 10% on either side of the observed prevalence value. Mean values with 95% CI were calculated for HAD scales, QOL domains and non-colonic features. Frequencies, percentages and χ2 tests for categorical variables; means, 95% CI's and independent sample t tests for normally distributed continuous data were used to assess the relationship between variables such as different types of IBS, different severities of patients with FI and IBS overall, as well as between new and follow-up patients. All statistical comparisons used the conventional two-sided 5% significance level. SPSS V.20 was used for all calculations.

data were used to assess the relationship between variables such as different types of IBS, different severities of patients with FI and IBS overall, as well as between new and follow-up patients. All statistical comparisons used the conventional two-sided 5% significance level. SPSS V.20 was used for all calculations. Results Demographics and clinical characteristics Five hundred consecutive patients with IBS completed questionnaires. Of these, 198 (40%) were new and 302 (60%) were follow-up patients. The mean age of the whole group was 46.3 (range 15–87) with 399 (80%) being female and 101 (20%) male. In terms of different types of IBS, the number of patients in the diarrhoea, constipation and alternating subgroups was 178 (35.6%), 140 (28%) and 182 (36.4%), respectively. With regard to IBS severity, 42 (8.4%) had mild, 164 (32.8%) moderate and 294 (58.8%) severe IBS (table 1). Table 1 Irritable bowel syndrome (IBS) severity distribution in new and follow-up patients as well as the whole group

Results Demographics and clinical characteristics Five hundred consecutive patients with IBS completed questionnaires. Of these, 198 (40%) were new and 302 (60%) were follow-up patients. The mean age of the whole group was 46.3 (range 15–87) with 399 (80%) being female and 101 (20%) male. In terms of different types of IBS, the number of patients in the diarrhoea, constipation and alternating subgroups was 178 (35.6%), 140 (28%) and 182 (36.4%), respectively. With regard to IBS severity, 42 (8.4%) had mild, 164 (32.8%) moderate and 294 (58.8%) severe IBS (table 1). Table 1 Irritable bowel syndrome (IBS) severity distribution in new and follow-up patients as well as the whole group IBS severity New patients Follow-up patients Total Mild IBS 11 (5.6%) 31 (10.3%) 42 (8.4%) Moderate IBS 54 (27.3%) 110 (36.4%) 164 (32.8%) Severe IBS 133 (67.2%) 161 (53.3%) 294 (58.8%) Prevalence and management of FI Two hundred and eighty-five patients (57%) reported FI; of these, 68 (23.9%) had mild, 99 (34.7%) moderate and 91 (31.9%) severe FI and 27 (9.5%) had FI in response to a laxative (table 2). Patients with FI were significantly older than non-incontinent patients (49.44 vs42.04, p<0.001), but showed no difference in gender (45.5% female vs 54.5% male, p=0.56). Considering the different types of IBS, the prevalence of FI was: 116/178 (65.2%) in diarrhoea, 53/140 (37.9%) in constipation and 116/182 (63.7%) in the alternating type. In constipated patients with FI, 35.8% had FI due to laxatives and in the remaining 64.2% it was spontaneous. FI was reported in 62% (26/42) of patients with mild IBS, 49.5% (81/164) of moderately severe patients and 61% (178/294) of severe patients, although it should be noted that the number of patients in the mild group was substantially lower than in the other two groups (table 3). The frequency of FI was 99 (50%) in new patients and 186 (61.6%) in follow-up patients (p=0.01). Twenty-eight per cent of patients with FI used loperamide (imodium) to control their FI, whereas 66% carried a change of clothes and 30% utilised incontinence pads on a regular basis.

than in the other two groups (table 3). The frequency of FI was 99 (50%) in new patients and 186 (61.6%) in follow-up patients (p=0.01). Twenty-eight per cent of patients with FI used loperamide (imodium) to control their FI, whereas 66% carried a change of clothes and 30% utilised incontinence pads on a regular basis. Table 2 Faecal incontinence (FI) severity distribution in new and follow-up patients as well as the whole group FI severity New patients with FI Follow-up patients with FI Total Mild FI 24 (24.2%) 44 (23.7%) 68 (23.9%) Moderate FI 32 (32.3%) 67 (36.0%) 99 (34.7%) Severe FI 32 (32.3%) 59 (31.7%) 91 (31.9%) In response to a laxative 11 (11.1%) 16 (8.6%) 27 (9.5%) Table 3 Comparison of faecal incontinence (FI) severity in patients with mild, moderate and severe irritable bowel syndrome (IBS)

otal Mild FI 24 (24.2%) 44 (23.7%) 68 (23.9%) Moderate FI 32 (32.3%) 67 (36.0%) 99 (34.7%) Severe FI 32 (32.3%) 59 (31.7%) 91 (31.9%) In response to a laxative 11 (11.1%) 16 (8.6%) 27 (9.5%) Table 3 Comparison of faecal incontinence (FI) severity in patients with mild, moderate and severe irritable bowel syndrome (IBS) IBS severity groups FI severity Number of patients with FI Percentage Mild IBS (n=42) Mild 9 21.5 Moderate 11 26.2 Severe 3 7.1 In response to a laxative 3 7.1 Total 26 61.9 Moderate IBS (n=164) Mild 30 18.3 Moderate 30 18.3 Severe 14 8.5 In response to a laxative 7 4.3 Total 81 49.4 Severe IBS (n=294) Mild 29 9.9 Moderate 58 19.7 Severe 74 25.2 In response to a laxative 17 5.8 Total 178 60.6 During the course of the study it became clear that the prevalence of FI was even more than anticipated. It was therefore decided to add a further question about whether the patient had disclosed the problem to anyone and, if so, to whom it had been revealed. From the total number of 176 patients, 14 (8%) had disclosed the problem only to their general practitioner (GP), 46 (26.1%) had told only a friend or member of their family, 75 (42.6%) had disclosed the problem to their GP and a member of their family and 41 (23.3%) had told nobody.

f so, to whom it had been revealed. From the total number of 176 patients, 14 (8%) had disclosed the problem only to their general practitioner (GP), 46 (26.1%) had told only a friend or member of their family, 75 (42.6%) had disclosed the problem to their GP and a member of their family and 41 (23.3%) had told nobody. Clinical associations In terms of the main IBS symptoms of pain and bloating, there was no significant association between these two features and FI (97.2% vs 95.1%, p=0.231 and 93% vs 90.2%, p=0.261, respectively). FI was significantly associated with urgency (table 4, p<0.001) but not with incomplete evacuation (table 4, p=0.824) and was much more likely to be associated with urinary incontinence (table 4, p<0.001). The prevalence of previous abdominal surgery in patients with FI was significantly higher than that observed in those without FI (66% vs 45%, p<0.001), and this difference reached significance for appendicectomy, cholecystectomy and hysterectomy (p=0.014, 0.001 and 0.003, respectively). Type of delivery and number of pregnancies were significantly associated with the prevalence of FI, meaning that the higher the number of pregnancies, the higher the risk of FI (p=0.024). Vaginal delivery was more likely to be associated with FI than caesarean section (63.1% vs 3.4%, p=0.004). Also, a history of forceps delivery was significantly higher in females with FI (70.2% vs 29.8%, p=0.039). Table 4 Frequency of different symptoms and previous surgery in patients with and without faecal incontinence (FI)

Clinical associations In terms of the main IBS symptoms of pain and bloating, there was no significant association between these two features and FI (97.2% vs 95.1%, p=0.231 and 93% vs 90.2%, p=0.261, respectively). FI was significantly associated with urgency (table 4, p<0.001) but not with incomplete evacuation (table 4, p=0.824) and was much more likely to be associated with urinary incontinence (table 4, p<0.001). The prevalence of previous abdominal surgery in patients with FI was significantly higher than that observed in those without FI (66% vs 45%, p<0.001), and this difference reached significance for appendicectomy, cholecystectomy and hysterectomy (p=0.014, 0.001 and 0.003, respectively). Type of delivery and number of pregnancies were significantly associated with the prevalence of FI, meaning that the higher the number of pregnancies, the higher the risk of FI (p=0.024). Vaginal delivery was more likely to be associated with FI than caesarean section (63.1% vs 3.4%, p=0.004). Also, a history of forceps delivery was significantly higher in females with FI (70.2% vs 29.8%, p=0.039). Table 4 Frequency of different symptoms and previous surgery in patients with and without faecal incontinence (FI) Patients with FI (285/500) Patients without FI (215/500) p Value Urgency 277 (97.5%) 167 (77.7%) <0.001 Incomplete evacuation 242 (84.9%) 181 (84.3%) 0.824 Urinary incontinence 98 (34.4%) 29 (13.6%) <0.001 Appendicectomy 55 (19.3%) 24 (11.2%) 0.014 Cholecystectomy 38 (13.3%) 9 (4.2%) 0.001 Hysterectomy (in females) 69 (30.3%) 29 (17.2%) 0.003 Sterilisation 27 (9.5%) 15 (7.0%) 0.319 HAD scores The mean score for anxiety in new patients with FI was significantly higher than that in those who were not incontinent (10.63 vs 9.18, p=0.041). In the follow-up group, the mean anxiety score was higher in patients with FI than in those without FI (10.73 vs 10.42), but this did not reach significance. The mean depression score in new and follow-up patients with FI was significantly higher than that in those without FI but did not reach the cut-off for being clinically significant (8.26 vs 5.91, p<0.001; 7.96 vs 6.40, p=0.002, table 5).

I than in those without FI (10.73 vs 10.42), but this did not reach significance. The mean depression score in new and follow-up patients with FI was significantly higher than that in those without FI but did not reach the cut-off for being clinically significant (8.26 vs 5.91, p<0.001; 7.96 vs 6.40, p=0.002, table 5). Table 5 Psychological, quality of life and non-colonic scores in new and follow-up patients with and without faecal incontinence (FI) New patients with FI (99/500) New patients without FI (99/500) p Value Follow-up patients with FI (186/500) Follow-up patients without FI (116/500) p Value Anxiety, mean scores (95% CI) 10.63 (9.61 to 11.64) 9.18 (8.23 to 10.13) 0.041 10.73 (10.01 to 11.44) 10.42 (9.52 to 11.32) 0.603 Depression, mean scores (95% CI) 8.26 (7.33 to 9.19) 5.91 (5.08 to 6.74) <0.001 7.96 (7.25 to 8.66) 6.40 (5.68 to 7.11) 0.002 Quality of life, mean scores (95% CI) 252.54 (231.83 to 273.24) 220.42 (200.76 to 240.07) 0.027 252.68 (238.35 to 267) 223.43 (208.34 to 238.52) 0.006 Non-colonic features, mean scores (95% CI) 239.54 (219.31 to 259.77) 210.32 (193.70 to 226.95) 0.028 249.31 (235.54 to 263.08) 209.22 (193.80 to 224.63) <0.001 Quality of life The QOL of new and follow-up patients who were suffering from FI was significantly worse than that of those without FI (p=0.027, p=0.006, table 5). Non-colonic features Both new and follow-up patients with FI reported more severe non-colonic features of IBS than did those without FI, with the difference between their total scores being statistically significant (p=0.028, p<0.001, table 5).

New patients with FI (99/500) New patients without FI (99/500) p Value Follow-up patients with FI (186/500) Follow-up patients without FI (116/500) p Value Anxiety, mean scores (95% CI) 10.63 (9.61 to 11.64) 9.18 (8.23 to 10.13) 0.041 10.73 (10.01 to 11.44) 10.42 (9.52 to 11.32) 0.603 Depression, mean scores (95% CI) 8.26 (7.33 to 9.19) 5.91 (5.08 to 6.74) <0.001 7.96 (7.25 to 8.66) 6.40 (5.68 to 7.11) 0.002 Quality of life, mean scores (95% CI) 252.54 (231.83 to 273.24) 220.42 (200.76 to 240.07) 0.027 252.68 (238.35 to 267) 223.43 (208.34 to 238.52) 0.006 Non-colonic features, mean scores (95% CI) 239.54 (219.31 to 259.77) 210.32 (193.70 to 226.95) 0.028 249.31 (235.54 to 263.08) 209.22 (193.80 to 224.63) <0.001 Quality of life The QOL of new and follow-up patients who were suffering from FI was significantly worse than that of those without FI (p=0.027, p=0.006, table 5). Non-colonic features Both new and follow-up patients with FI reported more severe non-colonic features of IBS than did those without FI, with the difference between their total scores being statistically significant (p=0.028, p<0.001, table 5). Discussion This study has shown that FI is extremely common in this population of patients with IBS and is nearly as prevalent as it is in the elderly in residential care. It might be anticipated that FI would be largely confined to the diarrhoea predominant variety of the condition, but this was not the case with over a third of the patients with constipation experiencing this problem, which was not expected. In 35.8% of constipated patients with FI, this was due to laxatives but in the remaining 64.2% it was spontaneous, and this latter observation is particularly important as it could have an impact on how laxatives are prescribed. The surprisingly high incontinence rate in the constipated group might at least be partly explained by pelvic floor damage. We have previously shown that patients with long-standing constipation can develop a similar level of pelvic floor damage as women who have given birth vaginally, presumably as a result of repetitive straining over many years.21 Managing bowel dysfunction in patients with an alternating bowel habit is always challenging as there is a dilemma of whether to advise laxatives, antidiarrhoeals or to avoid such medications completely. The fact that this group has an FI rate approximately the same as in those with diarrhoea emphasises how difficult it is to manage this problem. Consequently, it is important to ascertain whether an apparent alternating bowel habit is actually of a constipated patient with laxative-induced diarrhoea or an antidiarrhoeal-induced constipation in a diarrhoea predominant patient. From the management point of view, knowledge of the problem of FI and careful enquiry about the nature of the bowel dysfunction are critically important as they can affect the choice of medication as well as enable the physician to offer advice, especially as incontinence products are becoming increasingly available.

. From the management point of view, knowledge of the problem of FI and careful enquiry about the nature of the bowel dysfunction are critically important as they can affect the choice of medication as well as enable the physician to offer advice, especially as incontinence products are becoming increasingly available. The prevalence of incontinence in the follow-up patients was somewhat higher than that in the new patients and this is likely to be explained by the fact that patients who improve are discharged from the clinic, whereas those who continue to struggle with symptoms such as incontinence will remain under follow-up. It might be argued that mild incontinence, defined in this study as one episode a year or less, is trivial and not worth worrying about, although the sufferers would probably dispute this view given the nature and unpredictability of the problem. However, even if this figure is excluded, the prevalence of incontinence continues to be extremely high and its distribution between the different bowel habit subgroups remains similar. Most patients find it difficult to talk about FI and when asked the additional question about whether they had disclosed the problem to anyone, it is noteworthy that only half (50.6%) had disclosed the problem to their GP, 68.7% had told a friend or a member of their family and 23.3% had not told anybody. This is clearly a worrisome trend indicating that many of these individuals are failing to get any help or support with this issue.

ed the problem to anyone, it is noteworthy that only half (50.6%) had disclosed the problem to their GP, 68.7% had told a friend or a member of their family and 23.3% had not told anybody. This is clearly a worrisome trend indicating that many of these individuals are failing to get any help or support with this issue. It has been previously shown that bladder symptoms are very common in IBS, especially frequency, urgency and urge incontinence.22 23 Therefore, it is possibly not surprising that urinary incontinence was relatively common in patients with FI. Consequently, the burden that these two problems have on social functioning as well as the sex lives of these sufferers must be overwhelming and probably contributes to the high rates of sexual dysfunction24 and suicidal ideation that have previously been reported in these patients.15 It is now well described that when compared to the general population, patients with IBS undergo an excess of abdominal operations such as appendicectomy, cholecystectomy and hysterectomy.25 Not only are they unnecessary in some instances but also there is a strong clinical impression that they can make symptoms such as pain worse and more difficult to manage. It is therefore noteworthy that in this study a history of abdominal surgery was more common in patients with FI than in those without and that this reached significance for appendicectomy, cholecystectomy and hysterectomy.

clinical impression that they can make symptoms such as pain worse and more difficult to manage. It is therefore noteworthy that in this study a history of abdominal surgery was more common in patients with FI than in those without and that this reached significance for appendicectomy, cholecystectomy and hysterectomy. Loperamide is probably the most commonly recommended medication in patients with diarrhoea, urgency and FI, although it has little or no effect on pain and can actually make it worse. There are also claims that it can improve anal tone.26 Consequently, it was somewhat surprising that only 29% of faecally incontinent patients found this drug helpful, although the majority of users were in the diarrhoea group, which is possibly what might be anticipated. It is also worrying to think that 30% of these patients were having to wear incontinence pads on a regular basis and that their mean age was only 54.

nly 29% of faecally incontinent patients found this drug helpful, although the majority of users were in the diarrhoea group, which is possibly what might be anticipated. It is also worrying to think that 30% of these patients were having to wear incontinence pads on a regular basis and that their mean age was only 54. An obvious weakness of this study is that it was undertaken in a specialist IBS clinic where patients are unlikely to reflect the IBS population as a whole. However, a whole population study might be a difficult task as a definitive diagnosis of IBS, especially with respect to bowel habit subclassification, is most commonly only established in secondary care, with primary care physicians often preferring to make a symptom diagnosis such as constipation, diarrhoea or abdominal pain. One way of adjusting for this would be to exclude patients with severe symptom scores and even after this exercise, the prevalence of FI in this study was still 51.7% in moderate and mild patients who are more likely to reflect the general population of this condition. Conclusion The results of this study indicate that FI is a major problem in IBS and that patients are not necessarily going to disclose it without being specifically asked. If this issue is raised, they will at least realise that they are not alone in suffering from this problem, which can be specifically targeted with a variety of management strategies that could help the situation considerably.

and that patients are not necessarily going to disclose it without being specifically asked. If this issue is raised, they will at least realise that they are not alone in suffering from this problem, which can be specifically targeted with a variety of management strategies that could help the situation considerably. Contributors: SA, SR and PJW conceived and designed the study. They drafted the article and also approved the final version. SA and SR collected the data and performed the statistical analysis. PJW is the guarantor. Funding: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. Competing interests: In the past 5 years, PJW has received research grants or honoraria from Danone, Almirall, Shire, Abbott and Norgine who may in any way gain financial support from the publication of this manuscript either now or in the future. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Summary box What is already known about this subject? ▸  Hypomagnesaemia may result from diarrhoea or malabsorption or it can be induced by medications such as cisplatin preparations. ▸  The US Food and Drug Administration issued an alert stating that the long-term use of proton pump inhibitors (PPIs) may result in hypomagnesaemia. ▸  Statistically significant relations were observed between lower serum magnesium concentrations and PPI treatment in US inpatients. What are the new findings? ▸  Statistically significant relations were observed between lower serum magnesium concentrations and PPI treatment in Japanese outpatients. ▸  Serum magnesium levels did not differ among the groups of patients taking the three types of PPIs. ▸  PPI users with cirrhosis had significantly lower serum magnesium levels than non-users with cirrhosis and than patients without cirrhosis, regardless of PPI use. How might it impact on clinical practice in the foreseeable future? ▸  PPIs are frequently used in everyday medical practice. Although PPI-induced hypomagnesaemia is extremely rare, it may be life threatening in some patients.

▸  PPI users with cirrhosis had significantly lower serum magnesium levels than non-users with cirrhosis and than patients without cirrhosis, regardless of PPI use. How might it impact on clinical practice in the foreseeable future? ▸  PPIs are frequently used in everyday medical practice. Although PPI-induced hypomagnesaemia is extremely rare, it may be life threatening in some patients. Introduction Proton pump inhibitors (PPIs) are widely used to treat gastrointestinal disorders, such as gastric ulcers, duodenal ulcers and reflux oesophagitis.1 Because these drugs are well tolerated, increased numbers of patients receive long-term PPI treatment. However, PPI-associated hypomagnesaemia has been reported in several patients in the USA and Europe.2 Magnesium is the fourth most common cation in the body. Although found primarily in the muscles and bones and a constituent of numerous body structures, magnesium is also essential for enzymes involved in generating ATP, thus playing a key role in regulating numerous physiological functions.3 Hypomagnesaemia may result from diarrhoea or malabsorption or it can be induced by medications such as cisplatin preparations.4 Although hypomagnesaemia may cause serious conditions, such as tetany, spasms and arrhythmias, some patients are asymptomatic. Thus, the specific incidence of hypomagnesaemia associated with PPIs remains unknown.

result from diarrhoea or malabsorption or it can be induced by medications such as cisplatin preparations.4 Although hypomagnesaemia may cause serious conditions, such as tetany, spasms and arrhythmias, some patients are asymptomatic. Thus, the specific incidence of hypomagnesaemia associated with PPIs remains unknown. In May 2011, the US Food and Drug Administration issued an alert stating that the long-term use of PPIs may result in hypomagnesaemia.5 Serum magnesium concentrations were found to be significantly lower in the western inpatients and outpatients who received long-term PPI treatment compared with those who did not.6 7 Serum magnesium concentrations of patients receiving PPIs have only been rarely investigated in Japan, and there is only one case report.8 We therefore examined the effect of long-term use of PPIs on serum magnesium concentrations in Japanese outpatients.

ived long-term PPI treatment compared with those who did not.6 7 Serum magnesium concentrations of patients receiving PPIs have only been rarely investigated in Japan, and there is only one case report.8 We therefore examined the effect of long-term use of PPIs on serum magnesium concentrations in Japanese outpatients. Methods We studied 1742 individuals treated as outpatients at the Ishikawa Prefectural Central Hospital between October and November 2011. Our general hospital has a focus on digestive diseases. The study participants were 481 outpatients who met the inclusion criteria as follows: age >20 years and no history of hospitalisation within 1 week of blood sampling. Exclusion criteria were as follows: dialysis patients and patients who were administered magnesium oxide, diuretics or cisplatin preparations that may have affected serum magnesium. Cirrhosis was defined by clinical findings, blood examination data and image diagnosis (CT or ultrasound). Outpatients underwent blood sampling in the morning after at least a 12 h overnight fast. We measured serum magnesium levels in units of 0.1 mg/dL. Patient records were retrospectively reviewed to determine their underlying diseases (hypertension, diabetes, cirrhosis and dyslipidaemia), possibility of ingestion, oral medications used, long-term use of PPIs and serum magnesium concentrations. Patients administered a PPI for >1 year were regarded as PPI users, whereas those administered a PPI for <1 year or not at all were considered PPI non-users. At the time this study was conducted, only three PPIs were approved for clinical use in Japan: omeprazole, lansoprazole and rabeprazole. The dose of each drug was omeprazole 20 mg, lansoprazole 15 or 30 mg and rabeprazole 10 mg. Further, during this study, none of the patients were hospitalised for hypomagnesaemia. The protocol and consent form for this study were approved by the Institutional Review Board of Ishikawa Prefectural Central Hospital, and all patients provided written informed consent.

nsoprazole 15 or 30 mg and rabeprazole 10 mg. Further, during this study, none of the patients were hospitalised for hypomagnesaemia. The protocol and consent form for this study were approved by the Institutional Review Board of Ishikawa Prefectural Central Hospital, and all patients provided written informed consent. Statistical analysis Mean, SD and percentage with frequency were used to report continuous and discrete variables. χ2 test or two-sample t test was used to assess significant differences between two groups. Multiple linear regression analysis was used to determine whether serum Mg concentrations were significantly associated with PPI use; the mode was adjusted for potential confounders, including age, sex, diagnosis of cirrhosis, diabetes, hypertension, dyslipidaemia and PPI. A p value <0.05 was considered statistically significant. All statistical analyses were performed using SPSS II statistical software (SPSS Japan Inc, Tokyo, Japan).

d with PPI use; the mode was adjusted for potential confounders, including age, sex, diagnosis of cirrhosis, diabetes, hypertension, dyslipidaemia and PPI. A p value <0.05 was considered statistically significant. All statistical analyses were performed using SPSS II statistical software (SPSS Japan Inc, Tokyo, Japan). Results Patient characteristics are shown in table 1. Serum magnesium concentrations were significantly lower in PPI users than in PPI non-users (1.86±0.21 vs 1.91±0.19 mg/dL, p<0.01). The distribution of serum magnesium concentrations in PPI users and non-users is shown in figure 1. Of the 481 patients, 199 were PPI users and 282 were PPI non-users (administered a PPI or not at all: 263 patients; period of 6 months from 1 month: 15 patients; period of 1 year from 6 months: 4 patients). Their mean ages were 67.5±11.7 and 62.8±14.6 years, respectively (p<0.01). Serum magnesium stratified by age is shown in table 2. There were significant differences in serum magnesium in PPI users (1.86±0.04 mg/dL) and non-users (1.91±0.04 mg/dL) in patients aged >61 years. Table 1 Patient demographic and clinical characteristics

Results Patient characteristics are shown in table 1. Serum magnesium concentrations were significantly lower in PPI users than in PPI non-users (1.86±0.21 vs 1.91±0.19 mg/dL, p<0.01). The distribution of serum magnesium concentrations in PPI users and non-users is shown in figure 1. Of the 481 patients, 199 were PPI users and 282 were PPI non-users (administered a PPI or not at all: 263 patients; period of 6 months from 1 month: 15 patients; period of 1 year from 6 months: 4 patients). Their mean ages were 67.5±11.7 and 62.8±14.6 years, respectively (p<0.01). Serum magnesium stratified by age is shown in table 2. There were significant differences in serum magnesium in PPI users (1.86±0.04 mg/dL) and non-users (1.91±0.04 mg/dL) in patients aged >61 years. Table 1 Patient demographic and clinical characteristics PPI users (n=199) PPI non-users (n=282) p Value Age Year, mean±SD (range) 67.5±11.7 (31–91) 62.8±14.6 (20–87) <0.01* Sex Male/female 108 (54.2%)/91 (45.8%) 151 (53.5%)/131 (46.5%) NS† Cirrhosis Present/absent 55 (27.6%)/144 (72.4%) 57 (20.2%)/225 (79.8%) 0.06† Diabetes Present/absent 26 (13.1%)/173 (86.9%) 34 (12.1%)/248 (87.9%) NS† Hypertension Present/absent 60 (30.2%)/139 (69.8%) 57 (20.2%)/225 (79.8%) <0.01† Dyslipidaemia Present/absent 35 (17.6%)/164 (82.4%) 22 (7.8%)/260 (92.2%) <0.01† Serum Cr mg/dL, mean±SD (range) 0.79±0.30 (0.35–3.68) 0.73±0.26 (0.26–2.73) <0.01* Serum Mg mg/dL, mean±SD (range) 1.86±0.21 (1.2–2.5) 1.91±0.19 (1.1–2.5) <0.01* Serum Alb g/dL, mean±SD (range) 4.06±0.54 (2.5–5.1) 4.08±0.53 (2.4–5.2) NS* *Calculated using the t test.

35 (17.6%)/164 (82.4%) 22 (7.8%)/260 (92.2%) <0.01† Serum Cr mg/dL, mean±SD (range) 0.79±0.30 (0.35–3.68) 0.73±0.26 (0.26–2.73) <0.01* Serum Mg mg/dL, mean±SD (range) 1.86±0.21 (1.2–2.5) 1.91±0.19 (1.1–2.5) <0.01* Serum Alb g/dL, mean±SD (range) 4.06±0.54 (2.5–5.1) 4.08±0.53 (2.4–5.2) NS* *Calculated using the t test. †Calculated using the χ2 test. PPI, proton pump inhibitor; NS, no significance. Table 2 Serum magnesium stratified by age Age Serum Mg, mg/dL, mean±SD p Value 20–60 PPI-user (n=43) 1.87±0.04 0.44 Non-user (n=91) 1.90±0.04 61∼ PPI-user (n=156) 1.86±0.04 <0.01 Non-user (n=192) 1.91±0.04 Calculated using the t test. PPI, proton pump inhibitor. Figure 1 Distribution of serum magnesium concentrations in proton pump inhibitor (PPI) users and non-users. Their underlying diseases included hypertension in 30.2% and 20.2%, respectively (p<0.01), and dyslipidaemia in 17.6% and 7.8%, respectively (p<0.01). There were no significant differences between PPI users and non-users in the occurrence of cirrhosis or diabetes. Serum creatinine concentrations were significantly higher in PPI users than in PPI non-users (0.79±0.30 vs 0.73±0.26 mg/dL, p<0.01). Serum albumin concentrations did not differ significantly between PPI users and PPI non-users (table 1). Serum magnesium concentrations were similar in patients taking the PPI omeprazole (1.86±0.24 mg/dL), lansoprazole (1.86±0.17 mg/dL) and rabeprazole (1.86±0.20 mg/dL; table 3). Table 3 Serum magnesium concentrations in patients receiving three types of PPIs

Their underlying diseases included hypertension in 30.2% and 20.2%, respectively (p<0.01), and dyslipidaemia in 17.6% and 7.8%, respectively (p<0.01). There were no significant differences between PPI users and non-users in the occurrence of cirrhosis or diabetes. Serum creatinine concentrations were significantly higher in PPI users than in PPI non-users (0.79±0.30 vs 0.73±0.26 mg/dL, p<0.01). Serum albumin concentrations did not differ significantly between PPI users and PPI non-users (table 1). Serum magnesium concentrations were similar in patients taking the PPI omeprazole (1.86±0.24 mg/dL), lansoprazole (1.86±0.17 mg/dL) and rabeprazole (1.86±0.20 mg/dL; table 3). Table 3 Serum magnesium concentrations in patients receiving three types of PPIs OPZ (n=77) LPZ (n=39) RPZ (n=83) Serum Mg mg/dL, mean±SD (range) 1.86±0.24 (1.3–2.5) 1.86±0.17 (1.5–2.3) 1.86±0.20 (1.2–2.5) LPZ, lansoprazole; OPZ, omeprazole; PPI, proton pump inhibitor; RPZ, rabeprazole.

Their underlying diseases included hypertension in 30.2% and 20.2%, respectively (p<0.01), and dyslipidaemia in 17.6% and 7.8%, respectively (p<0.01). There were no significant differences between PPI users and non-users in the occurrence of cirrhosis or diabetes. Serum creatinine concentrations were significantly higher in PPI users than in PPI non-users (0.79±0.30 vs 0.73±0.26 mg/dL, p<0.01). Serum albumin concentrations did not differ significantly between PPI users and PPI non-users (table 1). Serum magnesium concentrations were similar in patients taking the PPI omeprazole (1.86±0.24 mg/dL), lansoprazole (1.86±0.17 mg/dL) and rabeprazole (1.86±0.20 mg/dL; table 3). Table 3 Serum magnesium concentrations in patients receiving three types of PPIs OPZ (n=77) LPZ (n=39) RPZ (n=83) Serum Mg mg/dL, mean±SD (range) 1.86±0.24 (1.3–2.5) 1.86±0.17 (1.5–2.3) 1.86±0.20 (1.2–2.5) LPZ, lansoprazole; OPZ, omeprazole; PPI, proton pump inhibitor; RPZ, rabeprazole. Linear regression analysis for each patient characteristic, with use or non-use of a PPI as an explanatory variable and serum magnesium concentration as a response variable, showed that the latter differed significantly in patients with and without liver cirrhosis and with and without PPI use (table 4). Multiple linear regression analysis showed that both of these factors were significantly predictive of magnesium concentration (table 5). Serum magnesium concentration was significantly lower in patients with and without cirrhosis (1.82±0.22 vs 1.91±0.19 mg/dL, p<0.01; figure 2). Among patients with cirrhosis, long-term PPI users had significantly lower serum magnesium level than PPI non-users (1.78±0.22 vs 1.87±0.22 mg/dL, p=0.03). Although among the non-cirrhosis patients, long-term PPI users did not have significantly lower serum magnesium levels than PPI non-users (1.89±0.20 vs 1.92±0.18 mg/dL).

patients with cirrhosis, long-term PPI users had significantly lower serum magnesium level than PPI non-users (1.78±0.22 vs 1.87±0.22 mg/dL, p=0.03). Although among the non-cirrhosis patients, long-term PPI users did not have significantly lower serum magnesium levels than PPI non-users (1.89±0.20 vs 1.92±0.18 mg/dL). Table 4 Univariate analysis of the association between patient characteristics and serum magnesium concentrations Variables Standardised coefficients p Value 95% CI Age 0.04 0.38 −0.00 0.01 Sex 0.03 0.52 −0.02 0.04 Cirrhosis −0.18 0.00 −0.12 −0.04 Diabetes −0.02 0.71 −0.06 0.04 Hypertension 0.00 0.92 −0.04 0.04 Dyslipidaemia −0.02 0.58 −0.07 0.04 PPI −0.12 0.01 −0.09 −0.01 Serum Cr 0.03 0.49 −0.03 0.07 Serum Alb 0.02 0.96 −0.03 0.03 PPI, proton pump inhibitor. Table 5 Multiple linear regression analysis of factors predictive of lower serum magnesium concentration Variables Standardised coefficients p Value 95% CI Cirrhosis −0.17 0.01 −0.12 −0.04 PPI −0.11 0.02 −0.08 −0.01 PPI, proton pump inhibitor. Figure 2 Serum magnesium levels in cirrhosis and non-cirrhosis (PPI, proton pump inhibitor) users. Discussion We have shown here that serum magnesium concentrations were lower in Japanese outpatients who did receive long-term PPI treatment than in those who did not. Although several cases of PPI-associated hypomagnesaemia have been reported in the USA and Europe,2 9–12 only one has been reported in Japan.8

Figure 2 Serum magnesium levels in cirrhosis and non-cirrhosis (PPI, proton pump inhibitor) users. Discussion We have shown here that serum magnesium concentrations were lower in Japanese outpatients who did receive long-term PPI treatment than in those who did not. Although several cases of PPI-associated hypomagnesaemia have been reported in the USA and Europe,2 9–12 only one has been reported in Japan.8 Hypomagnesaemia has been observed in >20% of hospitalised patients and in 65% of patients in intensive care units.13–15 Among US inpatients, those who received long-term PPI treatment had significantly lower serum magnesium levels than PPI non-users.6 Our results in Japanese outpatients were similar to those reported for US inpatients and indicate that PPI-induced hypomagnesaemia is also highly probable among outpatients. Magnesium is the fourth most common cation in the body, with a standard concentration range of 1.9–3.1 mg/dL. Maintaining serum magnesium levels depends primarily on food intake and effective function by the kidneys and small intestine.3 Hypomagnesaemia may be caused by diarrhoea; malabsorption syndrome; malabsorption caused by excessive alcohol consumption; and nephrotoxic agents such as cisplatin, amphotericin B and cyclosporine.4 Because magnesium is essential for the enzymatic reactions involved in generating ATP as well as for enzymes that are involved in nucleic acid metabolism, magnesium insufficiency can affect ATP generation and neural transmission.

l consumption; and nephrotoxic agents such as cisplatin, amphotericin B and cyclosporine.4 Because magnesium is essential for the enzymatic reactions involved in generating ATP as well as for enzymes that are involved in nucleic acid metabolism, magnesium insufficiency can affect ATP generation and neural transmission. The mechanism of PPI-induced hypomagnesaemia has not yet been determined. Renal magnesium wasting has not been observed to date in patients with PPI-induced hypomagnesaemia, suggesting that the latter is associated with the gastrointestinal system. Magnesium is absorbed in the small intestine through active and passive transport systems. Active transcellular process mediated by the transport proteins TRPM6 and TRPM7 are normally responsible for approximately 30% of the absorption, while a passive paracellular pathway through enterocyte tight junctions is responsible for the remaining 70%.16 17 The functions of these TRPM6 and TRPM7 channels may be impaired by a lack of protons or affected patients may be heterozygous for mutations in the TRPM6 genes.4 18 Alternatively, the passive paracellular transport mechanism may be much less efficient.

nterocyte tight junctions is responsible for the remaining 70%.16 17 The functions of these TRPM6 and TRPM7 channels may be impaired by a lack of protons or affected patients may be heterozygous for mutations in the TRPM6 genes.4 18 Alternatively, the passive paracellular transport mechanism may be much less efficient. Clinical symptoms have often been observed in patients with serum magnesium concentrations <1.2 mg/dL.16 Of our patients, one PPI user and one PPI non-user had serum magnesium concentrations below 1.2 mg/dL, although neither was clinically symptomatic. Although serum magnesium levels were significantly lower in PPI users than in non-users, few had concentrations low enough to trigger clinical symptoms. Patients with PPI-associated hypomagnesaemia had extremely low magnesium levels.3 4 6 7 11 19 Although additional patients should be examined, patients with extremely low magnesium levels may have factors that could further exacerbate PPI-induced hypomagnesaemia. Serum magnesium concentrations decreased with increasing age in PPI users (table 2). Age was not a significant factor in our multivariate analysis (table 5). Therefore, hypomagnesaemia should be carefully monitored even in patients receiving long-term PPI treatment.

Clinical symptoms have often been observed in patients with serum magnesium concentrations <1.2 mg/dL.16 Of our patients, one PPI user and one PPI non-user had serum magnesium concentrations below 1.2 mg/dL, although neither was clinically symptomatic. Although serum magnesium levels were significantly lower in PPI users than in non-users, few had concentrations low enough to trigger clinical symptoms. Patients with PPI-associated hypomagnesaemia had extremely low magnesium levels.3 4 6 7 11 19 Although additional patients should be examined, patients with extremely low magnesium levels may have factors that could further exacerbate PPI-induced hypomagnesaemia. Serum magnesium concentrations decreased with increasing age in PPI users (table 2). Age was not a significant factor in our multivariate analysis (table 5). Therefore, hypomagnesaemia should be carefully monitored even in patients receiving long-term PPI treatment. Serum magnesium levels did not differ among the groups of patients taking the three types of PPIs. Long-term administration of all these preparations, omeprazole, lansoprazole and rabeprazole, has been reported to result in hypomagnesaemia. Therefore, regardless of the preparation used, hypomagnesaemia should be carefully monitored in patients receiving long-term PPI treatment.

tients taking the three types of PPIs. Long-term administration of all these preparations, omeprazole, lansoprazole and rabeprazole, has been reported to result in hypomagnesaemia. Therefore, regardless of the preparation used, hypomagnesaemia should be carefully monitored in patients receiving long-term PPI treatment. Our study also showed that cirrhosis was associated with hypomagnesaemia. Because biosynthesis reactions and reactions involving trace elements occur in the liver, patients with cirrhosis have reduced trace element concentrations as well as low serum magnesium.20 We found that serum magnesium levels were lower in patients with cirrhosis than in those without. Further, PPI users with cirrhosis had significantly lower serum magnesium levels than non-users with cirrhosis and than patients without cirrhosis, regardless of PPI use (figure 2). Although the effect of long-term PPI administration on serum magnesium levels in patients with cirrhosis has not yet been examined in detail and PPI-induced hypomagnesaemia is rare, it may be life threatening in Japanese outpatients with cirrhosis.

and than patients without cirrhosis, regardless of PPI use (figure 2). Although the effect of long-term PPI administration on serum magnesium levels in patients with cirrhosis has not yet been examined in detail and PPI-induced hypomagnesaemia is rare, it may be life threatening in Japanese outpatients with cirrhosis. This study had several limitations. First, it was performed at a single institution. Since PPI-associated hypomagnesaemia is extremely rare, randomised comparative trials of serum magnesium concentrations and PPI use were virtually impossible. Second, the study included only outpatients who underwent blood sampling. Blood is sampled from only some outpatients, introducing a selection bias and preventing these results to be extrapolated to normal healthy adults. Third, a history of whether these patients took supplements containing magnesium was not evaluated completely. Although the study protocol excluded patients with a history of administration of medications that are believed to affect serum magnesium levels, supplements containing magnesium were not assessed and may have affected our results. Fourth, nutritional status (diet, alcohol intake and diarrhoea) that could affect serum magnesium levels was unknown in this study.

ts with a history of administration of medications that are believed to affect serum magnesium levels, supplements containing magnesium were not assessed and may have affected our results. Fourth, nutritional status (diet, alcohol intake and diarrhoea) that could affect serum magnesium levels was unknown in this study. Conclusion PPIs are frequently used in everyday medical practice. PPI-induced hypomagnesaemia may be life threatening in some patients. Some patients with PPI-induced hypomagnesaemia are asymptomatic, and this may be overlooked because of lack of information. Thus, we recommend that serum magnesium concentrations be measured periodically in outpatients receiving long-term PPI treatment. To the best of our knowledge, this study is the first to show an association between long-term PPI treatment and hypomagnesaemia in patients with cirrhosis. It is necessary to recognise the possibility of hypomagnesaemia when treating cirrhosis patients with PPIs. Contributors: YT had the original idea. KuT and SY contributed to the data analyses. KeT implemented the database. HD performed the laboratory analyses. YT wrote the initial draft and the final version of the manuscript. All the authors revised and approved the final version. Competing interests: None. Patient consent: Obtained. Ethics approval: The protocol and consent form for this study were approved by the Institutional Review Board of the Ishikawa Prefectural Central Hospital. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Summary box What is already known about this subject? ▸  Absorptive hypercalciuria (AH) is characterised by the hypersensitivity of calcium-sensing receptors of antral G cells. ▸  Patients with AH have normal fasting gastrinaemia and hypergastrinaemia stimulated by peptones and calcium. ▸  At present, no morphological and immunohistochemical studies on gastric biopsies of patients with AH have been published. What are the new findings? ▸  Patients with AH showed morphological G-cell hyperplasia, independent from therapies or Helicobacter pylori infection. ▸  We found an unexpected association between AH and fundic gland polyps, with a 10-fold raised prevalence when compared with the general population, a difference extremely statistically significant. How might it impact on clinical practice in the foreseeable future? ▸  It is to be elucidated if even in patients with sporadic fundic gland polyps, not treated with proton pump inhibitors, hypergastrinaemia may play a role. ▸  Alternatively, somatic β-catenin mutation of sporadic fundic gland polyps may stimulate G-cell hyperplasia. Introduction Idiopathic hypercalciuria is found in up to 40% of stone formers, but has an incidence of less than 10% in the overall population.1 The syndrome displays a large clinical variability with patients almost equally distributed between fasting or renal type (prevalent renal calcium loss) and absorptive type (prevalent increase of intestinal absorption).

und in up to 40% of stone formers, but has an incidence of less than 10% in the overall population.1 The syndrome displays a large clinical variability with patients almost equally distributed between fasting or renal type (prevalent renal calcium loss) and absorptive type (prevalent increase of intestinal absorption). Absorptive hypercalciuria (AH) is a clinical condition characterised by a large increase of calcium in the urine, formation of renal stones and osteoporosis in spite of normal levels in the blood. AH is characterised by hypersensitivity of calcium-sensing receptor (CaSR) of antral G cells, with normal fasting gastrinaemia and meal hypergastrinaemia.2 G-cell CaSR is the predominant chemosensor mediating gastrin secretion.3 To the best of our knowledge, there are no published data about the morphology and immunohistochemistry of gastric biopsies of patients with AH. So we studied the gastric biopsies of a group of 38 patients with an established diagnosis of AH,2 analysing their morphology, immunohistochemical features and associated lesions.

To the best of our knowledge, there are no published data about the morphology and immunohistochemistry of gastric biopsies of patients with AH. So we studied the gastric biopsies of a group of 38 patients with an established diagnosis of AH,2 analysing their morphology, immunohistochemical features and associated lesions. Patients and methods Criteria for patient selection were published elsewhere.2 All 38 patients had calcium urinary excretion higher than 250 mg/day (6.25 mmol/day, normal values, 2.5–6.25 mmol/day); they were on a free diet and were not taking calcium-sparing diuretics. After 1 month of a dietician-assisted calcium-free diet, they showed a reduction in daily calcium excretion (<100 mg/day or 2.5 mmol/day), and a decrease of calcium/creatinine values in fasting 2 h urine (<0.35 mmol Ca/mmol creatinine, normal values 0.1–0.2 mmol Ca/mmol creatinine), thus fulfilling the conventional criteria for AH diagnosis. They were all free of severe dyspepsia, and none had taken antiacid therapy. Whereas fasting gastrinaemia was comparable to controls (60–70 ng/L), patients with AH responded to both the calcium load test (1 g calcium gluconate Calcium Sandoz fortissimo) and peptone load test (10 mg Liebig meal extract diluted in 250 mL of 0.9% saline) with an abnormal rise in gastrinaemia with almost doubling values when compared with controls. The patients were referred to our institution with the principal aim of evaluating gastric antra and to find a possible correlation between meal hypergastrinaemia and the morphology of antral G cells.

They were all free of severe dyspepsia, and none had taken antiacid therapy. Whereas fasting gastrinaemia was comparable to controls (60–70 ng/L), patients with AH responded to both the calcium load test (1 g calcium gluconate Calcium Sandoz fortissimo) and peptone load test (10 mg Liebig meal extract diluted in 250 mL of 0.9% saline) with an abnormal rise in gastrinaemia with almost doubling values when compared with controls. The patients were referred to our institution with the principal aim of evaluating gastric antra and to find a possible correlation between meal hypergastrinaemia and the morphology of antral G cells. Endoscopically, all patients did not show significant lesions, except for sparse antral microerosions or hyperaemia; in five patients, small body-fundic polyps were seen that were biopsied. Five antral and body normal controls and biopsies, randomly taken from the antrum (in all 38 patients) and body-fundus (in 27 patients), plus the five body-fundus polyps were fixed in Bouin, embedded in paraffin, cut at 3 µ and stained with H&E and modified Giemsa. Helicobacter pylori were extensively searched for on modified Giemsa section at 40× on both antral and body biopsies. Further sections were cut for the immunohistochemical study of endocrine population using anti-Gastrin polyclonal antiserum (Dako), diluted 1:300 for 30′, room temperature, and an anti-Chromogranin A monoclonal antibodies cocktail (LK2H10 and PHE5 MoAbs, Neomarker) diluted 1:100 for 45′ room temperature. We used four normal antral and body-fundic biopsies as controls.

cal study of endocrine population using anti-Gastrin polyclonal antiserum (Dako), diluted 1:300 for 30′, room temperature, and an anti-Chromogranin A monoclonal antibodies cocktail (LK2H10 and PHE5 MoAbs, Neomarker) diluted 1:100 for 45′ room temperature. We used four normal antral and body-fundic biopsies as controls. Enterochromaffine like (ECL) cells (Chromogranin A+) were scored according to Solcia et al.4 We scored antral G cells as follows:5 6 ▸  Normal: 1–2 G cells for each gland, with uneven distribution; ▸  Simple hyperplasia: 4–5 cells for each gland; ▸  Linear hyperplasia: continuous chain-like distribution of G cells (figure 1). Figure 1 Scoring system of antral G cells.5 6 In a subset experiment, we counted all gastrin-positive cells in controls, simple hyperplasia and linear hyperplasia, evaluating at least 20 well-oriented gastric pits-neck-glands. Results were compared unpaired t test, calculated using the internet resource at http://graphpad.com/quickcalcs/ Gastritis was assessed following the approach by Owen, distinguishing chronic active and quiescent, superficial and deep (under the neck region) gastritis, assessing the presence of H. pylori, atrophy and intestinal metaplasia.7 We used strict criteria for the diagnosis of fundic gland polyps (FGPs),8–11 that is: ▸  Shortened gastric pits; ▸  Presence of irregular dilations, both superficial and deep, bordered by all three cell types of body mucosa—mucous, parietal and chief cells. All statistical analyses were performed using the GraphPad Quick calc resource at http://graphpad.com/

We used strict criteria for the diagnosis of fundic gland polyps (FGPs),8–11 that is: ▸  Shortened gastric pits; ▸  Presence of irregular dilations, both superficial and deep, bordered by all three cell types of body mucosa—mucous, parietal and chief cells. All statistical analyses were performed using the GraphPad Quick calc resource at http://graphpad.com/ Results Clinical features Table 1 shows anatomical and clinical features of the 38 patients with AH. They were mostly female (F/M: 4.4/1), with a median age of 58.2 years. Table 1 Clinical features of the 38 patients with AH are resumed AH patients 38 ♀ 31 ♂ 7 Average age 58.2 F/M 4.4/1 Helicobacter pylori Negative 12 Positive 26 Antral G cells* N − + 4 ++ 8 +++ − N − + 11 ++ 15 +++ − Body ECL cells* N 3 + 1 ++ − ++++ − N 11 + − ++ − +++ 2 Associated lesions 2 FGPs 3 FGPs Score key: N normal population, + simple hyperplasia, ++ linear hyperplasia, +++ micronodular hyperplasia. *The G-cell and ECL scores were evaluated before and after H. pylori eradication therapy, and they appeared unchanged. AH, absorptive hypercalciuria; FGP, fundic gland polyp. At least an antral biopsy was performed in all the patients, and body-fundic biopsies in 27 patients. None of the patients had been treated with proton pump inhibitor therapy previously.

*The G-cell and ECL scores were evaluated before and after H. pylori eradication therapy, and they appeared unchanged. AH, absorptive hypercalciuria; FGP, fundic gland polyp. At least an antral biopsy was performed in all the patients, and body-fundic biopsies in 27 patients. None of the patients had been treated with proton pump inhibitor therapy previously. Histology At first endoscopy, 26 patients showed active superficial gastritis with H. pylori colonisation. None of the 27 patients who had body-fundic biopsy showed atrophy or metaplastic change. Patients with H. pylori gastritis were treated with our standard eradication therapy (amoxicillin 1 g twice a day, clarithromycin 500 mg twice a day, omeprazole 20 mg twice a day, all drugs for 1 week). They underwent a second endoscopic examination with a mean interval of 20.6 months (minimum 4 to maximum 40 months). Their antral biopsies appeared H. pylori free with no residual significant inflammation.

oxicillin 1 g twice a day, clarithromycin 500 mg twice a day, omeprazole 20 mg twice a day, all drugs for 1 week). They underwent a second endoscopic examination with a mean interval of 20.6 months (minimum 4 to maximum 40 months). Their antral biopsies appeared H. pylori free with no residual significant inflammation. The immunohistochemical staining for gastrin was performed on 5 controls and 38 patients and repeated again after eradication on the 26 patients originally with gastritis. The G-cell score of the first examination and the second after eradication appeared unchanged, thus not being affected by H. pylori infection or proton pump inhibitors (PPI) therapy. The five controls showed normal G-cell values. All 38 patients showed a G-cell hyperplasia (simple in 15 and linear in 23; figure 2).The 12 patients, originally H. pylori free, showed simple hyperplasia in 4 cases and linear hyperplasia in 8 cases; the 26 patients, retested after H. pylori eradication, showed simple and linear hyperplasia in 11 and 15 cases, respectively. The different distribution of simple and linear hyperplasia between the patients originally H. pylori negative and patients eradicated, judged with Fisher's exact test, was not statistically significant (p=0.7281). Figure 2 Simple (A) and linear hyperplasia (B) highlighted by immunohistochemical stain for gastrin ((A) gastrin antiserum, 10× original magnification; (B) gastrin antiserum, 4× original magnification).

The immunohistochemical staining for gastrin was performed on 5 controls and 38 patients and repeated again after eradication on the 26 patients originally with gastritis. The G-cell score of the first examination and the second after eradication appeared unchanged, thus not being affected by H. pylori infection or proton pump inhibitors (PPI) therapy. The five controls showed normal G-cell values. All 38 patients showed a G-cell hyperplasia (simple in 15 and linear in 23; figure 2).The 12 patients, originally H. pylori free, showed simple hyperplasia in 4 cases and linear hyperplasia in 8 cases; the 26 patients, retested after H. pylori eradication, showed simple and linear hyperplasia in 11 and 15 cases, respectively. The different distribution of simple and linear hyperplasia between the patients originally H. pylori negative and patients eradicated, judged with Fisher's exact test, was not statistically significant (p=0.7281). Figure 2 Simple (A) and linear hyperplasia (B) highlighted by immunohistochemical stain for gastrin ((A) gastrin antiserum, 10× original magnification; (B) gastrin antiserum, 4× original magnification). Evaluating at least 20 well-oriented antral glands in controls, cases with simple and linear hyperplasia, mean G cells per gland were scored (table 2), and the differences were evaluated with an unpaired t test. The difference between controls and simple hyperplasia was extremely statistically significant (p=0.0003), whereas the difference between simple and linear hyperplasia was very statistically significant (p=0.0032).

mean G cells per gland were scored (table 2), and the differences were evaluated with an unpaired t test. The difference between controls and simple hyperplasia was extremely statistically significant (p=0.0003), whereas the difference between simple and linear hyperplasia was very statistically significant (p=0.0032). Table 2 The mean number of G cells per gland was scored in controls, simple and linear hyperplasia and the differences evaluated with unpaired t test were considered very statistically significant Controls Simple hyperplasia Linear hyperplasia Mean G cells per gland 1.18 3.48 7.12 SD 1.37 2.32 4.61 p=0.0003 p=0.0032 While studying the 27 body-fundic biopsies with an anti-Chromogranin antibody, we detected a normal ECL cell distribution in 25 cases, and a micronodular hyperplasia in only 2 cases. In both of them, the micronodular body hyperplasia was associated with linear G-cell hyperplasia. No signs of atrophy or parietal cell hypertrophy were present in all biopsies. In five patients, we found an association with FGPs, that is, polyps characterised by shortened gastric pits and superficial and deep dilations bordered by mucous, parietal and chief cells (figure 3). No associated signs of parietal cell hypertrophy were seen, both in FGPs and in the 27 body-fundic biopsies. None of the polyps showed any sign of dysplasia.

ion with FGPs, that is, polyps characterised by shortened gastric pits and superficial and deep dilations bordered by mucous, parietal and chief cells (figure 3). No associated signs of parietal cell hypertrophy were seen, both in FGPs and in the 27 body-fundic biopsies. None of the polyps showed any sign of dysplasia. Figure 3 Tiny polyps covered by pink glistening mucosa in the body-fundus were apparent at endoscopy (A); histologically, they showed superficial and deep cystic dilation, covered by mucus (arrow), parietal and chief cells, diagnostic for fundic gland polyps (B); antigastrin immunostaining evidence a persistent linear hyperplasia after the Helicobacter pylori eradication (C) ((B) H&E, ×10 original magnification; (C) gastrin antiserum, ×4 original magnification). Patient 1—A 59-year-old man, with antral active gastritis with H. pylori colonisation at the time of the first control, showed a normal body-fundic mucosa, without polyps or irregularities. Immunohistochemical stainings detected linear G-cell hyperplasia and a normal ECL cell population. On the next endoscopic control, 40 months after the eradication treatment, the gastric body presented some tiny sessile polyps (1–2 mm). The antral mucosa showed a mild hyperaemia H. pylori negative, and a persistent linear G-cell hyperplasia (figure 3).

ected linear G-cell hyperplasia and a normal ECL cell population. On the next endoscopic control, 40 months after the eradication treatment, the gastric body presented some tiny sessile polyps (1–2 mm). The antral mucosa showed a mild hyperaemia H. pylori negative, and a persistent linear G-cell hyperplasia (figure 3). Patient 2—A 59-year-old woman, with intense chronic active gastritis in the antral mucosa H. pylori positive, with simple G-cell hyperplasia associated to superficial active H. pylori positive gastritis in the body, normal endocrine population. The following posteradication control, 26 months after the first examination, detected a residual chronic quiescent H. pylori negative gastritis in the antral mucosa, with simple G-cell hyperplasia. Tiny sessile polyps with the typical FGPs histological features were seen in the gastric body. Patient 3—A 58-year-old woman showed at first chronic active H. pylori positive gastritis, with a linear G-cell hyperplasia, and a normal body-fundic mucosa. Thirty months later, the post-eradication endoscopic control evidenced a persistent linear G-cell hyperplasia in an already normal antral mucosa H. pylori negative, with FGPs in the body-fundus. Patient 4—A 37-year-old woman had at the first endoscopic control normal antral mucosa H. pylori negative, with linear G-cell hyperplasia and associated FGPs. Patient 5—A 60-year-old woman, with normal antral mucosa H. pylori negative at the first endoscopic control had a linear G-cell hyperplasia, and FGPs.

Patient 3—A 58-year-old woman showed at first chronic active H. pylori positive gastritis, with a linear G-cell hyperplasia, and a normal body-fundic mucosa. Thirty months later, the post-eradication endoscopic control evidenced a persistent linear G-cell hyperplasia in an already normal antral mucosa H. pylori negative, with FGPs in the body-fundus. Patient 4—A 37-year-old woman had at the first endoscopic control normal antral mucosa H. pylori negative, with linear G-cell hyperplasia and associated FGPs. Patient 5—A 60-year-old woman, with normal antral mucosa H. pylori negative at the first endoscopic control had a linear G-cell hyperplasia, and FGPs. The prevalence of FGPs in patients with AH (13.1%) was confronted with FGPs’ prevalence in our general endoscopic population (0.7%; 11) by mean of χ2 test with Yates correction. The difference between the two percentages was extremely statistically significant (p<0.0001).

Patient 5—A 60-year-old woman, with normal antral mucosa H. pylori negative at the first endoscopic control had a linear G-cell hyperplasia, and FGPs. The prevalence of FGPs in patients with AH (13.1%) was confronted with FGPs’ prevalence in our general endoscopic population (0.7%; 11) by mean of χ2 test with Yates correction. The difference between the two percentages was extremely statistically significant (p<0.0001). Discussion We studied morphologically and immunohistochemically the antral biopsies (and body-fundus biopsies in 27 cases) of 38 patients with AH. AH is characterised by a large increase of calcium in the urine, with renal stones and osteoporosis, with normal calcium levels in the blood, due to hypersensitivity of CaSRs located on the antral G-cell membrane. These patients show normal fasting gastrinaemia and meal hypergastrinaemia (Ca2+ and peptones) with median gastrin levels almost doubled compared with normal controls.2 It is now known that gastric CaSR and gastrin-secreting G cells are both activated by Ca2+, amino acids and elevated pH. This association strongly suggests an important role of CaSRs in the meal-stimulated gastrin secretion and regulation.3 The first aim of our histological and immunohistochemical work was to define if patients with AH and meal hypergastrinaemia had morphological antral G-cell alterations. None of the 38 patients had a history of significant dyspepsia or had received antisecretory therapies.

Discussion We studied morphologically and immunohistochemically the antral biopsies (and body-fundus biopsies in 27 cases) of 38 patients with AH. AH is characterised by a large increase of calcium in the urine, with renal stones and osteoporosis, with normal calcium levels in the blood, due to hypersensitivity of CaSRs located on the antral G-cell membrane. These patients show normal fasting gastrinaemia and meal hypergastrinaemia (Ca2+ and peptones) with median gastrin levels almost doubled compared with normal controls.2 It is now known that gastric CaSR and gastrin-secreting G cells are both activated by Ca2+, amino acids and elevated pH. This association strongly suggests an important role of CaSRs in the meal-stimulated gastrin secretion and regulation.3 The first aim of our histological and immunohistochemical work was to define if patients with AH and meal hypergastrinaemia had morphological antral G-cell alterations. None of the 38 patients had a history of significant dyspepsia or had received antisecretory therapies. According to literature data,12 and our own experience, normal controls showed 1–2 G cells per gland, with irregular distribution. All our 38 patients presented G-cell hyperplasia, either simple (4–5 cells per gland, uniformly distributed) or linear G-cell hyperplasia (continuous chain-like distribution). In a subset experiment, we scored the mean G cell per gland in our five controls, simple and linear hyperplasia (1.18; 3.48; 7.12, respectively), and we found that the difference between controls and simple hyperplasia was extremely statistically significant (p=0.0003), and the difference between simple and linear hyperplasia was very statistically significant (p=0.0032).

gland in our five controls, simple and linear hyperplasia (1.18; 3.48; 7.12, respectively), and we found that the difference between controls and simple hyperplasia was extremely statistically significant (p=0.0003), and the difference between simple and linear hyperplasia was very statistically significant (p=0.0032). Twenty-six patients initially presented with H. pylori gastritis. As H. pylori infection may induce a G-cell reactive hyperplasia,13 immunohistochemical stainings were performed again after eradication therapy, with a very long mean interval (20.6 months), with identical results in all cases. This very long interval seems to rule out a possible influence of both H. pylori and PPI therapy. None of the patients had shown significant body atrophy, another possible confounding condition for G-cell hyperplasia. Moreover, comparing the percentages of simple and linear hyperplasia between the 12 patients originally H. pylori negative and the 26 patients eradicated, there was not any statistically significant difference. This control between the two groups appears as a further proof of the primitivity of G-cell hyperplasia in patients with AH. So we could conclude that the observed G-cell hyperplasia was primitive and that it represented the morphological counterpart of meal hypergastrinaemia in AH.

Twenty-six patients initially presented with H. pylori gastritis. As H. pylori infection may induce a G-cell reactive hyperplasia,13 immunohistochemical stainings were performed again after eradication therapy, with a very long mean interval (20.6 months), with identical results in all cases. This very long interval seems to rule out a possible influence of both H. pylori and PPI therapy. None of the patients had shown significant body atrophy, another possible confounding condition for G-cell hyperplasia. Moreover, comparing the percentages of simple and linear hyperplasia between the 12 patients originally H. pylori negative and the 26 patients eradicated, there was not any statistically significant difference. This control between the two groups appears as a further proof of the primitivity of G-cell hyperplasia in patients with AH. So we could conclude that the observed G-cell hyperplasia was primitive and that it represented the morphological counterpart of meal hypergastrinaemia in AH. On the other hand, only 2 of 27 patients with a body biopsy had a micronodular endocrine cell hyperplasia. Evidently, in patients with AH, the meal hypergastrinaemia is not high enough to produce marked pictures of endocrine hyperplasia in the body-fundic mucosa, as seen in the Zollinger-Ellison syndrome.14

So we could conclude that the observed G-cell hyperplasia was primitive and that it represented the morphological counterpart of meal hypergastrinaemia in AH. On the other hand, only 2 of 27 patients with a body biopsy had a micronodular endocrine cell hyperplasia. Evidently, in patients with AH, the meal hypergastrinaemia is not high enough to produce marked pictures of endocrine hyperplasia in the body-fundic mucosa, as seen in the Zollinger-Ellison syndrome.14 Unexpectedly, we found in five patients with AH an association with FGPs. FGPs are tiny sessile polyps of the gastric body-fundic mucosa, so far described as sporadic, without other gastrointestinal associations,8 10 syndromic forms associated with familial adenomatous polyposis, classic or attenuated,15 16 with Zollinger-Ellison syndrome,14 17 and recently as part of the gastric adenocarcinoma-associated proximal polyposis of the stomach (GAPPS).6 18 In 2 of our 12 patients, with gastric antrum H. pylori negative, the polyps were already present at the first endoscopic examination. In the other three patients, FGPs appeared in the second endoscopic control, after H. pylori eradication. So even in patients with AH, we found an inverse relationship between polyps and H. pylori infection.19 20

ith gastric antrum H. pylori negative, the polyps were already present at the first endoscopic examination. In the other three patients, FGPs appeared in the second endoscopic control, after H. pylori eradication. So even in patients with AH, we found an inverse relationship between polyps and H. pylori infection.19 20 The prevalence of FGPs associated with patients with AH was more than 10-fold that of sporadic FGPs (13, 1% vs 0, 7% of sporadic FGPs), and this difference was extremely statistically significant. Aprile et al14 demonstrated in a multivariate analysis that in 106 patients with Zollinger-Ellison syndrome there was a statistically significant association between polyp formation and levels of gastrinaemia. Similarly, gastrin may a have a role in the increased prevalence of FGPs in patients with AH. In a preliminary experiment, we stained with a gastrin antiserum the gastric antrum biopsies of 79 patients with sporadic FGPs that had never been treated with PPI previously; we found that 75.9% of these patients presented a G-cell hyperplasia morphologically similar to patients with AH.5 Furthermore, similar findings were shown by Worthley et al18 in the gastric antrum of patients with GAPPS.6 Apart from this finding, our patients with AH had a negative family history for gastric or colonic cancer, and none of the FGPs showed any dysplastic change.

erplasia morphologically similar to patients with AH.5 Furthermore, similar findings were shown by Worthley et al18 in the gastric antrum of patients with GAPPS.6 Apart from this finding, our patients with AH had a negative family history for gastric or colonic cancer, and none of the FGPs showed any dysplastic change. Another important question to be addressed is to verify if among sporadic FGPs may exist a subset of patients with abnormal calcium excretion, stone formation or abnormal response to calcium or peptone load. At present, we have no data about this interesting question. In conclusion, we have seen that patients with AH, with sporadic FGPs and with GAPPS share in common an antral G-cell hyperplasia. This fact may represent a promising field of future research. Contributors: DP conceived the study design and wrote the draft. BJ gave a substantial contribution in draft preparation. TE and BS contributed substantially to the study design. OB, BA, AI and DM collected cases and also assisted in a critical reading of the manuscript. Competing interests: None. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Summary box What is already known about this subject? ▸  Optimal bowel preparation among hospitalised patients remains a challenge. ▸  Suboptimal bowel preparation results in higher chance of missed pathology and increased costs to the healthcare system. ▸  Outpatient split-dose bowel preparation has been associated with better quality bowel preparation. What are the new findings? ▸  This feasibility trial demonstrated one strategy for implementing split dose bowel prep in an inpatient setting. ▸  Inpatients who completed the split dose prep underwent more timely colonoscopy (less procedural delays and less additional laxative use). ▸  The split dose regimen was better tolerated than a non split bowel preparation. ▸  Larger studies evaluating the efficacy, tolerance, and uptake of split dose bowel preparation in the inpatient setting are needed. How might this impact on clinical practice? ▸  We introduce strategies to implement split-dose preparation for colonoscopy among inpatients. ▸  We highlight the importance of ancillary staff education and training in order to optimise uptake of new strategies and thus effectively disseminate efforts of implementation.

▸  Larger studies evaluating the efficacy, tolerance, and uptake of split dose bowel preparation in the inpatient setting are needed. How might this impact on clinical practice? ▸  We introduce strategies to implement split-dose preparation for colonoscopy among inpatients. ▸  We highlight the importance of ancillary staff education and training in order to optimise uptake of new strategies and thus effectively disseminate efforts of implementation. Introduction Optimal bowel preparation for adequate visualisation at the time of colonoscopy among hospitalised patients remains a significant challenge. Inpatients tend to be older, less ambulatory and have more comorbidities than outpatients, which may adversely affect compliance with the bowel regimen and contribute to inadequate bowel cleansing as reflected by several studies that have identified inpatient status as an independent predictor for poor bowel cleansing.1–3 The risk of failed bowel preparation for hospitalised patients is twofold higher than that of ambulatory outpatients.4 The implications of a poor prep include a higher chance of missed pathology, aborted or delayed procedures and the need for a repeat procedure; all of which contribute to patient inconvenience as well as increased costs to the healthcare system.5–8

Colon cleansing using a split-dose administration of 4 L polyethylene glyocol (PEG) is associated with higher compliance rates, better quality bowel preparations and higher colonoscopy completion rates.9–11 Despite evidence of improved outcomes with the use of a split-dose PEG solution, the majority of patients who are hospitalised receive the standard full dose (NON-SPLIT) 4 L PEG solution prior to inpatient colonoscopy.1 The reasons for this discrepancy remain unclear. Recognising this gap between evidence and clinical practice, we performed this multiphase study to evaluate the feasibility of implementing a split-dose bowel prep regimen for inpatient colonoscopy at our institution. Methods Study design This was a single-centre, prospective observational study conducted at the Malcom Randall VA Medical Center. Overview In this multiphase study, the first phase involved surveying nursing staff and providers to identify perceived barriers to administration of a split-dose bowel preparation and their concerns about patient tolerability. In the next phase, with the help of a multidisciplinary team, an intervention to increase adoption of the split-dose bowel regimen was developed, disseminated and implemented. In the third phase of the study, 100 colonoscopies, performed subsequent to implementation of the intervention, were reviewed to determine use of the intervention, as well as patient compliance with the bowel preparation, procedural delays and colonoscopy completion rates. The phases of the study are detailed in figure 1. Figure 1 Flow diagram depicting the phases of the study.

Overview In this multiphase study, the first phase involved surveying nursing staff and providers to identify perceived barriers to administration of a split-dose bowel preparation and their concerns about patient tolerability. In the next phase, with the help of a multidisciplinary team, an intervention to increase adoption of the split-dose bowel regimen was developed, disseminated and implemented. In the third phase of the study, 100 colonoscopies, performed subsequent to implementation of the intervention, were reviewed to determine use of the intervention, as well as patient compliance with the bowel preparation, procedural delays and colonoscopy completion rates. The phases of the study are detailed in figure 1. Figure 1 Flow diagram depicting the phases of the study. Phase A Study participants Recruitment A cross-sectional list of healthcare providers was obtained from the staff listing. All internal medicine (including hospitalists), critical care and general surgery staff were contacted to participate in a one-time electronic survey. A similar survey was also sent electronically to all medical ward nurses for completion. Survey completion was optional and anonymous.

s obtained from the staff listing. All internal medicine (including hospitalists), critical care and general surgery staff were contacted to participate in a one-time electronic survey. A similar survey was also sent electronically to all medical ward nurses for completion. Survey completion was optional and anonymous. Data collection and analysis The 13-item survey included structured questions related to bowel preparation, feasibility of split-dose administration and perceptions of patient compliance and tolerability. Responses were on a 1–5 bipolar scale (with a neutral point in the middle and two ends of the scale at opposite positions). The electronic survey could be completed in 5–6 min. Additional free text was allowed for clarification of responses. Descriptive statistics, including frequencies and percentages, were reported.

Responses were on a 1–5 bipolar scale (with a neutral point in the middle and two ends of the scale at opposite positions). The electronic survey could be completed in 5–6 min. Additional free text was allowed for clarification of responses. Descriptive statistics, including frequencies and percentages, were reported. Phase B Split-dose bowel preparation order set development To facilitate uptake and ease of ordering, an electronic ‘quick order set’ was developed. In a quick order set, series of orders can be grouped together with each order dialogue box having predefined, default values. Launching the order set allows for all of its components to be selected and prepared for electronic signature. This is the fastest way to enter orders and saves the ordering provider from having to enter each order manually.12 The objective was to develop an order set that contained explicit guidance on timing and volume of laxative administration for the nursing staff. Feedback from the nursing surveys was used to modify the order set prior to incorporation into the electronic health record system and for dissemination. Post implementation, all orders for inpatient colonoscopy were placed using this split-dose quick order set.

volume of laxative administration for the nursing staff. Feedback from the nursing surveys was used to modify the order set prior to incorporation into the electronic health record system and for dissemination. Post implementation, all orders for inpatient colonoscopy were placed using this split-dose quick order set. Phase C Study participants To evaluate the feasibility of implementation of the split-dose bowel preparation, consecutive inpatients scheduled to undergo colonoscopy between October 2012 and April 2013 were approached for study participation until 100 patients were enrolled. Patient interviews were conducted to obtain information regarding bowel preparation consumption, nursing instruction and adverse effects, and colonoscopy outcomes were obtained from chart review.

go colonoscopy between October 2012 and April 2013 were approached for study participation until 100 patients were enrolled. Patient interviews were conducted to obtain information regarding bowel preparation consumption, nursing instruction and adverse effects, and colonoscopy outcomes were obtained from chart review. Data collection After completing informed consent, the study team collected basic demographic information including body mass index, comorbid medical conditions, age, gender and prior surgical history. On the morning of colonoscopy, patients were interviewed by one of the investigators (DY, RS, BR, JBW) about completion of the bowel preparation, nursing instructions, quantity and timing of preparation consumption using a standardised patient survey. Patients were asked to report any adverse events or inconveniences (including nausea, vomiting, abdominal bloating/pain, lack of instructions and/or sleep disturbances), identify any obstacles for completing the laxative regimen (including taste, volume, instructions and time of the bowel preparation administration) and, finally, their willingness to repeat the same bowel preparation for future colonoscopies. All colonoscopy procedures were performed by a staff gastroenterologist, or a gastroenterology fellow under the supervision of an attending gastroenterologist. The quality of the bowel preparation was assessed using a standard scale.13 On the morning of the colonoscopy, patients were interviewed to determine the quantity and timing of bowel laxative consumption and side effects using a standardised patient survey. Additionally, patients were asked to recall what specific instructions they received regarding bowel cleansing.

using a standard scale.13 On the morning of the colonoscopy, patients were interviewed to determine the quantity and timing of bowel laxative consumption and side effects using a standardised patient survey. Additionally, patients were asked to recall what specific instructions they received regarding bowel cleansing. Study outcomes ‘Completion of the bowel preparation’ was defined as successful completion of at least three quarters (75%) of the PEG solution. This was evaluated by asking the patient, the nurse and/or the physician if the entire PEG solution (4 L) was consumed and, when applicable, by examining the amount of PEG remaining in the container. ‘Colonoscopy completion’ was defined as a completed colonoscopy conducted on the scheduled day. If clear liquid stool output void of sediment was not achieved by the time of the scheduled colonoscopy, then the colonoscopy was delayed while additional bowel purgatives were administered. In that situation the procedure was considered ‘delayed secondary to suboptimal bowel preparation’. Actual completion of the bowel regimen as a split-dose regimen was measured based on patient and nurse interview on the morning of the procedure by one of the investigators (DY, RS, BR, JBW).

urgatives were administered. In that situation the procedure was considered ‘delayed secondary to suboptimal bowel preparation’. Actual completion of the bowel regimen as a split-dose regimen was measured based on patient and nurse interview on the morning of the procedure by one of the investigators (DY, RS, BR, JBW). Statistical analyses All data were entered into an electronic database. The statistical analysis was performed using Statistical Analysis System software (SAS Institute Inc, Cary, North Carolina, USA). A paired t test was used for continuous variables and to compare means between the groups. χ2/Fisher's exact test was used for categorical variables and proportions in two by two contingency tables. A p value less than 0.05 was considered significant. In some instances, survey item responses were collapsed to ensure adequate numbers in each cell for comparison. When statistical testing was not performed between groups, percentages and frequencies were presented.

es and proportions in two by two contingency tables. A p value less than 0.05 was considered significant. In some instances, survey item responses were collapsed to ensure adequate numbers in each cell for comparison. When statistical testing was not performed between groups, percentages and frequencies were presented. Results Phase A A total of 65 physicians and 41 nurses completed the electronic survey. Based on survey responses, the most common bowel preparation for an inpatient colonoscopy was Go-Lytely (86.4%). The majority of nurses (80.5%) had never heard of a split-dose bowel regimen. Survey results indicated considerable nursing and provider concern over the feasibility of implementing an inpatient split-dose regimen. Concerns regarding patients’ ability to complete the bowel preparation and follow instructions were also reported (with providers expressing slightly more concern than patients). Providers also expressed concern about complexity of nursing instructions (see tables 1 and 2). Table 1 Perceived implementation barriers

Results Phase A A total of 65 physicians and 41 nurses completed the electronic survey. Based on survey responses, the most common bowel preparation for an inpatient colonoscopy was Go-Lytely (86.4%). The majority of nurses (80.5%) had never heard of a split-dose bowel regimen. Survey results indicated considerable nursing and provider concern over the feasibility of implementing an inpatient split-dose regimen. Concerns regarding patients’ ability to complete the bowel preparation and follow instructions were also reported (with providers expressing slightly more concern than patients). Providers also expressed concern about complexity of nursing instructions (see tables 1 and 2). Table 1 Perceived implementation barriers Very/somewhat concerned Neutral Not very/not at all concerned Patient will not want to awaken to finish second half of preparation Physician/PA 53.6% (30) 25.0% (14) 21.4% (12) Nursing staff 64.9% (24) 24.3% (9) 10.8% (4) Patients will have more difficulty following instructions Physician/PA 42.9% (24) 25.0% (14) 32.1% (18) Nursing staff 37.8% (14) 32.4% (12) 29.7% (11) Instructions too complicated for nurses Physician/PA 48.2% (27) 25.0% (14) 26.8% (15) Nursing staff 13.5% (5) 21.6% (8) 64.9% (24) Patients will not finish second dose in time Physician/PA 69.6% (39) 16.1% (9) 14.3% (8) Nursing staff 86.5% (32) 8.1% (3) 5.4% (2) Insufficient nursing support for second dose administration Physician/PA 62.5% (35) 25.0% (14) 12.5% (7) Nursing staff 51.4% (19) 21.6% (8) 27.0% (10) PA, physician assistant. Table 2 Perceived patient tolerability and outcome barriers

Very/somewhat concerned Neutral Not very/not at all concerned Patient will not want to awaken to finish second half of preparation Physician/PA 53.6% (30) 25.0% (14) 21.4% (12) Nursing staff 64.9% (24) 24.3% (9) 10.8% (4) Patients will have more difficulty following instructions Physician/PA 42.9% (24) 25.0% (14) 32.1% (18) Nursing staff 37.8% (14) 32.4% (12) 29.7% (11) Instructions too complicated for nurses Physician/PA 48.2% (27) 25.0% (14) 26.8% (15) Nursing staff 13.5% (5) 21.6% (8) 64.9% (24) Patients will not finish second dose in time Physician/PA 69.6% (39) 16.1% (9) 14.3% (8) Nursing staff 86.5% (32) 8.1% (3) 5.4% (2) Insufficient nursing support for second dose administration Physician/PA 62.5% (35) 25.0% (14) 12.5% (7) Nursing staff 51.4% (19) 21.6% (8) 27.0% (10) PA, physician assistant. Table 2 Perceived patient tolerability and outcome barriers Very/somewhat concerned Neutral Not very/not at all concerned More nausea Physician/PA 12.5% (7) 41.1% (23) 46.4% (26) Nursing staff 32.4% (12) 24.3% (9) 43.2% (16) More vomiting Physician/PA 10.7% (6) 41.1% (23) 48.2% (27) Nursing staff 32.4% (12) 24.3% (9) 43.2% (16) More abdominal pain Physician/PA 7.1% (4) 44.6% (25) 48.2% (27) Nursing staff 40.5% (15) 16.2% (6) 43.2% (16) Lower quality bowel prep Physician/PA 25.0% (14) 44.6% (25) 48.2% (27) Nursing staff 51.4% (19) 16.2% (6) 43.2% (16) PA, physician assistant.

41.1% (23) 48.2% (27) Nursing staff 32.4% (12) 24.3% (9) 43.2% (16) More abdominal pain Physician/PA 7.1% (4) 44.6% (25) 48.2% (27) Nursing staff 40.5% (15) 16.2% (6) 43.2% (16) Lower quality bowel prep Physician/PA 25.0% (14) 44.6% (25) 48.2% (27) Nursing staff 51.4% (19) 16.2% (6) 43.2% (16) PA, physician assistant. Phase B Taking into account information from the surveys, the study team developed a quick order set for split-dose bowel preparation. The timing of the first and second dose of the preparation was based on nursing concerns over adequate staff for bathroom or commode assistance and minimal overlap of preparation administration with nursing shift changes or medication administration times. The final order set included the following instructions: nursing order to administer first dose of the 4 L PEG solution at 1700 and the second dose at 2400 and nursing to instruct patients to drink 240 mL (8 ounce) increments every 15–20 min for a total of 2 L at each dose (based on nurse feedback). Charge nurses on the wards, nurse managers and nurse educators were responsible for educating staff about the rationale for and proper usage of the new quick order set (prior to implementation). With IT and pharmacy support, the quick order set was incorporated into the computerised medical record system (CPRS) in October 2012 and became the default order set that was primarily used for all inpatient colonoscopies.

f about the rationale for and proper usage of the new quick order set (prior to implementation). With IT and pharmacy support, the quick order set was incorporated into the computerised medical record system (CPRS) in October 2012 and became the default order set that was primarily used for all inpatient colonoscopies. Phase C One hundred and two consecutive inpatients undergoing colonoscopy were approached; only two patients declined to participate (citing disinterest in participation in research studies). Baseline characteristics are summarised in table 3. Despite use of the split-dose order set for all 100 inpatients, a little over a half (54%) of the inpatients consumed the bowel preparation as a split-dose (we refer to these patients as the SPLIT group), the remainder had the 4 L PEG laxative as the conventional non-split bowel regimen (these were referred to as the NON-SPLIT group). Since approximately half of the patients received the preparation as the conventional full dose administration, we chose to use the NON-SPLIT group of patients as an internal control group allowing for comparisons between groups. There were no significant differences between the two groups (SPLIT and NON-SPLIT). The majority of procedures were considered diagnostic colonoscopies. Overt gastrointestinal bleeding was the most frequent indication (43%), followed by gastrointestinal symptoms (17%), such as new onset abdominal pain and/or changes in bowel habits. Table 3 Baseline characteristics

Phase C One hundred and two consecutive inpatients undergoing colonoscopy were approached; only two patients declined to participate (citing disinterest in participation in research studies). Baseline characteristics are summarised in table 3. Despite use of the split-dose order set for all 100 inpatients, a little over a half (54%) of the inpatients consumed the bowel preparation as a split-dose (we refer to these patients as the SPLIT group), the remainder had the 4 L PEG laxative as the conventional non-split bowel regimen (these were referred to as the NON-SPLIT group). Since approximately half of the patients received the preparation as the conventional full dose administration, we chose to use the NON-SPLIT group of patients as an internal control group allowing for comparisons between groups. There were no significant differences between the two groups (SPLIT and NON-SPLIT). The majority of procedures were considered diagnostic colonoscopies. Overt gastrointestinal bleeding was the most frequent indication (43%), followed by gastrointestinal symptoms (17%), such as new onset abdominal pain and/or changes in bowel habits. Table 3 Baseline characteristics NON-SPLIT* (n=46) SPLIT† (n=54) Total (n=100) Comorbidities, n(%) Coronary artery disease 17 (37) 24 (44.4) 41 (41) Pulmonary disease 29 (63) 25 (46.3) 54 (54) Hypertension 29 (63) 40 (74.1) 69 (69) Neurological disease 14 (30.4) 21 (38.9) 35 (35) Diabetes mellitus 13 (28.3) 20 (37) 33 (33) Liver disease 3 (6.5) 1 (1.9) 4 (4) Renal disease 5 (10.9) 9 (16.7) 14 (14) History of abdominal surgeries 12 (26.1) 14 (25.9) 26 (26) Chronic narcotic use 13 (28.3) 11 (20.4) 24 (24) Indication Overt GI bleeding 21 (45.7) 22 (40.7) 43 (43) Occult GI bleeding 4 (8.7) 5 (9.3) 9 (9) Anaemia 5 (10.9) 7 (13.0) 12 (12) Weight loss 6 (13.0) 6 (11.1) 12 (12) History of colorectal polyps 1 (2.1) 6 (11.1) 7 (7) GI symptoms‡ 9 (19.6) 8 (14.8) 17 (17) *Age 63.7±13.2, BMI 28.0±7.3.

e 13 (28.3) 11 (20.4) 24 (24) Indication Overt GI bleeding 21 (45.7) 22 (40.7) 43 (43) Occult GI bleeding 4 (8.7) 5 (9.3) 9 (9) Anaemia 5 (10.9) 7 (13.0) 12 (12) Weight loss 6 (13.0) 6 (11.1) 12 (12) History of colorectal polyps 1 (2.1) 6 (11.1) 7 (7) GI symptoms‡ 9 (19.6) 8 (14.8) 17 (17) *Age 63.7±13.2, BMI 28.0±7.3. †Age 63.2±14.2, BMI 29.5±7.2. ‡GI symptoms: abdominal pain, nausea, vomiting and or change in bowel habit patterns. BMI, body mass index; GI, gastrointestinal. Completion of bowel preparation rates are depicted in figure 2. Approximately 97% of patients in the SPLIT group completed the bowel preparation (ie, consumed at least 75% of the PEG solution). Again using the NON-SPLIT group as a control group, we found that a lower percentage of patients in the SPLIT group (13%) required additional laxatives to aid in the cleansing process than the NON-SPLIT group (30.4%). Importantly, inpatient colonoscopy was more commonly delayed due to inadequate bowel preparation in the NON-SPLIT group (32.6%) compared to only 7% in the SPLIT group (p=0.002). Figure 3 depicts the need for additional laxatives and procedural delays in the two groups. Figure 2 Rates of completion of bowel preparation was significantly higher in the SPLIT group (96.3%). Figure 3 Suboptimal bowel preparation and colonoscopy delay. Lower rates of additional laxative use and less procedural delay among the SPLIT group.

Completion of bowel preparation rates are depicted in figure 2. Approximately 97% of patients in the SPLIT group completed the bowel preparation (ie, consumed at least 75% of the PEG solution). Again using the NON-SPLIT group as a control group, we found that a lower percentage of patients in the SPLIT group (13%) required additional laxatives to aid in the cleansing process than the NON-SPLIT group (30.4%). Importantly, inpatient colonoscopy was more commonly delayed due to inadequate bowel preparation in the NON-SPLIT group (32.6%) compared to only 7% in the SPLIT group (p=0.002). Figure 3 depicts the need for additional laxatives and procedural delays in the two groups. Figure 2 Rates of completion of bowel preparation was significantly higher in the SPLIT group (96.3%). Figure 3 Suboptimal bowel preparation and colonoscopy delay. Lower rates of additional laxative use and less procedural delay among the SPLIT group. Since survey data indicated nursing concerns about patient tolerability (nausea, vomiting, abdominal pain), our postbowel preparation interviews included assessment of side effects. Overall, the most common patient symptom was nausea (26%) in each group. Abdominal bloating and pain were reported more frequently in the NON-SPLIT group (30% vs 9%). Only one patient in the SPLIT group reported inability to wake up and/or to stay awake as a factor impeding the completion of the split-dose bowel preparation. Among patients in the SPLIT group, 90.7% favoured split-dose administration as their preference for future bowel preparation prior to colonoscopy, suggesting it was well tolerated. To evaluate the process of implementation, patients were explicitly asked about the nursing instructions they received; interestingly, a large percentage in the NON-SPLIT group (38/46, 83%) indicated that they had not received instructions to split the dose.

ation prior to colonoscopy, suggesting it was well tolerated. To evaluate the process of implementation, patients were explicitly asked about the nursing instructions they received; interestingly, a large percentage in the NON-SPLIT group (38/46, 83%) indicated that they had not received instructions to split the dose. Discussion Despite evidence demonstrating the superiority of split dosing for bowel cleansing, there has been reluctance in uniformly adopting this strategy for inpatient colonoscopy. Hospitalised patients represent a challenging population in which to achieve adequate bowel preparation, due to immobility, narcotic use and worse overall health. In this study, we demonstrate reasonable success with implementing split-dose administration of the 4 L bowel prep for hospitalised patients undergoing colonoscopy, which was the primary focus of our feasibility study.

hich to achieve adequate bowel preparation, due to immobility, narcotic use and worse overall health. In this study, we demonstrate reasonable success with implementing split-dose administration of the 4 L bowel prep for hospitalised patients undergoing colonoscopy, which was the primary focus of our feasibility study. Feasibility studies on implementation are focused on the extent, likelihood and manner in which an intervention can be fully implemented, often in an uncontrolled design,14 taking into account the contexts, settings and cultures that might translate the intervention into practice. Keeping this in mind, we incorporated several steps in our multiphase study including: (1) evaluating perceived barriers to implementation of a split dosing preparation; (2) adapting the intervention to maximise nursing uptake; (3) engaging stakeholders upfront to maximise buy-in; (4) working with a multidisciplinary team of nurse managers, nurse educators, pharmacy and IT support; and (5) measuring outcomes related to the intervention to demonstrate its success.

it dosing preparation; (2) adapting the intervention to maximise nursing uptake; (3) engaging stakeholders upfront to maximise buy-in; (4) working with a multidisciplinary team of nurse managers, nurse educators, pharmacy and IT support; and (5) measuring outcomes related to the intervention to demonstrate its success. One of the key steps in implementing a split-dose bowel preparation for inpatients related to engaging and working with nursing staff, as they played a significant role in ensuring that the intervention would be successful. To maximise uptake, we adapted the intervention for the nurses in our clinical setting. We adjusted the administration of the split-dose regimen to a time more suitable for the nursing staff (first dose at 17:00 and second dose at midnight) in response to concerns of interference with nursing shift changes, charting and medication administration, insufficient staff and perceived concerns about patients’ ability to complete the second dose prior to colonoscopy. Additionally, we relied on nurse educators to help disseminate information about the importance and value of split-dose preparations. Furthermore, we also used patient interviews to evaluate inpatients’ ability to complete the split-dose preparation and assessed patient tolerability as this was a concern expressed by nurses and providers and a potential barrier to its implementation.

ormation about the importance and value of split-dose preparations. Furthermore, we also used patient interviews to evaluate inpatients’ ability to complete the split-dose preparation and assessed patient tolerability as this was a concern expressed by nurses and providers and a potential barrier to its implementation. Despite these measures, 46% of inpatients did not receive the split-dose bowel preparation (as instructed in the quick order set). One of the contributing factors may have been the lack of information or education provided by the nurses: 94% in the SPLIT group reported receiving instruction about split dosing whereas only 17% in the NON-SPLIT group reported receiving instruction. This information, however, was based on patient reports and subject to recall bias as well as reporting bias. To overcome this limitation, we could have provided more repeated efforts at nursing education emphasising the value of the split-dose preparation. Additionally, providing nurse feedback with preliminary results from our colonoscopy interviews may also have helped increase uptake as our results showed decreased procedural delays and higher rates of preparation completion. Another explanation could be that nurses did not adopt the split-dose strategy because they questioned its clinical utility. This is consistent with diffusion of innovation theories, which describe a predictable process of adoption based on individuals’ level of readiness to accept new ideas.15

preparation completion. Another explanation could be that nurses did not adopt the split-dose strategy because they questioned its clinical utility. This is consistent with diffusion of innovation theories, which describe a predictable process of adoption based on individuals’ level of readiness to accept new ideas.15 Though our rate of uptake was modest, we demonstrated that among inpatients who did have the split bowel prep, there were fewer procedural delays, decreased use of additional laxatives to ensure complete clean-out as well as high-bowel preparation completion rates. The majority of inpatient colonoscopies in our cohort were performed for evaluation (and/or therapy) of GI bleeding or GI symptoms and timely colonoscopy was important for clinical decision-making and management in these patients. Adverse effects were similar across groups. Among the SPLIT group, patient adherence with the nocturnal dose was not an obstacle for bowel cleansing (only 1/56 patients could not wake up or remain awake to complete split-dose regimen). Moreover, the vast majority of inpatients expressed willingness to repeat the split-dose regimen for future colonoscopies. While our results appear to be in line with other reports indicating patient willingness to awaken during the night to complete a split-dose bowel preparation when educated about the benefits of split-dosing,16–21 these studies primarily involved outpatients. Thus, future studies in the inpatient setting are needed to corroborate our preliminary findings, to establish the optimal time interval between the end of purgative administration and colonoscopy to achieve an adequate bowel preparation, and to further determine how changes in split-dosing scheduling may affect patient and provider receptivity.

inpatient setting are needed to corroborate our preliminary findings, to establish the optimal time interval between the end of purgative administration and colonoscopy to achieve an adequate bowel preparation, and to further determine how changes in split-dosing scheduling may affect patient and provider receptivity. There are several limitations of our study. This was an uncontrolled, single centre study with a small number of patients, which limited our ability to make comparisons between the SPLIT-group and NON-SPLIT group. Also, despite explicit nursing orders regarding timing of preparation administration, there was variability in the actual administration of the two doses as well as the variability in the exact timing of colonoscopy, which could have impacted our findings. A limitation with respect to our dissemination efforts included reliance on nurse educators and charge nurses to provide education to the nurses; this was neither standardised nor monitored, thus leading to variability in how the importance of split-dose was communicated and perhaps affecting uptake. However, this direct communication by nursing may also have been an important factor influencing uptake and early adoption among nursing staff that did demonstrate uptake. Furthermore, while a patient survey following bowel preparation was obtained to assess tolerability and acceptability of split-dosing, a postintervention survey was not obtained from the nursing staff and/or providers. This may have been a valuable tool to help identify any potential barriers for split-dose bowel preparation implementation. Another limitation is that this study utilised an intervention (quick order set) that was specific for the VA hospital's electronic health record, and may not be available nor generalisable in other settings. However, we hope that the strategy and multistep process that we employed (not the specific intervention) may be of value to other individuals or institutions that are planning to implement a split-dose regimen for their inpatient colonoscopies.

alth record, and may not be available nor generalisable in other settings. However, we hope that the strategy and multistep process that we employed (not the specific intervention) may be of value to other individuals or institutions that are planning to implement a split-dose regimen for their inpatient colonoscopies. At our institution, this was the first attempt to implement a process for split-dose bowel preparation among hospitalised patients. While there have been studies evaluating the implementation of split dosing for outpatient colonoscopy, we found only one other study (presented in abstract form) that reported on outcomes associated with split-dose bowel preparation for inpatient colonoscopy, however, no data on implementation was provided.22 In conclusion, there is a lack of data on the optimal bowel preparation for inpatients undergoing colonoscopy. Overall, this multiphase study demonstrated one strategy for implementing split-dose bowel preparation for inpatient colonoscopy with reasonable success and suggests that split-dose bowel preparation may be better tolerated among inpatients. Based on our results, we emphasise the importance of ancillary staff education and training in order to optimise uptake of new strategies and thus effectively disseminate efforts of implementation. Larger studies evaluating the efficacy and tolerance of split dose bowel preparation as well as strategies for implementing a split dose bowel prep among inpatients are needed.

ncillary staff education and training in order to optimise uptake of new strategies and thus effectively disseminate efforts of implementation. Larger studies evaluating the efficacy and tolerance of split dose bowel preparation as well as strategies for implementing a split dose bowel prep among inpatients are needed. Contributors: DY and JBW performed the literature search. DY, RS, BR, SS and DC were involved in the design of the study and survey questionnaire. DY, RS, BR, JL and LM were involved in data and statistical analyses. DY, RS and BR wrote the first draft of the manuscript. SS provided critical appraisal of the manuscript. All authors were involved with the critical revision of the final manuscript. Funding: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. Dr. Sultan is supported by a VA Health Services Research and Development (HSR&D) Career Development Award. The findings and conclusions in this document are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs. Competing interests: None. Ethics approval: Institution's Review Board at the University of Florida and the Research and Development Committee at the Malcom Randall VA Medical Center in Gainesville, Florida. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Summary box What is already known about this subject? ▸  Diversity of gut microbiota is a vital part of intestinal defence, and once this is disturbed, the intestine is vulnerable to the detrimental effects of gut bacteria. ▸  Epidemiological data suggest that sepsis now represents a rising worldwide healthcare burden. Initial studies focused attention on bacterial translocation across the intestinal mucosa into the systemic milieu as a unifying mechanism to explain sepsis. ▸  More than half of patients who developed sepsis did not have an identifiable bacterial infection, and current studies could not fully explain this phenomenon. What are the new findings? ▸  Components of cultivable anaerobes from the Bacteroidetes phylotype are the primarily inducers of proinflammatory responses and these responses are primarily produced by myeloid dendritic cells. ▸  The common prophylactic therapy for sepsis which consists of neomycin and metronidazole given orally to patients the day before surgery is ineffective in clearing Bacteroidetes from the murine intestine. ▸  In response to acute intestinal injury, this prophylactic therapy induced increased systemic inflammation and premature death when compared with treatment-free mice, and this outcome is independent of the dissemination of cultivable anaerobes.

▸  The common prophylactic therapy for sepsis which consists of neomycin and metronidazole given orally to patients the day before surgery is ineffective in clearing Bacteroidetes from the murine intestine. ▸  In response to acute intestinal injury, this prophylactic therapy induced increased systemic inflammation and premature death when compared with treatment-free mice, and this outcome is independent of the dissemination of cultivable anaerobes. ▸  Intraperitoneal spread of gut-derived anaerobe of antibiotic-treated mice showed declined survival percentage compared with that with untreated mice, indicating that spillage of gut microbial products, rather than dissemination of gut microbes themselves, may underline the initiation of systemic inflammation leading to death in the early phase of intestinal barrier injury. How might it impact on clinical practice in the foreseeable future? ▸  How to bridge and extrapolate findings from mouse models to the setting of human sepsis is always a daunting challenge. Nevertheless, based on our data it would appear inadvisable to prescribe the perioperative therapy of neomycin and metronidazole to leaky gut-prone individuals. ▸  Novel treatments that will specifically target anaerobes from the Bacteroidetes could provide a successful infection-control approach for the treatment of systemic inflammation leading to sepsis and death. ▸  Therapeutic tools that block the activation of myeloid dendritic cells could be beneficial to treat sepsis.

How might it impact on clinical practice in the foreseeable future? ▸  How to bridge and extrapolate findings from mouse models to the setting of human sepsis is always a daunting challenge. Nevertheless, based on our data it would appear inadvisable to prescribe the perioperative therapy of neomycin and metronidazole to leaky gut-prone individuals. ▸  Novel treatments that will specifically target anaerobes from the Bacteroidetes could provide a successful infection-control approach for the treatment of systemic inflammation leading to sepsis and death. ▸  Therapeutic tools that block the activation of myeloid dendritic cells could be beneficial to treat sepsis. Introduction and rationale The gut microbiota is engaged in a dynamic interaction with the immune system, affecting different aspects of its development and function.1–4 The mechanisms through which the microbiota exerts its immune influences remain largely undefined, but include the recognition of secreted bacterial products by immune cells and elaboration of signalling molecules.3 5 The overall balance in the composition of the gut microbial communities is important in ensuring homeostasis or lack thereof in the gut and systemic sites.1 3 6 Perhaps, the most serious manifestation of a failure in the mammalian intestinal microbial symbiosis is the clinical state of systemic inflammation leading to sepsis. Translocation of gut bacteria, which are normally retained within the lumen of the intestinal tract across the intestinal barrier, is thought to be integral to this clinical process.7 8 However, more than half of patients who developed sepsis do not show evidence of bacterial contamination in the periphery,8–10 and there are no published studies that clearly explain this phenomenon.

the lumen of the intestinal tract across the intestinal barrier, is thought to be integral to this clinical process.7 8 However, more than half of patients who developed sepsis do not show evidence of bacterial contamination in the periphery,8–10 and there are no published studies that clearly explain this phenomenon. Most of the effects of intestinally derived bacteria on systemic immune cells are mediated by a controlled passage of bacterial products that act as immunomodulators.11 12 The advantage of these molecules over the bacteria themselves is that they can cross the intestinal epithelium and reach peripheral sites relatively easily. Thus, one possibility during the initial phase of sepsis is that uncontrolled spillage of secreted bacterial products, rather than gut bacteria themselves, may occur causing the initiation of systemic inflammatory responses. Therefore, we sought to understand the consequences of the interaction between secreted gut microbial components and systemic immune cells. Specifically, we asked whether systemic innate cells can mount an inflammatory response to secreted gut microbial components, which immune cells were the primary drivers of inflammation, and which bacterial populations were predominantly responsible. We show that proinflammatory responses in the periphery are generated by gut microbial components of the Bacteroidetes phylotype and these responses are primarily produced by CD11b+ CD11c+ dendritic cells (DCs).

ls were the primary drivers of inflammation, and which bacterial populations were predominantly responsible. We show that proinflammatory responses in the periphery are generated by gut microbial components of the Bacteroidetes phylotype and these responses are primarily produced by CD11b+ CD11c+ dendritic cells (DCs). It has been argued that aggressive perioperative infection-control strategies such as the use of preoperative antibiotics directly upregulate signalling and cellular-microbial cross-talk pathways within the gut with deleterious consequences for the host, such as microbial drug resistance and persistent rates of postoperative sepsis.7 This theory is reinforced by recent clinical studies which have demonstrated that surgical antibiotic use increases the risk of Clostridium difficile infection.13 Based on these observations, we tested the efficacy of preventive treatment for bowel surgery in clearing gut Bacteroidetes. A common oral antibiotic regimen is the Nichols-Condon bowel preparation and consists of neomycin and erythromycin administered the day prior to surgery.14 Metronidazole has been recently substituted for the erythromycin due to its seemingly increased efficacy against anaerobic microorganisms in the gut.15 16 Using doses similar to those prescribed to individual patients, we find that this prophylactic therapy, although is efficient for clearing most gut anaerobes, is ineffective for clearing those of the Bacteroidetes phylotype from the murine intestine. Moreover, we find that prophylactic treatment followed by intestinal barrier damage induced more rapid death kinetic compared with antibiotic-free mice. Together, our data further affirm the importance of a balanced gut microflora biodiversity in host immune homeostasis and suggest that prescribing oral antibiotics to patients, especially those with a leaky gut syndrome, have to be carefully weighed if outcomes from postoperative systemic inflammation leading to sepsis are to be improved.

urther affirm the importance of a balanced gut microflora biodiversity in host immune homeostasis and suggest that prescribing oral antibiotics to patients, especially those with a leaky gut syndrome, have to be carefully weighed if outcomes from postoperative systemic inflammation leading to sepsis are to be improved. Materials and methods Mice C57BL/6 and BALB/c mice were purchased from Taconic (Germantown, NY). All mice were maintained in a specific pathogen-free barrier unit at Rutgers University and were used between 6 and 12 weeks of age. Ethics and dissemination All experiments followed guidelines of Rutgers University Institutional Animal Care and Use Committee (IACUC). Approval for use of mice was obtained from Rutgers University according to criteria outlined in the Guide for the Care and Use of Laboratory Animals from the National Institutes of Health. Bacterial cultures The experimental setup is described in detail in the online supplementary methods. Assessment of anaerobes in the gut and extraintestinal tissues The experimental setup is described in detail in the online supplementary methods. Isolation of splenocytes and stimulation with bacterial-conditioned supernatants The experimental setup is described in detail in the online supplementary methods. Flow cytometry The experimental setup is described in detail in the online supplementary methods. DNA isolation from bacteria The experimental setup is described in detail in the online supplementary methods. Quantitative PCR The experimental setup is described in detail in the online supplementary methods.

Isolation of splenocytes and stimulation with bacterial-conditioned supernatants The experimental setup is described in detail in the online supplementary methods. Flow cytometry The experimental setup is described in detail in the online supplementary methods. DNA isolation from bacteria The experimental setup is described in detail in the online supplementary methods. Quantitative PCR The experimental setup is described in detail in the online supplementary methods. Treatment of mice with antibiotics The experimental setup is described in detail in the online supplementary methods. Systemic injection of bacterial-conditioned supernatants The experimental setup is described in detail in the online supplementary methods. Mouse model of intestinal injury induced by dextran sodium sulfate The experimental setup is described in detail in the online supplementary methods. Blood analysis The experimental setup is described in detail in the online supplementary methods. Survival assays We used time to death as well as moribund for all survival assays. We identified moribund mice as those mice that are unable to move when gently touched or unable to self-correct when placed on their side (ataxia). Moribund mice were euthanised in compliance with IACUC protocols at the Rutgers University. Statistics GraphPad Prism software (La Jolla, California, USA) was used to conduct unpaired, two-tailed Student t test for sample analysis. Results with p<0.05 were considered significant.

Survival assays We used time to death as well as moribund for all survival assays. We identified moribund mice as those mice that are unable to move when gently touched or unable to self-correct when placed on their side (ataxia). Moribund mice were euthanised in compliance with IACUC protocols at the Rutgers University. Statistics GraphPad Prism software (La Jolla, California, USA) was used to conduct unpaired, two-tailed Student t test for sample analysis. Results with p<0.05 were considered significant. Results Conditioned supernatants derived from gut anaerobes induce proinflammatory responses from specific antigen-presenting cell subsets The importance of innate cells as the first-line of defence against microorganisms and their microbial products is undisputable. However, while the impact of gut microbial products on gut innate cells is increasingly understood, the consequences of interaction between these components and distant peripheral innate cells remain unclear.

innate cells as the first-line of defence against microorganisms and their microbial products is undisputable. However, while the impact of gut microbial products on gut innate cells is increasingly understood, the consequences of interaction between these components and distant peripheral innate cells remain unclear. The use of 16S ribosomal RNA (rRNA) gene sequences has been by far the most common methodology to study gut bacterial phylogeny and taxonomy.17 However, whereas this method can be used for a number of bacterial identification purposes, the procedures that enable the acquisition of secreted bacterial components are mainly confined to the use of bacterial culture systems. Therefore, to address our question, we first cultured luminal contents of the small intestine and colon from C57BL/6 mice under aerobic and anaerobic conditions as described in Methods section,18–22 and conditioned supernatants were used to simulate splenocytes isolated from C57BL/6 mice. Four hours later, antigen-presenting cells (APC) were examined and production of proinflammatory cytokines was assessed in response to microbial products derived from cultivable aerobic and anaerobic supernatants (figure 1). We found that supernatants of anaerobes cultured from the small intestine and colon induced substantially tumour necrosis factor (TNF) α and pro-interleukin (IL) 1β than aerobic supernatants (figure 1A, B, and see online supplementary figure S1). Interestingly, the strong TNF-α response was found to be specific to CD11b+ CD11c+ DCs as neither CD11b− CD11c+ DCs nor CD11b+ CD11c− F4/80+ macrophages expressed more TNF-α in response to anaerobic than aerobic supernatants (figure 1A). Likewise, CD11b+ CD11c+ DCs expressed significantly more pro-IL-1β when exposed to conditioned medium from cultivable anaerobes than aerobes of the colon (see online supplementary figure S1). This suggests that proinflammatory responses generated by cultivable bacterial components are primarily induced by gut anaerobes and selectively produced by CD11b+ CD11c+ myeloid DC subset. Of note, the natural ecosystem of the gut microbiota is very complex and culturing luminal faecal contents does not enable the growth of all gut bacteria in culture systems since most species are not cultivable.17 This is not a pitfall of the experimental strategy but rather a limitation of current methodologies for gut bacterial culture procedures.

of the gut microbiota is very complex and culturing luminal faecal contents does not enable the growth of all gut bacteria in culture systems since most species are not cultivable.17 This is not a pitfall of the experimental strategy but rather a limitation of current methodologies for gut bacterial culture procedures. Figure 1 Proinflammatory responses are induced by components of gut anaerobes from the Bacteroidetes and produced by myeloid dendritic cells (DCs). (A) Bar graphs showing tumour necrosis factor (TNF) α expression as a % of CD11b+ CD11c+ DCs, CD11b− CD11c+ DCs, and CD11b+ CD11c− F4/80+ (n=4). (B) Dot plots showing expression of TNF-α by CD11b+ CD11c+ DCs and CD11b− CD11c+ DCs. Numbers indicate the % TNF-α cells. (C) Bar graphs showing that Firmicutes dominate the aerobic cultures and a large portion of aerobic Firmicutes were Enterococcus (n=4). (D) Bar graph showing that anaerobic cultures of small intestine and colon bacteria have relatively high proportions of Bacteroidetes (n=4–6). (E–G) Mice were gavaged with PBS (untreated) or a combination of neomycin and metronidazole (Neo/Met) used as prophylaxis for bowel surgery for 24 h, or injected intravenously with a combination of piperacillin and tazobactam (Pip/Taz) used as a chemotherapy for sepsis, every 24 h for three days. (E) Bar graph showing the relative number of cultivable anaerobes in antibiotic-treated mice to antibiotic-free animals (n=5–8). (F) Bar graph showing ratio of Bacteroidetes to Firmicutes in untreated and antibiotic-treated mice (n=4). (G) Bar graph showing the proportion of total gut bacteria represented by Bacteroidetes (n=4). *p<0.05, **p<0.005, ***p<0.0005 using two-tailed Student t test. SSC, side scatter.

o antibiotic-free animals (n=5–8). (F) Bar graph showing ratio of Bacteroidetes to Firmicutes in untreated and antibiotic-treated mice (n=4). (G) Bar graph showing the proportion of total gut bacteria represented by Bacteroidetes (n=4). *p<0.05, **p<0.005, ***p<0.0005 using two-tailed Student t test. SSC, side scatter. Aerobic and anaerobic bacterial cultures from the small intestine and colon are dominated by distinct phyla The gut of mammals is dominated by two major phylotypes of bacteria: Firmicutes and Bacteroidetes.23 24 Maintenance of a prescribed equilibrium between these two major groups is critical, as shifts in the balance between these two groups has been associated with obesity, irritable bowel syndrome, Crohn's disease, ulcerative colitis and colon cancer.24–29 Since products derived from cultivable anaerobes but not aerobes were potent inducer of proinflammatory cytokines by peripheral APCs, we next characterised their composition of Firmicutes and Bacteroidetes. To this end, DNA was isolated from both bacterial sources used to generate our conditioned bacterial supernatants and the relative proportions of Firmicutes versus Bacteroidetes was determined by quantitative PCR (QPCR) using specific primers for 16S rRNA gene.30 We found that Firmicutes dominate the aerobic cultures, with significantly more Firmicutes found in the cultures of aerobes compared with their anaerobic counterparts (figure 1C). To further extend this finding, we next examined the relative proportion of a major genus of Firmicutes in the gut, namely Enterococcus.31 We found that a large portion of aerobic Firmicutes cultured from the small intestine and colon were Enterococcus and that there was significantly more Enterococcus in the aerobic cultures compared with anaerobic counterparts (figure 1C). However, unlike aerobic cultures, anaerobic cultures from small intestine and colon revealed significantly higher proportions of Bacteroidetes spp. (figure 1D). Consistent with different effects in driving proinflammatory responses from peripheral APCs, these results suggest that aerobic and anaerobic cultures from small intestine and colon are dominated by distinct phyla, with aerobic cultures dominated by Firmicutes and anaerobic cultures mostly represented by Bacteroidetes.

Consistent with different effects in driving proinflammatory responses from peripheral APCs, these results suggest that aerobic and anaerobic cultures from small intestine and colon are dominated by distinct phyla, with aerobic cultures dominated by Firmicutes and anaerobic cultures mostly represented by Bacteroidetes. Changes of Bacteroidetes:Firmicutes ratios may dictate efficacy of antibiotic treatment for sepsis Our results suggest that secreted microbial components of cultivable anaerobes derived from the Bacteroidetes are potent inducers of proinflammatory cytokines by peripheral APCs. We next sought to understand whether Bacteroidetes could play a role in the induction of systemic inflammation leading to sepsis and death. To address this question, we first treated mice orally with a combination of neomycin and metronidazole (Neo/Met) used as antibiotic prophylaxis for bowel surgery32–34 and examined by QPCR the efficacy of this prophylactic therapy in clearing Bacteroidetes from the murine intestine (figure 1E, F). As expected, we found that numbers of cultivable anaerobes decreased significantly (∼70%) in antibiotic-treated mice compared with untreated littermates (figure 1E). However, while the total number of cultivable anaerobes decreased in response to antibiotic treatment, the ratio of Bacteroidetes to Firmicutes in the anaerobes that survived this treatment revealed a substantially increased (∼30-fold) proportion of Bacteroidetes (figure 1F). Moreover, when examining the proportion of total gut bacteria represented by Bacteroidetes, we found no decrease as a result of Neo/Met treatment (figure 1G), suggesting that therapy with Neo/Met is efficient at clearing most cultivable anaerobes but not Bacteroidetes spp.

fold) proportion of Bacteroidetes (figure 1F). Moreover, when examining the proportion of total gut bacteria represented by Bacteroidetes, we found no decrease as a result of Neo/Met treatment (figure 1G), suggesting that therapy with Neo/Met is efficient at clearing most cultivable anaerobes but not Bacteroidetes spp. To further extend this observation, we next asked whether a treatment for intra-abdominal infections following bowel surgery could be more effective at clearing gut Bacteroidetes. To this end, mice were treated with piperacillin/tazobactam (Pip/Taz) intravenously for 3 days and efficacy of this treatment in clearing Bacteroidetes from the intestine was examined by QPCR as previously described.30 35 Unlike treatment with Neo/Met, treatment with Pip/Taz significantly shifted the ratio of Bacteroidetes:Firmicutes in favour of Firmicutes (figure 1F) and drastically reduced both the number and percentage of Bacteroidetes in the gut to <3% of the total bacterial species (figure 1E, G). Thus, while the preventive treatment with Neo/Met was ineffective in clearing Bacteroidetes from the murine intestine, a more effective treatment with Pip/Taz was highly efficient at Bacteroidetes clearance. Together, these data indicate that efficacy of antibiotic treatment as a control strategy for systemic inflammation leading to sepsis may correlate with clearance of Bacteroidetes from the gut.

teroidetes from the murine intestine, a more effective treatment with Pip/Taz was highly efficient at Bacteroidetes clearance. Together, these data indicate that efficacy of antibiotic treatment as a control strategy for systemic inflammation leading to sepsis may correlate with clearance of Bacteroidetes from the gut. Prophylactic treatment increases systemic inflammation and causes more rapid death in response to intestinal barrier damage Since cultivable anaerobes formed by the Bacteroidetes are potent inducers of proinflammatory responses by peripheral APCs and the prophylactic Neo/Met treatment is ineffective at clearing these anaerobes (figure 1), we hypothesised that the preventive antibiotic treatment (the Nichols-Condon bowel preparation) may be detrimental rather than beneficial at reducing risks of systemic inflammation leading to sepsis and death in conditions of intestinal barrier damage. Caecal ligation and puncture has been proposed as a surgical murine model to study the pathogenesis of sepsis.36 However, because this model utilises the surgical perforation of the ligated caecum which results in immediate and consistent drainage of caecal bacteria into the peritoneal cavity,36 its applicability was unsuitable for this study since our intent is to understand the role of secreted bacterial products rather the bacteria themselves during the initial phase of intestinal barrier injury. Therefore, to test our hypothesis, we used a minimally invasive sepsis mouse model that couples antibiotic treatments with dextran sulfate sodium (DSS) induced intestinal barrier damage.37 DSS is toxic to colonic epithelial cells and typically causes a colitis-like disease in BALB-c mice characterised by weight loss and intestinal barrier injury.38 Briefly, BALB-c mice were treated with drinking water only (control) or drinking water supplemented with DSS continuously for 4–12 days (figure 2A, upper panel). By the end of the fourth day, significant physical and immunological changes were observed between both mouse groups (figure 2B first panel, C). DSS-treated BALB-c mice lost significant body weight compared with untreated controls (figure 2B first panel). In addition, percentages of DC-expressing proinflammatory TNF-α and pro-IL-1β were higher in the spleen of DSS-treated mice compared with untreated animals (figure 2C), suggesting the initiation of systemic inflammatory responses at day 4 post-DSS treatment.

ght compared with untreated controls (figure 2B first panel). In addition, percentages of DC-expressing proinflammatory TNF-α and pro-IL-1β were higher in the spleen of DSS-treated mice compared with untreated animals (figure 2C), suggesting the initiation of systemic inflammatory responses at day 4 post-DSS treatment. Next, we asked whether Nichols-Condon bowel preparation can further accelerate or rather attenuate this immune process. To this end, BALB-c mice were treated orally with Neo/Met; 24 h later, animals were provided with sterile drinking water (control) or drinking water supplemented with DSS continuously for 4 days (figure 2A, middle panel). In parallel experiments, mice were treated with Pip/Taz intravenously every 24 h for 3 days. At the end of the first day, mice were given autoclaved drinking water (control) or drinking water supplemented with DSS continuously for 4 days (figure 2A, lower panel). The goal of these experiments was to compare the outcomes of ineffective clearance of Bacteroidetes (Neo/Met) versus effective clearance of Bacteroidetes (Pip/Taz) in response to acute intestinal barrier damage. Consistent with previous results, the use of the prophylactic therapy induced significant weight loss in response to DSS compared with other animal groups (figure 2B). Plasma concentration of TNF-α is frequently elevated in septic patients and mouse models of sepsis and correlates with severity and outcome.39 40 In this context, our data revealed that plasma concentration of TNF-α did not decrease as a result of preventive therapy (figure 2D). Furthermore, mice treated with the prophylactic therapy died more rapidly in response to DSS treatment compared with wild type littermates that received DSS only or DSS in combination with the Pip/Taz therapy (figure 3A). Therefore, Neo/Met treatment used as prophylaxis for sepsis appears to be rather inefficient at reducing systemic inflammation and may even accelerate the initiation of this process resulting in increased mortality in intestinally-injured mice.

ceived DSS only or DSS in combination with the Pip/Taz therapy (figure 3A). Therefore, Neo/Met treatment used as prophylaxis for sepsis appears to be rather inefficient at reducing systemic inflammation and may even accelerate the initiation of this process resulting in increased mortality in intestinally-injured mice. Figure 2 Neo/Met therapy is inefficient for reducing systemic inflammation and causes faster death in intestinally-injured mice. (A) Mouse model of intestinal injury was developed using 6–8 week old BALB-c mice provided with drinking water only (control) or supplemented with DSS continuously for the duration of the experiments. Mice were orally injected with the Nichols-Condon bowel preparation which consists of a combination of (Neo/Met, 85 and 145 mg/kg, respectively). Twenty-four hours later, animals were transferred to new cages and provided with autoclaved drinking water (control) or drinking water supplemented with DSS (w/v) continuously for 4–12 days (Middle panel). Other mice were injected intravenously with Pip/Taz (1500 and 187.5 mg/kg, respectively) every 24 h for three days. At the end of the third day, mice were transferred to new cages and were given autoclaved drinking water (control) or drinking water supplemented with DSS continuously for 4–12 days (lower panel). (B) Weight loss of mice treated with PBS or antibiotics only (n=4), or DSS only (n=10), or DSS in combination with antibiotic therapies (n=10) for 4 days. (C) Bar graphs showing TNF-α and PROIL1-β production by splenic DCs in mice untreated (n=4) and DSS-treated animals (n=10). (D) Serum concentrations of TNF-α at day 4 post-DSS treatment. Bar graphs showing relative production of TNF-α in the blood (n=4–10). NS (p>0.01), *p<0.05, **p<0.005, using two-tailed Student t test (DCs, dendritic cells; DSS, dextran sulfate sodium; i.v., intravenously; Neo/Met, neomycin and metronidazole; NS, not significant; Pip/Taz, piperacillin and tazobactam; PROIL1-β, prointerleukin 1β; TNF-α, tumour necrosis factor α).

in the blood (n=4–10). NS (p>0.01), *p<0.05, **p<0.005, using two-tailed Student t test (DCs, dendritic cells; DSS, dextran sulfate sodium; i.v., intravenously; Neo/Met, neomycin and metronidazole; NS, not significant; Pip/Taz, piperacillin and tazobactam; PROIL1-β, prointerleukin 1β; TNF-α, tumour necrosis factor α). Figure 3 Spread of cultivable anaerobes occurs only at a later time point of intestinal injury and further accelerates death in Neo/Met-treated mice. (A) Survival of mice that received water or antibiotic only (n=4) or water supplemented with DSS (n=10), or DSS in combination with antibiotic therapies (n=20) for 4 days. (B) Amounts of cultivable anaerobes at day 4 post-DSS treatment in the spleen and liver of antibiotic-free and Neo/Met-treated mice (3 independent experiments of n=3). (C) Survival of mice that were injected i.p. with anaerobic broth (vehicle, n=5) or supernatants of cultivable anaerobes from the colon of untreated (n=10) or Neo/Met-treated (n=10) mice. (D) Amounts of cultivable anaerobes at day 12 post-DSS treatment in the spleen and liver of antibiotic-free and antibiotic-treated mice (3 independent experiments of n=3). (E) Survival of mice that received water or antibiotics only (n=4) or water supplemented with DSS (n=20), or DSS in combination with Neo/Met (n=20) for 12 days. NS (p>0.01), *p<0.05, using two-tailed Student t test (DSS, dextran sulfate sodium; i.p, intraperitoneal; Neo/Met, neomycin and metronidazole; NS, not significant; Pip/Taz, piperacillin/tazobactam). CFU, colony-forming unit; PBS, phosphate buffered saline.

(n=20), or DSS in combination with Neo/Met (n=20) for 12 days. NS (p>0.01), *p<0.05, using two-tailed Student t test (DSS, dextran sulfate sodium; i.p, intraperitoneal; Neo/Met, neomycin and metronidazole; NS, not significant; Pip/Taz, piperacillin/tazobactam). CFU, colony-forming unit; PBS, phosphate buffered saline. Intraperitoneal spread of gut anaerobe products prepared from colonic contents of Neo/Met-treated mice induces rapid death in host mice Translocation of bacteria, particularly anaerobes, across the intestinal mucosa into the systemic milieu was proposed as a unifying mechanism to explain early postoperative sepsis.7 10 41 To determine whether faster moribundus or death induced by the preventive therapy is caused by systemic dissemination of gut-derived microbes, we assessed the number of anaerobes in extraintestinal organs at day 4 post-DSS treatment. Our data showed no significant levels of cultivable anaerobes in the spleen and liver of Neo/Met-treated mice compared with untreated controls (figure 3B). These findings, however, do not exclude the possibility of the development of non-cultivable species in these tissues or the development of bacteraemia in the portal circulation that could have been cleared in the liver prior to the initiation of systemic inflammation. Nevertheless, these outcomes suggest that negative cultures in the spleen and liver are associated with a lack of significant level of cultivable anaerobes in these tissues in the early phase of intestinal injury. Since cultivable anaerobe products were potent inducers of proinflammatory cytokines by peripheral APCs (figure 1); we hypothesised that spread of gut anaerobe products may be an aetiological agent inducing the negative outcomes resulting from the prophylactic therapy. To test this hypothesis, we performed experiments to mimic the spread of gut-derived cultivable anaerobe products in extraintestinal organs (figure 3C). Briefly, healthy wild-type mice were injected intraperitoneally with bacteria-free supernatants prepared from cultivable colonic contents of untreated or Neo/Met-treated animals and outcomes from such treatment were assessed in vivo.

ead of gut-derived cultivable anaerobe products in extraintestinal organs (figure 3C). Briefly, healthy wild-type mice were injected intraperitoneally with bacteria-free supernatants prepared from cultivable colonic contents of untreated or Neo/Met-treated animals and outcomes from such treatment were assessed in vivo. Unlike injection of bacterial material from antibiotic-free mice, conditioned supernatants from Neo/Met-pretreated mice induced high mortality after only 48 h of treatment (figure 3C), indicating that early death induced by the prophylactic Neo/Met may be caused by the spread of gut anaerobe products, rather than the microbes themselves, in the initial phase of intestinal barrier damage.

itioned supernatants from Neo/Met-pretreated mice induced high mortality after only 48 h of treatment (figure 3C), indicating that early death induced by the prophylactic Neo/Met may be caused by the spread of gut anaerobe products, rather than the microbes themselves, in the initial phase of intestinal barrier damage. Prophylactic treatment for sepsis induces increased dissemination of gut anaerobes in response to chronic, but not acute, intestinal barrier injury During early intestinal barrier damage, we observed no detectable levels of cultivable anaerobes in the liver and spleen of untreated and antibiotic-treated mice (figure 3B). However, we could not eliminate the possibility that translocation of gut anaerobes across the intestinal mucosa into the systemic milieu may occur at later time points of intestinal injury. To test this possibility, BALB-c mice were treated with DSS to induce intestinal injury, coupled with Neo/Met therapy, for 12 days. Unlike in the early phase of intestinal injury, analysis at a later time point revealed the detection of gut cultivable anaerobes in the spleen and liver of treated mice (figure 3D), and this phenotype correlated with increased death rate in antibiotic-treated animals (figure 3E).

led with Neo/Met therapy, for 12 days. Unlike in the early phase of intestinal injury, analysis at a later time point revealed the detection of gut cultivable anaerobes in the spleen and liver of treated mice (figure 3D), and this phenotype correlated with increased death rate in antibiotic-treated animals (figure 3E). All together, the prophylactic therapy with Neo/Met followed by acute intestinal barrier damage appears to induce more rapid death in mice that may be caused by the spread of gut anaerobe products, rather than the anaerobes themselves. However, the dissemination of gut anaerobes occurs at later time points of intestinal barrier damage and this may further accelerate death. Discussion The intestinal tract houses the largest population of bacteria in the body and the majority of these microorganisms are anaerobes; of these ∼25% are species of Bacteroidetes.41 42 These bacteria maintain a complex and generally beneficial relationship with the host when retained in the gut.1 3 However, when Bacteroidetes species escape the gut, usually resulting from disruption of the intestinal barrier, they can cause significant pathology, including inflammation and abscess formation.41 43 44 In this context, our findings suggest the role of anaerobes from the Bacteroidetes phylotype are important inducers of systemic inflammation and underscore the importance of an overlooked aspect of these gut bacteria: their products in the induction of systemic inflammation leading to death.

abscess formation.41 43 44 In this context, our findings suggest the role of anaerobes from the Bacteroidetes phylotype are important inducers of systemic inflammation and underscore the importance of an overlooked aspect of these gut bacteria: their products in the induction of systemic inflammation leading to death. Sepsis with its complications is still a major challenge in contemporary medicine. Depending on the standards of medical care, the worldwide mortality rates in septic humans range from 30% to 70% (with an aggregate mortality rate of ∼50%).45 46 Despite more than 20 years of extensive research and development of numerous therapeutic approaches used in clinical settings, the incidence of sepsis and the number of sepsis-related deaths are rising at rates between 1.5% and 8% per year.45 46 For a long time, it was believed that sepsis was caused by an overwhelming immune response of the patient to invading microorganisms.14 47 As a precautionary measure, antibiotic regimens are given to patients before and/or during surgery.14 47 A common oral antibiotic regimen is the Nichols-Condon bowel preparation, which consists of neomycin and erythromycin given the day before surgery.14 32 Metronidazole was recently substituted for the erythromycin because of its seemingly increased efficacy against anaerobic microorganisms in the gut.48 A major outcome from our study is that secreted microbial components of cultivable Bacteroidetes are potent inducers of proinflammatory cytokines by peripheral DCs. Based on this finding; we tested the efficacy of the preventive therapy for bowel surgery in clearing Bacteroidetes and compared its efficacy to Pip/Taz treatment for intra-abdominal infection following bowel sepsis.49 50 While treatment for sepsis was highly efficient at Bacteroidetes clearance, we found that oral therapy with Neo/Met was ineffective for clearing Bacteroidetes from the murine intestine. One point of crucial consequences from this result was increased systemic inflammation and more rapid time death in mice treated with the prophylactic Neo/Met in condition coupled with intestinal barrier damage induced by DSS.

al therapy with Neo/Met was ineffective for clearing Bacteroidetes from the murine intestine. One point of crucial consequences from this result was increased systemic inflammation and more rapid time death in mice treated with the prophylactic Neo/Met in condition coupled with intestinal barrier damage induced by DSS. Still, the scope of this study is that the interplay between intestinally derived microbial products and the host may determine the outcome after intestinal barrier damage. Spillage of gut microbial products outside the intestinal mucosa may generate several possible scenarios (see online supplementary figure S3). Gut microbial products that do not derive from Bacteroidetes, such as aerobe Firmicutes, may be less harmful, in which case weak inflammatory responses are developed. As an example, bacterial products of Enterococcus faecalis, a representative of gut Firmicutes,31 had little to no effect on driving proinflammatory responses in vitro (see online supplementary figure S2). However, when inflammation is established by Bacteroidetes products, potent inflammatory responses may follow. These responses can be regulated by a network of proinflammatory cytokines including TNF-α and pro-IL-1β derived primarily from myeloid DCs. This initial response may determine the systemic expansion of the inflammatory response. Moreover, because Bacteroidetes products contain potent proteolytic enzymes of the gut epithelium such as the Bacteroides fragilis enterotoxin,41 they may at later stages, work in concert with other processes to increase the enzymatic destruction of mucosal barriers and facilitate the dissemination of gut microbes within the host leading to an immune collapse and faster death.

teolytic enzymes of the gut epithelium such as the Bacteroides fragilis enterotoxin,41 they may at later stages, work in concert with other processes to increase the enzymatic destruction of mucosal barriers and facilitate the dissemination of gut microbes within the host leading to an immune collapse and faster death. Cross-talks between the host and secreted bacterial products from intestinal microbiota are poorly understood, and we do not know yet to what extent the encounters between these secreted microbial products and the host can be modulated. In many settings the most cost-effective way to improve post intestinal barrier damage outcomes is to strengthen the quality of antibiotic treatments. However, in many parts of the world the spread of multidrug-resistant bacteria seriously threatens the success of antibiotic therapies.51 However because eliminating the need for antibiotics is unrealistic at this time, a better understanding of how antibiotic therapies impact the gut microbiota and host immunity is needed, so that strategies to mitigate the negative effects can be developed. How to bridge and extrapolate findings from mouse sepsis models to the setting of human sepsis is always a daunting challenge. Nevertheless, based on these findings, it would appear inadvisable to prescribe the perioperative therapy of Neo/Met to leaky gut-prone individuals. The authors thank Drs Yasmina Laouar and Yufang Shi for critical reading of the manuscript, and Mariel Watkins for technical assistance with faecal DNA isolation and QPCR.

Cross-talks between the host and secreted bacterial products from intestinal microbiota are poorly understood, and we do not know yet to what extent the encounters between these secreted microbial products and the host can be modulated. In many settings the most cost-effective way to improve post intestinal barrier damage outcomes is to strengthen the quality of antibiotic treatments. However, in many parts of the world the spread of multidrug-resistant bacteria seriously threatens the success of antibiotic therapies.51 However because eliminating the need for antibiotics is unrealistic at this time, a better understanding of how antibiotic therapies impact the gut microbiota and host immunity is needed, so that strategies to mitigate the negative effects can be developed. How to bridge and extrapolate findings from mouse sepsis models to the setting of human sepsis is always a daunting challenge. Nevertheless, based on these findings, it would appear inadvisable to prescribe the perioperative therapy of Neo/Met to leaky gut-prone individuals. The authors thank Drs Yasmina Laouar and Yufang Shi for critical reading of the manuscript, and Mariel Watkins for technical assistance with faecal DNA isolation and QPCR. Contributors: DS, AE and MT performed the experiments and analysed the data. DS and AL designed the experiments and analysed the data. AL wrote the manuscript. Funding: This work is supported by the Robert Wood Johnson Foundation (grant # 67038) to the Child Health Institute of New Jersey. Competing interests: None. Provenance and peer review: Not commissioned; externally peer reviewed.

Contributors: DS, AE and MT performed the experiments and analysed the data. DS and AL designed the experiments and analysed the data. AL wrote the manuscript. Funding: This work is supported by the Robert Wood Johnson Foundation (grant # 67038) to the Child Health Institute of New Jersey. Competing interests: None. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Summary box What is already known about this subject? ▸  Host interleukin-28B (IL-28B) polymorphisms were known to be associated with spontaneous hepatitis C virus (HCV) clearance and also response to treatment. HCV is the other factor contributing to clearance. When both host and viral factors are mixed involving in HCV clearance and disease progression, it is difficult to tell the important factors. What are the new findings? ▸  This is a the study on a unique group of patients with HCV1b-infection (n=105) accidentally transmitted from a single blood donor infected with genotype 1b in Guizhou province, southwest China. With the sole resource of the virus, the clear-known infected time, the similar ethnicity and environments, it is better to understand the host factors for HCV spontaneous clearance and disease progress. How might it impact on clinical practice in the foreseeable future? ▸  Add the knowledge of how the host factors may affect HCV clearance. Introduction Hepatitis C virus (HCV) infection affects hundreds of millions of people worldwide. It has been reported that about 20% of HCV-infected adults can spontaneously clear the virus, while 30% of patients with chronic infection progress to cirrhosis and hepatocellular carcinoma (HCC).1

How might it impact on clinical practice in the foreseeable future? ▸  Add the knowledge of how the host factors may affect HCV clearance. Introduction Hepatitis C virus (HCV) infection affects hundreds of millions of people worldwide. It has been reported that about 20% of HCV-infected adults can spontaneously clear the virus, while 30% of patients with chronic infection progress to cirrhosis and hepatocellular carcinoma (HCC).1 Viral and host factors are involved in HCV spontaneous clearance and disease progression. Virus factors include HCV genotypes, quasispecies, viral load and co-infection. Host factors include gender, age at infection or the ageing process, race, alanine aminotransferase (ALT) elevation and genetic factors.2 Recently, interleukin-28B (IL-28B) polymorphisms have been reported to be associated with spontaneous HCV clearance and also response to treatment.3–7

s, viral load and co-infection. Host factors include gender, age at infection or the ageing process, race, alanine aminotransferase (ALT) elevation and genetic factors.2 Recently, interleukin-28B (IL-28B) polymorphisms have been reported to be associated with spontaneous HCV clearance and also response to treatment.3–7 The purpose of the current study was to analyse a group of patients infected with the same HCV genotype 1b (HCV1b) source in order to focus on host parameters that may be involved in resolution or persistence of HCV infection. These patients in the current study are unique for several reasons. First, the sole resource of the HCV1b virus excludes virus genotypic differences. HCV1b is a difficult-to-treat genotype with interferon-based therapy. Second, the known date of infection provides data on the natural history of HCV infection over the course of 9–12 years. Third, the common ethnicity and similar environments of the patients reduce some variables into the analysis of factors involved in HCV spontaneous clearance and disease progression. Lastly, the broad age range of patients is helpful to study the importance of host age.

y of HCV infection over the course of 9–12 years. Third, the common ethnicity and similar environments of the patients reduce some variables into the analysis of factors involved in HCV spontaneous clearance and disease progression. Lastly, the broad age range of patients is helpful to study the importance of host age. Materials and methods Study subjects All patients had received blood transfusions, from 1998 to 2002, from a single blood donor who was subsequently found to have had HCV1b. All recipients were Chinese from Pingtang, Guizhou province, southwest China. Inclusion criteria were transfusion of blood or blood-products from the identified contaminated batches of the same donor. Patients who died from causes other than HCV-related liver disease, and patients we were unable to contact, were excluded. All patients with positive HCV RNA were tested and found to be genotype 1b. Patients were identified and blood samples were collected from 2010 to 2011, 9–12 years postinfection (median 10 years). The study was approved by the ethics committee from Guiyang Medical College and conformed to the ethical guidelines of the Declaration of Helsinki; informed consent had been obtained from each individual included in the study.

Patients were identified and blood samples were collected from 2010 to 2011, 9–12 years postinfection (median 10 years). The study was approved by the ethics committee from Guiyang Medical College and conformed to the ethical guidelines of the Declaration of Helsinki; informed consent had been obtained from each individual included in the study. Serum HCV antibody and RNA assays Serum biochemical parameters including ALT levels were measured by routine automated methods according to the manufacturer's instructions. Anti-HCV antibody levels were measured by electrochemiluminescence immunoassay (ECLIA) using Abbott Architect i2000 (ABBOTT, Wiesbaden, Germany) according to the manufacturer's instructions. HCV RNA was detected by the commercial quantitative reverse transcription PCR (RT-PCR; COBAS AMPLICOR, Roche Diagnostic Systems, Indianapolis, USA) according to the manufacturer's instructions. The lower limit of detection was 15 IU/mL. The HCV genotype was determined by Versant HCV Genotype 2.0 (LiPA; Siemens Healthcare Diagnostics, Tarrytown, New York, USA). Fibroscan detection Transient elastography was performed using FibroScan (Echosens, France). The examination was performed on the right lobe of the liver through the seventh or eighth intercostal space. The measurement depth was between 25 and 65 mm. As suggested by the manufacturer, only results obtained with 10 valid measurements, with a success-rate of at least 60% and with an IQR ≤30%, were considered reliable, as described previously.8–10

he right lobe of the liver through the seventh or eighth intercostal space. The measurement depth was between 25 and 65 mm. As suggested by the manufacturer, only results obtained with 10 valid measurements, with a success-rate of at least 60% and with an IQR ≤30%, were considered reliable, as described previously.8–10 PCR amplification and sequencing of IL-28B polymorphisms Genomic DNAs were isolated from 0.5 mL whole blood using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany). IL-28 rs12979860, rs8099917, rs10853728, rs12980275, rs4803219, rs4803223, rs8105790 and rs28416813 were amplified by PCR. The PCR protocol involved initial denaturation at 95°C for 10 min, 35 cycles of denaturation for 30 s at 95°C, annealing of primers for 30 s at 55°C and extension for 40 s at 72°C, followed by final extension at 72°C for 10 min. The amplified DNA fragments were separated on a 2% agarose gel, and purified with the QIAquick gel extraction kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions Nucleotide sequences were determined by Sanger sequencing using the Applied Biosystems Automated 3730 DNA Analyzer.

at 72°C for 10 min. The amplified DNA fragments were separated on a 2% agarose gel, and purified with the QIAquick gel extraction kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions Nucleotide sequences were determined by Sanger sequencing using the Applied Biosystems Automated 3730 DNA Analyzer. Exome capture and sequencing assay Exome capture Purified genomic DNA samples were randomly fragmented by Covaris, with the size of library fragments mainly distributed between 150 and 200 bp.11–13 Adapters were ligated to both ends and purified by the Agencourt AMPure SPRI beads according to the manufacturer's instructions. Fragments with insert sizes of about 250 bp were excised. Extracted DNA was amplified by ligation-mediated PCR (LM-PCR), purified and hybridised to the SureSelect Biotinylated RNA Library (BAITS) for enrichment. Hybridised fragments were bound to the strepavidin beads, whereas non-hybridised fragments were washed out after 24 h. Captured LM-PCR products were analysed using an Agilent 2100 Bioanalyzer to estimate the magnitude of enrichment. Each captured library was then loaded on a Hiseq2000 platform, to ensure that each sample meets the desired average sequencing depth for high-throughput sequencing. Raw image files were processed by Illumina basecalling Software V.1.7 with default parameters and the sequences of each individual were generated as 90 bp pair-end reads.11

rary was then loaded on a Hiseq2000 platform, to ensure that each sample meets the desired average sequencing depth for high-throughput sequencing. Raw image files were processed by Illumina basecalling Software V.1.7 with default parameters and the sequences of each individual were generated as 90 bp pair-end reads.11 Read mapping SOAPaligner (V.2.21) was used to align the sequencing reads to the NCBI human genome reference assembly (build 36.3) with a maximum of 3 mismatches and the parameters were set as -a -b -D -o -u -p -2 -m -x -s 40 -l 35 -v 3. To evaluate exome capture efficiency, the proportions of reads mapping to target regions and to their flanking regions (within 200 bp) were calculated for each individual. Reads that aligned to the designed target region (TR±200 bp) were collected for single nucleotide polymorphism (SNP) identification and subsequent analysis.11 Individual genotype calling On the basis of SOAP alignment results, the software SOAPsnp was used to call genotypes. The following parameters were set: -r 0.0005 -e 0.001 -t -u -2 -i -d -o -M -L 90 -s -T (http://soap.genomics.org.cn/ for details). To obtain an accurate genotype set, the genotype was filtered on the basis of the following criteria: the SNP should be observed in at least one individual with a quality score >20, and in a way that the number of reads containing mutant alleles was larger than the reads containing reference alleles. At the same time, at least 90% of all individuals got a quality score >20 and coverage >4.12 13 Our next steps are all based on this genotype set.

in at least one individual with a quality score >20, and in a way that the number of reads containing mutant alleles was larger than the reads containing reference alleles. At the same time, at least 90% of all individuals got a quality score >20 and coverage >4.12 13 Our next steps are all based on this genotype set. Principal component analysis To check the population stratification, we did a principal component analysis with EIGENSTRAT software. We chose the SNP from 1000 genomes (http://www.1000genomes.org/) to check our samples. Determination of associations We checked the genome-wide association results using PLINK software (the http://pngu.mgh.harvard.edu/~purcell/plink/) to perform a standard case/control association analysis using: plink --file mydata --assoc. All the top sites SNPs with statistical significance were also checked for Hardy-Weinberg equilibrium, and sites which had p<0.001 were filtered.12 13 Statistical analysis Clinical statistical analyses were performed using the Statistical Program for Social Sciences (V.11.5; SPSS). Continuous variables such as ALT, HCV RNA and time were presented as medians (range), and categorical variables as frequencies. Categorical variables were tested by χ2 test or Fisher's exact test. p Values <0.05 were considered to be statistically significant.

the Statistical Program for Social Sciences (V.11.5; SPSS). Continuous variables such as ALT, HCV RNA and time were presented as medians (range), and categorical variables as frequencies. Categorical variables were tested by χ2 test or Fisher's exact test. p Values <0.05 were considered to be statistically significant. Results Patient demographic data The demographic features of the patients are shown in figure 1 and table 1. A total of 105 receipts were found with the single same donor from medical records of the blood transfusions from 1998 to 2002. Twenty of them with a negative HCV antibody could not be contacted or died from causes other than liver disease. Among the 85 patients with a positive HCV antibody, the male-to-female ratio was 45:40 with a median age of 32 years (9–71 years). Twenty-seven cases (31.8%) were HCV RNA negative, and were considered to have achieved spontaneous clearance of HCV. The other 58 cases (68.2%) were HCV RNA positive with levels ranging from 3.0 log to 7.1 log IU/mL, median 4.8 log IU/mL. Table 1 Host characteristics of spontaneous clearance for HCV

Results Patient demographic data The demographic features of the patients are shown in figure 1 and table 1. A total of 105 receipts were found with the single same donor from medical records of the blood transfusions from 1998 to 2002. Twenty of them with a negative HCV antibody could not be contacted or died from causes other than liver disease. Among the 85 patients with a positive HCV antibody, the male-to-female ratio was 45:40 with a median age of 32 years (9–71 years). Twenty-seven cases (31.8%) were HCV RNA negative, and were considered to have achieved spontaneous clearance of HCV. The other 58 cases (68.2%) were HCV RNA positive with levels ranging from 3.0 log to 7.1 log IU/mL, median 4.8 log IU/mL. Table 1 Host characteristics of spontaneous clearance for HCV Characteristics Total (case, %) Spontaneous clearance Non-clearance OR (95% CI) p Value Case (case, %) 85 27, 31.8 58, 68.2 – – Gender (case, %) 0.462 Male 45, 52.9 15, 55.6 30, 51.7 1.11 (0.59 to 2.08) Female 40, 47.1 12, 44.4 28, 48.3 0.95 (0.71 to 1.27) Age, median (range) (case, %) 32 (9–71), years 30 (9–61), years 34 (10–71), years 0.160 ≤20 years 19, 22.4 10, 37.0 9, 15.5 2.04 (1.13 to 3.69) 0.028* 21–40 years 39, 45.9 11, 40.7 28, 48.3 0.81 (0.43 to 1.54) 0.340 41–60 years 21, 24.7 5, 18.5 16, 27.6 0.69 (0.30 to 1.60) 0.267 ≥61 years 6, 7.0 1, 3.7 5, 8.6 0.51 (0.08 to 3.12) 0.375 IL-28 allele frequency detected (case, %) 64 24 40 rs10853728 CC 36, 56.3 17, 70.8 19, 47.5 1.89 (0.91 to 3.91) 0.058 rs12979860 CC 53, 82.8 22, 91.7 31, 77.5 2.28 (0.63 to 8.32) 0.132 rs8099917 TT 53, 82.8 22, 91.7 31, 77.5 2.28 (0.63 to 8.32) 0.132 rs12980275 AA 53, 82.8 22, 91.7 31, 77.5 2.28 (0.63 to 8.32) 0.132 rs4803219 CC 55, 85.9 22, 91.7 33, 82.5 1.80 (0.51 to 6.37) 0.264 rs4803223 AA 47, 73.4 16, 66.7 31, 77.5 0.72 (0.38 to 1.38) 0.254 rs8105790 TT 54, 84.3 22, 91.7 32, 80.0 2.04 (0.57 to 7.33) 0.189 *p<0.05 was considered to be statistically significant.

rs12980275 AA 53, 82.8 22, 91.7 31, 77.5 2.28 (0.63 to 8.32) 0.132 rs4803219 CC 55, 85.9 22, 91.7 33, 82.5 1.80 (0.51 to 6.37) 0.264 rs4803223 AA 47, 73.4 16, 66.7 31, 77.5 0.72 (0.38 to 1.38) 0.254 rs8105790 TT 54, 84.3 22, 91.7 32, 80.0 2.04 (0.57 to 7.33) 0.189 *p<0.05 was considered to be statistically significant. HCV, hepatitis C virus; IL, interleukin. Figure 1 Data for the special HCV-infected group from one single blood donor. A total of 105 receipts, which were accidentally infected by a single HCV genotype 1b donor, from 1998 to 2002. AB, antibody; HCV, hepatitis C virus; HCC, hepatocellular carcinoma. Characteristics of HCV spontaneous clearance Among the 85 patients with a positive HCV antibody, host factors including the gender, age and IL-28 allele were analysed for an association with spontaneous clearance. It was shown that an age less than 20 years had the most significant association with spontaneous clearance (OR=2.04, 95% CI (1.13 to 3.69), p=0.028), while other ages did not have any significant relation to virus clearance (table 1). Gender had no relation with viral clearance (p=0.462; figure 2). Figure 2 Association of host factors including gender, age and interleukin-28 (IL-28) polymorphisms with hepatitis C virus (HCV) spontaneous clearance or disease progression by OR. (A) Association of age of infection less than 20 years and HCV spontaneous clearance. (B) Association of age of infection less than 40 years and HCV disease progression.

actors including gender, age and interleukin-28 (IL-28) polymorphisms with hepatitis C virus (HCV) spontaneous clearance or disease progression by OR. (A) Association of age of infection less than 20 years and HCV spontaneous clearance. (B) Association of age of infection less than 40 years and HCV disease progression. Eight IL-28B polymorphisms, alleles rs12979860, rs8099917, rs10853728, rs12980275, rs4803219, rs4803223, rs8105790 and rs28416813 were PCR amplified and sequenced. As shown in table 2, IL-28B rs10853728 showed a strong trend, which fell just short of statistical significance (p=0.058). For the most well-known SNP, IL-28B rs12979860, 22 of the 24 (91.7%) patients who had spontaneous clearance had the CC allele, while 31 of the 40 (77.5%) in the non-clearance group had the CC allele (p=0.132). The TT allele was not found in any of the patients (figure 3). Table 2 Characteristics of patients without HCV clearance and persisted infection with HCV RNA detectable

Eight IL-28B polymorphisms, alleles rs12979860, rs8099917, rs10853728, rs12980275, rs4803219, rs4803223, rs8105790 and rs28416813 were PCR amplified and sequenced. As shown in table 2, IL-28B rs10853728 showed a strong trend, which fell just short of statistical significance (p=0.058). For the most well-known SNP, IL-28B rs12979860, 22 of the 24 (91.7%) patients who had spontaneous clearance had the CC allele, while 31 of the 40 (77.5%) in the non-clearance group had the CC allele (p=0.132). The TT allele was not found in any of the patients (figure 3). Table 2 Characteristics of patients without HCV clearance and persisted infection with HCV RNA detectable Characteristics Patients without HCV clearance Case number (%) 58 HCV RNA, median (range) (number, %) 4.8 (3.0–7.1), log IU/mL ≤3.0 log IU/mL 2, 3.4 3.1–5.0 log IU/mL 29, 50.0 5.1–7.0 log IU/mL 25, 43.2 ≥7.1 log IU/mL 2, 3.4 ALT, median (range) (number, %) 30, (12–192), U/L ≤1.0 ULN 49, 84.5 1.1–3.0 ULN 3, 5.2 3.1–5.0 ULN 6, 10.3 Fibroscan, median (range) (number, %) 5.9, (3.7–35.3), kPa Failure/undetected 12, 20.7 ≤5.0 kPa 12/46, 26.1 5.0–7.3 kPa 21/46, 45.6 7.3–9.5 kPa 10/46, 21.7 >9.6 kPa 3/46, 6.6 Ultrasonography (number, %) Non-specific/undetected 41, 70.7 Enhanced/coarse echo of liver 13, 22.4 Splenomegaly 4, 6.9 ALT, alanine aminotransferase; HCV, hepatitis C virus; ULN, upper limit of normal value.

ure/undetected 12, 20.7 ≤5.0 kPa 12/46, 26.1 5.0–7.3 kPa 21/46, 45.6 7.3–9.5 kPa 10/46, 21.7 >9.6 kPa 3/46, 6.6 Ultrasonography (number, %) Non-specific/undetected 41, 70.7 Enhanced/coarse echo of liver 13, 22.4 Splenomegaly 4, 6.9 ALT, alanine aminotransferase; HCV, hepatitis C virus; ULN, upper limit of normal value. Figure 3 Interleukin-28 (IL-28) polymorphisms with hepatitis C virus (HCV) spontaneous clearance. IL-28 rs10853728 CC and HCV clearance (p=0.058). IL-28 single nucleotide polymorphisms (SNPs) and associations with spontaneous clearance of HCV. IL-28 SNPs, rs12979860 CC, rs8099917 TT and rs10853782 prevalence in Chinese patients.

nterleukin-28 (IL-28) polymorphisms with hepatitis C virus (HCV) spontaneous clearance. IL-28 rs10853728 CC and HCV clearance (p=0.058). IL-28 single nucleotide polymorphisms (SNPs) and associations with spontaneous clearance of HCV. IL-28 SNPs, rs12979860 CC, rs8099917 TT and rs10853782 prevalence in Chinese patients. Risk factors for HCV disease progression Nine to 12 years after transfusion, none of the 27 patients with spontaneous clearance had disease progression by liver function tests, abdominal ultrasonography and fibroscan tests. Among the 58 patients with positive HCV RNA, 9 cases (15.5%) had elevated serum ALT ranging from 41 to 192 IU/mL. A total of 13 cases (22.4%) developed chronic hepatitis with mild to moderate fibrosis as determined by clinical manifestations, fibroscan values higher than 7.1 kPa, and enhanced and coarse echogenicity of the liver by ultrasonography. A total of three patients (5.2%) developed cirrhosis with decreased albumin, fibroscan values higher than 9.5 kPa, and splenomegaly as determined by ultrasound.8–10 Neither decompensated cirrhosis nor HCC was found. According to the infected time and duration, it was estimated that the rate of HCV progression to mild or moderate fibrosis was 2.2% per year, and to cirrhosis was 0.5% per year. Multivariate regression analysis showed that gender (p=0.393), HCV RNA level (p=0.262) and IL-28B allele frequencies (p=0.565) were not statistically associated with disease progression. An age less than 40 years (OR=0.13, 95% CI (1.13 to 3.69), p=0.020) had a negative association with disease progression.

Risk factors for HCV disease progression Nine to 12 years after transfusion, none of the 27 patients with spontaneous clearance had disease progression by liver function tests, abdominal ultrasonography and fibroscan tests. Among the 58 patients with positive HCV RNA, 9 cases (15.5%) had elevated serum ALT ranging from 41 to 192 IU/mL. A total of 13 cases (22.4%) developed chronic hepatitis with mild to moderate fibrosis as determined by clinical manifestations, fibroscan values higher than 7.1 kPa, and enhanced and coarse echogenicity of the liver by ultrasonography. A total of three patients (5.2%) developed cirrhosis with decreased albumin, fibroscan values higher than 9.5 kPa, and splenomegaly as determined by ultrasound.8–10 Neither decompensated cirrhosis nor HCC was found. According to the infected time and duration, it was estimated that the rate of HCV progression to mild or moderate fibrosis was 2.2% per year, and to cirrhosis was 0.5% per year. Multivariate regression analysis showed that gender (p=0.393), HCV RNA level (p=0.262) and IL-28B allele frequencies (p=0.565) were not statistically associated with disease progression. An age less than 40 years (OR=0.13, 95% CI (1.13 to 3.69), p=0.020) had a negative association with disease progression. Whole-exome capture and sequencing Using exome capture and sequencing, 64 449 SNPs were identified in the sample population. A total of 17 081 coding genes were sequenced with coverage of each individual exome at an average depth of 33.9-fold. On average, about 95% of the target regions were covered by at least one read. More than 86% of the target regions were covered by more than 4 reads (figure 4). The population analysis showed that samples used for the association analysis had no significant population stratification.

me at an average depth of 33.9-fold. On average, about 95% of the target regions were covered by at least one read. More than 86% of the target regions were covered by more than 4 reads (figure 4). The population analysis showed that samples used for the association analysis had no significant population stratification. Figure 4 Exome capture and sequencing assay showing that single nucleotide polymorphisms (SNPs) are associated with spontaneous clearance of hepatitis C virus (HCV). (A) Depth distribution. (B) QQ plot to assess the discrepancy between the predicted value and the observed value. (C) A total of 64 449 SNPs were called from individuals, of which 400 were found to be associated with viral clearance by individual genotype calling. Two SNPs in tenascin-R (TNR), four in transmembrane protease serine 11A (TMPRSS11A), and one in serine peptidase inhibitor kunitz type 2 (SPINT2) showed the closest association (p<10−5).

449 SNPs were called from individuals, of which 400 were found to be associated with viral clearance by individual genotype calling. Two SNPs in tenascin-R (TNR), four in transmembrane protease serine 11A (TMPRSS11A), and one in serine peptidase inhibitor kunitz type 2 (SPINT2) showed the closest association (p<10−5). The top 20 SNPs which had the closest association are listed in table 3. There were SNPs from 11 exons from IL-28B rs2239818 and rs34842046, two from the tenascin receptor (TNR), one (rs3745948) from the serine peptidase inhibitor kunitz type 2 (SPINT2), one (rs7627615) from the 5-hydroxytryptamine receptor 3 family member E (HTR3E), four (rs1370840, rs11930532, rs28437478 and rs6552134) from the transmembrane protease serine 11A (TMPRSS11A), one (rs1263810) from the sal-like protein 2 (SALL2), three (rs9901726, rs2291604 and rs9900543) from the spermatogenesis associated 22 (SPATA 22), one (rs607332) from the nicotinamide mononucleotide adenyltransferase 2 (NMNAT2), two (novel) from the NCK-associated protein 1 (NCKAP1), one (rs2303225) from the MARVEL domain containing 3 (MARVELD3), three (novel) from the zinc finger protein 491, 440 and 439 (ZNF491, ZNF 440 and ZNF 439), and one (rs2307075) from the carbonate dehydratase II (CA2; p<10−4). Table 3 List of top 20 SNP differences in clearance and non-clearance patients

The top 20 SNPs which had the closest association are listed in table 3. There were SNPs from 11 exons from IL-28B rs2239818 and rs34842046, two from the tenascin receptor (TNR), one (rs3745948) from the serine peptidase inhibitor kunitz type 2 (SPINT2), one (rs7627615) from the 5-hydroxytryptamine receptor 3 family member E (HTR3E), four (rs1370840, rs11930532, rs28437478 and rs6552134) from the transmembrane protease serine 11A (TMPRSS11A), one (rs1263810) from the sal-like protein 2 (SALL2), three (rs9901726, rs2291604 and rs9900543) from the spermatogenesis associated 22 (SPATA 22), one (rs607332) from the nicotinamide mononucleotide adenyltransferase 2 (NMNAT2), two (novel) from the NCK-associated protein 1 (NCKAP1), one (rs2303225) from the MARVEL domain containing 3 (MARVELD3), three (novel) from the zinc finger protein 491, 440 and 439 (ZNF491, ZNF 440 and ZNF 439), and one (rs2307075) from the carbonate dehydratase II (CA2; p<10−4). Table 3 List of top 20 SNP differences in clearance and non-clearance patients Gene name SNP ID SNP in chromosome Minor allele Major allele Frequency in clearance Frequency in controls p Value OR Function 1 TNR rs2239818 chr1_173642258 A G 0.01 0.22 1.28E−05 0.04 Receptor binding; negative regulation of cell adhesion 2 SPINT2 rs3745948 chr19_43471696 T C 0.05 0.33 1.46E−05 0.12 Serine-type endopeptidase inhibitor activity 3 TNR rs34842046 chr1_173559295 T A 0.02 0.22 6.67E−05 0.08 Receptor binding; negative regulation of cell adhesion 4 HTR3E rs7627615 chr3_185301110 G A 0.35 0.04 8.81E−05 11.69 Receptor activity; extracellular ligand-gated ion channel activity 5 TMPRSS11A rs1370840 chr4_68492841 A G 0.31 0.02 0.000157 18.97 TMPRSS11A expression in normal oesophagus, liver, colon and lung, with downregulated expression in tumours 6 TMPRSS11A rs11930532 chr4_68494878 C T 0.29 0.02 0.000159 18.91 TMPRSS11A expression in normal oesophagus, liver, colon and lung, with downregulated expression in tumours 7 TMPRSS11A rs28437478 chr4_68493016 A T 0.29 0.02 0.000189 18.53 TMPRSS11A expression in normal oesophagus, liver, colon and lung, with downregulated expression in tumours 8 SALL2 rs1263810 chr14_21061466 G C 0.16 0.45 0.000206 0.23 DNA binding; sequence-specific DNA binding transcription factor activity 9 SPATA 22 rs9901726 chr17_3290284 T C 0.02 0.19 0.000223 0.09 SPATA22 10 SPATA 22 rs2291604 chr17_3299188 G C 0.02 0.19 0.000223 0.09 SPATA 22 11 NMNAT2 rs607332 chr1_181519836 A G 0.18 0.48 0.000227 0.24 Nucleotide binding; nicotinamide-nucleotide adenylyltransferase activity 12 NCKAP1 Novel chr2_183540400 T C 0.03 0.22 0.000266 0.11 Protein binding; plasma membrane; apoptosis 13 SPATA 22 rs9900543 chr17_3296435 C T 0.02 0.19 0.000316 0.09 SPATA 22 14 NCKAP1 Novel chr2_183501325 T C 0.03 0.22 0.000319 0.12 Protein binding; plasma membrane; apoptosis 15 MARVELD3 rs2303225 chr16_70232383 C G 0.42 0.11 0.000333 5.61 Membrane; integral to membrane; 16 TMPRSS11A rs6552134 chr4_68511802 G A 0.27 0.02 0.00048 16.25 TMPRSS11Aexpression in normal oesophagus, liver, colon and lung, with downregulated expression in tumours 17 ZNF491 Novel chr19_11778746 A C 0.01 0.15 0.000525 0.06 DNA binding; zinc ion binding 18 ZNF440 Novel chr19_11804552 A G 0.01 0.

gral to membrane; 16 TMPRSS11A rs6552134 chr4_68511802 G A 0.27 0.02 0.00048 16.25 TMPRSS11Aexpression in normal oesophagus, liver, colon and lung, with downregulated expression in tumours 17 ZNF491 Novel chr19_11778746 A C 0.01 0.15 0.000525 0.06 DNA binding; zinc ion binding 18 ZNF440 Novel chr19_11804552 A G 0.01 0. 15 0.000525 0.06 DNA binding; zinc ion binding 19 ZNF439 Novel chr19_11840305 G C 0.01 0.15 0.000525 0.06 DNA binding; zinc ion binding 20 CA2 rs2307075 chr8_86575480 C A 0.34 0.66 0.000528 0.27 Kidney development; morphogenesis of an epithelium; carbonate dehydratase activity; protein binding CA2, carbonate dehydratase II; HTR3E, 5-hydroxytryptamine receptor 3 family member E; MARVELD3, MARVEL domain containing 3; NCKAP1, NCK-associated protein 1; NMNAT2, nicotinamide mononucleotide adenylyltransferase 2; SALL2, sal-like protein 2; SNP, single nucleotide polymorphism; SPATA 22, spermatogenesis associated 22; SPINT2, serine peptidase inhibitor, kunitz type 2; TMPRSS11A, transmembrane protease, serine 11A; TNR, tenascin receptor; ZNF, zinc finger protein; From the function analysis, there were five SNPs of receptors, TNR, HTR3E and MARVELD3, and four SNPs from the transmembrane protease TMPRSS11A, which may affect HCV binding to the receptors and entry into hepatocytes (p<10−4).

15 0.000525 0.06 DNA binding; zinc ion binding 19 ZNF439 Novel chr19_11840305 G C 0.01 0.15 0.000525 0.06 DNA binding; zinc ion binding 20 CA2 rs2307075 chr8_86575480 C A 0.34 0.66 0.000528 0.27 Kidney development; morphogenesis of an epithelium; carbonate dehydratase activity; protein binding CA2, carbonate dehydratase II; HTR3E, 5-hydroxytryptamine receptor 3 family member E; MARVELD3, MARVEL domain containing 3; NCKAP1, NCK-associated protein 1; NMNAT2, nicotinamide mononucleotide adenylyltransferase 2; SALL2, sal-like protein 2; SNP, single nucleotide polymorphism; SPATA 22, spermatogenesis associated 22; SPINT2, serine peptidase inhibitor, kunitz type 2; TMPRSS11A, transmembrane protease, serine 11A; TNR, tenascin receptor; ZNF, zinc finger protein; From the function analysis, there were five SNPs of receptors, TNR, HTR3E and MARVELD3, and four SNPs from the transmembrane protease TMPRSS11A, which may affect HCV binding to the receptors and entry into hepatocytes (p<10−4). Discussion The differences between HCV genotypes coupled with the large diversity of host factors make an analysis of favourable and morbid outcomes in populations difficult to determine and analyse. Our study focused on a specific group of patients with HCV1b-infection due to a unique sole source reservoir of virus, the clear known time of infection, the similar ethnic and environmental background, and the broad scale of 85 patients including both sexes and almost all ages. With the same HCV source, this information may help us to better study the host factors involved in virus clearance and/or disease progress, in exclusion of HCV viral differences.

of infection, the similar ethnic and environmental background, and the broad scale of 85 patients including both sexes and almost all ages. With the same HCV source, this information may help us to better study the host factors involved in virus clearance and/or disease progress, in exclusion of HCV viral differences. There are only two studies comparable to the current study, one from Ireland and one from Germany. Both were reports of single-source infections with HCV1b involving homogeneous women of childbearing age who had received HCV-contaminated anti-D immunoglobulin to prevent Rh isoimmunisation from 1977 to 1979. The Irish group had 376 women with a mean age of 28±6 years, who had been infected for about 17 years.14 The German cohort had 1018 women with a median age of 24 years, who had been followed for 20 years.15 Both studies provided very important information on HCV spontaneous clearance and disease progression in large groups of patients with known durations of infection.16–22 However, owing to the relatively uniform age and gender, these two studies provide less data on how age and gender affect HCV spontaneous clearance and disease progression.

very important information on HCV spontaneous clearance and disease progression in large groups of patients with known durations of infection.16–22 However, owing to the relatively uniform age and gender, these two studies provide less data on how age and gender affect HCV spontaneous clearance and disease progression. IL-28B has recently been found to be a promising gene marker associated with treatment response and spontaneous clearance.3–7 However, data on the association between IL-28B genetic variants and spontaneous viral clearance in a Chinese study involving 376 HCV-infected paid plasma donors did not show an association with rs12979860. The other four SNPs, rs8099917, rs8105790, rs12980275 and rs10853728, were significantly associated with spontaneous HCV clearance.23 In our study, IL-28 rs10853728 showed a stronger association than the other seven SNPs, but the association fell short of statistical significance (p=0.058). This may be because the number of patients in this cohort was not large enough to provide statistical power to the differences. Also, in the highly prevalent IL-28B, favourable genotype area and, in particular, the Chinese population (>80%) may be different from the Caucasian population (40–50%).3 4

ificance (p=0.058). This may be because the number of patients in this cohort was not large enough to provide statistical power to the differences. Also, in the highly prevalent IL-28B, favourable genotype area and, in particular, the Chinese population (>80%) may be different from the Caucasian population (40–50%).3 4 TNR rs2239818 and rs34842046 are involved with receptor binding, extracellular matrix organisation and negative regulation of cell adhesion in the cell surface or extracellular region.24 25 Transmembrane protease TMPRSS11A rs1370840, rs11930532, rs28437478, rs6552134 and rs977728 are expressed in the normal liver, oesophagus, colon and lung, but downregulated in tumours.26 27 Neither TNR nor TMPRSS11A had been reported to be associated with spontaneous clearance of HCV. SPINT2 rs3745948 functions as an endopeptidase inhibitor within the plasma membrane, cytoplasm or extracellular region. Methylated SPINT2 and SRD5A2, combined with AFP and PIVKA-II, have been reported to be the most satisfactory panel to detect HCC in patients with chronically infected HCV.28 29 The mechanisms by which these three categories of SNPs result in the associations are not known. However, data from functional analyses suggest that they may affect HCV binding to the receptors and entering hepatocytes.

rted to be the most satisfactory panel to detect HCC in patients with chronically infected HCV.28 29 The mechanisms by which these three categories of SNPs result in the associations are not known. However, data from functional analyses suggest that they may affect HCV binding to the receptors and entering hepatocytes. There were two main limitations to the study. The first was the small number of patients. Though we cannot increase the sample size of this unique group since HCV were accidentally infected, the limited statistical power made us interpret the data more conservatively. Future studies, based on another larger scale cohort of patients with HCV and controls, are ongoing to verify those observations. Since this was a retrospective study, the other limitation was that we could not get the clinical data from the acute HCV phase about 10 years ago. The development of jaundice, other symptoms and laboratory findings during the acute phase could not be evaluated for spontaneous HCV clearance. In summary, this unique single-source HCV1b-infected patient population allows analysis of HCV1b with spontaneous clearance from a new perspective. Host gene SNPs within Tenascin-R, TMPRSS11A, and SPINT2 and IL-28 most likely play roles in the HCV spontaneous clearance and disease progression.

There were two main limitations to the study. The first was the small number of patients. Though we cannot increase the sample size of this unique group since HCV were accidentally infected, the limited statistical power made us interpret the data more conservatively. Future studies, based on another larger scale cohort of patients with HCV and controls, are ongoing to verify those observations. Since this was a retrospective study, the other limitation was that we could not get the clinical data from the acute HCV phase about 10 years ago. The development of jaundice, other symptoms and laboratory findings during the acute phase could not be evaluated for spontaneous HCV clearance. In summary, this unique single-source HCV1b-infected patient population allows analysis of HCV1b with spontaneous clearance from a new perspective. Host gene SNPs within Tenascin-R, TMPRSS11A, and SPINT2 and IL-28 most likely play roles in the HCV spontaneous clearance and disease progression. Contributors: HY contributed to patient management and analysis of the risk factors. SL was involved in local patient management. YX conducted local patients’ follow-up and maintained records. RC conducted the IL-28 SNPs analysis. YaS contributed to IL-28 detection. JR conducted the Exome capture and sequencing experiments and analysis. KW conducted SNPs data analysis. XJ contributed to SNPs comparison and analysis. YuS conducted bioinformatics analysis. HaZ was involved in HCV RNA detection. JL conducted patient follow-up and is a consultant. LW was involved in HCV RNA analysis and disease stage confirmation. HuZ was involved in study design and is a consultant. MC contributed to sample management and patient management. JJ contributed to the study design and final approval of the paper.

HCV RNA detection. JL conducted patient follow-up and is a consultant. LW was involved in HCV RNA analysis and disease stage confirmation. HuZ was involved in study design and is a consultant. MC contributed to sample management and patient management. JJ contributed to the study design and final approval of the paper. Funding: This work was supported by the Program for National Science and Technology Major Project (2013ZX10002004, 2012ZX10002003), Key Project from the Education Bureau of Beijing (KZ201210025024). Competing interests: None. Patient consent: Obtained. Ethics approval: The study was approved by the Ethics Committee from Guiyang Medical College, Guizhou, China. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Summary box What is already known about this subject? ▸  Bowel movements (BM) is perhaps the least understood and less studied of all human bodily functions. ▸  BM with less than three movements a week is considered as constipation, and more than three watery stools a day as diarrhoea. ▸  However, non-diarrhoeal increased frequency of BM (IFoBM-ND) in children has not been considered as a clinical symptom associated with viral infections. What are the new findings? ▸  This follow-up study of a cohort from birth up to 2 years of age reveals for the first time that non-polio enterovirus (NPEV) infections are associated with acute and persistent IFoBM-ND in children. ▸  NPEV-IFoBM-ND accounted for about 29% of NPEV infection episodes and 83% of total IFoBM-ND episodes. ▸  NPEV infections appear to result in resolution of about 74% of constipation episodes. How might it impact on clinical practice in the foreseeable future? ▸  The present finding of NPEV association with IFoBM in children, including acute and persistent IFoBM-ND and diarrhoea, could impact on future clinical understanding of IFoBM-ND and its recognition as a gastrointestinal disease symptom in clinical practice, and promote active research towards antienteroviral therapeutics.

e present finding of NPEV association with IFoBM in children, including acute and persistent IFoBM-ND and diarrhoea, could impact on future clinical understanding of IFoBM-ND and its recognition as a gastrointestinal disease symptom in clinical practice, and promote active research towards antienteroviral therapeutics. Background The frequency of bowel movements (BM) refers to the number of times a person discharges stool from the large intestine each day. There is no general rule for normal frequency of BM, which varies from three times a day to three times a week, and from person to person. However, it is commonly considered that a great majority of people have one BM each day.1–5 A variety of conditions may cause a significant increase in the frequency of BM, which include bacterial, parasitic or viral infections of the gastrointestinal tract, coeliac disease, cancer of the digestive tract, food allergies, diverticulitis, gall bladder disease or stones, lactose intolerance, inflammatory bowel disease (Crohn's disease and ulcerative colitis), irritable bowel syndrome (IBS), certain medications or drug overdose, etc.6–11 Increased frequency of BM can be either diarrhoeal, referred to as IFoBM-D, or non-diarrhoeal (IFoBM-ND). Frequent BM can be associated with a serious disease such as bowel obstruction and pancreatitis leading to a life-threatening condition. Other digestive as well as non-digestive symptoms including abdominal pain or cramping, bloody stool, diarrhoea, faecal incontinence, gas, abdominal swelling or bloating, nausea with or without vomiting, painful BM, urgent need to pass stool, body aches and fever might occur along with frequent BM.2

eatening condition. Other digestive as well as non-digestive symptoms including abdominal pain or cramping, bloody stool, diarrhoea, faecal incontinence, gas, abdominal swelling or bloating, nausea with or without vomiting, painful BM, urgent need to pass stool, body aches and fever might occur along with frequent BM.2 The majority of studies on the frequency of BM are based either on data recorded in sets of children of different age groups,5 12–14 or on the recording of BM only at specific months after birth15 or information provided by parents of children of specific age to a bowel-habit questionnaire.16 17 Further, significant differences in BM frequency appear to exist between children in developed and developing countries due to ethnic differences in dietary factors.4 Most of these studies suffer from severe variation in design and validity, and are only moderately sensitive.18 To the best of our knowledge, there is no systematic follow-up study on a day-to-day basis on frequency and type of BM, both IFoBM-D and IFoBM-ND, and analysis of the causes for change in BM frequency in a cohort of children from birth to 2 years of age.

tion in design and validity, and are only moderately sensitive.18 To the best of our knowledge, there is no systematic follow-up study on a day-to-day basis on frequency and type of BM, both IFoBM-D and IFoBM-ND, and analysis of the causes for change in BM frequency in a cohort of children from birth to 2 years of age. The genus enterovirus (EV), belonging to the family Picornaviridae, consists of a large group of viruses19 20 that are associated with a wide variety of diseases in humans including acute flaccid paralysis, aseptic meningitis, encephalitis, type-1 diabetes, hand-foot-and-mouth disease, uveitis, cardiomyopathy, aplastic anaemia, coagulopathy, multisystem haemorrhagic disease, sudden deafness, fatal illness with pulmonary hypertension in neonates, diarrhoea, etc.19–29 Poliovirus, coxsackievirus (CVA and CVB), echovirus (E), newer EVs and rhinovirus represent major pathogenic species of humans in the EV genus, and together present about 200 serotypes.19 20 A recent long-term epidemiological study in Bangalore, India, from our laboratory revealed a significant association of non-polio EVs (NPEVs) with acute diarrhoea in children.23 Further, a follow-up of 140 infants from birth for 6 months to 2 years of age revealed significant association of NPEV infections with acute diarrhoea as well as persistent diarrhoea (PD).30

alore, India, from our laboratory revealed a significant association of non-polio EVs (NPEVs) with acute diarrhoea in children.23 Further, a follow-up of 140 infants from birth for 6 months to 2 years of age revealed significant association of NPEV infections with acute diarrhoea as well as persistent diarrhoea (PD).30 During the large-scale purification of an echovirus 19 (E19) strain isolated from a child suffering from PD, called the persistent diarrhoea-associated virus (PDV; GenBank Accession number JX513514) for antibody generation, DCR had accidental infection resulting in a symptom, referred to as ‘non-diarrhoeal persistent IFoBM’ (PIFoBM-ND), which lasted for about 3 months. This symptom is often characterised by intense BM, an urgency to pass apparently normal stools more than twice a day instead of the normal once a day, abdominal bloating and discomfiture. This infection also resulted in a mild loss of motor function of the right hand, leading to stammering and incoherent handwriting. PDV was detected in the stools for more than 2 months, followed by E6 strain for about 12 days. The symptom subsided after a final mild diarrhoeal episode involving an enteropathogenic Escherichia coli infection for 3 days. Although IFoBM-ND shares some symptoms associated with IBS2 31–33 such as urgency for BM, discomfort, a feeling of incomplete evacuation (tenesmus) and abdominal bloating, as was experienced by DCR,30 comorbidities such as chronic abdominal pain, diarrhoea or constipation, headache, backache, fibromyalgia, chronic fatigue syndrome or psychiatric symptoms such as depression and anxiety frequently associated with IBS31–33 were not observed.

ete evacuation (tenesmus) and abdominal bloating, as was experienced by DCR,30 comorbidities such as chronic abdominal pain, diarrhoea or constipation, headache, backache, fibromyalgia, chronic fatigue syndrome or psychiatric symptoms such as depression and anxiety frequently associated with IBS31–33 were not observed. This unusual clinical symptom prompted us to undertake the present follow-up study of a cohort of infants from birth up to 2 years of age to understand the frequency of NPEV infections and their associated clinical outcomes including diarrhoea and IFoBM-ND. In a recent report, we described the frequency of NPEV infections, and association of 18% of the NPEV infection episodes with diarrhoea (12% acute and 6% persistent) in this cohort.23 We present an analysis of the apparently asymptomatic (referred to as non-diarrhoeal in our recent report23) NPEV infection episodes and their association with IFoBM-ND as observed in the cohort in the community setting.

f 18% of the NPEV infection episodes with diarrhoea (12% acute and 6% persistent) in this cohort.23 We present an analysis of the apparently asymptomatic (referred to as non-diarrhoeal in our recent report23) NPEV infection episodes and their association with IFoBM-ND as observed in the cohort in the community setting. Methods Definitions BM or bowel habit is referred to as discharge or passing of stool from the large intestine, or defaecation. Frequency of BM is the number of times an individual passes stools each day. BM ranges from three times a day to three times a week among different individuals. Less than three movements a week is considered as constipation34 and more than three watery stools a day is considered as diarrhoea.1 2 In this study, passing more than three apparently normal stools in a day or twice that of the existing frequency is considered as IFoBM-ND, the majority of acute IFoBM-ND (AIFoBM-ND) episodes lasting for less than 7 days and a few extending up to 14 days, and PIFoBM-ND episodes lasting for more than 14 days and up to three or more months. The corresponding IFoBM episodes involving watery stools are similarly considered as acute diarrhoea (AIFoBM-D) and PD (PIFoBM-D).35 36

acute IFoBM-ND (AIFoBM-ND) episodes lasting for less than 7 days and a few extending up to 14 days, and PIFoBM-ND episodes lasting for more than 14 days and up to three or more months. The corresponding IFoBM episodes involving watery stools are similarly considered as acute diarrhoea (AIFoBM-D) and PD (PIFoBM-D).35 36 The cohort The cohort, consisting of 140 infants, was recruited from birth over a period of 24 months between 2009 and 2010. All infants were normal and healthy at birth and remained normal during the study period. The major criteria for selection are the willingness of the parents to participate in the long-term study, and that they do not move out of the city during the study period. The mother was trained to carefully monitor and record the number and type of stools (diarrhoeal and non-diarrhoeal) daily, and to collect a stool sample in a sterile container on the designated date. The children were located within a radius of 6 km and the mothers were accessible by phone. R. M. V. Hospital, M. S. Ramaiah Teaching Hospital, Agadi Hospital and Arpita Clinic served as nodal centres for the 2-year follow-up study. More details on the cohort are described recently.23 30 Stool samples Stool samples were collected once in 14 days, and on alternate days during a diarrhoeal episode or an infection episode identified by NPEV or rotavirus (RV) positivity.23 30 The samples were transported from the residences of children to the laboratory on ice the same day.

The cohort The cohort, consisting of 140 infants, was recruited from birth over a period of 24 months between 2009 and 2010. All infants were normal and healthy at birth and remained normal during the study period. The major criteria for selection are the willingness of the parents to participate in the long-term study, and that they do not move out of the city during the study period. The mother was trained to carefully monitor and record the number and type of stools (diarrhoeal and non-diarrhoeal) daily, and to collect a stool sample in a sterile container on the designated date. The children were located within a radius of 6 km and the mothers were accessible by phone. R. M. V. Hospital, M. S. Ramaiah Teaching Hospital, Agadi Hospital and Arpita Clinic served as nodal centres for the 2-year follow-up study. More details on the cohort are described recently.23 30 Stool samples Stool samples were collected once in 14 days, and on alternate days during a diarrhoeal episode or an infection episode identified by NPEV or rotavirus (RV) positivity.23 30 The samples were transported from the residences of children to the laboratory on ice the same day. Oral polio vaccine period Almost every child in India receives oral polio vaccine (OPV) six to seven times during the first 2 years of life. Although a majority of OPV recipients shed vaccine strains for less than 7 days, a significant number of children were found to shed the virus up to 14 days and a few beyond 14 days.22 23 30 Hence, the clinical protocol should distinguish samples positive for OPV strains to reduce the burden of sequencing them from the large number of samples, which become EV positive in cell culture and reverse transcription PCR (RT-PCR). The 14-day period following OPV administration is referred to as the OPV period, and the majority of the samples positive for virus growth in rhabdomyosarcoma or HeLa cells, except the 100 random samples collected during this period, are considered OPV positive and are not subjected to RT-PCR and sequence analysis. However, all strains positive in the diarrhoea samples were identified by sequencing the VP1 gene.22 23 30

e samples positive for virus growth in rhabdomyosarcoma or HeLa cells, except the 100 random samples collected during this period, are considered OPV positive and are not subjected to RT-PCR and sequence analysis. However, all strains positive in the diarrhoea samples were identified by sequencing the VP1 gene.22 23 30 Identification of EV, RV and other agents in stool samples Methods for detection of NPEVs and other picornaviruses (kobuvirus, classevirus and cardiovirus), RV, diarrhoeagenic E. coli were described previously.22 23 30 The serotypes of NPEVs were identified by comparison of complete VP1 gene sequences with the prototype and other NPEV sequences in the GenBank database using the BLAST programme and by phylogenetic analysis.22 23 30 37 Nucleotide sequencing was carried out at Macrogen (Korea). Ethics Approvals from the Institutional Biosafety Committee and Ethical Committee were obtained for carrying out the work. Statistical analysis The data in Microsoft Office Excel format were subjected to statistical analyses using the Fisher exact test, the student t-test and computed p values as described.30 38

Identification of EV, RV and other agents in stool samples Methods for detection of NPEVs and other picornaviruses (kobuvirus, classevirus and cardiovirus), RV, diarrhoeagenic E. coli were described previously.22 23 30 The serotypes of NPEVs were identified by comparison of complete VP1 gene sequences with the prototype and other NPEV sequences in the GenBank database using the BLAST programme and by phylogenetic analysis.22 23 30 37 Nucleotide sequencing was carried out at Macrogen (Korea). Ethics Approvals from the Institutional Biosafety Committee and Ethical Committee were obtained for carrying out the work. Statistical analysis The data in Microsoft Office Excel format were subjected to statistical analyses using the Fisher exact test, the student t-test and computed p values as described.30 38 Results Age-dependent frequency of BM Owing to unexpected dropouts, the follow-up period for different children varied from 6 months to 2 years.30 For convenience, the study duration was divided into four periods of 6 months each. Analysis of the frequency of passing of stools and type of stools (diarrhoeal and normal) in the cohort recorded daily during the follow-up period revealed that the frequency of BM ranged between 0 and 16 during the first 6 months of life, between 0 and 12 during the 7–12-month period, between 0 and 8 for the 13–18-month period and between 0 and 4 during the 19–24-month period. The mean frequency among the cohort children for the four successive 6-month periods was 3.3, 1.6, 1.2 and 0.98, respectively (figure 1). Although the daily frequency of BM in a majority of the children ranged between 1 and 3, some infants passed stools once in 2–4 days or 5–10 days on more than one occasion (figure 2). Very few infants had 1 BM/day (<4% of total BM) during the first 6-month period, but the number of children having 1 BM/day increased with age with >92% having passing 1 BM/day by 2 years of age. The frequency of BM is high in all children below 6 months of age with 2, 3–4 and 5–16 BM/day representing 15%, 23% and 34%, respectively. Of note, while about 7% and 17% of the non-diarrhoeal BM occurred once in 2–4 days and 5–10 days, respectively, in children below 6 months of age, this category of BM represented less than 1–2% in children older than 6 months. The frequency of BM ranged between 0 and 16/day among the children in all age groups. Approximately 87% of IFoBM-ND episodes occurred during the first year with >57% occurring during the first 6 months. While about 82% and 10% of children in the age group of 12–18 months had 1 and 1–2 BM/day, respectively, about 97% of children in the age group of 19–24 months had only 1 BM/day. It should be noted that as a tradition, almost every infant in India is trained to pass stools every day in the morning by the time the child reaches 1–2 years of age.

children in the age group of 12–18 months had 1 and 1–2 BM/day, respectively, about 97% of children in the age group of 19–24 months had only 1 BM/day. It should be noted that as a tradition, almost every infant in India is trained to pass stools every day in the morning by the time the child reaches 1–2 years of age. Figure 1 Age-dependent average frequency of non-diarrhoeal bowel movements (BM-ND) in the cohort. The 2-year follow-up duration is divided for convenience into four 6-month periods. The frequency of non-diarrhoeal BM per day during the four successive 6-month periods from birth ranged between 0–16, 0–12, 0–8 and 0–4, respectively. Figure 2 Age-dependent changes in bowel movement (BM) frequencies. BM frequencies were grouped into six categories, 1 in a day (1/D), 2 in a day (2/D), 3–4 in a day (3–4/D), 5–17 in a day, 1 in 2–4 days (1/2–4D) and 1 in 5–10 days (1/5–10D). The per cent proportion of non-diarrhoeal BM of each category in each age group is calculated with reference to the total number of non-diarrhoeal BM of the respective category occurring in all the four age groups.

in a day (2/D), 3–4 in a day (3–4/D), 5–17 in a day, 1 in 2–4 days (1/2–4D) and 1 in 5–10 days (1/5–10D). The per cent proportion of non-diarrhoeal BM of each category in each age group is calculated with reference to the total number of non-diarrhoeal BM of the respective category occurring in all the four age groups. EV association with IFoBM-ND Since DCR experienced PIFoBM-ND for about 3 months following accidental infection with the E19 strain, PDV, isolated from a child with PD, the cohort children were carefully monitored each day for the frequency of passing of stools. The stool samples were examined for the presence of NPEVs, and the data were analysed for correlation between changes in frequency of BM and NPEV infection. Of 403 NPEV infection episodes observed among the 140 children during the follow-up period,23 30 115 were IFoBM-ND episodes. Further, 24 IFoBM-ND episodes that were negative for NPEV infection were also observed. Thus while NPEV-positive and NPEV-negative IFoBM-ND episodes represented 83% (115/139) and 17% (24/139) of total IFoBM episodes, respectively, the former accounted for 28.5% (115/403) of total NPEV infection episodes (table 1). The NPEV-negative IFoBM-ND episodes were all acute and represented about 17% of the episodes (table 1). Table 1 Analysis of IFoBM-ND episodes in comparison to diarrhoeal episodes

EV association with IFoBM-ND Since DCR experienced PIFoBM-ND for about 3 months following accidental infection with the E19 strain, PDV, isolated from a child with PD, the cohort children were carefully monitored each day for the frequency of passing of stools. The stool samples were examined for the presence of NPEVs, and the data were analysed for correlation between changes in frequency of BM and NPEV infection. Of 403 NPEV infection episodes observed among the 140 children during the follow-up period,23 30 115 were IFoBM-ND episodes. Further, 24 IFoBM-ND episodes that were negative for NPEV infection were also observed. Thus while NPEV-positive and NPEV-negative IFoBM-ND episodes represented 83% (115/139) and 17% (24/139) of total IFoBM episodes, respectively, the former accounted for 28.5% (115/403) of total NPEV infection episodes (table 1). The NPEV-negative IFoBM-ND episodes were all acute and represented about 17% of the episodes (table 1). Table 1 Analysis of IFoBM-ND episodes in comparison to diarrhoeal episodes Subjects, samples, episodes 0–6 months 6–12 months 12–18 months 18–24 months Total (%) 1 Cohort children followed (n=140) 140 130 112 70 140 2 Total samples 1724 1603 1381 864 5572 3 Total OPV-negative samples 1078 1430 1268 769 4545 4 Total NPEV-positive samples 260 179 135 84 638/4545 (14.0) 5 Total NPEV-negative samples 818 1251 1133 685 3907/4545 (86.0) 6 Total NPEV episodes 162 115 76 50 403/4545 (8.9) 7 NPEV+ve-acute episodes (%) 103/162 (63.6) 83/115 (72.2) 49/76 (64.5) 33/50 (66.0) 268/403 (66.5) Acute diarrhoeal episodes 19 13 10 6 48/403 (11.9) NPEV-AIFoBM-ND/total NPEV-IFoBM-ND episodes (%) 25/61 (41.0) 21/61 (34.4) 9/61 (14.8) 6/61 (9.8) 61/115 (53.0) NPEV-AIFoBM-ND/total NPEV episodes (%) 25/162 (15.4) 21/115 (18.3) 9/76 (11.8) 6/50 (12.0) 61/403 (15.1) NPEV-acute symptomatic (IFoBM+diarrhoea) 44 34 19 12 109/403 (27.1) Acute asymptomatic NPEV episodes 59 49 30 21 159/403 (39.5) 8 NPEV-positive persistent episodes 59 32 27 17 135 NPEV-positive PD episodes (%) 7/162 (4.3) 7/115 (6.1) 9/76 (11.8) 2/50) (4.0) 25/403 (6.2) NPEV-PIFoBM-ND/total NPEV-PIFoBM-ND (%) 26*/54 (48.2) 12/54 (22.2) 9/54 (16.7) 7/54 (13.0) 54/115 (47.0) NPEV-PIFoBM-ND/total NPEV episodes (%) 26/162 (16.0) 12/115 (10.4) 9/76 (11.8) 7/50 (14.0) 54/403 (13.4) NPEV-persistent symptomatic (PD+PIFoBM-ND) 33 19 18 9 79/403 (19.6) Persistent asymptomatic NPEV episodes 26 13 9 8 56/403 (13.9) 9 Non-NPEV-IFoBM-ND episodes 15 6 2 1 24/139 (17.3) 10 NPEV-IFoBM-ND/total NPEV-IFoBM-ND (%) 51/115 (44.4) 33/115 (28.7) 18/115 (15.6) 13/115 (11.3) 115/403 (28.5) NPEV-IFoBM-ND/total NPEV- +non-NPEV-IFoBM-ND (%) 51/139 (36.7) 33/139 (23.7) 18/139 (12.9) 13/139 (9.4) 115/139 (82.7) Total NPEV-IFoBM-ND/total non-diarrhoeal episodes (%) 51/330 (15.5) 33/330 (10.0) 18/330 (5.5) 13/330 (3.9) 115/330 (34.9) 11 NPEV diarrhoeal episodes/NPEV infection episodes (%) 26/162 (16.1) 20/115 (17.4) 19/76 (25.0) 8/50 (16.0) 73/403 (18.0) 12 NPEV symptomatic episodes/period (%) 77/162 (47.5) 53/115 (46.1) 37/76 (48.7) 21/50 (42.0) 188/403 (46.7) NPEV symptomatic episodes/total NPEV episodes (%) 77/403 (17.0) 53/403 (13.2) 37/403 (9.2) 21/403 (5.2) 13 NPEV asymptomatic episodes/period (%) 85/162 (52.5) 62/115 (53.9) 3

/76 (25.0) 8/50 (16.0) 73/403 (18.0) 12 NPEV symptomatic episodes/period (%) 77/162 (47.5) 53/115 (46.1) 37/76 (48.7) 21/50 (42.0) 188/403 (46.7) NPEV symptomatic episodes/total NPEV episodes (%) 77/403 (17.0) 53/403 (13.2) 37/403 (9.2) 21/403 (5.2) 13 NPEV asymptomatic episodes/period (%) 85/162 (52.5) 62/115 (53.9) 3 9/76 (51.3) 29/50 (58.0) 215/403 (53.3) NPEV asymptomatic episodes/total NPEV episodes (%) 85/403 (21.1) 62/403 (15.4) 39/403 (9.7) 29/40 (6.9) NPEV asymptomatic episodes/non-diarrhoeal episodes (%) 85/330 (25.8) 62/330 (18.8) 39/330 (11.8) 29/330 (8.8) 215/330 (65.2) 14 RV diarrhoea/RV episodes (%) 16/49 (32.7) 14/23 (60.9) 7/15 (46.7) 5/9 (55.6) 42/96 (43.7) 15 NPEV-RV-positive episodes 15 4 4 2 25/4545 (0.5) 16 NPEV+RV-diarrhoeal episodes/total diarrhoeal episodes (%) 9/193 (4.6) 1 (0.5) 1 (0.5) 1 (0.5) 12/193 (6.2) 17 Diarrhoea due to other agents/total diarrhoea (%) 25/193 (12.9) 20/193 (10.3) 13/193 (6.7) 8/193 (4.1) 66/193 (34.2) *One PIFoBM-ND episode was of NPEV-RV mixed infection. Details of diarrhoeal episodes, OPV samples, NPEV-positive and NPEV-negative samples have been described in detail previously.30 AIFoBM, acute increased frequency of bowel movements; IFoBM-ND, non-diarrhoeal IFoBM; NPEV, non-polio enterovirus; OPV, oral polio vaccine; PD, persistent diarrhoea; PIFoBM-ND, persistent IFoBM-ND; RV, rotavirus.

Details of diarrhoeal episodes, OPV samples, NPEV-positive and NPEV-negative samples have been described in detail previously.30 AIFoBM, acute increased frequency of bowel movements; IFoBM-ND, non-diarrhoeal IFoBM; NPEV, non-polio enterovirus; OPV, oral polio vaccine; PD, persistent diarrhoea; PIFoBM-ND, persistent IFoBM-ND; RV, rotavirus. While acute NPEV infections ranged between 64% and 72%, persistent infections were between 28% and 36% in all the age groups. Among 115 NPEV-positive IFoBM-ND episodes, acute and persistent episodes occurred at similar frequencies (61 acute (53%) and 54 persistent (47%)) with greater than 70% of both types of episodes occurring during the first year of life. AIFoBM-ND episodes accounted for about 15% of the total NPEV infection episodes and PIFoBM-ND episodes accounted for >13% (figure 3 and table 1). Figure 3 Per cent proportion of different non-polio enterovirus (NPEV)-associated symptoms with reference to total NPEV infection episodes in the cohort community during the 2-year follow-up period. Total number of samples, total NPEV-positive episodes, diarrhoeal episodes and oral polio vaccine samples in the follow-up study were described previously30 (IFoBM, increased frequency of bowel movements).

mptoms with reference to total NPEV infection episodes in the cohort community during the 2-year follow-up period. Total number of samples, total NPEV-positive episodes, diarrhoeal episodes and oral polio vaccine samples in the follow-up study were described previously30 (IFoBM, increased frequency of bowel movements). NPEV serotypes associated with IFoBM-ND We recently identified 24 NPEV serotypes associated with acute diarrhoea and 16 types with PD in this cohort.30 In the present analysis, of the 20 NPEV types detected, only 5 types E1, E6, E11, E19 and E24 were more frequently associated with IFoBM-ND. E6, E11 and E19 were frequently observed to persist for extended periods in PIFoBM-ND episodes, involving monotypic and sequential infections by different serotype strains (table 2). Note that many of the PIFoBM-ND episodes are repeated against different serotypes in table 2 due to sequential infections involving strains belonging to several NPEV serotypes during many of the episodes. Table 2 NPEV serotypes associated with acute diarrhoea, PD, AIFoBM, PIFoBM and asymptomatic infections, and the sequence of infections in PIFoBM episodes involving EV, RV and DEC

NPEV serotypes associated with IFoBM-ND We recently identified 24 NPEV serotypes associated with acute diarrhoea and 16 types with PD in this cohort.30 In the present analysis, of the 20 NPEV types detected, only 5 types E1, E6, E11, E19 and E24 were more frequently associated with IFoBM-ND. E6, E11 and E19 were frequently observed to persist for extended periods in PIFoBM-ND episodes, involving monotypic and sequential infections by different serotype strains (table 2). Note that many of the PIFoBM-ND episodes are repeated against different serotypes in table 2 due to sequential infections involving strains belonging to several NPEV serotypes during many of the episodes. Table 2 NPEV serotypes associated with acute diarrhoea, PD, AIFoBM, PIFoBM and asymptomatic infections, and the sequence of infections in PIFoBM episodes involving EV, RV and DEC EV-serotypes Strains in acute & PD episodes analysed Frequency of detection of different NPEV serotypes in 48 independent PIFoBM-ND episodes analysed AIFoBM-ND episodes analysed Total IFoBM-ND samples/episodes analysed Asymptomatic EV episodes analysed Total EV samples/episodes examined E1 2 10 (EV83, E1), (E1, 33), (E1, 11), (E32, 1, 2), (E7, 33, 1), (E1, 2), (E1, 25), (E11, 1), (E25, 25, 1, 1, 19, EPEC, RV), (E2, 1), 10 20 10 32 E2 2 3 (E2, 1), (E32, 1, 2), (E1, 2) – 3 – 5 E3 2+2 – 1 1 1 6 E5 1 – – – – 1 E6 5+10 13 (E6, 6, 6), (E6, 6, 6), (E6, 6, EPEC, RV), (E6, 6, 6), (E6, 6, 6, 6), (E21, 6, 6, OPV1), (E6, 19, 19), (E6, 19), (E6, 19, OPV1), (E11, 11, 6), (E21, 21, 6), (E11, 11, OPV1, 6, EAEC), (E19, 19, 19, 19, RV, E6, ETEC) 3 16 2 33 E7 2+8 2 (E7, 33, 1), (E21, 7) 2 4 1 15 E11 6+3 16 (E11, 11), (E11, 11), (E11, 11, 11, 11), (E11, 1), (E11, 11, 24), (E1, 11), (E25, 11), (E11, CVA9), (E32, 11), (E11, 11, 6), (E11, 11, OPV1, 6, EAEC), (E11, 11, 24, 24), (E11D, 32, OPV1), (OPV1, E11, 11), (E11D, 32, OPV1), (E11, RV) 6 22 8 39 E12 – – – – 2 2 E13 4+2 1 (E13, 13) 1 2 1 9 E14 2+1 4 (E14, 14), (E14, 24), (E1, 11, 14), (E14, 19), 2 6 2 11 E19 3+3 6 (E6,19,19), (E6, 19), (E14, 19) (E25, 25, 1, 19, 19, EPEC, RV), (E6, 19, RV, OPV1), (E19, 19, 19, 19, RV, E6, ETEC) 3 9 5 20 E20 2 2 (E20, 20), (E20, 33) 2 4 3 9 E21 – 3(E21, 21), (E21, 6, 6, OPV1), (E21, 21, 6) 1 4 2 6 E24 3+3 6 (E24, 24), (E24, 32), (E24, 32), (E11, 11, 24), (E14, 24), (E11, 11, 24, 24) 2 8 1 15 E25 – 3(E25, 25, 1, 1, 19, EPEC, RV), (E1, 25), (E25, 11) 2 5 2 7 E30 1 – – – 2 3 E32 1+1 4 (E32, 11), (E24, 32), (E32, 1, 2) (E11, 32, OPV1) 2 6 2 10 E33 4+2 3 (E7, 33, 1), (E1, 33), (E20, 33) 3 6 2 14 EV75 3+2 – – – – 5 EV76 – – – – 1 1 EV83 – 1 (EV83, E1) 1 2 1 3 CVA4 1+1 – – – 1 3 CVA9 1+1 – 1 1 1 4 CVA13 1 – – – 1 2 CVA24 1+1 – 1 1 – 3 CVB3 2 – 1 1 1 4 CVB4 1+2 1 1 1 5 CVB5 1+1 – – – – 2 OPV1 1+2 4 (OPV1, E11, 11), (E21, 6, 6, OPV1), (E11D, 32, OPV1), (E6, 19, OPV1) – 4 – 7 OPV2 1 – – – – 1 OPV3 1 – – – – 1 Total 54+45 =99 (80) 81 (12 mono+69 sequential) (48) 45 126 (91) 53 278 (222) The numbers in each row in each column represent the number of episodes from which the VP1 gene from strains belonging to the

+2 4 (OPV1, E11, 11), (E21, 6, 6, OPV1), (E11D, 32, OPV1), (E6, 19, OPV1) – 4 – 7 OPV2 1 – – – – 1 OPV3 1 – – – – 1 Total 54+45 =99 (80) 81 (12 mono+69 sequential) (48) 45 126 (91) 53 278 (222) The numbers in each row in each column represent the number of episodes from which the VP1 gene from strains belonging to the particular EV serotype was sequenced. Strains indicated in parenthesis in italic font refer to sequential infections by more than one NPEV serotype or agent in that order and those in normal font refer to monotypic infections. Strains from all the IFoBM-ND and asymptomatic NPEV infection episodes were not examined. Numbers in italic font in the bottom row include some sequentially collected samples during persistent episodes indicating representation of some episodes more than once, and the numbers in parenthesis in normal font in the bottom row indicate actual number of PIFoBM-ND episodes analysed. In the second column, numbers in normal and italic fonts represent samples from acute and persistent diarrhoeal episodes, respectively. AIFoBM, acute increased frequency of bowel movements; CVA, coxsackievirus A; CVB, coxsackievirus B; DEC, diarrhoeagenic Escherichia coli; E, echovirus; EAEC, enteroaggregative E. coli; EPEC, enteropathogenic E. coli; ETEC, enterotoxigenic E. coli; EV, enterovirus; IFoBM-ND, non-diarrhoeal IFoBM; NPEV, non-polio enterovirus; OPV, oral polio vaccine; PD, persistent diarrhoea; PIFoBM, persistent IFoBM; RV, rotavirus.

B; DEC, diarrhoeagenic Escherichia coli; E, echovirus; EAEC, enteroaggregative E. coli; EPEC, enteropathogenic E. coli; ETEC, enterotoxigenic E. coli; EV, enterovirus; IFoBM-ND, non-diarrhoeal IFoBM; NPEV, non-polio enterovirus; OPV, oral polio vaccine; PD, persistent diarrhoea; PIFoBM, persistent IFoBM; RV, rotavirus. Comparative analysis of diarrhoeal and IFoBM-ND episodes Analysis of IFoBM-ND and diarrhoeal episodes revealed that while diarrhoeal episodes accounted for 18%,30 IFoBM-ND accounted for about 29% (115/403) of the NPEV infection episodes. Thus, a significantly greater number of NPEV infections is associated with IFoBM-ND than with diarrhoea. While NPEV-associated diarrhoeal episodes accounted for about 38% (73/193), and RV-associated diarrhoea for 22% (42/193), those due to other agents represented about 34% (66/193) of the total diarrhoeal episodes (table 1).30 Taking into consideration our recent observation that about18–21% of diarrhoeal cases reporting to paediatricians are NPEV positive,23 about half of the NPEV-associated diarrhoeal episodes occurring in the community appear to remain unreported in India.23 30

ut 34% (66/193) of the total diarrhoeal episodes (table 1).30 Taking into consideration our recent observation that about18–21% of diarrhoeal cases reporting to paediatricians are NPEV positive,23 about half of the NPEV-associated diarrhoeal episodes occurring in the community appear to remain unreported in India.23 30 In our recent analysis of diarrhoeal episodes in the cohort, prior to analysis of IFoBM-ND symptoms, all NPEV infection episodes of IFoBM not associated with diarrhoea were considered as non-diarrhoeal or asymptomatic.30 Analysis revealed that about 35% (115/330) of the apparently asymptomatic NPEV infection episodes and 29% (115/403) of the total NPEV infection episodes were associated with IFoBM-ND symptoms. Thus, NPEV symptomatic infections (diarrhoea (18–21%)23 30 and IFoBM-ND (29%)), excluding NPEV-RV-positive mixed diarrhoeal infections, accounted for about 50% of total NPEV-positive infection episodes. Of interest, among the three highly prevalent NPEV serotypes (EV1, EV6 and EV11) associated with IFoBM-ND, E6 and E11 types were also frequently detected in diarrhoeal samples, but E1 is rarely associated with diarrhoea. Many serotypes were associated with symptomatic infections (diarrhoea and IFoBM-ND) and asymptomatic infections. While EV serotypes E21, E25 and EV83 were detected only in IFoBM-ND and asymptomatic infections, E5, EV75 and CVB5 were detected only in diarrhoeal samples30 (table 2).

1 is rarely associated with diarrhoea. Many serotypes were associated with symptomatic infections (diarrhoea and IFoBM-ND) and asymptomatic infections. While EV serotypes E21, E25 and EV83 were detected only in IFoBM-ND and asymptomatic infections, E5, EV75 and CVB5 were detected only in diarrhoeal samples30 (table 2). Similar to that observed in PD episodes,30 PIFoBM-ND episodes were characterised either by persistence of a single strain for a prolonged period or sequential infection by multiple strains during the persistent episode (table 2). While all AIFoBM-ND episodes involved monotype infections (table 2), 25% of PIFoBM-ND episodes were of monotype infections and 75% involved sequential infections by different strains. However, both types of infections occurred at a similar frequency in PD episodes.30 Among the agents examined, OPV strains, RV and diarrhoeagenic E. coli were detected in a small number of samples from a few episodes. When diarrhoeagenic E. coli was detected during some NPEV infection episodes, the stool was loose, but not diarrhoeal.

pes of infections occurred at a similar frequency in PD episodes.30 Among the agents examined, OPV strains, RV and diarrhoeagenic E. coli were detected in a small number of samples from a few episodes. When diarrhoeagenic E. coli was detected during some NPEV infection episodes, the stool was loose, but not diarrhoeal. Discussion Among all bodily functions, BM is the most poorly studied and least understood. Although infectious and non-infectious causes may be associated with IFoBM, bacterial, parasitic and viral infections are frequently associated with diarrhoea (IFoBM-D).1 2 23–30 34–36 To date, neither IFoBM-ND has been recognised as a clinical symptom nor a viral agent has been implicated to be associated with it. The present novel clinical finding of significant association of NPEV infection with IFoBM-ND in children strongly supports the personal experience and inconvenience suffered by DCR with the clinical symptom. This study represents the first description of a possible viral aetiology of IFoBM-ND, both AIFoBM-ND and PIFoBM-ND, in infants and children. NPEV-associated IFoBM-ND episodes accounted for about 29% of NPEV infection episodes and 83% of total IFoBM-ND episodes (table 1). In about 17% of IFoBM-ND episodes, no NPEV or other examined agents were detected, suggesting the association of non-infectious causes or unknown agents with these episodes.

s and children. NPEV-associated IFoBM-ND episodes accounted for about 29% of NPEV infection episodes and 83% of total IFoBM-ND episodes (table 1). In about 17% of IFoBM-ND episodes, no NPEV or other examined agents were detected, suggesting the association of non-infectious causes or unknown agents with these episodes. Of interest, 26 infants, included in the total number of NPEV-associated IFoBM-ND episodes, with a BM frequency of one in 4–10 days, had one or two normal stools per day following NPEV infection. However, no change in BM frequency was observed in nine children with constipation on NPEV infection. Thus, while about 74% of constipation episodes34 appear to have been resolved following NPEV infection, no effect was observed in 26% of constipation episodes on NPEV infection. To date, the clinical basis of constipation and their subsequent resolution in infants is poorly understood. The lack of change in BM frequency in these episodes could be due to the genetic diversity of the infecting NPEV isolates.

PEV infection, no effect was observed in 26% of constipation episodes on NPEV infection. To date, the clinical basis of constipation and their subsequent resolution in infants is poorly understood. The lack of change in BM frequency in these episodes could be due to the genetic diversity of the infecting NPEV isolates. Recently, we reported a significant association of NPEV infections with acute diarrhoea as well as PD, and that every child in this cohort had at least one NPEV infection, symptomatic (diarrhoeal) and/or asymptomatic, by 2 years of age.23 30 Analysis of NPEV strains associated with IFoBM-ND episodes in the present study revealed the patterns of NPEV association with AIFoBM-ND and PIFoBM-ND episodes to be similar to that observed for PD episodes. While monotype infections were associated with about 25% of PIFoBM-ND episodes, sequential infections by different EV serotype strains were associated with 75% of episodes. By contrast, infection involving monotype strains and multiple strains occurred at a similar frequency in PD episodes.30 OPV strains, RV and/or diarrhoeagenic E. coli were detected only in a small number of sequential samples from a few PIFoBM-ND episodes (table 2). It may be noted that two PIFoBM-ND episodes started as acute diarrhoea involving EV11 infection, but transformed into PIFoBM-ND after 4 days.

ilar frequency in PD episodes.30 OPV strains, RV and/or diarrhoeagenic E. coli were detected only in a small number of sequential samples from a few PIFoBM-ND episodes (table 2). It may be noted that two PIFoBM-ND episodes started as acute diarrhoea involving EV11 infection, but transformed into PIFoBM-ND after 4 days. Although no child in the cohort developed serious complications during NPEV infection episodes, 20 children had three to four episodes of common cold, 23 had minimal non-febrile illness and 19 had respiratory illness, 80% of which coincided with NPEV infections. These events associated equally with diarrhoeal and non-diarrhoeal NPEV episodes. Twenty-seven babies had non-specific vomiting during the first 3 months without correlation to EV infection. No hand-foot-and-mouth disease or meningitis symptoms were observed in the cohort during this period. All the children were normal and healthy and did not suffer from inflammatory bowel disease, IBS or coeliac disease.

seven babies had non-specific vomiting during the first 3 months without correlation to EV infection. No hand-foot-and-mouth disease or meningitis symptoms were observed in the cohort during this period. All the children were normal and healthy and did not suffer from inflammatory bowel disease, IBS or coeliac disease. A notable finding was that NPEV is the single most frequently and persistently detected infectious agent in PIFoBM-ND episodes than any other agent we have examined in this study, similar to that observed in PD episodes.30 In this context, EVs are known to cause persistent infections in humans, animal models and cells in vitro with some strains capable of persisting for several months.39–47 Although it is considered that <1% of EV infections result in severe disease with the majority being asymptomatic,19 our recent and present results reveal that about 50% of NPEV infections are associated with significant gastrointestinal symptoms involving diarrhoea and IFoBM-ND. Although the precise molecular basis for the symptomatic/asymptomatic nature of the isolates is not understood, mutations in the virus44–52 and host factors could influence the replication and clinical outcome of the infection.47–52 Our recent studies showed that Indian isolates exhibit a significant genetic diversity from strains isolated in other countries, forming distinct genogroups or subgenogroups within a genotype.22 23 30 This high level of genetic variation among isolates within a serotype could result in different isolates exhibiting a wide range of virulence phenotypes.

ian isolates exhibit a significant genetic diversity from strains isolated in other countries, forming distinct genogroups or subgenogroups within a genotype.22 23 30 This high level of genetic variation among isolates within a serotype could result in different isolates exhibiting a wide range of virulence phenotypes. In conclusion, this study revealed (1) an unrecognised but highly significant clinical symptom, IFoBM-ND, in children with possible EV aetiology, (2) a similarity in BM frequency between diarrhoeal and IFoBM-ND episodes in children, the two symptoms differing in the type of stools passed, with the former characterised by watery stools and the latter by apparently normal stools, (3) NPEV as the single most frequently and persistently detected infectious agent in PIFoBM-ND and (4) an association of approximately 50% of NPEV infection episodes with gastrointestinal diseases involving diarrhoea and IFoBM-ND. PD and IFoBM-ND episodes could result in high morbidity and/or mortality in rural community settings, imposing extreme trauma and economic burden on poor families in developing countries. Our recent and present findings of significant NPEV association with acute and persistent symptomatic infections involving diarrhoea and IFoBM-ND in children are of great clinical significance and child health concern, warranting urgent global attention for detailed future investigations on these ubiquitous viruses which remained neglected and unrecognised as enteric pathogens.

ion with acute and persistent symptomatic infections involving diarrhoea and IFoBM-ND in children are of great clinical significance and child health concern, warranting urgent global attention for detailed future investigations on these ubiquitous viruses which remained neglected and unrecognised as enteric pathogens. The authors are greatly indebted to the mothers of the cohort children for their interest, cooperation and support for this long-term study, without which this study could not have been possible. They thank A Raghavendra, Sudheendra Kumar, Harikrishna Reddy, E Surekha, G Ramya, R Bharat and V Vembuli for their technical support in processing of stool samples, for detection of NPEVs in cell culture primarily by DCR, analysis of stool samples for rotavirus by RNA PAGE and maintenance of sample records, and hence they were accorded coauthorship for their help in a previous manuscript. The authors are thankful to Mr Sanjay Kumar for collection and transport of stool samples from the residences of children according to schedule, and maintaining the details of information on OPV administration and samples at the hospitals. They also thank Dr A Srinivasa Rao, Georgia Regents University for help in statistical analysis of the data.

Mr Sanjay Kumar for collection and transport of stool samples from the residences of children according to schedule, and maintaining the details of information on OPV administration and samples at the hospitals. They also thank Dr A Srinivasa Rao, Georgia Regents University for help in statistical analysis of the data. Contributors: This project is an outcome of the personal clinical experiences of DCR, who is instrumental in designing, writing and securing funds for the implementation of the project's goals. DCR carried out an evaluation of a majority of the stool samples in cell culture, RNA extractions and RT-PCR, VP1 sequence and phylogenetic analyses, interpretation of data and writing of the manuscript. AMB and PPM were responsible for development and implementation of clinical protocols, selection, periodic monitoring and treatment of children during sickness and maintenance of clinical history of children, details of stool samples, and dates of OPV administration, and reviewed the manuscript. Funding: The work was supported by a grant (BT/01/COE/07/03) from the Department of Biotechnology, Government of India, from 2007 to 2012. Competing interests: None. Patient consent: Parental/guardian consent obtained. Ethics approval: Institutional Ethics committee, Indian Institute of Science. Provenance and peer review: Not commissioned, externally peer reviewed. Data sharing statement: Any data pertaining to OPV schedules, stool samples, NPEV strains and the cohort will be provided on request.

Summary box What is already known about this subject ▸  Diarrhoea is an important cause of morbidity especially in the young and the elderly. ▸  Effective and safe treatment is of particular importance to prevent complications often associated with episodes of acute diarrhoea. ▸  A combination of myrrh, coffee charcoal and chamomile has been used for decades for the support of the gastrointestinal function. ▸  Positive results have been obtained with the combination in maintaining remission in ulcerative colitis. What are the new findings ▸  A combination of myrrh, coffee charcoal and chamomile is effective in the treatment of acute diarrhoea symptoms in daily practice. ▸  Monotreatment with the herbal preparation is as efficacious as add-on treatment and treatment with other substances. ▸  Data indicates positive influence on the management of acute diarrhoea due to inflammatory bowel diseases and irritable bowel syndrome. How might it impact on clinical practice in the foreseeable future? ▸  Our findings indicate that symptoms associated with diarrhoea due to acute inflammatory disorders, inflammatory bowel diseases and irritable bowel syndrome can be effectively managed with the herbal preparation

▸  Data indicates positive influence on the management of acute diarrhoea due to inflammatory bowel diseases and irritable bowel syndrome. How might it impact on clinical practice in the foreseeable future? ▸  Our findings indicate that symptoms associated with diarrhoea due to acute inflammatory disorders, inflammatory bowel diseases and irritable bowel syndrome can be effectively managed with the herbal preparation Introduction Diarrhoea is a frequent symptom caused by different infectious as well as non-infectious aetiologies, reflecting different underlying gastrointestinal disorders, and is one of the most common diagnoses in family medicine.1 2 Here, the term ‘acute’ refers to symptoms lasting no longer than 2 weeks, ‘persistent’ or ‘transient’ refers to symptoms continuing for up to 4 weeks, while the term ‘chronic’ describes symptoms persisting longer than 4 weeks. Symptoms of acute diarrhoea usually refer to acute infective gastrointestinal disorders, but may also be present in other conditions such as inflammatory bowel diseases (IBD) and irritable bowel syndrome (IBS). Causative infectious agents in acute diarrhoea are bacteria and viruses usually transmitted through contaminated food, water or human contact. According to WHO, there are two billion cases of diarrhoeal disease worldwide every year.3 Children and the elderly are affected the most.1 Although relatively few patients die from diarrhoea in industrialised countries, it is still an important cause of morbidity and is responsible for substantial healthcare costs,3 while significantly contributing to mortality especially of children in developing countries.

hildren and the elderly are affected the most.1 Although relatively few patients die from diarrhoea in industrialised countries, it is still an important cause of morbidity and is responsible for substantial healthcare costs,3 while significantly contributing to mortality especially of children in developing countries. Acute diarrhoea is defined as the presence of three or more loose stools per day, a water content of more than 75% or a daily stool weight of more than 250 g. It is often accompanied by other symptoms such as vomiting, nausea or pain. Treatment starts with oral rehydration therapy to prevent or correct dehydration and, if necessary, is supplemented with antidiarrhoeal drugs.3 4

loose stools per day, a water content of more than 75% or a daily stool weight of more than 250 g. It is often accompanied by other symptoms such as vomiting, nausea or pain. Treatment starts with oral rehydration therapy to prevent or correct dehydration and, if necessary, is supplemented with antidiarrhoeal drugs.3 4 In Germany, a combination of myrrh, coffee charcoal and chamomile is well-known and has been used for more than 50 years for the treatment and support of gastrointestinal function.5 Myrrh is the aromatic resin from the Commiphora myrrha tree, which is valued for its astringent and antiseptic properties.6 7 Coffee charcoal has a high adsorptive capacity and is often utilised for treatment of acute diarrhoea.8 Chamomile flower extract possesses anti-inflammatory, antispasmodic and wound-healing effects and is used internally for symptomatic relief of gastrointestinal complaints.9 However, despite its long-term use and well-known safety profile, clinical data on the combination are scant. Recently, a randomised controlled trial demonstrated non-inferiority of the herbal preparation to the gold standard mesalazine in maintenance of remission in ulcerative colitis.10 The aim of this study was to verify the clinical efficacy and safety of the herbal preparation in the treatment of patients with symptoms associated with acute diarrhoea in daily practice.

ated non-inferiority of the herbal preparation to the gold standard mesalazine in maintenance of remission in ulcerative colitis.10 The aim of this study was to verify the clinical efficacy and safety of the herbal preparation in the treatment of patients with symptoms associated with acute diarrhoea in daily practice. Methods Study design This open prospective multicentre observational postmarketing study was conducted in 131 practices of family medicine and respiratory internal medicine in Germany between March 2012 and December 2013. The postmarketing study conforms to the current guidelines and was notified to the federal authority and the relevant institutions. According to German Drug Law, no approval by an institutional review board is required for observational studies. The duration of the observational period was defined according to the respective underlying disorder and was 7–14 days for patients with acute inflammatory disorders (AID) and 24–28 days for patients with IBD or IBS. At maximum, three visits were conducted: visit 1 (day 0 or baseline), visit 2 (day 7–14=last visit for patients with AID) and visit 3 (day 24–28=last visit for IBD and patients with IBS). The herbal medicinal product (Myrrhinil-Intest) used in this study is a combination of 100 mg myrrh powder, 50 mg coffee charcoal powder and 70 mg chamomile flower dry extract per coated tablet. No specifications were made regarding dosing, albeit most patients received the herbal preparation according to the Summary of Product Characteristics.5

t (Myrrhinil-Intest) used in this study is a combination of 100 mg myrrh powder, 50 mg coffee charcoal powder and 70 mg chamomile flower dry extract per coated tablet. No specifications were made regarding dosing, albeit most patients received the herbal preparation according to the Summary of Product Characteristics.5 Participants Participants aged at least 12 years with symptoms of acute diarrhoea due to AID of the gastrointestinal tract (enteritis, enterocolitis, gastroenteritis), IBD (ulcerative colitis, Crohn's disease) or IBS were included in this study after the appropriate therapy was chosen by the physician. Patients with IBS were not selected according to a specific IBS subtype. There were no preferences or limits regarding the choice of therapy. Prior to inclusion, informed patient consent was obtained. Assessment The primary outcome parameter was the pre-post change of symptoms, which was quantified as the change of the total mean score between beginning and study end. The total mean symptom score comprises the mean scores of the eight single symptoms (general well-being, stool frequency, stool consistency, blood or mucous in stool, flatulence, pain intensity, pain persistency, nausea and/or vomiting). This score was calculated separately for each subgroup (AID, IBD, IBS), each treatment option and, additionally, as an overall mean total symptom score which included all subgroups. Symptoms were recorded at each visit and assessed on 4-point Likert scales for the respective scores (table 1). Table 1 4-point Likert scales used to assess symptoms of acute diarrhoea

Assessment The primary outcome parameter was the pre-post change of symptoms, which was quantified as the change of the total mean score between beginning and study end. The total mean symptom score comprises the mean scores of the eight single symptoms (general well-being, stool frequency, stool consistency, blood or mucous in stool, flatulence, pain intensity, pain persistency, nausea and/or vomiting). This score was calculated separately for each subgroup (AID, IBD, IBS), each treatment option and, additionally, as an overall mean total symptom score which included all subgroups. Symptoms were recorded at each visit and assessed on 4-point Likert scales for the respective scores (table 1). Table 1 4-point Likert scales used to assess symptoms of acute diarrhoea Symptom score 0 1 2 3 General well-being Good Impaired Bad Very bad Stool frequency ≤2 3–5 6–8 ≥9 Stool consistency Hard or normal Somewhat loose Runny Watery Blood or mucous in stool None Some Moderate Many Flatulence Absent Mild Moderate Severe Pain intensity Absent Mild Moderate Severe Pain persistency Absent Sporadic Prolonged with relief after defecation Persistent Nausea and/or vomiting Absent Mild Moderate Severe Secondary efficacy objectives involved changes in score of single symptoms, duration of symptoms (according to diaries), physician's assessment of the clinical course and efficacy (5-point Likert scale from 0=worse to 4=complete resolution), and patient's satisfaction regarding the efficacy of the therapy (unsatisfactory, satisfactory, good, very good). Safety analysis included documentation of the frequency and severity of adverse events at each visit as well as daily in the patient's diary. Data on concomitant disorders, medication and therapy were collected at each visit. The physician-rated tolerability (poor, moderate, good, very good) was noted at each visit. Compliance and tolerability (good, not good) were checked via the patient's diary.

events at each visit as well as daily in the patient's diary. Data on concomitant disorders, medication and therapy were collected at each visit. The physician-rated tolerability (poor, moderate, good, very good) was noted at each visit. Compliance and tolerability (good, not good) were checked via the patient's diary. Statistical analyses Estimation of the sample size was based on the assumption that rare events (incidence of 1%) are detected with a probability of 95% within a population. Accordingly, a sample size of at least 600 patients was estimated to be required. An exact sample size calculation was not performed. It was planned to include approximately 1200 patients. For statistical analysis, three subgroups were formed according to the cause of acute diarrhoea (AID, IBD, IBS). Within each subgroup, patients were then divided into three groups according to the prescribed treatment. The herbal preparation was either prescribed as monotherapy or as add-on therapy (treatment group). A third group received treatments other than the herbal preparation (control group).

s in the AID (30.1%) and IBD (5.3%) groups. Patients with IBS applied spasmolytic/anticholinergic medication (7.9%) and probiotics (11%) more often. Other therapy consisted mainly of gastrointestinal preparations (AID: 44.8%, IBD: 2.0%, IBS: 12.3%) and homoeopathic medicine (AID: 15.8%) (see online supplementary data). Efficacy A reduction of the total mean symptom score was observed in all groups with all treatment options. The score decreased from 1.37±0.48 to 0.13±0.31, 1.54±0.50 to 0.40±0.43 and 1.21±0.42 to 0.43±0.35 in the AID, IBD and IBS groups, respectively. The pre-post change of the overall mean total symptom score was 1.34±0.48 to 0.20±0.35 (mono: 1.33±0.51 to 0.15±0.34, add-on: 1.39±0.41 to 0.30±0.37, other therapy: 1.31±0.43 to ±0.24±0.33). Covariance analysis of the primary efficacy criterion revealed a significant influence of the mean symptom score at baseline on the total mean symptom score. Thus, the change of clinical symptoms was adjusted to the same baseline criteria resulting in significant differences between disorder groups (AID, IBD, IBS; p=0.000) and therapy treatment (mono, add-on, other; p=0.023). Further analyses were conducted in the respective disorder groups, demonstrating no significant difference between treatment options in the AID (p=0.320) and IBD (p=0.554) groups. The pre-post change of the total mean symptom scores ranged from 1.21±0.03 to 1.25±0.01 in the AID group (mean difference: 0.000–0.035). In the IBD group, the change of the respective scores ranged from 1.08±0.08 to 1.27±0.22 (mean difference: 0.073–0.187). For IBS, significant differences (p=0.002) were observed between monotreatment and add-on treatment (mean difference: 0.140; 95% CI 0.036 to 0.245; p=0.009) and monotreatment and other therapy (mean difference: 0.217; 95% CI 0.085 to 0.349; p=0.001). Monotreatment of IBS with the herbal preparation resulted in a significantly higher total mean symptom score change compared to other treatment groups.

te diarrhoea (AID, IBD, IBS). Within each subgroup, patients were then divided into three groups according to the prescribed treatment. The herbal preparation was either prescribed as monotherapy or as add-on therapy (treatment group). A third group received treatments other than the herbal preparation (control group). Descriptive statistics were used for the analyses of all data. Adjusted analyses were conducted since the treatment according to physician's judgment might result in biased data. Thus, regression analyses were performed. The primary outcome parameter was rated using analysis of covariance. The duration of the disease was examined using Kaplan-Meier-method and Log-rank test.11 Change in the score of single symptoms was analysed using analysis of variance and F-test. The patients’ and physicians’ assessments (efficacy, tolerability) and other secondary efficacy parameters were calculated using contingency tables. Results During the observation period, data of 1062 patients were collected; 804 (75.5%) patients experienced acute diarrhoea due to AID, 53 (5%) patients due to IBD and 205 (19.3%) patients due to IBS. For subsequent visits, data were documented for 1045 patients at visit 2 and 255 patients at visit 3. Anthropometric data are presented in table 2. Table 2 Demographic data of the study population (n=number of patients)

Results During the observation period, data of 1062 patients were collected; 804 (75.5%) patients experienced acute diarrhoea due to AID, 53 (5%) patients due to IBD and 205 (19.3%) patients due to IBS. For subsequent visits, data were documented for 1045 patients at visit 2 and 255 patients at visit 3. Anthropometric data are presented in table 2. Table 2 Demographic data of the study population (n=number of patients) Demographic data AID (n=804) IBD (n=53) IBS (n=205) Total (n=1062) Age, years (mean±SD) 43.1±18.3 46.9±16.5 42.5±15.9 43.2±17.8 Female, n (%) 441 (55.0) 31 (58.5) 140 (68.3) 612 (57.7) Body weight, kg (mean±SD) 74.4±15.8 72.9±12.8 71.4±15.5 73.8±15.7 Body temperature, °C (mean±SD) 36.9±0.6 36.8±0.6 36.8±0.5 36.9±0.6 Risk factors for acute diarrhoea Alcohol, n (%) 90 (11.3) 8 (15.1) 38 (18.7) 136 (12.9) Travelling, n (%) 72 (9.1) 5 (9.4) 32 (15.7) 109 (10.4) Hormonal imbalance, n (%) 46 (5.8) 4 (7.5) 21 (10.3) 71 (6.7) Drug intake, n (%) 57 (7.2) 11 (20.8) 33 (16.3) 101 (9.6) Food intolerance, n (%) 101 (12.7) 14 (26.4) 109 (53.7) 224 (21.3) Smoking, n (%) 206 (25.9) 9 (17.0) 56 (27.5) 271 (25.8) Metabolic disorder, n (%) 88 (11.1) 5 (9.4) 10 (4.9) 103 (9.8) Stress, n (%) 282 (35.3) 35 (66.0) 155 (76.0) 472 (44.7) Concomitant illness, n (%) 232 (54.8) 36 (8.5) 155 (36.6) 423 (39.8) Concomitant medication, n (%) 240 (52.3) 50 (10.9) 169 (36.8) 459 (43.2) Treatment chosen by physician Monotreatment, n (%)* 524 (65.2) 21 (39.6) 86 (42.0) 631 (59.4) Add-on treatment, n (%)* 144 (17.9) 28 (52.8) 82 (40.0) 254 (23.9) Other therapy 136 (16.9) 4 (7.5) 37 (18.0) 177 (16.7) *Treatment with the herbal preparation.

tant medication, n (%) 240 (52.3) 50 (10.9) 169 (36.8) 459 (43.2) Treatment chosen by physician Monotreatment, n (%)* 524 (65.2) 21 (39.6) 86 (42.0) 631 (59.4) Add-on treatment, n (%)* 144 (17.9) 28 (52.8) 82 (40.0) 254 (23.9) Other therapy 136 (16.9) 4 (7.5) 37 (18.0) 177 (16.7) *Treatment with the herbal preparation. AID, acute inflammatory disorders; IBD, inflammatory bowel diseases; IBS, irritable bowel syndrome. Age ranged from 12 to 89 years: 35 patients were younger than 18 years (26 women, 19 men) and 22 patients were older than 80 years (14 women, 2 men). Most patients were 18–59 years old. Within the AID subgroup, gastroenteritis was the most frequent disease (table 3). Table 3 Specific medical history data of AID and IBD. No subdivision was possible for irritable bowel syndrome due to diverse symptomatology of unknown cause Monotreatment Add-on treatment Other therapy Total AID Enteritis, n (%) 175 (21.8) 19 (2.4) 15 (1.9) 209 (26.1) Enterocolitis, n (%) 35 (4.4) 19 (2.4) 5 (0.6) 59 (7.4) Gastroenteritis, n (%) 312 (38.9) 106 (13.2) 116 (14.5) 534 (66.6) IBD Ulcerative colitis, n (%) 10 (18.9) 14 (26.4) 3 (5.7) 27 (50.9) Acute, n (%) 8 (15.1) 7 (13.2) 3 (5.7) 18 (34.0) In remission, n (%) 2 (3.8) 7 (13.2) – 9 (17.0) Crohn's disease, n (%) 11 (20.8) 14 (26.4) 1 (1.9) 26 (49.1) Acute, n (%) 7 (13.2) 9 (17.0) – 16 (30.2) In remission, n (%) 4 (7.5) 5 (9.4) 1 (1.9) 10 (18.9) AID, acute inflammatory disorders; IBD, inflammatory bowel diseases.

te, n (%) 8 (15.1) 7 (13.2) 3 (5.7) 18 (34.0) In remission, n (%) 2 (3.8) 7 (13.2) – 9 (17.0) Crohn's disease, n (%) 11 (20.8) 14 (26.4) 1 (1.9) 26 (49.1) Acute, n (%) 7 (13.2) 9 (17.0) – 16 (30.2) In remission, n (%) 4 (7.5) 5 (9.4) 1 (1.9) 10 (18.9) AID, acute inflammatory disorders; IBD, inflammatory bowel diseases. Predominant aetiology for enteritis, enterocolitis and gastroenteritis was an assumed viral infection. Self-attributed common risk factors for an episode of acute diarrhoea were ‘stress’ (44.7%), ‘smoking’ (25.8%) and ‘food intolerance’ (21.3%). Concomitant diseases were documented in 233 patients; among these, the most common were metabolic disorders (113 patients, 26.7%) and cardiovascular disorders (103 patients, 24.3%). Most frequent concomitant medications were antihypertensive drugs (71 patients, 15.5%). The majority of patients received a dose of four tablets of the herbal preparation thrice daily. The most common preparations in the add-on herbal preparation group were gastrointestinal drugs in the AID (30.1%) and IBD (5.3%) groups. Patients with IBS applied spasmolytic/anticholinergic medication (7.9%) and probiotics (11%) more often. Other therapy consisted mainly of gastrointestinal preparations (AID: 44.8%, IBD: 2.0%, IBS: 12.3%) and homoeopathic medicine (AID: 15.8%) (see online supplementary data).

add-on treatment (mean difference: 0.140; 95% CI 0.036 to 0.245; p=0.009) and monotreatment and other therapy (mean difference: 0.217; 95% CI 0.085 to 0.349; p=0.001). Monotreatment of IBS with the herbal preparation resulted in a significantly higher total mean symptom score change compared to other treatment groups. Considering single symptoms, significant difference between groups was observed for the symptom ‘nausea/vomiting’ (p=0.003) in the AID group resulting in a better improvement with herbal preparation add-on treatment compared with monotreatment and other therapy (figure 1). At last visit, 94% of patients receiving monotreatment, 89.2% of patients receiving add-on treatment and 97% of patients receiving other therapy were free of symptoms. Median time until symptom resolution was 4 days for all groups. Paired comparison demonstrated a significantly smaller probability to be free of symptoms for patients applying the herbal preparation as add-on compared to monotreatment (p=0.019) or other therapy (p=0.005). No difference was observed between monotreatment and other therapy groups (p=0.138). Figure 1 Change of the score of each single symptom in the acute inflammatory disorders group (visit 1—last visit). Positive score difference values indicate an improvement.

Considering single symptoms, significant difference between groups was observed for the symptom ‘nausea/vomiting’ (p=0.003) in the AID group resulting in a better improvement with herbal preparation add-on treatment compared with monotreatment and other therapy (figure 1). At last visit, 94% of patients receiving monotreatment, 89.2% of patients receiving add-on treatment and 97% of patients receiving other therapy were free of symptoms. Median time until symptom resolution was 4 days for all groups. Paired comparison demonstrated a significantly smaller probability to be free of symptoms for patients applying the herbal preparation as add-on compared to monotreatment (p=0.019) or other therapy (p=0.005). No difference was observed between monotreatment and other therapy groups (p=0.138). Figure 1 Change of the score of each single symptom in the acute inflammatory disorders group (visit 1—last visit). Positive score difference values indicate an improvement. In the IBD group, stool frequency decreased significantly in the add-on treatment subgroup (p=0.026) compared to monotreatment and other therapy. At last visit, significant differences were observed for the symptom ‘flatulence’ (figure 2). Monotreatment resulted in significantly higher reduction of flatulence compared to add-on treatment or other therapy (p=0.038). The proportion of patients who were free of symptoms at the last visit was 80%, 64.3%, and 50% receiving monotreatment, add-on treatment or other therapy, respectively. Median time until symptom resolution was 16, 25 and 26 days for monotreatment, add-on treatment and other therapy, respectively. No difference was observed between groups regarding the probability to be free of symptoms.

t was 80%, 64.3%, and 50% receiving monotreatment, add-on treatment or other therapy, respectively. Median time until symptom resolution was 16, 25 and 26 days for monotreatment, add-on treatment and other therapy, respectively. No difference was observed between groups regarding the probability to be free of symptoms. Figure 2 Change of the score of each single symptom in the inflammatory bowel diseases group (visit 1—last visit). Positive score difference values indicate an improvement. In the IBS group, improvement of stool frequency, flatulence and pain intensity was noted during the observation period. Stool frequency improved significantly with add-on treatment and other therapy compared with monotreatment (p=0.023). Pain intensity was lower with add-on treatment compared with monotreatment and other treatment options (p=0.018). Flatulence improved with monotreatment during the observation (p=0.001) and at last visit (p=0.010). A further significantly improved symptom at study end was ‘stool texture’ with add-on treatment and other therapy (p=0.047; figure 3). At last visit, 54.2% of patients receiving monotreatment, 38.5% receiving add-on treatment and 35.1% receiving other therapy were free of symptoms. Median time until symptoms resolution was shortest for monotreatment (27 days). No difference was observed between groups regarding the probability to be free of symptoms. Figure 3 Change of the score of each single symptom in the irritable bowel syndrome group (visit 1—last visit). Positive score difference values indicate an improvement.

In the IBS group, improvement of stool frequency, flatulence and pain intensity was noted during the observation period. Stool frequency improved significantly with add-on treatment and other therapy compared with monotreatment (p=0.023). Pain intensity was lower with add-on treatment compared with monotreatment and other treatment options (p=0.018). Flatulence improved with monotreatment during the observation (p=0.001) and at last visit (p=0.010). A further significantly improved symptom at study end was ‘stool texture’ with add-on treatment and other therapy (p=0.047; figure 3). At last visit, 54.2% of patients receiving monotreatment, 38.5% receiving add-on treatment and 35.1% receiving other therapy were free of symptoms. Median time until symptoms resolution was shortest for monotreatment (27 days). No difference was observed between groups regarding the probability to be free of symptoms. Figure 3 Change of the score of each single symptom in the irritable bowel syndrome group (visit 1—last visit). Positive score difference values indicate an improvement. At study end, physicians predominantly assessed the clinical course as considerable improvement to complete resolution in all groups. The difference between respective treatment options was statistically not significant. At least 70% of the physicians rated the efficacy as good to very good. No differences regarding efficacy were observed in the IBD group. In the AID and IBS groups, good to very good efficacy occurred significantly often in the monotreatment group (p<0.001). Patient's satisfaction regarding efficacy was significantly higher with monotreatment in the AID and IBS groups (p<0.001) and with add-on treatment in IBD group (p=0.043). Tolerability was rated predominantly good to very good in all groups. Treatment with the herbal preparation was generally better tolerated (AID, IBS: p<0.001; IBD: p=0.027).

garding efficacy was significantly higher with monotreatment in the AID and IBS groups (p<0.001) and with add-on treatment in IBD group (p=0.043). Tolerability was rated predominantly good to very good in all groups. Treatment with the herbal preparation was generally better tolerated (AID, IBS: p<0.001; IBD: p=0.027). Safety Only two side effects were reported in relation to the herbal preparation. No serious adverse reactions in relation to the study medication were reported. In one case ‘vomiting’ was reported in a patient with IBD, who received 3×2 tablets daily, and ‘itching’ in another patient with IBS receiving 4×3 tablets daily. Both patients applied the herbal preparation as add-on therapy. No dose adjustment or additional treatment was necessary and both patients finished the study regularly. Discussion The aim of this observational study was to collect data regarding the efficacy, safety and tolerability of an herbal preparation containing a combination of myrrh, coffee charcoal and chamomile in the treatment of gastrointestinal disorders associated with symptoms of acute diarrhoea.

Safety Only two side effects were reported in relation to the herbal preparation. No serious adverse reactions in relation to the study medication were reported. In one case ‘vomiting’ was reported in a patient with IBD, who received 3×2 tablets daily, and ‘itching’ in another patient with IBS receiving 4×3 tablets daily. Both patients applied the herbal preparation as add-on therapy. No dose adjustment or additional treatment was necessary and both patients finished the study regularly. Discussion The aim of this observational study was to collect data regarding the efficacy, safety and tolerability of an herbal preparation containing a combination of myrrh, coffee charcoal and chamomile in the treatment of gastrointestinal disorders associated with symptoms of acute diarrhoea. Analysis of the change of the mean total symptom score indicates that monotreatment with the herbal preparation is as effective as add-on treatment or other therapy in the management of acute diarrhoea due to AID and IBD. However, the results obtained for the IBD subgroup must be interpreted with caution due to a small number of patients. Thus the trend observed in the IBD subgroup needs to be verified in a larger population. Regarding IBS, a significantly higher reduction of symptoms was observed with the monotreatment indicating better management with the herbal preparation. Further, the results of the secondary efficacy criteria suggest that monotreatment with 3×4 tablets herbal preparation per day is at least as effective as add-on and other therapy.

a significantly higher reduction of symptoms was observed with the monotreatment indicating better management with the herbal preparation. Further, the results of the secondary efficacy criteria suggest that monotreatment with 3×4 tablets herbal preparation per day is at least as effective as add-on and other therapy. The efficacy of the herbal preparation in the management of gastrointestinal symptoms is based on a proposed synergistic effect of the three active ingredients. Owing to a diverse spectrum of constituents, plants often possess various pharmacological effects. Myrrh was recently shown to reduce intestinal muscle tone and acetylcholine-induced contraction of inflamed rat ileum/jejunum preparations, thus contributing to reduction of intestinal motility and spasmolytic effects.12 Similar results were obtained for chamomile extract and its flavonoids in guinea pig ileum.13 14 Several constituents of the chamomile flower possess anti-inflammatory potential, for example, chamomile essential oil reduced TNBS-induced colitis in mice and mouse paw oedema.15 Chamazulen was reported to prevent leucotriene formation and chemical peroxidation of arachidonic acid, the latter likely through inhibition of COX-2.16 17 Chamomile extract also produced a 41% inhibition of rat paw oedema, and chamomile and myrrh showed a gastroprotective effect in rats.18 19 Anti-inflammatory and analgesic activity of myrrh extracts was recently demonstrated utilising the paw oedema mice model.20 Sesquiterpenes seem to be the main constituents responsible for the pharmacological effect of myrrh. These substances possess antibacterial and antifungal activity, an effect that might act supportive in infectious diarrhoea. Additionally, sesquiterpenes exert local anaesthetic activity by blocking the inward sodium current of excitable mammalian membranes.21 Coffee charcoal contributes its high adsorptive capacity and astringent activity of chlorogenic acid to the spectrum of activity of the herbal preparation.22

e in infectious diarrhoea. Additionally, sesquiterpenes exert local anaesthetic activity by blocking the inward sodium current of excitable mammalian membranes.21 Coffee charcoal contributes its high adsorptive capacity and astringent activity of chlorogenic acid to the spectrum of activity of the herbal preparation.22 Of special interest are the positive results obtained in patients with IBD such as ulcerative colitis and Crohn's disease and in patients with IBS since those subgroups frequently require treatment of acute diarrhoea episodes due to the underlying disease. The nature of IBD is complex and involves chronic, uncontrolled inflammation of the intestinal mucosa with the inability to downregulate this activity. Factors contributing to the development of mucosal inflammation range from environmental to genetic influences. IBD is also characterised by intestinal barrier defects that are associated with increased permeability of the epithelial surface.23 The mechanisms responsible for IBS are not yet clear, but disturbances of the intestinal epithelial barrier seem to play a major role in development of symptoms. Patients with diarrhoea-predominant IBS frequently display abnormal small intestinal permeability.24 Epithelial permeability is determined by tight junctions (TJ), a complex of molecules that regulate the paracellular transport of ions and act as a barrier. Altered regulation of TJ in the form of upregulation of the poreforming claudin-2 and downregulation of the sealing proteins claudin-5 and -8 in the sigmoid colon lead to barrier dysfunction in active Crohn's disease, while downregulation of the sealing proteins occludin, claudin-1 and -4 and upregulation of claudin-2 were observed in ulcerative colitis.25 26 Proinflammatory cytokines also play an important role in epithelial damage by affecting TJ regulation, for example, tumour necrosis factor α (TNFα) causes upregulation of claudin-2.27 Myrrh, a component of the herbal preparation, and the combination itself were shown to inhibit TNFα-induced decrease in epithelial resistance through inhibition of claudin-2 expression. Further, a redistribution of claudin-1 into subapical compartments was suppressed. Similarly, chamomile extract induced an increased expression of claudin-7.28 Thus, regulation of TJs might constitute a possible mechanism for the improvement observed in these subgroups in the study.

inhibition of claudin-2 expression. Further, a redistribution of claudin-1 into subapical compartments was suppressed. Similarly, chamomile extract induced an increased expression of claudin-7.28 Thus, regulation of TJs might constitute a possible mechanism for the improvement observed in these subgroups in the study. Treatment with the herbal preparation was generally better tolerated and was assessed as superior. Only two already known adverse reactions were described in 1062 patients resulting in a very low incidence of 0.2%. Both side effects were of short duration and both patients recovered. Allergic reactions to myrrh or composite plants including chamomile are not unlikely and might result in the symptom ‘itching’.29 30 ‘Vomiting’ occurred in a patient suffering from food intolerances and might also be connected to the underlying disorder.

side effects were of short duration and both patients recovered. Allergic reactions to myrrh or composite plants including chamomile are not unlikely and might result in the symptom ‘itching’.29 30 ‘Vomiting’ occurred in a patient suffering from food intolerances and might also be connected to the underlying disorder. Limitations generally encountered in non-interventional studies are lack of randomisation, blinding or standardised treatment protocol. Since the allocation to treatment was conducted by the physician in accordance with the patient, no balance between groups regarding baseline score or treatment data was expected. Thus, regression to the mean was performed to reduce potential bias resulting from different treatment procedures. Further, the number of patients with IBD treated with other therapy was smaller compared to monotreatment and add-on treatment. Thus, only a tendency regarding comparability to other therapies can be made for the IBD group. However, in contrast to data derived from clinical trials in a restricted population, the large body of data provides a more accurate description on efficacy and safety of the herbal preparation in the management of acute diarrhoea in a heterogeneous population encountered in daily practice. Conclusion The combination of myrrh, coffee charcoal and chamomile flower extract is effective, well-tolerated and safe for use in patients with symptoms of acute diarrhoea. Its efficacy is comparable to other therapies used in routine care and apparently more effective in patients with IBS.

Limitations generally encountered in non-interventional studies are lack of randomisation, blinding or standardised treatment protocol. Since the allocation to treatment was conducted by the physician in accordance with the patient, no balance between groups regarding baseline score or treatment data was expected. Thus, regression to the mean was performed to reduce potential bias resulting from different treatment procedures. Further, the number of patients with IBD treated with other therapy was smaller compared to monotreatment and add-on treatment. Thus, only a tendency regarding comparability to other therapies can be made for the IBD group. However, in contrast to data derived from clinical trials in a restricted population, the large body of data provides a more accurate description on efficacy and safety of the herbal preparation in the management of acute diarrhoea in a heterogeneous population encountered in daily practice. Conclusion The combination of myrrh, coffee charcoal and chamomile flower extract is effective, well-tolerated and safe for use in patients with symptoms of acute diarrhoea. Its efficacy is comparable to other therapies used in routine care and apparently more effective in patients with IBS. Supplementary Material Supplementary Materials Contributors: UA was involved in planning, conducting and reporting of the study. BS was the responsible statistician and performed the statistical analyses. RS and VM drafted the article. All authors contributed to subsequent and final drafts. UA is the guarantor.

Conclusion The combination of myrrh, coffee charcoal and chamomile flower extract is effective, well-tolerated and safe for use in patients with symptoms of acute diarrhoea. Its efficacy is comparable to other therapies used in routine care and apparently more effective in patients with IBS. Supplementary Material Supplementary Materials Contributors: UA was involved in planning, conducting and reporting of the study. BS was the responsible statistician and performed the statistical analyses. RS and VM drafted the article. All authors contributed to subsequent and final drafts. UA is the guarantor. Funding: This research was funded by Repha GmbH and the company was also the sponsor of the study. Competing interests: BS received remuneration from Repha GmbH for performing statistical analyses, RS for his consulting service and UA was compensated for the coordination of the study. VM has no competing interests. Patient consent: Obtained. Ethics approval Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Summary box What is already known about this subject? ▸  Previous studies suggest that the weight reduction effect of auricular acupuncture occurs via the vagus nerve. ▸  However, these different experimental models have not provided definitive answers regarding treatment efficacy or the underlying mechanism of action. ▸  The acupuncture sites used in this study lie along the auricular branch of the vagus nerve (Arnold's nerve), which is its only somatosensory branch. What are the new findings? ▸  This study examined changes in body weight and in the levels of several appetite-related peptides in response to auricular acupuncture. ▸  We found that the increase in early morning fasting ghrelin levels in the acupuncture group was suppressed after 1 week of auricular acupuncture treatment, whereas all of the participants in the placebo group showed an increase in early morning fasting ghrelin levels. ▸  Despite the small sample sizes, the results suggest that all participants in the auricular acupuncture group experienced weight loss during week 1 of the study. How might it impact on clinical practice in the foreseeable future? ▸  Although this study is preliminary and the number of cases is small, it is suggested that the results provide a scientific basis for the effect of acupuncture on appetite.

▸  Despite the small sample sizes, the results suggest that all participants in the auricular acupuncture group experienced weight loss during week 1 of the study. How might it impact on clinical practice in the foreseeable future? ▸  Although this study is preliminary and the number of cases is small, it is suggested that the results provide a scientific basis for the effect of acupuncture on appetite. Introduction The increasing prevalence of lifestyle-associated diseases is linked to an increased incidence of arteriosclerotic conditions such as cerebral infarction and myocardial infarction, which pose a major problem from a medical and an economic point of view. Overeating is a major cause of visceral fat accumulation, and overconsumption of calories and a lack of exercise are problems for people in the busy modern world. Auricular acupuncture, a traditional form of medicine practised worldwide, can suppress the appetite, although in the West it is also used for pain relief1 and to treat drug addiction2 and anxiety.3 Appetite suppression in response to acupuncture has been studied in rats, mice and humans; however, these different experimental models have not provided definitive answers regarding treatment efficacy or the underlying mechanism(s) of action.4 Progress is being made in understanding the mechanisms that regulate appetite, and several appetite-regulatory peptides were recently identified.5

s, mice and humans; however, these different experimental models have not provided definitive answers regarding treatment efficacy or the underlying mechanism(s) of action.4 Progress is being made in understanding the mechanisms that regulate appetite, and several appetite-regulatory peptides were recently identified.5 Previous studies suggest that the weight reduction effect of auricular acupuncture occurs via stimulation of the hunger and stomach points.6 Anatomically, the hunger point is located in the tragus, and the stomach point is located in the cuvum conchae, with both points falling within the distribution of the auricular branch of the vagus nerve. This study examined changes in feeding behaviour and in the levels of several appetite-related hormones in response to auricular acupuncture, and also attempted to identify the mechanism(s) by which this traditional medical treatment exerts its effects.

ithin the distribution of the auricular branch of the vagus nerve. This study examined changes in feeding behaviour and in the levels of several appetite-related hormones in response to auricular acupuncture, and also attempted to identify the mechanism(s) by which this traditional medical treatment exerts its effects. Materials and methods Subjects Ten healthy adult volunteers (nine female and one male; median age, 40 years (range, 24–44 years)) comprising students, office workers and housewives were recruited at the KOSAI Oriental Healthcare Center. The participants were randomly assigned to one of two groups (n=5 per group), an acupuncture group and a placebo group, using sealed envelopes. None of the participants had an endocrine disease or a history of diabetes, heart disease, allergy or immunological disease, or stroke. None were pregnant or lactating, or had given birth within the past 6 months, and all had been able to exercise to manage their weight within the past 3 months. Written consent was obtained from all participants after the study objectives, methods and safety issues were clearly and fully explained. The study protocol was approved by the Human Investigation Committee of the KOSAI Oriental Healthcare Center. All participants were free to withdraw from the study at any time.

ths. Written consent was obtained from all participants after the study objectives, methods and safety issues were clearly and fully explained. The study protocol was approved by the Human Investigation Committee of the KOSAI Oriental Healthcare Center. All participants were free to withdraw from the study at any time. Acupuncture Common auricular points used for the treatment of obesity include the hunger and stomach points, which regulate satiety and fullness.7 The hunger and stomach points were located using a Chinese auriculotherapy chart8 and Nogier's somatotopic inversion (figure 1).9 The precise locations of the acupuncture sites were identified using a Tormeter IW-ZEN device (Model 5505; Kuga Denshikiki Seisakusho, Japan), which measures autonomic nerve activity. This device detects zones of low electrical resistance in the skin; these zones correspond to acupuncture points.10 Briefly, the auriculotherapy electrode (tip diameter, 1.0 mm) was used to measure electrical resistance at the hunger point in the tragus and within the distribution of the auricular branch of the vagus nerve above the auricle (where the stomach point is located). The site with the least electrical resistance was used as the acupuncture point. Auricular acupuncture points were selected and treatment was performed by a certified and experienced Japanese acupuncturist at the KOSAI Oriental Healthcare Center.

f the vagus nerve above the auricle (where the stomach point is located). The site with the least electrical resistance was used as the acupuncture point. Auricular acupuncture points were selected and treatment was performed by a certified and experienced Japanese acupuncturist at the KOSAI Oriental Healthcare Center. Figure 1 Acupuncture stimulation sites (hunger point and stomach point) For the acupuncture group, indwelling intradermal needles were placed at the indicated sites (arrows) and fixed in position with medical tape. For the placebo group, intradermal needles were fixed with medical tape, but did not pierce the skin.

f the vagus nerve above the auricle (where the stomach point is located). The site with the least electrical resistance was used as the acupuncture point. Auricular acupuncture points were selected and treatment was performed by a certified and experienced Japanese acupuncturist at the KOSAI Oriental Healthcare Center. Figure 1 Acupuncture stimulation sites (hunger point and stomach point) For the acupuncture group, indwelling intradermal needles were placed at the indicated sites (arrows) and fixed in position with medical tape. For the placebo group, intradermal needles were fixed with medical tape, but did not pierce the skin. For the acupuncture group, the acupuncture site was cleaned with ethanol and an intradermal needle (3 mm in length, 0.12 mm in diameter; Seirin, Japan) was inserted horizontally to a depth of approximately 1–2 mm. Needles were inserted bilaterally. The needles were fixed with medical tape and left in position for 1 week. All needles were replaced on a weekly basis for a period of 4 weeks. A new ‘session’ began immediately after the needle was replaced. For the placebo group, the intradermal needle was fixed with medical tape but did not pierce the skin. Only the acupuncturist was aware of the group into which each participant had been assigned. The participants all firmly believed that they were receiving auricular acupuncture stimulation. At no point were the participants told whether or not they had received auricular acupuncture. The same acupuncture points were used for all participants, and all received acupuncture treatment once per week for four consecutive weeks (ie, four sessions) at the KOSAI Oriental Healthcare Center.

acupuncture stimulation. At no point were the participants told whether or not they had received auricular acupuncture. The same acupuncture points were used for all participants, and all received acupuncture treatment once per week for four consecutive weeks (ie, four sessions) at the KOSAI Oriental Healthcare Center. Observations Observations were conducted on a weekly basis for four consecutive weeks. Body weight, waist circumference and body mass index (BMI) were recorded before the start of the study and then on a weekly basis thereafter. Both groups kept a record of their diet during the study and a dietician assessed the food intake of each participant from pictures taken with a digital camera. No subjective measures were performed; however, changes in diet and food intake were monitored through the dietician's records. Both groups were instructed to carry on as normal during the observation period, although they were not allowed to receive acupuncture or to take any medications. Urinalysis and blood tests Fasting urine and blood samples were collected by a medical doctor at Tokyo Women's Medical University before the study began and again after 1 week of acupuncture treatment (before the needle was replaced). Samples were collected early in the morning before breakfast (between 6:30 and 8:00). A new acupuncture session began after the blood sample was collected.

ted by a medical doctor at Tokyo Women's Medical University before the study began and again after 1 week of acupuncture treatment (before the needle was replaced). Samples were collected early in the morning before breakfast (between 6:30 and 8:00). A new acupuncture session began after the blood sample was collected. In addition to general urinalysis, blood biochemistry and haematological tests were performed and the concentrations of insulin, C-peptide, leptin, adiponectin, adrenocorticotropic hormone (ACTH), active ghrelin and desacyl-ghrelin were measured. Blood ghrelin concentrations were measured using active ghrelin ELISA and desacyl-ghrelin ELISA kits (SCETI, Japan). All tests were performed by Mitsubishi Chemical Medience (Japan). The percentage changes in the levels of appetite-regulating factors were calculated as follows: ((value after acupuncture−value before acupuncture)/value before acupuncture)×100, where before acupuncture: before the study began, and after acupuncture: after 1 week and immediately before a new session began. Statistical analysis SAS software (V.9.1; SAS Institute, USA) was used for all statistical analyses. A paired t test was used to compare changes in body weight and changes in the levels of appetite-regulatory peptides both before the study and at 1 week after the start of acupuncture. A p value <0.05 was considered significant.

al analysis SAS software (V.9.1; SAS Institute, USA) was used for all statistical analyses. A paired t test was used to compare changes in body weight and changes in the levels of appetite-regulatory peptides both before the study and at 1 week after the start of acupuncture. A p value <0.05 was considered significant. Results Participant characteristics and general test results The median age of the participants in the acupuncture group (participants A–E) was 43 years (range, 40–44 years) and that of the participants in the placebo group (participants F–J) was 35 years (range, 24–43 years). The acupuncture group comprised four women and one man, and the placebo group comprised five women. The median body weight in the acupuncture group was 62.4 kg (range, 55.8–94.7 kg) and that in the placebo group was 60.2 kg (range, 53–82 kg; table 1). No abnormal haematology, blood biochemistry or lipid results were recorded for either group. Table 1 Physical findings Case Age Sex Height (cm) Weight (kg) BMI Waist circumference (cm) Blood pressure (mm Hg) Heart rate (bpm) Acupuncture group A 44 F 169 74.9 26.3 85 112/90 61 B 44 F 156 55.8 23 75.1 120/73 67 C 43 F 156 58.3 23.8 82 101/72 62 D 40 F 159 62.4 24.5 80.3 116/80 72 E 40 M 173 94.7 31.7 106.2 136/95 72 Placebo group F 24 F 172.5 60.2 20.3 79 113/81 72 G 34 F 161 54.4 20.8 68.2 110/73 84 H 43 F 159 53 21 76.1 95/57 66 I 43 F 157.8 65.2 26.4 86.2 117/71 85 J 35 F 170 82 28.4 97 117/59 76 BMI, body mass index.

3 67 C 43 F 156 58.3 23.8 82 101/72 62 D 40 F 159 62.4 24.5 80.3 116/80 72 E 40 M 173 94.7 31.7 106.2 136/95 72 Placebo group F 24 F 172.5 60.2 20.3 79 113/81 72 G 34 F 161 54.4 20.8 68.2 110/73 84 H 43 F 159 53 21 76.1 95/57 66 I 43 F 157.8 65.2 26.4 86.2 117/71 85 J 35 F 170 82 28.4 97 117/59 76 BMI, body mass index. Changes in body weight Figure 2 shows body weight before the study and after 1 week of treatment for all participants in the acupuncture group. Participant A lost 2.6 kg, which was the largest percentage change (−3.5%). The percentage changes in body weight for participants B, C, D, and E were −0.7%, −2.6%, −3%, and −1%, respectively. The difference in weight before treatment and after 1 week of treatment was significant for all participants in the acupuncture group (p=0.02). Figure 2 Percentage change in body weight at 1 week after the start of acupuncture treatment. Black columns denote the acupuncture group (A–E) and grey columns denote the placebo group (F–J). The percentage change in body weight was calculated as follows: ((value after stimulation−value before stimulation)/value before stimulation)×100 (%). A significant reduction in body weight was observed for all members of the acupuncture group (p=0.02). The percentage changes in body weight for participants in the placebo group over the same period were +1.3% (participant F), +0.3% (participant I) and 0% (participants G, H, and J). These changes were not significant (p=0.18).

Figure 2 Percentage change in body weight at 1 week after the start of acupuncture treatment. Black columns denote the acupuncture group (A–E) and grey columns denote the placebo group (F–J). The percentage change in body weight was calculated as follows: ((value after stimulation−value before stimulation)/value before stimulation)×100 (%). A significant reduction in body weight was observed for all members of the acupuncture group (p=0.02). The percentage changes in body weight for participants in the placebo group over the same period were +1.3% (participant F), +0.3% (participant I) and 0% (participants G, H, and J). These changes were not significant (p=0.18). Changes in appetite-regulatory peptide levels Tables 2 and 3 show the levels of biochemistry and appetite-regulatory peptides before treatment and at 1 week after the start of treatment. There were no clear differences in the fasting blood sugar, insulin, ACTH, adiponectin or leptin levels between the two groups. Figure 3 shows the percentage changes in active ghrelin levels 1 week after the start of acupuncture (with respect to baseline levels). The percentage changes in active ghrelin in the acupuncture group were −28% (participant A), +11% (participant B), +20% (participant C), −19% (participant D) and +2% (participant E), with no significant changes observed in active ghrelin levels at 1 week after acupuncture in any individual participant (p=0.89). In contrast, the percentage changes in active ghrelin levels in the placebo group at 1 week after the start of acupuncture were +434% (participant F), +14% (participant G), +149% (participant H), +64% (participant I) and +29% (participant J); all of these were significant (p=0.04). The leptin and adiponectin levels did not change significantly in either group. The dietician assessed reductions in food intake by estimating the nutritional value of each meal from photographs.

articipant G), +149% (participant H), +64% (participant I) and +29% (participant J); all of these were significant (p=0.04). The leptin and adiponectin levels did not change significantly in either group. The dietician assessed reductions in food intake by estimating the nutritional value of each meal from photographs. Table 2 Haematology and blood biochemistry

articipant G), +149% (participant H), +64% (participant I) and +29% (participant J); all of these were significant (p=0.04). The leptin and adiponectin levels did not change significantly in either group. The dietician assessed reductions in food intake by estimating the nutritional value of each meal from photographs. Table 2 Haematology and blood biochemistry Case Measurement time WCC (/μL) RBC (×104/μL) Hb (g/dL) Ht (%) MCV (fL) MCH (pg) MCHC (%) Platelet (μL) AST (IU/L) ALT (IU/L) TP (g/dL) BUN (mg/dL) HDL (mg/dL) LDL (mg/dL) FBS (mg/dL) Acupuncture group A Before 4800 387 11.4 35.9 93 29.5 31.8 22.3 14 12 7.7 18.1 51 125 98 1 week 4200 406 11.8 36.7 90 29.1 32.2 19.7 19 28 7.8 14.8 49 112 99 B Before 5400 449 13.6 42.2 94 30.3 32.2 18 27 31 6.4 13.6 55 34 92 1 week 4800 441 13.2 41.3 94 29.9 32 17 23 28 6.6 14.3 61 36 93 C Before 5300 411 11.5 38.3 93 28 30 37.2 21 9 7.6 13.1 71 115 96 1 week 5400 427 11.9 38.7 91 27.9 30.7 35.3 20 10 7.7 11.9 68 118 96 D Before 3200 428 13.2 41.4 97 30.8 31.9 22 22 18 7.1 12.2 51 94 81 1 week 3600 425 13.2 41.1 97 31.1 32.1 20.6 20 16 7 11.2 55 94 84 E Before 8300 488 14.7 45 92 30.1 32.7 28.7 25 32 7.7 9.5 62 118 98 1 week 7700 489 14.7 45.2 92 30.1 32.5 27.4 26 41 7.5 8.7 52 120 104 Placebo group F Before 4700 453 12.6 40.1 89 27.8 31.4 22.3 23 20 7.7 10.5 48 100 83 1 week 5600 469 13 41.5 88 27.7 31.3 20 18 15 7.8 11.5 55 107 85 G Before 5200 457 13.7 44.3 97 30 30.9 17.9 21 20 7.4 17.5 67 79 76 1 week 4400 458 13.7 44.2 97 29.9 31 16.9 19 21 7.6 12.8 71 75 80 H Before 5700 448 13.6 42 94 30.4 32.4 20.8 ND ND ND ND ND ND 89 1 week 5000 443 13.4 42.5 96 30.2 31.5 23.3 19 13 7.3 11 89 118 88 I Before 6600 436 8.8 31.7 73 20.2 27.8 46.4 13 15 7.6 13.1 46 146 90 1 week 6300 423 8.3 29.9 71 19.6 27.8 56 15 17 7.8 13.4 46 121 94 J Before 4500 455 13.9 41.6 91 30.5 33.4 18.5 17 12 7.3 13.9 43 130 ND 1 week 4900 433 13.6 40.2 93 31.4 33.8 19.3 15 13 7.1 11.7 45 124 88 ALT, alanine transaminase; AST, aspartate transaminase; BUN, Blood urea nitrogen; FBS, fasting blood sugar; Hb, haemoglobin; HDL, high-density lipoprotein; Ht, haematocrit; LDL, low-density lipoprotein; MCH, mean corpuscular haemoglobin; MCHC, MCH concentration; MCV, mean corpuscular volume; ND, not done because of insufficient blood samples; RBC, red blood cell; TP, total protein; WCC, white cell count.

fasting blood sugar; Hb, haemoglobin; HDL, high-density lipoprotein; Ht, haematocrit; LDL, low-density lipoprotein; MCH, mean corpuscular haemoglobin; MCHC, MCH concentration; MCV, mean corpuscular volume; ND, not done because of insufficient blood samples; RBC, red blood cell; TP, total protein; WCC, white cell count. Table 3 Changes in feeding-related peptides Case Measurement time Insulin (μU/mL) C-peptide (ng/mL) ACTH (pg/mL) Leptin (ng/mL) Active ghrelin (fmol/mL) Desacyl-ghrelin (fmol/mL) Adiponectin (μg/mL) Acupuncture group A Before 5.9 1.1 13 13.2 4.65 83.8 8.57 1 week 5.1 1 12 9.8 3.36 40.4 7.33 B Before 2.8 1 22 8 15.3 172 5.08 1 week 5.4 1.4 18 8.7 17.1 174 5.15 C Before 3.6 1 13 5.9 20.7 206 7.64 1 week 6.8 1.4 21 6.3 24.9 164 6.59 D Before 9.5 2.5 27 11.2 3.36 77.2 4.2 1 week 10.6 2.2 25 10.6 3.42 46.6 4.51 E Before 3 1.2 42 11.5 27.9 ND 13.6 1 week 3.7 1.3 47 12 22.5 387 11.4 Placebo group F Before 6.8 2.7 17 11.8 6.14 175 8.56 1 week 14.5 2.4 23 12.9 32.8 114 6.83 G Before 4.2 0.8 41 5.4 15.2 93 7.74 1 week 3.8 1.1 24 5 17.3 102 8.99 H Before ND ND 15 11.6 11.3 314 11.4 1 week 6.9 1.2 18 10.5 28.1 138 10.3 I Before 6.2 1.3 34 10.2 11 208 8.36 1 week 11.1 2 31 11.5 18 102 8.47 J Before 5.8 1.2 19 10.5 5.95 43.4 2.33 1 week 6.5 1.5 18 14.7 7.63 170 2.51 ACTH, adrenocorticotropic hormone; ND, not done because of insufficient blood samples.

.3 102 8.99 H Before ND ND 15 11.6 11.3 314 11.4 1 week 6.9 1.2 18 10.5 28.1 138 10.3 I Before 6.2 1.3 34 10.2 11 208 8.36 1 week 11.1 2 31 11.5 18 102 8.47 J Before 5.8 1.2 19 10.5 5.95 43.4 2.33 1 week 6.5 1.5 18 14.7 7.63 170 2.51 ACTH, adrenocorticotropic hormone; ND, not done because of insufficient blood samples. Figure 3 Percentage change in ghrelin levels at 1 week after the start of acupuncture. Black columns denote the acupuncture group (A–E) and grey columns denote the placebo group (F–J). The percentage change in ghrelin levels was calculated as follows: ((value after stimulation−value before stimulation)/value before stimulation)×100 (%). No member of the acupuncture group showed a significant increase in active ghrelin levels (p=0.89). Adverse effects No major adverse effects were observed during the study, and none of the participants withdrew from the study due to discomfort associated with the treatment. Discussion Auricular acupuncture therapy has been practised in China since ancient times; however, it has become internationally recognised due to the work of the French physician Paul Nogier.11 Acupuncture is widely used to treat many symptoms/conditions, and auricular acupuncture is the method most often used to treat obesity.12

uricular acupuncture therapy has been practised in China since ancient times; however, it has become internationally recognised due to the work of the French physician Paul Nogier.11 Acupuncture is widely used to treat many symptoms/conditions, and auricular acupuncture is the method most often used to treat obesity.12 The objective of this study was to identify the mechanism(s) by which auricular acupuncture therapy suppresses appetite in humans. To this end, we enrolled 10 volunteers: 5 had an indwelling intradermal needle inserted in the auricle and 5 had the needle taped to the auricle but not inserted. Body weight and the levels of several appetite-regulatory peptides were measured to examine treatment effects. Despite the small sample sizes, the results suggest that all participants in the auricular acupuncture group experienced weight loss during week 1 of the study. However, the sample was heterogeneous with respect to BMI. Therefore, future studies should examine the influence of acupuncture in groups of individuals of different body weight (ie, normal, overweight and obese individuals).

ipants in the auricular acupuncture group experienced weight loss during week 1 of the study. However, the sample was heterogeneous with respect to BMI. Therefore, future studies should examine the influence of acupuncture in groups of individuals of different body weight (ie, normal, overweight and obese individuals). The mechanism(s) underlying the weight loss observed after auricular acupuncture remains unclear, although the vagus nerve is thought to be involved.13 The acupuncture sites used in this study lie along the auricular branch of the vagus nerve (Arnold's nerve), which is its only somatosensory branch. Animal studies investigating the relationship between acupuncture and gastric peristalsis and feeding have been conducted.14 For example, Shiraishi et al investigated the effects of auricular acupuncture on neuronal activity in the feeding centre (lateral hypothalamic area: LHA) in a rat model of simple obesity, and on neuronal activity in the satiety centre (ventromedial nucleus of the hypothalamus: HVM) in a rat model of hypothalamic obesity (in which the satiety centre was disrupted). Using electrophysiological techniques, they found that auricular acupuncture stimulates the LHA via the auricular branch of the vagus nerve and inhibits the excitability of LHA neurons. They also reported evidence of signal transduction between the auricle and the hypothalamus, and noted that auricular acupuncture increased the activity of HVM neurons, thereby contributing to the induction or maintenance of satiety.15 In addition, Asamoto et al16 reported that auricular acupuncture increased HVM neuronal activity in rats, which led to a reduction in body weight. Therefore, auricular acupuncture may also suppress LHA neuronal activity via an afferent pathway of the vagus nerve and activate HVM neuronal activity in humans, which may lead to decreased food intake and subsequent weight loss.

puncture increased HVM neuronal activity in rats, which led to a reduction in body weight. Therefore, auricular acupuncture may also suppress LHA neuronal activity via an afferent pathway of the vagus nerve and activate HVM neuronal activity in humans, which may lead to decreased food intake and subsequent weight loss. In humans, the effects of auricular acupuncture often manifest as changes in body weight17 and subcutaneous fat levels,18 as assessed by modalities such as CT scan. In this study, we also measured the levels of appetite-regulatory peptides. We found that the increase in early morning fasting ghrelin levels in the acupuncture group was suppressed after 1 week of acupuncture, whereas all of the participants in the placebo group showed an increase in early morning fasting ghrelin levels. The serum ghrelin level change is affected by blood glucose and insulin levels.19 The changes in blood glucose and insulin levels were not significantly different between the acupuncture and placebo groups. The mechanism underlying this phenomenon may involve the interaction between feeding-related cytokines and the vagus nerve. Since the vagus nerve controls peristalsis, its nerve endings are distributed throughout the mucosa and submucosa of the gastrointestinal tract. The nerve also transmits neurochemical signals (triggered by gastrointestinal hormones) to the diencephalon and neocortex via the brainstem. Ghrelin is the only appetite-inducing peptide produced peripherally;20 the hormone is produced primarily by gastrointestinal endocrine cells located within the gastric body and it plays an important role in energy metabolism by increasing appetite and regulating gastrointestinal functions. Ghrelin transmits information about hunger and growth hormone levels to the nerve centre via afferent vagal pathways and through the blood.21 The concentration of ghrelin in the blood rises during fasting and decreases after eating. Ghrelin can pass through the blood–brain barrier and bind to ghrelin receptors in the arcuate nucleus of the hypothalamus.22 However, ghrelin receptors are also expressed at the vagal afferent terminals,23 and ghrelin excreted from the stomach transmits signals to the nerve centre via these receptors, which increases the appetite.24 Ghrelin is secreted when the stomach is empty, which promotes gastric peristalsis. The hormone then stimulates the brain via the circulation (it is present in the blood) and via vagal afferent pathways.

excreted from the stomach transmits signals to the nerve centre via these receptors, which increases the appetite.24 Ghrelin is secreted when the stomach is empty, which promotes gastric peristalsis. The hormone then stimulates the brain via the circulation (it is present in the blood) and via vagal afferent pathways. Electrical signals are then transmitted to the stomach via vagal efferent pathways, which again induce gastric peristalsis.

excreted from the stomach transmits signals to the nerve centre via these receptors, which increases the appetite.24 Ghrelin is secreted when the stomach is empty, which promotes gastric peristalsis. The hormone then stimulates the brain via the circulation (it is present in the blood) and via vagal afferent pathways. Electrical signals are then transmitted to the stomach via vagal efferent pathways, which again induce gastric peristalsis. Hsu et al25 conducted a randomised controlled trial of auricular acupuncture in 45 obese women. The acupuncture sites were the same as those used in this study and the results confirmed the biochemical effects of auricular acupuncture in humans. However, there was no clear difference between the acupuncture and placebo groups in terms of body weight before and after treatment. Furthermore, significant decreases in leptin levels and significant increases in ghrelin levels were observed in the acupuncture group. These results are different from those reported in this study. This difference may be attributable to the fact that we compared percentage changes in body weight and ghrelin levels in each individual participant before as well as after treatment, whereas Hsu et al compared the group values. Also, with regard to changes in body weight, Hsu et al compared values after 6 weeks and found no differences, whereas this study found that body weight decreased in the acupuncture group after 1 week of treatment; however, in three of the five participants, body weight returned to pretreatment levels after 4 weeks of treatment (which appeared to correspond with an increase in calorie intake after week 1). Therefore, it may be that the use of an indwelling needle led to an increase in the threshold level for acupuncture stimulation, leading to a reduction in the effect over time. In future, it may be necessary to conduct human studies using electroacupuncture therapy, as used in rats.26

e in calorie intake after week 1). Therefore, it may be that the use of an indwelling needle led to an increase in the threshold level for acupuncture stimulation, leading to a reduction in the effect over time. In future, it may be necessary to conduct human studies using electroacupuncture therapy, as used in rats.26 Conclusion The difference in weight before treatment and after 1 week of treatment was significant for all participants in the acupuncture group (p=0.02). The percentage changes in body weight in the placebo group were not significant (p=0.18). The percentage changes in active ghrelin in the acupuncture group were no significant changes observed in active ghrelin levels at 1 week after acupuncture in any individual participant (p=0.89), whereas all of the participants in the placebo group showed an increase in early morning fasting ghrelin levels (p=0.04). Thus, the sample number is small, but auricular acupuncture may reduce appetite by suppressing ghrelin production. The authors are grateful to the volunteers who participated in this study. They would also like to thank Yoshiko Fukushima, registered dietician, for analysing the content of the diets. Contributors: HI and OY designed the study, were responsible for the study concept and design and were involved in assessment, data entry and checking, statistical analysis and drafting of the manuscript. OY and TT extracted and analysed the data. OY offered support and advice at all stages of the study, and assisted with the revision of the manuscript. Competing interests: None. Patient consent: Obtained.

Contributors: HI and OY designed the study, were responsible for the study concept and design and were involved in assessment, data entry and checking, statistical analysis and drafting of the manuscript. OY and TT extracted and analysed the data. OY offered support and advice at all stages of the study, and assisted with the revision of the manuscript. Competing interests: None. Patient consent: Obtained. Ethics approval: The study protocol was approved by the Human Investigation Committee of the KOSAI Oriental Healthcare Center. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: The original data of this study can be accessed through OY via email.

Key messages What is already known about this subject ▸  Cancer of the pancreas and bile duct are highly lethal malignancies with short survival time ▸  Bile duct stenting for pancreatic and biliary malignancies are highly effective at relieving jaundice, but are not permanent ▸  Metal stents have a longer duration of efficacy than plastic stents but are much more expensive What are the new findings ▸  The presence of metastatic disease predicted a very short survival ▸  A Karnofsky score of <70 predicted a very short survival ▸  Use of a plastic or metal stent may be better selected based upon these easy to determine metrics How might it impact on clinical practice in the foreseeable future? ▸  Determination of the presence of metastatic disease is generally known to endoscopists at the time of evaluation of a patient with suspected pancreatic and biliary cancer. The Karnofsky score is very easy to determine. ▸  By using these factors at the time of stenting, one can possibly make a more cost effective determination of the best type of stent to use.

How might it impact on clinical practice in the foreseeable future? ▸  Determination of the presence of metastatic disease is generally known to endoscopists at the time of evaluation of a patient with suspected pancreatic and biliary cancer. The Karnofsky score is very easy to determine. ▸  By using these factors at the time of stenting, one can possibly make a more cost effective determination of the best type of stent to use. Introduction In patients with malignant obstructive jaundice, the endoscopists’ major decision centres on placement of a biliary plastic or self-expandable metal stent (SEMS). Current recommendations suggest that patients with an expected life expectancy of 6 months or less should undergo placement of a plastic stent.1 2 A number of studies have examined prognostic factors for survival in patients with pancreaticobiliary malignancy that include clinical, laboratory, radiographic, pathological and surgical variables.3–8 However, a majority of these factors have little relevance to the endoscopist at the time of decision-making. The use of simple criteria which are widely available, easily applied and predictable could be a major advancement to assist in choosing the appropriate biliary endoprosthesis. Prior studies suggest that a low Karnofsky score which is easy to calculate and the presence of liver metastases, generally known at the time of evaluation, have prognostic significance.9–16 We prospectively evaluated the utility of these two clinical tools in determining the most ideal endoprosthesis for biliary decompression.

dies suggest that a low Karnofsky score which is easy to calculate and the presence of liver metastases, generally known at the time of evaluation, have prognostic significance.9–16 We prospectively evaluated the utility of these two clinical tools in determining the most ideal endoprosthesis for biliary decompression. Methods Consecutive patients presenting with malignant obstructive jaundice and a mid or distal biliary stricture were included in this prospective observational cohort study beginning March 2008–June 2010 and with subsequent follow-up extending to 1 January 2014. Patients with hilar strictures were excluded. All patients signed informed consent both for the procedure and the study which was approved by our institutional review board (Protocol # X08020811). The diagnosis of cancer was generally confirmed by endoscopic ultrasound guided fine-needle aspiration (EUS-FNA) at the time of evaluation unless the patient had prior biopsy-proven malignancy. Prior to endoscopic retrograde cholangiopancreatography (ERCP), Karnofsky score was calculated (see online supplementary appendix 1) and radiographic reports were reviewed to both confirm the location of the tumour as well as evaluate for hepatic or extrahepatic metastases. All patients had previously undergone abdominal CT scanning, the majority (88%) with intravenous contrast. ERCP and biliary stenting was performed by two skilled endoscopists (CMW, SV).

iographic reports were reviewed to both confirm the location of the tumour as well as evaluate for hepatic or extrahepatic metastases. All patients had previously undergone abdominal CT scanning, the majority (88%) with intravenous contrast. ERCP and biliary stenting was performed by two skilled endoscopists (CMW, SV). Per protocol, patients with a Karnofsky score <80 and/or liver metastasis underwent placement of 10-French plastic stents (Cook Endoscopy, Winston Salem, North Carolina, USA) of variable length depending on the characteristic of the stricture. Patients with a Karnofsky score of 80 or more underwent placement of 10 mm uncovered SEMS (Boston Scientific, Natick, Massachusetts, USA) of variable length depending on the characteristic of the stricture. Sphincterotomy was not generally performed for stent placement (<5%). Patients were considered potential surgical candidates if they were referred by a surgeon who was planning Whipple resection or in whom at the time of initial evaluation comorbidity was limited, no vascular invasion was seen by EUS or CT, and metastases were absent; in these patients, a 10-French plastic stent was placed. These patients—termed preoperative candidates—are considered separately. Patients were excluded if a baseline CT scan was not performed prior to evaluation. Patients were also excluded if the stent type placed was not per protocol as some of the patients were specifically referred for a type of stent which may not have been according to protocol.

rative candidates—are considered separately. Patients were excluded if a baseline CT scan was not performed prior to evaluation. Patients were also excluded if the stent type placed was not per protocol as some of the patients were specifically referred for a type of stent which may not have been according to protocol. Following stent placement, patients were followed prospectively for a long-term until death or loss to follow-up. Patients were contacted every 4 months to determine if ERCP was repeated. All records for repeat procedures were reviewed. The use of chemotherapy and/or radiation therapy was documented. Main outcome measures Evaluating the utility of Karnofsky score and liver metastasis in predicting survival time, and determining the appropriate stent type for biliary decompression.

Following stent placement, patients were followed prospectively for a long-term until death or loss to follow-up. Patients were contacted every 4 months to determine if ERCP was repeated. All records for repeat procedures were reviewed. The use of chemotherapy and/or radiation therapy was documented. Main outcome measures Evaluating the utility of Karnofsky score and liver metastasis in predicting survival time, and determining the appropriate stent type for biliary decompression. Statistical analysis χ2 test and Fisher's exact test were used for each categorical variable to test their association among three groups. Since age for each group was normally distributed and had equal variance, one-way analysis of variance was conducted to test their mean difference. Kruskal-Wallis test was performed to assess whether one of three groups tends to have larger values than the others for Karnofsky score, which were not normally distributed. Cross-classifications were designed to identify potential interactions among potential predictors. The cumulative survival curve was depicted using the Kaplan-Meier method and patients alive at the last follow-up were censored. A priori planned analyses included evaluating factors independently associated with overall survival. This was performed by using univariate Cox proportional hazards regression analyses, and was conducted for the plastic and metal stent groups excluding the preoperative group. Analysis was performed using SAS software, V.9.3 (SAS Institute Inc, Cary, North Carolina, USA) and statistical difference was considered to be significant at the level of 0.05.

riate Cox proportional hazards regression analyses, and was conducted for the plastic and metal stent groups excluding the preoperative group. Analysis was performed using SAS software, V.9.3 (SAS Institute Inc, Cary, North Carolina, USA) and statistical difference was considered to be significant at the level of 0.05. Results Patient characteristics A total of 106 patients were evaluated, meeting the inclusion criteria of the study between March 2008 and June 2010 and followed until 1 January 2014. No patient considered a candidate for the study declined enrolment. At the time of analysis, 98 patients (92.5%) were dead and the overall survival of the cohort was poor with a median survival time of 8.7 months (IQR 2.4–19.2 months). Five patients were excluded as their physicians requested metallic stents though these patients were candidates for plastic stents, and three patients were excluded as only abdominal ultrasound was performed prior to ERCP as gallstone disease was suspected. The baseline characteristics of the final cohort of 98 patients and by the choice of stent are listed in table 1. Consistent with our patient population, most patients were older, Caucasian and had pancreatic adenocarcinoma. EUS-FNA performed before (60.2%) or after the ERCP confirmed a diagnosis of malignancy in all patients. A Whipple resection was performed in 21 patients (21.4%) and 19 (19.4%) received or had received chemoradiation. Twenty-four patients (preoperative group) considered surgical candidates underwent plastic stent placement preoperatively for biliary decompression and are considered separately. Median Karnofsky score in these 24 patients was 80 (range 80–100). The median survival of these 24 patients was 20.5 months (range 2–65 months) and 6 of these patients are alive.

oup) considered surgical candidates underwent plastic stent placement preoperatively for biliary decompression and are considered separately. Median Karnofsky score in these 24 patients was 80 (range 80–100). The median survival of these 24 patients was 20.5 months (range 2–65 months) and 6 of these patients are alive. Table 1 Baseline characteristics

oup) considered surgical candidates underwent plastic stent placement preoperatively for biliary decompression and are considered separately. Median Karnofsky score in these 24 patients was 80 (range 80–100). The median survival of these 24 patients was 20.5 months (range 2–65 months) and 6 of these patients are alive. Table 1 Baseline characteristics Total N=98 Metal stent N=31 (31.6%) Plastic stent N=43 (43.9%) Preoperative plastic stent N=24 (24.5%) p Value Age* 66.5±11.2 (34–88) 66.4±11.3 (34–87) 66.8±11.5 (45–88) 66.0±11.1 (46–84) 0.9613 Gender, N (%) 0.2062 Female 37 (37.8) 15 (48.4) 16 (37.2) 6 (25) Male 61 (62.2) 16 (51.6) 27 (62.8) 18 (75) Race, N (%) 0.2700 Caucasian 74 (75.5) 25 (80.7) 29 (67.4) 20 (83.3) African-American 21 (21.4) 6 (19.4) 11 (26.2) 4 (16.7) Other 3 (3.1) – 3 (7.1) – Diagnosis, N (%) <0.0001† Pancreatic cancer 57 (56.4) 26 (83.9) 16 (37.2) 13 (54.2) Cholangiocarcinoma 17 (16.8) 1 (3.2) 8 (18.6) 8 (33.3) Other  27 (26.7)‡ 4 (12.9) 19 (44.2) 3 (12.5) Chemoradiation, N (%) 0.0160† Prior 8 (8.2) 1 (3.2) 7 (16.3) – Current 11 (11.2) 4 (12.9) 7 (16.3) – Restent, N(%) 27 (27.6) 6 (19.4) 18 (41.9) 3 (12.5) 0.0168† Number of restenting session, N(%) 0.1639 1 21 (21.4) 6 (19.4) 12 (27.9) 3 (12.5) 2 3 (3.1) – 3 (7.0) – ≥3 3 (3.1) – 3 (7.0) – Karnofsky score§ 80 (20–100) 80 (80–100) 65 (20–80) 80 (80–100) <0.0001† Liver metastases, N (%) 16 (16.3) – 16 (37.2) – <0.0001† Other metastases, N (%) 18 (18.4) 3 (9.7) 15 (35.7) – 0.0019† Whipple resection, N (%) 21 (21.4) 5 (16.1) – 16 (66.7) <0.0001† Death, N (%) 90 (91.8) 29 (93.6) 43 (100) 18 (75) 0.0015† *Mean±SD (range).

ore§ 80 (20–100) 80 (80–100) 65 (20–80) 80 (80–100) <0.0001† Liver metastases, N (%) 16 (16.3) – 16 (37.2) – <0.0001† Other metastases, N (%) 18 (18.4) 3 (9.7) 15 (35.7) – 0.0019† Whipple resection, N (%) 21 (21.4) 5 (16.1) – 16 (66.7) <0.0001† Death, N (%) 90 (91.8) 29 (93.6) 43 (100) 18 (75) 0.0015† *Mean±SD (range). †Significant result at the 0.05 level of significance. ‡Metastatic lung cancer (5), metastatic colon cancer (4), metastatic breast cancer (3), gallbladder cancer (3) lymphoma (4), metastatic cancer to lymph nodes of unknown source (8). §Median (range).

ore§ 80 (20–100) 80 (80–100) 65 (20–80) 80 (80–100) <0.0001† Liver metastases, N (%) 16 (16.3) – 16 (37.2) – <0.0001† Other metastases, N (%) 18 (18.4) 3 (9.7) 15 (35.7) – 0.0019† Whipple resection, N (%) 21 (21.4) 5 (16.1) – 16 (66.7) <0.0001† Death, N (%) 90 (91.8) 29 (93.6) 43 (100) 18 (75) 0.0015† *Mean±SD (range). †Significant result at the 0.05 level of significance. ‡Metastatic lung cancer (5), metastatic colon cancer (4), metastatic breast cancer (3), gallbladder cancer (3) lymphoma (4), metastatic cancer to lymph nodes of unknown source (8). §Median (range). Survival time The overall mean follow-up time for the cohort of 98 patients was 13.6 months (SEMS, range 0.5–55.5 months; plastic stents, range 0.3–34 months; plastic stents preoperatively, range 2–65.2 months). As shown in figure 1, long-term survival among the three groups was significantly different (log rank χ2=36, p<0.0001). In figure 2, a Karnofsky score <80 was predictive of a shortened survival (median 3.1 months) compared with those with a score ≥80 (median 13.9 months; log rank χ2=28.7, p<0.0001). Likewise, patients with liver metastases (figure 3) had a shorter survival (median 1.8 months) compared with those without liver metastases (median 11.6 months; log rank χ2=17.1, p<0.0001). Metastases to organs other than the liver were noted in 18 patients in whom 3 (all with a Karnofsky score of ≥80) underwent SEMS (median survival 2.1 months) and 15 plastic stenting (median survival 3.3 months); the median survival for these 18 patients was 2.7 months (range 0.5–21.7). Patients who had a Whipple resection had a significantly longer survival (median 21.9 months) compared with those who did not have the surgery (median 5.3 months; log rank χ2=19.5, p<0.0001). Patients who did not receive chemoradiation had a prolonged survival compared with patients who received treatment (log rank χ2=12.9, p=0.0003). Only two patients in the entire cohort were lost to follow-up; one underwent Whipple resection (follow-up till 49 months), and the other a plastic stent (follow-up till 36 months).

nts who did not receive chemoradiation had a prolonged survival compared with patients who received treatment (log rank χ2=12.9, p=0.0003). Only two patients in the entire cohort were lost to follow-up; one underwent Whipple resection (follow-up till 49 months), and the other a plastic stent (follow-up till 36 months). Figure 1 Long-term survival based on study groups (metal: n=31, median survival 11.6 months; plastic: n=43, median survival 2.8 months; preOP: n=24, median survival 20.5 months; log rank χ2=36.0, p<0.0001; preOP, preoperative). Figure 2 Long-term survival based on Karnofsky score (K score ≥80: n=57, median survival 13.9 months; K score <80: n=41, median survival 3.1 months; log rank χ2=28.7, p<0.0001). Figure 3 Long-term survival based on presence of liver metastases (absence: n=82, median survival 11.6 months; presence: n=16, median survival 1.8 months; log rank χ2=17.1, p<0.0001). Of the 24 patients undergoing plastic stent placement ‘preoperatively’, 16 of these patients (66.7%) ultimately underwent Whipple resection. Of the remaining eight patients, the median survival was 19.9 months (range 2–37.6).

Figure 3 Long-term survival based on presence of liver metastases (absence: n=82, median survival 11.6 months; presence: n=16, median survival 1.8 months; log rank χ2=17.1, p<0.0001). Of the 24 patients undergoing plastic stent placement ‘preoperatively’, 16 of these patients (66.7%) ultimately underwent Whipple resection. Of the remaining eight patients, the median survival was 19.9 months (range 2–37.6). Stent patency Of these 98 patients, 67 underwent plastic stent placement for a low Karnofsky score and/or liver metastases, and 31 SEMS placements. The overall reintervention rate for these three groups was 27.5% (n=27) and 41.9% (n=18) for the plastic stent group; 19.4% (n=6) for the SEMS cohorts; and 12.5% (n=3) for the preoperative plastic stent group (table 1). No patient with pancreatic cancer and liver metastases underwent restenting. No patient with a Karnofsky score <50 underwent restenting. The majority of patients with liver metastasis in whom plastic stents were placed had non-pancreatic malignancies and metastasis from other organs, such as colon or lung. In all but three patients, the patients for restenting presented with recurrent jaundice or abnormal liver tests. No patient died from stent occlusion.

e majority of patients with liver metastasis in whom plastic stents were placed had non-pancreatic malignancies and metastasis from other organs, such as colon or lung. In all but three patients, the patients for restenting presented with recurrent jaundice or abnormal liver tests. No patient died from stent occlusion. Univariate analysis In univariate Cox proportional hazards regression analyses, the patient's age, dichotomised Karnofsky score (≥80 vs <80), presence of liver metastases, stent type, chemotherapy or radiation therapy, Whipple resection, reintervention, and number of restent were tested to evaluate an association with survival. The univariate analysis showed that patients with higher Karnofsky score or with metal stent placed or who underwent Whipple resection were associated with longer survival, but presence of liver metastases and history of chemoradiation therapy were negatively related with long-term survival (table 2). Table 2 Univariate analysis of prognostic factors for overall survival in metal and plastic stent group Predictor HR 95% CI for HR p Value Age 1.01 (0.99 to 1.04) 0.1671 Karnofsky score (≥80 vs <80) 0.40 (0.25 to 0.67) 0.0004* Liver metastases (yes vs no) 2.36 (1.32 to 4.22) 0.0037* Stent type (metal vs plastic) 0.36 (0.22 to 0.60) <0.0001* Whipple resection (yes vs no) 0.19 (0.06 to 0.61) 0.0053* Chemoradiation (yes vs no) 1.89 (1.10 to 3.24) 0.0221* Restent (yes vs no) 1.08 (0.65 to 1.78) 0.7640 Number of restent 1.06 (0.80 to 1.42) 0.6787 *Significant result at the 0.05 level of significance.

* Stent type (metal vs plastic) 0.36 (0.22 to 0.60) <0.0001* Whipple resection (yes vs no) 0.19 (0.06 to 0.61) 0.0053* Chemoradiation (yes vs no) 1.89 (1.10 to 3.24) 0.0221* Restent (yes vs no) 1.08 (0.65 to 1.78) 0.7640 Number of restent 1.06 (0.80 to 1.42) 0.6787 *Significant result at the 0.05 level of significance. Discussion The choice of a plastic stent or SEMS for biliary decompression in malignant obstructive jaundice is predominantly based on the endoscopist's preference. We have shown that by using simple clinical tools, a potentially better choice can be made regarding which type of endoprosthesis to deploy. For those in whom longer survival was predicted and a SEMS placed, reintervention was required in only six patients (19.4%). These findings are consistent with the rate of reinterventions in selected groups reported in prior studies in patients with pancreaticobiliary malignancy.17 18

type of endoprosthesis to deploy. For those in whom longer survival was predicted and a SEMS placed, reintervention was required in only six patients (19.4%). These findings are consistent with the rate of reinterventions in selected groups reported in prior studies in patients with pancreaticobiliary malignancy.17 18 To the best of our knowledge, there are no studies that specifically address the determination of which endoprosthesis to use at the time of initial evaluation. Many prior studies have demonstrated the longer patency rates of SEMS as compared with plastic stents. Generally, metal stents remain patent for a mean of 8–12 months compared with 3–6 months for large plastic stents.17 19 20 While the poor survival of patients with pancreaticobiliary malignancy is well recognised, there are some patients who have had a more prolonged survival and some patients who undergo Whipple resection. Given the cost of SEMS, appropriate use in those in whom surgery would not be likely or when an extended survival is predicted is appropriate. The use of plastic stents has been recommended for use in those with distant metastasis as well.17 In our cohort of 18 patients with metastases to organs other than the liver, median survival was poor. Three of these patients underwent metal stent placement because of a high Karnofsky score and the range of survival for these three patients was 0.9–11.6 months.