Browse the corpus

k between this gene and predisposition to IPF. That the p.A279T mutation was present in IPF and COPD cohorts with similar frequencies makes it unlikely that it contributes directly to the pathogenesis of IPF. Complex functional assays will be necessary to identify any mechanisms whereby these mutations exert an effect. MUC5B Since the rs35705950 promoter polymorphism is over-represented in individuals with IPF, we resequenced all but the largest exon of MUC5B to determine if mutations in the coding sequence of the gene might also associate with IPF. Although we found individual rare deleterious variants in IPF and COPD cohorts, the similar collapsed frequencies in both cohorts makes it unlikely that mutations in coding regions of this gene contribute appreciably to the risk of IPF. The MAF of the MUC5B promoter polymorphism was approximately 30% in our IPF cohort which is similar to previous reports.12 38 Since this frequency is similar in patients with IPF with and without other candidate gene mutations, this polymorphism is not associated with increased risk for disease in association with the exonic mutations in ABCA3, SFTPA2, SFTPC or TERT.

imate the genetic contribution of a gene to IPF, further investigation of genetic susceptibility to IPF using exome sequencing of affected and unaffected family members could elucidate not only the relationship between childhood ILD and adult IPF, but also the unidentified genetic factors leading to disease expression. The authors would like to thank Michael J Holtzman, MD, for access to the DNA archive for COPD, supported by a SCCOR in Chronic Obstructive Pulmonary Disease, HL084922. The authors would also like to thank Jessica Hoisington-Lopez from the Center for Genome Sciences and Systems Biology for her sequencing expertise, and the NHLBI GO Exome Sequencing Project and its ongoing studies which produced and provided exome variant calls for comparison: the Lung GO Sequencing Project (HL-102923), the WHI Sequencing Project (HL-102924), the Broad GO Sequencing Project (HL-102925), the Seattle GO Sequencing Project (HL-102926) and the Heart GO Sequencing Project (HL-103010). Study data were collected and managed using REDCap (Research Electronic Data Capture), hosted at Washington University.

Key messages Rare mutations in SFTPA2, SFTPC, (surfactant proteins) and TERT (telomerase) were found in individuals with idiopathic pulmonary fibrosis (IPF) but not those with chronic obstructive pulmonary disease. Mutations associated with childhood interstitial lung disease are present in adults with IPF supporting that these mutations predispose to a spectrum of interstitial lung diseases manifesting at all ages. Genetic analysis and counselling should be considered as part of the IPF evaluation. Introduction Idiopathic pulmonary fibrosis (IPF) is the most common chronic interstitial lung disease (ILD) in adults.1 With an average presenting age of 60–70 years, the clinical course is one of progressive decline in lung function or precipitous deterioration following an acute exacerbation. With a median survival of only 2–3 years after diagnosis, it is one of the most common indications for lung transplantation in adults.1 2 Despite an association with smoking and environmental exposures, the pathogenesis of this disease remains elusive.1 Familial IPF, defined by disease in two or more first-degree relatives, is estimated to occur in 2–20% of all IPF cases and can present at earlier ages.1 3

dications for lung transplantation in adults.1 2 Despite an association with smoking and environmental exposures, the pathogenesis of this disease remains elusive.1 Familial IPF, defined by disease in two or more first-degree relatives, is estimated to occur in 2–20% of all IPF cases and can present at earlier ages.1 3 Previous studies investigating genetic predisposition in sporadic and familial pulmonary fibrosis have identified rare mutations in the gene encoding telomerase (TERT, Gene ID 7015) and in the surfactant-associated genes, surfactant protein A2 (SFTPA2, Gene ID 729238), surfactant protein C (SFTPC, Gene ID 6440) and the ATP-binding cassette member A3 (ABCA3, Gene ID 21).4–11 Additionally, a polymorphism (rs35705950) in the mucin 5B gene (MUC5B, Gene ID 727897) promoter is significantly more prevalent in individuals with both sporadic and familial IPF.12 13 Mutations in SFTPC, ABCA3, and the thyroid transcription factor gene (NKX2-1, Gene ID 7080) are associated with childhood ILD.9 14–19 In studies of individuals and families with mutations in SFTPC, disease penetrance varies in severity, age of onset, histopathology and clinical characteristics suggesting that childhood ILD and IPF in adulthood may be part of the same spectrum of disease.9 10 18–21

ne ID 7080) are associated with childhood ILD.9 14–19 In studies of individuals and families with mutations in SFTPC, disease penetrance varies in severity, age of onset, histopathology and clinical characteristics suggesting that childhood ILD and IPF in adulthood may be part of the same spectrum of disease.9 10 18–21 Since no previous studies investigated all of these candidate genes simultaneously in IPF, we resequenced these genes in a cohort of patients with IPF to test the hypothesis that mutations in the coding regions of these six genes occur at a higher frequency in IPF than in chronic obstructive pulmonary disease (COPD) or in population-based cohorts, and to determine if interactions among these genes and the MUC5B promoter polymorphism modified disease phenotype.

th IPF to test the hypothesis that mutations in the coding regions of these six genes occur at a higher frequency in IPF than in chronic obstructive pulmonary disease (COPD) or in population-based cohorts, and to determine if interactions among these genes and the MUC5B promoter polymorphism modified disease phenotype. Methods Subject selection IPF cohort Individuals with IPF were included if they had a usual interstitial pneumonia (UIP) pattern on lung explant or biopsy, or had a clinical diagnosis of IPF validated by the research team using American Thoracic Society/European Respiratory Society 2011 criteria.1 Individuals with non-IPF/UIP idiopathic interstitial pneumonia, hypersensitivity pneumonitis, occupational lung disease, drug-induced lung disease and connective tissue disorders were excluded. Recruitment included all individuals with IPF regardless of race, gender, or age presenting from 2009 to 2012. This cohort included archived DNA samples acquired through a waiver of consent from patients in the Washington University/Barnes-Jewish Hospital Adult Lung Transplant Programme who underwent lung transplantation for IPF (n=105) and prospectively recruited individuals followed in the Division of Pulmonary and Critical Care Medicine (n=27) for a total n=132, which was a sample size of convenience. Clinical data, including demographic information, age at onset of symptoms and diagnosis, family history, smoking history, exposure history and results from radiological studies and lung biopsies, were acquired through medical record review.

are Medicine (n=27) for a total n=132, which was a sample size of convenience. Clinical data, including demographic information, age at onset of symptoms and diagnosis, family history, smoking history, exposure history and results from radiological studies and lung biopsies, were acquired through medical record review. COPD cohort Archived DNA samples from individuals with COPD through the COPD SCCOR Project at Washington University/Barnes-Jewish Hospital (5 P50 HL084922) from patients recruited from 2005 to 2011 were used as a disease-based control. Through permission for sample sharing, we obtained the DNA and linked de-identified clinical information including age, gender, race and smoking history from existing databases through a waiver of consent. The clinical characteristics of each cohort are summarised in table 1. Table 1 Characteristics of the IPF and COPD cohorts IPF (n=132) COPD (n=192) p Value (IPF vs COPD) Age 54.6±9.2 60.2±8.1 0.5 Gender Female 34 (26%) 93 (48%) Male 98 (74%) 99 (52%) <0.001 Race Caucasian 119 (90%) 179 (93%) 0.43 African–American 6 (5%) 10 (5%) Asian 4 (3%) 0 (0%) Hispanic 3 (2%) 2 (1%) Multiple races 0 (0%) 1 (0.5%) Tobacco use Yes 84 (64%) 189 (98%) No 47 (36%) 2 (1%) <0.001 Unknown 1 (1%) 1 (0.5%) COPD, chronic obstructive pulmonary disease; IPF, idiopathic pulmonary fibrosis. Exome sequencing project database Data from the NHLBI Exome Sequencing Project (ESP, version ESP6500SI-V2, http://evs.gs.washington.edu/EVS/) were used to provide race-stratified, population-based frequencies of variants in candidate genes (accessed 1 July 2013).

IPF (n=132) COPD (n=192) p Value (IPF vs COPD) Age 54.6±9.2 60.2±8.1 0.5 Gender Female 34 (26%) 93 (48%) Male 98 (74%) 99 (52%) <0.001 Race Caucasian 119 (90%) 179 (93%) 0.43 African–American 6 (5%) 10 (5%) Asian 4 (3%) 0 (0%) Hispanic 3 (2%) 2 (1%) Multiple races 0 (0%) 1 (0.5%) Tobacco use Yes 84 (64%) 189 (98%) No 47 (36%) 2 (1%) <0.001 Unknown 1 (1%) 1 (0.5%) COPD, chronic obstructive pulmonary disease; IPF, idiopathic pulmonary fibrosis. Exome sequencing project database Data from the NHLBI Exome Sequencing Project (ESP, version ESP6500SI-V2, http://evs.gs.washington.edu/EVS/) were used to provide race-stratified, population-based frequencies of variants in candidate genes (accessed 1 July 2013). Gene selection Based on our hypothesis that childhood ILD and IPF may be part of a spectrum of disease, we selected genes that were known at the time we designed the study between associate with IPF only (SFTPA2 and TERT) and genes associated with both childhood ILD and IPF (ABCA3, SFTPC and NKX2-1). Mutations in ABCA3 are expressed in a recessive manner, while mutations in SFTPA2, SFTPC, NKX2-1 and TERT are dominantly expressed. The presence of either one or two copies of the MUC5B promoter variant is associated with IPF, however, exonic mutations in MUC5B have not been studied. Therefore, we also interrogated MUC5B, though the mode of inheritance is unknown. Although mutations in the telomerase RNA component gene (TERC, Gene ID 7012) are also associated with IPF, the mutations occur in the part of the gene that encodes the RNA component and we therefore could not apply our definition of ‘mutation’ based on alteration in protein function.6 22 23

e of inheritance is unknown. Although mutations in the telomerase RNA component gene (TERC, Gene ID 7012) are also associated with IPF, the mutations occur in the part of the gene that encodes the RNA component and we therefore could not apply our definition of ‘mutation’ based on alteration in protein function.6 22 23 Sample preparation and sequencing We performed DNA purification on preserved saliva samples obtained from the patients with IPF using Oragene DNA self-collection kits and purification protocol (DNA Genotek Inc, Kanata, Ontario, Canada). The DNA from each specimen was quantified using the Qubit Fluorometer and equimolar amounts of each patient's DNA were pooled. Using this pooled DNA, all exons of ABCA3, MUC5B (except exon 49 due to its large size (∼11 kb) and the presence of significant repetitive sequence that prevented standard sequencing approach), NKX2-1, SFTPA2, SFTPC and TERT were amplified, approximately 45 kb of sequence per individual. Equal amounts of amplified product were pooled for next generation sequencing using the Illumina Genome Analyzer/MiSeq platform. Negative and positive controls were inserted into the pooled sample to optimise detection of rare variants with high specificity and sensitivity.24 25

mately 45 kb of sequence per individual. Equal amounts of amplified product were pooled for next generation sequencing using the Illumina Genome Analyzer/MiSeq platform. Negative and positive controls were inserted into the pooled sample to optimise detection of rare variants with high specificity and sensitivity.24 25 We used SPLINTER (Short IN/DEL Prediction by Large deviation Inference and Non-linear True frequency Estimation by Recursion) to identify rare single nucleotide variants and small insertions/deletions.25 We defined a mutation as a small insertion or deletion or a non-synonymous single nucleotide variant previously identified in children or adults with respiratory disease and/or predicted to disrupt encoded protein function in at least 2 of 3 prediction algorithms: SIFT, PolyPhen2, or Mutation Taster.26–28 Since IPF is a rare disease, we excluded variants with a prevalence of >1% in the general population as catalogued in the NHLBI Exome Sequencing Project (ESP, version ESP6500SI-V2, http://evs.gs.washington.edu/EVS/ (accessed July 1, 2013)). We then validated using an independent genotyping strategy through Sanger resequencing or Taqman Genotyping Assays. Taqman genotyping was also used for the MUC5B promoter polymorphism. The validation of variants allowed us to confirm the frequency of each variant and to link it to an individual.

ccessed July 1, 2013)). We then validated using an independent genotyping strategy through Sanger resequencing or Taqman Genotyping Assays. Taqman genotyping was also used for the MUC5B promoter polymorphism. The validation of variants allowed us to confirm the frequency of each variant and to link it to an individual. Data analysis Study data were collected and managed using REDCap electronic data capture tools hosted at Washington University School of Medicine's Institute for Clinical and Translational Studies.29 Since it is unlikely that an individual carries more than one rare mutation at a single gene locus, the number of mutations in a single gene can be collapsed for statistical purposes and compared using a univariate test.30 We used Fisher's Exact tests to compare frequencies of individual variants and collapsed frequencies across a gene, and to test the difference of other categorical features (gender, race, tobacco use). We used t tests to test the difference in age of onset between mutated and non-mutated IPF samples.

proximately 30% in our IPF cohort which is similar to previous reports.12 38 Since this frequency is similar in patients with IPF with and without other candidate gene mutations, this polymorphism is not associated with increased risk for disease in association with the exonic mutations in ABCA3, SFTPA2, SFTPC or TERT. Conclusion The presence of childhood ILD-associated gene mutations in adults with IPF supports the hypothesis that these mutations predispose to a spectrum of fibrotic lung diseases manifesting from infancy to adulthood. Therefore, it is possible that childhood ILD and adult IPF represent variable expression of a common underlying gene-based pathogenic mechanism. While these mutations increase disease risk, the environmental modifiers that cause injury, disrupt repair and ultimately result in disease remain to be identified, along with the full complement of gene mutations predisposing to fibrotic lung disease. While possible that exome sequencing may only capture coding region variants and disregard non-coding elements of genome leading to an underestimate the genetic contribution of a gene to IPF, further investigation of genetic susceptibility to IPF using exome sequencing of affected and unaffected family members could elucidate not only the relationship between childhood ILD and adult IPF, but also the unidentified genetic factors leading to disease expression.

univariate test.30 We used Fisher's Exact tests to compare frequencies of individual variants and collapsed frequencies across a gene, and to test the difference of other categorical features (gender, race, tobacco use). We used t tests to test the difference in age of onset between mutated and non-mutated IPF samples. Results Mutations in IPF-associated and childhood ILD-associated genes In the IPF population, we found 15 mutations in 14 different individuals in all the genes interrogated except NKX2-1. There were no differences in age of onset, sex or tobacco use in the IPF individuals with or without a mutation (table 2). No individuals had single rare mutations in two different genes, although one individual was homozygous for ABCA3 p.E292V. There was one mutation found in SFTPA2 in the IPF cohort and none in the COPD cohort but the collapsed frequency was not statistically significant (table 3). Table 2 Characteristics of IPF individuals with and without mutations IPF Mutation (n=14) Number of mutation (n=118) p Value Age 49.6±14 55.2±8.4 0.51 Gender Female 3 (21%) 31 (26%) 1.0 Male 11 (79%) 87 (74%) Tobacco use Yes 8 (57%) 76 (64%) 0.57 No 6 (43%) 41 (35%) Unknown 0 (0%) 1 (1%) Outcome Alive w/o lung transplant 3 (21%) 20 (17%) Lung transplant* 11 (79%) 95 (81%) Died w/o lung transplant 0 (0%) 3 (3%) *Individuals who had a transplant and then died are included in the ‘transplant’ numbers. IPF, idiopathic pulmonary fibrosis. Table 3 Mutations identified in IPF and COPD cohorts of European descent

IPF Mutation (n=14) Number of mutation (n=118) p Value Age 49.6±14 55.2±8.4 0.51 Gender Female 3 (21%) 31 (26%) 1.0 Male 11 (79%) 87 (74%) Tobacco use Yes 8 (57%) 76 (64%) 0.57 No 6 (43%) 41 (35%) Unknown 0 (0%) 1 (1%) Outcome Alive w/o lung transplant 3 (21%) 20 (17%) Lung transplant* 11 (79%) 95 (81%) Died w/o lung transplant 0 (0%) 3 (3%) *Individuals who had a transplant and then died are included in the ‘transplant’ numbers. IPF, idiopathic pulmonary fibrosis. Table 3 Mutations identified in IPF and COPD cohorts of European descent Gene Mutation Amino acid change IPF* (ED) (n=119) (n (MAF)) COPD (n=178) (n (MAF)) ESP (ED) (MAF), % p Value (IPF vs COPD) dbSNP number SFTPA2 c.532G>A V178M 1 (0.4%)† 0 0.01 0.4 SFTPC c.218T>C I73T 3 (1.3%)† 0 0 rs121917834 c.329T>G L110R 1 (0.4%) 0 0 c.334G>A A112T 1 (0.4%) 0 0 Collapsed frequency 2.1% 0.01 ABCA3 c.875A>T E292V 4 (1.68%) 3 (0.84%) 0.45 rs149989682 c.4420G>A R1474W 0 3 (0.84%) 0.36 rs146709251 Collapsed frequency 1.68% 1.68% 1.0 NKX2-1 0 0 NA TERT c.323G>C ‡R108P 1 (0.42%) 0 0 c.994C>T ‡L332F 1 (0.42%) 0 0 c.1775A>G ‡H592R 1 (0.42%)† 0 0 c.2110C>T P704S 2 (0.84%) 0 0 rs199422297 Collapsed frequency 2.1% 0.01 *All individuals with mutations were of European descent. †One individual each with a family history. ‡Novel mutations. ED, European descent; MAF, minor allele frequency.

NKX2-1 0 0 NA TERT c.323G>C ‡R108P 1 (0.42%) 0 0 c.994C>T ‡L332F 1 (0.42%) 0 0 c.1775A>G ‡H592R 1 (0.42%)† 0 0 c.2110C>T P704S 2 (0.84%) 0 0 rs199422297 Collapsed frequency 2.1% 0.01 *All individuals with mutations were of European descent. †One individual each with a family history. ‡Novel mutations. ED, European descent; MAF, minor allele frequency. All mutations in IPF and COPD cohorts were found only in individuals of European descent, so the analyses in table 3 focused on this subset. Mutations in SFTPC were found only in the IPF cohort and included the common disease-causing allele p.I73T as well as two other mutations (p.L110R and p.A112T) previously identified in children with ILD but not in adults with IPF (table 3, unpublished data A Hamvas and L M Nogee, 2014).15 None of the SFTPC mutations was present in individuals with COPD or in the ESP database. The collapsed frequency of SFTPC mutations in the IPF cohort was statistically significant when compared to the COPD cohort. Rare mutations in TERT were also present in the IPF cohort and not found in COPD. The collapsed frequency of TERT mutations in the IPF cohort was statistically greater than that of the COPD cohort (table 3).

All mutations in IPF and COPD cohorts were found only in individuals of European descent, so the analyses in table 3 focused on this subset. Mutations in SFTPC were found only in the IPF cohort and included the common disease-causing allele p.I73T as well as two other mutations (p.L110R and p.A112T) previously identified in children with ILD but not in adults with IPF (table 3, unpublished data A Hamvas and L M Nogee, 2014).15 None of the SFTPC mutations was present in individuals with COPD or in the ESP database. The collapsed frequency of SFTPC mutations in the IPF cohort was statistically significant when compared to the COPD cohort. Rare mutations in TERT were also present in the IPF cohort and not found in COPD. The collapsed frequency of TERT mutations in the IPF cohort was statistically greater than that of the COPD cohort (table 3). To determine if the individuals in the IPF and COPD cohorts were representative of the general population, we also examined the frequencies of common variants in these genes. Although none of these common variants are predicted to be deleterious, we cannot exclude the possibility that some of these variants could be in linkage with our functional mutations, but our sequencing methodology did not permit identifying on which allele these variants resided. The common non-synonymous variants in SFTPC (p.T138N and p.S186N) and SFTPA2 (p.A91P) were present in similar frequencies in the IPF and COPD cohorts compared to the general population, as were the common synonymous variants in ABCA3 (p.F353F, p.P585P and p.S1372S) (see online supplementary table S1). The frequency of SFTPA2 p.Q223K was less than the ESP frequency, which could be due to sample size. The frequencies of the common TERT variant p.A279T, previously associated with aplastic anaemia, were similar in IPF and COPD cohorts and lower than the 3% population-based frequency in the ESP, suggesting that this mutation is not contributing to pulmonary disease (see online supplementary table S1).18

to sample size. The frequencies of the common TERT variant p.A279T, previously associated with aplastic anaemia, were similar in IPF and COPD cohorts and lower than the 3% population-based frequency in the ESP, suggesting that this mutation is not contributing to pulmonary disease (see online supplementary table S1).18 Familial IPF Of the 132 individuals in the IPF cohort, 26 (20%) reported a family history of IPF. In those 26 individuals, only 3 (12%) had a mutation identified in one of the 6 genes, one each in SFTPA2, SFTPC and TERT. The remaining 23 individuals (88%) did not have a mutation in any of these candidate genes that explained their family history. Interestingly, the other 11 individuals in the IPF cohort with an identified mutation did not have a family history of IPF, suggesting (1) that the mutations were spontaneous in these individuals, (2) the mutations were inherited but not penetrant in the parent, (3) in the case of novel mutations, they were not disease causing or (4) family history of pulmonary fibrosis had not been fully ascertained or recognised through family member recall.

gesting (1) that the mutations were spontaneous in these individuals, (2) the mutations were inherited but not penetrant in the parent, (3) in the case of novel mutations, they were not disease causing or (4) family history of pulmonary fibrosis had not been fully ascertained or recognised through family member recall. Mutations in MUC5B Collapsed frequencies of rare coding region mutations in MUC5B were similar in the IPF and COPD cohorts and in the ESP database. Consequently, mutations in the coding sequence did not appear to be associated with attributable risk for IPF and were analysed separately from mutations in the other genes (table 4). Interestingly, p.V5436M was significantly under-represented in the IPF cohort (p=0.002), but the clinical significance, if any, of this observation is unclear (see online supplementary table S1). Table 4 MUC5B mutations in individuals of European descent Gene Mutation Amino acid prediction IPF (n=119) (n (MAF)) COPD (n=178) (n (MAF)) ESP (ED) (MAF), % dbSNP number MUC5B c.356G>A R119H 0 2 (0.6%) 0.03 rs199733278 c.1520C>T T507M 1 (0.4%) 0 0.09 rs201605309 c.1855C>T R619W 2 (0.8%) 2 (0.5%) 0.38 rs56411917 c.1994C>T *A665V 1 (0.4%) 0 c.14896G>A G4966S 0 3 (0.8%) 0.3 rs56217440 Collapsed freq 1.7% 2.0% *Novel mutation. ED, European descent; IPF, idiopathic pulmonary fibrosis; MAF, minor allele frequency.

Gene Mutation Amino acid prediction IPF (n=119) (n (MAF)) COPD (n=178) (n (MAF)) ESP (ED) (MAF), % dbSNP number MUC5B c.356G>A R119H 0 2 (0.6%) 0.03 rs199733278 c.1520C>T T507M 1 (0.4%) 0 0.09 rs201605309 c.1855C>T R619W 2 (0.8%) 2 (0.5%) 0.38 rs56411917 c.1994C>T *A665V 1 (0.4%) 0 c.14896G>A G4966S 0 3 (0.8%) 0.3 rs56217440 Collapsed freq 1.7% 2.0% *Novel mutation. ED, European descent; IPF, idiopathic pulmonary fibrosis; MAF, minor allele frequency. We also tested both cohorts for the previously reported IPF-associated MUC5B promoter polymorphism (rs35705950; table 5). The 28% minor allele frequency (MAF) in our IPF cohort is significantly greater than the MAF in either the COPD cohort (13%, p<0.001) or the previously reported population-based frequency of 9%, p<0.001.12 31 The similar prevalence of the rs35705950 polymorphism in individuals with or without mutations in the other six genes studied in either cohort (p=1.0) suggests that the polymorphism does not interact with mutations in these genes to modify disease penetrance (table 5, online supplementary table S2). However, the small sample size of this study limits the ability to truly detect interactions. Table 5 MUC5B promoter variant (rs35705950) genotypes and minor allele frequencies (MAF) in individuals with IPF-associated mutations

We also tested both cohorts for the previously reported IPF-associated MUC5B promoter polymorphism (rs35705950; table 5). The 28% minor allele frequency (MAF) in our IPF cohort is significantly greater than the MAF in either the COPD cohort (13%, p<0.001) or the previously reported population-based frequency of 9%, p<0.001.12 31 The similar prevalence of the rs35705950 polymorphism in individuals with or without mutations in the other six genes studied in either cohort (p=1.0) suggests that the polymorphism does not interact with mutations in these genes to modify disease penetrance (table 5, online supplementary table S2). However, the small sample size of this study limits the ability to truly detect interactions. Table 5 MUC5B promoter variant (rs35705950) genotypes and minor allele frequencies (MAF) in individuals with IPF-associated mutations IPF COPD Mutation (n=14) No Mutation (n=118) Total (n=132) Mutation (n=6) No Mutation (n=186) Total (n=192) G/G 6 (43%) 57 (48%) 63 (48%) 6 (100%) 140 (75%) 146 (76%) G/T 8 (57%) 57 (48%) 65 (49%) 0 (0%) 44 (24%) 44 (23%) T/T 0 (0%) 4 (3%) 4 (3%) 0 (0%) 2 (1%) 2 (1%) MAF 8 (29%)* 65 (28%)* 73 (28%)† 0 (0%) 48 (13%) 48 (13%)† p Value IPF versus population (MAF 9%)=6.7 e-12; COPD versus population=0.09.12 p Value: COPD no mutation versus mutation=0.38. *p Value IPF mutation versus no mutation=1.0. †p Value of IPF versus COPD=1.9 e-06. COPD, chronic obstructive pulmonary disease; IPF, idiopathic pulmonary fibrosis.

IPF COPD Mutation (n=14) No Mutation (n=118) Total (n=132) Mutation (n=6) No Mutation (n=186) Total (n=192) G/G 6 (43%) 57 (48%) 63 (48%) 6 (100%) 140 (75%) 146 (76%) G/T 8 (57%) 57 (48%) 65 (49%) 0 (0%) 44 (24%) 44 (23%) T/T 0 (0%) 4 (3%) 4 (3%) 0 (0%) 2 (1%) 2 (1%) MAF 8 (29%)* 65 (28%)* 73 (28%)† 0 (0%) 48 (13%) 48 (13%)† p Value IPF versus population (MAF 9%)=6.7 e-12; COPD versus population=0.09.12 p Value: COPD no mutation versus mutation=0.38. *p Value IPF mutation versus no mutation=1.0. †p Value of IPF versus COPD=1.9 e-06. COPD, chronic obstructive pulmonary disease; IPF, idiopathic pulmonary fibrosis. Discussion We demonstrated that previously described rare, dominantly expressed mutations in SFTPC, SFTPA2 and TERT occur in patients with sporadic and familial IPF.4–10 However, our study is the first in which all genes associated with both IPF and/or childhood ILD were resequenced simultaneously in a cohort of individuals with both sporadic and familial IPF. The lack of mutations in more than one gene in each individual suggests that interactions among these genes do not contribute significantly to disease expression, although significantly larger cohorts are necessary to confirm this preliminary observation and possibly identify further rare variants. Furthermore, recently published genome-wide association studies have identified other candidate loci associated with increased risk of IPF and should be interrogated in future studies but had not been identified at the time that gene selection for this study occurred.13 32

n and possibly identify further rare variants. Furthermore, recently published genome-wide association studies have identified other candidate loci associated with increased risk of IPF and should be interrogated in future studies but had not been identified at the time that gene selection for this study occurred.13 32 In this study, 14 (11%) of the individuals in the IPF cohort had mutations in IPF-associated genes. This frequency may be either overestimated or underestimated based on our definition of a ‘mutation’ which relied on a previous association of the gene with disease, or computationally predicted effects on protein function. We used a stringent definition for altered function by requiring that 2 of 3 algorithms predicted functionality. However, only testing in a relevant model system and replication in an independent cohort can confirm the functional significance of these mutations. Additionally, our IPF cohort was weighted toward individuals who were referred for lung transplantation which could bias towards younger individuals, and thus, could also be enriching the cohort for genetically-based disease. The significant gender difference between the IPF and COPD cohorts is consistent with the demographics of each disease and mutations were identified in individuals of both genders in each cohort but any possible gender affect on rare variant discovery would have to be investigated with a larger patient population.

ase. The significant gender difference between the IPF and COPD cohorts is consistent with the demographics of each disease and mutations were identified in individuals of both genders in each cohort but any possible gender affect on rare variant discovery would have to be investigated with a larger patient population. IPF and childhood ILD-associated genes Surfactant-associated genes As anticipated from other studies, we found mutations in SFTPC in one individual with familial IPF (4%) and four individuals with sporadic IPF (4%). Previous studies have demonstrated the prevalence of SFTPC mutations to be as high as 25% in familial pulmonary fibrosis (without a distinction from IPF, specifically) and as low as 0.7% in sporadic IPF.7 8 The SFTPC mutations identified in our cohort have previously been identified in children and adults with ILD and/or IPF. Two of these mutations, p.I73T and p.L110R, result in aberrant prosurfactant protein C protein products in vitro.33 Since we did not find any mutations in SFTPC in the COPD cohort, and they are not present in the general population, this enrichment in the IPF cohort further strengthens the evidence that these dominantly expressed mutations are disease-causing and demonstrates that even ‘sporadic’ IPF may have a genetic basis due to spontaneous mutation.

any mutations in SFTPC in the COPD cohort, and they are not present in the general population, this enrichment in the IPF cohort further strengthens the evidence that these dominantly expressed mutations are disease-causing and demonstrates that even ‘sporadic’ IPF may have a genetic basis due to spontaneous mutation. Dominant mutations in SFTPA2 have been identified in kindreds with familial IPF and lung cancer.4 In our study, p.V178M was present in one individual with familial IPF and was seen only once in ESP. However, no other affected family members in this kindred were available to determine if this mutation segregates with disease, so it is difficult to determine if this mutation is disease-causing.

IPF and lung cancer.4 In our study, p.V178M was present in one individual with familial IPF and was seen only once in ESP. However, no other affected family members in this kindred were available to determine if this mutation segregates with disease, so it is difficult to determine if this mutation is disease-causing. Homozygous recessive or compound heterozygous ABCA3 mutations cause neonatal respiratory failure and childhood ILD, and single ABCA3 mutations (p.E292V and p.D123N) interacting with SFTPC p.I73T have been reported in two families, one with childhood ILD and one with IPF.10 34 Our identification of one individual who was homozygous for ABCA3 p.E292V adds to the recent identification of a kindred with pulmonary fibrosis due to a p.G964D thus further supporting the possibility that adult-onset fibrotic lung disease due to homozygous or compound heterozygous mutations in ABCA3 may occur.11 In contrast to our previous study that demonstrated an enrichment of single ABCA3 mutations in newborns with RDS, suggesting a developmental interaction that increased risk or severity of disease, we did not find the frequency of single ABCA3 mutations in the IPF cohort to be greater than the 3–5% in the general population. This is consistent with a prior study showing single E292V mutations were not a major risk factor for pulmonary disease in the general population, and also suggests that interactions with other modifiers are less likely to increase risk for disease.35–37

IPF cohort to be greater than the 3–5% in the general population. This is consistent with a prior study showing single E292V mutations were not a major risk factor for pulmonary disease in the general population, and also suggests that interactions with other modifiers are less likely to increase risk for disease.35–37 Since mutations in NKX2-1 are extremely rare in the general population, the lack of identifiable mutations in NKX2-1 in our IPF and COPD cohorts may simply be a function of the limited sample size of our cohort, thus making it difficult to know the true prevalence of NKX2-1 mutations in patients with IPF. TERT Previous studies of TERT and the RNA component of telomerase (TERC) have shown heterozygous mutations to be present in 8–15% of familial IPF cohorts.6 In our study population, novel mutations in TERT were only seen in the IPF cohort, solidifying the link between this gene and predisposition to IPF. That the p.A279T mutation was present in IPF and COPD cohorts with similar frequencies makes it unlikely that it contributes directly to the pathogenesis of IPF. Complex functional assays will be necessary to identify any mechanisms whereby these mutations exert an effect.

Project (HL-102923), the WHI Sequencing Project (HL-102924), the Broad GO Sequencing Project (HL-102925), the Seattle GO Sequencing Project (HL-102926) and the Heart GO Sequencing Project (HL-103010). Study data were collected and managed using REDCap (Research Electronic Data Capture), hosted at Washington University. Contributors: MAC prepared the samples, performed all validation, analysed the results, wrote the first draft of paper and took part in revising the manuscript. AS, HJH and TDR recruited patients, assisted in project development and writing of the manuscript. RDM assisted in project development, data interpretation and writing of the manuscript. QZ analysed the results. DJW and FSC assisted in project development, result analysis and writing. AH designed and supervised the study, assisted in project development, analysed results and completed writing of the manuscript. All authors agreed on the final version of the manuscript. Funding: This research was supported by the National Institutes of Health (NIH) CTSA Grant # UL1 TR000448 through the Washington University Institute for Clinical and Translational Sciences (AH), NIH HL065174 (FSC, AH) HL082747 (FSC, AH) and HL089968 (HJH, AS, FSC). Competing interests: None. Ethics approval: Washington University Human Research Protection Office. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Key messages We believe this study is the first to compare the appraisal and help-seeking experiences of patients with symptoms indicative of lung cancer (such as cough, dyspnoea, haemoptysis, chest and back pain) between people subsequently diagnosed with lung cancer and people diagnosed with other non-cancer conditions. The study was guided by the Aarhus statement recommendations on improving design and reporting of studies on early cancer diagnosis. Recruiting at the time of referral to specialist respiratory services and interviewing patients before or close to diagnosis reduced risk of post hoc rationalisation and recall bias. Almost half of the study group visited the general practitioner (GP) two or more times before an appropriate investigation or referral was made, enabling us to report on the patient perception of GP advice on symptom monitoring and on when to reconsult.

Recruiting at the time of referral to specialist respiratory services and interviewing patients before or close to diagnosis reduced risk of post hoc rationalisation and recall bias. Almost half of the study group visited the general practitioner (GP) two or more times before an appropriate investigation or referral was made, enabling us to report on the patient perception of GP advice on symptom monitoring and on when to reconsult. Introduction Lung cancer is the most common cause of cancer death in the UK. While there has been a fall in incidence in men, there has been a slow, steady increase of cases in women. It continues to have one of the poorest 5-year relative survival rates of all cancers,1 only 7.8% for men and 9.3% for women.2 Most cases of lung cancer present symptomatically, and poor survival rates are primarily due to later stage disease, the biology of the disease, lack of screening and fewer treatment options.3 4 Diagnosis in primary care is challenging as the majority of patients who present to their general practitioner (GP) with respiratory symptoms will not have lung cancer.5 Survival rates in the UK from lung cancer are poorer than in other European countries6 and it may be that negative beliefs about barriers to symptomatic presentation contribute to this.7 Understanding how patients recognise possible signs of lung cancer and the decisions they make about seeking help for their symptoms can inform the development of interventions to reduce the time to diagnose lung cancer and potentially improve survival.

fs about barriers to symptomatic presentation contribute to this.7 Understanding how patients recognise possible signs of lung cancer and the decisions they make about seeking help for their symptoms can inform the development of interventions to reduce the time to diagnose lung cancer and potentially improve survival. The timeliness of patient help-seeking for potential cancer symptoms is influenced by a number of factors. A UK population survey, using the validated Cancer Awareness Measure, identified four key barriers to timely help-seeking: perceived service barriers, that is, difficulty getting an appointment; practical barriers, that is, lacking the time or transport to attend the consultation; emotional barriers that is, fear of receiving bad news, and failing to interpret the symptom/s as requiring medical attention.8 Studies explicitly investigating the help-seeking experiences of people diagnosed with lung cancer have reported similar barriers.9–15 Furthermore, the recognition of new respiratory symptoms is particularly difficult for patients who have a lung comorbidity such as chronic obstructive pulmonary disease (COPD).5 Many patients at risk of lung cancer have a history of smoking, which can further prolong help-seeking owing to perceived risks of being stigmatised and of not being worthy of medical help.10

mptoms is particularly difficult for patients who have a lung comorbidity such as chronic obstructive pulmonary disease (COPD).5 Many patients at risk of lung cancer have a history of smoking, which can further prolong help-seeking owing to perceived risks of being stigmatised and of not being worthy of medical help.10 Raising awareness of cancer symptoms may help to promote timely help-seeking. Public health campaigns such as the ‘Be Clear on Cancer’ lung cancer campaign16 have been used to raise awareness of the signs of lung cancer. At a community level, there is evidence of increased awareness of lung cancer symptoms and increase in referrals to specialist respiratory services during similar campaigns.17 18 However, understanding the possible cause of a symptom does not always directly translate into seeking a consultation to discuss such symptoms.19 People have to make complex decisions about when it is appropriate to seek help, and in part these decisions are influenced by perception of personal risk.20 Decisions to seek help are also shaped by public and professional perception of what is a reasonable time to wait for symptoms to resolve spontaneously. While there is general consensus that ‘red flag symptoms’ such as haemoptysis21 should be presented and referred urgently, National Institute for Clinical Evidence (NICE) recommends that other respiratory symptoms such as cough and dyspnoea should have been present for at least 3 weeks before investigation by chest X-ray.22 Therefore, it is important that we more fully understand the reasoning behind patients’ help-seeking decisions for a range of respiratory symptoms in order that interventions, particularly those aimed at promoting presentation of symptoms, can be developed to improve timeliness of help-seeking. To date, studies exploring patient appraisal and help-seeking for symptoms suggestive of lung cancer have only reported the experiences of those diagnosed with lung cancer, and have interviewed or surveyed patients often several months or years after diagnosis. These studies may be biased by post hoc rationalisation and recall bias. Patients are more likely to recall their appraisal and help-seeking decisions fully if they are interviewed as close as possible to the time they were experienced, and preferably before their diagnosis is known.

al months or years after diagnosis. These studies may be biased by post hoc rationalisation and recall bias. Patients are more likely to recall their appraisal and help-seeking decisions fully if they are interviewed as close as possible to the time they were experienced, and preferably before their diagnosis is known. Our aim was to understand the symptom evaluation, or ‘appraisal’, and help-seeking decisions of patients with symptoms suggestive of lung cancer. In this paper, we report the results from an interview study that recruited people with respiratory symptoms referred to specialist respiratory services for consideration of possible cancer, irrespective of their subsequent diagnosis. This method enabled us to explore the complex processes and events that shaped patient appraisal and help-seeking from when they first noticed a symptom, to first consultation with a healthcare professional (HCP), through until they were referred. Methods Design and definitions This in-depth, face-to-face interview study was nested within the SYMPTOM Lung Study (http://discovery-programme.org/symptom_study.php). The SYMPTOM study was a prospective cohort study investigating associations between symptoms and other factors on the total diagnostic interval and stage of diagnosis among patients with symptoms suggestive of lung, colorectal and pancreatic cancer. This interview study used qualitative methods to explore the factors that affected patient appraisal and help-seeking for respiratory symptoms. Ethical approval was obtained from Cambridgeshire 3 Research Ethics Committee (10/H0306/50).

among patients with symptoms suggestive of lung, colorectal and pancreatic cancer. This interview study used qualitative methods to explore the factors that affected patient appraisal and help-seeking for respiratory symptoms. Ethical approval was obtained from Cambridgeshire 3 Research Ethics Committee (10/H0306/50). The study design, including data collection and analysis, was underpinned by the theoretical approach of the Model of Pathways to Treatment (figure 1).23 24 The model enables explicit consideration of patient, disease and healthcare factors that impact on patients’ appraisal of symptoms and decisions to seek help. Using a theoretical framework and the definitions of events along the patient pathway reflects best practice as defined in the Aarhus Statement.25 Detecting bodily change, perceiving a reason to seek help and first consulting a HCP are key ‘milestones’ or events in the pathway to treatment, and represent the ‘time to presentation’ (TTP).25 We define TTP as the interval between the patient-reported date of first noticing a symptom and their first consultation with an HCP, usually their GP. However, as one-third of patients with lung cancer consult their GP three times before referral,26 for participants who are not referred after the first consultation, we have also defined their further symptom appraisal and decisions to seek help again as the ‘Re-appraisal Interval’. This definition has resonance with the iterative nature of the Appraisal and Help-seeking Intervals as illustrated in the Pathways to Treatment model, figure 1.

ot referred after the first consultation, we have also defined their further symptom appraisal and decisions to seek help again as the ‘Re-appraisal Interval’. This definition has resonance with the iterative nature of the Appraisal and Help-seeking Intervals as illustrated in the Pathways to Treatment model, figure 1. Figure 1 Model of pathways to treatment (HCP, healthcare professional). Study setting and recruitment Recruitment to the study was undertaken when patients were referred to specialist respiratory clinics in secondary and tertiary care at five hospitals in the East and North East of England. Patients aged 40 years and over, referred to hospital via urgent (2 week wait), routine and diagnostic routes, for whom GPs had reported symptoms potentially suggestive of lung cancer, were mailed an introduction letter and SYMPTOM questionnaire, and invited to join the SYMPTOM study. They were also invited to take part in this interview study. We purposively sampled participants by region, selecting for variation of age, gender, educational level and diagnoses (cancer/other non-cancer). Data collection Interviews were undertaken by NH, KM and LB between November 2011 and January 2013. Interviews were conducted as soon as possible after referral to specialist respiratory care, the majority within 10 weeks. Twelve participants were unaware of their diagnosis at interview (2 lung cancer). All diagnoses were confirmed from review of secondary care medical records.

between November 2011 and January 2013. Interviews were conducted as soon as possible after referral to specialist respiratory care, the majority within 10 weeks. Twelve participants were unaware of their diagnosis at interview (2 lung cancer). All diagnoses were confirmed from review of secondary care medical records. Interviews were preceded by an explanation of the research process and signing of a consent form; consent was also rechecked at the end of each interview. We then used open-ended questions to explore the participant's appraisal of symptoms and help-seeking decisions with questions developed from our experiences of undertaking similar interviews with people recently diagnosed with cancer.20 Concepts explored included description and appraisal of the initial symptom/s and any self-care; the appraisal of subsequent symptom/s and the trigger/s that prompted consultation with an HCP; and the outcome of this consultation, and any further appraisal and help-seeking decisions. We probed the participant's understanding of symptoms, previous knowledge and understanding of lung cancer, and social influences on their decisions. A specifically developed calendar landmarking instrument was used to assist with participant recall, which helped to clarify the sequence and dates of events and intervals during their pathway to diagnosis.27

ing of symptoms, previous knowledge and understanding of lung cancer, and social influences on their decisions. A specifically developed calendar landmarking instrument was used to assist with participant recall, which helped to clarify the sequence and dates of events and intervals during their pathway to diagnosis.27 Interviews lasted between 40 and 65 minutes and were usually conducted in the participant's home; one person chose to be interviewed in university offices and another in their workplace. In several of the interviews a relative, usually spouse, was present. Relatives contributed to the interview by confirming participant's comments or adding detail to the account. Interviews continued until saturation of data, when no new themes were identified in three consecutive interviews.28 Audio-recordings of interviews were professionally transcribed verbatim and anonymised. Analysis Analysis was an iterative process that started after the first few interviews. Framework analysis methods were used to ensure a systematic and rigorous progression through the five analytic steps: familiarisation with data; developing thematic framework; indexing data to framework; mapping and questioning the data; and theoretical interpretation.29 The thematic framework was developed by NH and KM in consultation with FW and LB; our study patient representative (MJ) also contributed to all stages of analysis. Data management was assisted by NVivo V.9.

ematic framework; indexing data to framework; mapping and questioning the data; and theoretical interpretation.29 The thematic framework was developed by NH and KM in consultation with FW and LB; our study patient representative (MJ) also contributed to all stages of analysis. Data management was assisted by NVivo V.9. During the mapping and questioning of data, we examined transcripts of participants diagnosed with lung cancer and compared them with those who presented with similar symptoms but who were diagnosed with other non-cancer conditions. We explicitly considered symptom appraisal, decisions about help-seeking, re-appraisal and further help-seeking. Following thematic analysis, data were charted by patient characteristics (age, gender, smoking, lung comorbidity and geographical region) to seek patterns or non-confirming cases between those diagnosed with lung cancer and those diagnosed with other non-cancer conditions. Results Patient characteristics Seven hundred and seventy-six (81% n=963) of the participants recruited into the main SYMPTOM Lung study expressed an interest in taking part in the interview study. We undertook purposive sampling among this group, and only seven people declined to be interviewed because they were feeling unwell or were too busy with hospital appointments.

nty-six (81% n=963) of the participants recruited into the main SYMPTOM Lung study expressed an interest in taking part in the interview study. We undertook purposive sampling among this group, and only seven people declined to be interviewed because they were feeling unwell or were too busy with hospital appointments. We undertook 35 interviews; 17 participants were diagnosed with lung cancer and 18 with other conditions. Table 1 shows the sociodemographic characteristics of the cohort, as reported in the SYMPTOM Lung study. The characteristics of the interview study cohort were similar to those of the SYMPTOM Lung study, although there were more smokers in the interview study. We purposefully sampled for people with a cancer diagnosis. In the interview study nearly half (49%) of the cohort had lung cancer compared with only 16% in the questionnaire study. Both cohorts had similar sociodemographic characteristics (ethnicity, education, employment). In the interview study nine reported they had existing lung comorbidity (4 COPD, 3 asthma, 2 other lung conditions). Table 1 Participant characteristics

We undertook 35 interviews; 17 participants were diagnosed with lung cancer and 18 with other conditions. Table 1 shows the sociodemographic characteristics of the cohort, as reported in the SYMPTOM Lung study. The characteristics of the interview study cohort were similar to those of the SYMPTOM Lung study, although there were more smokers in the interview study. We purposefully sampled for people with a cancer diagnosis. In the interview study nearly half (49%) of the cohort had lung cancer compared with only 16% in the questionnaire study. Both cohorts had similar sociodemographic characteristics (ethnicity, education, employment). In the interview study nine reported they had existing lung comorbidity (4 COPD, 3 asthma, 2 other lung conditions). Table 1 Participant characteristics Diagnosis Lung cancer N=17 Other conditions* N=18 Interview study N=35 SYMPTOM Lung study N=963 Location (England) East 8 8 16 North East 9 10 19 Sex Female 6 9 15 (42.9%) 441 (45.8%) Male 11 9 20 (57.1%) 522 (54.2%) Age mean years (range) 69 (57–83) 61 (41–88) 65 (41–88) 66 (40–95) Ethnicity White 17 16 33 (94.3%) 931 (96.7%) Other – 2 2 (5.7%) 32 (3.3%) Smoking Current smoker 5 4 9 (25.7%) 115 (12.2%) Ex-smoker 10 6 16 (45.7%) 517 (54.8%) Never smoked 2 8 10 (28.6%) 311 (33.0%) Education Up to higher education 10 13 23 (65.7%) 655 (68.0%) Higher education or equivalent 7 5 12 (34.3%) 308 (32.0%) Employment† In employment 2 6 8 (23.5%) 274 (30.4%) Disabled or unemployed 2 4 6 (17.6%) 56 (6.2%) Retired 12 8 20 (58.8%) 570 (63.3%) IMD quintile‡ Least deprived 1 5 6 11 (31.4%) 317 (32.9%) 2 4 3 7 (20%) 216 (22.4%) 3 2 3 5 (14.3%) 160 (16.6%) 4 2 4 6 (17.1%) 126 (13.1%) Most deprived 5 4 2 6 (17.1%) 143 (14.9%) Number of GP appointments§ 1 9 7 16 2 or more 6 10 16 Type of referral to specialist services¶ Urgent (2WW) 14 8 22 Routine 0 8 8 Comorbidities Respiratory COPD 1 3 4 Other lung conditions 1 4 5 Other Previous cancer** 2 1 3 Diabetes 3 1 4 Heart disease 4 4 8 Arthritis 3 7 10 Anxiety/depression 3 2 5 IBS 2 2 4 *Other conditions: pneumonia (3), COPD (2), asthma (1), pleurisy (1), fibrosis asbestosis (1), eosinophilic bronchitis (1), allergic rhinitis (1), acid reflux (1), calcified lung lesion (1), weight gain (1), post-infection sensitivity (1), nil abnormal found (3) and unreported (1).

10 Anxiety/depression 3 2 5 IBS 2 2 4 *Other conditions: pneumonia (3), COPD (2), asthma (1), pleurisy (1), fibrosis asbestosis (1), eosinophilic bronchitis (1), allergic rhinitis (1), acid reflux (1), calcified lung lesion (1), weight gain (1), post-infection sensitivity (1), nil abnormal found (3) and unreported (1). †Missing data interview study n=1: questionnaire study n=63. ‡The IMD is a key data set on deprivation. The Indices measure levels of deprivation across seven distinct domains: Income Deprivation, Employment Deprivation, Health Deprivation and Disability, Education Skills and Training Deprivation, Barriers to Housing and Services, Living Environment Deprivation and Crime. §Three participants referred following MRI or CT for other conditions. ¶5 Missing: missing data=3; referred through secondary care hospitals to tertiary centres=2. **Previous cancer sites: kidney (1) and bowel (2). COPD, chronic obstructive pulmonary disease; GP, general practitioner; IBS, irritable bowel syndrome; IMD, Index of Multiple Deprivation. Of the 17 participants diagnosed with lung cancer, one had a lung metastasis from a primary colon cancer; eight had disease staged with potential for treatment with curative intent (stage I n=4; stage II n=4), while eight had more advanced disease (stage III n=1; stage IV n=7). The 18 participants with other conditions mainly had inflammatory conditions, see table 1.

, one had a lung metastasis from a primary colon cancer; eight had disease staged with potential for treatment with curative intent (stage I n=4; stage II n=4), while eight had more advanced disease (stage III n=1; stage IV n=7). The 18 participants with other conditions mainly had inflammatory conditions, see table 1. Duration of respiratory symptom/s Seven participants did not seek help because of respiratory symptom/s; however, they did retrospectively recall respiratory symptoms and these data were included in our analysis. Three were referred to respiratory clinics following CT scan or MRI for non-respiratory symptoms, which reported lung abnormalities (2 lung cancer), and four had their respiratory condition opportunistically detected during consultations with GPs for other health concerns (2 lung cancer). Most of the remaining 28 participants could recall when they first noticed their symptom/s and the date they went to the GP, though many were unable to recall the exact date they decided to seek help (ie, the start of the Help-seeking Interval in figure 1). We report the Appraisal Interval and first Help-seeking Interval as the TTP. The TTP ranged from 2 to 334 days; just under half consulted within 30 days of first noticing a symptom. Table 2 presents the characteristics of participants in the quartiles with the shortest and longest TTP, demonstrating similarities in age, gender, initial symptom attributions and trigger for help-seeking between those diagnosed with lung cancer and those diagnosed with other non-cancer conditions.

noticing a symptom. Table 2 presents the characteristics of participants in the quartiles with the shortest and longest TTP, demonstrating similarities in age, gender, initial symptom attributions and trigger for help-seeking between those diagnosed with lung cancer and those diagnosed with other non-cancer conditions. Table 2 Comparison of attribution and symptom/s that triggered help-seeking between participants with the shortest and longest quartile in the interval from first noticing a symptom to first presentation (ordered by first TTP)

noticing a symptom. Table 2 presents the characteristics of participants in the quartiles with the shortest and longest TTP, demonstrating similarities in age, gender, initial symptom attributions and trigger for help-seeking between those diagnosed with lung cancer and those diagnosed with other non-cancer conditions. Table 2 Comparison of attribution and symptom/s that triggered help-seeking between participants with the shortest and longest quartile in the interval from first noticing a symptom to first presentation (ordered by first TTP) TTP (days) Sex Age Smoking status Comorbidity Dyspnoea Haemoptysis Dry cough Productive cough Chest pain Back pain Tightness in chest Increased tiredness Initial patient attribution Lung cancer Time between noticing a system and seeking help—shortest quartile 2 F 45–49 Never Asthma, IBS ● ○ Sinus infection 7 M 65–69 Never Arthritis ● Gallstones 7 F 80–84 Ex – ● Flu 7 F 65–69 Ex – ○ ● ● Chest infection ✓ 14 F 50–54 Ex Anxiety depression ● Chest infection 14 M 70–74 Ex – ● Getting older ✓ 15 F 55–59 Current – ● ● Chest infection 21 M 65–69 Ex – ● Muscle pain ✓ Time between noticing a symptom and seeking help—longest quartile 49 M 80–84 Current – ● ○ Getting older ✓ 49 M 60–64 Ex – ● Muscle pain ✓ 60 F 55–59 Never – ● Allergy 58 M 60–64 Ex – ● Over exertion ✓ 60 M 55–59 Current – ● ○ Cardiac problems 120 F 40–44 Never – ○ ● ○ ○ Chest infection 150 M 65–69 Ex Heart disease ● ○ Cancer ✓ 180 M 80–04 Ex Heart disease, diabetes ● Asbestos in lung ● Indicates symptom/s that triggered help-seeking. ○ Indicates symptom/s reported that did not trigger help-seeking.

● Over exertion ✓ 60 M 55–59 Current – ● ○ Cardiac problems 120 F 40–44 Never – ○ ● ○ ○ Chest infection 150 M 65–69 Ex Heart disease ● ○ Cancer ✓ 180 M 80–04 Ex Heart disease, diabetes ● Asbestos in lung ● Indicates symptom/s that triggered help-seeking. ○ Indicates symptom/s reported that did not trigger help-seeking. F, female; M, male; IBS, irritable bowel syndrome; TTP, time to presentation. For the 16 (6 lung cancer) who consulted their GP two or more times before referral to specialist respiratory services, we also report the number of days between their first consultation and their referral: the Re-appraisal Interval. The time between first consultation and referral was between 10 and 182 days; for seven participants it was over 60 days (2 lung cancer). There were similarities in the Re-appraisal Interval between participants diagnosed with lung cancer and those diagnosed with other non-cancer conditions, although overall, patients diagnosed with lung cancer had fewer GP visits and were mainly referred under the 2 week wait referral system (table 3). Table 3 Re-appraisal Interval: time between first noticing a symptom to first consultation and first consultation until referral to specialist respiratory services, by time intervals (ordered by first TTP)

For the 16 (6 lung cancer) who consulted their GP two or more times before referral to specialist respiratory services, we also report the number of days between their first consultation and their referral: the Re-appraisal Interval. The time between first consultation and referral was between 10 and 182 days; for seven participants it was over 60 days (2 lung cancer). There were similarities in the Re-appraisal Interval between participants diagnosed with lung cancer and those diagnosed with other non-cancer conditions, although overall, patients diagnosed with lung cancer had fewer GP visits and were mainly referred under the 2 week wait referral system (table 3). Table 3 Re-appraisal Interval: time between first noticing a symptom to first consultation and first consultation until referral to specialist respiratory services, by time intervals (ordered by first TTP) Time from noticing a symptom to first consultation (TTP) (days) Time from first consultation to referral (Re-appraisal) (days) Number of consultations Type of referral Gender and age Diagnosis or lung cancer staging (T N M)* Diagnosed with cancer 1 3 60 2 Urgent M, 75–79 IB 0 0 2 7 14 2 Urgent M, 75–79 IV 2 IA 3 21 14 2 Urgent M, 65–69 IV 2 IA 4 49 35 2 Urgent M, 80–84 IIA I 0 5 50 102 4 Urgent F, 70–74 IV 3 IA 6 90 133 4 Urgent F, 55–59 IIB 0 0 Diagnosed with other conditions 7 2 28 3 Routine F, 45–49 Nil abnormal detected 8 2 35 4 Routine F, 50–54 Emphysema 9 7 182 3 NA F, 80–84 Interstitial pneumonia 10 14 80 3 Routine F, 50–54 Asthma 11 15 63 2 Urgent F, 55–59 Emphysema 12 30 138 4 Urgent F, 65–69 Eosinophilic bronchitis 13 60 10 2 Urgent M, 55–59 Tracheobronchitis and GORD 14 60 34 3 NA F, 55–59 Seasonal allegoric rhinitis 15 60 60 2 Routine M, 55–59 COPD 16 120 126 4 Routine F, 40–44 Erythema suggestive of GORD *T=tumour size; N=Regional lymph; M=Distant metastasis; Stage of cancer is indicative of whether treatment will be curative in intent.

Urgent M, 55–59 Tracheobronchitis and GORD 14 60 34 3 NA F, 55–59 Seasonal allegoric rhinitis 15 60 60 2 Routine M, 55–59 COPD 16 120 126 4 Routine F, 40–44 Erythema suggestive of GORD *T=tumour size; N=Regional lymph; M=Distant metastasis; Stage of cancer is indicative of whether treatment will be curative in intent. COPD, chronic obstructive pulmonary disease; F, female; GORD, Gastro-oesophageal reflux disease; M, male; NA, not applicable; TTP, time to presentation. Qualitative themes We found many similarities in the appraisal and help-seeking of participants who were diagnosed with lung cancer compared to those diagnosed with other conditions. Only two participants reported that they immediately thought of cancer because of their symptoms (1 lung cancer). We report the data in three sections: The appraisal of symptoms describes how they are difficult to recognise, and that lung comorbidities often masked respiratory changes, the alternative explanations offered for the cause of symptoms, and the ways in which symptoms were self-managed prior to help-seeking. Deciding to seek help describes the factors that helped to endorse a decision to seek a GP consultation, including symptoms such as signs of acute illness, recognising a changing symptom, symptoms not responding as expected, the influence of family members and the social network, increased awareness of lung cancer and the impact of smoking on help-seeking.

ctors that helped to endorse a decision to seek a GP consultation, including symptoms such as signs of acute illness, recognising a changing symptom, symptoms not responding as expected, the influence of family members and the social network, increased awareness of lung cancer and the impact of smoking on help-seeking. Symptom re-appraisal and returning to the GP describes the factors that shaped the decision to seek a further consultation including symptom change, increasing impact of symptom on activities, increasing concern about personal risk of having lung cancer and the role of the GP in endorsing such help-seeking. Extracts from interviews illustrate the results; each quotation is contextualised by the participant's gender, age band, smoking history, TTP in days, diagnosis (cancer, non-cancer) and if diagnosis was not known at time of interview. The appraisal of symptoms The first interval in the patient pathway is the appraisal and self-management of symptom/s. If a symptom was difficult to recognise, attributed to ageing or smoking, or appraised as a condition that could be self-managed, participants often remained for some time in the appraisal interval. We found no evidence of differences between those diagnosed with lung cancer and those diagnosed with other non-cancer conditions across any of the factors, as shown in table 4. Table 4 Time to Presentation (TTP): factors that impacted on symptom appraisal and initial help-seeking across people diagnosed with lung cancer and those diagnosed with other conditions

The appraisal of symptoms The first interval in the patient pathway is the appraisal and self-management of symptom/s. If a symptom was difficult to recognise, attributed to ageing or smoking, or appraised as a condition that could be self-managed, participants often remained for some time in the appraisal interval. We found no evidence of differences between those diagnosed with lung cancer and those diagnosed with other non-cancer conditions across any of the factors, as shown in table 4. Table 4 Time to Presentation (TTP): factors that impacted on symptom appraisal and initial help-seeking across people diagnosed with lung cancer and those diagnosed with other conditions Participants diagnosed with lung cancer Participants diagnosed with other conditions Factors that increased timeliness of help-seeking Symptom sign of acute illness You do the usual, you take your Lemsip's and your Beechams Powders and when it doesn't clear up after a week you think well you need some antibiotics or something slightly stronger. So that was when I went to the doctors (M, 75–79, smoking not recorded, 7 days) I went to the doctor actually quite quickly because I had to fly and I knew that I would have trouble in the aeroplane if I didn't get some medication (F, 50–54, ex-smoker, 14 days) Symptom caused concern I went up a slight slope, and when I got to the top I was absolutely panting and really short of breath. … And I sat till I got my breath back, then I carried on normally, but when I got back home, on thinking about it, it wasn't right, and so that was the trigger point (M, 70–74, ex-smoker, 15 days) I found myself very short of breath the onset was very quick and I thought ‘I don't think I should feel like this’, and I thought if I'm puffing like this rang the surgery and said ‘do you think you could make me an appointment for the next day because I'm really short of breath (F, 80–84, ex-smoker, 7 days, diagnosis not known) Influence of family Wife persuaded me to go to the doctor about it. I wasn't too worried about it, it wasn't significant to me (M, 75–79, ex-smoker, 10 days, diagnosis not known) It didn't get any better, it got worse.

or the next day because I'm really short of breath (F, 80–84, ex-smoker, 7 days, diagnosis not known) Influence of family Wife persuaded me to go to the doctor about it. I wasn't too worried about it, it wasn't significant to me (M, 75–79, ex-smoker, 10 days, diagnosis not known) It didn't get any better, it got worse. So (daughter's name) would say, mother, that cough is awful and husband was saying the same (F, 65–69, never smoked, 30 days, diagnosis not known) Factors that prolonged TTP Competing responsibilities I was so damn busy, too busy to think of this…I was busy with all sorts of things, work and everything, and eventually I made the appointment (F, 70–74, ex-smoker, 50 days) It started about the end of October in Australia and I saw my doctor about, the Monday before Christmas, … the cough had got worst whilst I was away (F, 55–59, never smoked, 60 days) Misattribution of symptoms I thought it was my asthma getting worse because I was getting more breathless I honestly truly believed it was my asthma getting worse (F, 55–59, ex-smoker, 90 days) Breathlessness 2 or 3 months, but I mean it's just something which is coming on, so you don't get over concerned (M, 55–59, smoker, 60 days, diagnosis not known) Healthcare factors I didn't go because I couldn't explain it. I thought I don't want to waste his time just saying “I've, somehow I feel a bit strange” or whatever it is (M, 75–79, ex-smoker, 56 days) Whenever I'm out of this situation (immigrant visa expired) then I will go and you know talk to the GP about it (M, 40–44, never smoked, 334 days, diagnosis not known) F, female; M, male; TTP, time to presentation.

is time just saying “I've, somehow I feel a bit strange” or whatever it is (M, 75–79, ex-smoker, 56 days) Whenever I'm out of this situation (immigrant visa expired) then I will go and you know talk to the GP about it (M, 40–44, never smoked, 334 days, diagnosis not known) F, female; M, male; TTP, time to presentation. Difficult to recognise The insidious nature of some respiratory symptoms made it difficult to recognise a change in respiratory function “breathlessness has been creeping up” (M, 55–59, smoker, 60 days, non-cancer). An elderly man diagnosed with lung cancer explained he felt “not quite right, yucky, bit weary and tired”. He said he could not describe his symptoms sufficiently to justify visiting the doctor and waited 8 weeks before consulting his GP. Having a lung comorbidity, such as COPD, could make it difficult to notice a change in symptom/s and some participants were only alerted to their respiratory symptom during consultations for other conditions:I've always been breathless for 30-odd, it's 34 years since I got this heart disease, then I have COPD and emphysema. (M, 55–59, smoker, opportunistically detected, cancer) Seven participants had not sought help for their symptoms although, in retrospect, those diagnosed following a CT scan or MRI for another reason, could often recall respiratory symptoms. These symptoms had not caused them concern as this woman diagnosed with lung cancer said about her cough:P. I suppose it's been going on a while Int. Would that be a few weeks or less than that? P. Oh, more than that it was a dry tickly cough

Seven participants had not sought help for their symptoms although, in retrospect, those diagnosed following a CT scan or MRI for another reason, could often recall respiratory symptoms. These symptoms had not caused them concern as this woman diagnosed with lung cancer said about her cough:P. I suppose it's been going on a while Int. Would that be a few weeks or less than that? P. Oh, more than that it was a dry tickly cough Int. Ok, have you had any breathlessness? P. A little, but that could be for other reasons…had it not been picked up on the scan I still wouldn't have gone to the GP. (F, 65–69, never smoked, opportunistically detected, cancer, diagnosis not known) Alternative explanations for respiratory symptoms Alternative explanations provided for the cause of symptoms were usually contextualised by anticipated changes in bodily function, or as a reaction to specific activities. For example, older people may expect to be more breathless on exertion: “I noticed maybe a few time I was not able to walk as fast as I used to be able to do, but I didn't think much about it…I am 72 I am going to begin to slow down” (M, 70–74, ex-smoker, 14 days, cancer). Not all smokers mentioned smoking when discussing their appraisal or help-seeking but the few who did explained their symptom could be due to smoking. Another explanation, particularly for pain, was muscle strain. The explanation was justified within the context of activity: “I do a lot of work around the garden, picking things up, obviously muscles go into a bit of a spasm, so not to worry” (M, 65–69, ex-smoker, 21 days, cancer).

eir symptom could be due to smoking. Another explanation, particularly for pain, was muscle strain. The explanation was justified within the context of activity: “I do a lot of work around the garden, picking things up, obviously muscles go into a bit of a spasm, so not to worry” (M, 65–69, ex-smoker, 21 days, cancer). Self-management of symptoms During the appraisal interval some participants tried to contain and self-manage their symptoms. By drawing on their knowledge about the cause of similar symptoms, their judgements about the cause of their symptom influenced subsequent decisions to seek healthcare:It started in February and I just thought a tickly cough, it just went on and on and I thought it was a cold and then I thought, perhaps it's the beginning of hay fever so I just kept leaving it, trying hay fever tablets. (F, 40–44, never smoked, 121 days, non-cancer) An alternative explanation for self-managing the symptom was occasionally situated within a dislike of attending the doctor's surgery:I don't particularly enjoy going to the doctors so I went to the chemist and gradually built up a line of medication from the chemist these cough bottle I decided I would take rather than going to the doctors. (F, 65–69, never smoked, 30 days, non-cancer, diagnosis not known)

a dislike of attending the doctor's surgery:I don't particularly enjoy going to the doctors so I went to the chemist and gradually built up a line of medication from the chemist these cough bottle I decided I would take rather than going to the doctors. (F, 65–69, never smoked, 30 days, non-cancer, diagnosis not known) Deciding to seek help The decision to consult an HCP and move from appraising symptoms to seeking help was triggered by patient, disease and, to lesser extent, healthcare factors. Appraising symptom/s as a sign of acute illness prompted timely help-seeking. Family members and friends recognised symptoms in the participants and encouraged them to seek help. Increasing awareness of the signs of lung cancer, usually due to the ‘Be Clear on Cancer Campaign’, prompted help-seeking. Participants did not report that their smoking behaviour was an inhibitor to help-seeking. Competing responsibilities and limited access to healthcare prolonged the TTP for only a few participants, see table 4. Symptoms as a sign of acute illness Just over half of the participants consulted their GP within 30 days of first noticing their symptom, usually when it was attributed to a chest infection, see table 4. Drawing on previous experience and knowledge they quickly sought a consultation anticipating a prescription to alleviate symptoms:I thought it might be chest infection in which case you know, antibiotics and away you go. (F, 65–69, ex-smoker, 7 days, cancer) I thought that I needed maybe some antibiotics if I'd got a chest infection. (F, 50–54, smoker, 2 days, non-cancer)

Symptoms as a sign of acute illness Just over half of the participants consulted their GP within 30 days of first noticing their symptom, usually when it was attributed to a chest infection, see table 4. Drawing on previous experience and knowledge they quickly sought a consultation anticipating a prescription to alleviate symptoms:I thought it might be chest infection in which case you know, antibiotics and away you go. (F, 65–69, ex-smoker, 7 days, cancer) I thought that I needed maybe some antibiotics if I'd got a chest infection. (F, 50–54, smoker, 2 days, non-cancer) When acute illness was experienced help-seeking was rapid, and those diagnosed with cancer and those diagnosed with other non-cancer conditions both said they had been well before their acute symptoms. When asked to clarify how long she had been feeling breathless one woman explained “maximum a week, up until then was still swimming 3 times a week, I'd walk for a good hour every day, I'd play table tennis just carry on as normal” (F, 65–69, ex-smoker, 7 days, cancer). Symptoms not appraised as being due to a lung condition were often also attributed to an acute illness based on previous experience. For example, one man with a history of intermittent pain in the lower back and recent gallstones attributed his pain to a kidney infection and sought help promptly.

When acute illness was experienced help-seeking was rapid, and those diagnosed with cancer and those diagnosed with other non-cancer conditions both said they had been well before their acute symptoms. When asked to clarify how long she had been feeling breathless one woman explained “maximum a week, up until then was still swimming 3 times a week, I'd walk for a good hour every day, I'd play table tennis just carry on as normal” (F, 65–69, ex-smoker, 7 days, cancer). Symptoms not appraised as being due to a lung condition were often also attributed to an acute illness based on previous experience. For example, one man with a history of intermittent pain in the lower back and recent gallstones attributed his pain to a kidney infection and sought help promptly. Recognising a changing symptom Having a lung comorbidity meant patients were already living with chronic respiratory symptoms yet they were still able to disentangle and identify the ‘unusual’ or changing symptom. Even when a symptom was distinct from what was normally experienced some undertook ‘watchful waiting’ to see if the change was repeated before seeking help:I've got COPD but I never coughed up blood before. I thought it will clear up but after two weeks it didn't so I thought I had better get it checked. (M 70–74, ex-smoker, 14 days, non-cancer)

tinct from what was normally experienced some undertook ‘watchful waiting’ to see if the change was repeated before seeking help:I've got COPD but I never coughed up blood before. I thought it will clear up but after two weeks it didn't so I thought I had better get it checked. (M 70–74, ex-smoker, 14 days, non-cancer) When other factors highlighted the change, such as the failure of self-management or usual treatment to control the symptom, then the need for timely help-seeking was reinforced:It wasn't like my normal asthma cough, I'd use my inhaler it had no impact at all … still continue coughing. (F, 45–49, never smoked, 2 days, non-cancer) Symptoms not responding as expected When symptoms were appraised as being self-limiting and not requiring medical intervention, such as due to allergy or muscle strain, help-seeking was only instigated when the symptom continued or failed to respond in the expected way. A cough appraised as an allergy that did not improve when the seasons changed triggered an initial consultation.A seasonal allergy… (then it's) December, it's freezing cold and I was continuing to cough so I rang the doctor. (F, 55–59, never smoked, 60 days, non-cancer)

ntinued or failed to respond in the expected way. A cough appraised as an allergy that did not improve when the seasons changed triggered an initial consultation.A seasonal allergy… (then it's) December, it's freezing cold and I was continuing to cough so I rang the doctor. (F, 55–59, never smoked, 60 days, non-cancer) Similarly, with an appraisal of muscle pain there was the expectation that the pain would subside. When this did not happen or the pain increased there was an urgency to seek help:It (back pain) got worse and worse, absolute agony…we had a couple of nights on holiday and I thought if it is still aching when I get back I'll call the doctor…we actually got back at 3 in the morning and I actually managed to get an appointment that day. (M, 65–69, ex-smoker, 21 days, cancer) Influence of family members and the social network A persistent cough was reported by approximately a third of participants (7 lung cancer). The cough was usually publicly noticeable, for instance, the tickly cough which made it difficult to talk and led to retching, and the expectorant cough with sputum and tissues. Family members, friends and work colleagues readily commented on a persistent cough:Our friends were saying “Oh you know that cough is dreadful… it's just going on and on.” (F, 55–59, never smoked, 60 days, non-cancer) A friend said perhaps you ought to get that cough looked at. (F, 70–74, ex-smoker, 50 days, cancer)

Influence of family members and the social network A persistent cough was reported by approximately a third of participants (7 lung cancer). The cough was usually publicly noticeable, for instance, the tickly cough which made it difficult to talk and led to retching, and the expectorant cough with sputum and tissues. Family members, friends and work colleagues readily commented on a persistent cough:Our friends were saying “Oh you know that cough is dreadful… it's just going on and on.” (F, 55–59, never smoked, 60 days, non-cancer) A friend said perhaps you ought to get that cough looked at. (F, 70–74, ex-smoker, 50 days, cancer) Older participants tended to attribute increasing breathless to a normal part of aging and their help-seeking was also often prompted by relatives, illustrating the importance of raising awareness of cancer symptoms in the older population:My daughter said “I've never walked with you and you've lagged behind…” then she started to question about this breathing. (M, 80–84, smoker, 49 days, cancer, diagnosis not known) When symptoms were ignored by the participant, the family was proactive in endorsing and sometimes organising the first consultation:Probably I've been ignoring it for a few weeks, you know… the wife took it out of my hands she phoned the doctor. (M, 55–59, smoker, 60 days, non-cancer, diagnosis not known) Even when participants lived alone, those who spoke to family and friends about their symptoms were encouraged to consult a GP, “brother said ‘I'm telling you now to go to the doctor” (M, 70–74, ex-smoker, 14 days, non-cancer).

When symptoms were ignored by the participant, the family was proactive in endorsing and sometimes organising the first consultation:Probably I've been ignoring it for a few weeks, you know… the wife took it out of my hands she phoned the doctor. (M, 55–59, smoker, 60 days, non-cancer, diagnosis not known) Even when participants lived alone, those who spoke to family and friends about their symptoms were encouraged to consult a GP, “brother said ‘I'm telling you now to go to the doctor” (M, 70–74, ex-smoker, 14 days, non-cancer). Another man who lived alone explained “my friend went up to see the doctor and said I was in a bad way and the GP came here” (M, 60–64, ex-smoker, internal hospital referral for sleep apnoea tests, non-cancer).

Even when participants lived alone, those who spoke to family and friends about their symptoms were encouraged to consult a GP, “brother said ‘I'm telling you now to go to the doctor” (M, 70–74, ex-smoker, 14 days, non-cancer). Another man who lived alone explained “my friend went up to see the doctor and said I was in a bad way and the GP came here” (M, 60–64, ex-smoker, internal hospital referral for sleep apnoea tests, non-cancer). Increased awareness of lung cancer A cough was often the symptom that triggered help-seeking. Public awareness of a cough as a sign of lung cancer was raised by the ‘Be Clear on Cancer’ Lung cancer campaign, which ran across England during the period of interviewing.16 Five participants had seen the information and they focused on the headline message about cough being a warning sign of cancer but could not recall the other symptoms. Two did not have a cough so did not feel the advert applied to them. One said it had raised their awareness of their cough and triggered their help-seeking, although it had not specifically heightened their concern of lung cancer:Until that advert came on, I never really took much notice of this cough… then it dawned on me I had this cough for a couple of weeks, so I waited till the third week and I went. (F, 55–59, smoker, 15 days, non-cancer, diagnosis not known)

eeking, although it had not specifically heightened their concern of lung cancer:Until that advert came on, I never really took much notice of this cough… then it dawned on me I had this cough for a couple of weeks, so I waited till the third week and I went. (F, 55–59, smoker, 15 days, non-cancer, diagnosis not known) The remaining two reported that the information had little impact, as they had already made the decision to book a GP consultation. Furthermore, one reported the campaign had slightly prolonged their help-seeking:Had that not been screened I would have been to the doctors the week before, because I was saying to myself, it's just ‘cos that's on you're getting worried, so it had the adverse effect. (M, 65–69, ex-smoker, 150 days, cancer) Impact of smoking on help-seeking Those who had a history of smoking did not initially seek medical help with a concern of possible lung cancer and being a smoker did not impact on TTP, as in table 2. Rather, the cancer risk implicit in smoking was considered more when participants reappraised their symptoms following the first consultation. Nonetheless, when there was a change in the smoker's cough the decision was made to seek help:Because if you are a smoker, see first thing in the morning it gets you, well I was waking up through the night with this dry cough, but it never lasted long and that's when I really started ooh have I got cancer, like lung cancer. (M 55–59, smoker, 60 days, non-cancer, diagnosis not known)

cision was made to seek help:Because if you are a smoker, see first thing in the morning it gets you, well I was waking up through the night with this dry cough, but it never lasted long and that's when I really started ooh have I got cancer, like lung cancer. (M 55–59, smoker, 60 days, non-cancer, diagnosis not known) Symptom re-appraisal and returning to the GP Participants were generally happy with the outcome of their initial consultation, but when the prescribed treatment was not effective or their symptom changed, they started a process of re-appraisal and made decisions about whether to seek another GP consultation. There were similarities in the help-seeking decisions of those diagnosed with lung cancer and those diagnosed with other non-cancer conditions (see table 5). Table 5 Re-appraisal: factors that triggered further help-seeking following initial consultation across people diagnosed with lung cancer and those diagnosed with other conditions

Symptom re-appraisal and returning to the GP Participants were generally happy with the outcome of their initial consultation, but when the prescribed treatment was not effective or their symptom changed, they started a process of re-appraisal and made decisions about whether to seek another GP consultation. There were similarities in the help-seeking decisions of those diagnosed with lung cancer and those diagnosed with other non-cancer conditions (see table 5). Table 5 Re-appraisal: factors that triggered further help-seeking following initial consultation across people diagnosed with lung cancer and those diagnosed with other conditions Participants diagnosed with lung cancer Participants diagnosed with other conditions Concern about symptom I thought it strange to have a muscular problem in my back that hadn't got better in a couple of weeks so I thought it might be something, not necessarily sinister, well I suppose a slipped disc or a bone problem (M, 65–69, ex-smoker, 21 days, 2 GP visits 14 days) I coughed up a bright red blood clot. It was while I was waiting to see doctor (for a persistent cough), I phoned them up and I said, “Look, I need to see someone and I need to see someone today” (F, 55–59, smoker, 15 days, 2 GP visits 63 days, diagnosis not known) Increasingly affected ability to undertake activity I take the dog out every day and that was getting less and less, the walk, so I knew something was wrong and I was getting to the stage where I didn't want to take her (F, 55–59, ex-smoker, 90 days, 4 GP visits 133 days) Coughing with the fan heater and that kind of dry heat and because of my job I go into hot houses and then back to my car, back to hot houses all the time and that set the coughing off (F, 40–44, 120 days, 4 GP visits 126 days) Increasing concern that symptom indicative of lung cancer I read an article in the paper about a man who had a cough for weeks, found he had lung cancer and I remember thinking ‘oh god’ have I got that (F, 70–74, ex-smoker, 50 days, 4 GP visits 102 days) My husband decided I'd got lung cancer.

ys, 4 GP visits 126 days) Increasing concern that symptom indicative of lung cancer I read an article in the paper about a man who had a cough for weeks, found he had lung cancer and I remember thinking ‘oh god’ have I got that (F, 70–74, ex-smoker, 50 days, 4 GP visits 102 days) My husband decided I'd got lung cancer. (Laughs) Well a friend of ours had exactly these symptoms and had lung cancer F, 55–59, never smoked, 60 days, 3 GP visits 34 days) Increasing concern about candidacy for lung cancer (Negative case) I was expecting a clean bill of health, I mean I've never shown any sign of illness, I'm 75 now and apart from the odd bit of flu I've never really been ill all my life (M, 75–79, smoking not recorded,7 days, 2 GP visits 14 days) Because of the potential for methotrexate to be a problem and the fact that I had had a cough for longer than 3 weeks (F, 56, never smoked, 60 days, 3 GP visits 34 days, non-cancer) I thought I am a smoker, I've been a smoker for a lot of years that (cancer) crossed my mind (F, 55–59, smoker, 15 days 2 GP visits 63 days, diagnosis not known) Further help-seeking endorsed by GP The doctor said it looks like being muscular I'll give you some painkillers and come back in couple of weeks if it hasn't gone. So two weeks exactly I went back (M, 65–69, ex-smoker, 21 days 2 GP visits 14 days) The doctor asked me to go back in a fortnight which I did (F, 81, ex-smoker, 7 days, 3 GP visits 182 days, non-cancer, diagnosis not known) The doctor did say, try the steroids and the one inhaler to start with, see how you go, because you can come back (F, 40–45, never smoked, 121 days, 3 GP visits 126 days) F, female; GP, general practitioner; M, male.

fortnight which I did (F, 81, ex-smoker, 7 days, 3 GP visits 182 days, non-cancer, diagnosis not known) The doctor did say, try the steroids and the one inhaler to start with, see how you go, because you can come back (F, 40–45, never smoked, 121 days, 3 GP visits 126 days) F, female; GP, general practitioner; M, male. In this section each quotation is contextualised by the participant's gender, age band, smoking history, TTP in days, re-appraisal time in days and diagnosis. Symptom change triggering further help-seeking Following their initial consultation, an awareness of symptom changes often triggered a further GP consultation. For some it was the recognition of new symptoms:We all sort of joked about my cough, but no, I was seriously very, very tired, and I suddenly thought “My God, I feel I've aged five years in five months…” and I thought no, wait a minute, you are breathless as well. (F, 70–74, ex-smoker, TTP 50 days, Re-appraisal 102 days, cancer) For others it was the persistence or an increase in the severity of one or more of their symptoms that made them decide to seek further help:The coughing was worse and it started really hurting my lungs and I started really being nervous about it because every time when I coughed it felt like as if I was ripping my lungs apart, it was really painful. (F, 50–54, ex-smoker, TTP 14 days, Re-appraisal 80 days, non-cancer, diagnosis not known)

seek further help:The coughing was worse and it started really hurting my lungs and I started really being nervous about it because every time when I coughed it felt like as if I was ripping my lungs apart, it was really painful. (F, 50–54, ex-smoker, TTP 14 days, Re-appraisal 80 days, non-cancer, diagnosis not known) When the symptom was recognised as a warning sign, such as coughing up blood, the urgency to return to the GP increased for some but not all people (table 5). For example, if the blood in the sputum could be attributed to other causes, the urgency to seek further help was reduced, as illustrated by a man who was using inhalers following his first consultation:I started to cough I noticed flecks of blood. I took no notice of it and thought possible it was the stuff I was breathing in, or that it is just a burst blood vessel in my lungs. (M, 80–84, smoker TTP 49 days, Re-appraisal 35 days, cancer) Increasing impact of symptom on activities During the re-appraisal, participants described how symptom/s increasingly curtailed their ability to undertake work and leisure activities. This increasing impact of symptoms on activities acted as a trigger to reconsult:I was trying to go to the shop and it was getting to be a bit of an embarrassment because in the shop, if I coughed me eyes would water and I'd choke or be sick. So I'd have to leave me shopping, come out the exit, try to pull myself together, then go back in. (F, 65–69, never smoked, TTP 30 days, Re-appraisal 138 days, non-cancer, diagnosis not known)

it was getting to be a bit of an embarrassment because in the shop, if I coughed me eyes would water and I'd choke or be sick. So I'd have to leave me shopping, come out the exit, try to pull myself together, then go back in. (F, 65–69, never smoked, TTP 30 days, Re-appraisal 138 days, non-cancer, diagnosis not known) Increasing concern at personal risk of lung cancer During the process of re-appraisal of symptoms several participants became increasingly concerned their symptoms were indicative of lung cancer. Personal predisposition for lung cancer was considered within the context of family history and previous environmental exposure to carcinogenic materials, particularly smoking. Drawing on their knowledge of lung cancer from seeing relatives or friends with the disease, the match or mismatch in symptoms shaped their re-appraisal and subsequent decision to seek further help:My dad was coughing sputum, he coughed a lot of sputum, but I haven't, so they were totally different. My dad had a cough, I didn't have a cough, so there was a lot of differences… so the furthest thing from my mind it was going to be lung cancer. (F, 55–59, ex-smoker, TTP 90 days, Re-appraisal 133 days, cancer) I thought it was lung cancer ‘cos my mam died of lung cancer… she used to always have like a dry cough, and I was having a bit of a dry cough. (M, 55–59, smoker, TTP 60 days, Re-appraisal 60 days, non-cancer, diagnosis not known)

Increasing concern at personal risk of lung cancer During the process of re-appraisal of symptoms several participants became increasingly concerned their symptoms were indicative of lung cancer. Personal predisposition for lung cancer was considered within the context of family history and previous environmental exposure to carcinogenic materials, particularly smoking. Drawing on their knowledge of lung cancer from seeing relatives or friends with the disease, the match or mismatch in symptoms shaped their re-appraisal and subsequent decision to seek further help:My dad was coughing sputum, he coughed a lot of sputum, but I haven't, so they were totally different. My dad had a cough, I didn't have a cough, so there was a lot of differences… so the furthest thing from my mind it was going to be lung cancer. (F, 55–59, ex-smoker, TTP 90 days, Re-appraisal 133 days, cancer) I thought it was lung cancer ‘cos my mam died of lung cancer… she used to always have like a dry cough, and I was having a bit of a dry cough. (M, 55–59, smoker, TTP 60 days, Re-appraisal 60 days, non-cancer, diagnosis not known) Heightened concern about personal risk for lung cancer due not only to smoking but also to environmental exposure seemed to facilitate further help-seeking:I've been smoking all my life more or less, I did work with asbestos when I was an apprentice and you think well maybe this is what it is all about maybe you have got cancer. (M, 55–59, smoker, TTP 60 days, Re-appraisal 10 days, non-cancer, diagnosis not known)

nmental exposure seemed to facilitate further help-seeking:I've been smoking all my life more or less, I did work with asbestos when I was an apprentice and you think well maybe this is what it is all about maybe you have got cancer. (M, 55–59, smoker, TTP 60 days, Re-appraisal 10 days, non-cancer, diagnosis not known) GPs role in endorsing further help-seeking The decision to reconsult was not always an easy one. A minority of participants reported they felt the GP was dismissive of their symptom/s and concerns, “I said no, no, you're not listening to me I've had this cough for ages and ages, it's not just an overnight thing” (F, 65–69, never smoked, TTP 30 days, Re-appraisal 138 days, non-cancer, diagnosis not known). Explicit advice on when to return seemed to legitimise a further consultation even when new symptoms had not developed (table 5). However, the absence of advice on further monitoring of symptoms and the appropriateness of a further consultation led some to revert to self-managing symptom/s:Doctor gave me some antibiotics and he said it would do the trick. It didn't… so I left it, I left it and I left it and I continued another line of cough mixtures and stuff. (F, 65–69, never smoked, TTP 30 days, Re-appraisal 138 days, non-cancer, diagnosis not known)

tation led some to revert to self-managing symptom/s:Doctor gave me some antibiotics and he said it would do the trick. It didn't… so I left it, I left it and I left it and I continued another line of cough mixtures and stuff. (F, 65–69, never smoked, TTP 30 days, Re-appraisal 138 days, non-cancer, diagnosis not known) Management of comorbidities often led to regular contact with their GP, but it was apparent that opportunities to investigate symptoms suggestive of lung cancer were not always maximised.Because I had the x-ray in July Dr didn't think it (pain) would be anything but we would keep an eye on it. …I was going to the doctors maybe once a fortnight, once a month and every time it was for something else and I forgot to mention it even though it was happening. (F, 55–59, ex-smoker, TTP 90 days, Re-appraisal 133 days, cancer) Consultations for comorbidities were not only with the GP; symptoms suggestive of lung cancer were mentioned to other HCPs but there was little evidence of effective advice on symptom monitoring being provided to the participants:I go six monthly to the nurse in the clinic and I mentioned to her I was spitting blood and she said “well make an appointment with the doctor” but I just went out and never bothered. (M, 85–89, ex-smoker, 180 days, non-cancer, diagnosis not known) When this patient returned a few months later and reported his haemoptysis again the nurse phoned the doctor directly and he had a chest X-ray within 24 h.

Consultations for comorbidities were not only with the GP; symptoms suggestive of lung cancer were mentioned to other HCPs but there was little evidence of effective advice on symptom monitoring being provided to the participants:I go six monthly to the nurse in the clinic and I mentioned to her I was spitting blood and she said “well make an appointment with the doctor” but I just went out and never bothered. (M, 85–89, ex-smoker, 180 days, non-cancer, diagnosis not known) When this patient returned a few months later and reported his haemoptysis again the nurse phoned the doctor directly and he had a chest X-ray within 24 h. Discussion Main findings We believe this is the first study to explore the appraisal and help-seeking decisions of patients responding to symptoms suggestive of lung cancer. Unsurprisingly, there are few differences in the appraisal of symptoms, decisions to seek help, TTPs and re-appraisal intervals between people diagnosed with lung cancer and other conditions. We found evidence of complex reasoning and decision-making supporting the processes of deciding whether and when to seek help. While patients often recognised new symptoms or subtle symptom changes, even against a background of the expected symptoms of lung and cardiac comorbidities or being a smoker, they often did not feel the need to seek help because they developed alternative explanations based on their previous experiences or believed they could self-manage the symptoms. Help-seeking was triggered by recognising the symptoms as signs of acute or serious illness, the progression or persistence of existing symptoms or new symptoms, the influence of family members and their social network, particularly due to the visibility of symptoms, and sometimes current public health messages. Half the sample received treatment for other conditions, such as for acute respiratory illness, and had not been referred after their first GP consultation. Symptom monitoring and re-appraisal followed; returning to the GP was again influenced by the progression or persistence of existing symptoms or new symptoms, an impact of symptoms on daily living activities, and increasing concern about underlying serious disease and cancer. We found little evidence that patients received adequate advice from their GPs about symptom monitoring or reasons to return for review.

progression or persistence of existing symptoms or new symptoms, an impact of symptoms on daily living activities, and increasing concern about underlying serious disease and cancer. We found little evidence that patients received adequate advice from their GPs about symptom monitoring or reasons to return for review. Strengths and limitations The major strength of this study is that we interviewed people during their pathway to diagnosis and treatment, often before they received their diagnosis; therefore we were able to compare accounts of people diagnosed with lung cancer and other non-cancer conditions. We sought to interview people as early in their disease development as possible. Interviewing 12 people before they received their diagnosis helped to reduce post hoc rationalisation and recall bias,9 11–13 and the remaining interviews were conducted as soon as possible after diagnosis (range: 1 day to 16 weeks), with 15 (43%) occurring within 4 weeks of diagnosis. Importantly, we were able to include people with advanced stage cancer before they became too ill to participate in research.

alisation and recall bias,9 11–13 and the remaining interviews were conducted as soon as possible after diagnosis (range: 1 day to 16 weeks), with 15 (43%) occurring within 4 weeks of diagnosis. Importantly, we were able to include people with advanced stage cancer before they became too ill to participate in research. In accordance with the guidelines outlined in the Aarhus Statement on improving design and reporting of studies on early cancer diagnosis,25 we used a rigorous study design with the theoretical Model of Pathways to Treatment23 underpinning the interview schedule as well as the analysis. The calendar-landmarking instrument helped some participants recall dates and symptom changes. Purposive sampling from two areas of England ensured data were reported from people with differing socioeconomic backgrounds and differing exposures to carcinogenic environments. For example, some people in the North East spoke of personal risk due to proximity to, or employment in, heavy industry. Our broad range of scientific and clinical expertise helped ensure consensus in the findings, and we benefitted from the input of our lay member at all stages of the research process, including interpretation of the data.

le in the North East spoke of personal risk due to proximity to, or employment in, heavy industry. Our broad range of scientific and clinical expertise helped ensure consensus in the findings, and we benefitted from the input of our lay member at all stages of the research process, including interpretation of the data. We acknowledge that the experiences of patients from these two regions may not be representative of those from other regions of the UK and that we cannot know the experiences of those who did not take part, although the interview sample has similar demographic characteristics to the main SYMPTOM Lung study (see table 1). Furthermore, we are only able to report the perspectives of patients who had been referred through primary care and had consulted their GP, but we have not been able to access the appraisal or help-seeking experiences of those who presented first at accident and emergency department.30 While we used a calendar-landmarking instrument during the interviews to ensure, as far as possible, the accuracy of time intervals, some people were unable to recall precise dates and we took the dates from their responses to the questionnaire in the main SYMPTOM lung study. The nature of qualitative data collection is such that we can only report the experiences patients chose to divulge; it may be that they did not share experiences that they considered to be private or embarrassing.

ise dates and we took the dates from their responses to the questionnaire in the main SYMPTOM lung study. The nature of qualitative data collection is such that we can only report the experiences patients chose to divulge; it may be that they did not share experiences that they considered to be private or embarrassing. Comparison with existing literature We recruited people with symptoms suggestive of lung cancer, and included patients who were not aware of their diagnosis at the time of interview, those who were diagnosed with other non-cancer conditions as well as people diagnosed with early and later stage lung cancer. We are therefore able to add new insights to the existing literature on the experiences and decision-making processes of people with lung cancer9–13 in order to develop new targeted interventions to promote timelier lung cancer diagnosis. We found that initial alternative explanations for symptoms were based on patients’ previous experiences, for example, a dry cough during summer was attributed to hay fever, and slight breathlessness in winter attributed to a chest infection. This is concordant with findings from a UK interview study of patients with operable and inoperable lung cancer, which reported that they minimised, normalised or misattributed symptoms.12 However, that study also reported their patients lacked agency in seeking help; in contrast, patients in our study tended to either adopt ‘watchful waiting’ to see if symptoms improved, or to self-manage symptoms with over the counter drugs.

r, which reported that they minimised, normalised or misattributed symptoms.12 However, that study also reported their patients lacked agency in seeking help; in contrast, patients in our study tended to either adopt ‘watchful waiting’ to see if symptoms improved, or to self-manage symptoms with over the counter drugs. Other studies interviewing people with lung cancer have highlighted that lung and cardiac comorbidities, such as COPD and asthma, can delay symptom appraisal and timely help-seeking for lung cancer.9 12 We found that many patients, especially those with comorbidities, showed a complex and sophisticated ability to distinguish minor changes in respiratory function from pre-existing symptoms due either to their comorbidities or smoking habits. They quickly noticed either change in their normal symptoms or the effectiveness of usual medication. The recognition of changing effectiveness of medication is a novel finding and highlights the way in which patients draw on information from different experiences and events to make a judgement on whether or not a symptom requires medical care.31

change in their normal symptoms or the effectiveness of usual medication. The recognition of changing effectiveness of medication is a novel finding and highlights the way in which patients draw on information from different experiences and events to make a judgement on whether or not a symptom requires medical care.31 The nature of cough has been discussed in the context of those with lung cancer32 and we add to this discussion by reporting the impact of cough on help-seeking prior to diagnosis. The very public nature of a chronic cough prompted members of the family and the wider social network to encourage and endorse help-seeking. The impact of cough on daily activities, such as talking on the phone and shopping, also triggered initial and subsequent help-seeking. Living alone has been reported as a factor in prolonging help-seeking,9 33 but in line with an earlier interview study by Tod et al,34 we found that if the participant's symptom was observed by, or discussed with, family and friends, then timely help-seeking often took place.

ggered initial and subsequent help-seeking. Living alone has been reported as a factor in prolonging help-seeking,9 33 but in line with an earlier interview study by Tod et al,34 we found that if the participant's symptom was observed by, or discussed with, family and friends, then timely help-seeking often took place. As we interviewed patients during a national ‘Be Clear on Cancer’ lung cancer campaign16 we are able to comment on the possible impact of the campaign on patient's symptom appraisal and help-seeking behaviour. Some patients who had seen the campaign had a heightened awareness of respiratory symptoms, particularly a persistent cough; for them the campaign endorsed the need to seek help in line with findings from other studies reporting the impact of similar health campaigns.17 18 However, in our study there were two patients who reported that an increased awareness had slightly prolonged their help-seeking as they were concerned they were over-reacting.

hem the campaign endorsed the need to seek help in line with findings from other studies reporting the impact of similar health campaigns.17 18 However, in our study there were two patients who reported that an increased awareness had slightly prolonged their help-seeking as they were concerned they were over-reacting. This is one of the first studies to explicitly explore patient processes during the time between first consulting a HCP, and further consultations leading to investigations and referral. During the Re-appraisal Interval participants continued to consider symptom change, severity and duration as triggers to return to the GP, but rarely reported receiving guidance from their HCP on symptom monitoring or when to reconsult. There were examples of substantial time intervals between GP consultations.5 Those who were given explicit advice on symptom monitoring and when to return to the GP (so-called safety netting),35 promptly sought further consultations enabling GPs to continue diagnostic investigations in a timely way. We found some examples where patients with comorbidities accessed regular healthcare but the HCP did not seem to use the opportunity to follow-up on previously disclosed symptoms suggestive of lung cancer.

ting),35 promptly sought further consultations enabling GPs to continue diagnostic investigations in a timely way. We found some examples where patients with comorbidities accessed regular healthcare but the HCP did not seem to use the opportunity to follow-up on previously disclosed symptoms suggestive of lung cancer. Despite the evidence suggesting that people delay help-seeking due to fear of cancer, we found that patients were often unconcerned that symptoms may be indicative of lung cancer even when they had increased risk due to current or recent smoking habits.20 36 37 In our study, concerns about candidacy for lung cancer were only discussed within the context of reappraising symptoms and when alternative explanations failed to respond as expected to initial treatment. Implications for clinicians and policymakers Our findings provide further evidence for targeted public health campaigns that are tailored to specific groups such as smokers and people with other chest conditions who may have difficultly detecting symptom change, focusing not only on recognition of new, changing and persistent symptoms but also on recognising reduced effectiveness of medications. The role of family and social networks in recognising and discussing a symptom, and then endorsing help-seeking, could contribute more prominently to public health initiatives to raise community awareness of appropriate help-seeking for timely diagnosis of lung cancer and other serious lung conditions.

of medications. The role of family and social networks in recognising and discussing a symptom, and then endorsing help-seeking, could contribute more prominently to public health initiatives to raise community awareness of appropriate help-seeking for timely diagnosis of lung cancer and other serious lung conditions. The vast majority of people who seek help for respiratory symptoms will not have lung cancer but our findings indicate that those with cancer and other non-cancer diagnoses all undertake similar complex reasoning and decision-making when deciding whether and when to seek help. There is a need for further research into the ways in which people make these complex decisions around assessing the seriousness and severity of symptoms, the triggers to seek a medical consultation, and their explanations about which symptoms can be self-managed. Understanding the social context in which risks of ill health are assessed would provide more opportunities for the development of targeted and evidence-based interventions to promote timely help-seeking.

toms, the triggers to seek a medical consultation, and their explanations about which symptoms can be self-managed. Understanding the social context in which risks of ill health are assessed would provide more opportunities for the development of targeted and evidence-based interventions to promote timely help-seeking. Brindle and colleagues recently questioned whether GP elicitation of normalised symptoms could reduce delay in lung cancer diagnosis.13 Our findings confirm that GPs appear to miss opportunities, particularly among people at higher risk such as smokers, ex-smokers and those with chronic chest conditions; this may be due to their gatekeeper role and current guidelines. While acknowledging that the majority of patients presenting with respiratory symptoms will not have lung cancer, there nevertheless remains a need for vigilance and a systematic application of safety-netting procedures such as explicit oral and written instructions detailing expected symptom progression over time; recognition of changing, persistent and new symptoms that should prompt a further appointment; and a specified follow-up time.35

theless remains a need for vigilance and a systematic application of safety-netting procedures such as explicit oral and written instructions detailing expected symptom progression over time; recognition of changing, persistent and new symptoms that should prompt a further appointment; and a specified follow-up time.35 The authors thank the patients who so kindly and freely gave their time and experiences to this study. They thank Helen Morris, the SYMPTOM study manager, our other study patient, and public representatives Victor Boulter and Sue Ballard, and the NHS trusts and their nurses who recruited patients and collected data. The authors would also like to acknowledge the contribution made by the Discovery Programme Steering Committee comprising: Roger Jones (chair); Greg Rubin; Alison Clutterbuck; Ardiana Gjini; Joanne Hartland; Maire Justice; Jenny Knowles; Richard Neal; Peter Rose. Contributors: FMW is the guarantor. All authors approve the final version of the manuscript. All authors contributed to study design, study conduct and study management. LB, NH and KM collected the data and with FMW analysed and interpreted the data. LB, NH and FMW wrote the first draft of the report, and JE, JB, MJ, WH and KM made revisions and agreed the on the final draft. Funding: This report presents independent research funded by the National Institute for Health Research Programme Grants for Applied Research programme (RP-PG-0608–10045). Competing interests: None. Patient consent: Obtained. Ethics approval: Cambridgeshire 3 Research Ethics Committee (10/H0306/50).

Contributors: FMW is the guarantor. All authors approve the final version of the manuscript. All authors contributed to study design, study conduct and study management. LB, NH and KM collected the data and with FMW analysed and interpreted the data. LB, NH and FMW wrote the first draft of the report, and JE, JB, MJ, WH and KM made revisions and agreed the on the final draft. Funding: This report presents independent research funded by the National Institute for Health Research Programme Grants for Applied Research programme (RP-PG-0608–10045). Competing interests: None. Patient consent: Obtained. Ethics approval: Cambridgeshire 3 Research Ethics Committee (10/H0306/50). Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages Lung Clearance Index (LCI) can be time-consuming, limiting its use clinically. Investigation of the flexibility of current multiple breath washout test end points is an important area for future research. LCI1/20 is a repeatable and sensitive test that is shorter than LCI1/40, potentially offering a more feasible research and clinical measure.

Key messages Lung Clearance Index (LCI) can be time-consuming, limiting its use clinically. Investigation of the flexibility of current multiple breath washout test end points is an important area for future research. LCI1/20 is a repeatable and sensitive test that is shorter than LCI1/40, potentially offering a more feasible research and clinical measure. Introduction Lung Clearance Index (LCI) derived from multiple breath washout (MBW) is a sensitive measure of ventilation inhomogeneity1 2 and a robust surrogate outcome measure of the severity of lung disease in cystic fibrosis (CF)3 which has begun to be incorporated into clinical trials.4 5 It also shows promise as a sensitive outcome measure in idiopathic bronchiectasis6 and asthma.7 A drawback of the test is that it can be time-consuming, especially in patients with advanced disease, limiting its feasibility within the clinical environment. By convention a MBW test involves performing a minimum of three inert tracer gas washout runs, ending the washout when end-tidal tracer gas concentration falls below 1/40th of the initial concentration.8 The end point of 1/40th is based on historic studies and has not been systematically validated.8 9 The European Respiratory Society/American Thoracic Society (ERS/ATS) consensus statement highlights investigation of the flexibility of current MBW test end points as an important area for future research, which could potentially improve the utility of this test.8 Assessment of the clinimetric properties of shortened LCI in CF using nitrogen (N2) MBW testing (100% as the inert gas), have reported good diagnostic performance in children with mild disease, offering a measure of ventilation inhomogeneity which may be more practical in the clinical setting.10 However, there are no studies to assess the performance of shortened LCI using sulfur hexafluoride (SF6) MBW (another common MBW method), or studies of shortened LCI in adult patients with more moderate to advanced disease. Differences in gas diffusion and molecular mass of the inert gases used mean that results of the two types of test are not comparable.11 Study of the sensitivity of shortened MBW tests using SF6 could be useful in improving the clinical utility of these tests.

I in adult patients with more moderate to advanced disease. Differences in gas diffusion and molecular mass of the inert gases used mean that results of the two types of test are not comparable.11 Study of the sensitivity of shortened MBW tests using SF6 could be useful in improving the clinical utility of these tests. In this study we aimed to assess and compare the repeatability, sensitivity, specificity and test duration of LCI derived from washout to 1/30th (LCI1/30), 1/20th (LCI1/20) and 1/10th of the initial concentration (LCI1/10) to ‘standard’ LCI derived from washout to 1/40th initial concentration (LCI1/40), using 0.2% SF6 as the tracer gas, in school age—adolescent children and adults with CF and healthy controls.

test duration of LCI derived from washout to 1/30th (LCI1/30), 1/20th (LCI1/20) and 1/10th of the initial concentration (LCI1/10) to ‘standard’ LCI derived from washout to 1/40th initial concentration (LCI1/40), using 0.2% SF6 as the tracer gas, in school age—adolescent children and adults with CF and healthy controls. Methods Subject recruitment Cross-sectional data from 30 people with CF (n=15 aged 6–17 years old; n=15 aged ≥18 years old) and 30 healthy control participants (n=15 aged 6–17 years old; n=15 aged ≥18 years old) with three valid and repeatable MBW tests were analysed. Thirty anonymised CF and 30 healthy control data sets, as consecutively listed in a database of results collected in a large prospective project investigating the clinimetric and clinical relevance of LCI in CF were used. People with CF were recruited at a routine outpatient visit to the Northern Ireland paediatric and adult CF centres at Belfast Health and Social Care Trust (BHSCT), when clinically stable (no pulmonary exacerbation requiring intravenous antibiotics in the previous 4 weeks), between October 2010 and June 2013. Control participants were recruited by means of email circulation among people employed in Queen's University Belfast (QUB) and BHSCT between September 2011 and August 2012. All adult participants provided written informed consent. All child participants provided child or young person assent and parental consent.

2013. Control participants were recruited by means of email circulation among people employed in Queen's University Belfast (QUB) and BHSCT between September 2011 and August 2012. All adult participants provided written informed consent. All child participants provided child or young person assent and parental consent. MBW testing The MBW test to measure LCI was carried out using a modified Innocor device and 0.2% SF6 using the open-circuit technique in accordance with the standard operating procedure developed by the UK CF Gene Therapy Consortium (UKCFGTC; see online supplementary file 1) as described and validated by Horsley et al2 and used in a recent CF clinical trial and observational study.4 12 Participants breathed through a mouthpiece at tidal volumes, while in a seated position and wearing a nose clip. Participants breathed 0.2% SF6 in air via a flowpast circuit until washin was complete, at which point the flowpast was disconnected and the participant breathed room air until the end tidal expired SF6 concentration fell below 1/40th of the initial concentration before disconnection. Three washouts were performed for each participant. Analysis of MBW data was performed using the SimpleWashout programme developed by Dr Nicholas Bell (UKCFGTC) and used with his permission (see online supplementary file 1). For each washout, four values for functional residual capacity (FRC) and LCI were calculated: FRC1/40 and LCI1/40 were derived from washout data from flowpast disconnection until the first breath with end tidal SF6 concentration below 1/40th (≤0.005%) of the starting SF6 concentration (0.2%).

ine supplementary file 1). For each washout, four values for functional residual capacity (FRC) and LCI were calculated: FRC1/40 and LCI1/40 were derived from washout data from flowpast disconnection until the first breath with end tidal SF6 concentration below 1/40th (≤0.005%) of the starting SF6 concentration (0.2%). FRC1/30 and LCI1/30 were derived from washout data from flowpast disconnection until the first breath with end tidal SF6 concentration below 1/30th (≤0.007%) of the starting SF6 concentration (0.2%). FRC1/20 and LCI1/20 were derived from washout data from flowpast disconnection until the first breath with end tidal SF6 concentration below 1/20th (≤0.01%) of the starting SF6 concentration (0.2%). FRC1/10 and LCI1/10 were derived from washout data from flowpast disconnection until the first breath with end tidal SF6 concentration below 1/10th (≤0.02%) of the starting SF6 concentration (0.2%). Mean LCI and FRC values and test duration (minutes) for each end point were calculated from each of the three washouts in each testing session. Spirometry Spirometry was measured according to American Thoracic Society/European Respiratory Society ATS/ERS guidelines13 using a Microlab (ML3500 MK8) spirometer (CareFusion, Kent, UK). Predicted values were calculated from reference ranges for all ages.14 Statistical analysis Data were analysed using PASW Statistics (V.18, IBM software, USA) and Prism (V.5.01 GraphPad Software Inc.) packages. CF and control participant characteristics were summarised using descriptive statistics.

Spirometry Spirometry was measured according to American Thoracic Society/European Respiratory Society ATS/ERS guidelines13 using a Microlab (ML3500 MK8) spirometer (CareFusion, Kent, UK). Predicted values were calculated from reference ranges for all ages.14 Statistical analysis Data were analysed using PASW Statistics (V.18, IBM software, USA) and Prism (V.5.01 GraphPad Software Inc.) packages. CF and control participant characteristics were summarised using descriptive statistics. Intravisit repeatability of LCI1/40, LCI1/30 LCI1/20 and LCI1/10 was assessed using the coefficient of variation (CV%) of all three tests and Bland-Altman plots15 comparing tests one and three, for people with CF and healthy controls. Results between people with CF and healthy controls were compared using an independent samples t test. Mean LCI and FRC values for each end point from people with CF or healthy controls were compared using paired samples t tests. The relationship between mean LCI1/40 and mean LCI1/30, LCI1/20 and LCI1/10 was assessed using scatter plots and the Spearman's rank correlation coefficient. This analysis was also used to assess the relationship between FEV1% predicted and all LCI end points. Sensitivity of all LCI end points compared with FEV1% predicted was assessed using scatter plots and limits of normal of respective tests calculated from healthy controls (mean +1.96 SD). Sensitivity and specificity were further analysed using receiver operating characteristic (ROC) curves, and by comparing area under the ROC curve (AUCROC) and 95% CI for LCI1/40, LCI1/30, LCI1/20, LCI1/10 and FEV1% predicted. Mean test duration (minutes) of LCI1/40, LCI1/30, LCI1/20 and LCI1/10 was compared using a paired samples t test. As multiple comparisons were being made, a Bonferroni adjustment was incorporated. A p value of <0.01 was considered statistically significant.

CI for LCI1/40, LCI1/30, LCI1/20, LCI1/10 and FEV1% predicted. Mean test duration (minutes) of LCI1/40, LCI1/30, LCI1/20 and LCI1/10 was compared using a paired samples t test. As multiple comparisons were being made, a Bonferroni adjustment was incorporated. A p value of <0.01 was considered statistically significant. Results LCI1/40, LCI1/30, LCI1/20, LCI1/10 and FEV1% predicted were significantly different between the CF and control group. However, there was no difference between the CF and control group in age, sex, LCI1/40 CV%, LCI1/30 CV%, LCI1/20 CV%, LCI1/10 CV% or test duration of any LCI end point (table 1). Table 1 CF and healthy control participant characteristics

Results LCI1/40, LCI1/30, LCI1/20, LCI1/10 and FEV1% predicted were significantly different between the CF and control group. However, there was no difference between the CF and control group in age, sex, LCI1/40 CV%, LCI1/30 CV%, LCI1/20 CV%, LCI1/10 CV% or test duration of any LCI end point (table 1). Table 1 CF and healthy control participant characteristics People with CF Healthy controls p Value (CF vs controls) N 30 30 – M/F 14/16 16/14 0.61 Age (years) 20.7 (11.1) (6–51) 20.8 (10.7) (7–44) 0.93 FEV1 (% predicted) 79.3 (17.9) (46.0–116.0) 92.9 (11.3) (68.0–116.0) 0.009 FRC1/40 (L) 1.95 (0.76) 2.23 (1.0) 0.22 LCI1/40 (number of turnovers) 9.0 (2.4) 6.4 (0.5) <0.0001 LCI1/40 CV% 5.2 (2.8) 4.3 (2.0) 0.16 LCI1/40 triplicate test duration (min) 21.8 (9.7) (8.0–57.1) 19.4 (6.5) (10.1–31.1) 0.26 FRC1/30 (L) 1.93 (0.76) 2.24 (1.02) 0.18 LCI1/30 (number of turnovers) 7.7 (1.6) 5.9 (0.4) <0.0001 LCI1/30 CV% 4.6 (3.2) 4.1 (2.8) 0.53 LCI1/30 triplicate test duration (min) 20.9 (9.1) (7.3–53.2) 18.9 (6.2) (9.9–29.9) 0.34 FRC1/20 (L) 1.90 (0.75) 2.24 (1.0) 0.15 LCI1/20 (number of turnovers) 6.6 (1.2) 5.2 (0.4) <0.0001 LCI1/20 CV% 5.7 (3.4) 4.6 (2.3) 0.14 LCI1/20 triplicate test duration (min) 20.0 (8.9) (6.7–51.7) 18.3 5.9 (9.6–28.3) 0.38 FRC1/10 (L) 1.84 (0.73) 2.22 (1.00) 0.09 LCI1/10 (number of turnovers) 4.7 (0.6) 4.0 (0.3) <0.0001 LCI1/10 CV% 5.4 (3.8) 4.3 (2.8) 0.25 LCI1/10 triplicate test duration (min) 18.6 (8.1) (5.7–46.9) 17.1 (5.5) (8.8–26.6) 0.41 All values summarised as mean (SD)±(range).

(min) 20.0 (8.9) (6.7–51.7) 18.3 5.9 (9.6–28.3) 0.38 FRC1/10 (L) 1.84 (0.73) 2.22 (1.00) 0.09 LCI1/10 (number of turnovers) 4.7 (0.6) 4.0 (0.3) <0.0001 LCI1/10 CV% 5.4 (3.8) 4.3 (2.8) 0.25 LCI1/10 triplicate test duration (min) 18.6 (8.1) (5.7–46.9) 17.1 (5.5) (8.8–26.6) 0.41 All values summarised as mean (SD)±(range). CF, cystic fibrosis; CV, coefficient of variation, F, female; FEV1, forced expiratory volume in 1 s; FRC, functional residual capacity; LCI, Lung Clearance Index; M, male. There was no difference between the CF and control group in any of the FRC values (FRC1/40, FRC1/30, FRC1/20 or FRC1/10). Within the CF group, as expected, FRC was incrementally lower with each earlier end point. Although FRC1/40, and FRC1/30 were not significantly different (1.95 vs 1.93, p=0.07) FRC1/40 and FRC1/20 (1.95 vs 1.90, p=0.002) and FRC1/40 and FRC1/10 (1.95 vs 1.84) were significantly different. Intravisit repeatability LCI1/40 CV%, LCI1/30 CV%, LCI1/20 CV% and LCI1/10 CV% in people with CF were not significantly different to values in healthy controls (table 1). There was also no significant difference between LCI1/40 CV% and the CV% of any other LCI end point in people with CF. A Bland-Altman plot15 of the mean versus the difference between tests one and three of LCI1/40, LCI1/30, LCI1/20 and LCI1/10 for people with CF showed no evidence of greater variability in participants with more advanced disease (ie, a higher LCI reading; figure 1A–D).

CV% of any other LCI end point in people with CF. A Bland-Altman plot15 of the mean versus the difference between tests one and three of LCI1/40, LCI1/30, LCI1/20 and LCI1/10 for people with CF showed no evidence of greater variability in participants with more advanced disease (ie, a higher LCI reading; figure 1A–D). Figure 1 (A) Lung Clearance Index (LCI)1/40; (B) LCI1/30; (C) LCI1/20; (D) LCI1/10 first and third test in people with cystic fibrosis (dotted horizontal lines represent the bias and 95% limits of agreement). For LCI1/40, the 95% limits of agreement between the two measurements were −1.70 to 1.33 lung turnovers, compared with −1.33 to 1.07 (LCI1/30), −1.43 to 0.96 (LCI1/20) and −0.83 to 0.66 (LCI1/10) lung turnovers. Therefore the intravisit repeatability of the LCI1/40, LCI1/30, LCI1/20 and LCI1/10 measurements was 1.5, 1.2, 1.2 and 0.7 lung turnovers respectively. Relationship between shortened LCI and ‘standard’ LCI1/40 In people with CF, LCI1/30 (r=0.98, p<0.0001), LCI1/20 (r=0.95, p<0.0001) and LCI1/10 (r=0.88, p<0.0001) correlated significantly with LCI1/40 (figure 2). Figure 2 Shortened Lung Clearance Index (LCI) versus standard LCI1/40 in people with cystic fibrosis. Sensitivity and specificity The upper limit of normal for LCI1/40, LCI1/30, LCI1/20 and LCI1/10 was 7.3, 6.7, 5.9 and 4.6 lung turnovers, respectively, (control mean +1.96 SD). The lower limit of normal of 80% for FEV1% predicted was used, as this is the level that is historically used in clinical practice.

Figure 2 Shortened Lung Clearance Index (LCI) versus standard LCI1/40 in people with cystic fibrosis. Sensitivity and specificity The upper limit of normal for LCI1/40, LCI1/30, LCI1/20 and LCI1/10 was 7.3, 6.7, 5.9 and 4.6 lung turnovers, respectively, (control mean +1.96 SD). The lower limit of normal of 80% for FEV1% predicted was used, as this is the level that is historically used in clinical practice. The sensitivity of LCI1/40, LCI1/30, LCI1/20 to differentiate between people with CF and healthy controls was identical (67%). The sensitivity of LCI1/10 and FEV1% predicted was lower (53% and 47%, respectively). In people with CF, LCI1/40 (r=−0.73, p<0.0001), LCI1/30 (r=−0.70, p<0.0001), LCI1/20 (r=−0.69, p<0.0001) and LCI1/10 (r=−0.62, p=0.0003) correlated significantly with FEV1% predicted (figures 3A–D). Using LCI1/40, 6/30 (20%) people with CF had an abnormal LCI in the presence of a normal FEV1% predicted (figure 3A). Similarly, using LCI1/30 or LCI1/20, 7/30 (23%) people with CF had an abnormal LCI in the presence of a normal FEV1% predicted (figure 3B–C). LCI1/10 was less sensitive, detecting 5/30 (17%) with an abnormal LCI in the presence of a normal FEV1% predicted (figure 3D). Figure 3 Forced expiratory volume in 1 s (FEV1)% predicted versus (A) LCI1/40; (B) LCI1/30; (C) LCI1/20 (D) LCI1/10 (dotted horizontal lines represent the limits of normal for FEV1% predicted (80% predicted) and LCI (LCI1/40:7.3; LCI1/30:6.7; LCI1/20:5.9; LCI1/10:4.6)).

The sensitivity of LCI1/40, LCI1/30, LCI1/20 to differentiate between people with CF and healthy controls was identical (67%). The sensitivity of LCI1/10 and FEV1% predicted was lower (53% and 47%, respectively). In people with CF, LCI1/40 (r=−0.73, p<0.0001), LCI1/30 (r=−0.70, p<0.0001), LCI1/20 (r=−0.69, p<0.0001) and LCI1/10 (r=−0.62, p=0.0003) correlated significantly with FEV1% predicted (figures 3A–D). Using LCI1/40, 6/30 (20%) people with CF had an abnormal LCI in the presence of a normal FEV1% predicted (figure 3A). Similarly, using LCI1/30 or LCI1/20, 7/30 (23%) people with CF had an abnormal LCI in the presence of a normal FEV1% predicted (figure 3B–C). LCI1/10 was less sensitive, detecting 5/30 (17%) with an abnormal LCI in the presence of a normal FEV1% predicted (figure 3D). Figure 3 Forced expiratory volume in 1 s (FEV1)% predicted versus (A) LCI1/40; (B) LCI1/30; (C) LCI1/20 (D) LCI1/10 (dotted horizontal lines represent the limits of normal for FEV1% predicted (80% predicted) and LCI (LCI1/40:7.3; LCI1/30:6.7; LCI1/20:5.9; LCI1/10:4.6)). ROC curve analysis (figure 4) showed that while all LCI values and FEV1% predicted had statistically significant levels of sensitivity and specificity in determining people with CF vs control participants, LCI1/40, LCI1/30 and LCI1/20 had comparable and higher sensitivity and specificity compared with LCI1/10 and FEV1% predicted (table 2). Table 2 AUCROC and 95% CI for LCI1/40, LCI1/30, LCI1/20 LCI1/10 and inverse FEV1% predicted (1.0 indicating best performance, 0.5 indicating poor performance)

ROC curve analysis (figure 4) showed that while all LCI values and FEV1% predicted had statistically significant levels of sensitivity and specificity in determining people with CF vs control participants, LCI1/40, LCI1/30 and LCI1/20 had comparable and higher sensitivity and specificity compared with LCI1/10 and FEV1% predicted (table 2). Table 2 AUCROC and 95% CI for LCI1/40, LCI1/30, LCI1/20 LCI1/10 and inverse FEV1% predicted (1.0 indicating best performance, 0.5 indicating poor performance) AUCROC 95% CI p Value LCI1/40 0.89 0.80 to 0.97 <0.0001 LCI1/30 0.87 0.77 to 0.96 <0.0001 LCI1/20 0.87 0.78 to 0.96 <0.0001 LCI1/10 0.83 0.72 to 0.94 <0.0001 FEV1% predicted (inverse) 0.73 0.60 to 0.86 0.002 AUC, area under the curve; FEV1, forced expiratory volume in 1 s; LCI, Lung Clearance Index, ROC, receiver operating characteristic Figure 4 Receiver operating characteristic ROC curve of Lung Clearance Index (LCI)1/40, LCI1/30, LCI1/20, LCI1/10 and inverse forced expiratory volume in 1 s % predicted: sensitivity and specificity to the presence of cystic fibrosis CF. Test duration Test duration of LCI1/30, LCI1/20 and LCI1/10 was significantly shorter compared with washout duration of LCI1/40 in people with CF (p<0.0001) and in healthy controls (p<0.0001; table 1). In people with CF, the mean (95% CI) time saving per triplicate MBW test was 1 (0.8 to 1.3) minutes or 5% with LCI1/30, 1.9 (1.4 to 2.3) minutes or 9% with LCI1/30 and 3.3 (2.6 to 4.2) minutes or 15% with LCI1/10.

with washout duration of LCI1/40 in people with CF (p<0.0001) and in healthy controls (p<0.0001; table 1). In people with CF, the mean (95% CI) time saving per triplicate MBW test was 1 (0.8 to 1.3) minutes or 5% with LCI1/30, 1.9 (1.4 to 2.3) minutes or 9% with LCI1/30 and 3.3 (2.6 to 4.2) minutes or 15% with LCI1/10. Discussion This study is the first to show that SF6 MBW tests can be reliably shortened. Results show that in children and adults with CF, LCI shortened to 1/30th or 1/20th (LCI1/30 or LCI1/20) of the initial concentration have comparable intravisit repeatability and sensitivity to ‘standard’ LCI at 1/40th of the starting concentration (LCI1/40) providing additional information to FEV1% predicted and offering a time saving. Although repeatable, LCI shortened to 1/10th of the starting concentration (LCI1/10) was less sensitive to lung disease, compared with the other LCI end points. It was, however, still more sensitive than FEV1% predicted.

entration (LCI1/40) providing additional information to FEV1% predicted and offering a time saving. Although repeatable, LCI shortened to 1/10th of the starting concentration (LCI1/10) was less sensitive to lung disease, compared with the other LCI end points. It was, however, still more sensitive than FEV1% predicted. The ‘standard’ end point of 1/40th is based on historic studies using nitrogen washout (2.5%) and has not been systematically validated for MBW tests using SF6.8 9 This study aimed to assess the performance of earlier arbitrary end points compared with the ‘standard’ end point in the SF6 washout, in an attempt to improve the clinical utility of the MBW test by reducing test duration. Like LCI1/40, LCI1/30 and LCI1/20 target the flatter tail of the washout curve, making it unsurprising that similar information can be obtained (see online supplementary file 2). In contrast, when using LCI1/10, the end point occurs before the washout curve flattens. This supports the theory that most information is contained in the tail of the washout curve.16 Therefore a cut-off before this point may provide less information about lung disease severity, as highlighted by the lower sensitivity of LCI1/10 in this study. Yammine et al10 assessed the repeatability and sensitivity of shortened N2 MBW to measure LCI, at a number of earlier end points including 1/20th of the starting concentration and as early as 1/5th of the starting concentration, in 68 children with CF (n=44 with mild disease). In agreement with results from our study, they reported good performance of shortened LCI at 1/20th of the starting concentration compared with ‘standard’ LCI. Furthermore, as with the results from our study, Yammine et al10 found that while the earlier LCI end points had good intravisit repeatability, they were less sensitive and specific to the presence of CF. The authors concluded that shortened N2 MBW to 1/20th of the starting concentration could offer a more feasible measure for use in clinical practice. In this study we extended these observations to investigate and confirm the utility of shortened ventilation indices in older patients with more advanced disease.

of CF. The authors concluded that shortened N2 MBW to 1/20th of the starting concentration could offer a more feasible measure for use in clinical practice. In this study we extended these observations to investigate and confirm the utility of shortened ventilation indices in older patients with more advanced disease. Theoretically, in cases of severe flow asynchrony between the best and the least ventilated lung units, the end-tidal concentrations in subsequent breathing cycles can enhance the contribution of the least ventilated units toward the end of the washout where LCI is measured.16 This may lead to increased variation in results taken at an earlier end point, such as 1/20th of the starting concentration. In contrast, LCI from an earlier end point may be more precise, as an end point where the washout curve slope is greater may avoid random breath-by-breath variability that could be observed at a later end point. In this study, LCI1/30 and LCI1/20 had marginally better sensitivity to the presence of CF (23% vs 20%), compared with LCI1/40, even though there was no significant difference in variability (LCI CV%) between ‘standard’ LCI (LCI1/40) and LCI at an earlier end point (LCI1/30, LCI1/20, LCI1/10). LCI1/30 and LCI1/20 also had good diagnostic performance and superior sensitivity compared with FEV1% predicted. Importantly, these findings indicate that no further additional information was obtained using LCI1/40 compared with the shortened versions, LCI1/30 and LCI1/20.

arlier end point (LCI1/30, LCI1/20, LCI1/10). LCI1/30 and LCI1/20 also had good diagnostic performance and superior sensitivity compared with FEV1% predicted. Importantly, these findings indicate that no further additional information was obtained using LCI1/40 compared with the shortened versions, LCI1/30 and LCI1/20. Test duration for LCI1/30, LCI1/20 and LCI1/10 was significantly shorter than LCI1/40 in people with CF, with an average 5%, 9% and 15% time saving, respectively. As LCI1/20 offers a greater time saving than LCI1/30, while maintaining reliability and sensitivity, it provides the most attractive measure and may enhance the feasibility of MBW in the research and clinical setting. The proportion of time saved in this study is smaller than that reported in the study by Yammine et al10 as MBW tests using an exogenous gas such as SF6 require a washin and a washout phase. However, as these results are from a retrospective analysis, the time saving measurement does not take into account the total time saving per testing session. Finishing a test earlier would allow for the second and third test of the triplicate to start sooner, resulting in a larger time saving. Although the washin time is unchanged, any shortening of LCI test duration could be especially useful in younger children where long assessment periods are not feasible, and in patients with advanced disease, where the washin and/or washout periods can be prolonged. The use of MBW equipment using SF6 has been successfully used in multicentre studies and remains popular as it has advantages in terms of tracer gas estimation (measured directly rather than by subtraction as with N2 MBW) avoiding the potential confounding effects of 100% O2 on breathing pattern. This study is the first to show that MBW using SF6 can be reliably shortened.

y used in multicentre studies and remains popular as it has advantages in terms of tracer gas estimation (measured directly rather than by subtraction as with N2 MBW) avoiding the potential confounding effects of 100% O2 on breathing pattern. This study is the first to show that MBW using SF6 can be reliably shortened. Considering potential limitations of shortened MBW tests, one study highlights that advanced analysis of washout curves to determine the relative contribution of convective and acinar airways to ventilation heterogeneity (phase III analysis) usually requires six lung-volume turnovers.17 However, recent work suggests that the same information may be obtainable in three lung-volume turnovers18 in which case use of LCI1/20 would still enable full phase III analysis. Regardless, phase III indices may have limited utility in CF, as demonstrated by Horsley et al.19 The retrospective nature of this study is a limitation. However, the study did endeavour to avoid selection bias by use of anonymsied patient data sets as consecutively listed in a database and represents the first exploratory study to report on the clinimetric properties of an earlier end point in SF6 MBW. More data from across the disease severity range in CF are required to define normal ranges of shortened LCI. Conclusions LCI1/20 is a repeatable and sensitive test with equal diagnostic performance to LCI1/40 that is shorter, potentially offering a more feasible research and clinical measure.

The retrospective nature of this study is a limitation. However, the study did endeavour to avoid selection bias by use of anonymsied patient data sets as consecutively listed in a database and represents the first exploratory study to report on the clinimetric properties of an earlier end point in SF6 MBW. More data from across the disease severity range in CF are required to define normal ranges of shortened LCI. Conclusions LCI1/20 is a repeatable and sensitive test with equal diagnostic performance to LCI1/40 that is shorter, potentially offering a more feasible research and clinical measure. Supplementary Material Web supplement Web supplement The authors would like to thank the patients and families who participated in this study. Contributors: KO'N recruited patients, collected clinical data, conducted MBW tests and performed lung function assessment; KO'N, DH, JSE and NB analysed data; IB provided statistical analysis support; KO'N, DH, JMB, NB and JSE wrote the paper. Funding: Work funded by HSC Research and Development, Public Health Agency, Northern Ireland and the Medical Research Council through a US-Ireland Partnership Grant. Competing interests: None. Patient consent: Obtained. Ethics approval: This study was approved by the Office for Research Ethics Committees Northern Ireland (ORECNI) (REC reference number: 10/NIR01/41) and co-sponsored by BHSCT and QUB (research office reference number: 10067SE-OPMS). Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages Can primary mesenchymal stem cells be identified and isolated from human lung tissue and where are they localised? Primary mesenchymal stem cells in human lungs are CD105+/CD90+ cell that are perivascularly located tissue-resident cells. The identification and phenotypic characterisation of primary lung MSC is an important first step in identifying the role of MSC in normal lung physiology and pulmonary diseases. Introduction Mesenchymal stem cells (MSC) are multipotent cells, which have been isolated from a variety of tissues such as bone marrow, cord blood, fat and lungs.1–4 In vitro, clonogenic cells, which are denoted as colony-forming units, fibroblast (CFU-F),5 can be assayed as plastic adherent cells giving rise to fibroblastic colonies. These CFU-F cells are considered to reflect the primary MSC and, on further proliferation in culture, their descendants make up the extensively studied cultured mesenchymal stromal cells.6 7 MSC have a number of intriguing properties, that is, high proliferation potential, differentiation capacity into different mesenchymal cell types such as adipocytes, osteoblasts and chondrocytes, capacities to produce a variety of different cytokines, chemokines, extracellular matrix and microvesicles, as well as potent immunomodulatory function.5 8–11

operties, that is, high proliferation potential, differentiation capacity into different mesenchymal cell types such as adipocytes, osteoblasts and chondrocytes, capacities to produce a variety of different cytokines, chemokines, extracellular matrix and microvesicles, as well as potent immunomodulatory function.5 8–11 Owing to these features, MSC have become promising candidates for cellular therapies. For example, with regard to pulmonary diseases, MSC have been demonstrated to reduce inflammation and fibrosis in lung injury mouse models.12 13 MSC have also been proposed as potential precursors for fibroblasts/myofibroblasts, thereby contributing to the development of fibrotic obliteration of small airways.14 15 Fibrotic obliteration of the peripheral airways (obliterative bronchiolitis) is the hallmark of chronic rejection after lung transplantation, affecting approximately 50–60% of patients within 5 years after transplantation and constituting the main cause of long-term morbidity and mortality. Since obliterative bronchiolitis is a patchy disease, the term bronchiolitis obliterans syndrome (BOS) is used clinically and is defined by a decline in lung function and is measured by spirometry and further graded from one to three based on the lung function results.16 17

n cause of long-term morbidity and mortality. Since obliterative bronchiolitis is a patchy disease, the term bronchiolitis obliterans syndrome (BOS) is used clinically and is defined by a decline in lung function and is measured by spirometry and further graded from one to three based on the lung function results.16 17 Previous studies have demonstrated that donor-derived MSC are present in the bronchoalveolar lavage (BAL) fluid from lung-transplanted patients.18 Furthermore, MSC were successfully culture isolated from human lung tissue.2 19 However, little is known about cellular identity and properties of primary ‘bona fide’ MSC in the lung, which is the key information required to investigate the function and physiological/pathophysiological role of MSC in situ. We demonstrate—for the first time—the identity, phenotype and localisation of primary MSC in lung tissues of lung-transplanted patients from central and peripheral transbronchial biopsies. Our results show that lung MSC are tissue-resident cells with typical MSC properties except for in vivo bone formation, thus indicating that pulmonary MSC represent a population of tissue-specific stromal progenitor cells. Furthermore, our results indicate that lung-derived MSC—in contrast to what has been reported for MSC from BAL—do not correlate with the development of BOS in lung-transplanted patients.

pt for in vivo bone formation, thus indicating that pulmonary MSC represent a population of tissue-specific stromal progenitor cells. Furthermore, our results indicate that lung-derived MSC—in contrast to what has been reported for MSC from BAL—do not correlate with the development of BOS in lung-transplanted patients. Methods Isolation of lung cells from lung biopsies Central (bronchial wall) and peripheral transbronchial (parenchymal tissue) biopsies were collected at different time points after transplantation (range 3 months–16 years), and biopsies were preferentially taken from the right lower lobe. Lung function was measured by standard spirometry and BOS grades were defined according to the International Society for Heart and Lung Transplantation guidelines (see online supplementary table S1).20 All patients gave their written informed consent to participate in the study (2005-560). Fresh lung biopsies were cut into smaller pieces and then further dissociated by enzymatic digestion with collagenase type I, 300 U/mL (Gibco BRL, Paisley, USA), hyaluronidase, 1 mg/mL (Fisher Scientific) and DNAse (Qiagen, Solna, Sweden) in Dulbecco’s phosphate buffered saline (DPBS). After washing, lung cells were seeded in NH expansion medium (Miltenyi Biotec, Bergisch Gladbach, Germany) supplemented with 1% antibiotic antimycotic solution (Sigma Aldrich, Stockholm, Sweden) at 37°C, 5% CO2 for generation of lung-derived MSC. Medium was changed after 3 days and weekly thereafter. MSC were passaged with 0.05% trypsin-EDTA (Invitrogen, Lidingö, Sweden) at 70–90% confluence.

ec, Bergisch Gladbach, Germany) supplemented with 1% antibiotic antimycotic solution (Sigma Aldrich, Stockholm, Sweden) at 37°C, 5% CO2 for generation of lung-derived MSC. Medium was changed after 3 days and weekly thereafter. MSC were passaged with 0.05% trypsin-EDTA (Invitrogen, Lidingö, Sweden) at 70–90% confluence. CFU-F assay The frequency of CFU-F in lung cells was determined as described previously.21 22 Briefly, lung cells were isolated as above and plated in a median of 4 (range 1.0–6.0) wells of a six-well plate at 3×104 and 4×104 cells/well. Complete medium changes were performed after 3 and 7 days. On day 14, cells were washed with DPBS, fixed with methanol and stained with 0.5% or 0.1% crystal violet. CFU-F were enumerated microscopically (Nikon TMS microscope (Nikon, Tokyo, Japan) equipped with Infinity1 camera (BergmanLabora AB, Danderyd, Sweden)) as colonies with ≥40 fibroblast-like cells. Wells containing more than 50 colonies were counted as more than 150 CFU-F/100 000 cells.

d with 0.5% or 0.1% crystal violet. CFU-F were enumerated microscopically (Nikon TMS microscope (Nikon, Tokyo, Japan) equipped with Infinity1 camera (BergmanLabora AB, Danderyd, Sweden)) as colonies with ≥40 fibroblast-like cells. Wells containing more than 50 colonies were counted as more than 150 CFU-F/100 000 cells. MSC in vitro differentiation assay Cultured lung-derived MSC were differentiated towards adipocytes, osteoblasts and chondrocytes as described.21 22 Briefly, for adipocyte differentiation, cells were cultured under standard conditions until confluency. Thereafter, cells were cultured for another 21 days in AdipoDiff medium (Miltenyi Biotec), and stained with Oil-Red-O following fixation (Sigma Aldrich). For osteoblast differentiation, MSC were cultured in Dulbecco's modified Eagle's medium (DMEM) high glucose/l-glutamine (PAA) containing β-glycerophosphate (Sigma Aldrich), l-ascorbic acid-2-phosphate (Wako Chemicals, Neuss, Germany) and dexamethasone (Sigma Aldrich) for 21 days and calcium deposition was visualised by alizarin red staining (Sigma Aldrich). For chondrocyte differentiation, cell pellets were cultured in complete chondrocyte induction medium consisting of l-ascorbic acid-2-phosphate, pyruvic acid sodium salt (Sigma Aldrich), l-proline (Sigma Aldrich), ITS+ culture supplement (BD Bioscience, Erembodegem, Belgium) and TGF-β3 (R&D Systems, Abingdon, UK). After 28 days, pellets were fixed and embedded in OCT compound (Tissue-Tek, Sakura, Zoeterwoude, The Netherlands). The pellets were sectioned and stored at −80°C. For analysis, sections were stained with goat antihuman aggrecan (R&D Systems) and a corresponding secondary antibody (Donkey anti-goat IgG, Texas Red (JackssonEurope, Suffolk, UK)). Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI). Pictures of adipocyte and osteoblast differentiation were taken with a Nikon Eclipse TE2000-E microscope equipped with a Nikon DS-U2/L2 USB camera and the chondrocyte differentiation was documented with an Axiovert 200M fluorescence microscope and an AxioCam HRm camera.

tained with 4′,6-diamidino-2-phenylindole (DAPI). Pictures of adipocyte and osteoblast differentiation were taken with a Nikon Eclipse TE2000-E microscope equipped with a Nikon DS-U2/L2 USB camera and the chondrocyte differentiation was documented with an Axiovert 200M fluorescence microscope and an AxioCam HRm camera. Gene expression analysis of cultured lung MSC Cultured lung MSC (passage 4) were harvested with 0.05% trypsin-EDTA and total RNA was isolated with the RNeasy mini kit (Qiagen, GmBH, Hilden, Germany). The amount of RNA was measured using NanoDrop ND-100 (Nano Drop Technologies, Delaware, Maryland, USA) and 1 μg RNA was used for cDNA synthesis. Superscript II (Invitrogen, Carlsbad, California, USA) was used to reverse-transcribe RNA and the cDNA was stored at −80°C. Then the cDNA was mixed with primers (see online supplementary material) and Fast CYBR Green Master Mix 2X (Applied Biosystems). Reverse transcription-PCR reactions were performed with the StepOnePlus Real Time PCR system (Applied Biosystems) starting at 50°C for 2 min and 95°C for 2 min, followed by 45 cycles of 95°C for 15 s, 60°C for 25 s and 73°C for 30 s, and finished by 95°C for 15 s, 70°C for 15 s and 98°C for 15 s. Relative quantifications were performed according to 2−Δ ΔCt method using GAPDH as a housekeeping gene and centrally derived MSCs as reference gene.

50°C for 2 min and 95°C for 2 min, followed by 45 cycles of 95°C for 15 s, 60°C for 25 s and 73°C for 30 s, and finished by 95°C for 15 s, 70°C for 15 s and 98°C for 15 s. Relative quantifications were performed according to 2−Δ ΔCt method using GAPDH as a housekeeping gene and centrally derived MSCs as reference gene. MSC in vivo differentiation assay In vivo transplantation of lung-derived MSC was performed as described previously.23 Briefly, central and peripheral lung MSC from two patients were harvested, incubated with hydroxyapatite/tricalcium phosphate (HA/TCP) ceramic powder overnight at 37°C in 5% CO2 and 500 000 cells were implanted subcutaneously into 8-week-old female non-obese diabetic/severe combined immunodeficient (NOD/SCID) mice (four implants per culture). Implants were removed after 8 weeks, fixed, decalcified and paraffin embedded. Implants from one animal had to be removed 4 days earlier due to development of thymic lymphoma. Sections were stained with H&E and analysed as described.24 Photomicrographs were taken with a Leica DM4500B microscope equipped with a motorised stage from Märzhäuser Wetzlar GmbH and a Leica DFC300 camera, and controlled by the Surveyor software (Objective Imaging).

days earlier due to development of thymic lymphoma. Sections were stained with H&E and analysed as described.24 Photomicrographs were taken with a Leica DM4500B microscope equipped with a motorised stage from Märzhäuser Wetzlar GmbH and a Leica DFC300 camera, and controlled by the Surveyor software (Objective Imaging). Immunophenotyping Cells were detached with 0.05% trypsin-EDTA and blocked with phosphate buffered saline (PBS) containing 1% fetal bovine serum (FCS; Gibco BRL), 3.3 mg/mL human immunoglobulin (Gammanorm; Octapharma, Stockholm, Sweden) and 0.1% sodium azide. Cells were labelled with different combinations of the following direct conjugated antibodies: anti-CD105-FITC, anti-CD34-FITC, anti-HLA-DR-FITC, anti-CD31-FITC, anti-CD73-PE, anti-CD14-PE, anti-19-PE, anti-HLAclass I-PE, anti-CD146-PE, anti-CD90-APC, anti-CD45-APC (all from BD Bioscience) and anti-CD271-APC (Miltenyi Biotec). Corresponding isotype controls were all from Becton Dickinson. Before analysis, cells were stained with 7-amino-actinomycin D (1 μL/mL; Sigma Aldrich) for dead cell exclusion. Samples were analysed on an FACSCalibur (BD Bioscience). Data acquisition and analysis were performed using CellQuest (BD Bioscience) and FlowJo software (Tree star, Ashland, Oregon, USA).

Becton Dickinson. Before analysis, cells were stained with 7-amino-actinomycin D (1 μL/mL; Sigma Aldrich) for dead cell exclusion. Samples were analysed on an FACSCalibur (BD Bioscience). Data acquisition and analysis were performed using CellQuest (BD Bioscience) and FlowJo software (Tree star, Ashland, Oregon, USA). Fluorescence-activated cell sorting of primary lung cells Following blocking with PBS containing 1% FCS and 3.3 mg/mL human immunoglobulin, single lung cell suspensions were stained with combinations of the following antibodies: anti-CD90-APC/FITC and anti-CD105-FITC/APC (BD Bioscience) or anti-CD146-PE with anti-CD271-APC. After washing, cells were incubated with 7-amino-actinomycin D (1 μL/mL) and sorted on an FACSAria I or FACSAria III cell sorter (both from BD Bioscience). Single stained cells were used to set up sorting gates and doublets were excluded by gating on SSC-H versus SSC-W and FSH-H versus FSC-W. Sorted cells were collected in NH medium supplemented with 1% antibiotic antimycotic solution and gentamicin (Gibco BRL) and assayed for CFU-F as described above; colonies with ≥20 cells were counted as CFU-F. In total, 11 sorting experiments were performed on CD105 and CD90 stained cells. However, five of the experiments were not evaluable because of insufficient cell yield after sorting.

antimycotic solution and gentamicin (Gibco BRL) and assayed for CFU-F as described above; colonies with ≥20 cells were counted as CFU-F. In total, 11 sorting experiments were performed on CD105 and CD90 stained cells. However, five of the experiments were not evaluable because of insufficient cell yield after sorting. Karyotyping and X and Y chromosomes fluorescence in situ hybridisation of lung-derived MSC Central and peripheral MSC from seven sex-mismatched lung-transplanted patients were harvested, fixed in methanol:glacial acetic acid (3:1) and spread on slides. After spreading, the slides were kept at 60°C overnight and then treated with 2T (2×saline–sodium citrate (SCC) buffer with 0.05% Tween-20 (Sigma Aldrich)) at 60°C. Slides were then incubated with 20 mg/mL pepsin in 0.01M HCl at 37°C for 10 min, washed and incubated with 1% formaldehyde in PBS, washed and dehydrated. LSI SRY and CEP X probes (Vysis, Downers Grove, Illinois, USA) were added to the slides and denaturation was performed at 74°C. Hybridisation was carried out overnight at 37°C in a moist chamber. Posthybridisation washes were carried out in 0.4T (0.4×SSC with 0.05% Tween-20) in a 72°C water bath. Slides were dehydrated and mounted with DABCO/DAPI (DABCO, Sigma Aldrich; DAPI, ROCHE, Bromma, Sweden). One hundred nuclei were analysed per sample.

. Hybridisation was carried out overnight at 37°C in a moist chamber. Posthybridisation washes were carried out in 0.4T (0.4×SSC with 0.05% Tween-20) in a 72°C water bath. Slides were dehydrated and mounted with DABCO/DAPI (DABCO, Sigma Aldrich; DAPI, ROCHE, Bromma, Sweden). One hundred nuclei were analysed per sample. Karyotyping of the MSC cultures was performed as described.25 Briefly, cultured MSC derived from central biopsies of two sex-mismatched lung-transplanted patients were arrested in metaphase with 0.04 µg/mL colcemid. Then the cells were fixed, prepared on slides and chromosomes were G-banded using Wright’s stain. Twenty-five metaphases were analysed per sample. Images were taken with a Zeiss Axioplan 2 microscope (Carl Zeiss AG, Oberkochen, Germany) using CytoVision software (Leica Biosystems, Nussloch, Germany).

04 µg/mL colcemid. Then the cells were fixed, prepared on slides and chromosomes were G-banded using Wright’s stain. Twenty-five metaphases were analysed per sample. Images were taken with a Zeiss Axioplan 2 microscope (Carl Zeiss AG, Oberkochen, Germany) using CytoVision software (Leica Biosystems, Nussloch, Germany). Immunohistochemistry Paraffin-embedded tissue samples were sectioned into 4.5 μm thick sections. Sections were rehydrated and antigen retrieval was performed in high pH buffer (EnVision FLEX target retrieval solution (K8004, Dako, Glostrup, Denmark)) in microwave. Endogenous peroxidase activity was quenched with 0.5% H2O2. CD105 was detected by an anti-CD105 mouse monoclonal antibody (Clone 4G11, Leica Microsystems, Newcastle, UK) diluted in an EnVisionFLEX antibody diluent (K8006, Dako) and the signal was enhanced with the Tyramid Signal Amplification kit (T20912, Molecular Probes) and visualised with EnVision labelled ploymere HRP (K4001, Dako). CD90 expression was detected with an anti-CD90 rabbit monoclonal antibody (Clone EP56, Epitomics, California, USA) diluted in an EnVisionFLEX antibody diluent and visualised with a goat antirabbit Alexa Fluor 555-conjugated secondary antibody. Nuclei were stained with DAPI and sections were mounted. Staining was absent in sections using isotype control antibodies (see online supplementary figure S5A–F). Pictures were taken with a Nikon TE200E microscope equipped with a high-resolution camera Nikon DXM 1200C using the NIS-Elements AR V.3.0 system (Nikon).

dy. Nuclei were stained with DAPI and sections were mounted. Staining was absent in sections using isotype control antibodies (see online supplementary figure S5A–F). Pictures were taken with a Nikon TE200E microscope equipped with a high-resolution camera Nikon DXM 1200C using the NIS-Elements AR V.3.0 system (Nikon). Statistics CFU-F data were analysed statistically by using the Mann-Whitney U test to compare central and peripheral CFU-F numbers. Results are presented as median and range. Analyses were performed with the GraphPad Prism software V.5.0c. Two Kaplan-Meier analyses with BOS as the outcome variable and time, status and groups CFU-F central (<50 = 0 vs >50=1) and CFU-F peripheral (<10=0 vs >10=1), respectively, were conducted.26 Furthermore, data were analysed for correlation by a predictive model built on multivariate logistic regression. Kaplan-Meier and regression analyses and correlation analysis were conducted with the IBM SPSS Statistics V.19.0 software package. p-Values ≤0.05 were considered as significant. A detailed description of the statistical analysis is provided in the online supplementary material.

uilt on multivariate logistic regression. Kaplan-Meier and regression analyses and correlation analysis were conducted with the IBM SPSS Statistics V.19.0 software package. p-Values ≤0.05 were considered as significant. A detailed description of the statistical analysis is provided in the online supplementary material. Results Primary MSC can be isolated from central and peripheral transbronchial biopsies in lung-transplanted patients Primary mesenchymal stem/progenitor cells are considered to be reflected by in vitro clonogenic cells—denoted as CFU-F. In order to determine the mesenchymal progenitor frequency in lung biopsies, single-cell suspensions from central and peripheral transbronchial biopsies of 27 lung-transplanted patients (in some of the patients, biopsies were collected at more than one time point after transplantation) were assayed in standard CFU-F assays (figure 1). Colony formation was observed in all but two central biopsies (n=30 cultures) and in 24 of 31 cultures from peripheral transbronchial biopsies. As shown in figure 1A, mesenchymal progenitor frequency was significantly higher (p=0.0254) in central biopsies (median 49.86 CFU-F/105 seeded cells; range 0.0–166.7) compared with peripheral tissues (median 12.05; range 0.0–166.7). The median intra-assay variations of centrally derived cells were 9.8% (range 0–173.2) and 8.6% (range 0–173.2) at a seeding density of 3×104 and 4×104 cells/well, respectively. For peripherally derived cells, the median intra-assay variations were 9.7 (range 0–173.2) and 5.1 (range 0–173.2) at a seeding density of 3×104 and 4×104 cells/well, respectively. The mesenchymal colonies assayed from lung tissues showed typical CFU-F morphology as shown in figure 1B.

d 4×104 cells/well, respectively. For peripherally derived cells, the median intra-assay variations were 9.7 (range 0–173.2) and 5.1 (range 0–173.2) at a seeding density of 3×104 and 4×104 cells/well, respectively. The mesenchymal colonies assayed from lung tissues showed typical CFU-F morphology as shown in figure 1B. Figure 1 Lung-derived cultured mesenchymal stem cells (MSC) show a typical surface marker profile and have a higher colony-forming unit, fibroblast (CFU-F) frequency in lung MSC derived from central biopsies compared with peripheral transbronchial biopsies. Primary cells were isolated from the central and peripheral transbronchial lung biopsies of 27 lung-transplanted patients. Single-cell suspensions were assayed for CFU-F content under standard conditions for 14 days. (A) Primary lung cells derived from central biopsies displayed a significantly higher CFU-F frequency compared with peripheral transbronchially derived cells (p=0.0254). Median CFU-F frequencies of individual biopsies are shown as colonies/100 000 seeded cells. Closed circles represent CFU-F numbers from biopsies of individual patients undergoing lung biopsies once after transplantation, whereas open circles represent CFU-F numbers of repetitive biopsies of patients undergoing bronchoscopies at different time points after transplantation. The Mann-Whitney U test was used to calculate p values. *p<0.05. Colonies with ≥40 cells were counted as CFU-F. (B) A typical lung-derived colony stained with crystal violet. (C) Cultured lung MSC (passages 3 and 4) were harvested, stained with surface marker antibodies and analysed with flow cytometry. A representative surface marker profile from central lung-derived MSC of one patient is presented. Sample: red shaded area. Isotype control: grey line.

ed colony stained with crystal violet. (C) Cultured lung MSC (passages 3 and 4) were harvested, stained with surface marker antibodies and analysed with flow cytometry. A representative surface marker profile from central lung-derived MSC of one patient is presented. Sample: red shaded area. Isotype control: grey line. The Kaplan-Meier analysis revealed no significant differences between the number of central CFU-F in relation to time to event (BOS; mean 2.79 years, 95% CI 1.52 to 4.05 and mean 3.01 years, 95% CI 0.99 to 5.03, respectively, p=0.508; see online supplementary figure S1A). In the second Kaplan-Meier analysis, no differences between the number of peripheral CFU-F in relation to time to event (BOS) were seen (mean 2.97 years, 95% CI 1.69 to 4.26 and mean 3.25 years, 95% CI 1.33 to 5.17, respectively, p=0.541; see online supplementary figure S1B). Furthermore, a logistic regression (enter model) analysis demonstrated that none of the tested independent variables were significantly correlated with the number of CFU-F in central or peripheral biopsies (see online supplementary table S2).

CI 1.33 to 5.17, respectively, p=0.541; see online supplementary figure S1B). Furthermore, a logistic regression (enter model) analysis demonstrated that none of the tested independent variables were significantly correlated with the number of CFU-F in central or peripheral biopsies (see online supplementary table S2). Lung-derived mesenchymal stromal cells show a typical MSC surface marker profile and multilineage differentiation capacity in vitro but lack bone differentiation capacity in vivo On further proliferation in culture, CFU-F give rise to the so-called cultured mesenchymal stromal cells.1 18 27 Immunophenotyping of lung-derived cultured mesenchymal stromal cells (passages 3 and 4) using multicolour flow cytometry showed a typical MSC surface marker profile, that is, lung-derived MSC were positive for CD105, CD73, CD90, HLA-class I and CD146, but lacked expression of CD45, CD34, CD14, CD19, HLA-DR, CD31 and CD271 (figure 1C). There were no differences in surface marker expression between central and peripheral transbronchial biopsy MSC (see online supplementary figure S2).

that is, lung-derived MSC were positive for CD105, CD73, CD90, HLA-class I and CD146, but lacked expression of CD45, CD34, CD14, CD19, HLA-DR, CD31 and CD271 (figure 1C). There were no differences in surface marker expression between central and peripheral transbronchial biopsy MSC (see online supplementary figure S2). In vitro differentiation experiments revealed multilineage differentiation potential of lung-derived MSC from central and peripheral transbronchial biopsies, that is, cells differentiated into adipocytes, osteoblasts and chondrocytes. Adipocyte and osteoblast differentiation was observed in four of six and five of six cultures of MSC derived from central and peripheral biopsies, respectively (figure 2A,B). Chondrocyte differentiation was observed for central biopsy-derived cells (in one of three cultures; figure 2C). No differentiation was seen in the negative control (figure 2D–F). In vitro differentiation capacity was further confirmed by gene expression analysis (figure 2G–I). Expression of PPARG (peroxisome proliferator-activated receptor γ), ALPL (alkaline phosphatase) and ACAN (aggrecan) was seen in centrally and peripherally derived cultures. The in vivo differentiation potential of lung-derived MSC was investigated by xenotransplantation of cultured MSC together with HA/TCP carrier particles subcutaneously into NOD/SCID mice. Lung-derived MSC formed adipocytes and stromal tissues in vivo. However, bone formation was clearly impaired or even absent in most of the slides evaluated. Generally, only small areas of possible bone tissue were detectable in lung MSC transplanted animals (figure 3A,B). Control telomerase-immortalised bone marrow MSC, on the other hand, showed clear bone formation (figure 3C). HA/TCP carrier particles without cells served as negative controls (figure 3D).

ides evaluated. Generally, only small areas of possible bone tissue were detectable in lung MSC transplanted animals (figure 3A,B). Control telomerase-immortalised bone marrow MSC, on the other hand, showed clear bone formation (figure 3C). HA/TCP carrier particles without cells served as negative controls (figure 3D). Figure 2 Lung-derived mesenchymal stem cells (MSC) display in vitro multilineage potential. MSC derived from central and peripheral transbronchial lung biopsies obtained from six lung-transplanted patients were seeded in an appropriate differentiation-induction medium. Lung-derived MSC differentiated into adipocytes (A), osteoblasts (B) and chondrocytes (C). Adipocytes were stained with Oil-Red-O staining (A and D), osteoblasts with alizarin red (B and E) and chondrocytes with an antiaggrecan antibody (C and F). Control cells were cultured in normal growth medium (D–F). Scale bare for adipocytes, osteoblast and chondrocyte pictures represents 20, 100 and 50 μm, respectively. To confirm in vitro differentiation, we performed reverse transcription-PCR on cultured MSC obtained from three lung-transplanted patients (G–I). Fold change in mRNA expression of peripheral transbronchially derived cells (white) is shown for peroxisome proliferator-activated receptor γ (PPARG; G), alkaline phosphatase (ALPL; H) and aggrecan (ACAN; I) using centrally derived cells (black) as the reference gene.

tained from three lung-transplanted patients (G–I). Fold change in mRNA expression of peripheral transbronchially derived cells (white) is shown for peroxisome proliferator-activated receptor γ (PPARG; G), alkaline phosphatase (ALPL; H) and aggrecan (ACAN; I) using centrally derived cells (black) as the reference gene. Figure 3 Lung-derived mesenchymal stem cells (MSC) display impaired bone formation capacity in vivo. Cultured MSC derived from two lung-transplanted patients were subcutaneously transplanted (together with hydroxyapatite/tricalcium phosphate (HA/TCP) particles) into NOD/SCID mice. After 8 weeks, the transplants were removed, fixated and stained with H&E (A, B). Lung-derived MSC clearly formed adipocytes (a), stroma with invading haematopoietic cells (s) and an extracellular matrix (ECM). However, bone formation was hardly observable (*) compared with the bone formation (b) seen in the positive control (C, telomerase-immortalised human MSC). Arrow indicates a megakaryocyte. HA implants without cells served as negative controls (D). Scale bars indicate 100 μm.

cells (s) and an extracellular matrix (ECM). However, bone formation was hardly observable (*) compared with the bone formation (b) seen in the positive control (C, telomerase-immortalised human MSC). Arrow indicates a megakaryocyte. HA implants without cells served as negative controls (D). Scale bars indicate 100 μm. Lung MSC are tissue-resident In order to examine if lung MSC were tissue resident and originated from donor lungs, we performed fluorescence in situ hybridisation (FISH) analysis on cultured central and peripheral transbronchial-derived cells (passages 3 and 4) from sex-mismatched lung-transplanted patients (n=7; figure 4). MSC were isolated from biopsies taken as soon as 3 months, and as late as nearly 16 years after transplantation (see online supplementary table S3). All evaluated MSC samples showed donor sex karyotype (median 97%; range 93–100%; figure 4, see online supplement table S3). There was no difference between MSC derived from central biopsies (figure 4A, C) compared with peripheral transbronchial biopsies (figure 4B, D). In addition, G-band analysis was performed on passage 4 MSC samples, confirming the results of the FISH analysis and furthermore demonstrating that cultured lung-derived MSC had a normal karyotype (figure 4E).

derived from central biopsies (figure 4A, C) compared with peripheral transbronchial biopsies (figure 4B, D). In addition, G-band analysis was performed on passage 4 MSC samples, confirming the results of the FISH analysis and furthermore demonstrating that cultured lung-derived MSC had a normal karyotype (figure 4E). Figure 4 Lung mesenchymal stem cells (MSC) are donor derived and tissue resident. Cultured MSC isolated from central (A and C) and peripheral transbronchial (B and D) biopsies of seven sex-mismatched lung-transplanted patients were harvested and analysed by fluorescence in situ hybridisation. LSI SRY (orange) and CEP X (green) probes were used in order to distinguish donor-derived cells from recipient cells, with the presence of Y and X chromosomes indicated by the red and green signal, respectively. A and B show that the lung-derived MSC from male patient originated from the female lung donor. C and D show male donor MSC in a female recipient. Furthermore, karyotyping was performed on MSC from central biopsies of two patients demonstrating a normal karyotype as exemplified in E.

red and green signal, respectively. A and B show that the lung-derived MSC from male patient originated from the female lung donor. C and D show male donor MSC in a female recipient. Furthermore, karyotyping was performed on MSC from central biopsies of two patients demonstrating a normal karyotype as exemplified in E. Primary pulmonary MSC are enriched in the CD90/CD105 cell fraction Next, we wanted to investigate the phenotype of the primary lung MSC and evaluate if these cells could be isolated directly from lung tissues by fluorescence-activated cell sorting (FACS) based on surface markers previously described for MSC isolation. First, primary lung cells were sorted based on the expression of CD146 and CD271 (n=4), a surface marker combination that has previously been used by us and others for the isolation of primary MSC from human bone marrow.21 28 However, and in contrast to bone marrow, lung CFU-F were not only found in the CD271 single positive fractions but also in double negative cells in three of four experiments. In addition, CFU-F were even occasionally found in double positive and CD146 single positive cells, respectively, indicating that CD271/CD146 is not a suitable surface marker combination to identify lung CFU-F (see online supplementary figure S3A–E). On the other hand, when primary lung mononuclear cells were sorted based on CD90/CD105 expression (n=6), CFU-F were consistently found in the double positive fraction (median 8.05 colonies/1000 primary cells, range 0.35–37.59), but only very occasionally in the CD90−/CD105 (median 0.84 colonies/1000 primary cells, range 0.0–46.67) and CD90/CD105− fractions (median 0 colonies/1000 primary cells, range 0.0–8.49). Moreover, CFU-F were not detected in CD90−/CD105− double negative cells (figure 5A).

colonies/1000 primary cells, range 0.35–37.59), but only very occasionally in the CD90−/CD105 (median 0.84 colonies/1000 primary cells, range 0.0–46.67) and CD90/CD105− fractions (median 0 colonies/1000 primary cells, range 0.0–8.49). Moreover, CFU-F were not detected in CD90−/CD105− double negative cells (figure 5A). Figure 5 Primary mesenchymal stem cells (MSC) are enriched in the CD90/CD105 double positive cells in human lung tissue and are located perivascularly. (A) Single-cell suspensions were isolated from central and peripheral transbronchial lung biopsies of lung-transplanted patients, stained with direct conjugated antibodies against CD90 (Thy-1) and CD105 (Endoglin), and sorted by fluorescent-activated cell sorting. CD90/CD105, CD90−/CD105, CD90/CD105− and CD90−/CD105− cell populations were sorted and assayed for colony-forming unit, fibroblast (CFU-F) content. Colonies were enumerated after 13 and 14 days in culture, showing that the majority of CFU-F was contained in the CD90/CD105 double positive cell fraction. Colonies with ≥20 cells were counted as CFU-F. Statistical analysis was performed using a non-parametric Mann–Whitney U test. Horizontal lines indicate medians for each sorted cell fraction. *p<0.05, **p<0.01. (B) Paraffin embedded sections from lung-transplanted patients were stained with antibodies against CD90 and CD105. CD90/CD105 double positive cells were observed in perivascular regions. Arrow indicates a CD90/CD105 cell. (C–F) Enlargement of a CD90 (F) and CD105 (E) double positive cell, the nuclei is stained with 4′,6-diamidino-2-phenylindole (D). EP, small airway epithelium. Scale bar represents 20 μm.

es against CD90 and CD105. CD90/CD105 double positive cells were observed in perivascular regions. Arrow indicates a CD90/CD105 cell. (C–F) Enlargement of a CD90 (F) and CD105 (E) double positive cell, the nuclei is stained with 4′,6-diamidino-2-phenylindole (D). EP, small airway epithelium. Scale bar represents 20 μm. In situ localisation of CD90/CD105 cells in human lung tissue In order to investigate the in situ localisation of resident lung CD90/CD105 CFU-F, sections from lung-transplanted patients were analysed by antibody staining. Generally, only very few CD90/CD105 double positive cells were found, which correlates well with the low number of these cells in the sorting experiments as well as with CFU-F frequencies in lung tissues. In accordance with the reported MSC locations in other organs, double positive CD90/CD105 lung cells were found to be located perivascularly (figure 5B).

uble positive cells were found, which correlates well with the low number of these cells in the sorting experiments as well as with CFU-F frequencies in lung tissues. In accordance with the reported MSC locations in other organs, double positive CD90/CD105 lung cells were found to be located perivascularly (figure 5B). Discussion MSC have not only been implicated in the development of lung diseases, but they have also been proposed as a future cell-based therapy for lung diseases, such as obliterative bronchiolitis. Despite this potentially important role of MSC in lung physiology, relatively little is known about the cellular identity and biological function of the primary human lung MSC in health and disease. Thus far, only culture-derived stromal cells from BAL fluid,18 fetal and adult healthy lung tissues have been investigated.2 19 29 We therefore aimed to isolate primary pulmonary MSC in lung tissue from transplanted patients in order to investigate their origin and anatomical localisation within the lung tissue, as well as to characterise their mesenchymal properties in vivo and in vitro.

d adult healthy lung tissues have been investigated.2 19 29 We therefore aimed to isolate primary pulmonary MSC in lung tissue from transplanted patients in order to investigate their origin and anatomical localisation within the lung tissue, as well as to characterise their mesenchymal properties in vivo and in vitro. Our results demonstrate that mesenchymal stem/progenitor cells, as assessed by CFU-F assays, were present in central and peripheral transbronchial biopsies. CFU-F frequencies were significantly higher in the central biopsies indicating, in accordance with what has been reported previously,30 that there are differences in lung structure between central and peripheral sections. Interestingly, there was no correlation between CFU-F number and BOS grade, indicating that lung-derived tissue-resident MSC could not be used as a predictor for BOS development. Badri et al,31 on the other hand, reported previously that the measurement of CFU-F in BAL fluid provided predictive information regarding future BOS onset in lung-transplanted patients. This finding leads to the assumption that CFU-F numbers in transplanted lung tissues might be independent from CFU-F numbers in BAL fluids. However, as we have not directly compared these two cell sources, we cannot draw any definite conclusions at this point of time. Alternatively, lung tissue and BAL MSC might have biological differences and therefore react differently to pathological stress situations. Certainly, this is an important consideration which is worth addressing in future experiments.

hese two cell sources, we cannot draw any definite conclusions at this point of time. Alternatively, lung tissue and BAL MSC might have biological differences and therefore react differently to pathological stress situations. Certainly, this is an important consideration which is worth addressing in future experiments. One of the fundamental properties of MSC is their multilineage differentiation capacity, which is commonly evaluated only by in vitro assays. Our data show that pulmonary MSC possessed standard multilineage differentiation potential in vitro, thus formally fulfilling the requirements for MSC definition, but they lacked proper bone formation in vivo. This indicates that lung MSC and bone marrow-derived MSC are biologically different as bone marrow MSC robustly formed bone in xenotransplantation experiments.21 Based on the fact that the differentiation spectrum of bone marrow-derived MSC also differs from other MSC preparations, such as adipocyte-derived and umbilical vein-derived stromal cells, it is reasonable to conclude that MSC are tissue-specific.32 Furthermore, our observation that lung-derived MSC possessed a full three-lineage potential only in vitro, which is most likely due to supraphysiological levels of differentiation-inducing signals in standard in vitro cultures, stresses the importance of in vivo experiments to assess proper differentiation potential.

One of the fundamental properties of MSC is their multilineage differentiation capacity, which is commonly evaluated only by in vitro assays. Our data show that pulmonary MSC possessed standard multilineage differentiation potential in vitro, thus formally fulfilling the requirements for MSC definition, but they lacked proper bone formation in vivo. This indicates that lung MSC and bone marrow-derived MSC are biologically different as bone marrow MSC robustly formed bone in xenotransplantation experiments.21 Based on the fact that the differentiation spectrum of bone marrow-derived MSC also differs from other MSC preparations, such as adipocyte-derived and umbilical vein-derived stromal cells, it is reasonable to conclude that MSC are tissue-specific.32 Furthermore, our observation that lung-derived MSC possessed a full three-lineage potential only in vitro, which is most likely due to supraphysiological levels of differentiation-inducing signals in standard in vitro cultures, stresses the importance of in vivo experiments to assess proper differentiation potential. We analysed lung-derived MSC by XY chromosome analysis and our data clearly demonstrate that pulmonary MSC in lung-transplanted patients were donor derived. Interestingly, different origins of pulmonary MSC have been discussed, with one of the hypotheses being that MSC are recruited from the bone marrow through the circulation.13 33 However, our results in accordance with data previously reported by Lama et al18 provide compelling evidence that lung-derived MSC are tissue-resident cells. Some of the analysed MSC were isolated from biopsies taken as long as nearly 16 years after transplantation, indicating that lung-resident MSC either are very long-lived cells or that they might possess self-renewal capacity, which is certainly an important point given the recent data providing evidence for in vivo self-renewal of murine and human MSC.28 34

from biopsies taken as long as nearly 16 years after transplantation, indicating that lung-resident MSC either are very long-lived cells or that they might possess self-renewal capacity, which is certainly an important point given the recent data providing evidence for in vivo self-renewal of murine and human MSC.28 34 Until now, only culture-derived lung MSC have been studied and the phenotype of the primary lung MSC has thus far been elusive. We identify—for the first time—the phenotype of primary lung-derived MSC. Using FACS sorting of primary lung cells, we could demonstrate that the CD90/CD105 cell fraction was highly enriched for CFU-F. Based on our previous results on the isolation of CFU-F in primary bone marrow-derived cells, we initially started sorting based on CD271 and CD146 expression,21 but this combination did not allow to enrich for lung CFU-F, which furthermore strengthens the tissue specificity of MSC. Owing to the fact that only limited numbers of cells were available for sorting due to the small size of the tissue biopsies, it was not possible to reanalyse the sorted cell fractions. Therefore, colonies growing from single positive CD105 cells might be the result of sorting impurities. Here, the sorting of a larger number of cells, for example, from lung explants would be necessary, which is certainly an important consideration for future experiments.

ossible to reanalyse the sorted cell fractions. Therefore, colonies growing from single positive CD105 cells might be the result of sorting impurities. Here, the sorting of a larger number of cells, for example, from lung explants would be necessary, which is certainly an important consideration for future experiments. On the basis of the identification of the phenotype of primary lung MSC, we then could address the important question of where these cells are located in situ in the lung tissue. Here, our experiments showed that CD90/CD105 double positive cells were found exclusively perivascularly. The perivascular location of the pulmonary MSC is thus in accordance with what has been demonstrated in other organs such as bone marrow.28 35 36

uestion of where these cells are located in situ in the lung tissue. Here, our experiments showed that CD90/CD105 double positive cells were found exclusively perivascularly. The perivascular location of the pulmonary MSC is thus in accordance with what has been demonstrated in other organs such as bone marrow.28 35 36 Several preclinical studies have evaluated the therapeutic potential of MSC in different lung disease models. Ortiz et al12, for example, demonstrated that the administration of bone marrow-derived MSC into bleomycin-treated mice resulted in decreased inflammation and collagen deposition. Furthermore, Rojas et al13 reported an increased production of growth factors, a better survival and reduced inflammation in bleomycin-induced lung fibrosis in mice treated with bone marrow-derived MSC. Bone marrow-derived MSC have already been used clinically in humans for the treatment of inflammatory diseases such as severe graft-versus-host disease in haematopoietic stem cell transplantation.37 Autologous bone marrow-derived MSC have also been used to improve graft function in solid organ transplantation and MSC are furthermore under active investigation in non-transplant settings for the treatment of various diseases, such as myocardial infarction38 and multiple sclerosis.39 Furthermore, bone marrow-derived MSC have been used in clinical trials for treatment of chronic obstructive pulmonary disease where it was proved to be a safe treatment.40 Clinical MSC treatment has been reported to be effective and safe and has therefore also been suggested as a potential treatment option for inflammatory lung diseases. Based on our data showing differences in the differentiation potential of bone marrow and lung MSC, we would suggest that pulmonary-derived MSC might be more appropriate for the treatment of lung diseases. It is tempting to speculate that lung-derived MSC could be used as a cell-based therapy for treatment of lung diseases like chronic rejection. Lung-derived MSC for clinical use could, for example, be generated from normal, non-diseased human adult or fetal lungs, the latter of which is an excellent source for MSC.41 However, the availability of both cell sources is limited and, furthermore, the use of fetal tissues raises ethical considerations. Alternative approaches such as iPS-derived MSC might be alternatives and should thus be investigated.

ased human adult or fetal lungs, the latter of which is an excellent source for MSC.41 However, the availability of both cell sources is limited and, furthermore, the use of fetal tissues raises ethical considerations. Alternative approaches such as iPS-derived MSC might be alternatives and should thus be investigated. Cultured lung-derived MSC have been proved to inhibit T-cell proliferation;42 however, the in vitro and in vivo immunomodulatory effects of lung MSC have to be studied in more detail in comparison with bone marrow-derived cells before any final conclusions can be drawn. In summary, our study demonstrates that tissue-resident MSC can be isolated from central and peripheral transbronchial lung biopsies of lung-transplanted patients. Furthermore, we show—for the first time—that primary lung mesenchymal progenitor cells are highly enriched in the CD90/CD105 double positive cell fraction and that these cells are located perivascularly. Cultured lung MSC displayed multilineage potential both in vivo and in vitro, but showed impaired bone formation in vivo, indicating that lung-derived MSC are phenotypically and functionally different from MSC from other sources, such as bone marrow, and might therefore be a better alternative for cell-based therapy of lung diseases like obliterative bronchiolitis.

ial both in vivo and in vitro, but showed impaired bone formation in vivo, indicating that lung-derived MSC are phenotypically and functionally different from MSC from other sources, such as bone marrow, and might therefore be a better alternative for cell-based therapy of lung diseases like obliterative bronchiolitis. Supplementary Material Web supplement Web supplement Web supplement Web supplement The authors would like to thank Nicholas Ditzel, Linda Magnusson, Lena Thiman, Marie Wildt, Cecilia Andersson and Maria Weitoft for excellent technical assistance, and Michael Bzorek for help with the immunostaining. Contributors: SR performed and designed the research, analysed and interpreted the data and wrote the manuscript. AAS was involved in the design of the research and interpreted the results. JCB was involved in the design and also optimised and interpreted the results in some of the cell culture experiments. HL performed the fluorescence-activated cell sorting experiments. MK interpreted and analysed the animal experiments. FM interpreted and analysed the karyotype and fluorescence in situ hybridisation results. AW performed the statistics. LE and LB were involved in the design of the research. LH and IS recruited and characterised the patients and also collected the materials. GW-T and SS designed the research, interpreted the results, funded the study and wrote the manuscript. All authors read the manuscript.

sation results. AW performed the statistics. LE and LB were involved in the design of the research. LH and IS recruited and characterised the patients and also collected the materials. GW-T and SS designed the research, interpreted the results, funded the study and wrote the manuscript. All authors read the manuscript. Funding: This study was supported by the Heart & Lung Foundation, the Swedish Medical Research Council (11 550 and 2010-3298 awarded to GWT and SS, respectively), the Evy and Gunnar Sandberg Foundation, Greta and John Kock, the Alfred Österlund Foundation, the Anna-Greta Crafoord Foundation, the Konsul Bergh Foundation, the Royal Physiographical Society in Lund and the Medical Faculty of Lund University. Competing interests: None. Patient consent: Obtained. Ethics approval: The local ethics committee in Lund, Sweden. All animal procedures were approved by the local ethical committees on animal experiments (2012-D4-2934-00006, C1). Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Key messages Peptidoglycan exerts a significant proinflammatory cytokine response in primary human alveolar epithelium but not in primary human nasal epithelium. The Toll-like receptor regulator Toll-interacting protein is widely expressed in the human respiratory tract. Introduction Hospital-acquired infections (HAIs) are common and associated with significant morbidity and mortality.1 Pneumonia is associated with the highest mortality among the HAIs.1 2 The pathogenesis of hospital-acquired pneumonia is thought to involve recurrent microaspiration of mircoorganisms which have asymptomatically colonised the patient's oropharynx/nasopharynx during the course of hospital admission.2 Why the nasal epithelium should tolerate these microorganisms well, while the alveolar epithelium mounts such a florid inflammatory response, remains poorly understood. A better understanding of this paradox has been hampered by difficulties in accessing primary cells from the human nose and alveoli.

Introduction Hospital-acquired infections (HAIs) are common and associated with significant morbidity and mortality.1 Pneumonia is associated with the highest mortality among the HAIs.1 2 The pathogenesis of hospital-acquired pneumonia is thought to involve recurrent microaspiration of mircoorganisms which have asymptomatically colonised the patient's oropharynx/nasopharynx during the course of hospital admission.2 Why the nasal epithelium should tolerate these microorganisms well, while the alveolar epithelium mounts such a florid inflammatory response, remains poorly understood. A better understanding of this paradox has been hampered by difficulties in accessing primary cells from the human nose and alveoli. We therefore sought to characterise the effects of key virulence factors from Staphylococcus aureus and Pseudomonas aeruginosa (recognised as key pathogens in nosocomial pneumonia)2 on human primary nasal and alveolar epithelial cells. An additional aim was to determine whether Toll-interacting protein (TOLLIP, an endogenous inhibitor of Toll-like receptor (TLR) signalling)3 4 was expressed in the human respiratory tract and, if so, whether there was differential expression in nasal and alveolar epithelium. This protein has been implicated as a key regulator of inflammatory responses in the large intestine, contributing to the dampening of TLR responses to microbe-associated molecular patterns derived from the extensive community of commensal organisms.5 6 However, remarkably little is known about TOLLIP expression in the human respiratory tract.

licated as a key regulator of inflammatory responses in the large intestine, contributing to the dampening of TLR responses to microbe-associated molecular patterns derived from the extensive community of commensal organisms.5 6 However, remarkably little is known about TOLLIP expression in the human respiratory tract. The primary hypothesis for this study was that primary alveolar cells would mount a brisk response to inflammatory stimuli, associated with minimal or absent TOLLIP expression, whereas primary nasal cells would exhibit a blunted response to inflammatory stimuli, associated with abundant TOLLIP expression. Methods Derivation of cells Primary human nasal epithelial cells, bronchial epithelial cells and type II alveolar epithelial cells were obtained from patients undergoing elective pneumonectomy or lobectomy for cancer. Methods for obtaining and culturing the nasal and alveolar cells have been described elsewhere.7 8 Bronchial epithelial cells were obtained using a cytology brush passed through an endotracheal tube during the surgical procedure. Cells were seeded onto plates coated with type I rat tail collagen (Sigma-Aldrich, St Louis, Missouri, USA) and allowed to achieve confluence. Cells were studied at passage 2. Informed written consent was provided by all participants providing primary cells. The human colonic carcinoma cell line T84 and the human nasal carcinoma cell line RPMI 2650 were from LGC Promochem (Manassas, Virginia, USA; ATCC numbers CCL-248 and CCL-30 respectively). A549 cells (derived from a human alveolar cell carcinoma) were available in-house.

Methods Derivation of cells Primary human nasal epithelial cells, bronchial epithelial cells and type II alveolar epithelial cells were obtained from patients undergoing elective pneumonectomy or lobectomy for cancer. Methods for obtaining and culturing the nasal and alveolar cells have been described elsewhere.7 8 Bronchial epithelial cells were obtained using a cytology brush passed through an endotracheal tube during the surgical procedure. Cells were seeded onto plates coated with type I rat tail collagen (Sigma-Aldrich, St Louis, Missouri, USA) and allowed to achieve confluence. Cells were studied at passage 2. Informed written consent was provided by all participants providing primary cells. The human colonic carcinoma cell line T84 and the human nasal carcinoma cell line RPMI 2650 were from LGC Promochem (Manassas, Virginia, USA; ATCC numbers CCL-248 and CCL-30 respectively). A549 cells (derived from a human alveolar cell carcinoma) were available in-house. Cell stimulation experiments Confluent cells were treated with 100 ng/mL of ultrapure lipopolysaccharide (LPS) derived from P. aeruginosa strain PA01 (a gift from Professor Ian Poxton, University of Edinburgh), 10 μg/mL of S. aureus peptidoglycan (PGN; Fluka, Sigma-Aldrich), 10 μg/mL of S. aureus lipoteichoic acid (LTA; Sigma-Aldrich), 10 ng/mL of recombinant human tumour necrosis factor (TNF; R&D Systems, Minneapolis, USA), 1 μΜ CpG-C DNA (ODN 2395; HyCult Biotechnology b.v., Uden, the Netherlands) or medium alone (all final concentrations). Cells were incubated for 24 h at 37°C and supernatants were removed and stored at −80°C until estimation of interleukin (IL)-1β, IL-6, IL-8, IL-10, IL-12p70 and TNF assayed using the BD Cytometric Bead Array (CBA) Human Inflammatory Cytokine kit (BD Biosciences), with analysis performed using a BD FACSArray Bioanalyzer System.

re incubated for 24 h at 37°C and supernatants were removed and stored at −80°C until estimation of interleukin (IL)-1β, IL-6, IL-8, IL-10, IL-12p70 and TNF assayed using the BD Cytometric Bead Array (CBA) Human Inflammatory Cytokine kit (BD Biosciences), with analysis performed using a BD FACSArray Bioanalyzer System. RNA extraction, reverse transcriptase PCR and real-time quantitative PCR Total RNA was extracted using the total RNA isolation kit Nucleospin RNAII (Macherey-Nagel, Duren, Germany). 1 μg RNA was reverse transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Carlsbard, California, USA). Primers and probes are summarised in a table in the online supplementary section. A Taqman Low Density Array (TLDA; Applied Biosystems) was used to assess the stability of potential housekeeping genes. Based on the normalisation score, Cyclophilin A (PPIA) had the lowest variability rate in the samples assayed. Results were normalised using a TaqMan endogenous control (Applied Biosystems).

RNA extraction, reverse transcriptase PCR and real-time quantitative PCR Total RNA was extracted using the total RNA isolation kit Nucleospin RNAII (Macherey-Nagel, Duren, Germany). 1 μg RNA was reverse transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Carlsbard, California, USA). Primers and probes are summarised in a table in the online supplementary section. A Taqman Low Density Array (TLDA; Applied Biosystems) was used to assess the stability of potential housekeeping genes. Based on the normalisation score, Cyclophilin A (PPIA) had the lowest variability rate in the samples assayed. Results were normalised using a TaqMan endogenous control (Applied Biosystems). Diluted cDNA (1:100) was used as a template for the PCR reaction and samples were loaded onto the Applied Biosystems 7900HT Fast Real-Time PCR System. The specificity of the reactions was controlled using ‘no template’ and ‘no reverse transcription’ controls. Results were normalised to the human PPIA gene using the standard curve method. Standard curves for the genes of interest were prepared using the plasmids pcDNA3-TLR9-YFP, Addgene plasmid 13642, pcDNA3-TLR4-YFP, Addgene plasmid 13018 and pUC19/human IL-8 Addgene plasmid 17610. Pooled DNA was used in the standard curves for PPIA, TOLLIP and TLR2.

the human PPIA gene using the standard curve method. Standard curves for the genes of interest were prepared using the plasmids pcDNA3-TLR9-YFP, Addgene plasmid 13642, pcDNA3-TLR4-YFP, Addgene plasmid 13018 and pUC19/human IL-8 Addgene plasmid 17610. Pooled DNA was used in the standard curves for PPIA, TOLLIP and TLR2. Immunocytochemistry and confocal microscopy Confluent cells were detached using trypsin/EDTA solution (10 min at 37°C), and centrifuged. Resuspended cells were seeded onto glass coverslips for 15 min and incubated overnight at 37°C. Medium was replaced with ice-cold methanol for 10 min, the cells were washed and then blocking was performed using 2% goat serum for 30 min. Cells were dried and antibodies were applied overnight as appropriate: murine monoclonal IgG1 against human cytokeratin 18, murine monoclonal IgG2a against human cytokeratin 19, murine monoclonal IgG2a against human TLR2 (all Invitrogen), polyclonal rabbit antihuman TLR4 IgG and polyclonal rabbit antihuman TOLLIP IgG (Abcam). Controls comprised murine isotype monoclonal antibodies (Invitrogen) or, where polyclonal primaries were used, non-immune rabbit IgG (Invitrogen). The following day cells were washed with phosphate buffered saline and secondary antibodies applied for 1 h. Secondary antibodies comprised AlexaFluor488-conjugated goat antimouse IgG (Invitrogen) or goat antirabbit IgG conjugated to AlexaFluor 488 (Invitrogen) as appropriate. Cells were washed, dried and Vectashield with DAPI (Vector Laboratories, Burlingame, California, USA) added. Cells were visualised using a Leica TCS SP5 confocal microscope (Leica Microsystems CMS GmbH, Mannheim, Germany), and photomicrographs taken.

irabbit IgG conjugated to AlexaFluor 488 (Invitrogen) as appropriate. Cells were washed, dried and Vectashield with DAPI (Vector Laboratories, Burlingame, California, USA) added. Cells were visualised using a Leica TCS SP5 confocal microscope (Leica Microsystems CMS GmbH, Mannheim, Germany), and photomicrographs taken. Coculture of cell lines with S. aureus The cell lines RPMI 2650 or A549 were seeded at a density of 1×106 cells per well. On the same day 5 mL of Modified Eagles Medium (MEM; Sigma-Aldrich) was inoculated with S. aureus strain Newman, and incubated overnight at 37°C with continuous shaking. The following day an aliquot was inoculated in 5 mL of MEM and allowed to reach logarithmic phase. Bacteria were washed and resuspended in MEM to achieve an optical density of approximately 0.1. Known volumes were (A) added directly to cells and (B) plated onto tryptic soy agar, so that viable bacterial concentrations could be determined by quantifying colony forming units (CFU) the next day. After infection, cells were incubated for a further 4 h at 37°C prior to cell lysis and RNA extraction as above.

y 0.1. Known volumes were (A) added directly to cells and (B) plated onto tryptic soy agar, so that viable bacterial concentrations could be determined by quantifying colony forming units (CFU) the next day. After infection, cells were incubated for a further 4 h at 37°C prior to cell lysis and RNA extraction as above. Statistics Friedman's test was used to provide a global indication of whether any significant difference existed across the conditions applied to cultured cells. Post hoc analysis comparing unstimulated and stimulated cells was performed using Dunn's test. Comparisons of numerical data between groups were carried out using the Mann-Whitney U test. Comparison of proportions between groups was carried out using Fisher's exact test. Correlations were analysed using Spearman's test. All statistical analyses were performed using GraphPad Prism software (GraphPad Software, La Jolla, California, USA). Statistical significance was considered to be at the p<0.05 level.

. Comparison of proportions between groups was carried out using Fisher's exact test. Correlations were analysed using Spearman's test. All statistical analyses were performed using GraphPad Prism software (GraphPad Software, La Jolla, California, USA). Statistical significance was considered to be at the p<0.05 level. Results Primary nasal cells were successfully cultured from 6 patients, and primary alveolar cells from 7 (in two cases nasal and alveolar cell were cultured from the same patient). The two groups of patients were similar in their baseline characteristics, although there were more women in the group providing alveolar cells (results from the patients providing nasal cells appear first in all the following comparisons: median age 65 vs 60 years; smoking history100% vs 71%; women 50% vs 86%; mean forced expiratory volume in 1 s 85% vs 84% of predicted; mean diffusing capacity for carbon monoxide (Tco) 63% vs 75% of predicted; no significant difference for any of the comparisons). The patients were admitted for resection of non-small cell lung cancer, with the exception of two patients admitted for resection of solitary metastases. Characterisation by quantitative reverse transcriptase PCR (qRT-PCR) demonstrated that cultured nasal epithelial cells consistently expressed the epithelial cell markers cytokeratin 18 and 19 and alveolar epithelial cells expressed the type II pneumocyte markers SP-C and AQP-3 (data not shown, methods described in the online supplementary section).

ve reverse transcriptase PCR (qRT-PCR) demonstrated that cultured nasal epithelial cells consistently expressed the epithelial cell markers cytokeratin 18 and 19 and alveolar epithelial cells expressed the type II pneumocyte markers SP-C and AQP-3 (data not shown, methods described in the online supplementary section). A range of bacterial virulence factors was applied to primary cells and the cytokine responses were examined by CBA and qRT-PCR. All of the cytokines examined could be produced by primary nasal epithelial cells. However, none of the measured cytokines were significantly upregulated by exposure to PGN, LTA, LPS or CpG (table 1). In contrast, exposure to TNF induced a significant upregulation of IL-8 and IL-6 secretion (but not the other cytokines studied). Table 1 Constitutive and stimulated cytokine production by primary nasal epithelial cells

A range of bacterial virulence factors was applied to primary cells and the cytokine responses were examined by CBA and qRT-PCR. All of the cytokines examined could be produced by primary nasal epithelial cells. However, none of the measured cytokines were significantly upregulated by exposure to PGN, LTA, LPS or CpG (table 1). In contrast, exposure to TNF induced a significant upregulation of IL-8 and IL-6 secretion (but not the other cytokines studied). Table 1 Constitutive and stimulated cytokine production by primary nasal epithelial cells Stimulant Basal Staphylococcus aureus PGN S. aureus LTA Pseudomonas aeruginosa LPS CpG TNF IL-1β (pg/mL) 7.1 0–18.7 7.7 0–23.8 4.2 0–21.9 3.6 0–16.4 6 0–17.3 8.1 0–165 IL-6 (pg/mL) 29.7 13.7–313 140 21.6–695 52.1 6.3–459 139 7.9–279 45 4.7–535 956 ** 67.5–3173 IL-8 (pg/mL) 504 192–1557 1363 378–3821 663 297–2309 740 131–4295 520 11.8–2531 7817 *** 2033–48 688 IL-10 (pg/mL) 9.2 4–18.7 12.5 4–21.6 7.1 0–14.5 6.4 0–18.6 6.5 0–21.1 13 0–67 IL-12 (pg/mL) 13.2 3.6–19.8 12.1 0–21 8.8 0–16.1 10.3 0–21.4 10.4 0–26.7 10.4 0–23.3 TNF (pg/mL) 10 1.7–15 6.2 2–24.3 7.2 0–11.8 6.5 3–16.1 6.3 0–17.5 † Data are expressed as median (upper line, italic) and range (lower line, normal text). n=6 for all conditions. PGN and LTA were applied at 10 μg/mL, LPS at 100 ng/mL, CpG at 1 μM and TNF at 10 ng/mL. Statistical analysis was by Friedman's test and Dunn's post hoc test. *p<0.05, **p<0.01, ***p<0.001 relative to basal levels, by Dunn's post hoc test. †TNF was used as a positive control; TNF was not measured in TNF-stimulated cells.

Stimulant Basal Staphylococcus aureus PGN S. aureus LTA Pseudomonas aeruginosa LPS CpG TNF IL-1β (pg/mL) 7.1 0–18.7 7.7 0–23.8 4.2 0–21.9 3.6 0–16.4 6 0–17.3 8.1 0–165 IL-6 (pg/mL) 29.7 13.7–313 140 21.6–695 52.1 6.3–459 139 7.9–279 45 4.7–535 956 ** 67.5–3173 IL-8 (pg/mL) 504 192–1557 1363 378–3821 663 297–2309 740 131–4295 520 11.8–2531 7817 *** 2033–48 688 IL-10 (pg/mL) 9.2 4–18.7 12.5 4–21.6 7.1 0–14.5 6.4 0–18.6 6.5 0–21.1 13 0–67 IL-12 (pg/mL) 13.2 3.6–19.8 12.1 0–21 8.8 0–16.1 10.3 0–21.4 10.4 0–26.7 10.4 0–23.3 TNF (pg/mL) 10 1.7–15 6.2 2–24.3 7.2 0–11.8 6.5 3–16.1 6.3 0–17.5 † Data are expressed as median (upper line, italic) and range (lower line, normal text). n=6 for all conditions. PGN and LTA were applied at 10 μg/mL, LPS at 100 ng/mL, CpG at 1 μM and TNF at 10 ng/mL. Statistical analysis was by Friedman's test and Dunn's post hoc test. *p<0.05, **p<0.01, ***p<0.001 relative to basal levels, by Dunn's post hoc test. †TNF was used as a positive control; TNF was not measured in TNF-stimulated cells. IL, interleukin; LPS, lipopolysaccharide; LTA, lipoteichoic acid; TNF, tumour necrosis factor; PGN, peptidoglycan.

Stimulant Basal Staphylococcus aureus PGN S. aureus LTA Pseudomonas aeruginosa LPS CpG TNF IL-1β (pg/mL) 7.1 0–18.7 7.7 0–23.8 4.2 0–21.9 3.6 0–16.4 6 0–17.3 8.1 0–165 IL-6 (pg/mL) 29.7 13.7–313 140 21.6–695 52.1 6.3–459 139 7.9–279 45 4.7–535 956 ** 67.5–3173 IL-8 (pg/mL) 504 192–1557 1363 378–3821 663 297–2309 740 131–4295 520 11.8–2531 7817 *** 2033–48 688 IL-10 (pg/mL) 9.2 4–18.7 12.5 4–21.6 7.1 0–14.5 6.4 0–18.6 6.5 0–21.1 13 0–67 IL-12 (pg/mL) 13.2 3.6–19.8 12.1 0–21 8.8 0–16.1 10.3 0–21.4 10.4 0–26.7 10.4 0–23.3 TNF (pg/mL) 10 1.7–15 6.2 2–24.3 7.2 0–11.8 6.5 3–16.1 6.3 0–17.5 † Data are expressed as median (upper line, italic) and range (lower line, normal text). n=6 for all conditions. PGN and LTA were applied at 10 μg/mL, LPS at 100 ng/mL, CpG at 1 μM and TNF at 10 ng/mL. Statistical analysis was by Friedman's test and Dunn's post hoc test. *p<0.05, **p<0.01, ***p<0.001 relative to basal levels, by Dunn's post hoc test. †TNF was used as a positive control; TNF was not measured in TNF-stimulated cells. IL, interleukin; LPS, lipopolysaccharide; LTA, lipoteichoic acid; TNF, tumour necrosis factor; PGN, peptidoglycan. Alveolar cell responses were assessed in parallel with nasal cells. LPS and LTA failed to significantly alter secretion of any of the cytokines (table 2). However, in contrast to the nasal cells, exposure to PGN significantly increased production of all cytokines studied in alveolar cells from every patient studied, with the exception of IL-12, suggesting a differential TLR2 response in primary human alveolar versus nasal epithelial cells. Similarly to the response of primary nasal cells, TNF-mediated stimulation induced significant elevations in secretion of IL-6, IL-8 and IL-10 from alveolar cells, suggesting no major differences in signalling downstream of the TNF receptor between these two cell types.

ary human alveolar versus nasal epithelial cells. Similarly to the response of primary nasal cells, TNF-mediated stimulation induced significant elevations in secretion of IL-6, IL-8 and IL-10 from alveolar cells, suggesting no major differences in signalling downstream of the TNF receptor between these two cell types. Table 2 Constitutive and stimulated cytokine production by primary type II alveolar epithelial cells Stimulant Basal Staphylococcus aureus PGN S. aureus LTA Pseudomonas aeruginosa LPS CpG TNF IL-1β (pg/mL) 5 2.5–8 17.2 ** 5–152 3.4 1.6–12.5 6.3 2.2–14 7.5 1.7–11 11 1.2–85.3 IL-6 (pg/mL) 236 8.3–1276 927 * 121–9060 333 7.6–716 214 8.2–533 228 12.6–803 1205 ** 34.1–4029 IL-8 (pg/mL) 2273 707–11 226 7444 * 1283–100 000 2002 843–21 914 1507 649–13 548 2919 636–40 775 31 721 *** 9450–78 198 IL-10 (pg/mL) 15 2.6–1276 25.4 ** 3.5–5000 23.2 3.6–716 19.2 3–504 20.2 0–803 26 * 3.5–4029 IL-12 (pg/mL) 8 5.4–19.7 7.3 6.6–31.2 8.3 4.9–30 12.7 3.5–25 12.0 2.7–28.6 7 2.7–30.3 TNF (pg/mL) 10 3.6–21.2 29 * 6.5–279 5 0–21.4 12 2.3–26.7 7.0 0–15.7 † Data are expressed as median (upper line, italic) and range (lower line, normal text). n=7 for all conditions. PGN and LTA were applied at 10 μg/mL, LPS at 100 ng/mL, CpG at 1 μM and TNF at 10 ng/mL. Statistical analysis was by Friedman's test and Dunn's post hoc test. *p<0.05, **p<0.01, ***p<0.001 relative to basal levels, by Dunn's post hoc test. †TNF was used as a positive control; TNF was not measured in TNF-stimulated cells. IL, interleukin; LPS, lipopolysaccharide; LTA, lipoteichoic acid; TNF, tumour necrosis factor; PGN, peptidoglycan.

Stimulant Basal Staphylococcus aureus PGN S. aureus LTA Pseudomonas aeruginosa LPS CpG TNF IL-1β (pg/mL) 5 2.5–8 17.2 ** 5–152 3.4 1.6–12.5 6.3 2.2–14 7.5 1.7–11 11 1.2–85.3 IL-6 (pg/mL) 236 8.3–1276 927 * 121–9060 333 7.6–716 214 8.2–533 228 12.6–803 1205 ** 34.1–4029 IL-8 (pg/mL) 2273 707–11 226 7444 * 1283–100 000 2002 843–21 914 1507 649–13 548 2919 636–40 775 31 721 *** 9450–78 198 IL-10 (pg/mL) 15 2.6–1276 25.4 ** 3.5–5000 23.2 3.6–716 19.2 3–504 20.2 0–803 26 * 3.5–4029 IL-12 (pg/mL) 8 5.4–19.7 7.3 6.6–31.2 8.3 4.9–30 12.7 3.5–25 12.0 2.7–28.6 7 2.7–30.3 TNF (pg/mL) 10 3.6–21.2 29 * 6.5–279 5 0–21.4 12 2.3–26.7 7.0 0–15.7 † Data are expressed as median (upper line, italic) and range (lower line, normal text). n=7 for all conditions. PGN and LTA were applied at 10 μg/mL, LPS at 100 ng/mL, CpG at 1 μM and TNF at 10 ng/mL. Statistical analysis was by Friedman's test and Dunn's post hoc test. *p<0.05, **p<0.01, ***p<0.001 relative to basal levels, by Dunn's post hoc test. †TNF was used as a positive control; TNF was not measured in TNF-stimulated cells. IL, interleukin; LPS, lipopolysaccharide; LTA, lipoteichoic acid; TNF, tumour necrosis factor; PGN, peptidoglycan. Given the differential secretion of IL-8 in response to PGN, the effect of this bacterial TLR agonist on IL-8 mRNA production was also analysed. No significant increase in IL-8 expression was observed in either cell type (data not shown), suggesting that at least some of the effect of PGN on IL-8 secretion in alveolar cells may be post-transcriptional.

f IL-8 in response to PGN, the effect of this bacterial TLR agonist on IL-8 mRNA production was also analysed. No significant increase in IL-8 expression was observed in either cell type (data not shown), suggesting that at least some of the effect of PGN on IL-8 secretion in alveolar cells may be post-transcriptional. Given that PGN mediates its effects largely through TLR2-mediated recognition and signalling, expression of TLR2 in primary nasal and alveolar epithelial cells was also assessed by qRT-PCR (figure 1A). TLR2 expression was significantly greater in alveolar epithelial cells than in nasal cells (p=0.0043). In contrast, no significant differences in expression of TLR4 and TLR9 were observed between these two cell types (data not shown). Interestingly, TLR2 expression correlated significantly with IL-8 secretion in nasal and epithelial cells, both under basal (p=0.0144) and PGN-stimulated (p=0.0074) conditions (figure 1B). Figure 1 TLR2 expression is significantly higher in alveolar epithelium than in nasal epithelium, and correlates with IL-8 secretion. (A) Comparison of TLR2 expression in primary nasal and alveolar epithelium, in the presence or absence of PGN. *p<0.05, **p<0.01 using the Mann-Whitney U test. (B) Correlation between TLR2 expression and IL-8 secretion in primary cells, in the presence or absence of PGN. Dots represent nasal epithelial cells, grey triangles represent alveolar cells. *p<0.05, **p<0.01 using Spearman's rank correlation coefficient. TLR, Toll-like receptor; IL, interleukin; PGN, peptidoglycan.

. (B) Correlation between TLR2 expression and IL-8 secretion in primary cells, in the presence or absence of PGN. Dots represent nasal epithelial cells, grey triangles represent alveolar cells. *p<0.05, **p<0.01 using Spearman's rank correlation coefficient. TLR, Toll-like receptor; IL, interleukin; PGN, peptidoglycan. In addition to differential expression of TLR2, the expression of the TLR regulator TOLLIP was evaluated. TOLLIP expression has been clearly defined in the T84 colonic carcinoma cell line6; therefore, we initially characterised our novel TOLLIP qRT-PCR assay in this setting. A band of the expected size was consistently detected, and was absent in negative controls (figure 2A). TOLLIP expression was quantified in cultured primary nasal and type II alveolar epithelial cells (from n=5 and n=6, respectively) treated under identical conditions. Basal TOLLIP mRNA expression was observed in nasal and alveolar cells but was found to be significantly higher (p<0.05) in the primary nasal epithelial cells (figure 2B).

n was quantified in cultured primary nasal and type II alveolar epithelial cells (from n=5 and n=6, respectively) treated under identical conditions. Basal TOLLIP mRNA expression was observed in nasal and alveolar cells but was found to be significantly higher (p<0.05) in the primary nasal epithelial cells (figure 2B). Figure 2 TOLLIP expression in nasal and alveolar epithelium. (A) T84 cells were plated at two different cell densities: 5×105 per well (lanes 1, 2); 2×106, (lanes 3, 4). Lane 5 represents a negative control without the reverse transcriptase. GAPDH was used as a housekeeping gene. (B) TOLLIP expression was quantified in primary nasal and alveolar epithelium. *p<0.05 by Mann-Whitney U test. (C and D) Cell lines were infected with Staphylococcus aureus strain Newman. RNA extraction was performed followed by RT-PCR. Panel C shows RPMI 2650 cells—and panel D A549 cells—infected with S. aureus. Lanes: (1) positive control for TOLLIP from cell line T84; (2 and 3) unstimulated; (4 and 5) cells with S. aureus at 1.1×105 cfu/mL; (6 and 7) cells with S. aureus at 1.6×105 cfu/mL. GAPDH was used as a housekeeping gene. Band size for TOLLIP 347 bp and for GAPDH 727 bp (TOLLIP, Toll-interacting protein; RT-PCR, reverse transcriptase PCR).

ontrol for TOLLIP from cell line T84; (2 and 3) unstimulated; (4 and 5) cells with S. aureus at 1.1×105 cfu/mL; (6 and 7) cells with S. aureus at 1.6×105 cfu/mL. GAPDH was used as a housekeeping gene. Band size for TOLLIP 347 bp and for GAPDH 727 bp (TOLLIP, Toll-interacting protein; RT-PCR, reverse transcriptase PCR). Owing to the difficulties in obtaining sufficient numbers of primary cells, and the difficulties inherent in applying live bacteria to cells, the effect of S. aureus on TOLLIP expression was studied in cell lines. Clear evidence for basal TOLLIP expression was observed in nasal and alveolar cell lines, and 4 h exposure to S. aureus did not appear to influence this (figure 2C, D), suggesting a non-inducible expression in these cell types. Primary nasal and bronchial epithelial cells demonstrated a broadly similar pattern of TOLLIP protein expression, with diffuse punctate staining throughout the cytoplasm, and a suggestion (in a proportion of cells) of peripheral accentuation of staining around the cell membrane (figure 3A–D). Punctate staining was also visible in type II alveolar epithelial cells (figure 3E, F).

roadly similar pattern of TOLLIP protein expression, with diffuse punctate staining throughout the cytoplasm, and a suggestion (in a proportion of cells) of peripheral accentuation of staining around the cell membrane (figure 3A–D). Punctate staining was also visible in type II alveolar epithelial cells (figure 3E, F). Figure 3 TOLLIP is found in primary human nasal, bronchial and alveolar epithelial cells. Primary nasal (A and B), bronchial (C and D) and type II alveolar epithelial cells (E and F) were fixed, blocked with 2% goat serum and incubated with a rabbit polyclonal antibody against TOLLIP (A, C and E) or isotype control (B, D and F). Nuclei were stained with DAPI (blue). Secondary antibody was antirabbit IgG conjugated with Alexa 488 (green). Images were analysed using confocal microscopy. Three nasal samples, one bronchial and one alveolar were analysed. Scale bar equals 50 μm in A–D, and 10 μm in E and F (TOLLIP, Toll-interacting protein). Discussion To our knowledge, this study is among the first to compare the differential response of primary human nasal and alveolar epithelial cells to a range of identical inflammatory stimuli, and the first to systematically describe TOLLIP expression and localisation in the human respiratory tract.

Figure 3 TOLLIP is found in primary human nasal, bronchial and alveolar epithelial cells. Primary nasal (A and B), bronchial (C and D) and type II alveolar epithelial cells (E and F) were fixed, blocked with 2% goat serum and incubated with a rabbit polyclonal antibody against TOLLIP (A, C and E) or isotype control (B, D and F). Nuclei were stained with DAPI (blue). Secondary antibody was antirabbit IgG conjugated with Alexa 488 (green). Images were analysed using confocal microscopy. Three nasal samples, one bronchial and one alveolar were analysed. Scale bar equals 50 μm in A–D, and 10 μm in E and F (TOLLIP, Toll-interacting protein). Discussion To our knowledge, this study is among the first to compare the differential response of primary human nasal and alveolar epithelial cells to a range of identical inflammatory stimuli, and the first to systematically describe TOLLIP expression and localisation in the human respiratory tract. The findings suggest that primary nasal epithelial cells have a relatively limited repertoire of responsiveness to inflammatory stimuli, generating a statistically significant (but still numerically modest) increase in the proinflammatory cytokines IL-6 and IL-8, only in response to stimulation with TNF, but not TLR agonists. This responsiveness to TNF is consistent with findings elsewhere.7 Other studies have suggested that primary human nasal epithelial cells have a relatively restricted nasal cytokine responsiveness to stimulation, broadly in keeping with findings here.9 10 However, unlike our results, both these studies found responsiveness of IL-8 to a variety of stimuli, while a further study found that both IL-6 and IL-8 were increased in response to LPS.11

elial cells have a relatively restricted nasal cytokine responsiveness to stimulation, broadly in keeping with findings here.9 10 However, unlike our results, both these studies found responsiveness of IL-8 to a variety of stimuli, while a further study found that both IL-6 and IL-8 were increased in response to LPS.11 In contrast to the relative quiescence of primary nasal cells, we found that primary alveolar epithelial cells were characterised by a more florid response to PGN and TNF that spanned a wider range of cytokines. These observations appear consistent with the hypothesis that bacterial virulence factors are better tolerated by the nose. Our data suggest that S. aureus PGN induces a particularly florid inflammatory response in alveolar epithelial cells. It may be particularly relevant that, in our hands, the levels of expression of TLR2 (which recognises PGN) correlated closely with responsiveness, as assessed by IL-8 secretion. The implication seems to be not only that alveolar epithelium expresses more ‘target’ for PGN, but that PGN can upregulate TLR2 expression more effectively on alveolar epithelium. This may go some way to explaining the differential responsiveness of nasal and alveolar epithelium, and perhaps why the lung mounts such a striking inflammatory response to S. aureus, a common ‘coloniser’ of the human nose.12 It is far less clear why PGN produced a proinflammatory response in our alveolar epithelial cells while LTA and LPS did not. In the case of LPS, the lack of responsiveness could not be attributed to an absence of appropriate receptors, as TLR4 is well described on alveolar epithelial cells, and other groups have described LPS responsiveness in alveolar epithelium.13 14 The apparently selective and florid response of alveolar cells to PGN in our hands is intriguing. It is tempting to speculate that membrane-based TLR regulators may recognise different virulence factors preferentially, and/or that PGN effects intracellular TLR regulators in a different way from other virulence factors in primary alveolar epithelial cells. However, this must remain purely speculative until further data are available.

speculate that membrane-based TLR regulators may recognise different virulence factors preferentially, and/or that PGN effects intracellular TLR regulators in a different way from other virulence factors in primary alveolar epithelial cells. However, this must remain purely speculative until further data are available. To investigate further potential reasons for differential innate immune responsiveness between the nose and lung, we drew on data describing an excess of TOLLIP in the large intestine, where bacterial tolerance is essential. We believe this to be the first systematic characterisation of TOLLIP's presence and location in primary cells from the human respiratory tract. TOLLIP has been cloned from a human lung cDNA library,15 and expression has been described in pooled human lung tissue,16 but the purpose of those studies did not include cellular localisation. TOLLIP mRNA and TOLLIP protein have been detected in commercially available human small airway epithelial cells.17 TOLLIP mRNA has also been described in pleural effusions.18 Our findings in figure 3 complement those in small airway epithelial cells by suggesting that TOLLIP is produced throughout the length of the human respiratory tract. These observations are at variance with our initial hypothesis. However, the finding of higher TOLLIP mRNA expression in primary nasal epithelial cells in comparison to type II alveolar epithelial cells broadly supports the hypothesis. The observation that TOLLIP is constitutively and ubiquitously expressed in human respiratory epithelium is consistent with a potential role as a key regulator of inflammatory responses.3 4 19 However, we must stress that we found no evidence for differential TOLLIP responsiveness to bacterial virulence factors in nasal and alveolar cell lines.

P is constitutively and ubiquitously expressed in human respiratory epithelium is consistent with a potential role as a key regulator of inflammatory responses.3 4 19 However, we must stress that we found no evidence for differential TOLLIP responsiveness to bacterial virulence factors in nasal and alveolar cell lines. TOLLIP binds to IL-1 receptor-associated kinase (IRAK-1), preventing proinflammatory signalling. On stimulation of cells with LPS or IL-1, a receptor complex rapidly forms, incorporating TOLLIP bound to IRAK-1. Sufficient phosphorylation of IRAK-1 allows its dissociation from TOLLIP, and proinflammatory signalling (for example, through nuclear factor κ B) rapidly ensues. TOLLIP is therefore well placed to regulate inflammatory processes. TOLLIP's ready availability in organs regularly exposed to bacteria, such as the gut, nose and lung, seems potentially important in this regard. Interestingly, TOLLIP has been implicated in LPS hyporesponsiveness in human monocytes and human primary intestinal epithelial cells.20 21

Three measurements of blood pressure were taken during the health check. Men were categorised as having hypertension if mean systolic blood pressure using the second and third blood pressure measurements was greater than 139 mm Hg or diastolic blood pressure was greater than 89 mm Hg and/or participants reported taking any antihypertensive medications either daily or sometimes. Angina was determined from self-reported angina symptoms measured using the Rose angina questionnaire.27 Classification of diabetes mellitus, MI and stroke were all assessed from self-report of these conditions. Self-reported health Self-reported physical and mental health were measured using the SF-12, a standardised instrument for measuring self-reported health, derived as a shorter alternative to the SF-36.25 In accordance with standard practice, two summary scales were derived from these questions: a physical health component score and a mental health component score. Both of these summary scores ranged from 0 to 100.

to regulate inflammatory processes. TOLLIP's ready availability in organs regularly exposed to bacteria, such as the gut, nose and lung, seems potentially important in this regard. Interestingly, TOLLIP has been implicated in LPS hyporesponsiveness in human monocytes and human primary intestinal epithelial cells.20 21 The functional importance of TOLLIP as a regulator of acute inflammation is supported by emerging clinical data. For example, in the Chinese Han population, increased susceptibility to sepsis is conferred by polymorphisms in the TOLLIP gene that result in reduced TOLLIP function.22 Similarly, functional polymorphisms in a Vietnamese population have been associated with susceptibility to tuberculosis.23 In a Caucasian population, TOLLIP gene polymorphisms have been weakly associated with increased susceptibility to atopic dermatitis.24 Observational data suggest that TOLLIP expression is reduced in tissue from coeliac disease and necrotising enterocolitis.25 26 While the data here are some way from having direct clinical relevance, validation of a florid alveolar response to PGN in other cohorts might yield avenues for further exploration. In particular, selective administration of anti-TLR2 or specific TLR regulators early in the florid proinflammatory phase of staphylococcal pneumonia seems theoretically attractive in a condition with continued high mortality despite modern antibiotics and supportive care. The association between TLR2 expression and IL-8 secretion in unstimulated and PGN-stimulated cells is potentially relevant in this regard.

florid proinflammatory phase of staphylococcal pneumonia seems theoretically attractive in a condition with continued high mortality despite modern antibiotics and supportive care. The association between TLR2 expression and IL-8 secretion in unstimulated and PGN-stimulated cells is potentially relevant in this regard. Comparison of responses in primary human cells increases the relevance of this study. However, we recognise that there are several potential limitations. First, all of our patients had cancer and most had a long history of smoking, which is known to affect cytokine secretion by epithelial cells throughout the respiratory tract.27 28 We cannot exclude the possibility that smoking or systemic effects of patients’ illness may have altered cytokine production or cellular responsiveness. Second, numbers of patients were small, reflecting low availability and technical issues in obtaining cells. While recognising this limitation, we felt that studying primary human cells would be by far the most relevant way to advance this area. Furthermore, consistent effects in studies of this nature help to generate hypotheses for further investigation. Third, as in any model system, we obviously cannot be certain that isolated, cultured epithelial cells behave as they would in their complex native environment. Finally, while epithelial cells are numerically dominant in the nose and alveoli, we cannot exclude the possibility that our stimuli might induce effects in other, less well-represented cells in these regions. Furthermore, in rodents it has been suggested that type I alveolar epithelial cells (notoriously hard to isolate from humans) respond more floridly to inflammatory stimuli than do type II cells.29

annot exclude the possibility that our stimuli might induce effects in other, less well-represented cells in these regions. Furthermore, in rodents it has been suggested that type I alveolar epithelial cells (notoriously hard to isolate from humans) respond more floridly to inflammatory stimuli than do type II cells.29 In summary, primary human alveolar epithelial cells appear to mount a more exuberant inflammatory response to PGN and TNF than do primary human nasal epithelial cells. PGN's effects may relate to the relative abundance and regulation of TLR2 in the upper and lower airway. TOLLIP is produced throughout the human respiratory tract. TOLLIP is expressed in greater levels in nasal cells than in alveolar epithelial cells, but differential TOLLIP expression in nasal and lung cells in response to bacterial virulence factors was not observed. These data suggest that relative expression of TLR2 and TOLLIP might play a role in the tolerant nature of the nasal epithelium to bacteria. Further studies are required to address a range of remaining questions—these include, but are by no means limited to: whether other TLR regulators are differentially expressed (constitutively or inducibly) in nasal versus alveolar epithelium; whether bacterial virulence factors differentially influence TLR regulator expression within alveolar epithelial cells (favouring a proinflammatory effect of PGN but not the other virulence factors measured here) and whether PGN can evade membrane-based TLR regulators on alveolar cells.

asal versus alveolar epithelium; whether bacterial virulence factors differentially influence TLR regulator expression within alveolar epithelial cells (favouring a proinflammatory effect of PGN but not the other virulence factors measured here) and whether PGN can evade membrane-based TLR regulators on alveolar cells. The authors are grateful to Professor Ian Poxton, University of Edinburgh, for providing ultrapure LPS, and to Dr Peter Barlow, Napier University, Edinburgh, for advice in performing experiments. Contributors: OLM-N designed the study, obtained clinical samples, performed experiments, analysed data and wrote the paper. TSW, MB, BJM and ROJ performed experiments and contributed to writing the manuscript. ACM performed statistical analysis and contributed to writing the manuscript. WSW, DJD and AJS designed the study, analysed data and wrote the paper. Funding: The study was funded by the UK Medical Research Council (clinical training fellowship to OLM-N), and by the Sir Jules Thorn Charitable Trust. DJD is an MRC Senior Research Fellow (G1002046). Competing interests: None. Ethics approval: The study was approved by the Lothian Research Ethics Committee (reference 08/S1102/32). Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Chronic obstructive pulmonary disease (COPD) is the second most common cause of emergency admission to hospital in the UK and one of the most costly inpatient conditions treated by the National Health Service (NHS).1 One in eight people over the age of 35 has undiagnosed COPD and, in recent years, there have been public health campaigns to identify those ‘missing millions’ in the UK.2 COPD is a diagnosis based on clinical symptoms, confirmed by the presence of obstructive spirometry3 and there is unreliable evidence for the initiation of large scale screening approaches to identify individuals at high risk. Therefore, a mechanism by which individuals can easily be identified as being high risk, from general practice (GP) databases, and be invited for spirometry, has important public health implications.

here is unreliable evidence for the initiation of large scale screening approaches to identify individuals at high risk. Therefore, a mechanism by which individuals can easily be identified as being high risk, from general practice (GP) databases, and be invited for spirometry, has important public health implications. Electronic health records (EHR) are an increasingly popular resource in which to conduct research. Owing to the large volume of patients encompassed, they provide tremendous statistical power to answer many clinical questions. However, the devil is in the detail, as the outcomes obtained from this type of research are only as good as the methods used to identify those outcomes to begin with. A recent paper by Jones et al4 highlighted the very important issue of missed opportunities for identifying individuals with COPD at an earlier point in time in primary care. However, in analysing these data, the authors did not appreciate the fact that Read codes used within GP databases change over time. Some of the codes in the code list they used to identify people with COPD did not exist at the time they attributed a GP to have missed an opportunity to make a COPD diagnosis. As a result, the number of diagnostic opportunities missed was overestimated.

e fact that Read codes used within GP databases change over time. Some of the codes in the code list they used to identify people with COPD did not exist at the time they attributed a GP to have missed an opportunity to make a COPD diagnosis. As a result, the number of diagnostic opportunities missed was overestimated. In this issue, Haroon et al5 have published a case–control study undertaken between 2000 and 2006 in which they have developed and validated a clinical risk score for use in primary care to identify people at risk of COPD using data from the Clinical Practice Research Datalink (CPRD). The clinical score incorporates smoking status, a previous diagnosis of asthma, lower respiratory tract infections (LRTI) and salbutamol prescriptions in the previous 3 years. The score developed, when applied to patients over the age of 35 years and used at their suggested threshold of 2.5, would have a positive predictive value of 22.6%, a negative predictive value of 97.6% and an overall screening yield of 3.5% for identifying people with COPD. In real terms, for every 29 records screened, 5 patients would require a clinical assessment to identify one patient with COPD. Development of clinical risk scores is important for population level screening, and the authors of this paper should be commended for their statistical rigour and the clarity of their methodology. However, there are several important caveats, which require the reader to be cautious in interpreting the results.

h COPD. Development of clinical risk scores is important for population level screening, and the authors of this paper should be commended for their statistical rigour and the clarity of their methodology. However, there are several important caveats, which require the reader to be cautious in interpreting the results. First, cases were identified between 2000 and 2006, spanning the implementation of the first quality and outcomes framework (QOF) for COPD in 2004,6 from which point the accuracy of coding associated with a COPD diagnosis improved significantly. While the authors acknowledge one of the weaknesses of their paper is not using a validated COPD codelist, thus bringing the accuracy of the diagnosis of COPD into question, this unfortunately leaves the reader questioning whether the outcome of this paper is actually just a score that predicts which patients will be mislabelled as having COPD in the future. While validation studies are laborious, they are crucial in ensuring accuracy of interrogation of EHR; use of a validated COPD definition would have strengthened this paper.7

tioning whether the outcome of this paper is actually just a score that predicts which patients will be mislabelled as having COPD in the future. While validation studies are laborious, they are crucial in ensuring accuracy of interrogation of EHR; use of a validated COPD definition would have strengthened this paper.7 Second, smoking is the predominant risk factor for COPD in the UK. Within EHR, recording of smoking is not always complete or accurate, particularly pre 2004. In this study, nearly 20% of the cases had never smoked, calling into question the accuracy of defining the cases. Likewise, misclassification of the controls may also have occurred, as the controls who smoked and had symptoms may well have had COPD and not had spirometry. How this affects the clinical risk tool will depend on the extent to which any misclassification is differential. Third, owing to the fluctuating nature of coding, the lack of longevity of specific Read codes, the introduction of new Read codes and new coding systems being introduced (such as SNOMED-CT), tools such as this have a limited lifetime of usage. Furthermore, with changes in clinical practice and updates to QOF, this algorithm would benefit from validation in a more recent data set. Finally, the number and timing of several of the variables used in the score are likely to affect accuracy. For example, a patient's risk of COPD is likely to be different if they have had one LRTI in the proceeding 3 years as opposed to more.

QOF, this algorithm would benefit from validation in a more recent data set. Finally, the number and timing of several of the variables used in the score are likely to affect accuracy. For example, a patient's risk of COPD is likely to be different if they have had one LRTI in the proceeding 3 years as opposed to more. Certainly what this paper does provide is an excellent preliminary analysis from which future models could be developed, but this risk score in its current format is by no means a definitive tool. Undoubtedly, a validation tool that could be incorporated into GP software to identify people at risk of COPD would be extremely useful. Competing interests: None. Provenance and peer review: Not commissioned; internally peer reviewed. Data sharing statement: No additional data are available.

Key messages The prevalence of respiratory symptoms in Izhevsk was high and associated with substantial cardiovascular comorbidity. The prevalence of smoking was extremely high even among those with severe breathlessness, suggesting the need for improved smoking cessation treatment in those with respiratory symptoms. This is one of the very few papers investigating the burden of respiratory disease in the Russian Federation and shows the need for further research into this neglected area. Introduction Chronic respiratory diseases are a major cause of morbidity and mortality worldwide.1 2 Tobacco smoking is an important risk factor for lung disease.3 Prevalence of smoking in Russia is very high among men, estimated at 60.2% in the 2009 Global Adult Tobacco Survey,4 and has shown no evidence of a decline over time,5 6 thus suggesting that the burden of chronic respiratory disease in Russia is likely to be substantial. In addition, occupational exposure to vapours, dusts and fumes are likely to be important risk factors for respiratory diseases among industrial workers in Russia.7

and has shown no evidence of a decline over time,5 6 thus suggesting that the burden of chronic respiratory disease in Russia is likely to be substantial. In addition, occupational exposure to vapours, dusts and fumes are likely to be important risk factors for respiratory diseases among industrial workers in Russia.7 Despite a high prevalence of smoking in Russia, very little attention has been paid to the burden of chronic respiratory problems in the Russian population. We are aware of only two studies investigating the prevalence of respiratory symptoms in the general population. Vietri et al8 found in the analyses of 5920 adults, aged 40 years or older in the 2011 Russia National Health and Wellness Survey, that 54% of respondents reported one or more symptoms of chronic respiratory disease (45% reported shortness of breath, 27% reported coughing up mucous, and 18% reported wheezing). Chuchalin et al9 found in a cross-sectional survey of 7164 adults from 12 regions in Russia that 25.7% reported an attack of wheezing or whistling with breathlessness and 8.6% reported symptoms of chronic bronchitis. Both previous studies suggest that the prevalence of chronic respiratory disease in Russia is high.

g). Chuchalin et al9 found in a cross-sectional survey of 7164 adults from 12 regions in Russia that 25.7% reported an attack of wheezing or whistling with breathlessness and 8.6% reported symptoms of chronic bronchitis. Both previous studies suggest that the prevalence of chronic respiratory disease in Russia is high. Russia has one of the highest levels of cardiovascular disease mortality in the world.10 There is evidence that respiratory disease and cardiovascular disease are linked, with associations found between cardiovascular morbidity and mortality and chronic obstructive pulmonary disease (COPD),11–16 asthma,17–19 and idiopathic pulmonary fibrosis.20 Cardiovascular disease is one of the main causes of death in both patients with COPD11 14 16 21 and asthma.22 23 The reasons for links between respiratory and cardiovascular diseases are not well understood. Some of these associations may be due to smoking as a common cause; however, systemic inflammation has also been suggested as a potential mechanism.11 16 Despite the high prevalence of cardiovascular disease, to our knowledge, no previous studies have investigated the association between respiratory and cardiovascular morbidity in Russia. The aim of this study was to investigate the prevalence of respiratory symptoms (chronic productive cough and breathlessness) and their associations with cardiovascular risk factors and self-reported comorbidities among working-age men in the city of Izhevsk, Russia.

Russia has one of the highest levels of cardiovascular disease mortality in the world.10 There is evidence that respiratory disease and cardiovascular disease are linked, with associations found between cardiovascular morbidity and mortality and chronic obstructive pulmonary disease (COPD),11–16 asthma,17–19 and idiopathic pulmonary fibrosis.20 Cardiovascular disease is one of the main causes of death in both patients with COPD11 14 16 21 and asthma.22 23 The reasons for links between respiratory and cardiovascular diseases are not well understood. Some of these associations may be due to smoking as a common cause; however, systemic inflammation has also been suggested as a potential mechanism.11 16 Despite the high prevalence of cardiovascular disease, to our knowledge, no previous studies have investigated the association between respiratory and cardiovascular morbidity in Russia. The aim of this study was to investigate the prevalence of respiratory symptoms (chronic productive cough and breathlessness) and their associations with cardiovascular risk factors and self-reported comorbidities among working-age men in the city of Izhevsk, Russia. Methods Study sample The study sample was 938 working-age men (25–60 years old) who were residents in the city of Izhevsk, an industrial city in Russia located 1300 km south east of Moscow, West of the Ural Mountains.

The aim of this study was to investigate the prevalence of respiratory symptoms (chronic productive cough and breathlessness) and their associations with cardiovascular risk factors and self-reported comorbidities among working-age men in the city of Izhevsk, Russia. Methods Study sample The study sample was 938 working-age men (25–60 years old) who were residents in the city of Izhevsk, an industrial city in Russia located 1300 km south east of Moscow, West of the Ural Mountains. These men were participants in the Izhevsk Family Study 2 (IFS-2). This was a follow-up study of 1941 men living in Izhevsk originally recruited as population controls in a case–control study of the association between hazardous alcohol consumption and premature mortality in 2003–2006.24 The original controls were living men frequency matched by age to cases (men aged 25–54 years who had died in the past year) selected at random from the population list of Izhevsk in 2002. In IFS-2 (2008–2010), 1515 of these men were followed up, interviewed by a trained interviewer and invited to attend a health check. The health check conducted by a doctor included questions on medical history, three measurements of blood pressure and collection of a blood sample. There were four different doctors who conducted the health check. Participants were also given a self-completed questionnaire which included the 12-Item Short-Form Health Survey (SF-12) questions.25 For the majority of men, the health check took place in a polyclinic, but those who were unable to attend the clinic were offered the option of having this done in their own home.

check. Participants were also given a self-completed questionnaire which included the 12-Item Short-Form Health Survey (SF-12) questions.25 For the majority of men, the health check took place in a polyclinic, but those who were unable to attend the clinic were offered the option of having this done in their own home. Out of the 1515/2041 men successfully reinterviewed as part of the IFS-2 study, 1052 attended the health check. For the purposes of this study, the sample was restricted to 983 men, with complete data collected on self-reported respiratory symptoms. A flow chart of participation in the study is shown in figure 1. Figure 1 Flow chart of participants. IFS-2 was approved by the Ethics Committees of the London School of Hygiene & Tropical Medicine and Izhevsk Medical Academy. Informed written consent was obtained from all participants to take part in the health check. Definition of respiratory symptoms No spirometry data were collected at the health check; therefore, it was not possible to investigate prevalence of a specific disease such as COPD or asthma. However, data were collected on self-reported respiratory symptoms (chronic productive cough and breathlessness). Men who reported ‘Don't know/difficult to answer’ to any of the questions were coded as missing data on self-reported symptoms. The impact of this was assessed in sensitivity analyses.

ch as COPD or asthma. However, data were collected on self-reported respiratory symptoms (chronic productive cough and breathlessness). Men who reported ‘Don't know/difficult to answer’ to any of the questions were coded as missing data on self-reported symptoms. The impact of this was assessed in sensitivity analyses. Chronic productive cough was defined by one positive answer to the following questions: ‘Do you usually cough up any phlegm from your chest first thing in morning in winter?’, and/or ‘Do you usually cough up any phlegm from your chest, during the day or at night in winter?’, and/or ‘Do you bring up phlegm like this on most days for as much as 3 months each year?’. Breathlessness was defined as a positive answer to one or more of the questions: (1) ‘Are you troubled by shortness of breath when hurrying on level ground or walking up a slight hill?’, and/or (2) ‘Do you get short of breath walking with people of your own age group on level ground?’, and/or (3) ‘Do you have to stop for breath when walking at your own pace on level ground?’, and/or (4) ‘Have you had attacks of wheezing or whistling in your chest at any time in the last 12 months?’, and/or (5) ‘Have you at any time in the past 12 months been woken at night by an attack of shortness of breath?’.

nd?’, and/or (3) ‘Do you have to stop for breath when walking at your own pace on level ground?’, and/or (4) ‘Have you had attacks of wheezing or whistling in your chest at any time in the last 12 months?’, and/or (5) ‘Have you at any time in the past 12 months been woken at night by an attack of shortness of breath?’. Breathlessness questions 1, 2 and 3 are part of the UK Medical Research Council (MRC) breathlessness scale, a validated tool for measuring disability associated with breathlessness,26 and were used to further divide men with respiratory symptoms into those with breathlessness grade 2 or less, and those with breathlessness grade 3 or higher. Question 4 was used to identify prevalence of wheeze and question 5, prevalence of nocturnal dyspnoea. Definition of exposures Comorbidities Comorbidities included were hypertension, diabetes, angina, previous myocardial infarction (MI) and previous stroke. All data on comorbidities were collected at the health check. Three measurements of blood pressure were taken during the health check. Men were categorised as having hypertension if mean systolic blood pressure using the second and third blood pressure measurements was greater than 139 mm Hg or diastolic blood pressure was greater than 89 mm Hg and/or participants reported taking any antihypertensive medications either daily or sometimes. Angina was determined from self-reported angina symptoms measured using the Rose angina questionnaire.27 Classification of diabetes mellitus, MI and stroke were all assessed from self-report of these conditions.

andardised instrument for measuring self-reported health, derived as a shorter alternative to the SF-36.25 In accordance with standard practice, two summary scales were derived from these questions: a physical health component score and a mental health component score. Both of these summary scores ranged from 0 to 100. Cardiovascular risk factors Other exposures of interest were body mass index (BMI; kg/m2), total cholesterol, high-density lipoprotein cholesterol (HDL-C), and triglycerides measured in mmol/L, and B-type natriuretic peptide (BNP) measured in pg/mL. BNP is a biomarker of heart failure which is released with abnormal cardiac wall strain. Levels may also be raised for other reasons, including age, myocardial ischaemia, atrial fibrillation and renal dysfunction.28 29 BNP is an independent predictor of cardiovascular morbidity and mortality,30 31 and has been found to be strongly associated with hazardous alcohol consumption in this population.32 The distributions of triglycerides and BNP were skewed; therefore, these variables were transformed using the natural logarithm. Potential confounding variables Age Age was categorised into 5-year age groups and used as a continuous variable in analyses. Education Three categories of education were used: incomplete secondary, secondary, and higher or incomplete higher education. Owing to data sparsity, education was used as a continuous variable in analyses.

Potential confounding variables Age Age was categorised into 5-year age groups and used as a continuous variable in analyses. Education Three categories of education were used: incomplete secondary, secondary, and higher or incomplete higher education. Owing to data sparsity, education was used as a continuous variable in analyses. Smoking Self-reported smoking status was categorised into five groups: never smoked, ex-smoker, current smoker 1–5 cigarettes per day, current smoker 11–20 cigarettes per day and current smoker >20 cigarettes per day. Alcohol consumption Alcohol consumption was considered as a potential confounder as it is strongly associated with smoking status and is also an important predictor of general health status and cardiovascular mortality in Russia.24 33 34

Smoking Self-reported smoking status was categorised into five groups: never smoked, ex-smoker, current smoker 1–5 cigarettes per day, current smoker 11–20 cigarettes per day and current smoker >20 cigarettes per day. Alcohol consumption Alcohol consumption was considered as a potential confounder as it is strongly associated with smoking status and is also an important predictor of general health status and cardiovascular mortality in Russia.24 33 34 Two aspects of alcohol consumption in the past 12 months were considered: quantity of alcohol consumed and hazardous drinking pattern. Quantity of alcohol consumed was measured using total volume of ethanol consumed in the past year calculated from questions on the frequency and usual quantity of beer, wine and spirits consumed. Using an approach developed in our previous work, men were also classified as non-drinkers, non-hazardous drinkers and hazardous drinkers.24 35 Men were classified as hazardous drinkers if they reported twice weekly or more frequency of hangover and/or excessive drunkenness and/or sleeping in clothes at night because of drunkenness and/or failing their family or personal obligations because of drinking and/or drinking non-beverage alcohols (sources of ethanol not intended for drinking such as eau de cologne) and/or one or more episodes of zapoi (a period of 2 or more days of drinking during which a participant is withdrawn from normal social life).

d/or failing their family or personal obligations because of drinking and/or drinking non-beverage alcohols (sources of ethanol not intended for drinking such as eau de cologne) and/or one or more episodes of zapoi (a period of 2 or more days of drinking during which a participant is withdrawn from normal social life). Medication use Men were asked to report all medications they were taking at the time of the health check coded by a cardiologist. Use of medication for a respiratory problem was assessed by self-reported use of medications classified as bronchodilators, glucocorticoids and mucolytics. Data were not specifically collected on route of admission (ie, oral or inhaled). Statistical analysis The prevalence of chronic cough and breathlessness was directly standardised by age using the 2013 European Standard Population aged 25–59 years. Logistic regression models were fitted with each of the cardiovascular comorbidities as the outcome (hypertension, diabetes, angina, MI and stroke) and respiratory symptoms categorised by degree of breathlessness as the exposure variable.

Statistical analysis The prevalence of chronic cough and breathlessness was directly standardised by age using the 2013 European Standard Population aged 25–59 years. Logistic regression models were fitted with each of the cardiovascular comorbidities as the outcome (hypertension, diabetes, angina, MI and stroke) and respiratory symptoms categorised by degree of breathlessness as the exposure variable. It was necessary to limit, where possible, the number of confounders which could be included in the models because of sparsity of data for four of the outcomes (diabetes, angina, MI and stroke). Models were adjusted for age and education as a priori confounders and then additionally for smoking status. Smoking was considered as a potential confounder because it is strongly associated with both respiratory and cardiovascular diseases. However, smoking status could also be affected by both respiratory and cardiovascular diseases, and as people may stop smoking due to illness, adjustment could also lead to spurious associations due to reverse causality. Models were, therefore, considered with and without adjustment for smoking status. Although alcohol consumption was considered a priori as a potential confounder, no strong associations were found between the two different measures of alcohol use and respiratory symptoms in univariate analysis; therefore, alcohol consumption was not included as a confounder in the final model.

ent for smoking status. Although alcohol consumption was considered a priori as a potential confounder, no strong associations were found between the two different measures of alcohol use and respiratory symptoms in univariate analysis; therefore, alcohol consumption was not included as a confounder in the final model. The associations between self-reported physical and mental health components of the SF-12 and serum lipids (total cholesterol, HDL-C and triglycerides), BNP and BMI were assessed using linear regression models with self-reported respiratory symptoms as exposure variables. Models were adjusted for age, education and smoking status. Sensitivity analysis There were 63 men who answered “don't know” to at least one question on their respiratory symptoms. In the main analyses, these men were excluded as missing. However, a sensitivity analysis was carried out first recategorising ‘don't know’ answers to ‘no’ for that particular question, and then recategorising ‘don't know’ answers to ‘yes’ for that particular question. Analyses were rerun under these two alternative scenarios. Results Overall data were available on symptoms of chronic cough and breathlessness for 983 men. There were no differences in these men with respect to age, education, smoking status or alcohol use from data available on respiratory symptoms compared with all the 1515 men who were interviewed in the IFS-2 study.

Sensitivity analysis There were 63 men who answered “don't know” to at least one question on their respiratory symptoms. In the main analyses, these men were excluded as missing. However, a sensitivity analysis was carried out first recategorising ‘don't know’ answers to ‘no’ for that particular question, and then recategorising ‘don't know’ answers to ‘yes’ for that particular question. Analyses were rerun under these two alternative scenarios. Results Overall data were available on symptoms of chronic cough and breathlessness for 983 men. There were no differences in these men with respect to age, education, smoking status or alcohol use from data available on respiratory symptoms compared with all the 1515 men who were interviewed in the IFS-2 study. In the population, overall prevalence of chronic cough was 44.0% (n=432), breathlessness (grade 2 or above) was 33.9% (n=333), breathlessness (grade 3 or above) was 8.9% (n=87), wheeze was 4.3% (n=42), and nocturnal dyspnoea was 3.2% (n=31).There were 204 men (20.8%) who reported symptoms of both cough and breathlessness (age-standardised prevalence 20.9% (95% CI 18.4% to 23.5%)). Among men with chronic cough, 49 (24.0%) reported grade 3 or above breathlessness. The prevalence of wheeze among those with chronic cough and breathlessness was 16.7% (34 men) and nocturnal dyspnoea 11.4% (23 men). The prevalence of respiratory symptoms by smoking status is shown in table 1. Table 1 Prevalence of self-reported respiratory symptoms by smoking status

In the population, overall prevalence of chronic cough was 44.0% (n=432), breathlessness (grade 2 or above) was 33.9% (n=333), breathlessness (grade 3 or above) was 8.9% (n=87), wheeze was 4.3% (n=42), and nocturnal dyspnoea was 3.2% (n=31).There were 204 men (20.8%) who reported symptoms of both cough and breathlessness (age-standardised prevalence 20.9% (95% CI 18.4% to 23.5%)). Among men with chronic cough, 49 (24.0%) reported grade 3 or above breathlessness. The prevalence of wheeze among those with chronic cough and breathlessness was 16.7% (34 men) and nocturnal dyspnoea 11.4% (23 men). The prevalence of respiratory symptoms by smoking status is shown in table 1. Table 1 Prevalence of self-reported respiratory symptoms by smoking status Respiratory symptom Non-smoker Ex-smoker Current smoker All men N (%) N (%) N (%) N (%) Chronic productive cough 24 (12.6) 40 (21.7) 368 (60.6) 432 (44.0) MRC breathlessness grade 2 34 (17.8) 39 (21.2) 173 (28.5) 246 (25.1) MRC breathlessness grade 3 and above 11 (5.8) 18 (9.8) 58 (9.6) 87 (8.9) Wheeze 5 (2.6) 4 (2.2) 33 (5.4) 42 (4.3) Nocturnal dyspnoea 6 (3.1) 6 (3.3) 19 (3.1) 31 (3.2) Chronic cough and breathlessness* 9 (4.7) 17 (9.2) 178 (29.3) 204 (20.8) Total 191 (100) 184 (100) 607 (100) 982† (100) *Breathlessness defined as MRC breathless grade 2 or above and/or wheeze and/or nocturnal dyspnoea. †Smoking status missing for one man. MRC, Medical Research Council.

Respiratory symptom Non-smoker Ex-smoker Current smoker All men N (%) N (%) N (%) N (%) Chronic productive cough 24 (12.6) 40 (21.7) 368 (60.6) 432 (44.0) MRC breathlessness grade 2 34 (17.8) 39 (21.2) 173 (28.5) 246 (25.1) MRC breathlessness grade 3 and above 11 (5.8) 18 (9.8) 58 (9.6) 87 (8.9) Wheeze 5 (2.6) 4 (2.2) 33 (5.4) 42 (4.3) Nocturnal dyspnoea 6 (3.1) 6 (3.3) 19 (3.1) 31 (3.2) Chronic cough and breathlessness* 9 (4.7) 17 (9.2) 178 (29.3) 204 (20.8) Total 191 (100) 184 (100) 607 (100) 982† (100) *Breathlessness defined as MRC breathless grade 2 or above and/or wheeze and/or nocturnal dyspnoea. †Smoking status missing for one man. MRC, Medical Research Council. The level of use of medications for respiratory problems was very low. Ten men reported using bronchodilators (seven with symptoms of chronic cough and breathlessness). Four men reported using glucocorticoids (two with symptoms of chronic cough and breathlessness) and seven men reported use of mucolytics (three with symptoms of chronic cough and breathlessness).

ory problems was very low. Ten men reported using bronchodilators (seven with symptoms of chronic cough and breathlessness). Four men reported using glucocorticoids (two with symptoms of chronic cough and breathlessness) and seven men reported use of mucolytics (three with symptoms of chronic cough and breathlessness). The majority of men with chronic cough and breathlessness were current smokers (87.3%).There was no evidence of a systematic difference in smoking behaviour by grade of breathlessness (p=0.54) despite a higher percentage of ex-smokers in those with breathlessness grade 3 or above compared with those with grade 2 breathlessness (6.5% vs 14.3%). The distribution of men with and without symptoms of chronic cough and breathlessness by sociodemographic factors, smoking, cardiovascular risk factors, self-reported health and comorbidities is shown in table 2. There was strong evidence of an association between self-reported symptoms of chronic cough and breathlessness and education. The proportion of men with higher education decreased as severity of breathlessness increased. Different patterns emerged for the two measures of alcohol use: total volume consumed was highest among men with chronic cough and MRC breathlessness of grade 2 or less, but lowest among men with chronic cough and MRC breathlessness grade 3 or less. Conversely, the proportion of hazardous drinkers in the sample increased with severity of breathlessness. However, there was only weak evidence for an association with either measure of alcohol use and self-reported symptoms of chronic cough and breathlessness.

with chronic cough and MRC breathlessness grade 3 or less. Conversely, the proportion of hazardous drinkers in the sample increased with severity of breathlessness. However, there was only weak evidence for an association with either measure of alcohol use and self-reported symptoms of chronic cough and breathlessness. Table 2 Distribution of sample by chronic cough and MRC breathlessness grade, sociodemographic variables, smoking, cardiovascular risk factors and health status

with chronic cough and MRC breathlessness grade 3 or less. Conversely, the proportion of hazardous drinkers in the sample increased with severity of breathlessness. However, there was only weak evidence for an association with either measure of alcohol use and self-reported symptoms of chronic cough and breathlessness. Table 2 Distribution of sample by chronic cough and MRC breathlessness grade, sociodemographic variables, smoking, cardiovascular risk factors and health status No respiratory symptoms Chronic cough and breathlessness ≤Grade 2 breathlessness ≥Grade 3 breathlessness N, mean or median (%,SD or IQR) N, mean or median (%, SD or IQR) N, mean or median (%, SD or IQR) Age (years) 25–29 10 (1.3) 3 (1.9) 0 (0.0) 30–34 58 (7.5) 12 (7.7) 2 (4.1) 35–39 63 (8.1) 20 (12.9) 4 (8.2) 40–44 96 (12.3) 14 (9.0) 3 (6.1) 45–49 160 (20.5) 29 (18.7) 3 (6.1) 50–54 187 (24.0) 34 (21.9) 12 (24.5) 55–60 205 (26.3) 43 (27.7) 25 (51.0) χ2 (df) 23.13 (12) p=0.03 Education Incomplete secondary 31 (4.0) 9 (5.8) 3 (6.1) Secondary 556 (71.4) 123 (79.4) 43 (87.8) Higher or incomplete higher 192 (24.7) 23 (14.8) 3 (6.1) χ2 (df) 15.5 (4) p=0.004 Smoking status (missing=1) Never smoked 182 (23.4) 7 (4.5) 2 (4.1) Ex-smoker 167 (21.5) 10 (6.5) 7 (14.3) Current smoker (1–10/day) 81 (10.4) 25 (16.1) 7 (14.3) Current smoker (11–20/day) 274 (35.2) 82 (52.9) 25 (51.0) Current smoker (>20/day) 74 (9.5) 31 (20.0) 8 (16.3) χ2 (df) 75.9 (8) p<0.001 Total volume of ethanol (litres per year) (missing=13) Median (IQR) 4.0 (1.2–10.1) 5.9 (1.7–12.5) 3.1 (1.2–17.3) Kruskal-Wallis p=0.13 Hazardous drinker (missing=12) Non-drinker 104 (13.5) 18 (11.7) 6 (12.2) Drinker (non-hazardous) 602 (78.4) 117 (76.0) 34 (69.4) Drinker (hazardous) 62 (8.1) 19 (12.3) 9 (18.4) χ2 (df) 8.03 (4) p=0.09 Body mass index (missing=4) Mean (SD) 26.2 (4.2) 26.4 (4.7) 27.8 (5.2) Test for trend p=0.04 Total cholesterol mmol/L (missing=66) Mean (SD) 5.44 (1.02) 5.27 (0.96) 5.22 (1.01) Test for trend p=0.03 HDL-C mmol/L (missing=66) Mean (SD) 1.44 (0.44) 1.40 (0.48) 1.36 (0.40) Test for trend p=0.14 Log triglycerides (missing=82) Mean (SD) 0.29 (0.51) 0.26 (0.50) 0.30 (0.52) Test for trend p=0.73 Log BNP pg/mL (missing=74) Mean (SD) 2.42 (0.87) 2.59 (1.03) 2.99 (0.96) Test for trend p<0.001 SF-12 (missing=15) Mean Physical health score (SD) 48.8 (7.7) 45.2 (7.7) 36.1 (9.3) Test for trend p<0.001 Mean Mental health score (SD) 49.2 (8.6) 47.9 (8.8) 43.6 (10.3) Test for trend p<0.001 Hypertension (missing=2) Yes 477 (61.4) 97 (62.6) 41 (83.7) χ2 (df) 9.79(2) p=0.007 Diabetes (Missing=7) Yes 9 (1.2) 3 (2.0) 4 (8.3) χ2 (df) 14.54 (2) p=0.001 Angina (missing=9) Yes 43 (5.6) 24 (15.6) 14 (29.2) χ2 (df) 45.7 (2

<0.001 Mean Mental health score (SD) 49.2 (8.6) 47.9 (8.8) 43.6 (10.3) Test for trend p<0.001 Hypertension (missing=2) Yes 477 (61.4) 97 (62.6) 41 (83.7) χ2 (df) 9.79(2) p=0.007 Diabetes (Missing=7) Yes 9 (1.2) 3 (2.0) 4 (8.3) χ2 (df) 14.54 (2) p=0.001 Angina (missing=9) Yes 43 (5.6) 24 (15.6) 14 (29.2) χ2 (df) 45.7 (2 ) p<0.001 Myocardial infarction (missing=2) Yes 11 (1.4) 2 (1.3) 5 (10.2) χ2 (df) 20.1 (2) p<0.001 Stroke (missing=3) Yes 8 (1.0) 1 (0.70) 4 (8.2) χ2 (df) 18.6 (2) p<0.001 Total 779 (100) 155 (100) 49 (100) BNP, B-type natriuretic peptide; HDL-C, high-density lipoprotein cholesterol; SF-12, 12-Item Short-Form Health Survey. The cross-sectional associations between self-reported chronic cough and breathlessness and comorbidities are shown in table 3. The odds of hypertension, diabetes, angina, MI and stroke were all raised in men with self-reported cough with breathlessness of grade 3 or above compared with men with no respiratory symptoms. There was strong evidence that the odds of diabetes, angina and MI, and good evidence that the odds of hypertension and stroke increased with severity of breathlessness. These associations remained on adjusting for smoking. Table 3 Association between self-reported chronic cough and breathlessness symptoms and other comorbidities

The cross-sectional associations between self-reported chronic cough and breathlessness and comorbidities are shown in table 3. The odds of hypertension, diabetes, angina, MI and stroke were all raised in men with self-reported cough with breathlessness of grade 3 or above compared with men with no respiratory symptoms. There was strong evidence that the odds of diabetes, angina and MI, and good evidence that the odds of hypertension and stroke increased with severity of breathlessness. These associations remained on adjusting for smoking. Table 3 Association between self-reported chronic cough and breathlessness symptoms and other comorbidities No respiratory symptoms Chronic cough and breathlessness Self-reported comorbidity Odds ratio (95% CI) ≤Grade 2 breathlessness ≥Grade 3 breathlessness Test for trend Hypertension (n=980) Model 1 1 (ref) 1.12 (0.77 to 1.62) 2.84 (1.29 to 6.26) p=0.04 Model 2 1 (ref) 1.19 (0.81 to 1.74) 3.03 (1.36 to 6.74) p=0.02 Diabetes (n=975) Model 1 1 (ref) 1.86 (0.49 to 7.03) 7.27 (2.05 to 25.85) p=0.008 Model 2 1 (ref) 2.91 (0.68 to 12.37) 10.55 (2.69 to 41.37) p=0.002 Angina (n=973) Model 1 1 (ref) 3.29 (1.91 to 5.65) 6.74 (3.31 to 13.75) p<0.001 Model 2 1 (ref) 3.73 (2.09 to 6.66) 7.54 (3.61 to 15.73) p<0.001 Myocardial infarction (n=980)* Model 1 1 (ref) 0.91 (0.20 to 4.20) 5.18 (1.65 to 16.22) p=0.04 Model 2 1 (ref) 1.40 (0.27 to 7.17) 7.61 (2.10 to 27.49) p=0.008 Stroke (n=979)* Model 1 1 (ref) 0.71 (0.09 to 5.77) 7.76 (2.09 to 28.88) p=0.03 Model 2 1 (ref) 0.60 (0.07 to 5.01) 6.61 (1.73 to 25.34) p=0.06 Model 1: adjusted for age and education.

=980)* Model 1 1 (ref) 0.91 (0.20 to 4.20) 5.18 (1.65 to 16.22) p=0.04 Model 2 1 (ref) 1.40 (0.27 to 7.17) 7.61 (2.10 to 27.49) p=0.008 Stroke (n=979)* Model 1 1 (ref) 0.71 (0.09 to 5.77) 7.76 (2.09 to 28.88) p=0.03 Model 2 1 (ref) 0.60 (0.07 to 5.01) 6.61 (1.73 to 25.34) p=0.06 Model 1: adjusted for age and education. Model 2: model 1 + smoking status. *For myocardial infarction and stroke, the smoking variable was recoded so that the category >20 cigarettes/day merged with 11–20 cigarettes per day; otherwise 113 observations dropped due to perfect prediction.

=980)* Model 1 1 (ref) 0.91 (0.20 to 4.20) 5.18 (1.65 to 16.22) p=0.04 Model 2 1 (ref) 1.40 (0.27 to 7.17) 7.61 (2.10 to 27.49) p=0.008 Stroke (n=979)* Model 1 1 (ref) 0.71 (0.09 to 5.77) 7.76 (2.09 to 28.88) p=0.03 Model 2 1 (ref) 0.60 (0.07 to 5.01) 6.61 (1.73 to 25.34) p=0.06 Model 1: adjusted for age and education. Model 2: model 1 + smoking status. *For myocardial infarction and stroke, the smoking variable was recoded so that the category >20 cigarettes/day merged with 11–20 cigarettes per day; otherwise 113 observations dropped due to perfect prediction. The cross-sectional associations between self-reported chronic cough and breathlessness symptoms, self-reported health, serum lipids, BNP, and BMI are shown in table 4. There was strong evidence for a substantial decrease in physical self-reported health score with severity of breathlessness symptoms as men with cough and grade 3 breathlessness had a mean physical health score 11.58 points (95% CI 9.34 to 13.83) lower than men who did not report any respiratory symptoms. There was also a corresponding but smaller inverse trend in self-reported mental health (men with grade 3 breathlessness reported a mean mental health score 5.90 points lower than men with no respiratory symptoms (95% CI 8.51 to 3.29)). There was some evidence for a negative association with total cholesterol (test for trend p=0.06) and HDL-C (test for trend p=0.01) and respiratory symptoms with evidence of a decrease with increasing severity of breathlessness. Mean BMI was higher in men with chronic cough and breathlessness; there was strong evidence for a positive trend with severity of breathlessness (test for trend p<0.001). There was also good evidence of a positive trend in log BNP with severity of breathlessness (test for trend p=0.006). There was no evidence for an association between respiratory symptoms and triglycerides (test for trend p=0.56).

g evidence for a positive trend with severity of breathlessness (test for trend p<0.001). There was also good evidence of a positive trend in log BNP with severity of breathlessness (test for trend p=0.006). There was no evidence for an association between respiratory symptoms and triglycerides (test for trend p=0.56). Table 4 Association between self-reported chronic cough and breathlessness symptoms and self-reported health, lipids, BNP and body mass index

g evidence for a positive trend with severity of breathlessness (test for trend p<0.001). There was also good evidence of a positive trend in log BNP with severity of breathlessness (test for trend p=0.006). There was no evidence for an association between respiratory symptoms and triglycerides (test for trend p=0.56). Table 4 Association between self-reported chronic cough and breathlessness symptoms and self-reported health, lipids, BNP and body mass index No respiratory symptoms Chronic cough and breathlessness Test for trend ≤Grade 2 breathlessness ≥Grade 3 breathlessness Coefficient (95% CI) SF-12 PCS (n=967) Model 1 1 (ref) −3.49 (−4.80 to −2.18) −11.56 (−13.78 to −9.33) p<0.001 Model 2 1 (ref) −3.60 (−4.95 to −2.25) −11.58 (−13.83 to −9.34) p<0.001 SF-12 MCS (n=967) Model 1 1 (ref) −1.36 (−2.88 to 0.16) −5.97 (−8.55 to −3.39) p<0.001 Model 2 1 (ref) −1.33 (−2.90 to 0.24) −5.90 (−8.51 to −3.29) p<0.001 Total cholesterol (n=916) Model 1 1 (ref) −0.16 (−0.33 to 0.02) −0.25 (−0.55 to 0.05) p=0.02 Model 2 1 (ref) −0.13 (−0.31 to 0.06) −0.23 (−0.53 to 0.08) p=0.06 HDL-C (n=916) Model 1 1 (ref) −0.05 (−0.13 to 0.03) −0.10 (−0.23 to 0.03) p=0.07 Model 2 1 (ref) −0.08 (−0.16 to 0.002) −0.13 (−0.26 to 0.01) p=0.01 Log triglycerides (n=900) Model 1 1 (ref) −0.02 (−0.11 to 0.07) 0.02 (−0.14 to 0.17) p=0.88 Model 2 1 (ref) 0.01 (−0.08 to 0.11) 0.04 (−0.11 to 0.20) p=0.56 Log BNP (n=908) Model 1 1 (ref) 0.16 (0.003 to 0.31) 4.14 (0.16 to 0.67) p=0.001 Model 2 1 (ref) 0.11 (−0.05 to 0.26) 0.37 (0.12 to 0.63) p=0.006 Body mass index (n=978) Model 1 1 (ref) 0.37 (−0.38 to 1.12) 1.86 (0.60 to 3.12) p=0.01 Model 2 1 (ref) 0.95 (0.19 to 1.70) 2.29 (1.05 to 3.53) p<0.001 Model 1: adjusted for age and education.

del 1 1 (ref) 0.16 (0.003 to 0.31) 4.14 (0.16 to 0.67) p=0.001 Model 2 1 (ref) 0.11 (−0.05 to 0.26) 0.37 (0.12 to 0.63) p=0.006 Body mass index (n=978) Model 1 1 (ref) 0.37 (−0.38 to 1.12) 1.86 (0.60 to 3.12) p=0.01 Model 2 1 (ref) 0.95 (0.19 to 1.70) 2.29 (1.05 to 3.53) p<0.001 Model 1: adjusted for age and education. Model 2: model 1 + smoking status. BNP, B-type natriuretic peptide; HDL-C, high-density lipoprotein cholesterol; MCS, mental health component score; PCS, physical health component score; SF-12, 12-Item Short-Form Health Survey. Sensitivity analyses When men who answered ‘don't know’ to any particular question were included, the sample size was 1046. Assuming that all men who answered ‘don't know’ did not have the symptom, the prevalence of cough was 45.4%, any breathlessness was 34.5%, breathlessness grade 3 or above was 9.3%, and for both cough and breathlessness was 21.9%. Under the alternative scenario that all men who answered ‘don't know’ did have that symptom, the prevalence of cough was 45.8%, any breathlessness was 37.2%, breathlessness grade 3 or above was 11.6%, and for both cough and breathlessness was 23.5%. The substantive associations between self-reported cough and breathlessness and the other health outcomes were not altered under either scenario with the exception that in both the sensitivity analyses, no associations were seen between self-reported respiratory symptoms and total cholesterol (data not shown).

s 23.5%. The substantive associations between self-reported cough and breathlessness and the other health outcomes were not altered under either scenario with the exception that in both the sensitivity analyses, no associations were seen between self-reported respiratory symptoms and total cholesterol (data not shown). Discussion In this sample of working-age men living in the industrial city of Izhevsk, the age-standardised prevalence of self-reported symptoms of chronic cough and breathlessness was 20.9% (age-standardised prevalence of chronic cough with grade 3 or above breathlessness was 3.7%). The prevalence of chronic cough was 44.0% and breathlessness grade 2 or above was 33.9%. This is high compared with a prevalence of self-reported cough and/or breathlessness from the BREATHE study of 12.2% among 31 418 men living in 11 countries in the Middle East and North Africa.36 The prevalence of cough and breathlessness among men in Izhvesk was higher than for any of the countries included in the BREATHE study (range 6.1% in the United Arab Emirates to15.9% in Pakistan). Chronic cough and breathlessness are both symptoms of COPD and the high prevalence of these symptoms suggests prevalence of COPD is likely to be high in this population. This is consistent with the one previous study which has investigated prevalence of COPD in Russia using spirometry; the study estimated overall prevalence in both men and women of 15.3%9 compared with an international prevalence of 10.1% in the BOLD study (11.8% in men).37 Further studies in Russia using spirometry are needed to provide more reliable estimates of the burden of COPD.

ted prevalence of COPD in Russia using spirometry; the study estimated overall prevalence in both men and women of 15.3%9 compared with an international prevalence of 10.1% in the BOLD study (11.8% in men).37 Further studies in Russia using spirometry are needed to provide more reliable estimates of the burden of COPD. In this study, self-reported symptoms of cough and breathlessness were associated with poorer self-reported physical and mental health as assessed by the SF-12. There was a particularly strong association with physical health; men with chronic cough and grade 3 breathlessness or above had a physical health score 11.6 points (95% CI 9.3 to 13.8) lower than men without any respiratory symptoms, indicating the severe impact of impaired respiratory function on quality of life for these men. Men with chronic cough and grade 3 or above breathlessness also had much higher odds of several comorbidities (hypertension, diabetes, angina pectoris, MI and stroke) compared with men without respiratory symptoms, even after adjusting for smoking status. These strong associations are consistent with the findings from other countries that respiratory diseases, in particular COPD, are associated with substantial cardiovascular morbidity.11–13 15 17–20 The association between respiratory symptoms and cardiovascular morbidity has not been investigated previously in Russia. The strong associations with other health conditions found in this study have implications for management of patients with respiratory symptoms; these patients could benefit from identification and treatment of comorbidities. Along with increased odds of cardiovascular comorbidities, men in Izhvesk with respiratory symptoms also had on average a worse cardiovascular risk profile with higher BMI and lower HDL-C levels; however, total cholesterol was also lower and no association was found with triglycerides. There was also strong evidence of higher levels of BNP with respiratory symptoms.

rbidities, men in Izhvesk with respiratory symptoms also had on average a worse cardiovascular risk profile with higher BMI and lower HDL-C levels; however, total cholesterol was also lower and no association was found with triglycerides. There was also strong evidence of higher levels of BNP with respiratory symptoms. Some of this association may be due to misclassification of heart failure as a respiratory problem in this study since breathlessness is a symptom of both; however, this finding is consistent with previous studies which have found both higher levels of BNP among patients with COPD compared with healthy controls38 and that heart failure is a common comorbidity among those with COPD.39 There were only weak associations found in this study between alcohol consumption and respiratory symptoms despite strong associations between hazardous alcohol consumption and blood pressure, cardiovascular disease and BNP.32 33 40–43 Given the very high levels of cardiovascular mortality in Russia, the identification of patients with respiratory problems as a potentially high-risk group who may benefit from primary and secondary prevention could be particularly important. Conversely, respiratory function should also be investigated among those with cardiovascular disease.

ls of cardiovascular mortality in Russia, the identification of patients with respiratory problems as a potentially high-risk group who may benefit from primary and secondary prevention could be particularly important. Conversely, respiratory function should also be investigated among those with cardiovascular disease. In this study the majority of men (87.3%) who reported respiratory symptoms continued to smoke. This was the case even for men experiencing more severe breathlessness (81.6% current smokers). Smoking cessation is a key part of treatment for many respiratory conditions. This is the single most effective treatment for COPD with an important role in slowing the disease progression.14 44 An estimated 62% of patients with COPD in the UK are current smokers estimated from a random sample of patients with COPD in the UK population registered with general practitioner practices in the Clinical Practice Research Datalink.45 Although not directly comparable, the findings from Izhvesk are consistent with findings from this study and studies in clinical populations that a large proportion of those with COPD continue to smoke despite their disease.46 Smoking cessation treatment has been found to be effective for reducing smoking in patients with COPD46–49 and should be offered routinely to those reporting respiratory symptoms.

gs from this study and studies in clinical populations that a large proportion of those with COPD continue to smoke despite their disease.46 Smoking cessation treatment has been found to be effective for reducing smoking in patients with COPD46–49 and should be offered routinely to those reporting respiratory symptoms. There were several limitations to this study. First spirometry was not carried out at the health check. Self-reported symptoms of cough and breathlessness were used here to make some inference about chronic respiratory disease burden overall and its associations with cardiovascular comorbidities in this population. We would like to use these findings to make some inferences particularly about COPD, but this should be done with caution given that the case definition used here is less specific than COPD as determined by spirometry, and that some men with cough and breathlessness in this study may have had other lung conditions such as asthma or tuberculosis. In addition, all comorbidities were also self-reported which is an additional source of measurement error. A further limitation in considering the prevalence estimates is that the study population was a small subset of men aged 25–59 years living in the city of Izhevsk; therefore, the results on prevalence are not generalisable to women, older men or to men living elsewhere in Russia and estimates of prevalence are likely to be imprecise. However, the finding that prevalence of chronic cough and breathlessness is high is consistent with findings from other studies in Russia.8 9 There may also have been some selection bias in the study since only 65% of men interviewed at IFS-2 went on to attend the health check. However, no differences were found between men included in this study and all men taking part in IFS-2.

thlessness is high is consistent with findings from other studies in Russia.8 9 There may also have been some selection bias in the study since only 65% of men interviewed at IFS-2 went on to attend the health check. However, no differences were found between men included in this study and all men taking part in IFS-2. A further limitation was that no data were collected about clinical diagnoses and data collected on treatment were limited. Men were asked about what medications they were taking. However, this question did not refer specifically to inhalers and it is possible many men did not report these. Only 10 men reported the use of bronchodilators in the sample overall, but it is not possible to disentangle whether this is due to very low levels of treatment with bronchodilators or under-reporting of use of these medications. There seems to be a very low level of use of bronchodilators given that some men are likely to have been prescribed bronchodilators for asthma, though it is worth noting that bronchodilators could be expensive for a substantial proportion of the population as Russian patients need to pay the full cost of their medications. It was, however, possible to look at the prevalence of smoking cessation, an important aspect of treatment for respiratory disease. The high proportion of current smokers with respiratory symptoms supports findings from medication data that treatment levels in this population are low. Finally there was a low prevalence of several of the comorbidities included, in particular stroke, diabetes and MI; therefore, results are based on small numbers and a limited number of potential confounders were included. Despite these limitations, the finding of a high level of self-reported chronic cough and breathlessness in this study is important given the very limited data available on the burden of chronic respiratory conditions in the Russian Federation.

small numbers and a limited number of potential confounders were included. Despite these limitations, the finding of a high level of self-reported chronic cough and breathlessness in this study is important given the very limited data available on the burden of chronic respiratory conditions in the Russian Federation. In conclusion, in this sample of working-age men living in Izhevsk, the prevalence of chronic cough and breathlessness was high and associated with worse self-reported physical and mental health and higher levels of cardiovascular comorbidities. Despite the importance of smoking cessation in treatment for respiratory diseases, the majority of men with chronic cough and breathlessness in this population continued to smoke, thus showing that this is not being tackled sufficiently. Chronic respiratory disease is an important public health issue for Russia, but the low levels of research published suggest this is a neglected area. More research is needed, specifically around the levels of diagnosis and treatment in Russia. The association between respiratory symptoms and cardiovascular comorbidity is of particular importance given the extremely high levels of cardiovascular mortality in Russia. The authors thank Vladimir Shkolnikov and Arytom Gil for reviewing a draft of this paper. Contributors: SC and JKQ conceived the idea for this paper. SC, JKQ and DAL interpreted the results of the analyses. DAL designed the Izhevsk Family Studies. MV was involved in data collection. SC analysed the data and wrote the first draft of the paper to which all authors contributed.

The authors thank Vladimir Shkolnikov and Arytom Gil for reviewing a draft of this paper. Contributors: SC and JKQ conceived the idea for this paper. SC, JKQ and DAL interpreted the results of the analyses. DAL designed the Izhevsk Family Studies. MV was involved in data collection. SC analysed the data and wrote the first draft of the paper to which all authors contributed. Funding: The Izhevsk Family Study 2 was funded by the Wellcome Trust (Programme Grant 078557). Competing interests: None declared. Ethics approval: IFS-2 was approved by the Ethics Committees of the London School of Hygiene & Tropical Medicine and Izhevsk Medical Academy. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: Data from the Izhevsk Family Studies are available at http://www.ifsmetadata.info/.

Key messages Engaging in optimal levels of recommended physical activity was associated with significantly better levels of asthma control in adults. Undertaking physical activity during winter, rather than summer, was a stronger predictor of better asthma control. This study provides further evidence of the potential benefits of physical activity as an adjunct treatment in adult patients with asthma. Introduction The evolution of diagnostic and treatment strategies for asthma means that, in theory, all patients with asthma could be well controlled.1 2 However, asthma remains poorly controlled in the majority of patients, for example, 60% of Canadian3 and 75% of European4 patients report having poor control. This inability to decrease asthma symptomology appears to be driving the global burden of asthma, and is associated with increased healthcare utilisation and greater functional impairment.1 5 Identification of factors which might aid or improve patients’ control of their asthma over and above pharmacotherapy may help reduce the burden of asthma.

asthma symptomology appears to be driving the global burden of asthma, and is associated with increased healthcare utilisation and greater functional impairment.1 5 Identification of factors which might aid or improve patients’ control of their asthma over and above pharmacotherapy may help reduce the burden of asthma. Increased levels of leisure time physical activity (LTPA) have been linked to improved health and health outcomes across a variety of chronic diseases,6 including asthma.7 8 For example, increasing levels of LTPA in older women with asthma were associated with a decreased likelihood of reporting an asthma exacerbation over 1 year,8 and an intervention to increase LTPA was associated with a 12-month improvement in asthma control and quality of life.9 Additionally, physically inactive adults with asthma were more likely to be hospitalised overnight and seek multiple physician visits compared to active adults with asthma and active adults without asthma.7

intervention to increase LTPA was associated with a 12-month improvement in asthma control and quality of life.9 Additionally, physically inactive adults with asthma were more likely to be hospitalised overnight and seek multiple physician visits compared to active adults with asthma and active adults without asthma.7 Outside of asthma, it would seem that the impact of LTPA on health outcomes might differ as a function of different exercise intensities and patterns. For example, moderate-intensity aerobic exercise seems to be as beneficial for reducing blood pressure as high-intensity aerobic exercise,10 and there was a marked difference in activity levels as a function of season (lower levels in winter vs summer) in patients with diabetes.11 However, the role of different intensities and seasonal patterns of LTPA has not been previously assessed in adult patients with asthma. The identification of optimal patterns of LTPA for asthma control would be important for the development of future interventions. The purpose of this study was to assess associations between LTPA (total activity and different seasonal patterns and intensities of activity) and levels of asthma control in a Canadian cohort of adult asthmatics.

Outside of asthma, it would seem that the impact of LTPA on health outcomes might differ as a function of different exercise intensities and patterns. For example, moderate-intensity aerobic exercise seems to be as beneficial for reducing blood pressure as high-intensity aerobic exercise,10 and there was a marked difference in activity levels as a function of season (lower levels in winter vs summer) in patients with diabetes.11 However, the role of different intensities and seasonal patterns of LTPA has not been previously assessed in adult patients with asthma. The identification of optimal patterns of LTPA for asthma control would be important for the development of future interventions. The purpose of this study was to assess associations between LTPA (total activity and different seasonal patterns and intensities of activity) and levels of asthma control in a Canadian cohort of adult asthmatics. Materials and methods Study participants and procedure This study was a subanalysis of an existing longitudinal cohort study assessing the impact of psychosocial factors on asthma12 13 and included 643 adult patients with physician-diagnosed asthma who were interviewed at the first follow-up appointment. At baseline, patients with physician-diagnosed asthma14 were recruited from the outpatient asthma clinic of Hôpital du Sacré-Coeur de Montréal between June 2003 and January 2007. Eligible patients were between the ages of 18 and 75 years, and either French-speaking or English-speaking. Patients with occupational asthma, a significant comorbid medical condition (eg, cardiovascular disease, chronic obstructive pulmonary disease) or evidence of severe psychopathology (eg, schizophrenia) were excluded. For the follow-up interview, all patients were recontacted by phone. The follow-up interview included assessments of asthma control (verbal administration of the asthma control questionnaire (ACQ) and LTPA: verbal administration of the 12-month physical activity recall. In addition, patients were then sent a questionnaire pack to complete and return in a self-addressed envelope. Those patients who provided any follow-up data were included in the current analyses.

tration of the asthma control questionnaire (ACQ) and LTPA: verbal administration of the 12-month physical activity recall. In addition, patients were then sent a questionnaire pack to complete and return in a self-addressed envelope. Those patients who provided any follow-up data were included in the current analyses. Ethics statement This project was approved by the Research Ethics Board of Hôpital du Sacré-Cœur de Montréal (#2003-10-198; 2010-95), and written consent was obtained from all participants. Measures Asthma control questionnaire The ACQ15 is a seven-item questionnaire that assesses levels of asthma control in the past week according to standard criteria specified by international guidelines,1 and yields a mean score out of 6. Six of the seven items are self-reported by the patient with one item (forced expiratory volume in 1 s (FEV1)% predicted) calculated from pulmonary function testing. As this questionnaire was administered over the phone, assessment of FEV1% predicted was not possible, so the current analyses reflects only the first six items (which has been shown to be a valid and reliable method of administering this instrument16). Previous research has shown that scores of ≥0.8 indicate poorly controlled asthma17 18 and a change or difference of ≥0.5 in the ACQ has clinical significance.19

ssible, so the current analyses reflects only the first six items (which has been shown to be a valid and reliable method of administering this instrument16). Previous research has shown that scores of ≥0.8 indicate poorly controlled asthma17 18 and a change or difference of ≥0.5 in the ACQ has clinical significance.19 Asthma quality of life questionnaire The asthma quality of life questionnaire (AQLQ) is a self-report questionnaire, consisting of 32 items, that assesses asthma-related quality of life across four life domains, symptoms, activity limitations, environmental stimuli and emotional distress, each of which may be affected by asthma,20 and yields a mean score out of 7. The AQLQ has demonstrated excellent measurement properties and has been validated in Canadian French,21 and was completed as part of the questionnaire pack that was mailed to participants.

tations, environmental stimuli and emotional distress, each of which may be affected by asthma,20 and yields a mean score out of 7. The AQLQ has demonstrated excellent measurement properties and has been validated in Canadian French,21 and was completed as part of the questionnaire pack that was mailed to participants. Adapted 12-month physical activity recall interview The 12-month physical activity recall interview (12M-PAR)22 23 is a self-report questionnaire which assesses the average amount of LTPA that individuals perform over the course of a year. This scale includes activities at three intensity levels: moderate (eg, brisk walking), hard (eg, dancing) and very hard (eg, swimming), and generates estimates of energy expenditure (metabolic equivalent (MET)-h/week). Of note, 10 MET-h/week approximates to current guidelines for healthy activity levels of 30 min of moderate activity on most (ie, 5) days of the week.23 For this study, the questionnaire was administered over the phone and was adapted to ask about the summer (May–October) and winter (November–April) months separately, due to the extreme difference in temperature and weather conditions in Montreal. Other asthma measures All self-reported clinical data, including medical/asthma history, skin prick atopy status24 and medication dosage (including the inhaled corticosteroid (ICS) dose, which was used as a measure of severity1), were verified by medical chart review. In addition, baseline spirometry data were used to assess pulmonary function for descriptive purposes.

ta, including medical/asthma history, skin prick atopy status24 and medication dosage (including the inhaled corticosteroid (ICS) dose, which was used as a measure of severity1), were verified by medical chart review. In addition, baseline spirometry data were used to assess pulmonary function for descriptive purposes. Analyses To ensure accuracy of data entry, all records were double entered and compared using PROC COMPARE (SAS). Missing data were imputed using PROC MI (SAS) and five independent data sets were generated.25 The specific variables which were included in the imputation and the amount of missing data are indicated in table 1. For the analyses, estimations for model coefficients were generated using PROC MIANALYZE according to Rubin's rules.26 Table 1 Demographic and medical/asthma characteristics presented as participating or not in LTPA

Analyses To ensure accuracy of data entry, all records were double entered and compared using PROC COMPARE (SAS). Missing data were imputed using PROC MI (SAS) and five independent data sets were generated.25 The specific variables which were included in the imputation and the amount of missing data are indicated in table 1. For the analyses, estimations for model coefficients were generated using PROC MIANALYZE according to Rubin's rules.26 Table 1 Demographic and medical/asthma characteristics presented as participating or not in LTPA No LTPA n=245 (38%) Some LTPA n=398 (62%) Mean or % 95% CI Mean or % 95% CI n F p Value Demographics Age (years)* 53 51 to 55 54 52 to 55 643 0.12 0.732 Sex (% male) 40 34 to 46 39 34 to 44 643 0.07 0.798 Ethnicity (% white) 93 90 to 96 92 90 to 95 639 0.17 0.682 Cohabitating (% yes) 69 64 to 75 68 64 to 73 643 0.08 0.782 Employed (% yes) 52 45 to 59 47 42 to 53 589 1.17 0.280 BMI (kg/m2)* 28.9 28.0 to 29.7 27.3 26.7 to 28.0 570 7.83 0.005 Current smoker (% yes) 13 9 to 17 6 3 to 9 582 8.22 0.004 Asthma characteristics ACQ (0–6)* 1.1 1.0 to 1.2 0.9 0.8 to 1.0 636 6.07 0.014 Good ACQ (% yes) 47 41 to 54 56 51 to 61 636 4.51 0.034 AQLQ (1–7)* 5.3 5.1 to 5.4 5.4 5.3 to 5.5 582 2.85 0.092 Asthma duration (years)* 25 22 to 27 24 22 to 26 633 0.19 0.664 Atopic (% yes) 60 54 to 66 63 58 to 68 639 0.50 0.481 ICS dose (fluticasone equivalents)* 660 588 to 733 571 513 to 628 507 3.61 0.058 Bronchodilator use (# times in the past week)* 3.9 2.8 to 5.0 3.4 2.5 to 4.2 637 0.58 0.447 Bolding indicates values that are statistically significant.

to 27 24 22 to 26 633 0.19 0.664 Atopic (% yes) 60 54 to 66 63 58 to 68 639 0.50 0.481 ICS dose (fluticasone equivalents)* 660 588 to 733 571 513 to 628 507 3.61 0.058 Bronchodilator use (# times in the past week)* 3.9 2.8 to 5.0 3.4 2.5 to 4.2 637 0.58 0.447 Bolding indicates values that are statistically significant. *These values are reported as means. ACQ, asthma control questionnaire; AQLQ, asthma quality of life questionnaire; BMI, body mass index; ICS, inhaled corticosteroid; LTPA, leisure time physical activity.

to 27 24 22 to 26 633 0.19 0.664 Atopic (% yes) 60 54 to 66 63 58 to 68 639 0.50 0.481 ICS dose (fluticasone equivalents)* 660 588 to 733 571 513 to 628 507 3.61 0.058 Bronchodilator use (# times in the past week)* 3.9 2.8 to 5.0 3.4 2.5 to 4.2 637 0.58 0.447 Bolding indicates values that are statistically significant. *These values are reported as means. ACQ, asthma control questionnaire; AQLQ, asthma quality of life questionnaire; BMI, body mass index; ICS, inhaled corticosteroid; LTPA, leisure time physical activity. To assess the association between LTPA (continuous variable) and the asthma measures (ACQ and AQLQ scores), a series of general linear models were conducted. Logistic regression models were conducted to assess the relationship between a five-group LTPA variable (group 1=no LTPA and groups 2–5=4 quartiles of those doing any LTPA) and good asthma control (ACQ score ≤0.8). Finally, a series of exploratory analyses were conducted to assess the association between winter and summer LTPA, and moderate, hard and very hard LTPA (all as continuous variables) and ACQ score. All models were adjusted for age, sex, season of assessment and ICS dose, which were determined a priori. In order to examine the robustness of our findings, an additional series of analyses were conducted additionally adjusting for comorbid medical characteristics (current smoking and BMI) that have been associated with worse asthma outcomes27 28 and that differed between those who did and those who did not engage in any LTPA. All tests were two-sided and the significance level was set at p<0.05. Data analysis was performed using SAS V.9.3 (SAS Institute, Cary, North Carolina, USA).

rent smoking and BMI) that have been associated with worse asthma outcomes27 28 and that differed between those who did and those who did not engage in any LTPA. All tests were two-sided and the significance level was set at p<0.05. Data analysis was performed using SAS V.9.3 (SAS Institute, Cary, North Carolina, USA). Results Sample characteristics The final sample of 643 patients included 387 (60%) women and they had a mean (SD) age of 53.4 (15.4) years. The mean (SD) duration of asthma for the sample was 20.0 (15.7) years. The mean (SD) (range) amount of LTPA was 4.8 (7.9) (0.0–48.1) MET-h/week, with 245 (38%) doing no LTPA at all. The mean sample (SD) (range) for ACQ and AQLQ scores was 1.0 (1.0) (0.0–5.2) and 5.4 (1.2) (1.0–7.0), respectively. Overall, these data are consistent with a tertiary care asthma population. Demographic and medical/asthma history characteristics Demographic and medical characteristics as a function of having engaged in no LTPA versus any LTPA are presented in table 1. Relative to patients engaging in no LTPA, those doing any LTPA had a lower BMI and were less likely to be current smokers. With regard to asthma, patients engaging in some LTPA had better asthma control, tended to have better quality of life, and were prescribed less ICS.

LTPA versus any LTPA are presented in table 1. Relative to patients engaging in no LTPA, those doing any LTPA had a lower BMI and were less likely to be current smokers. With regard to asthma, patients engaging in some LTPA had better asthma control, tended to have better quality of life, and were prescribed less ICS. Association between LTPA and asthma control As seen in table 2, unadjusted LTPA was negatively associated with ACQ score (β=−0.012). However, when adjusting for covariates, the association was no longer statistically significant. As seen in table 3, for analyses where LTPA was split into five groups, there were significant trends for higher levels of physical activity to be associated with better control. It should be noted that those in the highest quartile of physical activity were nearly 2.5 times more likely (ie, adjusted OR=2.47) to have good control compared with those who did not engage in any LTPA. Table 2 Associations between different forms of leisure time physical activity and asthma control

Association between LTPA and asthma control As seen in table 2, unadjusted LTPA was negatively associated with ACQ score (β=−0.012). However, when adjusting for covariates, the association was no longer statistically significant. As seen in table 3, for analyses where LTPA was split into five groups, there were significant trends for higher levels of physical activity to be associated with better control. It should be noted that those in the highest quartile of physical activity were nearly 2.5 times more likely (ie, adjusted OR=2.47) to have good control compared with those who did not engage in any LTPA. Table 2 Associations between different forms of leisure time physical activity and asthma control Unadjusted Adjusted 1* Adjusted 2† β (% CI) 95 p Value β (95% CI) p Value β (95% CI) p Value Whole sample −0.012 (−0.022 to −0.002) 0.017 −0.009 (−0.019 to 0.001) 0.068 −0.008 (−0.018 to 0.002) 0.115 Season Winter −0.027 (−0.048 to −0.006) 0.011 −0.022 (−0.043 to −0.001) 0.040 −0.019 (−0.041 to 0.002) 0.072 Summer −0.019 (−0.037 to −0.001) 0.039 −0.013 (−0.030 to 0.006) 0.151 −0.011 (−0.029 to 0.006) 0.207 Intensity Moderate −0.011 (−0.037 to 0.015) 0.425 −0.010 (−0.035 to 0.016) 0.459 −0.008 (−0.033 to 0.018) 0.555 Hard −0.052 (−0.137 to 0.033) 0.231 −0.044 (−0.130 to 0.041) 0.309 −0.040 (−0.126 to 0.045) 0.351 Very hard −0.012 (−0.024 to −0.001) 0.031 −0.009 (−0.020 to 0.002) 0.114 −0.008 (−0.020 to 0.03) 0.170 Bolding indicates values that are statistically significant. *Adjusted for age, sex, inhaled corticosteroid dose and season recruited in.

Unadjusted Adjusted 1* Adjusted 2† β (% CI) 95 p Value β (95% CI) p Value β (95% CI) p Value Whole sample −0.012 (−0.022 to −0.002) 0.017 −0.009 (−0.019 to 0.001) 0.068 −0.008 (−0.018 to 0.002) 0.115 Season Winter −0.027 (−0.048 to −0.006) 0.011 −0.022 (−0.043 to −0.001) 0.040 −0.019 (−0.041 to 0.002) 0.072 Summer −0.019 (−0.037 to −0.001) 0.039 −0.013 (−0.030 to 0.006) 0.151 −0.011 (−0.029 to 0.006) 0.207 Intensity Moderate −0.011 (−0.037 to 0.015) 0.425 −0.010 (−0.035 to 0.016) 0.459 −0.008 (−0.033 to 0.018) 0.555 Hard −0.052 (−0.137 to 0.033) 0.231 −0.044 (−0.130 to 0.041) 0.309 −0.040 (−0.126 to 0.045) 0.351 Very hard −0.012 (−0.024 to −0.001) 0.031 −0.009 (−0.020 to 0.002) 0.114 −0.008 (−0.020 to 0.03) 0.170 Bolding indicates values that are statistically significant. *Adjusted for age, sex, inhaled corticosteroid dose and season recruited in. †Adjusted for age, sex, inhaled corticosteroid dose, season recruited in, body mass index and current smoking or not. Table 3 Exercise, asthma control and asthma quality of life levels by quartiles of LTPA No exercise Quartile 1 Quartile 2 Quartile 3 Quartile 4 Trend n 245 99 100 99 100 LTPA, Mean±SD (range) (MET-h/week) 0±0 1.0±0.4 (0.1–1.7) 2.8±0.7 (1.8–4.2) 7.1±1.9 (4.3–10.9) 20.1±8.9 (11.1–48.1) Total ACQ score† (0–6) 1.10 0.92 1.03 0.92 0.84 p=0.037 Predicted good control* Ref 0.87 (0.38–2.00) 0.98 (0.43–2.22) 0.84 (0.37–1.91) 2.47 (1.06–5.73) Total AQLQ score† (1–7) 5.27 5.29 5.23 5.49 5.49 p=0.058 Bolding indicates values that are statistically significant. All analyses adjusted for age, sex, ICS dose and season recruited in.

No exercise Quartile 1 Quartile 2 Quartile 3 Quartile 4 Trend n 245 99 100 99 100 LTPA, Mean±SD (range) (MET-h/week) 0±0 1.0±0.4 (0.1–1.7) 2.8±0.7 (1.8–4.2) 7.1±1.9 (4.3–10.9) 20.1±8.9 (11.1–48.1) Total ACQ score† (0–6) 1.10 0.92 1.03 0.92 0.84 p=0.037 Predicted good control* Ref 0.87 (0.38–2.00) 0.98 (0.43–2.22) 0.84 (0.37–1.91) 2.47 (1.06–5.73) Total AQLQ score† (1–7) 5.27 5.29 5.23 5.49 5.49 p=0.058 Bolding indicates values that are statistically significant. All analyses adjusted for age, sex, ICS dose and season recruited in. *Good control=ACQ score ≤0.8. †As the analyses were conducted using multiple imputation, it is not appropriate to include SDs with the mean scores. ACQ, asthma control questionnaire; AQLQ, asthma quality of life questionnaire; ICS, inhaled corticosteroid; LTPA, leisure time physical activity; MET, metabolic equivalent.

All analyses adjusted for age, sex, ICS dose and season recruited in. *Good control=ACQ score ≤0.8. †As the analyses were conducted using multiple imputation, it is not appropriate to include SDs with the mean scores. ACQ, asthma control questionnaire; AQLQ, asthma quality of life questionnaire; ICS, inhaled corticosteroid; LTPA, leisure time physical activity; MET, metabolic equivalent. Exploratory analyses between LTPA patterns and asthma control As expected, participants generally engaged in more LTPA during the summer months compared to the winter months. We observed a significant negative relationship between winter and summer LTPA and ACQ scores (table 2), with winter exercise showing a stronger relationship to ACQ than summer exercise. However, in fully adjusted models, neither winter nor summer LTPA remained statistically significantly associated with control. When intensity of exercise was assessed separately, there was a statistically significant negative relationship between very hard LTPA and ACQ, but there were no relationships between moderate and hard LTPA and control (table 2). As with the seasonal variations, adjustment for covariates resulted in a reduction in the size of the effect for very hard LTPA, with a loss of statistical significance. An additional analysis was conducted to see if there was a difference in ACQ scores according to the season of recruitment, for which there was none (β=0.055, 95% CI −0.105 to 0.215, p=0.50), indicating that those recruited during summer (mean=0.97) had similar ACQ scores to those recruited in winter (mean=1.02).

Exploratory analyses between LTPA patterns and asthma control As expected, participants generally engaged in more LTPA during the summer months compared to the winter months. We observed a significant negative relationship between winter and summer LTPA and ACQ scores (table 2), with winter exercise showing a stronger relationship to ACQ than summer exercise. However, in fully adjusted models, neither winter nor summer LTPA remained statistically significantly associated with control. When intensity of exercise was assessed separately, there was a statistically significant negative relationship between very hard LTPA and ACQ, but there were no relationships between moderate and hard LTPA and control (table 2). As with the seasonal variations, adjustment for covariates resulted in a reduction in the size of the effect for very hard LTPA, with a loss of statistical significance. An additional analysis was conducted to see if there was a difference in ACQ scores according to the season of recruitment, for which there was none (β=0.055, 95% CI −0.105 to 0.215, p=0.50), indicating that those recruited during summer (mean=0.97) had similar ACQ scores to those recruited in winter (mean=1.02). Association between LTPA and asthma quality of life LTPA was not significantly associated with AQLQ scores in unadjusted (β=0.011, 95% CI −0.001 to 0.024, p=0.08) or adjusted (β=0.006, 95% CI −0.006 to 0.019, p=0.30) analyses. However, it should be noted that the nature of the relationship between the two variables was in the hypothesised direction, that is, increased LTPA was associated with higher scores on the AQLQ. As seen in table 2, when LTPA was split into five groups, there was no significant association between LTPA and AQLQ scores.

However, it should be noted that the nature of the relationship between the two variables was in the hypothesised direction, that is, increased LTPA was associated with higher scores on the AQLQ. As seen in table 2, when LTPA was split into five groups, there was no significant association between LTPA and AQLQ scores. Discussion This study assessed the cross-sectional association between LTPA and asthma control in a cohort of adult patients with asthma. Results showed that although patients engaging in higher levels of LTPA appeared to have better asthma control, this effect was lost once important covariates were adjusted for. In addition, those participants who engaged in the most amount of exercise (about 30 min of moderate exercise per day, most days of the week) were almost 2.5 times more likely to be well controlled compared with those who did not engage in any exercise, a result which held even with adjustment for covariates. Further interrogation of the data revealed that the amount of activity engaged in during the winter months was a stronger predictor of asthma control than summer activity.

s more likely to be well controlled compared with those who did not engage in any exercise, a result which held even with adjustment for covariates. Further interrogation of the data revealed that the amount of activity engaged in during the winter months was a stronger predictor of asthma control than summer activity. In general, these findings are in line with the previous literature that has found decreases in asthma exacerbations and healthcare use in patients who engage in LTPA.7 8 However, our study adds to the extant literature by assessing seasonal variations in activity and its association with asthma control. Asthmatics who exercised tended to perform more activity in the summer months than in the winter months, which is consistent with other studies.29–31 Although this result was unsurprising, the finding that winter activity was more strongly associated with control than summer activity was unexpected. We know of no studies that have found a differential relationship between physical activity and disease morbidity as a function of patterns of seasonal engagement. How engaging in physical activity over the winter may be associated with better control is not known. However, it is possible that asthmatics who continued to exercise during the winter may have accumulated more activity over the year, and that continuous and cumulative exercise may be a key element in the positive role of physical activity on asthma. Given the cross-sectional nature of the study, we cannot discount the possibility of reverse causality and that those with better control were able to exercise more during the winter. For example, it has been shown that for any specific intensity of exercise, there is a great decrease in FEV132 and overall worse bronchial hyper-responsiveness33 in winter compared with summer. These, coupled with a greater probability of having exercise-induced bronchoconstriction during the colder, dryer months of winter,34 mean that there is the possibility of patients having generally poorer control in winter, which could then translate into less activity.

-responsiveness33 in winter compared with summer. These, coupled with a greater probability of having exercise-induced bronchoconstriction during the colder, dryer months of winter,34 mean that there is the possibility of patients having generally poorer control in winter, which could then translate into less activity. Further investigation into the factors which might account for the differential relationships between seasons is warranted. For example, patients with asthma who significantly reduced their exercise levels during the winter may have negative beliefs about exercise-induced symptoms (eg, increased risk of bronchoconstriction while exercising in cold temperatures), which could be driven by worsening bronchial obstruction and hyper-responsiveness (as detailed above), and/or an increase in perceived barriers to exercise (eg, concerns of the status of the roads, cost of membership to a fitness centre, etc). Understanding such facets may help to improve aspects of asthma treatment, for example, altered medication dosing in winter months, and self-management.

tion could be residual confounding by severity of maternal asthma; perhaps mothers with more severe asthma were less likely to breastfeed, but more likely to have children with low lung volumes. Although the analysis adjusted for presence of maternal asthma, we did not have information on asthma severity in the mother. Strengths and limitations A strength of this study is the large range of lung measurements, including spirometry and plethysmography and in vivo measurements reflecting alveolar dimension.13 14 Furthermore, the study was able to consider important potential confounding including early onset of wheeze, through which we attempted to control for possible reverse causation.12 The main limitation is the small sample size which was due to the complexity and cost of the techniques involved. While the study had been adequately powered to assess potential age-related changes of alveolar dimensions 13 and differences between term and preterm children,14 it was probably too small to detect minor differences resulting from duration of breastfeeding in such a distal outcome as lung function at school age. This was further complicated by the attempt to determine if the association is influenced by maternal asthma, performing an analysis with an interaction term; there were only seven participants of mothers with asthma who had been breastfed over 3 months (table 1).

eness (as detailed above), and/or an increase in perceived barriers to exercise (eg, concerns of the status of the roads, cost of membership to a fitness centre, etc). Understanding such facets may help to improve aspects of asthma treatment, for example, altered medication dosing in winter months, and self-management. When interpreting the potential seasonal variations in the LTPA-asthma control relationship, one must be cognisant of the impact of weather changes on asthma in general. A number of studies have identified that asthma symptoms seem to be worse and asthma-related hospitalisations more frequent during winter relative to summer.35–38 Consistent with our findings, it may be that due to the higher symptom burden, exercising during winter is a more effective adjunct therapy to medications than during summer. Furthermore, there is some evidence that the severity of asthma-related hospitalisations may be higher during the summer,35 and it would seem that more asthma-related deaths occur during the summer compared with the winter.39 As such, in the context of our study, any benefits of engaging in LTPA on asthma during summer may not be related to asthma control but may be associated with a reduction in the severity of asthma exacerbations, which were not assessed in this study. While these studies may help explain the current findings, it should be noted that such seasonal variations may be less pronounced as patients get older, with minimal difference seen in seasonal variations in hospitalisations for middle-aged patients.38 40 Given that our cohort was primarily middle-aged, this could account for the lack of seasonal difference in ACQ score for our study.

be noted that such seasonal variations may be less pronounced as patients get older, with minimal difference seen in seasonal variations in hospitalisations for middle-aged patients.38 40 Given that our cohort was primarily middle-aged, this could account for the lack of seasonal difference in ACQ score for our study. This study has some limitations that warrant discussion. First, we used a self-reported physical activity questionnaire to measure LTPA. However, such questionnaires have been shown to be reasonably consistent with other measures of activity.29 30 Another limitation is that assessments of asthma control were conducted at only one time point; as such, season variations in LTPA and asthma control may not have been temporally assessed. However, we did not find seasonal differences in ACQ scores at recruitment and the season of recruitment was adjusted for in analyses. It should also be noted that the study was conducted in Montreal, which has a significant temperature gradient across the seasons, where winter can average −20°C and summer can be above 20°C. As such, it may not be possible to generalise these results to more temperate climates. Finally, this study is cross-sectional, so it is impossible for us to be able to define the exact direction of the relationship between asthma control and LTPA. As mentioned above, it is possible that our results reflect the possibility that those with better asthma control engage in more physical activity. However, recent studies have shown a prospective relationship between physical activity and a lower risk for asthma exacerbations, decreased healthcare usage, and improvements in quality of life.7–9

possible that our results reflect the possibility that those with better asthma control engage in more physical activity. However, recent studies have shown a prospective relationship between physical activity and a lower risk for asthma exacerbations, decreased healthcare usage, and improvements in quality of life.7–9 Despite some limitations, this study also has a number of important strengths. It was conducted with a well characterised sample of objectively diagnosed asthmatics. The sample size was relatively large and included both men and women. Conclusions In summary, this study found evidence for an association between LTPA and asthma control in a large tertiary-care sample of Canadian adults with asthma, with recommended levels of LTPA being associated with better levels of asthma control. Furthermore, it would seem that engaging in LTPA during winter, rather than summer, was a stronger driver of the relationship. As this study was not designed to examine the mechanisms linking LTPA to asthma control, future studies should examine the pathways by which patients with asthma may benefit from exercising regularly across all seasons, but especially during the winter months. Finally, longitudinal and intervention studies are needed to determine the minimal dose of LTPA and the most cost-efficient strategies to improve asthma control in this kind of population.

ys by which patients with asthma may benefit from exercising regularly across all seasons, but especially during the winter months. Finally, longitudinal and intervention studies are needed to determine the minimal dose of LTPA and the most cost-efficient strategies to improve asthma control in this kind of population. Contributors: SLB conceived of the study, wrote the initial draft of the manuscript, conducted all statistical data analyses, and obtained funding for the study. KLL participated in the design and coordination of the study, obtained funding for the study, and cowrote the manuscript. CL, GM, GN and VP provided critical feedback on the drafts of the manuscript and helped with the interpretation of the data. All authors read and approved the final manuscript. Funding: Funding support for this study was provided by grants from the Social Sciences and Humanities Research Council (SSHRC), Fonds de recherche du Québec—santé (FRQS), Canadian Institutes of Health Research (CIHR: MOP93807), and the Michel Auger Foundation of Hôpital du Sacré-Coeur de Montréal. Additional investigator salary award support was obtained from FRQS and CIHR (SLB and KLL). Competing interests: None declared. Ethics approval: This project was approved by the Research Ethics Board of Hôpital du Sacré-Cœur de Montréal (#2003-10-198; 2010-95). Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: The data from this study are not openly available. However, those interested in getting access should contact the corresponding author.

Key messages This study is the first to report data on the possible association between breastfeeding and alveolar size. While the study does not provide strog evidence of an association, it suggests that breastfeeding may be associated with increased lung volumes and alveolar size, particularly in children of mothers with asthma. Introduction Breastfeeding has many beneficial effects for children and mothers.1 The impact on respiratory health is, however, less clear. Studies have shown that breastfed children have fewer and less severe respiratory infections than their non-breastfed peers.2–5 The influence of breastfeeding on lung development has been investigated by few researchers, with mixed results.6–11 Most studies reported larger normalised lung volumes in breastfed children, usually higher forced vital capacity (FVC) or forced expiratory volume at 1 s (FEV1).6 8–11 In a recent study, we found that children breastfed over 3 months had higher forced mid-expiratory flows (FEF50) at school-age.12 There was also evidence of an effect modification by maternal asthma: in children whose mothers had asthma, the increase in FEF50 was greater and accompanied by increased FEV1 and FVC.12

–11 In a recent study, we found that children breastfed over 3 months had higher forced mid-expiratory flows (FEF50) at school-age.12 There was also evidence of an effect modification by maternal asthma: in children whose mothers had asthma, the increase in FEF50 was greater and accompanied by increased FEV1 and FVC.12 From these studies, it remains unclear if the reported increase of volumes with breastfeeding involves different lung compartments proportionally. For instance, increased lung volumes might be explained by more alveoli or by larger alveoli due to structural differences or hyperinflation. Proportional increases in all lung volumes would suggest genuinely larger lungs, while increases limited to residual volume (RV), functional residual capacity (FRC) and total lung capacity (TLC) would suggest the presence of hyperinflation. We used the novel technique of 3He MR, which measures freedom of gas diffusion in alveolar-acinar compartments and provides a surrogate measure of alveolar size, to determine whether the reported increase in lung volume is associated with proportionately increased alveolar size (suggesting lung growth) or disproportionately increased alveolar size (suggesting hyperinflation). Using this technique, we recently reported evidence for continued alveolarisation throughout childhood and adolescence13 and evidence for catch-up alveolarisation in ex-preterm children.14 We have now reanalysed this data set to investigate the association between breastfeeding and plethysmographic lung volumes and alveolar dimensions. No previous studies have investigated the effect of breastfeeding on these measurements.

adolescence13 and evidence for catch-up alveolarisation in ex-preterm children.14 We have now reanalysed this data set to investigate the association between breastfeeding and plethysmographic lung volumes and alveolar dimensions. No previous studies have investigated the effect of breastfeeding on these measurements. The aim of our study were thus to determine (1) whether plethysmographic lung volumes and alveolar size are associated with duration of breastfeeding, and (2) whether the potential association between breastfeeding and lung volumes differs depending on mother's history of asthma. Materials and methods The study was approved by the Ethics Committee of the Leicestershire Health Authority and written consent had been obtained from all participants and their parents. Study population For this study, we analysed a data set of 111 children and young adults (aged 11–21 years) from Leicestershire, UK, who had participated in a complex study on alveolar dimensions, measured by 3He MR, the ‘Helium study’.13 14 The original study was not designed specifically for analysing the association with breastfeeding. We included for analysis participants who had been born at 32 weeks of gestation or later, had a birth weight of over 1500 g, who had no congenital anomalies or chronic lung diseases, such as cystic fibrosis or bronchopulmonary dysplasia, and in whom prospectively collected information on breastfeeding was available.

. We included for analysis participants who had been born at 32 weeks of gestation or later, had a birth weight of over 1500 g, who had no congenital anomalies or chronic lung diseases, such as cystic fibrosis or bronchopulmonary dysplasia, and in whom prospectively collected information on breastfeeding was available. The participants had been recruited from the Leicester Respiratory Cohorts, consisting of two population-based cohorts based on random samples of children born in Leicestershire between 1987 and 1989 (the 1990 cohort), and between 1993 and 1997 (the 1998 cohort). The Leicestershire Health Authority Child Health Database, later called the Leicester Specialist Community Child Health Services Database, had been used to draw the sample and obtain baseline routine data.15 Parents of participating children received the first set of respiratory questionnaires in 1990 and 1998, respectively, when the children were 0–5 years old. The participants were then followed up repeatedly with questionnaires, every 2–4 years. In total, 410 participants from the Leicester Respiratory Cohorts had been invited to the Helium study. A total of 125 (30.5%) participated, and we obtained 3HeMR measurements in 114 participants. Of the 114 with 3He MR measurements, 111 were eligible for the present study, based on criteria described above and availability of breastfeeding data.

n such a distal outcome as lung function at school age. This was further complicated by the attempt to determine if the association is influenced by maternal asthma, performing an analysis with an interaction term; there were only seven participants of mothers with asthma who had been breastfed over 3 months (table 1). Another limitation is a possible recall bias in reporting duration of breastfeeding. The participants reported duration of breastfeeding when the children were 1–4 years old. However, the question had shown an excellent short-term repeatability in our cohort, with a Cohen's κ of 0.96 21 and there is independent evidence suggesting that long-term recall of breastfeeding is excellent.22

rom the Leicester Respiratory Cohorts had been invited to the Helium study. A total of 125 (30.5%) participated, and we obtained 3HeMR measurements in 114 participants. Of the 114 with 3He MR measurements, 111 were eligible for the present study, based on criteria described above and availability of breastfeeding data. Health records and questionnaire data We extracted perinatal data for all participants from Leicestershire Health Authority Child Health Database, including mother's ethnicity, birth weight and gestational age. We had assessed respiratory symptoms, and individual and family-related exposures initially in 1990 and 1998, and in several follow-up questionnaires in 2001, 2003 and 2006. At the time of the helium study, all participants and their parents were asked to complete a new questionnaire, providing information on health history, environmental exposures and demographic data. Breastfeeding was defined as total duration of breastfeeding, regardless of exclusivity. We recorded duration of breastfeeding in the first questionnaire (applied at age 0–5 years). The question asked if the child had been breastfed and, if yes, for how long, with the following response options: less than a month; 1–3 months; 4–6 months and more than 6 months. Owing to the small sample size, we combined the responses into one of three categories: (1) no breastfeeding, (2) breastfeeding for less than or equal to 3 months and (3) breastfeeding for more than 3 months.

ow long, with the following response options: less than a month; 1–3 months; 4–6 months and more than 6 months. Owing to the small sample size, we combined the responses into one of three categories: (1) no breastfeeding, (2) breastfeeding for less than or equal to 3 months and (3) breastfeeding for more than 3 months. We considered as potential confounders those factors that might be a common cause of both exposure (duration of breastfeeding) and outcomes (lung and alveolar volumes) or be on the pathway of a common cause. For our analysis these included: Townsend score—an area-based deprivation score16—birth weight, preterm status, smoking during pregnancy, maternal asthma, maternal ethnicity and early-onset wheeze, described in more detail below. Participants were considered preterm if their gestational age, extracted from the birth registry, was under 37 weeks. Information on mother's history of smoking during pregnancy (yes/no) was collected in the initial questionnaire and updated at the time of measurements. Maternal history of asthma (yes/no) was self-reported based on the question: “Has the child's mother ever suffered from any of the following conditions :(…) wheezing; asthma?” We coded maternal asthma as ‘yes’ if the mother reported wheezing or asthma. Maternal ethnicity was extracted from birth records and categorised into White and south-Asian; the south-Asian group, which includes people of Indian/Sri Lanka, Pakistani or Bangladesh origin, is the largest ethnic minority in Leicestershire and one of the sampling strata of the Leicester Respiratory Cohorts. We defined early-onset wheeze as any wheezing with onset during the first year of life. Early-onset wheeze was included in the analysis in an attempt to control for a possible reverse causation, when early-onset wheeze, a possible precursor of future lung problems resulting in decreased lung function, might influence the duration of breastfeeding. In order to avoid over-adjustment, we did not include in the analysis variables that theoretically might be on the causal path, such as participant's history of asthma.17

early-onset wheeze, a possible precursor of future lung problems resulting in decreased lung function, might influence the duration of breastfeeding. In order to avoid over-adjustment, we did not include in the analysis variables that theoretically might be on the causal path, such as participant's history of asthma.17 Physiological measurements At the time of the study, height and weight were measured, and lung mechanics were assessed by performing spirometry and full-body plethysmography (Jaeger Masterscreen Body, Wuerzburg, Germany). The highest values of forced expired volume in 1 s (FEV1) and forced vital capacity (FVC) were reported, together with forced expiratory flow at 50% vital capacity (FEF50) from the manoeuvre with the largest sum of FEV1 and FVC. From plethysmography, the mean value of functional residual capacity (FRC), the largest vital capacity (VC), and the residual volume (RV) and total lung capacity (TLC) associated with the largest VC were reported. The measurements conformed to the American Thoracic Society/European Respiratory Society task force specifications,18 19 and were reviewed for repeatability and quality control by a specialist respiratory physiologist (CSB).

esidual volume (RV) and total lung capacity (TLC) associated with the largest VC were reported. The measurements conformed to the American Thoracic Society/European Respiratory Society task force specifications,18 19 and were reviewed for repeatability and quality control by a specialist respiratory physiologist (CSB). To assess the average dimensions of the alveoli, we used the hyperpolarised 3Helium MR technique (3HeMR), which measures the restricted diffusion of hyperpolarised 3He in a constrained space such as the alveoli. The measurement provides a surrogate of alveolar dimension, the apparent diffusion coefficient (ADC). The participant is required to lie supine and at the end of a normal expiration, she/he inhales a bolus of 600 mL of a mixture of hyperpolarised 3He and 4He, and holds the breath for approximately 10 s. 4He is an abundant, naturally occurring non-radioactive helium isotope, while 3He is a rare non-radioactive isotope of helium that can be polarised.20 The MR measurements were performed with a 0.15 T permanent magnet system (Intermagnetics General Corporation, New York, New York, USA) using a modified rapid acquisition with refocused echoes (RARE) sequence (64 echoes, echo time=14 ms, acquisition time=896 ms, fixed gradient strength (b=0.3 s.cm−2), and slice select and phase gradients turned off).13 The technique is non-invasive and radiation-free, but expensive and used in only few centres worldwide. At least three technically satisfactory ADC values were obtained for each participant and the mean value was reported. We corrected ADC values for differences in 3He concentration and bolus size in relation to FRC.13 14 The measurements were highly repeatable, with a within-subject coefficient of variation of 3.1%.13

. At least three technically satisfactory ADC values were obtained for each participant and the mean value was reported. We corrected ADC values for differences in 3He concentration and bolus size in relation to FRC.13 14 The measurements were highly repeatable, with a within-subject coefficient of variation of 3.1%.13 Statistical analysis The association between duration of breastfeeding and lung volumes and alveolar size was analysed using multivariable linear regression. The outcomes of interest were spirometric measures (FVC, FEV1 and FEF50), plethysmographic measures (FRC, VC, RV and TLC), and alveolar size (ADC). To assess relative (percent) differences between levels of breastfeeding, we first log-transformed the outcomes, performed the regressions and then exponentiated estimated coefficients. Thus, the (exponentiated) regression constant and slope represent geometric means and relative differences, respectively. The analysis was performed in three steps: first we ran a model controlling only for variables needed to standardise lung function measurements, namely age, sex and height (basic model). Second, we ran a model including potential confounders, namely preterm status, birth weight, Townsend score, smoking during pregnancy, maternal asthma, maternal ethnicity and early-onset wheeze (adjusted model); third, we ran a model in which we added an interaction term between breastfeeding and maternal asthma to test for effect modification by maternal asthma (effect-modification model). The analyses of ADC were adjusted, additionally, for the natural logarithm of FRC, because of an expected increase in ADC with FRC.13 Thus, regression coefficients for the ADC models reflect relative differences in ADC over and above those accounted for by changes in FRC due to lung growth. These relative differences in ADC due to the risk factors in the regression model are also expressed in terms of estimated differences in mean alveolar volume using the relationship ADC ratio=(volume ratio)0.415 derived in Narayanan et al.13

n ADC over and above those accounted for by changes in FRC due to lung growth. These relative differences in ADC due to the risk factors in the regression model are also expressed in terms of estimated differences in mean alveolar volume using the relationship ADC ratio=(volume ratio)0.415 derived in Narayanan et al.13 Based on the third model with an interaction term between breastfeeding and maternal asthma, we calculated the means of the outcome measures for children of mothers with and without asthma. All analyses were performed using Stata V.12 (Stata Corporation, Austin, Texas, USA). Results Sample characteristics Detailed data on sample characteristics are presented (table 1). Among participants 24 (22%) had never been breastfed, 37 (33%) had been breastfed 3 months or less, and 50 (45%) had been breastfed over 3 months; the mothers of 22 (20%) reported having asthma. Girls, children of mothers without asthma and of mothers who did not smoke during pregnancy tended to be breastfed longer. We did not find evidence for differences in spirometry or plethysmographic measurements and ADC between categories of breastfeeding. Table 1 Characteristics of study population, by duration of breastfeeding (N=151)

Results Sample characteristics Detailed data on sample characteristics are presented (table 1). Among participants 24 (22%) had never been breastfed, 37 (33%) had been breastfed 3 months or less, and 50 (45%) had been breastfed over 3 months; the mothers of 22 (20%) reported having asthma. Girls, children of mothers without asthma and of mothers who did not smoke during pregnancy tended to be breastfed longer. We did not find evidence for differences in spirometry or plethysmographic measurements and ADC between categories of breastfeeding. Table 1 Characteristics of study population, by duration of breastfeeding (N=151) Breastfeeding categories Total (N=111) Never (N=24) ≤3 months (N=37) >3 months (N=50) p Value* Age† (years) 14.1 (2.5) 14.7 (2.7) 14.3 (2.9) 13.5 (2.0) 0.111 Height† (cm) 160.8 (10.9) 164.0 (13.5) 158.7 (10.8) 160.9 (9.4) 0.190 Weight† (kg) 53.8 (13.1) 58.2 (17.2) 52.8 (13.1) 52.5 (10.5) 0.189 Gestational age† (weeks) 38.5 (2.3) 39.0 (1.9) 38.4 (2.5) 38.9 (2.0) 0.371 Birth weight (g) 3235.8 (599.2) 3111.2 (622.2) 3212.7 (605.1) 3313.8 (583.9) 0.386 Sex‡ Male 51 (46.0) 23 (63.9) 14 (32.6) 35 (48.1) 0.021 Female 60 (54.0) 13 (36.1) 29 (67.4) 37 (51.4) Ethnicity‡ South-Asian 24 (21.6) 5 (20.8) 9 (37.5) 10 (41.7) 0.884 White 87 (78.4) 19 (21.8) 28 (32.2) 40 (45.9) Smoking during pregnancy‡ No 100 (90.1) 18 (18.0) 36 (36.0) 46 (46.0) 0.014 Yes 11 (9.9) 6 (54.5) 1 (9.1) 4 (36.4) Early-onset wheeze‡ No 88 (79.3) 19 (21.6) 31 (35.2) 38 (43.2) 0.676 Yes 23 (20.7) 5 (21.7) 6 (26.1) 12 (52.2) Maternal asthma‡ No 88 (79.3) 15 (17.05) 31 (35.23) 42 (47.73) 0.052 Yes 22 (19.8) 9 (40.91) 6 (27.27) 7 (31.82) Missing 1 (0.9) 0 (0.0) 0 (0.0) 1 (100.0) FVC§ (L) 3.5 (0.7) 3.76 (0.92) 3.41 (0.74) 3.53 (0.63) 0.183 FEV1§ (L) 3.0 (0.6) 3.23 (0.75) 2.94 (0.62) 3.04 (0.52) 0.197 FEF50§ (L/s) 3.6 (0.6) 3.78 (0.74) 3.50 (0.62) 3.61 (0.53) 0.220 FRC§ (L) 2.2 (0.5) 2.37 (0.59) 2.17 (0.51) 2.19 (0.40) 0.227 RV§ (L) 1.2 (0.2) 1.31 (0.29) 1.22 (0.25) 1.22 (0.19) 0.233 TLC§ (L) 4.7 (0.9) 4.94 (1.15) 4.50 (0.94) 4.64 (0.78) 0.196 ADC¶ (cm2/s) 0.096 (0.012) 0.095 (0.012) 0.096 (0.012) 0.095 (0.013) 0.444 *The p value is based on a χ2 test.

50 (0.62) 3.61 (0.53) 0.220 FRC§ (L) 2.2 (0.5) 2.37 (0.59) 2.17 (0.51) 2.19 (0.40) 0.227 RV§ (L) 1.2 (0.2) 1.31 (0.29) 1.22 (0.25) 1.22 (0.19) 0.233 TLC§ (L) 4.7 (0.9) 4.94 (1.15) 4.50 (0.94) 4.64 (0.78) 0.196 ADC¶ (cm2/s) 0.096 (0.012) 0.095 (0.012) 0.096 (0.012) 0.095 (0.013) 0.444 *The p value is based on a χ2 test. †Mean (SD). ‡N (%); the percentages represent breastfeeding frequencies within levels of confounder (row percentages). §The means of the outcome variables are predicted values adjusted for age, sex, height, weight using linear regression. ¶The ADC analysis was performed using the natural logarithm, adjusting for age, sex, height and ln FRC; the estimates were back-transformed from the logarithmic scale to the original scale, therefore the estimated means are geometric means. ADC, apparent diffusion coefficient; ETS, exposure to tobacco smoke; FEF50, forced mid-expiratory flow; FEV1, forced expiratory volume at 1 s; FRC, functional residual capacity; FVC, forced vital capacity; RV, residual volume; TLC, total lung capacity; VC, vital capacity.

¶The ADC analysis was performed using the natural logarithm, adjusting for age, sex, height and ln FRC; the estimates were back-transformed from the logarithmic scale to the original scale, therefore the estimated means are geometric means. ADC, apparent diffusion coefficient; ETS, exposure to tobacco smoke; FEF50, forced mid-expiratory flow; FEV1, forced expiratory volume at 1 s; FRC, functional residual capacity; FVC, forced vital capacity; RV, residual volume; TLC, total lung capacity; VC, vital capacity. Breastfeeding and lung volumes Basic and adjusted model Table 2 presents the results of the basic and adjusted models. When compared with participants who had not been breastfed, the TLC of participants breastfed >3 months was on average larger by over 6% in the basic and adjusted model. We found no strong evidence for an association between duration of breastfeeding and any of the other lung function outcomes, regardless of adjustment for potential confounders. Although participants breastfed for >3 months tended to have larger lung volumes and flows than non-breastfed participants, 95% CIs (95% CI) for differences included 0 and the p values were all >0.1. Table 2 Association between breastfeeding and lung function measurements in all participants, basic model and fully adjusted model*

Breastfeeding and lung volumes Basic and adjusted model Table 2 presents the results of the basic and adjusted models. When compared with participants who had not been breastfed, the TLC of participants breastfed >3 months was on average larger by over 6% in the basic and adjusted model. We found no strong evidence for an association between duration of breastfeeding and any of the other lung function outcomes, regardless of adjustment for potential confounders. Although participants breastfed for >3 months tended to have larger lung volumes and flows than non-breastfed participants, 95% CIs (95% CI) for differences included 0 and the p values were all >0.1. Table 2 Association between breastfeeding and lung function measurements in all participants, basic model and fully adjusted model* Lung function (unit) Basic model† Adjusted model‡ estimates (CI) p Value estimates (CI) p Value FVC§ (L) No BF (mean) 3.31 (3.12 to 3.49) 3.39 (3.18 to 3.61) BF ≤3 months (% difference) 3.07 (−4.30 to 10.44) 0.407 4.20 (−3.03 to 11.43) 0.245 BF >3 months (% difference) 4.44 (−2.65 to 11.52) 0.210 4.20 (−2.40 to 10.80) 0.203 FEV1§ (L) No BF (mean) 2.82 (2.66 to 2.98) 2.93 (2.73 to 3.13) BF ≤3 months (% difference) 4.33 (−3.43 to 12.08) 0.264 5.42 (−2.34 to 13.17) 0.160 BF >3 months (% difference) 5.66 (−1.80 to 13.12) 0.126 5.41 (−1.67 to 12.48) 0.124 FEF50§ (L/s) No BF (mean) 3.20 (2.86 to 3.55) 3.46 (2.98 to 3.95) BF ≤3 months (% difference) 9.16 (−6.04 to 24.35) 0.217 9.71 (−7.14 to 26.56) 0.237 BF >3 months (% difference) 9.92 (−4.54 to 24.39) 0.159 9.67 (−5.62 to 24.95) 0.194 FRC§ (L) No BF (mean) 1.99 (1.87 to 2.12) 1.96 (1.79 to 2.13) BF ≤3 months (% difference) 3.03 (−5.57 to 11.64) 0.483 4.72 (−5.12 to 14.55) 0.336 BF >3 months (% difference) 6.58 (−1.89 to 15.06) 0.116 5.76 (−3.36 to 14.87) 0.203 RV§ (L) No BF (mean) 1.11 (1.01 to 1.21) 1.08 (0.96 to 1.20) BF ≤3 months (% difference) −1.27 (−12.62 to 10.09) 0.828 −0.59 (−12.75 to 11.57) 0.924 BF >3 months (% difference) 8.32 (−3.53 to 20.18) 0.152 11.27 (−1.22 to 23.75) 0.062 TLC§ (L) No BF (mean) 4.31 (4.11 to 4.51) 4.38 (4.15 to 4.62) BF ≤3 months (% difference) 3.00 (−3.13 to 9.13) 0.330 4.36 (−1.59 to 10.31) 0.143 BF >3 months (% difference) 6.13 (0.12 to 12.14) 0.039 6.57 (0.99 to 12.14) 0.017 ADC§ (cm2/s) No BF (mean) 0.09 (0.09 to 0.10) 0.09 (0.08 to 0.10) BF ≤3 months (% difference) Absolute ADC value 3.00 (−3.58 to 9.57) 0.365 0.59 (−6.31 to 7.49) 0.867 Average alveolar size¶ 7.38 (−8.45 to 24.71) 1.45 (−14.51 to 19.03) BF >3 months (% difference) Absolute ADC value 0.69 (−5.43 to 6.81) 0.824 1.88 (−4.56 to 8.32) 0.563 Average alveolar size¶ 1.69 (−12.52 to 17.17) 4.64 (−10.72 to 21.18) *The analyses were performed using the natural logarithm of the outcome variable; the estimates were back-transformed from the logarithmic scale to the original scale, therefore the means represent the geometric means, not the arithmetic ones.

) 0.563 Average alveolar size¶ 1.69 (−12.52 to 17.17) 4.64 (−10.72 to 21.18) *The analyses were performed using the natural logarithm of the outcome variable; the estimates were back-transformed from the logarithmic scale to the original scale, therefore the means represent the geometric means, not the arithmetic ones. †In the basic model we adjusted for age, sex and height. ‡In the adjusted model we included, additionally, preterm status, birth weight, Townsend score, smoking during pregnancy, maternal asthma, maternal ethnicity and early-onset wheeze (wheezing history with onset during the first year of life). §All analysis were performed using the natural logarithm; the estimates were back-transformed from the logarithmic scale to the original scale, therefore the coefficients are multiplicative (they represent ratios; eg, in the basic model, participants breastfed ≤3 months had an ADC absolute value 3% higher than participants who were not breastfed). ¶Calculated with the formula ADC ratio=(volume ratio)0.415, see Narayanan et al.13 ADC, apparent diffusion coefficient; BF, breastfeeding; FEF50, forced mid-expiratory flow; FEV1, forced expiratory volume at 1 s; FRC, functional residual capacity; FVC, forced vital capacity; RV, residual volume; TLC, total lung capacity.

§All analysis were performed using the natural logarithm; the estimates were back-transformed from the logarithmic scale to the original scale, therefore the coefficients are multiplicative (they represent ratios; eg, in the basic model, participants breastfed ≤3 months had an ADC absolute value 3% higher than participants who were not breastfed). ¶Calculated with the formula ADC ratio=(volume ratio)0.415, see Narayanan et al.13 ADC, apparent diffusion coefficient; BF, breastfeeding; FEF50, forced mid-expiratory flow; FEV1, forced expiratory volume at 1 s; FRC, functional residual capacity; FVC, forced vital capacity; RV, residual volume; TLC, total lung capacity. Effect-modification model The results from the model that tested an effect modification by maternal asthma are presented in table 3 and figure 1. We did not find significant evidence for an effect modification by maternal asthma. Among participants born to mothers with asthma there was a tendency towards larger lung volumes and alveolar size in those who were breastfed >3 months compared with those who were not breastfed but in offspring of mothers without asthma these differences were close to zero. When compared with participants who had not been breastfed, the FRC of participants breastfed >3 months was on average larger by 2.4% in offspring of non-asthmatic mothers but larger by 15.2% in those born to asthmatic mothers. Similar differences between children of asthmatic and non-asthmatic mothers were found for the other lung volumes (table 3). The ADC of participants breastfed over 3 months was 0.5% lower in those born to non-asthmatic mothers but 11.0% higher in those born to asthmatic mothers. Using the formula ADC ratio=(volume ratio)0.415, we determined that among participants breastfed for over 3 months the average alveolar size was smaller by 1.2% in participants of non-asthmatic mothers but 28.6% larger in participants of asthmatic mothers compared to participants who were not breastfed (table 3).

hers. Using the formula ADC ratio=(volume ratio)0.415, we determined that among participants breastfed for over 3 months the average alveolar size was smaller by 1.2% in participants of non-asthmatic mothers but 28.6% larger in participants of asthmatic mothers compared to participants who were not breastfed (table 3). Table 3 Association between breastfeeding and lung function measurements by maternal asthma, fully adjusted model with interaction

hers. Using the formula ADC ratio=(volume ratio)0.415, we determined that among participants breastfed for over 3 months the average alveolar size was smaller by 1.2% in participants of non-asthmatic mothers but 28.6% larger in participants of asthmatic mothers compared to participants who were not breastfed (table 3). Table 3 Association between breastfeeding and lung function measurements by maternal asthma, fully adjusted model with interaction Lung function (unit) Mothers without asthma Mothers with asthma estimate (CI) estimate (CI) p-interaction FVC* (L) No BF (mean) 3.44 (3.18 to 3.70) 3.35 (3.01 to 3.69) BF ≤3 months (% difference) 2.38 (– 6.37 to 11.13) 8.54 (– 5.68 to 22.75) 0.471 BF >3 months (% difference) 3.09 (– 4.93 to 11.10) 6.59 (– 7.92 to 21.09) 0.689 FEV1* (L) No BF (mean) 2.97 (2.73 to 3.20) 2.90 (2.59 to 3.21) BF ≤3 months (% difference) 5.27 (– 4.18 to 14.71) 3.80 (– 10.48 to 18.07) 0.869 BF >3 months (% difference) 3.23 (– 5.19 to 11.65) 12.88 (– 3.25 to 29.01) 0.308 FEF50* (L/s) No BF (mean) 3.54 (2.97 to 4.12) 3.40 (2.66 to 4.14) BF ≤3 months (% difference) 11.10 (– 9.47 to 31.66) 1.15 (– 27.30 to 29.61) 0.591 BF >3 months (% difference) 4.20 (– 13.21 to 21.60) 31.27 (– 7.11 to 69.65) 0.198 FRC* (L) No BF (mean) 1.99 (1.79 to 2.20) 1.98 (1.71 to 2.26) BF ≤3 months (% difference) 1.87 (– 9.90 to 13.64) 9.90 (– 9.60 to 29.40) 0.490 BF >3 months (% difference) 2.45 (– 8.35 to 13.24) 15.20 (– 6.01 to 36.40) 0.299 RV* (L) No BF (mean) 1.12 (0.97 to 1.26) 1.08 (0.89 to 1.27) BF ≤3 months (% difference) −4.98 (– 19.15 to 9.18) 6.67 (– 17.75 to 31.10) 0.414 BF >3 months (% difference) 4.73 (– 9.51 to 18.97) 31.65 (0.36 to 62.93) 0.117 TLC* (L) No BF (mean) 4.44 (4.17 to 4.72) 4.38 (4.02 to 4.75) BF ≤3 months (% difference) 2.07 (– 5.08 to 9.23) 9.32 (– 2.45 to 21.08) 0.303 BF >3 months (% difference) 4.65 (– 2.04 to 11.34) 11.36 (– 1.08 to 23.80) 0.365 ADC* (cm2/s) No BF (mean) 0.09 (0.08 to 0.10) 0.09 (0.08 to 0.10) BF ≤3 months (% difference) absolute ADC value 1.41 (– 6.87 to 9.70) −4.00 (– 16.12 to 8.13) 0.437 average alveolar size† 3.41 (– 15.82 to 24.98) 1.69 (– 12.52 to 17.17) BF >3 months (% difference) absolute ADC value −0.47 (– 7.89 to 6.96) 11.04 (– 3.61 to 25.70) 0.174 average alveolar size† −1.20 (– 17.98 to 17.70) 28.58 (– 8.45 to 73.50) In this model we adjusted for age, sex, height, preterm status, birth weight, Townsend score, smoking during pregnancy, maternal asthma, maternal ethnicity, and early-onset wheeze (wheezing history with onset during the first year of life) and included an interaction term between breastfeeding and maternal asthma.

.50) In this model we adjusted for age, sex, height, preterm status, birth weight, Townsend score, smoking during pregnancy, maternal asthma, maternal ethnicity, and early-onset wheeze (wheezing history with onset during the first year of life) and included an interaction term between breastfeeding and maternal asthma. The means and coefficients for the groups of children of mother with and without asthma were calculated using the regression coefficients for breastfeeding, maternal asthma and the interaction term, using the command lincom in Stata, which computes point estimates, CIs and p-values for linear combinations of coefficients. *All analyses were performed using the natural logarithm; the estimates were back-transformed from the logarithmic scale to the original scale, therefore the coefficients are multiplicative (they represent ratios; eg, in children of mothers without asthma, participants breastfed ≤3 months had an ADC value 1.4% higher than participants who were not breastfed) †The average alveolar size was calculated with the formula ADC ratio=(volume ratio)0.415, see Narayanan et al.13 ADC, apparent diffusion coefficient; BF, breastfeeding; FEF50, forced mid-expiratory flow; FEV1, forced expiratory volume at 1 second; FRC, functional residual capacity; FVC, forced vital capacity; p-int, p-interaction; RV, residual volume; TLC, total lung capacity.

†The average alveolar size was calculated with the formula ADC ratio=(volume ratio)0.415, see Narayanan et al.13 ADC, apparent diffusion coefficient; BF, breastfeeding; FEF50, forced mid-expiratory flow; FEV1, forced expiratory volume at 1 second; FRC, functional residual capacity; FVC, forced vital capacity; p-int, p-interaction; RV, residual volume; TLC, total lung capacity. Figure 1 Association between breastfeeding and lung function measurements by maternal asthma, fully adjusted model with interaction. The graph represents the adjusted means and CIs for each breastfeeding category (from left to right: none, ≤3 months and >3 months), in the entire sample (blue diamonds) and stratified by children of mothers with no asthma (green squares) and children of mothers with asthma (red triangles). The estimates come from the adjusted model with interaction. Discussion Findings and interpretation In this study we did not find evidence of an association between breastfeeding duration and lung volumes or alveolar dimensions at school age, except for larger TLC values in children breastfed over 3 months. However, we observed a consistent trend towards larger alveoli and larger lung volumes, both spirometric and plethysmographic, in children of asthmatic mothers who had been breastfed over 3 months compared to those not breastfed. While the findings were comparable to previous reports,12 they did not reach statistical significance for any of the tested outcomes in this small study.

and larger lung volumes, both spirometric and plethysmographic, in children of asthmatic mothers who had been breastfed over 3 months compared to those not breastfed. While the findings were comparable to previous reports,12 they did not reach statistical significance for any of the tested outcomes in this small study. Several authors have found positive associations between breastfeeding and lung function in school-age children.6–11 We have reported previously in a larger study that breastfed children had increased FEF50 compared with non-breastfed children. This increase was larger in participants born to mothers with asthma, with evidence for a dose–response relationship with duration of breastfeeding. Furthermore, it was accompanied by increases in FVC and FEV1. In the present study we found differences that were similar or larger in magnitude than in our earlier study, but they failed to reach statistical significance. This is not surprising as the two studies are not independent: 81% of the children in the present study were also included in the previous study. However, because of the costly and time consuming nature of 3HeMR measurements, the sample size in the present study was more modest and thus, the statistical power reduced.

This is not surprising as the two studies are not independent: 81% of the children in the present study were also included in the previous study. However, because of the costly and time consuming nature of 3HeMR measurements, the sample size in the present study was more modest and thus, the statistical power reduced. Our results were suggestive of an increase in all volumes associated with breastfeeding in children of asthmatic mothers. The precision of our estimates was too small to distinguish between the hypotheses of congruent volume increases in all components versus hyperinflation. The relative increases were largest for mean alveolar volume, suggesting that volume increases were not accompanied by a proportional increase in alveolar number. If we consider that these findings do reflect real differences in the population, it is difficult to speculate why this might be so, and why this difference is seen only in children of mothers with asthma. A possible explanation is that there were influences associated with secretion of lung growth factors in breast milk of asthmatic mothers. Another possible explanation could be residual confounding by severity of maternal asthma; perhaps mothers with more severe asthma were less likely to breastfeed, but more likely to have children with low lung volumes. Although the analysis adjusted for presence of maternal asthma, we did not have information on asthma severity in the mother.

of breastfeeding. The participants reported duration of breastfeeding when the children were 1–4 years old. However, the question had shown an excellent short-term repeatability in our cohort, with a Cohen's κ of 0.96 21 and there is independent evidence suggesting that long-term recall of breastfeeding is excellent.22 Diffusion-weighted 3HeMR uses the degree of restriction to diffusion of 3He as a proxy for dimensions of the enclosing structure. It follows that the diffusion displacement, s, should be of a similar order of magnitude to the distance between the barriers. If ‘s’ is too small, the 3-He molecules are not restricted by the barriers and ADC approximates free diffusion coefficient, D. If ‘s’ is too large, it is affected by the structures outside the barriers. Parra-Robles et al23 contended that the diffusion time employed in this study would result in ‘s’ that would be sensitive to structures outside the alveoli. However, ‘s’ in our case is only 1.58 times larger than they suggested (because of the square root relationship)24 Also, while it was true that some of the 3He atoms in our study do sample the space outside an individual alveolus and may move to the alveolar duct space, the measurements still reflect alveolar dimensions as the alveolar duct does not have an independent wall. The ultrastructure of the periphery of the lung is made up of alveolar septae. As long as the alveolar duct dimension does not increase or decrease independent of alveolar dimensions, our ADC measurements are valid proxies of alveolar dimensions. This is explored in further detail in our reply24 to Parra-Robles et al23

The ultrastructure of the periphery of the lung is made up of alveolar septae. As long as the alveolar duct dimension does not increase or decrease independent of alveolar dimensions, our ADC measurements are valid proxies of alveolar dimensions. This is explored in further detail in our reply24 to Parra-Robles et al23 The relationship between ADC and volume was derived by measuring ADC in children at different levels of inflation.13 Using this relationship to extrapolate alveolar volume ratio between participants from ADC ratio assumes similar alveolar geometry across participants, and therefore alveolar volume ratio should be interpreted with caution. A hypothesis that could not be explored in this study is that a possible association of breastfeeding with lung function measurement is age dependent, that is, the association might be present (or stronger) at younger age, compared with older ages. Unfortunately, despite the large age range of our participants, the size of the sample against which we could test this hypothesis is too small; for example, only five participants were 10 years or younger.

age dependent, that is, the association might be present (or stronger) at younger age, compared with older ages. Unfortunately, despite the large age range of our participants, the size of the sample against which we could test this hypothesis is too small; for example, only five participants were 10 years or younger. While not providing a definite answer to the research question, the study offered hints that in children of mothers with asthma, those who are breastfed might have larger lung volumes, and it opened the path towards investigating the possible mechanisms involved. Further investigations with larger sample sizes are essential to answer this important question. Although the differences in lung volumes and alveolar size might be small for individual children, if these are confirmed in a larger study it would have important consequences for public health since the proportion of women with asthma is high. Contributors: All coauthors contributed to this study, and to the writing and revising of the manuscript. Funding: Swiss National Science Foundation 3200B0-12234; Wellcome Trust UK 081367/B/06/Z; Asthma UK 07/048; Swiss National Science Foundation fellowship PZ00P3_147987. Competing interests: None declared. Patient consent: Obtained. Ethics approval: Ethics Committee of the Leicestershire Health Authority. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Key messages Vitamin D deficiency prior to hospitalization is associated with development of acute respiratory failure. Vitamin D deficient patients who develop acute respiratory failure have heightened mortality. Introduction In the USA, every year, 190 600 patients develop acute respiratory distress syndrome (ARDS), with 74 500 associated deaths and 2.2 million days in the intensive care unit (ICU).1 Incidences of ARDS among patients with acute illness and mechanically ventilated patients are 7% and 11–23%, respectively.2 Although critical care outcomes in general and those associated with acute respiratory failure have improved over time, long-term outcomes of ICU survivors has gained prominence.3 4 Among the more than 100 000 patients who survive ARDS each year, many develop cognitive abnormalities, depression, post-traumatic stress disorders and have poor health-related quality of life.1 5

ciated with acute respiratory failure have improved over time, long-term outcomes of ICU survivors has gained prominence.3 4 Among the more than 100 000 patients who survive ARDS each year, many develop cognitive abnormalities, depression, post-traumatic stress disorders and have poor health-related quality of life.1 5 Recent evidence suggests that vitamin D is a key regulator of the innate and adaptive immune system.5 Serum 25-hydroxyvitamin D (25(OH)D) is the major circulating metabolite of vitamin D, the standard measure of vitamin D status and is used to assess therapeutic response to supplementation. Low serum 25(OH)D levels are associated with increased risk of viral and bacterial infections as well as sepsis possibly due to effects on innate and adaptive immunity.6 7 Additionally, vitamin D modulates inflammation, fibrosis and airway destruction in the lung which are major steps in ARDS pathogenesis.8 9 Further, decreased muscle strength and mass is associated with vitamin D deficiency.10 Recent data from a randomised controlled trial on high-dose vitamin D in critically ill patients (VITdAL-ICU trial) shows as a secondary outcome the improved survival in patients with severe vitamin D deficiency (25(OH)D≤12 ng/mL) who receive vitamin D.11 Data are lacking with regards to the association of prehospitalisation vitamin D inadequacy and incident ARDS during critical illness.

in critically ill patients (VITdAL-ICU trial) shows as a secondary outcome the improved survival in patients with severe vitamin D deficiency (25(OH)D≤12 ng/mL) who receive vitamin D.11 Data are lacking with regards to the association of prehospitalisation vitamin D inadequacy and incident ARDS during critical illness. Given that vitamin D inadequacy, defined by measuring circulating 25(OH) D3 levels, is increasingly prevalent in the general population of the USA and associated with critical care outcomes, we performed a two-centre observational study of adult patients among whom 25(OH)D concentrations had been measured for routine clinical reasons within 1 year before hospitalisation.7 12–23 The objective of this study was to test the hypothesis that vitamin D status before hospital admission is inversely associated with the risk of developing acute respiratory failure in a critically-ill patient cohort.

centrations had been measured for routine clinical reasons within 1 year before hospitalisation.7 12–23 The objective of this study was to test the hypothesis that vitamin D status before hospital admission is inversely associated with the risk of developing acute respiratory failure in a critically-ill patient cohort. Materials and methods Data on all patients admitted to two teaching hospitals in Boston between 4 August 1998 and 12 January 2011 were obtained through a computerised registry. During the study period, there were 79 927 patient ICU admissions to the hospitals under study. Approval for the study was granted by the Partners Human Research Committee (Institutional Review Board). Requirement for consent was waived as the data were analysed anonymously. The cohort included patients ≥18 years of age who received critical care and had serum 25(OH)D measured between 7 and 365 days before hospitalisation. We excluded patients with congestive heart failure (International Classification of Diseases Ninth Edition (ICD-9) 428.0–428.9) diagnosed following hospital admission.24 The exposure of interest was serum 25(OH)D level obtained 7 to 365 days prior to the date of hospital admission and categorised as 25(OH)D <10 ng/mL; 10–19.9 ng/mL; 20–29.9 ng/mL and ≥30 ng/mL.14 We define prehospital serum 25(OH)D concentration <20 ng/mL as vitamin D inadequacy, as suggested by a recent Institute of Medicine report.25 In cases where a patient had serum 25(OH)D measured more than once in the year prior to hospitalisation, the serum 25(OH)D measured closest to the date of hospital admission was utilised.

prehospital serum 25(OH)D concentration <20 ng/mL as vitamin D inadequacy, as suggested by a recent Institute of Medicine report.25 In cases where a patient had serum 25(OH)D measured more than once in the year prior to hospitalisation, the serum 25(OH)D measured closest to the date of hospital admission was utilised. Race was either self-determined or designated by a patient representative/healthcare proxy. Patient admission ‘type’ was defined as ‘medical’ or ‘surgical’ and incorporates the Diagnostic-Related Grouping (DRG) methodology.26 We utilised the Deyo-Charlson Index to assess the burden of chronic illness which is well studied and validated.27 Sepsis was defined by the presence of any of the following ICD-9-clinical modification (CM) codes 038.0–038.9, 020.0, 790.7, 117.9, 112.5 or 112.8128, 3 days prior to critical care initiation to 7 days after critical care initiation, a definition validated in our administrative data.7 To determine neighbourhood socioeconomic disadvantage we used geocoded residential address data29 from electronic health records; we then linked the zip+4 data to the Area Deprivation Index developed by Singh et al30 and linked to the 2000 US census by Kind et al.31

, a definition validated in our administrative data.7 To determine neighbourhood socioeconomic disadvantage we used geocoded residential address data29 from electronic health records; we then linked the zip+4 data to the Area Deprivation Index developed by Singh et al30 and linked to the 2000 US census by Kind et al.31 During the study period between 1998 and 2011, the chemiluminescence assay, the radioimmunoassay or liquid chromatography-mass spectroscopy (LC-MS) were employed at different times as a 25(OH)D assay method. Dates, times and type of 25(OH)D assay were recorded. The 25(OH)D assays were tested for imprecision by the clinical laboratories at the two hospitals. Imprecision testing with human serum specimens showed within-run coefficients of variation (CVs) of ≤4.5% for the chemiluminescence assay, ≤10.8% for the radioimmunoassay, and ≤8.6% for LC-MS.6

25(OH)D assay were recorded. The 25(OH)D assays were tested for imprecision by the clinical laboratories at the two hospitals. Imprecision testing with human serum specimens showed within-run coefficients of variation (CVs) of ≤4.5% for the chemiluminescence assay, ≤10.8% for the radioimmunoassay, and ≤8.6% for LC-MS.6 The primary end point was acute respiratory failure defined by Cooke et al24 and identified by the presence of ICD-9, codes for respiratory failure or pulmonary oedema (518.4, 518.5, 518.81 and 518.82) and mechanical ventilation (96.7×), excluding congestive heart failure (428.0–428.9) following hospital admission.24 Inclusion of mechanical ventilation codes and the exclusion of heart failure codes increases the specificity of the ICD-9 code combination for ARDS.32 In subanalyses, the acute respiratory failure end point was validated against the Berlin Definition of ARDS.33 The secondary end point was 90-day all-cause mortality obtained from the Social Security Administration Death Master File. One hundred per cent of the cohort had at least 90-day follow-up. The censoring date was 5 January 2012. Power calculations and statistical analysis In the cohort under study, the incidence of acute respiratory failure was 18%. By assuming acute respiratory failure incidence in patients with prehospital serum 25(OH)D >30 ng/mL to be 18% and an OR of 1.5 for incident acute respiratory failure in patients with prehospital serum 25(OH)D ≤30 ng/mL, we needed 359 patients with 25(OH)D >30 ng/mL and 359 patients with 25(OH)D ≤30 ng/mL to achieve 80% power and 5% significance.

y failure incidence in patients with prehospital serum 25(OH)D >30 ng/mL to be 18% and an OR of 1.5 for incident acute respiratory failure in patients with prehospital serum 25(OH)D ≤30 ng/mL, we needed 359 patients with 25(OH)D >30 ng/mL and 359 patients with 25(OH)D ≤30 ng/mL to achieve 80% power and 5% significance. Unadjusted associations between vitamin D groups and outcomes were estimated by bivariable logistic regression analysis. Adjusted ORs were estimated by multivariable logistic regression models with inclusion of covariate terms chosen based on the biological plausibility of possible confounding of the vitamin D—acute respiratory failure association. For the primary model, specification of each continuous covariate was adjudicated by the empiric association with the primary outcome using Akaike's Information Criterion; the overall model fit was assessed using the Hosmer Lemeshow test. Models for secondary analyses (90-day mortality) were specified identically to the primary model in order to achieve greatest analogy. Sensitivity analysis were performed for patients with 25(OH)D measured at various time points prior to hospital admission. Locally weighted scatter plot smoothing (LOWESS) was used to graphically represent the relationship between prehospital 25(OH)D concentration and rate of acute respiratory failure. A multivariable Cox’s proportional hazards model was used to illustrate the survival among patients with acute respiratory failure as vitamin D intake increases. All p values presented are two-tailed; values below 0.05 were considered to be significant. All analyses are performed using STATA 12.0MP (College Station, Texas, USA).

iable Cox’s proportional hazards model was used to illustrate the survival among patients with acute respiratory failure as vitamin D intake increases. All p values presented are two-tailed; values below 0.05 were considered to be significant. All analyses are performed using STATA 12.0MP (College Station, Texas, USA). To validate the accuracy of ICD-9-CM and current procedural terminology (CPT)-defined acute respiratory failure assignment in our study, 206 of the 1985 cohort patients were chosen at random. Subject charts were retrospectively reviewed by two clinician investigators blinded to the subject 25(OH)D level and ICD-9-CM/CPT acute respiratory failure code assignment to determine if patients had clinical criteria for ARDS in the first 14 days of an ICU admission according to the Berlin definition: (1) acute respiratory failure not fully explained by cardiac failure or fluid overload, per the intensivist of record; (2) bilateral opacities consistent with pulmonary oedema on the chest radiograph or the CT scan; and (3) onset within 1 week after a known clinical insult or new or worsening respiratory symptoms.33 The validation criteria were met if the criteria for mild, moderate or severe ARDS, as defined by The Berlin Definition of ARDS, were achieved in the first 14 days of an ICU admission: mild if PaO2/FiO2=201–300 mm Hg, moderate if PaO2/FiO2=101–200 mm Hg and severe if PaO2/FiO2 ≤100 mm Hg with PEEP level ≥5 cm H2O was present in all ARDS cases. Diagnosis of ARDS was established by consensus of the two clinician investigators and resolved by a third in case of discrepancies.

4 days of an ICU admission: mild if PaO2/FiO2=201–300 mm Hg, moderate if PaO2/FiO2=101–200 mm Hg and severe if PaO2/FiO2 ≤100 mm Hg with PEEP level ≥5 cm H2O was present in all ARDS cases. Diagnosis of ARDS was established by consensus of the two clinician investigators and resolved by a third in case of discrepancies. Results Table 1 shows demographic characteristics of the parent ICU cohort and the study cohort. Differences between the parent and study cohorts included gender, patient type and comorbidity. In the study cohort (N=1985), most patients were women, white and had medical-related DRG with a mean age of 63 years. The mean (SD) 25(OH)D in the study cohort was 29.0 (15.5) ng/mL which did not statistically differ by season of 25(OH)D draw (p=0.12, χ2). The majority of 25(OH)D measurements in the study cohort occurred within 6 months of hospital admission (20% within 1 month, 48% within 3 months and 72% within 6 months). Most study cohort patients had 25(OH)D levels ≥20 ng/mL (8% with 25(OH)D <10 ng/mL; 24% with 10–19.9 ng/mL; 24% with 20–29.9 ng/mL and 44% with ≥30 ng/mL). The 90-day mortality of the study cohort was 15.7%. Table 1 Characteristics of total ICU cohort (N=77 927) and study cohort (N=1985)

Results Table 1 shows demographic characteristics of the parent ICU cohort and the study cohort. Differences between the parent and study cohorts included gender, patient type and comorbidity. In the study cohort (N=1985), most patients were women, white and had medical-related DRG with a mean age of 63 years. The mean (SD) 25(OH)D in the study cohort was 29.0 (15.5) ng/mL which did not statistically differ by season of 25(OH)D draw (p=0.12, χ2). The majority of 25(OH)D measurements in the study cohort occurred within 6 months of hospital admission (20% within 1 month, 48% within 3 months and 72% within 6 months). Most study cohort patients had 25(OH)D levels ≥20 ng/mL (8% with 25(OH)D <10 ng/mL; 24% with 10–19.9 ng/mL; 24% with 20–29.9 ng/mL and 44% with ≥30 ng/mL). The 90-day mortality of the study cohort was 15.7%. Table 1 Characteristics of total ICU cohort (N=77 927) and study cohort (N=1985) Total ICU cohort Study cohort N 79 927 1985 Age-mean(SD) 61.8 (18.3) 63.2 (16.2) Sex number (%) Female 33 556 (42) 1083 (55) Male 46 371 (58) 902 (45) Race number (%) White 63 712 (80) 1602 (81) Non-white 16 215 (20) 383 (19) Patient type number (%) Medical 39 355 (49) 1372 (69) Surgical 40 572 (51) 613 (31) Deyo-charlson index number (%) 0–3 56 312 (70) 1227 (62) 4–6 19 060 (24) 553 (28) >6 4555 (6) 205 (10) Body mass index-mean(SD)* 27.3 (6.9) 27.3 (7.7) Area deprivation index-mean(SD)† 86.0 (32.8) 81.8 (30.4) Sepsis number (%) 10 215 (13) 231 (12) Acute respiratory failure number (%) 12 308 (15) 351 (18) 90-day mortality number (%) 13 860 (17) 311 (16) Greater Area Deprivation Index means a greater disadvantage.

205 (10) Body mass index-mean(SD)* 27.3 (6.9) 27.3 (7.7) Area deprivation index-mean(SD)† 86.0 (32.8) 81.8 (30.4) Sepsis number (%) 10 215 (13) 231 (12) Acute respiratory failure number (%) 12 308 (15) 351 (18) 90-day mortality number (%) 13 860 (17) 311 (16) Greater Area Deprivation Index means a greater disadvantage. *5527 of the total ICU cohort and 435 study cohort patients had body mass index determined. †Area Deprivation Index is an ecological measure of socioeconomic disadvantage. ICU, intensive care unit. In the study cohort, the most common Major Diagnostic Codes in the cohort by system were circulatory 31.7%, digestive 14%, musculoskeletal 8.9%, kidney 8.4%, endocrine 7.7%, respiratory 4.8%, injuries 4.4% and infectious 4.2%. The mean (SD) time from hospital admission to acute respiratory failure diagnosis was 1.46 (4.43) days and the mean (SD) time from hospital admission to ICU admission was 1.90 (6.19) days. Validation of ICD-9-CM/CPT-defined acute respiratory failure assignment, as per the Berlin Definition of ARDS, showed that ICD-9-CM/CPT-defined acute respiratory failure assignment had a sensitivity of 59.4% (95% CI 46.4 to 71.5%), a specificity of 95.1% (95% CI 90.1 to 97.9%), a positive predictive value of 84.4% (95% CI 69.9 to 93.0) and a negative predictive value of 83.9% (95% CI 77.0 to 89.0).

he Berlin Definition of ARDS, showed that ICD-9-CM/CPT-defined acute respiratory failure assignment had a sensitivity of 59.4% (95% CI 46.4 to 71.5%), a specificity of 95.1% (95% CI 90.1 to 97.9%), a positive predictive value of 84.4% (95% CI 69.9 to 93.0) and a negative predictive value of 83.9% (95% CI 77.0 to 89.0). Patient characteristics of the study cohort were stratified according to preadmission vitamin D groups (table 2). Factors that significantly differed between stratified groups included age, gender, calcium and haematocrit. Factors that did not significantly differ between stratified groups included Deyo-Charlson index, patient type (surgical vs medical) and season of 25(OH)D measurement. Table 3 indicates that age, gender and 25(OH)D are significant predictors of acute respiratory failure in our model. Table 2 Patient characteristics by prehospital vitamin D status

Patient characteristics of the study cohort were stratified according to preadmission vitamin D groups (table 2). Factors that significantly differed between stratified groups included age, gender, calcium and haematocrit. Factors that did not significantly differ between stratified groups included Deyo-Charlson index, patient type (surgical vs medical) and season of 25(OH)D measurement. Table 3 indicates that age, gender and 25(OH)D are significant predictors of acute respiratory failure in our model. Table 2 Patient characteristics by prehospital vitamin D status Prehospital 25(OH)D, ng/mL <10.0 10–19.9 20–29.9 ≥30 Total N 164 477 473 871 1985 p Value Age-mean (SD) 57.3 (16.9) 61.3 (16.6) 62.4 (15.9) 65.9 (15.4) 63.2 (16.2) 0.001* Sex number (%) 0.01 Female 83 (51) 243 (51) 246 (52) 511 (59) 1083 (55) Male 81 (49) 234 (49) 227 (48) 360 (41) 902 (45) Race number (%) <0.0001 White 125 (76) 366 (77) 371 (78) 740 (85) 1602 (81) Non-white 39 (24) 111 (23) 102 (22) 39 (15) 383 (19) African–American 17 (10) 43 (9) 34 (7) 34 (4) 128 (6) Patient type number (%) 0.27 Medical 115 (70) 343 (72) 331 (70) 583 (67) 1372 (69) Surgical 49 (30) 134 (28) 142 (30) 288 (33) 613 (31) Deyo-Charlson index number (%) 0.087 0–3 88 (54) 285 (60) 294 (62) 560 (64) 1227 (62) 4–6 59 (36) 135 (28) 137 (29) 222 (25) 553 (28) >6 17 (10) 57 (12) 42 (9) 89 (10) 205 (10) Diabetes mellitus number (%) 49 (30) 193 (40) 159 (34) 274 (31) 675 (34) 0.005 Body mass index-mean (SD)† 27.9 (8.2) 28.1 (9.9) 28.0 (6.5) 25.9 (5.9) 27.3 (7.7) 0.064* Season of 25(OH)D draw number (%) 0.18 Spring 43 (26) 132 (28) 141 (30) 242 (28) 558 (28) Summer 35 (21) 114 (24) 120 (25) 229 (26) 498 (25) Winter 38 (23) 119 (25) 103 (22) 162 (19) 422 (21) Fall 48 (29) 112 (23) 109 (23) 238 (27) 507 (26) Calcium ≥10.5 mg/dL number (%) 26 (16) 66 (14) 60 (13) 74 (9) 226 (11) 0.003 Haematocrit <30% number (%) 31 (19) 98 (21) 88 (19) 121 (14) 338 (17) 0.009 >90 days between 25(OH)D and Admission number (%) 56 (34) 192 (40) 264 (56) 516 (59) 1028 (52) <0.001 Area deprivation index-mean (SD) 87.1 (28.1) 84.6 (27.1) 84.1 (28.3) 77.9 (33.2) 81.8 (30.4) <0.001* Sepsis number (%) 28 (17) 73 (15) 53 (11) 77 (9) 231 (12) 0.001 Sputum culture positivity number (%) 30 (21) 92 (22) 71 (17) 119 (15) 312 (17) 0.01 Acute respiratory failure number (%) 41 (25) 102 (21) 89 (19) 119 (14) 351 (18) <0.001 90-day mortality number (%) 34 (21) 111 (23) 76 (16) 90 (10) 311 (16) <0.001 p Values determined by χ2 unless designated by (*) then p value determined by ANOVA.

re positivity number (%) 30 (21) 92 (22) 71 (17) 119 (15) 312 (17) 0.01 Acute respiratory failure number (%) 41 (25) 102 (21) 89 (19) 119 (14) 351 (18) <0.001 90-day mortality number (%) 34 (21) 111 (23) 76 (16) 90 (10) 311 (16) <0.001 p Values determined by χ2 unless designated by (*) then p value determined by ANOVA. Columns may not add up to 100% due to rounding. Greater Area Deprivation Index means a greater disadvantage. †435 cohort patients had body mass index determined. Table 3 Multivariable-adjusted associations between covariates and acute respiratory failure Acute respiratory failure Absent* 1634 Acute respiratory failure Present* 352 p Value† OR‡ 95% CI p Value Age (per 1 year) 64 (16) 59 (17) <0.001 0.99 0.98 to 0.99 <0.001 Sex 0.002 Male 716 186 1.37 1.09 to 1.74 0.008 Female 918 165 1 Referent Race 0.22 Non-white 307 (19) 76 (22) 1.10 0.82 to 1.47 0.53 White 1327 (81) 275 (78) 1 Referent Patient type 0.14 Surgical 493 (70) 120 (66) 1.22 0.95 to 1.57 0.11 Medical 1141 (30) 231 (34) 1 Referent Deyo-Charlson index 0.012 0–3 1034 (63) 193 (55) 1 Referent 4–6 435 (27) 118 (34) 1.45 1.13 to 1.88 0.004 ≥6 165 (10) 40 (11) 1.30 0.89 to 1.90 0.18 25(OH)D§ <0.001 <10 ng/mL 123 (8) 41 (12) 1.84 1.22 to 2.77 0.004 10–19.9 ng/mL 375 (23) 102 (29) 1.60 1.19 to 2.15 0.002 20–29.9 ng/mL 384 (24) 89 (25) 1.37 1.01 to 1.86 0.04 ≥30 ng/mL 752 (46) 119 (34) 1.00 Referent Estimates for each variable are adjusted for all other variables in the table. *Number (%) shown except for age which is shown as Mean (SD). †p Values determined by χ2 except age determined by Kruskal-Wallis.

Acute respiratory failure Absent* 1634 Acute respiratory failure Present* 352 p Value† OR‡ 95% CI p Value Age (per 1 year) 64 (16) 59 (17) <0.001 0.99 0.98 to 0.99 <0.001 Sex 0.002 Male 716 186 1.37 1.09 to 1.74 0.008 Female 918 165 1 Referent Race 0.22 Non-white 307 (19) 76 (22) 1.10 0.82 to 1.47 0.53 White 1327 (81) 275 (78) 1 Referent Patient type 0.14 Surgical 493 (70) 120 (66) 1.22 0.95 to 1.57 0.11 Medical 1141 (30) 231 (34) 1 Referent Deyo-Charlson index 0.012 0–3 1034 (63) 193 (55) 1 Referent 4–6 435 (27) 118 (34) 1.45 1.13 to 1.88 0.004 ≥6 165 (10) 40 (11) 1.30 0.89 to 1.90 0.18 25(OH)D§ <0.001 <10 ng/mL 123 (8) 41 (12) 1.84 1.22 to 2.77 0.004 10–19.9 ng/mL 375 (23) 102 (29) 1.60 1.19 to 2.15 0.002 20–29.9 ng/mL 384 (24) 89 (25) 1.37 1.01 to 1.86 0.04 ≥30 ng/mL 752 (46) 119 (34) 1.00 Referent Estimates for each variable are adjusted for all other variables in the table. *Number (%) shown except for age which is shown as Mean (SD). †p Values determined by χ2 except age determined by Kruskal-Wallis. ‡OR for acute respiratory failure. §25(OH)D estimates adjusted for age, race, sex, Deyo-Charlson Index and patient type (medical vs surgical).

Acute respiratory failure Absent* 1634 Acute respiratory failure Present* 352 p Value† OR‡ 95% CI p Value Age (per 1 year) 64 (16) 59 (17) <0.001 0.99 0.98 to 0.99 <0.001 Sex 0.002 Male 716 186 1.37 1.09 to 1.74 0.008 Female 918 165 1 Referent Race 0.22 Non-white 307 (19) 76 (22) 1.10 0.82 to 1.47 0.53 White 1327 (81) 275 (78) 1 Referent Patient type 0.14 Surgical 493 (70) 120 (66) 1.22 0.95 to 1.57 0.11 Medical 1141 (30) 231 (34) 1 Referent Deyo-Charlson index 0.012 0–3 1034 (63) 193 (55) 1 Referent 4–6 435 (27) 118 (34) 1.45 1.13 to 1.88 0.004 ≥6 165 (10) 40 (11) 1.30 0.89 to 1.90 0.18 25(OH)D§ <0.001 <10 ng/mL 123 (8) 41 (12) 1.84 1.22 to 2.77 0.004 10–19.9 ng/mL 375 (23) 102 (29) 1.60 1.19 to 2.15 0.002 20–29.9 ng/mL 384 (24) 89 (25) 1.37 1.01 to 1.86 0.04 ≥30 ng/mL 752 (46) 119 (34) 1.00 Referent Estimates for each variable are adjusted for all other variables in the table. *Number (%) shown except for age which is shown as Mean (SD). †p Values determined by χ2 except age determined by Kruskal-Wallis. ‡OR for acute respiratory failure. §25(OH)D estimates adjusted for age, race, sex, Deyo-Charlson Index and patient type (medical vs surgical). Primary outcome Preadmission 25(OH)D was strongly associated with acute respiratory failure during hospitalisation (table 4). The odds of acute respiratory failure was 2.1-fold, 1.7-fold and 1.5-fold higher in patients with 25(OH)D values in the <10 ng/mL, 10–19.9 ng/mL, 20–29.9 ng/mL groups, respectively, compared to those with 25(OH)D ≥30 ng/mL. 25(OH)D in the cohort remains a significant predictor of odds of acute respiratory failure following adjustment for age, race, sex, Deyo-Charlson Index and patient type (medical vs surgical). The adjusted odds of acute respiratory failure was 1.8-fold, 1.6-fold and 1.4-fold higher in patients with 25(OH)D values in the <10 ng/mL, 10–19.9 ng/mL, 20–29.9 ng/mL groups, respectively, compared to those with 25(OH)D ≥30 ng/mL (table 4). LOWESS plot (figure 1) showed a near inverse linear association between vitamin D status and an acute respiratory failure rate for 25(OH)D concentrations of approximately 40 ng/mL; beyond 40 ng/mL, the association appears flat.

/mL, 20–29.9 ng/mL groups, respectively, compared to those with 25(OH)D ≥30 ng/mL (table 4). LOWESS plot (figure 1) showed a near inverse linear association between vitamin D status and an acute respiratory failure rate for 25(OH)D concentrations of approximately 40 ng/mL; beyond 40 ng/mL, the association appears flat. Table 4 Unadjusted and adjusted associations between prehospital 25(OH)D level and acute respiratory failure OR 95% CI p Value Unadjusted 25(OH)D <10 ng/mL 2.11 1.41 to 3.15 <0.0001 10–19.9 ng/mL 1.72 1.28 to 2.30 <0.0001 20–29.9 ng/mL 1.46 1.08 to 1.98 0.01 ≥30 ng/mL 1.00 Referent Adjusted 25(OH)D <10 ng/mL 1.84 1.22 to 2.77 0.004 10–19.9 ng/mL 1.60 1.19 to 2.15 0.002 20–29.9 ng/mL 1.37 1.01 to 1.86 0.04 ≥30 ng/mL 1.00 Referent CI, OR for acute respiratory failure. Referent in each case is 25(OH)D ≥30 ng/mL. Estimates adjusted for age, race, sex, Deyo-Charlson Index and patient type (medical vs surgical). Figure 1 Vitamin D status versus acute respiratory failure Rate. 25(OH)D=25-hydroxyvitamin D. Locally weighted scatter plot smoothing (LOWESS) utilised to represent the near inverse linear association between prehospital 25(OH)D concentration and acute respiratory failure rate. With bandwidth parameter=0.99, 1919 cohort patients were utilised to construct the curve.

ry failure Rate. 25(OH)D=25-hydroxyvitamin D. Locally weighted scatter plot smoothing (LOWESS) utilised to represent the near inverse linear association between prehospital 25(OH)D concentration and acute respiratory failure rate. With bandwidth parameter=0.99, 1919 cohort patients were utilised to construct the curve. Individually running the adjusted model with and without terms for assay, year, season or sepsis did not materially affect the acute respiratory failure estimates reported above (see online supplementary table S1); thus, the vitamin D-acute respiratory failure relationship was not materially confounded by these covariates. Further, utilising 25(OH)D as a continuous variable, we find that for every increase in preadmission 25(OH)D of 5 ng/mL, the odds of acute respiratory failure during hospitalisation decreases by 8% (OR 0.92, 95% CI 0.88 to 0.96; p<0.001) when adjusted for age, gender, race, type and Deyo-Charlson index. We also performed a sensitivity analysis of the timing of prehospital 25(OH)D draw date. Sensitivity analysis of the effects of excluding patients with preadmission 25(OH)D levels obtained ≥30, ≥60, ≥90 or ≥180 days prior to admission did not materially change the association between vitamin D and ICD-9-CM defined acute respiratory failure (see online supplementary table S2). A modest attenuation of the effect size was observed as the time difference increased, but the vitamin D acute respiratory failure associations retained statistical significance.

d not materially change the association between vitamin D and ICD-9-CM defined acute respiratory failure (see online supplementary table S2). A modest attenuation of the effect size was observed as the time difference increased, but the vitamin D acute respiratory failure associations retained statistical significance. In cohort patients with acute respiratory failure (n=381), vitamin D status was associated with risk of mortality. Though limited by sample size, patients with prehospital 25(OH)D values of <10 ng/mL, 10–19.9 ng/mL and 20–29.9 ng/mL had odds of 90-day mortality of 2.02 (95% CI 0.96 to 4.27; p=0.064), 2.57 (95% CI 1.49 to 4.45; p<0.001) and 1.57 (95% CI 0.89 to 2.80; p=0.12), respectively, compared to those with 25(OH)D ≥30 ng/mL following adjustments for age, race, sex, Deyo-Charlson Index and patient type (medical vs surgical; table 5). Cox proportional-hazards regression analysis demonstrated that in patients with acute respiratory failure, the risk of death declines with each 5 ng/mL increase in prehospital 25(OH)D adjusted for age, race, sex, Deyo-Charlson Index and patient type (medical vs surgical; HR 0.92, 95% CI 0.87 to 0.97; p=0.004). Table 5 Unadjusted and adjusted associations between prehospital 25(OH)D level and 90-day mortality in patients with acute respiratory failure

In cohort patients with acute respiratory failure (n=381), vitamin D status was associated with risk of mortality. Though limited by sample size, patients with prehospital 25(OH)D values of <10 ng/mL, 10–19.9 ng/mL and 20–29.9 ng/mL had odds of 90-day mortality of 2.02 (95% CI 0.96 to 4.27; p=0.064), 2.57 (95% CI 1.49 to 4.45; p<0.001) and 1.57 (95% CI 0.89 to 2.80; p=0.12), respectively, compared to those with 25(OH)D ≥30 ng/mL following adjustments for age, race, sex, Deyo-Charlson Index and patient type (medical vs surgical; table 5). Cox proportional-hazards regression analysis demonstrated that in patients with acute respiratory failure, the risk of death declines with each 5 ng/mL increase in prehospital 25(OH)D adjusted for age, race, sex, Deyo-Charlson Index and patient type (medical vs surgical; HR 0.92, 95% CI 0.87 to 0.97; p=0.004). Table 5 Unadjusted and adjusted associations between prehospital 25(OH)D level and 90-day mortality in patients with acute respiratory failure OR 95% CI p Value Unadjusted 25(OH)D <20 ng/mL 2.19 1.31 to 3.63 0.003 20–29.9 ng/mL 1.37 0.77 to 2.44 0.285 ≥30 ng/mL 1.00 Referent Adjusted 25(OH)D <20 ng/mL 2.48 1.45 to 4.25 0.001 20–29.9 ng/mL 1.48 0.82 to 2.70 0.195 ≥30 ng/mL 1.00 Referent CI, OR for 90-day mortality. N=381, 90-day mortality 39.9%. Referent in each case is 25(OH)D ≥30 ng/mL. <20 ng/mL cut point chosen for power issues. Estimates adjusted for age, race, sex, Deyo-Charlson Index and patient type (medical vs surgical). 100% of the cohort had vital status follow-up at 90 days.

OR 95% CI p Value Unadjusted 25(OH)D <20 ng/mL 2.19 1.31 to 3.63 0.003 20–29.9 ng/mL 1.37 0.77 to 2.44 0.285 ≥30 ng/mL 1.00 Referent Adjusted 25(OH)D <20 ng/mL 2.48 1.45 to 4.25 0.001 20–29.9 ng/mL 1.48 0.82 to 2.70 0.195 ≥30 ng/mL 1.00 Referent CI, OR for 90-day mortality. N=381, 90-day mortality 39.9%. Referent in each case is 25(OH)D ≥30 ng/mL. <20 ng/mL cut point chosen for power issues. Estimates adjusted for age, race, sex, Deyo-Charlson Index and patient type (medical vs surgical). 100% of the cohort had vital status follow-up at 90 days. Effect modification Analyses based on fully-adjusted models were performed to evaluate the 25(OH)D-acute respiratory failure association, and p for interaction was determined to explore for any evidence of effect modification. We individually tested for effect modification by hospital, time between 25(OH)D draw and hospital admission and season of 25(OH)D draw by adding an interaction term to the multivariate models. Effect modification analysis showed that the association between 25(OH)D and acute respiratory failure was not modified by the hospital that provided care (p interaction=0.19), or the time between 25(OH)D draw and hospital admission (p interaction=0.30). When we investigated effect modification by season of 25(OH)D draw, we identified a nearly statistically significant interaction term for 25(OH)D and acute respiratory failure (p interaction=0.07). When the cohort is analysed by restricting this to the season of 25(OH)D draw, the directionality of the association between 25(OH)D and acute respiratory failure remains, but the effect size is strongest with 25(OH)D drawn in the spring season and weakest in the fall.

nd acute respiratory failure (p interaction=0.07). When the cohort is analysed by restricting this to the season of 25(OH)D draw, the directionality of the association between 25(OH)D and acute respiratory failure remains, but the effect size is strongest with 25(OH)D drawn in the spring season and weakest in the fall. Discussion The present, two-centre study aimed to determine whether suboptimal vitamin D status prior to hospital admission was associated with acute respiratory failure in the critically ill. In unadjusted and adjusted analyses, we found increased odds of acute respiratory failure in patients with preadmission 25(OH)D <20 ng/mL. In addition, our mortality analysis suggests that patients with 25(OH)D levels <20 ng/mL before hospital admission who develop acute respiratory failure have a higher risk for mortality compared to patients with prehospital 25(OH)D levels ≥30 ng/mL.

e respiratory failure in patients with preadmission 25(OH)D <20 ng/mL. In addition, our mortality analysis suggests that patients with 25(OH)D levels <20 ng/mL before hospital admission who develop acute respiratory failure have a higher risk for mortality compared to patients with prehospital 25(OH)D levels ≥30 ng/mL. ARDS is due to diffuse alveolar damage mediated by inflammatory cytokines such as tumour-necrosis factor, interleukin (IL)-1, IL-6 and IL-8 and subsequent neutrophil recruitment and release of oxygen free radicals that damage capillary endothelium and alveolar epithelium.34 35 Immunomodulatory and proinflammatory associations of vitamin D deficiency are well known. Vitamin D induces expression of the gene for cathelicidin, which promotes intracellular killing of bacteria.36 In the lung, cathelicidin likely plays a role in mucosal defence as it is produced by neutrophils, macrophages and airway epithelium, and upregulated in response to infection and inflammation.37 These biological observations indicate the potential importance of vitamin D status to innate immunity and ARDS.

bacteria.36 In the lung, cathelicidin likely plays a role in mucosal defence as it is produced by neutrophils, macrophages and airway epithelium, and upregulated in response to infection and inflammation.37 These biological observations indicate the potential importance of vitamin D status to innate immunity and ARDS. Prior studies have reported robust associations of low vitamin D and sepsis, bloodstream infections and mortality in critically ill adults.13 The VITdAL-ICU trial secondary outcome data shows high-dose vitamin D supplementation reduces mortality in patients with severe vitamin D deficiency (25(OH)D≤12 ng/mL), but was not designed to study respiratory outcomes.11 In the current study, the reason for an increase in mortality in acute respiratory failure patients with vitamin D inadequacy is likely to be multifactorial. Other issues that may be important for critical illness outcomes include vitamin D related effects on vascular endothelial growth factor, endothelin and the renin-angiotensin-aldosterone system.38–40 Further, comorbidities including incident hypertension, glucose intolerance, the metabolic syndrome, obesity and cardiovascular disease are all associated with low 25(OH)D and higher mortality.41–45

in D related effects on vascular endothelial growth factor, endothelin and the renin-angiotensin-aldosterone system.38–40 Further, comorbidities including incident hypertension, glucose intolerance, the metabolic syndrome, obesity and cardiovascular disease are all associated with low 25(OH)D and higher mortality.41–45 The present study is not without potential limitations. Observational studies may be limited by bias, confounding, and/or reverse causation.46 Importantly, causality cannot be determined in our study. Our utilisation of ICD-9 and CPT coding likely reflects the measured incidence of acute respiratory failure in the cohort rather than the actual incidence. Ascertainment bias may exist in our study as the patient cohort under study had vitamin D status measurements for reasons that may be absent in other critically ill patients. Despite adjustment for multiple potential confounders, residual confounding may contribute to the observed differences in outcomes and vitamin D status could simply be a marker of baseline healthy behaviours. Specifically, we are unable to adjust for immobilisation, excessive alcohol intake, smoking status, genetic factors, hypertension, low-density lipoprotein-cholesterol, education level and low milk consumption all of which can alter 25(OH)D.47 48 The percentage of cohort patients who are female is 55% while the percentage of African-Americans is only 6%, which may limit generalisability. The 25(OH)D-mortality association appears to be preserved when 25(OH)D is obtained within 30 days of admission. Despite this observation, vitamin D levels at the time of hospitalisation may be different from the levels when prehospital values were drawn. We also do not have any information as to why 25(OH)D concentrations were obtained in the cohort. Finally, our observations may represent a healthy user effect rather than causality.49

this observation, vitamin D levels at the time of hospitalisation may be different from the levels when prehospital values were drawn. We also do not have any information as to why 25(OH)D concentrations were obtained in the cohort. Finally, our observations may represent a healthy user effect rather than causality.49 The present study has several strengths. To the best of our knowledge, our study is the first large sample study to evaluate an association between prehospital vitamin D and incident acute respiratory failure. Crucially, we have sufficient statistical power to detect a clinically relevant difference in acute respiratory failure between the vitamin D groups. The exclusion of ICD-9 codes related to heart failure and requirement of mechanical ventilation ICD-9 codes for our exposure increases the specificity of the codes for ARDS.24 The Deyo-Charlson Index allowed us to account for chronic medical comorbidities. Finally, by measuring vitamin D status at least 7 days prior to hospitalisation, we attempted to uncouple the influence of acute illness and inflammation on 25(OH)D levels.

or our exposure increases the specificity of the codes for ARDS.24 The Deyo-Charlson Index allowed us to account for chronic medical comorbidities. Finally, by measuring vitamin D status at least 7 days prior to hospitalisation, we attempted to uncouple the influence of acute illness and inflammation on 25(OH)D levels. Conclusion In summary, prehospital vitamin D inadequacy (25(OH)D <20 ng/mL) is associated with incident acute respiratory failure during critical illness and death in patients with acute respiratory failure. Despite our observations, supplementation of vitamin D in critically ill patients cannot be advocated for prevention or treatment of acute respiratory failure as our study does not establish causation. The data presented in combination with the hypothesis generating VITdAL-ICU trial data11 does provide an impetus to perform randomised controlled trials to determine whether vitamin D supplementation therapy might have benefit in decreasing the incidence of or improving outcomes of ARDS in critically ill patients with low 25(OH)D. Acknowledgements: This manuscript is dedicated to the memory of our dear friend and colleague Nathan Edward Hellman, MD, PhD. The authors thank Shawn Murphy and Henry Chueh and the Partners HealthCare Research Patient Data Registry group for facilitating use of their database.

Conclusion In summary, prehospital vitamin D inadequacy (25(OH)D <20 ng/mL) is associated with incident acute respiratory failure during critical illness and death in patients with acute respiratory failure. Despite our observations, supplementation of vitamin D in critically ill patients cannot be advocated for prevention or treatment of acute respiratory failure as our study does not establish causation. The data presented in combination with the hypothesis generating VITdAL-ICU trial data11 does provide an impetus to perform randomised controlled trials to determine whether vitamin D supplementation therapy might have benefit in decreasing the incidence of or improving outcomes of ARDS in critically ill patients with low 25(OH)D. Acknowledgements: This manuscript is dedicated to the memory of our dear friend and colleague Nathan Edward Hellman, MD, PhD. The authors thank Shawn Murphy and Henry Chueh and the Partners HealthCare Research Patient Data Registry group for facilitating use of their database. Contributors: DRT and KBC designed research. DRT, TM and KBC conducted research. KBC, KMM and FKG provided essential materials. KBC performed statistical analysis. DRT, TM, AAL, KA, SAQ, KAL-S, KMM, SWP, FKG, CAC, Jr, EG and KBC wrote the paper. KBC had primary responsibility for final content. All authors read and approved the final manuscript. Competing interests: None declared. Ethics approval: Partners HealthCare Human Research Committee (Institutional Review Board). Provenance and peer review: Not commissioned; externally peer reviewed.

Contributors: DRT and KBC designed research. DRT, TM and KBC conducted research. KBC, KMM and FKG provided essential materials. KBC performed statistical analysis. DRT, TM, AAL, KA, SAQ, KAL-S, KMM, SWP, FKG, CAC, Jr, EG and KBC wrote the paper. KBC had primary responsibility for final content. All authors read and approved the final manuscript. Competing interests: None declared. Ethics approval: Partners HealthCare Human Research Committee (Institutional Review Board). Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Key messages Electronic monitoring devices (EMDs) offer the most accurate solution for recording adherence to inhaled medication, yet little is known about the impact they could have on the healthcare system. This study examined the perceptions of healthcare providers and stakeholders in asthma care towards EMDs to examine what they believed their benefits could be, as well as their costs and barriers. The respondents felt that EMDs could promote better asthma control and better health for patients, and could also support decision-making and discussions with patients during clinical consultations. However, they also had concerns regarding cost, data governance and felt that more evidence was required for the effectiveness of these devices. Introduction The 2014 National Review of Asthma Deaths found that children in the UK are still dying from avoidable asthma attacks, with widespread overprescribing of rescue medication a major issue.1 Moreover, adherence rates to inhaled steroids are below 75% in children and adolescents2 and asthma still costs the National Health Service (NHS) £1 billion a year.3

a Deaths found that children in the UK are still dying from avoidable asthma attacks, with widespread overprescribing of rescue medication a major issue.1 Moreover, adherence rates to inhaled steroids are below 75% in children and adolescents2 and asthma still costs the National Health Service (NHS) £1 billion a year.3 One option to improve compliance and potentially reduce reliever overuse is electronic monitoring devices (EMDs). A recent article referred to EMDs as the ‘21st century gold standard’ for measuring inhaler use.4 These devices are now considered the optimal method for accurately and reliably recording objective data on adherence for clinical and research practice.2 5 Recent trials have extensively tested the validity and accuracy of exemplar EMDs over prolonged periods to fully demonstrate their efficacy for use in clinical research.6 7 These devices have also been associated with improved adherence to inhaler therapy.8 9 However, a recent review of currently available EMDs highlighted a lack of consideration for patient attitudes and stakeholder involvement as key issues facing these devices going forward.10

efficacy for use in clinical research.6 7 These devices have also been associated with improved adherence to inhaler therapy.8 9 However, a recent review of currently available EMDs highlighted a lack of consideration for patient attitudes and stakeholder involvement as key issues facing these devices going forward.10 Initial evidence now suggests that adolescents feel positively towards the monitoring and reminding capabilities of EMDs for helping them to demonstrate their adherence and for ensuring they remember to take their inhaler on time, with their main concerns surrounding the bulkiness and unusual appearance of the devices.11 However, in order to improve asthma care at a population level, EMDs would also require interaction from a variety of healthcare professionals; all of whom have clinical requirements that need accounting for if the provision of a medical device is to ultimately be successful.10 12 13 The stakeholders who should have access to the data produced by EMDs have not been determined. However, if data indicating a patient's inhaler use was putting their health at risk were available to a healthcare professional, but was not acted on, they could be open to criticism for not intervening.10 This potentially major issue is highlighted by a recent clinical trial using EMDs where researchers reported ‘extreme overuse’ of β-agonists in 26% of their 152 participants, with these patients actuating 32 or more doses of reliever inhaler a day.14

not acted on, they could be open to criticism for not intervening.10 This potentially major issue is highlighted by a recent clinical trial using EMDs where researchers reported ‘extreme overuse’ of β-agonists in 26% of their 152 participants, with these patients actuating 32 or more doses of reliever inhaler a day.14 To understand the potential benefits and issues that EMDs could create for the asthma care system, it is important to involve asthma stakeholders in the introduction of these devices, to help ensure they are ultimately safe and effective. The aim of the present study was to use a three-round ‘Delphi’ Survey to engage with multiple asthma care stakeholders and decision-makers in order to understand what they perceive as the key pros and cons for the introduction of EMDs into everyday asthma management. It is envisaged that this information could be used to help inform and guide the future development and successful introduction of these devices into the NHS. Methods Defining stakeholders The targeted study population—stakeholders—can be defined as ‘individuals, organisations or communities that have a direct interest in the process and outcomes of a project, research or policy endeavour’.13 For this research, stakeholders with an active interest in the future of asthma care were identified as being: Healthcare providers: General practitioners (GPs), nurses, consultants and pharmacists; Payers and purchasers: NHS clinical commissioning groups (CCGs); Advisory boards: British Thoracic Society (BTS) asthma specialist advisory group.

Methods Defining stakeholders The targeted study population—stakeholders—can be defined as ‘individuals, organisations or communities that have a direct interest in the process and outcomes of a project, research or policy endeavour’.13 For this research, stakeholders with an active interest in the future of asthma care were identified as being: Healthcare providers: General practitioners (GPs), nurses, consultants and pharmacists; Payers and purchasers: NHS clinical commissioning groups (CCGs); Advisory boards: British Thoracic Society (BTS) asthma specialist advisory group. Recruitment As the targeted population was anticipated to be difficult to recruit from, several routes were used for contacting potential participants. Paper survey Handed out to the respiratory teams at Queen’s Medical Centre and City Hospital in Nottingham, UK; Posted to members of local CCGs where they had listed their interests or specialties as ‘asthma’, ‘respiratory’ or ‘paediatrics’; Sent to members of the BTS asthma specialist advisory group; Handed out to delegates at the East Midlands Asthma Day (2015) at the University of Nottingham, UK. Online survey Emailed to doctors and nurses in Leicestershire, UK; Emailed to members of the Respiratory Effectiveness Group (REG)—an investigator-led, not-for-profit research initiative (http://effectivenessevaluation.org); Shared on Twitter by the REG (@RespirEffect); Article and link to survey posted on the Respiratory Futures website (http://www.respiratoryfutures.org.uk).

Online survey Emailed to doctors and nurses in Leicestershire, UK; Emailed to members of the Respiratory Effectiveness Group (REG)—an investigator-led, not-for-profit research initiative (http://effectivenessevaluation.org); Shared on Twitter by the REG (@RespirEffect); Article and link to survey posted on the Respiratory Futures website (http://www.respiratoryfutures.org.uk). Study design A ‘Delphi’ method was chosen as it is considered an effective research tool for collecting the judgments of experts in different physical locations.15 Through multiple rounds of surveys, this method also allows participants to view the anonymous opinions of others, then refine and adjust their own views dependent on their level of agreement.16 The three rounds of the Delphi Survey used in this study are explained further. Delphi round one Respondents were first given a short explanation of EMDs for asthma in case they were unaware of their purpose; this was carefully worded to avoid biases for or against the devices (see online supplementary appendix 1). They were then asked to provide at least six pros and six cons that they felt the introduction of EMDs could have for asthma care. 10.1136/bmjresp-2016-000159.supp1supplementary appendix

Delphi round one Respondents were first given a short explanation of EMDs for asthma in case they were unaware of their purpose; this was carefully worded to avoid biases for or against the devices (see online supplementary appendix 1). They were then asked to provide at least six pros and six cons that they felt the introduction of EMDs could have for asthma care. 10.1136/bmjresp-2016-000159.supp1supplementary appendix Delphi round two The responses from round one were collected and similar points were merged. For example, responses from participants such as ‘actual cost of device’, ‘increased inhaler costs’, ‘who funds the device?’ could all be grouped under the same point—‘cost of devices’. Participants were then provided with a randomised list of every unique point that was raised in round one, for pros and cons. Their task was then to rank each point for its level of importance, with ‘1’ being least important and ‘10’ being most important. Delphi round three The ratings from round two were then collated and analysed. Lists of the pros and cons ranked by their rated level of importance were then presented back to participants in round three and they were asked for final qualitative feedback on whether they agreed with the order, if they felt anything was missing, and if they felt that their views towards EMDs had changed over the process.

of the pros and cons ranked by their rated level of importance were then presented back to participants in round three and they were asked for final qualitative feedback on whether they agreed with the order, if they felt anything was missing, and if they felt that their views towards EMDs had changed over the process. Materials Surveys were available in paper and online formats to allow for different methods of contacting potential participants. The online version of the survey was run using Qualtrics (https://www.qualtrics.com). The information included in both versions of the survey was the same. Results Study population It was not possible to accurately estimate the number of potential participants compared with the number who responded because of the multiple different routes of recruitment that were used. Participants recruited into the study all took part in round one of the Delphi, with dropout meaning a reduced number responded to round two and a further reduced number completed round three. Details of the sample at each stage are provided in table 1. Although there is no concrete recommendation for the sample size required for a Delphi Study, they are rarely conducted with <10 participants.17 The sample used here therefore falls within recommended guidelines. Table 1 Sample size across the three rounds of the Delphi Survey with occupation demographics included

Participants recruited into the study all took part in round one of the Delphi, with dropout meaning a reduced number responded to round two and a further reduced number completed round three. Details of the sample at each stage are provided in table 1. Although there is no concrete recommendation for the sample size required for a Delphi Study, they are rarely conducted with <10 participants.17 The sample used here therefore falls within recommended guidelines. Table 1 Sample size across the three rounds of the Delphi Survey with occupation demographics included Delphi round Total Consultants GPs Nurses Pharmacists No occupation provided CCG/advisory board members (from the existing sample) Round one 31 8 6 9 1 7 5 Round two 18 8 3 6 1 0 4 Round three 10 3 1 6 0 0 1 CCG, clinical commissioning group; GP, general practitioner. Round one results After collecting the responses for round one, the 154 pros and 159 cons of EMDs provided by the respondents were analysed to find occasions where different participants had raised the same point. Responses were then grouped and given names to accurately report every unique point raised by the participants. A total of 29 pros and 32 cons were found from the data in survey one and are displayed in tables 2 and 3, respectively. The number of times each point was raised independently is shown in the tables and provides an indication of the factors that were brought up most frequently. Table 2 The 29 pros participants gave for electronically monitoring inhaler use, with the number of times each point was raised

A total of 29 pros and 32 cons were found from the data in survey one and are displayed in tables 2 and 3, respectively. The number of times each point was raised independently is shown in the tables and provides an indication of the factors that were brought up most frequently. Table 2 The 29 pros participants gave for electronically monitoring inhaler use, with the number of times each point was raised Pros Sum 1. An accurate record of adherence for clinicians/nurses/general practitioners to use in auditing and review 24 2. Reminding the patient to use their inhaler 17 3. For identifying patterns of inhaler use, for example, days, times, school, holidays, etc 12 4. Aiding discussions between the clinician and patient, for example, visual evidence 11 5. Improve compliance 8 6. Reducing costs through less wasted medication and less time in hospital 8 7. Relating an accurate record of a patient’s inhaler use to their health outcomes and asthma control 7 8. Data for research 6 9. Patient can see their inhaler use from home and know if they are underusing/overusing 6 10. Patient has proof of their adherence to share with their clinician—increasing trust 6 11. Increase patient involvement and motivation for treating their condition 5 12. More informed decision-making for clinicians 5 13. Better asthma control and improved quality of life 4 14. Adding the ability to alert when the inhaler is about to run out would be beneficial 3 15. Adding the ability to monitor inhaler technique would be beneficial 3 16. Can be used to identify inhaler types that are less likely to be used—to ultimately find the most widely accepted and used inhaler types 3 17. Increasing patient independence, accountability and self-management for their asthma 3 18. Parents can check on their child's inhaler use 3 19. The patient's awareness of monitoring by their clinician may improve their compliance 3 20. ‘Cool’ technology may appeal to patients 2 21. Could reduce exacerbations 2 22. GPS would be beneficial in identifying triggers for a patient’s asthma, for example, pollen, pollution 2 23. Could be used with other monitoring techniques, for example, peak flow 1 24. Helpful for identifying dose dumping 1 25. Increasing patient confidence in their care 1 26. Long term—could be used to develop bio feedback 1 27. Promote competition 1 28. Useful data for emergency situations 1 29. Useful to monitor adherence of different groups of patients on different treatments 1 Table 3 The 32 cons participants gave for electronically monitoring inhaler use, with the number of times each point was raised

be used to develop bio feedback 1 27. Promote competition 1 28. Useful data for emergency situations 1 29. Useful to monitor adherence of different groups of patients on different treatments 1 Table 3 The 32 cons participants gave for electronically monitoring inhaler use, with the number of times each point was raised Cons Sum 1. Cost of devices 32 2. Bulkiness and appearance may put patients off 14 3. Patient may not like being ‘watched’ 12 4. Accuracy and reliability of the device, as well as potential technical issues 9 5. Concerns over the time and workload this would add to the consultation process 9 6. Concerns if this is only compatible with MDIs 7 7. Records actuation but not inhalation, technique, nor identifies if canister is empty or inhaler is shared 7 8. Whose responsibility is downloading, processing and interpreting the data and discussing with patients? 7 9. How is data stored and who has access? 6 10. An EMD may be required for more than one inhaler per patient 5 11. Evidence of the effectiveness of EMDs is required 4 12. Cleaning and maintenance of the device 3 13. Concerns about the role of pharma companies 3 14. Could interfere with inhalation technique or not be compatible with spacer 3 15. Data overload 3 16. Ease of use—another thing patients have to learn 3 17. Elderly patients may struggle with the technology or have a negative attitude towards it 3 18. May make no difference to already unengaged patients 3 19. May put patients off coming to clinic particularly if they have failed 3 20. Paternalistic approach 3 21. Added cost/time/workload of training clinicians and staff on how to use device, how to teach patients and how to interpret results 2 22. Are there better alternatives, for example, Tele-health or Medication Possession Ratio (MPR)? 2 23. Over-reliance on data—also need to determine reasons for non-adherence 2 24. Patient may forget to bring device with them to clinic 2 25. Patient resistance or refusal to use the device 2 26. Patients may find the reminders a nuisance 2 27. Bad for the environment—plastic and batteries 1 28. Could create potential conflicts between the patient and their clinician or parents 1 29. Many who get this device may do so as there are adherence concerns and therefore will show (inevitably) that adherence is poor 1 30. More benefits for researchers than patients, meaning patients may fail to see worth 1 31. Non-adopters lead to selection biases 1 32.

ween the patient and their clinician or parents 1 29. Many who get this device may do so as there are adherence concerns and therefore will show (inevitably) that adherence is poor 1 30. More benefits for researchers than patients, meaning patients may fail to see worth 1 31. Non-adopters lead to selection biases 1 32. This will not address intentional non-adherence 1 EMD, electronic monitoring device. Round two results Participants in round two rated each point for its relative importance on a scale where 10 was most important and 1 was least important. To analyse these data, the number of 10, 9 and 8s were counted for each issue. The most important issues were then selected using the following criteria: Total number of 10 and 9s; If still equal between two or more issues—the total number of 10, 9 and 8s. This calculation was done for the full sample of participants (N=18) as well as for the two largest subgroups—consultants (n=8) and nurses (n=6). The top five most important pros and cons for the full sample are shown in tables 4 and 5, with the data for consultants and nurses shown in table 6. Table 4 The top five pros rated most important by the participants (10=most important, 1=least important)

This calculation was done for the full sample of participants (N=18) as well as for the two largest subgroups—consultants (n=8) and nurses (n=6). The top five most important pros and cons for the full sample are shown in tables 4 and 5, with the data for consultants and nurses shown in table 6. Table 4 The top five pros rated most important by the participants (10=most important, 1=least important) Five most important pros (N=18) 10s, 9s 10s, 9s, 8s 1. Better asthma control and improved quality of life 9 13 2. Aiding discussions between the clinician and patient, for example, visual evidence 8 15 3. The patient's awareness of monitoring by their clinician may improve their compliance 7 14 4. Increase patient involvement and motivation for treating their condition 7 12 5. More informed decision-making for clinicians 7 11 Table 5 The top five cons rated most important by the participants (10=most important, 1=least important)

t's awareness of monitoring by their clinician may improve their compliance 7 14 4. Increase patient involvement and motivation for treating their condition 7 12 5. More informed decision-making for clinicians 7 11 Table 5 The top five cons rated most important by the participants (10=most important, 1=least important) Five most important cons (N=18) 10s, 9s 10s, 9s, 8s 1. Evidence of the effectiveness of electronic monitoring devices is required 9 11 2. Records actuation but not inhalation, technique, nor identifies if the canister is empty or the inhaler is shared 6 11 3. Whose responsibility is downloading, processing and interpreting the data and discussing with patients? 6 10 4. Could interfere with inhalation technique or not be compatible with spacer 6  9 5. Patient may forget to bring device with them to clinic 6  7 Table 6 The top three most important pros and cons for the two occupation groups with the largest samples—consultants and nurses (10=most important, 1=least important)

? 6 10 4. Could interfere with inhalation technique or not be compatible with spacer 6  9 5. Patient may forget to bring device with them to clinic 6  7 Table 6 The top three most important pros and cons for the two occupation groups with the largest samples—consultants and nurses (10=most important, 1=least important) 10s, 9s 10s, 9s, 8s Three most important pros Consultants (n=8) Aiding discussions between the clinician and the patient, for example, visual evidence 5 7 More informed decision-making for clinicians 5 6 The patient's awareness of monitoring by their clinician may improve their compliance 4 5 Nurses (n=6) Better asthma control and improved quality of life 4 5 Improve compliance 3 5 Relating an accurate record of a patient's inhaler use to their health outcomes and asthma control 3 3 Three most important cons Consultants (n=8) Evidence of the effectiveness of electronic monitoring devices is required 4 4 Patient may not like being ‘watched’ 2 5 Could interfere with inhalation technique or not be compatible with spacer 2 4 Nurses (n=6) Whose responsibility is downloading, processing and interpreting the data and discussing with patients? 4 5 An electronic monitoring device may be required for more than one inhaler per patient 4 4 Accuracy and reliability of the device, as well as potential technical issues 4 4 As can be observed from the data, the points rated most important in round two are not the same points that were raised most frequently in round one. ‘Evidence of the effectiveness of EMDs is required’ was raised on only four occasions in the first round of the survey, yet was given a rating of 9 or 10 by half of the sample, making it the most important con facing EMDs, according to the results.

are not the same points that were raised most frequently in round one. ‘Evidence of the effectiveness of EMDs is required’ was raised on only four occasions in the first round of the survey, yet was given a rating of 9 or 10 by half of the sample, making it the most important con facing EMDs, according to the results. For consultants and nurses, there was a difference between the points they rated as most important. The pros rated important by the consultants appear to focus more on supporting the consultation process; through providing them with visual evidence to show patients—such as graphs on their adherence over a period of time, and by allowing them to make more informed decisions about a patient's care plan. In contrast, the pros rated highly by nurses focus more on the patient and their health, asthma outcomes and medication use. For cons, consultants rated ‘evidence of effectiveness of EMDs is required’ as most important, while nurses were more concerned with where responsibility would lie for downloading, interpreting and discussing the data with patients. Interestingly, ‘records actuation but not inhalation, technique, nor identifies if canister is empty or inhaler is shared’ was rated as the second most important con overall (see table 5), yet was not rated in the top three cons for either consultants or nurses (see table 6). This was due to high ratings of 10 from two GPs, who were a smaller participant group.

nhalation, technique, nor identifies if canister is empty or inhaler is shared’ was rated as the second most important con overall (see table 5), yet was not rated in the top three cons for either consultants or nurses (see table 6). This was due to high ratings of 10 from two GPs, who were a smaller participant group. Round three results In round three, the remaining 10 participants reviewed the ranked list of pros and cons formed from the ratings given in round two and provided their final comments and feedback. Respondents provided new feedback on some of the pros and cons that had been formed in round one. For example, one consultant felt that ‘better asthma control and improved quality of life’ was not necessarily “…implicit in the use of smart inhalers [EMDs ]: it is what we hope will happen….” The same consultant felt that a pro that was missing from the list and should be considered important would be “Collecting data for making large scale decisions in terms of health services delivery and research.” Furthermore, the same consultant felt that a con that was currently missing but could be considered related to ‘cost of devices’ was ‘institutional procurement process’ and stated, “it is a complete pain to acquire things that aren't either drugs for a patient or equipment for the institute.”

es delivery and research.” Furthermore, the same consultant felt that a con that was currently missing but could be considered related to ‘cost of devices’ was ‘institutional procurement process’ and stated, “it is a complete pain to acquire things that aren't either drugs for a patient or equipment for the institute.” One GP disagreed with EMDs being regarded as a ‘paternalistic approach’ and instead commented, “this is genuinely treating patients as an adult and empowering them.” The same participant felt that ‘data overload’ should be rated as more important, stating that this is “a genuine fear which needs to be addressed.” Elsewhere, a nurse felt that more should be made of the technology appealing to patients, “…particularly as we are often dealing with a younger adult age group.” Many of the respondents spoke about how taking part in the research had changed their views towards EMDs. For example, one consultant said that their attitude had moved from “not at all interested [in using EMDs] to possibly interested for selected patients.” Whereas a nurse stated that their views were different because they now had a “…better understanding about how they work and the benefits to the patients.” However, another nurse who had previous experience with EMDs still had reservations about the devices, and believed they could often be prone to errors and faults, stating “these can be frustrating for both patients and clinicians, and may cause problems when reviewing results.”

work and the benefits to the patients.” However, another nurse who had previous experience with EMDs still had reservations about the devices, and believed they could often be prone to errors and faults, stating “these can be frustrating for both patients and clinicians, and may cause problems when reviewing results.” Overall review of the pros and cons associated with electronically monitoring inhaler use Pros and cons were obtained which relate to the patient and their health, the job and workload of the clinician, advancing asthma research and the practicalities of EMDs more generally. An in-depth review of these factors is provided here, focusing on the factors where the most in-depth supporting quotes from participants were received. Patient-related factors Participants identified several pros and cons of EMDs that could directly affect the health and care of patients with asthma. The factors where the most in-depth comments were received are discussed here and a full list is provided in table 7. Table 7 The pros and cons that are related to the patient

Patient-related factors Participants identified several pros and cons of EMDs that could directly affect the health and care of patients with asthma. The factors where the most in-depth comments were received are discussed here and a full list is provided in table 7. Table 7 The pros and cons that are related to the patient Patient-related pros Patient-related cons ▸ Reminding the patient to use their inhaler ▸ Bulkiness and appearance may put patients off ▸ Improve compliance ▸ Patient may not like being ‘watched’ ▸ Patient can see their inhaler use from home and know if they are underusing/overusing ▸ An electronic monitoring device may be required for more than one inhaler per patient ▸ Patient has proof of their adherence to share with their clinician—increasing trust ▸ Ease of use—another thing patients have to learn ▸ Increase patient involvement and motivation for treating their condition ▸ Elderly patients may struggle with the technology or have a negative attitude towards it ▸ Better asthma control and improved quality of life ▸ May make no difference to already unengaged patients ▸ Adding the ability to alert when the inhaler is about to run out would be beneficial ▸ May put patients off coming to clinic particularly if they have failed ▸ Increasing patient independence, accountability and self-management for their asthma ▸ Paternalistic approach ▸ Parents can check on their child's inhaler use ▸ Patient may forget to bring the device with them to clinic ▸ Patient's awareness of monitoring by their clinician may improve their compliance ▸ Patient resistance or refusal to use the device ▸ ‘Cool’ technology may appeal to patients ▸ Patient may find the reminders a nuisance ▸ Could reduce exacerbations ▸ Could create potential conflicts between the patient and their clinician or parents ▸ Increasing patient confidence in their care ▸ Many who get this device may do so as there are adherence concerns and this will show (inevitably) that adherence is poor ▸ Promote competition ▸ More benefits for researchers than patients, meaning patients may fail to see worth ▸ This will not address intentional non-adherence The respondents often spoke about how EMDs could have the potential to improve adherence, with some continuing to describe how this could lead to improved asthma control and quality of life.

benefits for researchers than patients, meaning patients may fail to see worth ▸ This will not address intentional non-adherence The respondents often spoke about how EMDs could have the potential to improve adherence, with some continuing to describe how this could lead to improved asthma control and quality of life. Comments included “compliance enhanced,” “Effective use of treatment,” “more likely to comply” and “Improved care, because of increased adherence.” Many of the respondents felt that the reminders an EMD could provide would have a direct effect on how often patients were using their inhalers. For example, one nurse stated, “The EMD would be useful in patients who forget to take their inhaler, and acts as a reminder, and get them into a habit of taking their inhaler regularly,” whereas another nurse commented, “Some children would love the fact that devices are being used to remind them….”

ng their inhalers. For example, one nurse stated, “The EMD would be useful in patients who forget to take their inhaler, and acts as a reminder, and get them into a habit of taking their inhaler regularly,” whereas another nurse commented, “Some children would love the fact that devices are being used to remind them….” Participants also commented on the positive effect that EMDs could have for motivating patients and creating a sense of independence and accountability for their condition. A nurse commented that “ the patients can individualise the EMD by selecting their own tune and time of the alarm, which by involving the patient in choice may increase adherence to medication” while a pharmacists and CCG member stated “Uploadable results may improve patient involvement and interest in their treatment.” Some comments were also received from respondents who felt that the technology could be an attractive feature, such as “The device may appeal to patients who like the new technology and gadget feel.”

pharmacists and CCG member stated “Uploadable results may improve patient involvement and interest in their treatment.” Some comments were also received from respondents who felt that the technology could be an attractive feature, such as “The device may appeal to patients who like the new technology and gadget feel.” Participants also felt that the data recorded by EMDs could foster a trusting relationship between the patient and their clinician. For example, one respiratory consultant stated, “patients know that clinicians ‘believe ’ them re concordance” while a GP similarly said that it “may encourage asthma review as have ‘done well ’ and will receive praise.” Some respondents also felt that a patient simply being aware of their inhaler use being monitored could have a positive impact, with a respiratory consultant commenting, “Positive observation effect: people won't know when they are being watched so are more likely to behave.” While participants recognised the benefit EMDs could have for patients with asthma, they equally had concerns for the potential barriers that may need to be addressed or overcome for benefits to come to fruition.

Participants also felt that the data recorded by EMDs could foster a trusting relationship between the patient and their clinician. For example, one respiratory consultant stated, “patients know that clinicians ‘believe ’ them re concordance” while a GP similarly said that it “may encourage asthma review as have ‘done well ’ and will receive praise.” Some respondents also felt that a patient simply being aware of their inhaler use being monitored could have a positive impact, with a respiratory consultant commenting, “Positive observation effect: people won't know when they are being watched so are more likely to behave.” While participants recognised the benefit EMDs could have for patients with asthma, they equally had concerns for the potential barriers that may need to be addressed or overcome for benefits to come to fruition. First, many comments pertaining to the bulkiness and size of the device were received. One GP and CCG member stated that the “EMD may make the whole inhaler bulky—not user friendly and may discourage patients to clip it onto their inhaler,” while a respiratory consultant with first-hand experience of EMDs warned, “they were not popular with patients, many liked the idea but found the reality of the device too bulky and cumbersome.” Others additionally felt that the usability of the device could be a potential issue, with a nurse and CCG member commenting, “this may be another device the patient has to get to grips with along with the various types of inhaler they may be using.” This could be a particular problem in older patients; a GP and CCG member asked, “Are they accessible to the older generation who may not be quite so ok with using technology?”

member commenting, “this may be another device the patient has to get to grips with along with the various types of inhaler they may be using.” This could be a particular problem in older patients; a GP and CCG member asked, “Are they accessible to the older generation who may not be quite so ok with using technology?” As opposed to the potential benefits for some patients in encouraging good adherence, respondents felt there was potential for EMDs to have a negative effect on a patient's attitude. A GP commented, “Intrusion may not be well received—‘Big Brother’” and a nurse stated, “people are highly resistant to being watched.” Some felt this could even effect whether patients attend clinic visits, with a respiratory consultant stating, “Patients may feel their clinician is checking up on them and may be put off coming for review.” Comments were also received stating that this was a “paternalistic approach to medicine” and a “restriction of a patient's autonomy.” However, this was not a universal opinion. Clinician-related factors With the data collected by an EMD having wide-ranging uses and implications for a clinician and their care of patients; the participants identified many different pros and cons related to healthcare professionals. The factors where the most in-depth comments were received are discussed here and a full list is provided in table 8. Table 8 The pros and cons that are related to the clinician

Clinician-related factors With the data collected by an EMD having wide-ranging uses and implications for a clinician and their care of patients; the participants identified many different pros and cons related to healthcare professionals. The factors where the most in-depth comments were received are discussed here and a full list is provided in table 8. Table 8 The pros and cons that are related to the clinician Clinician-related pros Clinician-related cons ▸ An accurate record of adherence for clinicians/nurses/general practitioners to use in auditing and review ▸ Concerns over the time and workload this would add to the consultation process ▸ For identifying patterns of inhaler use, for example, days, times, school, holidays, etc ▸ Whose responsibility is downloading, processing and interpreting the data and discussing with patients? ▸ Aiding discussions between the clinician and the patient, for example, visual evidence ▸ Data overload ▸ Relating an accurate record of a patient's inhaler use to their health outcomes and asthma control ▸ Added cost/time/workload of training clinicians and staff on how to use device, how to teach patients and how to interpret results ▸ More informed decision-making for clinicians ▸ Over-reliance on data—also need to determine reasons for non-adherence ▸ GPS would be beneficial in identifying triggers for a patient's asthma, for example, pollen, pollution ▸ Helpful for identifying dose dumping ▸ Useful data for emergency situations Many of the participants spoke about the devices providing an accurate record of adherence that could be used by a clinician in their auditing of a patient. For example, one respiratory consultant said that EMDs would create “a tool to monitor actual adherence achieved against the goals previously agreed.” Some respondents specifically mentioned how this would be a more accurate estimate of inhaler use than asking the patient directly, with one CCG member saying it would “improve assessment of compliance—rather than rely on a patient telling their clinician.”

ctual adherence achieved against the goals previously agreed.” Some respondents specifically mentioned how this would be a more accurate estimate of inhaler use than asking the patient directly, with one CCG member saying it would “improve assessment of compliance—rather than rely on a patient telling their clinician.” Participants then spoke of how this accurate record of adherence would be useful for identifying patterns of use, with one nurse and CCG member stating, “Being able to demonstrate patterns of use… is particularly important for patients who forget to use their inhalers or are unsure of the frequency of use for ‘as required’ inhalers.” Other participants then additionally spoke about how a true record of inhaler use could be related to health outcomes and asthma control, “the data could be used to establish if their asthma control is poor due to reduced adherence to medication or their medication requires escalation to improve control.”

d’ inhalers.” Other participants then additionally spoke about how a true record of inhaler use could be related to health outcomes and asthma control, “the data could be used to establish if their asthma control is poor due to reduced adherence to medication or their medication requires escalation to improve control.” Furthermore, participants commented on how this detailed picture of a patient's medication management would ultimately lead to more informed decision-making, with one respiratory consultant stating it would provide “objective data for clinicians to make decisions, not report” and another consultant adding, “Changes in medication…could be based on accurate reports of current medication regimen administered.” There was also a shared feeling that the data could be used to aid discussions between a clinician and patient. One nurse provided the insightful comment that “We do practice open conversations, but seeing things written down or pictorially presented may be more effective in devising an action plan together.” Much the same as with factors related to the patient, the participants had concerns about the potential barriers EMDs may have to overcome to provide true benefit for clinicians. First, there were concerns expressed by participants about the time and added workload EMDs could create. One nurse said, “The time to download and examine the data collected from the EMDs in clinic may be limited,” while a GP and CCG member stated, “Doctors don’t have enough time to follow up adherence.”

Much the same as with factors related to the patient, the participants had concerns about the potential barriers EMDs may have to overcome to provide true benefit for clinicians. First, there were concerns expressed by participants about the time and added workload EMDs could create. One nurse said, “The time to download and examine the data collected from the EMDs in clinic may be limited,” while a GP and CCG member stated, “Doctors don’t have enough time to follow up adherence.” Questions were also asked about who should actually take responsibility for downloading, processing and interpreting the data and then discussing it with the patient. One participant stated, “some clinicians may not appreciate that the time taken to discuss results together is actually saving time in the future. Who is the best professional to discuss data with patient/family? GP, nurse, specialist doctor or other?” Related to this was a comment about the cost of training, “clinicians will require training on the use and interpreting of the results.” Research-related factors With a proportion of the recruited sample having strong ties to academic research, points were raised in relation to using EMDs for asthma research going forward, as well as further research required now. The factors where the most in-depth comments were received are discussed here and a full list is provided in table 9. Table 9 The pros and cons that are related to research

Research-related factors With a proportion of the recruited sample having strong ties to academic research, points were raised in relation to using EMDs for asthma research going forward, as well as further research required now. The factors where the most in-depth comments were received are discussed here and a full list is provided in table 9. Table 9 The pros and cons that are related to research Research-related pros Research-related cons ▸ Data for research ▸ Evidence of the effectiveness of electronic monitoring devices is required ▸ Can be used to identify inhaler types that are less likely to be used—to ultimately find the most widely accepted and used inhaler types ▸ Are there better alternatives, for example, Tele-health or Medical Possession Ratio (MPR)? ▸ Long-term could be used to develop bio-feedback ▸ Non-adopters lead to selection biases ▸ Useful to monitor adherence of different groups of people on different treatments One respiratory consultant felt that EMDs could be useful “to monitor concordance of different groups of patients on different treatment regimens” to examine how adherence varies from medication to medication. Along a similar line, a GP commented that EMDs could be used to “identify devices which are less likely to be used and weed them out, and which ones are easier and more likely to be used.”

nce of different groups of patients on different treatment regimens” to examine how adherence varies from medication to medication. Along a similar line, a GP commented that EMDs could be used to “identify devices which are less likely to be used and weed them out, and which ones are easier and more likely to be used.” Another GP spoke of the potential for location monitoring, stating it could be useful “to identify triggers…but may not be accurate if trigger is mobile.” Elsewhere, a nurse recognised the benefit EMDs could have for research more generally, “to validate study results by showing compliance with a particular treatment.” Although research exists that demonstrates the accuracy of EMDs, some participants felt that more evidence is needed. One respiratory consultant asked, “Has the data they collect been validated? Personal use with an older generation of these devices [on a research study] has been poor. Ranged from recording puffs for patients who had never used them to recording nothing at all when we knew they were definitely used.” Another participant asked, “does the EMD have a sustainable effect as a reminder?” Practical factors Many of the points raised by the participants were applicable to the healthcare industry in general and focused on the practicalities of the devices. The factors where the most in-depth comments were received are discussed here and a full list is provided in table 10. Table 10 The pros and cons that are related to practical factors

Practical factors Many of the points raised by the participants were applicable to the healthcare industry in general and focused on the practicalities of the devices. The factors where the most in-depth comments were received are discussed here and a full list is provided in table 10. Table 10 The pros and cons that are related to practical factors Practical-related pros Practical-related cons ▸ Reducing costs through less wasted medication and less time in hospital ▸ Cost of devices ▸ Could be used with other monitoring techniques, for example, peak flow ▸ Accuracy and reliability of the device, as well as potential technical issues ▸ Concerns if this is only compatible with MDIs ▸ Records actuation but not inhalation, technique, nor identifies if canister is empty or inhaler is shared ▸ How is data stored and who has access? ▸ Cleaning and maintenance of the device ▸ Concerns about the role of pharma companies ▸ Could interfere with inhalation technique or not be compatible with spacer ▸ Bad for the environment—plastic and batteries Some participants recognised that better adherence and asthma control could lead to a reduction in hospital admissions, wasted medication and healthcare costs. One nurse and CCG member stated, “Ensuring better adherence to inhaler treatment will have a positive effect on wider health economics, eg, more patients being well managed.”

nts recognised that better adherence and asthma control could lead to a reduction in hospital admissions, wasted medication and healthcare costs. One nurse and CCG member stated, “Ensuring better adherence to inhaler treatment will have a positive effect on wider health economics, eg, more patients being well managed.” While the potential positive financial effects of EMDs were recognised by some participants, many had concerns about the initial cost of the devices. One nurse asked, “How much do they cost? Is this cost incurred by the department, GP practise or patient?” while a GP stated, “The cost of each EMD may be expensive and may be limited to the more severe asthmatics having them.” The points were typically questioning of the cost rather than dismissive of it, highlighting a need for better clarity on initial costings versus potential savings. There were also some concerns regarding the reliability of the devices with a nurse asking, “How robust are they? They need to be up to being thrown around and chucked in bags or drawers and left in wet steamy bathrooms!” Participants also raised issue with the fact the devices typically only record actuation and do not record inhalation, technique nor detect if the inhaler is shared. One participant stated, “I would want assurance that the device was not open to abuse, eg, removable in order to conceal overuse or inhaler sharing, or pressing device and not inhaling.”

ith the fact the devices typically only record actuation and do not record inhalation, technique nor detect if the inhaler is shared. One participant stated, “I would want assurance that the device was not open to abuse, eg, removable in order to conceal overuse or inhaler sharing, or pressing device and not inhaling.” Discussion Implications for EMDs and asthma care By examining the pros and cons of electronically monitoring inhaler use from the perspective of stakeholders, this research provides a knowledge base for EMD developers. The data gathered should be used as a guide to help inform future directions for improving the design of the devices and for providing greater clarity to stakeholders for how they could realistically be integrated into everyday asthma care. For example, participants here expressed concerns about the time these devices could add to the consultation process, as well as the training that would be required to use them. EMD developers should work alongside healthcare professionals to ensure devices and accompanying software are simple and intuitive to reduce their impact on consultation time.

pressed concerns about the time these devices could add to the consultation process, as well as the training that would be required to use them. EMD developers should work alongside healthcare professionals to ensure devices and accompanying software are simple and intuitive to reduce their impact on consultation time. Limitations and future research The sample used here was sufficient to gather a comprehensive list of the main pros and cons associated with EMDs, but was limited in the number of participants who gave ratings for each point's importance. Furthermore, the sample contained more consultants and GPs than nurses, meaning a balance of different viewpoints may not have been fully achieved. Future research should look to distribute the list of pros and cons to a larger sample, such as a respiratory organisation, to gather a larger data set for the issues stakeholders rate as most crucial to address. Future research should also address data governance. While participants here did question whose responsibility it would be to download, interpret and discuss the data with patients, this should be examined further. Working with stakeholders to establish how data could realistically be handled and managed within the NHS will help to reduced fears of ‘data overload’ and ensure that the data can be used appropriately and effectively to ultimately benefit patient care.

discuss the data with patients, this should be examined further. Working with stakeholders to establish how data could realistically be handled and managed within the NHS will help to reduced fears of ‘data overload’ and ensure that the data can be used appropriately and effectively to ultimately benefit patient care. Conclusion Through collecting the views of stakeholders with a direct interest in the future of asthma care, a comprehensive list of the pros and cons associated with introducing EMDs into the NHS has been obtained. The stakeholders raised points relating to the care and well-being of patients, the work of a clinician, research and more practical factors such as device reliability and cost. Factors that particularly need addressing by EMD developers include assurances over the cost of the devices and data governance. The findings provide a guide for where EMDs could add most benefit and additionally identify the key issues and challenges that need to be overcome for their introduction to be successful. SH is based in the Horizon Centre for Doctoral Training at the University of Nottingham and funded through the Engineering and Physical Sciences Research Council (EPSRC) (grant number EP/G037574/1). Thanks go to the Respiratory Effectiveness Group (REG) (http://effectivenessevaluation.org), Respiratory Futures, Dr Dermot Ryan (Leicester, UK) and Dr John Blakey (Liverpool, UK) in helping to disseminate the surveys.

d through the Engineering and Physical Sciences Research Council (EPSRC) (grant number EP/G037574/1). Thanks go to the Respiratory Effectiveness Group (REG) (http://effectivenessevaluation.org), Respiratory Futures, Dr Dermot Ryan (Leicester, UK) and Dr John Blakey (Liverpool, UK) in helping to disseminate the surveys. Contributors: SH designed the study, collected the data, interpreted the data and wrote the paper. AL, SS and DS all reviewed the design of the study, as well as the final manuscript. Funding: Engineering and Physical Sciences Research Council, 10.13039/501100000266, EP/G037574/1. Competing interests: SH received a small payment for assisting Adherium (Smartinhaler) at their exhibition stands at two UK conferences. Ethics approval: University of Nottingham Engineering Ethics Board. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: All data are provided in full in the results section of this paper.

Background Solitary pulmonary nodule (SPN) is defined as a discrete well-defined intraparenchymal lesion <3 cm. SPNs are present in 20–50% of individuals considered to be at high risk for lung cancer and present an important diagnostic problem. A small proportion of patients with a SPN will have early stage lung cancer with a high 5-year survival rate following surgical resection. In the National Lung Screening Trial (NLST) an SPN was identified in 24% of the 53 454 asymptomatic current or former smokers with a 30 pack year history and of these >95% were benign.

mall proportion of patients with a SPN will have early stage lung cancer with a high 5-year survival rate following surgical resection. In the National Lung Screening Trial (NLST) an SPN was identified in 24% of the 53 454 asymptomatic current or former smokers with a 30 pack year history and of these >95% were benign. CT is the first-line investigation for SPN. Fleischner Society guidelines have been widely used for many years although recently published British Thoracic Society (BTS) guidelines using the prediction scores and volumetry are now being more widely used in the UK. In general, nodules of 8 mm or less in diameter are assessed for interval growth using serial CT surveillance. For nodules >8 mm a number of imaging options are available determined by the patient's performance status and the probability for malignancy. For patients that are potentially suitable for treatment with curative intent, positron emission tomography-CT (PET-CT) is the investigation of choice. PET-CT provides both anatomical (CT) and functional information (PET) following intravenous administration of small quantities of the radioactive glucose analogue 18-flurodeoxyglucose (18FDG). 18FDG-PET characterises SPNs on the basis of increased glucose metabolism in malignant lesions. However, false positives can occur in certain benign conditions including granulomatous disease and infective/inflammatory lesions. False-negative scans are seen in tumours with low glucose metabolism including lung adenocarcinoma with a lepidic or mucinous component and carcinoid tumours with no atypia. In order to minimise patient anxiety, risk of overdiagnosis, cumulative radiation burden and cost of performing multiple tests a streamlined diagnostic pathway that involves the least number of investigations for accurate nodule characterisation is essential. In Paul Barnett's study looking at the management of patients with SPNs for the Veterans Affairs Positron Emission Tomography Imaging Cooperative study group, the average US Medicare expenditure for clinical management of an incidental SPN was $50 233 (£30 363) when the nodule was malignant and $22 461 (£13 577) when benign.

l Barnett's study looking at the management of patients with SPNs for the Veterans Affairs Positron Emission Tomography Imaging Cooperative study group, the average US Medicare expenditure for clinical management of an incidental SPN was $50 233 (£30 363) when the nodule was malignant and $22 461 (£13 577) when benign. Dynamic contrast-enhanced CT (DCE-CT) is an imaging technique that involves the acquisition of a series of CT images through the nodule before and at fixed time points after the injection of an iodinated contrast medium. A region of interest is placed within the nodule and the mean enhancement value (Hounsfield Units) is calculated at each time point. DCE-CT characterises SPNs on the basis of increased enhancement in malignant nodules reflecting the presence of tumour neovascularisation. Malignant nodules typically demonstrate increased contrast enhancement >15 HU compared with benign nodules. Pooled analysis from 10 DCE-CT studies (incorporating 1167 nodules) reported a sensitivity, specificity, PPV and NPV of 93%, 76%, 80% and 95%, respectively, for differentiating malignant from benign nodules.1 A single study that compared the relative cost-effectiveness of DCE-CT to conventional CT surveillance and PET-based strategies for nodule evaluation, demonstrated savings of up to £2000 per patient compared with routine nodule surveillance. Furthermore, a strategy whereby patients only underwent 18FDG-PET if DCE-CT suggested malignancy had similar effectiveness to 18FDG-PET alone but was consistently less expensive.2

e and PET-based strategies for nodule evaluation, demonstrated savings of up to £2000 per patient compared with routine nodule surveillance. Furthermore, a strategy whereby patients only underwent 18FDG-PET if DCE-CT suggested malignancy had similar effectiveness to 18FDG-PET alone but was consistently less expensive.2 To date, only three studies have directly compared the diagnostic performances of 18FDG-PET and DCE-CT in the same cohort of patients. However, pooled data from these three studies evaluating 217 SPNs demonstrated that 18FDG-PET and DCE-CT had sensitivities of 92% and 87% and specificities of 90% and 83%, respectively.3–5 As yet, no large comparative multicentre trial of DCE-CT as a standalone technique or in combination with 18FDG-PET or integrated PET-CT for nodule characterisation has been performed.

ing 217 SPNs demonstrated that 18FDG-PET and DCE-CT had sensitivities of 92% and 87% and specificities of 90% and 83%, respectively.3–5 As yet, no large comparative multicentre trial of DCE-CT as a standalone technique or in combination with 18FDG-PET or integrated PET-CT for nodule characterisation has been performed. Study design The SPUtNIk trial is a multicentre prospective cohort observational study designed to (1) assess the diagnostic performances of DCE-CT and 18FDG-PET-CT for the characterisation of SPNs in the National Health Service (NHS) setting and (2) use decision analytic modelling to assess the likely costs and health outcomes resulting from incorporation of DCE-CT into management strategies for patients with SPNs. Secondary objectives include assessment of the incremental value of incorporating the CT appearances of a SPN into the interpretation of integrated PET-CT examinations and assessing whether combining DCE-CT with 18FDG-PET-CT is more accurate and/or cost-effective for characterising SPNs than either test used alone or in series. The study consists of a cohort of 375 patients and the full protocol is available as an online supplementary material. 10.1136/bmjresp-2016-000156.supp1Supplementary data

Study design The SPUtNIk trial is a multicentre prospective cohort observational study designed to (1) assess the diagnostic performances of DCE-CT and 18FDG-PET-CT for the characterisation of SPNs in the National Health Service (NHS) setting and (2) use decision analytic modelling to assess the likely costs and health outcomes resulting from incorporation of DCE-CT into management strategies for patients with SPNs. Secondary objectives include assessment of the incremental value of incorporating the CT appearances of a SPN into the interpretation of integrated PET-CT examinations and assessing whether combining DCE-CT with 18FDG-PET-CT is more accurate and/or cost-effective for characterising SPNs than either test used alone or in series. The study consists of a cohort of 375 patients and the full protocol is available as an online supplementary material. 10.1136/bmjresp-2016-000156.supp1Supplementary data The primary inclusion criterion is the presence of a dominant indeterminate soft tissue SPN identified on CT that measures ≥8 and ≤30 mm on axial plane (with no ancillary evidence strongly indicative of malignancy) which is being considered for further evaluation with 18FDG-PET-CT. Potential participants will be identified either at local multidisciplinary meetings or at time of referral for investigation of a SPN. The study involves a single DCE-CT (radiation dose 25 mSv) being performed in addition to the patient's standard SPN management. The DCE-CT is performed within 21 days of the 18FDG-PET-CT examination. Exclusion criteria include history of malignancy within 2 years, confirmed aetiology of the SPN at the time of the qualifying CT, biopsy of the nodule before the DCE-CT, contraindication to potential surgical resection or radiotherapy and contraindication to any of the imaging investigations.

18FDG-PET-CT examination. Exclusion criteria include history of malignancy within 2 years, confirmed aetiology of the SPN at the time of the qualifying CT, biopsy of the nodule before the DCE-CT, contraindication to potential surgical resection or radiotherapy and contraindication to any of the imaging investigations. Site accreditation and quality assessment for 18FDG-PET-CT and DCE-CT will be performed using established procedures by the PET core laboratory at St Thomas' Hospital, London and Mount Vernon Hospital, London, respectively. The 18FDG-PET-CT and DCE-CT image acquisition and data analysis will follow a standardised protocol. The 18FDG-PET-CT images and attenuation correction CT images will initially be classified according to a five-point characterisation scale. Further quantitative analysis will consist of measurements of FDG uptake expressed as the maximum standardised uptake value. The diagnostic performance of 18FDG-PET will be assessed with and without incorporation of the CT appearances. The presence of incidental extrathoracic findings on PET-CT will also be recorded. DCE-CT interpretation will be performed by thoracic radiologists at each participating site. For both DCE-CT and PET-CT, central review of 10% of cases will be performed by an expert radiologist/nuclear medicine physician to ensure quality assurance.

of incidental extrathoracic findings on PET-CT will also be recorded. DCE-CT interpretation will be performed by thoracic radiologists at each participating site. For both DCE-CT and PET-CT, central review of 10% of cases will be performed by an expert radiologist/nuclear medicine physician to ensure quality assurance. Following PET-CT and DCE-CT examinations subsequent SPN management is determined by the local specialist lung multidisciplinary meeting. The reference standard will comprise pathological and/or imaging follow-up data at 24 months. All patients without definitive pathological findings will undergo repeat CT examinations of the chest at 3, 9 and 24 months (with or without biopsy) in accordance with Fleischner guidelines. Clinical information, including information relating to costs will be extracted using a standardised data collection form to inform the economic analysis. A subanalysis using the BTS guidelines algorithm will also be undertaken. In parallel with the main study, a quantitative substudy (IPCARD-SPN: validated in a population of general practitioner referred chest radiograph attendees) will aim to (1) identify the positive and negative predictive values of symptoms that distinguish between malignant and non-malignant SPN and (2) ascertain whether or not the inclusion of symptoms found to distinguish between malignant and non-malignant nodules increases the diagnostic value of DCE-CT and 18FDG-PET-CT.

will aim to (1) identify the positive and negative predictive values of symptoms that distinguish between malignant and non-malignant SPN and (2) ascertain whether or not the inclusion of symptoms found to distinguish between malignant and non-malignant nodules increases the diagnostic value of DCE-CT and 18FDG-PET-CT. Trial outcome measures Primary outcome measures will include diagnostic test characteristics (sensitivity, specificity, accuracy) for 18FDG-PET-CT and DCE-CT in relation to a subsequent clinical diagnosis of lung cancer. The outcome measures used in the economic model will include accuracy, estimated life expectancy and quality-adjusted life years (QALYs). Costs will be estimated from an NHS perspective. Incremental cost-effectiveness ratios will compare management strategies with DCE-CT to strategies without DCE-CT, where DCE-CT is expected to cost less than half that of 18FDG-PET-CT. Secondary outcome measures will include diagnostic test characteristics for 18FDG-PET-CT with incorporation of CT appearances and combined DCE-CT/18FDG-PET-CT. The incidence of incidental extrathoracic findings on 18FDG-PET/CT, subsequent investigations and costs will also be determined.

Trial outcome measures Primary outcome measures will include diagnostic test characteristics (sensitivity, specificity, accuracy) for 18FDG-PET-CT and DCE-CT in relation to a subsequent clinical diagnosis of lung cancer. The outcome measures used in the economic model will include accuracy, estimated life expectancy and quality-adjusted life years (QALYs). Costs will be estimated from an NHS perspective. Incremental cost-effectiveness ratios will compare management strategies with DCE-CT to strategies without DCE-CT, where DCE-CT is expected to cost less than half that of 18FDG-PET-CT. Secondary outcome measures will include diagnostic test characteristics for 18FDG-PET-CT with incorporation of CT appearances and combined DCE-CT/18FDG-PET-CT. The incidence of incidental extrathoracic findings on 18FDG-PET/CT, subsequent investigations and costs will also be determined. Recruitment will be terminated when 375 patients have been recruited and undergone DCE-CT. The study has two possible end points for each patient: either the diagnosis of lung cancer via biopsy or a diagnosis of benign or non-lung cancer via either biopsy or failure of the imaged nodule to progress (increase in size) during the 2-year follow-up period. The end of the study will be reached when the last study patient reaches either of these two end points or withdraws full consent for continuing in the study.

or a diagnosis of benign or non-lung cancer via either biopsy or failure of the imaged nodule to progress (increase in size) during the 2-year follow-up period. The end of the study will be reached when the last study patient reaches either of these two end points or withdraws full consent for continuing in the study. Conclusions With the potential adoption of a CT-based lung cancer screening programme in the UK, the number of patients with a SPN requiring further investigation could increase substantially. Furthermore, SPNs are a common finding on CT examinations undertaken for diagnostic or staging purposes. Novel cost-effective approaches to the assessment of SPNs will be of value to the NHS. This study will provide accurate data on the diagnostic performances of DCE-CT and 18FDG-PET/CT in the NHS for the characterisation of SPNs: the decision analytic modelling will assess the likely costs and health outcomes resulting from incorporation of DCE-CT into management strategies for patients with SPNs. The trial is funded by the NIHR HTA Programme (grant no: 09/22/117) and is being run by Southampton Clinical Trials Unit, directed by Professor Gareth Griffiths and part funded by CRUK. NRQ and RCR are part funded by the Cambridge Biomedical Research Centre and the Cancer Research Network: Eastern.

This study will provide accurate data on the diagnostic performances of DCE-CT and 18FDG-PET/CT in the NHS for the characterisation of SPNs: the decision analytic modelling will assess the likely costs and health outcomes resulting from incorporation of DCE-CT into management strategies for patients with SPNs. The trial is funded by the NIHR HTA Programme (grant no: 09/22/117) and is being run by Southampton Clinical Trials Unit, directed by Professor Gareth Griffiths and part funded by CRUK. NRQ and RCR are part funded by the Cambridge Biomedical Research Centre and the Cancer Research Network: Eastern. Collaborators: The SPUtNIk Group: KAM, FJG, SG, SH, JJ, Andrew Clegg, NRQ, RCR, Ashley Groves, FVG, FWP, WLW, Richard Riley, LB, PM, CM, CP, AB, Andrea Lodge, John Buscombe, Theresa Green, Amanda Stone, Neal Navani, Robert Shortman, Gabriella Azzopardi, JS, RE, AS, Lucy Pike, DS, SD, AJF, Sarah Doffman, Janice Bush, Jane Lyttle, Kenneth Jacob, SH, Joris van der Horst, Joseph Sarvesvaran, Barbara McLaren, Lesley Gomersall, Lutfi Kurban, Ravi Sharma, Kathleen Collie, KK, Steve O'Hickey, Jayne Tyler, Sue King, John O'Brien, Rajiv Srivastava, Hugh Lloyd-Jones, Sandra Beech, Matthew Callister, Andrew Scarsbrook, Victoria Ashford-Turner, PC, Elaine Smith, Susan Mbale, Nick Adams, Osei Kankam, Gail Pottinger.

Barbara McLaren, Lesley Gomersall, Lutfi Kurban, Ravi Sharma, Kathleen Collie, KK, Steve O'Hickey, Jayne Tyler, Sue King, John O'Brien, Rajiv Srivastava, Hugh Lloyd-Jones, Sandra Beech, Matthew Callister, Andrew Scarsbrook, Victoria Ashford-Turner, PC, Elaine Smith, Susan Mbale, Nick Adams, Osei Kankam, Gail Pottinger. Contributors: KAM, FJG, SG, SH, JJ, Andrew Clegg, AMG, Tony Frew, FVG, FWP, WLW and Richard Riley designed the study and obtained funding. Andrea Lodge, Theresa Green, Amanda Stone, Rob Shortman, Gabriella Azzopardi, Kathleen Collie, Barbara Mclaren, June Innes, Kenneth Jacob, Jane Lyttle, Victoria Ashford-Turner, Susan Mbale, Jayne Tyler, Hugh Lloyd-Jones, Sandra Beech and Gail Pottinger were involved in data acquisition. KAM, FJG, SH, SG, JJ, NRQ, RCR, Andrew Clegg, AMG, FVG, FWP and RR were involved in data analysis and interpretation. RCR, FJG, NRQ, KAM, SG, SH, JJ, JM were involved in manuscript draft and revision for intellectual content. All members of the SPUtNIk Group were involved in approval of the final version. FJG is the guarantor of the study. Funding: Health Technology Assessment Programme (09/22/117). Competing interests: None declared. Patient consent: Obtained. Ethics approval: South West-Exeter. Provenance and peer review: Not commissioned; internally peer reviewed. Data sharing statement: Details of the methods used are freely available to any researcher wishing to use them.

Key messages Integrated respiratory clinics delivering joint care by specialists and primary care teams can improve clinical outcomes and reduce care costs for patients with airways disease. Introduction Long-term respiratory conditions in the UK are very common; over 6 million people live with the two most common conditions, asthma and chronic obstructive pulmonary disease (COPD).1 Treating respiratory diseases costs the UK National Health Service (NHS) an estimated £4.7 billion annually.2 Respiratory disease is the third biggest cause of death in the UK with ∼800 000 patients dying annually.2 A high proportion of these costs are generated by a relatively small group of patients with more severe disease or with complex problems that include multimorbidity, at-risk behaviours and socioeconomic disadvantage.3 4 These patients often struggle to engage with the structured, proactive care approach to chronic disease management advocated for asthma and COPD, resulting in repeated emergency healthcare use of primary and secondary care.4 An integrated approach to the management of complex patients, incorporating specialist and primary care teams' expertise, may be effective in improving outcomes for such high-risk patients. However, the evidence is mixed5–8 and there is a need for evaluations of models of integrated care in routine, ‘real-world’ clinical settings.

d approach to the management of complex patients, incorporating specialist and primary care teams' expertise, may be effective in improving outcomes for such high-risk patients. However, the evidence is mixed5–8 and there is a need for evaluations of models of integrated care in routine, ‘real-world’ clinical settings. Over the past two decades, there has been a shift in the locus of care for the majority of patients with chronic respiratory diseases in the UK towards the community.9 Respiratory diseases are among the most common causes of primary care consultations, accounting for 24 million consultations annually.10 Increasing numbers of complex respiratory patients are being managed in the primary care setting by generalist teams, with a focus on avoidance of admissions to hospital.9 Specialist secondary care is restricted to those patients admitted to hospital in a crisis or referred because of uncontrolled disease.1 10

ually.10 Increasing numbers of complex respiratory patients are being managed in the primary care setting by generalist teams, with a focus on avoidance of admissions to hospital.9 Specialist secondary care is restricted to those patients admitted to hospital in a crisis or referred because of uncontrolled disease.1 10 There is evidence of significant and unwarranted variability in the standards of respiratory management in the primary and secondary care sectors. Marked variations in outcomes for patients with respiratory disease have also been shown, regionally and between individual general practitioner (GP) practices.1 There is evidence linking the quality of care provided in general practice with unplanned admissions to secondary care,11 and decreased admission rates have been reported in a number of long-term conditions (including COPD and asthma) where GPs were financially incentivised to provide high-quality care.12 Moreover, higher levels of professional education, nurse staffing and clinical recording in primary care are all associated with an improvement in the quality of clinical care for patients with COPD.13 However, a ‘skills gap’ may exist in some primary care settings, where GPs and other healthcare professionals lack advanced training in the management of these common conditions, particularly in the case of patients with multimorbidity, uncertain diagnosis or complex problems.14 Patients with more severe or complicated disease may receive suboptimal care, which may in turn lead to poor outcomes.14 Such patients may fail to reach a specialist assessment that could potentially improve outcomes, either because they are not offered referral to a specialist clinic or because they decline going to a hospital clinic for such an assessment. Therefore, a community-based integrated care approach, harnessing specialist skills and the overall holistic perspective of the generalist primary care teams, is a promising and attractive solution which is being explored by newly commissioned services.

se they decline going to a hospital clinic for such an assessment. Therefore, a community-based integrated care approach, harnessing specialist skills and the overall holistic perspective of the generalist primary care teams, is a promising and attractive solution which is being explored by newly commissioned services. Potential benefits of joint specialist–generalist clinics in the community include not only improvement in quality of care for each of the individual patients seen, but also on-site education for the primary care teams, leaving a legacy of improved skills and greater confidence in managing complex disease.14 Such clinics may potentially increase patient and staff satisfaction, reduce secondary care use and consequently reduce the financial burden of respiratory disease on the local health economy.15

ucation for the primary care teams, leaving a legacy of improved skills and greater confidence in managing complex disease.14 Such clinics may potentially increase patient and staff satisfaction, reduce secondary care use and consequently reduce the financial burden of respiratory disease on the local health economy.15 Local context The UK region of Wessex is situated on the south coast of England and represents a diverse population of around 2.8 million people, ranging from inner city deprivation to remote rural populations. Local clinical audit data have demonstrated marked variation between local regional administrative groups (eg, a 1.9-fold difference in COPD admission rates and a 2.8-fold difference in asthma admission rates) and between individual GP practices (eg, a 4.7-fold difference in COPD admission rates). Improving respiratory care is an agreed local priority and the basis for newly commissioned integrated respiratory services. West Hampshire Clinical Commissioning Group (CCG), the Wessex Academic Health Sciences Network (WAHSN) and the Wessex Collaboration for Leadership in Applied Health Research and Care (CLAHRC) collaborated to prospectively evaluate a service pilot of an integrated model for managing complex or poorly controlled asthma and COPD across the organisational silos of primary and secondary care with a view to subsequent regional roll-out.

sex Collaboration for Leadership in Applied Health Research and Care (CLAHRC) collaborated to prospectively evaluate a service pilot of an integrated model for managing complex or poorly controlled asthma and COPD across the organisational silos of primary and secondary care with a view to subsequent regional roll-out. Methods Setting Two practices were enrolled in the pilot evaluation, one rural and one urban. Patients with poorly controlled COPD and asthma were identified by searches of the practice clinical computer systems, using routinely recorded clinical data. The pilot was registered with the WHCCG as a Quality Improvement project and consultation with the Health Research Authority confirmed the project to qualify as a service evaluation.

controlled COPD and asthma were identified by searches of the practice clinical computer systems, using routinely recorded clinical data. The pilot was registered with the WHCCG as a Quality Improvement project and consultation with the Health Research Authority confirmed the project to qualify as a service evaluation. Criteria for patient inclusion and joint-clinic arrangements COPD patients with poor outcomes were identified using four parameters, those measured in the DOSE Index, a multidimensional assessment tool used to predict outcomes in COPD16 whose items are routinely collected in primary care in the UK (Dyspnoea: MRC Breathlessness Score; Obstruction: % predicted FEV1; smoking status; exacerbations). Patients with poorly controlled asthma were identified based on clinical consensus and on at-risk factors identified by the UK National Review of Asthma Deaths (NRAD)17 (shown in box 1). Patients were identified through the use of electronic audit and case-finding tools and a manual review of routine patient records by respiratory nurse specialists (RNS). Clinically pertinent information on each patient was prepared in advance of the clinic including the number of exacerbations, hospital admissions and inhaler usage in the 12 months prior to the appointment. Additionally, clinical staff from the GP practice were able to include respiratory patients they considered ‘at risk’ or for whom they required further advice or considered complex. Box 1 Inclusion and exclusion criteria for identifying patients for clinics Inclusion criteria for identifying patients with complex asthma High use of short-acting β2-agonist/inhaled corticosteroid combinations (<12 in 12 months).

hey considered ‘at risk’ or for whom they required further advice or considered complex. Box 1 Inclusion and exclusion criteria for identifying patients for clinics Inclusion criteria for identifying patients with complex asthma High use of short-acting β2-agonist/inhaled corticosteroid combinations (<12 in 12 months). High SABA use (>12 in 12 months). High SABA use, low ICS use (>12 SABA and <12 ICS in 12 months). High oral corticosteroid use (>2 prescriptions for prednisolone in 12 months). High doses of ICS monotherapy (≥800 mcg of budesonide (or equivalent)). Inclusion criteria for identifying patients with complex chronic obstructive pulmonary disease Exacerbations within the last 12 months. Acute hospital admissions related to respiratory within the last 12 months. MRC≥3. FEV1<50% predicted. At the GP/practice nurse's discretion. Exclusion criteria Pregnancy. Housebound. Under secondary care for respiratory issue (or other related, eg, cardiac for breathlessness). Active cancer. At the GP's discretion.

Inclusion criteria for identifying patients with complex chronic obstructive pulmonary disease Exacerbations within the last 12 months. Acute hospital admissions related to respiratory within the last 12 months. MRC≥3. FEV1<50% predicted. At the GP/practice nurse's discretion. Exclusion criteria Pregnancy. Housebound. Under secondary care for respiratory issue (or other related, eg, cardiac for breathlessness). Active cancer. At the GP's discretion. Patients identified were invited to attend for a joint specialist-practice team clinic in the GP premises, including diagnostic review and treatment monitoring from the joint specialist–generalist nursing teams and a clinical assessment from the specialist–generalist medical teams. The practices agreed to provide administrative support which included facilitating electronic searches on their databases, sending out invitations to patients, booking patients into clinics and reminding patients of their appointment date and time, much of which is considered routine clinical practice. There were no financial incentives for the practices. Clinical and information governance arrangements were agreed: the specialist respiratory team was endorsed by the CCG as a guest of the surgeries and the clinical responsibility for the patients remained with the GP. No patient identifiable information was removed from the practice premises.

centives for the practices. Clinical and information governance arrangements were agreed: the specialist respiratory team was endorsed by the CCG as a guest of the surgeries and the clinical responsibility for the patients remained with the GP. No patient identifiable information was removed from the practice premises. The intervention: collaborative clinics The clinical intervention consisted of a collaborative clinic at the patients' own GP surgery; clinical interventions were driven by clinician–patient interaction rather than predefined protocol. This comprised a 60 min appointment including a 20 min initial nurse assessment, including spirometry and other near-patient diagnostic and monitoring tests appropriate, by a RNS and a practice nurse (PN) or nurse practitioner (NP), followed by a 20 min joint assessment by a respiratory physician (RP) working alongside a practice clinician (GP and/or PN/NP) and a 20 min follow-up education session by an RNS. A personalised disease management and action plan was agreed jointly between the RP, practice clinician and patient. Practical tasks, for example, prescriptions, were carried out by the practice clinician. Relatives and carers were actively encouraged to attend with the patient. The GP retained clinical control and responsibility for the patient.

ement and action plan was agreed jointly between the RP, practice clinician and patient. Practical tasks, for example, prescriptions, were carried out by the practice clinician. Relatives and carers were actively encouraged to attend with the patient. The GP retained clinical control and responsibility for the patient. Outcome measures Longitudinal follow-up data Data on exacerbations, medication usage, emergency department (ED)/hospital admissions and GP practice visits were collected from the standard practice clinical electronic records by an RNS 9 months postclinic (February–November 2015). Data were collected through a virtual review of routine medical records for 53 patients who still remained on the GP registers. This was compared to a seasonally matched 9-month period (February–November 2014) prior to clinic. Cost-effectiveness analysis A cost-effectiveness analysis (CEA) of this project was undertaken: the intervention was the clinic with the outcomes measured in the longitudinal follow-up (postclinic) data; the comparator was the preclinic data. The CEA was performed using standard NHS templates for costing including cost of medication (short-acting β2-agonist, SABA and inhaled corticosteroid, ICS),18 cost of exacerbation (prescription of antibiotic and oral corticosteroids),18 cost of scheduled and unscheduled GP and PN visits,19 cost of intervention19 (which included clinician and administration time and consumables) and cost of hospital admissions.20

ort-acting β2-agonist, SABA and inhaled corticosteroid, ICS),18 cost of exacerbation (prescription of antibiotic and oral corticosteroids),18 cost of scheduled and unscheduled GP and PN visits,19 cost of intervention19 (which included clinician and administration time and consumables) and cost of hospital admissions.20 Statistical analysis: paired non-parametric test Descriptive statistics have been used in the analysis of the outcome measures. Non-parametric tests (paired Wilcoxon signed-rank tests) have been calculated on the non-parametric data to compare the preclinic and postclinic data after outcome variables display non-normality features; the significance threshold was 0.05. Patient and practice feedback Patients and staff were able to provide unstructured written feedback on the service at the end of their appointment. Feedback was not formally requested from the integrated care team, however a debriefing and education session was held following the completion of the clinics which provided an outlet for feedback from the team members.

f were able to provide unstructured written feedback on the service at the end of their appointment. Feedback was not formally requested from the integrated care team, however a debriefing and education session was held following the completion of the clinics which provided an outlet for feedback from the team members. Results GP practice demographics Two GP practices were included in this pilot project. The first was a large, rural market town practice with a practice population of 12 598, staffed by 10 GPs, 3 NPs, 2 PNs and 3 healthcare assistants (HCAs). The practice population deprivation is in the second least deprived decile in the UK; the COPD and asthma prevalence at baseline was 1.8% and 6.7%, respectively (compared to a Wessex average of 1.7% for COPD and 6.3% for asthma). The second was a small, suburban practice with a practice population of 3604 staffed by 1 GP, 1 NP, 2 PNs and 1 HCA. The practice population deprivation is in the fourth least deprived decile in the UK; the baseline COPD and asthma prevalence was 1.7% and 4.6%, respectively. Patient demographics Eighty-two patients were invited to the intervention clinics, shown in figure 1. Fifty-six of these patients responded and booked an appointment. Of all patients who booked an appointment, 98.2% (55) attended the clinics. Demographics for these patients are shown in table 1. Table 1 Demographics of patients attending clinics

Patient demographics Eighty-two patients were invited to the intervention clinics, shown in figure 1. Fifty-six of these patients responded and booked an appointment. Of all patients who booked an appointment, 98.2% (55) attended the clinics. Demographics for these patients are shown in table 1. Table 1 Demographics of patients attending clinics N (%) Patients attended 55 Mean age 60 Age range 19–82 Females 33 (60%) Patients with follow-up data 53 Baseline diagnosis 55 Asthma 36 (65%) COPD 7 (13%) Asthma and COPD overlap syndrome (ACOS) 9 (16%) ‘Other respiratory’ 1 (2%) No respiratory diagnosis 2 (4%) Smoking data 53 Current smoker 11 (21%) Ex smoker 22 (42%) Never smoker 20 (38%) Body mass index (BMI) data 53 BMI <20 0 BMI 20–24 13 (25% BMI 25–29 20 (38%) BMI 30–39 18 (24%) BMI ≥40 2 (4%) Figure 1 Patient identification, attendance and diagnosis data. ACOS, Asthma and COPD overlap syndrome; COPD, chronic obstructive pulmonary disease. Clinic data Diagnostic review and accuracy Of the 55 patients seen in the complex clinics, 13 (23.6%) received a change in diagnosis after review by the clinical project team (figure 2A). Reasons for this change in diagnosis included inaccurate initial diagnosis; condition changing over time, for example, asthma to Asthma and COPD overlap syndrome (ACOS); a correct diagnosis had been made previously but not been recorded in the electronic record; the necessary differential diagnostic tests had not been performed to identify the condition.

included inaccurate initial diagnosis; condition changing over time, for example, asthma to Asthma and COPD overlap syndrome (ACOS); a correct diagnosis had been made previously but not been recorded in the electronic record; the necessary differential diagnostic tests had not been performed to identify the condition. Figure 2 Impacts of clinical review. (A) Summary of changes in clinical diagnosis after clinic review. (B) Summary of inhaled short-acting bronchodilator and inhaled corticosteroid prescription in 9 months prior to and after clinical review. ACOS, Asthma and COPD overlap syndrome; COPD, chronic obstructive pulmonary disease; OSA, obstructive sleep apnoea. Onward referrals Of the 55 patients seen, only 4 (7.3%) were referred on to secondary care for further investigations. Thirty-seven (67.3%) required further appointments with their GP/PN for follow-up care such as medication reviews or inhaler technique checks. The remaining 14 (25.4%) did not require any follow-up and returned to the usual routine care with their GP/PN.

%) were referred on to secondary care for further investigations. Thirty-seven (67.3%) required further appointments with their GP/PN for follow-up care such as medication reviews or inhaler technique checks. The remaining 14 (25.4%) did not require any follow-up and returned to the usual routine care with their GP/PN. Clinical outcome: longitudinal follow-up data (postclinic). Medication usage SABA inhalers The number of prescribed SABA inhalers over a 9-month period (February–November 2014) prior to the joint clinic ranged from 0 to 30, with a median of 5. Following the joint clinic, the number of prescribed SABA inhalers ranged from 0 to 26, with a median of 3 over a 9-month period (February–November 2015). The total number of SABA prescribed over the 9-month periods reduced by 33.3% (427 to 285), p<0.001 (see figure 2B). The frequency of prescriptions is shown in figure 3A. Figure 3 Change in frequency of SABA and ICS prescription after clinical intervention. (A) Frequency of SABA prescriptions preclinic and postclinic (n=53). (B) Frequency of ICS inhaler prescriptions for patients with asthma preclinic and postclinic (n=28).

Clinical outcome: longitudinal follow-up data (postclinic). Medication usage SABA inhalers The number of prescribed SABA inhalers over a 9-month period (February–November 2014) prior to the joint clinic ranged from 0 to 30, with a median of 5. Following the joint clinic, the number of prescribed SABA inhalers ranged from 0 to 26, with a median of 3 over a 9-month period (February–November 2015). The total number of SABA prescribed over the 9-month periods reduced by 33.3% (427 to 285), p<0.001 (see figure 2B). The frequency of prescriptions is shown in figure 3A. Figure 3 Change in frequency of SABA and ICS prescription after clinical intervention. (A) Frequency of SABA prescriptions preclinic and postclinic (n=53). (B) Frequency of ICS inhaler prescriptions for patients with asthma preclinic and postclinic (n=28). ICS inhalers The total number of ICS-containing inhalers prescribed for all patients increased by 23.3% (295–364), p<0.05 in the 9 months following the specialist clinic compared to the previous seasonally matched 9-month period. A subanalysis of the 28 patients with asthma, in whom ICS adherence is particularly important, was performed, shown in figure 3B. The number of ICS inhalers prescribed over a 9-month period prior to the joint clinic ranged from 0 to 24, with a median of 5. The number of ICS inhalers prescribed over a 9-month period following the joint clinic ranged from 0 to 18, with a median of 6.5.

e is particularly important, was performed, shown in figure 3B. The number of ICS inhalers prescribed over a 9-month period prior to the joint clinic ranged from 0 to 24, with a median of 5. The number of ICS inhalers prescribed over a 9-month period following the joint clinic ranged from 0 to 18, with a median of 6.5. Healthcare usage In the seasonally matched 9-month period preclinic and postclinic, respiratory exacerbations (defined as a non-scheduled contact caused by an acute deterioration in respiratory symptoms resulting in a prescription of oral steroids and/or antibiotics, with a continuation of symptoms without improvement requiring multiple courses of treatment classed as one exacerbation) reduced by 67.6% (from 37 to 12), p<0.01 (figure 4). Non-elective respiratory GP visits (defined as an urgent, unplanned respiratory appointment with the GP) reduced by 78.5% (from 42 to 9), p≤0.01. Elective respiratory GP visits (defined as planned respiratory appointments with the GP for reasons such as medication reviews and clinical review following resolution of an exacerbation) reduced by 28.6% (from 28 to 20), p<0.05. Respiratory PN visits reduced by 47.7% (from 65 to 34), p<0.01. In total, the number of visits to the GP surgery for respiratory issues was reduced by just over half (from 135 to 63), p<0.01. Figure 4 Number of exacerbations, emergency hospital admissions and primary care visits before and after clinics. ED, emergency department; GP, general practitioner.

Healthcare usage In the seasonally matched 9-month period preclinic and postclinic, respiratory exacerbations (defined as a non-scheduled contact caused by an acute deterioration in respiratory symptoms resulting in a prescription of oral steroids and/or antibiotics, with a continuation of symptoms without improvement requiring multiple courses of treatment classed as one exacerbation) reduced by 67.6% (from 37 to 12), p<0.01 (figure 4). Non-elective respiratory GP visits (defined as an urgent, unplanned respiratory appointment with the GP) reduced by 78.5% (from 42 to 9), p≤0.01. Elective respiratory GP visits (defined as planned respiratory appointments with the GP for reasons such as medication reviews and clinical review following resolution of an exacerbation) reduced by 28.6% (from 28 to 20), p<0.05. Respiratory PN visits reduced by 47.7% (from 65 to 34), p<0.01. In total, the number of visits to the GP surgery for respiratory issues was reduced by just over half (from 135 to 63), p<0.01. Figure 4 Number of exacerbations, emergency hospital admissions and primary care visits before and after clinics. ED, emergency department; GP, general practitioner. In the seasonally matched 9-month period preclinic and postclinic, the number of admissions or ED attendances for respiratory issues reduced from 3 to 0, p<0.05. At the time of the follow-up review in November 2015, none of the patients who attended the specialist clinics had died.

Figure 4 Number of exacerbations, emergency hospital admissions and primary care visits before and after clinics. ED, emergency department; GP, general practitioner. In the seasonally matched 9-month period preclinic and postclinic, the number of admissions or ED attendances for respiratory issues reduced from 3 to 0, p<0.05. At the time of the follow-up review in November 2015, none of the patients who attended the specialist clinics had died. Health economic evaluation Respiratory-related costs per patient over a 9-month period decrease from £458.11 to £226.25 following the intervention, a reduction of £231.86 per patient, equating to an annualised saving of £309.15 (table 2). The overall cost of the intervention for the 55 patients seen in two practices was £16 325 equating to £296.82 per patient. The incremental cost-effectiveness of the intervention versus no intervention is £142.89 per exacerbation avoided (considered as cost per patient in the timeframe of 9 months). Table 2 Health economic evaluation of clinics: cost per patient over a 9-month period Cost preclinic Cost postclinic GP visit £85.27 £35.33 PN visit £52.00 £27.20 SABA £29.37 £19.60 ICS £116.16 £143.32 Exacerbations £2.44 £0.79 Hospital admissions £172.88 £0.00 Total cost £458.11 £226.25 GP, general practitioner; PN, practice nurse; SABA, short-acting β2-agonist. Patient feedback Of the 55 patients who attended, 50 provided unstructured written feedback on the service. All described attending the clinics as a positive experience in terms of patient experience of the clinic and interventions made.

Cost preclinic Cost postclinic GP visit £85.27 £35.33 PN visit £52.00 £27.20 SABA £29.37 £19.60 ICS £116.16 £143.32 Exacerbations £2.44 £0.79 Hospital admissions £172.88 £0.00 Total cost £458.11 £226.25 GP, general practitioner; PN, practice nurse; SABA, short-acting β2-agonist. Patient feedback Of the 55 patients who attended, 50 provided unstructured written feedback on the service. All described attending the clinics as a positive experience in terms of patient experience of the clinic and interventions made. Feedback from practices The informal feedback received from the integrated team members was also extremely positive. Specialist and generalist clinicians found it a useful learning experience with an immediate impact on practice. For example, GPs and PNs stated that they approached inhaler technique training differently after the clinic. RPs mentioned changing the way in which they disseminated information to primary care to include more detailed information.

clinicians found it a useful learning experience with an immediate impact on practice. For example, GPs and PNs stated that they approached inhaler technique training differently after the clinic. RPs mentioned changing the way in which they disseminated information to primary care to include more detailed information. Discussion Main findings We present findings from a prospective evaluation of a local service development consisting of an integrated model of care aiming to improve outcomes for complex respiratory patients in the community. The model involved identification of patients with poor outcomes and their review in joint clinics delivered collaboratively by visiting specialist teams and the host primary care teams. This initiative was acceptable to patients and healthcare professionals, including members of the integrated care team. The project demonstrated that informatics approaches to identification of patients with poorly controlled respiratory disease and follow-up is possible and may be conducted using electronic searches of routine clinical data from medical records. Moreover, this intervention can improve both markers of efficient medication use such as improved adherence to preventer inhaled therapy and reduced use of rescue medication and in improved markers of clinical outcomes such as frequency of exacerbation and unscheduled primary care consultations. The health economics analysis demonstrates significant cost savings over usual respiratory care that may cover the additional costs of providing the service, although this will require confirmation in larger studies with a more rigorous health economic evaluation and longer follow-up periods.

rimary care consultations. The health economics analysis demonstrates significant cost savings over usual respiratory care that may cover the additional costs of providing the service, although this will require confirmation in larger studies with a more rigorous health economic evaluation and longer follow-up periods. The acceptability of the intervention to patients is not only demonstrated through the positive feedback but also by the high response and attendance levels. Some of the patients who attended had repeatedly failed to attend the practice for their routine asthma/COPD annual check-up, a risk factor for exacerbations and premature mortality.17 The high level of attendance to the joint respiratory clinics suggests that patients may be more motivated to attend specialist respiratory clinics located in GP surgeries rather than, conventionally, at their local hospital thus potentially providing a more acceptable, efficient and cost-effective service. However, several issues remain—more work is required to identify optimal mechanisms to identify patients with inadequate control and at greatest risk of poor outcome. Furthermore, work is required to address the real unmet need driven by frequent no-attenders to clinical review in routine primary care and in additional services such as this. This population similarly is not involved in current service evaluations and focused research in this group is needed.

st risk of poor outcome. Furthermore, work is required to address the real unmet need driven by frequent no-attenders to clinical review in routine primary care and in additional services such as this. This population similarly is not involved in current service evaluations and focused research in this group is needed. The intervention resulted in a change in primary diagnosis for almost a quarter of the patients who attended the clinic. Improving diagnostic accuracy relied particularly on the provision of quality assured spirometry and where needed additional near-patient tests (eg, fraction of exhaled nitric oxide testing). Accurate diagnosis is fundamental to the appropriate provision of preventive therapy,21 enabling appropriate management tailored to each condition, for example, ICS in asthma and pulmonary rehabilitation in COPD, and may help to avoid treatment that is of no benefit or harmful.22 The shift shown in prescription patterns after clinic review was to one more closely aligned with national guidelines for both COPD. The majority of patients had either a change in diagnosis and/or management made without the need for further input from secondary care, which demonstrates the practicability of this integrated care model in primary care.

tterns after clinic review was to one more closely aligned with national guidelines for both COPD. The majority of patients had either a change in diagnosis and/or management made without the need for further input from secondary care, which demonstrates the practicability of this integrated care model in primary care. Prevention of exacerbations is a key goal in improving control of COPD and asthma.23 24 Furthermore, reducing hospital admissions has potential cost savings for the NHS. Emergency hospital admissions for ambulatory care sensitive conditions (ACSC) cost the NHS £1.42 billion annually; 34% of these are for respiratory-related disease.25 The results from this project show a reduction in healthcare usage both in terms of exacerbations, hospital admissions and GP surgery attendance. A 68% reduction in exacerbations and 79% reduction in non-elective GP visits are particularly worthy of note and indicate better disease control. The health economic analysis demonstrated that there was a cost saving in terms of healthcare usage following the intervention. Further research with a longer follow-up period is needed to perform an analysis of whether the outcomes from the intervention, for example, reductions in healthcare usage, are sustainable and whether the model will have further cost savings long term as benefits continue to accrue. Further research will also require the collection of quality of life data in order to calculate the cost-effectiveness in terms of cost per quality-adjusted life year (QALY).

e, reductions in healthcare usage, are sustainable and whether the model will have further cost savings long term as benefits continue to accrue. Further research will also require the collection of quality of life data in order to calculate the cost-effectiveness in terms of cost per quality-adjusted life year (QALY). This integrated model allowed for shared learning interorganisationally, between primary and secondary care, and intraorganisationally, within the multidisciplinary team. Furthermore, this initiative placed a strong emphasis on the importance of patient and carer education. The emphasis on education for patients, carers and healthcare professionals was intended to leave a legacy of patients who are able to self-manage more effectively and healthcare professionals who are upskilled in the management of patients with respiratory disease and more technically assured in performing specialist respiratory assessments. Interpretation of findings in relation to previously published work This project has demonstrated that this integrated model of care has the potential to improve outcomes for patients with complex respiratory conditions. This is supported by evidence from various NHS organisations including NHS England26 and the King's Fund,14 which suggest that ensuring specialist support in the delivery of care outside hospital has the potential to improve patient experience and access to care.

tcomes for patients with complex respiratory conditions. This is supported by evidence from various NHS organisations including NHS England26 and the King's Fund,14 which suggest that ensuring specialist support in the delivery of care outside hospital has the potential to improve patient experience and access to care. Strengths and limitations The benefits of this intervention in terms of improving patient outcomes and experience, access to care and reduction in healthcare usage were significant and have the potential to be replicated across the NHS, but require replication in large, well-designed trials. Formal research in a controlled setting (eg, an appropriately powered cluster randomised trial, comparing outcomes in matched practices receiving and not receiving the outreach intervention) is needed. The intervention in this pilot was a complex and multifactorial one. Consequently, understanding which aspects of the process were key drivers to improvement in an individual is difficult to ascertain; this issue is exacerbated by the relatively small sample size and that patients with asthma and patients with COPD were reviewed. Future larger scale studies and more discrete interventions may be required to tease out relative benefits of diagnostic change from treatment alterations or improved adherence. A more pragmatic approach would be to improve the evidence base behind the improved delivery of guideline adherent care in an integrated team and hence to provide the rationale for commissioners to adopt this model at scale if proven to be cost-effective.

gnostic change from treatment alterations or improved adherence. A more pragmatic approach would be to improve the evidence base behind the improved delivery of guideline adherent care in an integrated team and hence to provide the rationale for commissioners to adopt this model at scale if proven to be cost-effective. The limited sample size in this pilot restricts the power to detect significant change in hospital admissions, a major driver to costs in respiratory care. However, a reduction was observed, which will again need further investigation in larger controlled trials. Implications for future research, policy and practice Future research in a controlled environment to prove a causal relationship between this integrated respiratory care model and an improvement in respiratory outcomes in a cluster randomised trial is required. If validated this model will then have the potential to be replicated across the NHS and improve respiratory outcomes for patients throughout the UK. Conclusions A range of evidence demonstrates that the current reactive model predicated on acute hospitals is unsustainable in the face of an ageing population, with increasingly complex chronic conditions. This service-development evaluation demonstrates that patients with respiratory disease at risk of suboptimal outcomes can proactively be identified for management by an integrated team in the community without the need for extensive, expensive secondary care technologies and warrants further evaluation at scale to determine its impact in other regions to fully determine health economic outcomes.

disease at risk of suboptimal outcomes can proactively be identified for management by an integrated team in the community without the need for extensive, expensive secondary care technologies and warrants further evaluation at scale to determine its impact in other regions to fully determine health economic outcomes. The authors are thankful to the Wessex AHSN, Wessex CLAHRC, West Hampshire CCG and the patients and staff from the two GP surgeries involved who gave their time and supported this project. The authors wish to highlight the support of Beverley Meeson and Kimm Lawson in facilitating the work within the CCG and Rachel Dominey for help with project management. Contributors: KG, KL, JL, CA, CR, MT and TMW conducted the clinics. TMW, MT, KL and AJC helped develop the clinical model. KG and KL collected and analysed the data and wrote the first draft of the paper. SXL provided the statistical analysis and RO conducted the health economic analysis. All authors contributed to subsequent drafts of the paper and approved the final draft. TMW is the guarantor. Funding: This paper presents independent research by the National Institute for Health Research Collaboration for Leadership in Applied Health Research and Care Wessex (NIHR CLAHRC Wessex). CLAHRC Wessex acknowledge the participation and resources of our partner organisations. Further details can be found at www.clahrc-wessex.nihr.ac.uk and http://wessexahsn.org.uk/programmes/1/respiratory.

te for Health Research Collaboration for Leadership in Applied Health Research and Care Wessex (NIHR CLAHRC Wessex). CLAHRC Wessex acknowledge the participation and resources of our partner organisations. Further details can be found at www.clahrc-wessex.nihr.ac.uk and http://wessexahsn.org.uk/programmes/1/respiratory. Disclaimer: The views and opinions expressed are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health. Competing interests: None declared. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: Access to data can be requested from the corresponding author. Ethics approval: The pilot was registered with the WHCCG as a Quality Improvement project and consultation with the Health Research Authority confirmed the project to qualify as a service evaluation.

Key messages Cough and sputum in COPD have a substantial impact on patients' health status, yet there are relatively few studies that have investigated the effect of bronchodilators on these symptoms. In this paper, we analyse data from three Phase III studies to elucidate the effect of aclidinium bromide on cough and sputum. The results suggest that in addition to improving lung function, LAMAs, such as aclidinium, can improve cough and sputum expectoration compared with placebo in patients with COPD. As cough and sputum impact negatively on overall patient wellbeing, controlling these symptoms may represent an important additional therapeutic benefit of this class of drugs. Introduction Chronic obstructive pulmonary disease (COPD) is characterised by persistent and progressive airflow limitation and an enhanced inflammatory response to noxious stimuli.1 The resulting lung injury leads to breathlessness and other characteristic symptoms of COPD, including cough and sputum.

As cough and sputum impact negatively on overall patient wellbeing, controlling these symptoms may represent an important additional therapeutic benefit of this class of drugs. Introduction Chronic obstructive pulmonary disease (COPD) is characterised by persistent and progressive airflow limitation and an enhanced inflammatory response to noxious stimuli.1 The resulting lung injury leads to breathlessness and other characteristic symptoms of COPD, including cough and sputum. In patients with COPD, chronic cough and sputum production are associated with lung-function decline,2 more frequent exacerbations and hospitalisations, and increased risk of death.3 4 Accumulation of mucus in small airways is also associated with disease progression,5 and a productive cough has been shown to be independently associated with increased mortality in smokers with mild-to-moderate airflow obstruction.6 Cough symptoms also impact adversely on the health status of patients with COPD to a similar degree to that observed in bronchiectasis, asthma and chronic cough.7 The importance of cough and sputum symptoms in defining a patient's overall well-being is reflected in the inclusion of these items in the COPD Assessment Test, a patient-reported outcomes tool designed to assess overall COPD-related health status.8 9

egree to that observed in bronchiectasis, asthma and chronic cough.7 The importance of cough and sputum symptoms in defining a patient's overall well-being is reflected in the inclusion of these items in the COPD Assessment Test, a patient-reported outcomes tool designed to assess overall COPD-related health status.8 9 Between disease exacerbations, when COPD is considered stable, there may still be marked daily variability in patients' perceptions of symptom severity. In a pan-European cross-sectional study in patients with COPD, cough and phlegm were reported to be most troublesome in the morning.10 However, in a recent observational study of COPD symptoms, despite overall night-time symptoms being less prevalent than in the morning and during the day, cough was still the most common symptom at night.11

s-sectional study in patients with COPD, cough and phlegm were reported to be most troublesome in the morning.10 However, in a recent observational study of COPD symptoms, despite overall night-time symptoms being less prevalent than in the morning and during the day, cough was still the most common symptom at night.11 Despite the evidence of a clear association between cough and adverse clinical outcomes, its significance in patients with COPD is often underappreciated.12 13 In addition, almost nothing is known about the effect of current first-line COPD treatments on symptoms of cough and sputum, and the need for studies to address this has been highlighted.12 Aclidinium bromide is a long-acting muscarinic antagonist (LAMA) that inhibits the action of acetylcholine at M3 receptors in the lungs, indirectly leading to airway smooth muscle relaxation. Aclidinium is approved as a maintenance bronchodilator treatment in patients with COPD.14–16 Several phase III studies have shown that aclidinium 400 µg twice daily improves lung function and symptoms in patients with moderate-to-severe airflow limitation.17–23 In this manuscript, we report our analysis of the data from three of these studies, ACCORD COPD I, ATTAIN and a 6-week active-comparator study, which was undertaken to determine the effect of the approved dose of aclidinium (400 µg twice daily metered dose; equivalent to aclidinium 322 μg delivered dose) on cough and sputum symptoms in patients with stable moderate-to-severe COPD. The three phase III studies reported here were selected on the basis that they had similar inclusion/exclusion criteria and included end points that assessed the efficacy of aclidinium 400 µg twice daily on cough and sputum symptoms. Four additional phase III studies of aclidinium did not record cough data so could not be included in this analysis.

ported here were selected on the basis that they had similar inclusion/exclusion criteria and included end points that assessed the efficacy of aclidinium 400 µg twice daily on cough and sputum symptoms. Four additional phase III studies of aclidinium did not record cough data so could not be included in this analysis. Methods In these analyses, only data from patients randomised to placebo, aclidinium 400 µg twice daily (the dose approved for use in patients with COPD) or tiotropium 18 µg once daily (also the approved dose) were evaluated. The purpose of this additional analysis was to assess the impact of aclidinium on cough and sputum symptoms across three clinical studies, including the relationship between symptoms and time of day. All end points were preplanned, with the exception of post hoc analyses assessing the correlation between Evaluating Respiratory Symptoms (E-RS; formerly known as EXAcerbations of Chronic pulmonary disease Tool) cough and sputum domain score and cough severity score in the active-comparator study and change from baseline in E-RS total and cough and sputum domain scores in patients who had ≥1 exacerbation event in the ATTAIN study.

ng Respiratory Symptoms (E-RS; formerly known as EXAcerbations of Chronic pulmonary disease Tool) cough and sputum domain score and cough severity score in the active-comparator study and change from baseline in E-RS total and cough and sputum domain scores in patients who had ≥1 exacerbation event in the ATTAIN study. Study design ACCORD COPD I (ClinicalTrials.gov identifier: NCT00891462) and ATTAIN (ClinicalTrials.gov identifier: NCT01001494) were multinational, randomised, double-blind, placebo-controlled phase III studies.17 18 Following screening and a 2-week run-in period, patients were randomised (1:1:1) to receive aclidinium 200 µg, aclidinium 400 µg (metered dose; equivalent to aclidinium 322 µg delivered dose) or placebo twice daily via the Genuair™/Pressair®i inhaler for 12 weeks in ACCORD COPD I and 24 weeks in ATTAIN. The third study was a randomised, double-blind, double-dummy, placebo-controlled and active-controlled phase IIIb study (ClinicalTrials.gov identifier: NCT01462929).19 Following a 2–3-week run-in period, patients were randomised (2:2:1) to receive aclidinium 400 µg twice daily (metered dose; equivalent to aclidinium 322 µg delivered dose), tiotropium 18 µg once daily in the morning via HandiHaler® or placebo for 6 weeks.

hase IIIb study (ClinicalTrials.gov identifier: NCT01462929).19 Following a 2–3-week run-in period, patients were randomised (2:2:1) to receive aclidinium 400 µg twice daily (metered dose; equivalent to aclidinium 322 µg delivered dose), tiotropium 18 µg once daily in the morning via HandiHaler® or placebo for 6 weeks. In all three studies, inhaled albuterol/salbutamol (108/100 µg/puff) was permitted as relief medication as long as it was discontinued 6 hours prior to study visits. Additional permitted medications included inhaled corticosteroids, oral or parenteral corticosteroids (≤10 mg/day of prednisone or 20 mg every other day), oral sustained-release theophyllines and oxygen therapy (<15 hours/day), provided that treatment was stable for ≥4 weeks before screening. Other long-acting bronchodilators and anticholinergic drugs were washed out prior to screening and were not allowed during the treatment periods. All studies were conducted in accordance with the Declaration of Helsinki, International Conference on Harmonisation/Good Clinical Practice Guidelines and local regulations. The protocols were approved by institutional review boards/independent ethics committees at each site, and all patients gave written informed consent.

studies were conducted in accordance with the Declaration of Helsinki, International Conference on Harmonisation/Good Clinical Practice Guidelines and local regulations. The protocols were approved by institutional review boards/independent ethics committees at each site, and all patients gave written informed consent. Study populations Detailed inclusion/exclusion criteria for the three studies have been reported previously.17–19 Briefly, each study enrolled male and female patients (≥40 years old) with a diagnosis of stable COPD and moderate-to-severe airflow obstruction (postbronchodilator forced expiratory volume in 1 s (FEV1) ≥30% and <80% of the predicted value and FEV1/forced vital capacity ratio <70%)1 who were current or former smokers with a smoking history of ≥10 pack-years. The presence of cough or sputum symptoms at baseline was not a specific inclusion criterion in any of the studies. Exclusion criteria included any respiratory tract infection or COPD exacerbation within 6 weeks prior to screening (3 months if exacerbation resulted in hospitalisation), any clinically relevant respiratory conditions, including a history or current diagnosis of asthma and a history of hypersensitivity to inhaled anticholinergics or other inhaled medications. Study assessments A summary of patient-reported outcome measures used to capture symptoms in each clinical trial and corresponding end points are shown in online supplementary table S1. Baseline values for all end points were calculated as the average scores over the 2–3-week screening period prior to randomisation.

Exclusion criteria included any respiratory tract infection or COPD exacerbation within 6 weeks prior to screening (3 months if exacerbation resulted in hospitalisation), any clinically relevant respiratory conditions, including a history or current diagnosis of asthma and a history of hypersensitivity to inhaled anticholinergics or other inhaled medications. Study assessments A summary of patient-reported outcome measures used to capture symptoms in each clinical trial and corresponding end points are shown in online supplementary table S1. Baseline values for all end points were calculated as the average scores over the 2–3-week screening period prior to randomisation. 10.1136/bmjresp-2016-000148.supp1supplementary tables

Study assessments A summary of patient-reported outcome measures used to capture symptoms in each clinical trial and corresponding end points are shown in online supplementary table S1. Baseline values for all end points were calculated as the average scores over the 2–3-week screening period prior to randomisation. 10.1136/bmjresp-2016-000148.supp1supplementary tables Daily symptoms In ATTAIN and the active-comparator study, daily respiratory symptoms were assessed using the E-RS algorithm.24–26 The EXACT is a 14-item electronic daily diary used to quantify and measure exacerbations of COPD. It is completed by patients at night with a recall period of ‘today’ and captures symptoms of COPD including cough and sputum production. The E-RS total score (range 0–40) is a derivative tool which uses the 11 EXACT items that relate specifically to respiratory symptoms, with higher scores indicating more severe symptoms; the E-RS cough and sputum domain score is the sum of the three EXACT items that relate specifically to cough and sputum symptoms (range 0–11). Responder criteria for E-RS total score and E-RS cough and sputum domain scores have been proposed as a change of ≥−2.0 units in the E-RS total score and ≥−0.7 in the E-RS cough and sputum domain score.27 The E-RS was not used in the ACCORD study.

tems that relate specifically to cough and sputum symptoms (range 0–11). Responder criteria for E-RS total score and E-RS cough and sputum domain scores have been proposed as a change of ≥−2.0 units in the E-RS total score and ≥−0.7 in the E-RS cough and sputum domain score.27 The E-RS was not used in the ACCORD study. Morning and night-time cough and sputum symptoms Cough and sputum symptoms during the morning and night-time were assessed in the three phase III studies using questionnaires developed by the study sponsors.28 29 In ATTAIN, a 6-item night-time and morning symptoms of COPD questionnaire, completed by patients at approximately the same time every morning using an electronic patient diary, was used to assess the number of days patients experienced a range of morning or night-time symptoms, including coughing and bringing up phlegm or mucus. The questionnaire included one item that asked patients if they experienced symptoms during the night and one item that asked about their symptoms since they got out of bed to start the day.

umber of days patients experienced a range of morning or night-time symptoms, including coughing and bringing up phlegm or mucus. The questionnaire included one item that asked patients if they experienced symptoms during the night and one item that asked about their symptoms since they got out of bed to start the day. In the active-comparator study, morning symptoms were assessed using a 9-item COPD symptom questionnaire, completed daily by patients between 7:00 am and 11:00 am using an electronic diary. Early morning was defined as the time from when patients got out of bed to start the day until they started their daily activities. One item of the questionnaire was related to the presence of a range of early-morning symptoms, including cough and phlegm, with five items related to the severity of these symptoms. Patients assessed the severity of their overall morning symptoms (5-point scale: 1=‘I did not experience any symptoms’; 2=‘mild’; 3=‘moderate’; 4=‘severe’; 5=‘very severe’) and the severity of individual symptoms, including cough and difficulty bringing up phlegm (5-point scale: 0=‘no symptoms’; 1=‘mild’; 2=‘moderate’; 3=‘severe’; 4=‘very severe’).

of their overall morning symptoms (5-point scale: 1=‘I did not experience any symptoms’; 2=‘mild’; 3=‘moderate’; 4=‘severe’; 5=‘very severe’) and the severity of individual symptoms, including cough and difficulty bringing up phlegm (5-point scale: 0=‘no symptoms’; 1=‘mild’; 2=‘moderate’; 3=‘severe’; 4=‘very severe’). In ACCORD COPD I, night-time symptoms were assessed using an 11-item COPD night-time symptoms questionnaire, adapted from an existing COPD symptom questionnaire30 to include additional items assessing the frequency of COPD symptoms, such as night-time breathlessness, cough, sputum production and wheezing, during the previous night. Patients completed the questionnaire daily in the morning using an electronic patient diary (the recall period was ≤24 hours). The frequency of night-time symptoms was assessed on a 5-point scale: 0=‘never’; 1=‘1–2 times’; 2=‘3–4 times’; 3=‘5–6 times’; 4=‘7 or more times’. The severity and impact of night-time symptoms were assessed on a 5-point scale: 0=‘no symptoms’; 1=‘symptoms present but caused little/no discomfort’; 2=‘mild symptoms that were unpleasant but caused little/no discomfort’; 3=‘moderate symptoms that caused discomfort but did not affect daily activities’; 4=‘severe symptoms that interfered with normal daily activities’.

assessed on a 5-point scale: 0=‘no symptoms’; 1=‘symptoms present but caused little/no discomfort’; 2=‘mild symptoms that were unpleasant but caused little/no discomfort’; 3=‘moderate symptoms that caused discomfort but did not affect daily activities’; 4=‘severe symptoms that interfered with normal daily activities’. End points Predefined efficacy end points included: changes from baseline in E-RS total score and E-RS cough and sputum domain score over the study period (ATTAIN and active-comparator study); the percentage of days with morning or night-time symptoms over the study period (ATTAIN); changes from baseline in the percentage of days without morning symptoms and the severity of morning cough and difficulty bringing up phlegm over the study period (active-comparator study) and changes from baseline at week 12 in COPD night-time symptoms (ACCORD COPD I). To investigate the reliability of different measures of cough symptoms used in these analyses, a post hoc analysis assessed the correlation between changes from baseline in E-RS cough and sputum domain scores in those patients who had ≥1 exacerbation event in the ATTAIN study, and the severity of morning cough (based on the symptom questionnaires) at week 6 in the active-comparator study. These were selected as both measures assess the improvement from baseline in symptom severity. Safety and tolerability were assessed in all three studies by recording adverse events. Additional safety assessments included a physical examination, laboratory tests, vital signs and ECGs.

End points Predefined efficacy end points included: changes from baseline in E-RS total score and E-RS cough and sputum domain score over the study period (ATTAIN and active-comparator study); the percentage of days with morning or night-time symptoms over the study period (ATTAIN); changes from baseline in the percentage of days without morning symptoms and the severity of morning cough and difficulty bringing up phlegm over the study period (active-comparator study) and changes from baseline at week 12 in COPD night-time symptoms (ACCORD COPD I). To investigate the reliability of different measures of cough symptoms used in these analyses, a post hoc analysis assessed the correlation between changes from baseline in E-RS cough and sputum domain scores in those patients who had ≥1 exacerbation event in the ATTAIN study, and the severity of morning cough (based on the symptom questionnaires) at week 6 in the active-comparator study. These were selected as both measures assess the improvement from baseline in symptom severity. Safety and tolerability were assessed in all three studies by recording adverse events. Additional safety assessments included a physical examination, laboratory tests, vital signs and ECGs. Statistical analyses Demographic and baseline characteristics were assessed in the intent-to-treat (ITT) population (all treated patients who had baseline and at least one postbaseline FEV1 assessment) and are reported as mean (SD) or percentage, as appropriate. Efficacy analyses were performed in the ITT population. Changes from baseline in E-RS total and cough and sputum domain scores (ATTAIN and active-comparator study), percentage of days with morning or night-time cough symptoms (ATTAIN) and changes from baseline in the percentage of days without morning symptoms and the severity of morning symptoms (active-comparator study) were analysed using an analysis of covariance (ANCOVA) model, with treatment group and sex as factors and age and corresponding baseline as covariates. Changes from baseline in the frequency and severity of night-time symptoms (ACCORD COPD I) were analysed using an ANCOVA model with treatment as a factor and the corresponding baseline as a covariate. Data are reported as least squares mean (SEM), least squares mean differences (95% CIs) or percentages, as appropriate. For the post hoc analysis, Pearson coefficients were used to evaluate the correlation between improvements in E-RS cough and sputum domain score and the scores from the cough severity question in symptom questionnaires.

s least squares mean (SEM), least squares mean differences (95% CIs) or percentages, as appropriate. For the post hoc analysis, Pearson coefficients were used to evaluate the correlation between improvements in E-RS cough and sputum domain score and the scores from the cough severity question in symptom questionnaires. Additional post hoc analyses assessed the change from baseline in E-RS total and cough and sputum domain scores in those patients who had ≥1 exacerbation event identified using the EXACT in the ATTAIN study. An EXACT-identified event was defined as a persistent increase from baseline in total EXACT score of ≥9 points for ≥3 days or ≥12 points for ≥2 days.26 31 Results Patient population The ITT populations in ACCORD COPD I, ATTAIN and the active-comparator study included 559, 819 and 414 patients, respectively. Demographics and baseline clinical characteristics of the study populations have been reported previously;17–19 the demographics and baseline clinical characteristics in the placebo, aclidinium 400 μg and tiotropium arms are shown in table 1. E-RS scores and symptom questionnaire scores at baseline in the placebo, aclidinium 400 μg and tiotropium study arms are shown in online supplementary table S2. Table 1 Demographics and baseline clinical characteristics (ITT population)

Results Patient population The ITT populations in ACCORD COPD I, ATTAIN and the active-comparator study included 559, 819 and 414 patients, respectively. Demographics and baseline clinical characteristics of the study populations have been reported previously;17–19 the demographics and baseline clinical characteristics in the placebo, aclidinium 400 μg and tiotropium arms are shown in table 1. E-RS scores and symptom questionnaire scores at baseline in the placebo, aclidinium 400 μg and tiotropium study arms are shown in online supplementary table S2. Table 1 Demographics and baseline clinical characteristics (ITT population) ACCORD COPD I ATTAIN Active-comparator study Characteristic Placebo (N=185) Aclidinium 400 µg twice daily (N=190) Placebo (N=273) Aclidinium 400 µg twice daily (N=269) Placebo (N=85) Aclidinium 400 µg twice daily (N=171) Tiotropium 18 µg once daily (N=158) Age (years), mean (SD) 65.0 (9.2) 64.9 (9.5) 62.0 (8.0) 62.9 (8.4) 62.2 (8.2) 61.8 (8.2) 62.8 (7.9) Gender (male), n (%) 95 (51.4) 100 (52.6) 189 (69.2) 182 (67.7) 48 (56.5) 114 (66.7) 116 (73.4) Current smoker, n (%) 87 (47.0) 80 (42.1) 144 (52.8) 148 (55.0) 47 (55.3) 93 (54.4) 84 (53.2) Smoking history (pack-years), mean (SD) 52.9 (28.1) 57.2 (28.5) 38.9 (18.3) 41.7 (21.1) 39.6 (15.4) 41.5 (22.4) 45.0 (21.8) Postbronchodilator FEV1,* mean (SD), L 1.6 (0.6) 1.5 (0.5) 1.6 (0.5) 1.6 (0.5) 1.6 (0.5) 1.6 (0.5) 1.7 (0.5) Postbronchodilator FEV1% predicted,* mean (SD) 54.7 (13.4) 54.1 (12.9) 56.6 (12.8) 56.2 (12.2) 55.5 (11.8) 55.8 (13.3) 56.0 (13.2) Severity of airflow limitation,†,‡ n (%) Moderate 111 (60.0) 118 (62.1) 178 (65.9) 184 (68.7) 58 (68.2) 108 (63.2) 104 (66.2) Severe 72 (38.9) 68 (35.8) 92 (34.1) 84 (31.3) 27 (31.8) 63 (36.8) 53 (33.8) ≥1 COPD exacerbation in previous year,‡ n (%) 52 (28.1) 43 (22.6) 88 (32.6) 97 (36.2) 19 (22.4) 61 (35.7) 47 (29.7) Concomitant use of ICS, n (%) 70 (37.6) 81 (42.6) 145 (53.1) 128 (47.6) 36 (42.4) 82 (48.0) 67 (42.4) *At screening visit.

6.2) Severe 72 (38.9) 68 (35.8) 92 (34.1) 84 (31.3) 27 (31.8) 63 (36.8) 53 (33.8) ≥1 COPD exacerbation in previous year,‡ n (%) 52 (28.1) 43 (22.6) 88 (32.6) 97 (36.2) 19 (22.4) 61 (35.7) 47 (29.7) Concomitant use of ICS, n (%) 70 (37.6) 81 (42.6) 145 (53.1) 128 (47.6) 36 (42.4) 82 (48.0) 67 (42.4) *At screening visit. †Moderate COPD: 50% ≤postbronchodilator FEV1 <80% predicted and FEV1/FVC <0.70; severe COPD: 30% ≤postbronchodilator FEV1 <50% predicted and FEV1/FVC <0.70. ‡Patients with available data. COPD, chronic obstructive pulmonary disease; FEV1, forced expiratory volume in 1 s; FVC, forced vital capacity; ICS, inhaled corticosteroid; ITT, intent-to-treat. Safety and tolerability Safety and tolerability outcomes have previously been reported for each study.17–19 In summary, aclidinium is well tolerated with the most common adverse events being nasopharyngitis, headache, COPD exacerbation and cough. No clinically significant differences in other safety assessments were observed. No new safety and tolerability findings were anticipated based on these additional analyses. Daily COPD symptoms Treatment with aclidinium 400 µg significantly reduced total daily COPD symptoms compared with placebo, as assessed by E-RS total score over 24 weeks in ATTAIN (p<0.001; figure 1A) and 6 weeks in the active-comparator study (p<0.001; figure 1A).19 In the active-comparator study, E-RS total score was also significantly reduced with tiotropium compared with placebo (p<0.05; figure 1A).

COPD symptoms compared with placebo, as assessed by E-RS total score over 24 weeks in ATTAIN (p<0.001; figure 1A) and 6 weeks in the active-comparator study (p<0.001; figure 1A).19 In the active-comparator study, E-RS total score was also significantly reduced with tiotropium compared with placebo (p<0.05; figure 1A). Figure 1 Change from baseline in (A) E-RS total score and (B) E-RS cough and sputum domain score over the study period in ATTAIN and the active-comparator study. Data are reported as LS mean+SE. E-RS total score ranged from 0 to 40; E-RS cough and sputum domain score ranged from 0 to 11. Higher scores indicate more severe symptoms. *p<0.05, **p<0.01, ***p<0.001 vs placebo. E-RS, Evaluating Respiratory Symptoms, formerly known as EXAcerbations of Chronic pulmonary disease Tool; LS, least squares; MCID, minimum clinically important difference. Daily cough and sputum symptoms, assessed by E-RS cough and sputum domain score, were also significantly reduced with aclidinium 400 µg versus placebo in ATTAIN and the active-comparator study (both p<0.01; figure 1B). There was no significant difference between tiotropium and placebo treatments on cough and sputum symptoms in the active-comparator study (p=0.109; figure 1B).

d sputum domain score, were also significantly reduced with aclidinium 400 µg versus placebo in ATTAIN and the active-comparator study (both p<0.01; figure 1B). There was no significant difference between tiotropium and placebo treatments on cough and sputum symptoms in the active-comparator study (p=0.109; figure 1B). Post hoc analysis of a patient subpopulation with ≥1 exacerbation event identified by the EXACT (n=178) indicated that E-RS total score and E-RS cough and sputum scores were similar at baseline between aclidinium 400 μg and placebo in ATTAIN. After 24 weeks, treatment differences (95% CI) between aclidinium 400 μg and placebo in this group were significant for E-RS total score (−1.9 [−3.1 to −0.6]; p<0.01) and E-RS cough and sputum scores (−0.5 [−0.9 to −0.2]; p<0.01). Morning and night-time cough and sputum symptoms In ATTAIN, the percentage of days with any morning or night-time symptoms over the study period was significantly lower in patients treated with aclidinium 400 µg compared with placebo (both p<0.001; figure 2). Aclidinium treatment also significantly reduced the percentage of days with morning or night-time cough symptoms compared with placebo (both p<0.01; figure 2). Similarly, the percentage of days with morning or night-time bringing up phlegm or mucus was also significantly lower over the study period in patients treated with aclidinium 400 µg compared with placebo (p<0.01; figure 2).

f days with morning or night-time cough symptoms compared with placebo (both p<0.01; figure 2). Similarly, the percentage of days with morning or night-time bringing up phlegm or mucus was also significantly lower over the study period in patients treated with aclidinium 400 µg compared with placebo (p<0.01; figure 2). Figure 2 Percentage of days with (A) morning symptoms and (B) night-time symptoms over the study period in ATTAIN. Data are reported as least squares mean. **p<0.01, ***p<0.001 vs placebo. In the active-comparator study, both aclidinium and tiotropium significantly increased the change from baseline in the percentage of days without any morning symptoms over the study period versus placebo (treatment differences [95% CI] 8.9% [4.1% to 13.8%] with aclidinium and 5.6% [0.6% to 10.6%] with tiotropium; p<0.001 and p<0.05 vs placebo, respectively). Similarly, there was a significant increase in the percentage of days without morning cough symptoms in the aclidinium group compared with placebo (treatment difference [95% CI] 7.2% [1.1% to 13.4%]; p<0.05); there was no significant difference with tiotropium versus placebo (treatment difference [95% CI] 5.5% [−0.8% to 11.8%]; p=0.084). While the change from baseline in the percentage of days without difficulty bringing up phlegm was numerically higher with aclidinium (7.7%) and tiotropium (4.8%) compared with placebo (2.0%), the differences between the active treatments and placebo did not reach statistical significance (p=0.100 for aclidinium and p=0.425 for tiotropium).

baseline in the percentage of days without difficulty bringing up phlegm was numerically higher with aclidinium (7.7%) and tiotropium (4.8%) compared with placebo (2.0%), the differences between the active treatments and placebo did not reach statistical significance (p=0.100 for aclidinium and p=0.425 for tiotropium). Patients' assessment of the overall severity of their morning symptoms over the study duration was significantly reduced with aclidinium (−0.22; p<0.001) and tiotropium (−0.12; p<0.05) compared with placebo in the active-comparator study.19 When the severity of morning cough and difficulty bringing up phlegm was assessed, there was a significant reduction in the severity of both symptoms with aclidinium versus placebo over 6 weeks (p<0.05; figure 3). There was no significant change from baseline in the severity of either cough or difficulty bringing up phlegm in patients treated with tiotropium compared with placebo. Figure 3 Change from baseline in the severity of morning cough and severity of difficulty bringing up phlegm in the morning over the study period in the active-comparator study. Data are reported as LS mean+SE. Assessed on a 5-point scale: 0=‘no symptoms’ to 4=‘very severe symptoms’. *p<0.05, **p<0.01 vs placebo. LS, least squares.

ge from baseline in the severity of morning cough and severity of difficulty bringing up phlegm in the morning over the study period in the active-comparator study. Data are reported as LS mean+SE. Assessed on a 5-point scale: 0=‘no symptoms’ to 4=‘very severe symptoms’. *p<0.05, **p<0.01 vs placebo. LS, least squares. ACCORD COPD I investigated the prevalence and severity of night-time cough and sputum symptoms.17 After 12 weeks of treatment, aclidinium 400 µg significantly reduced the frequency of night-time cough compared with placebo (p<0.001; figure 4). The severity and impact of night-time cough symptoms was also significantly reduced at week 12 with aclidinium 400 µg compared with placebo (p<0.05; figure 4). In addition, the frequency of night-time sputum production was significantly lower in patients treated with aclidinium 400 µg compared with placebo (p<0.001; figure 4). Figure 4 Change from baseline in the severity of night-time cough and the frequency of night-time sputum production at week 12 in ACCORD COPD I.17 Data are reported as mean+SE. Symptom frequency assessed on a 5-point scale: 0=‘never’ to 4=‘7 or more times’. Symptom severity assessed on a 5-point scale: 0=‘no symptoms’ to 4=‘severe symptoms that interfered with normal daily activities’. *p<0.05, ***p<0.001 vs placebo.

utum production at week 12 in ACCORD COPD I.17 Data are reported as mean+SE. Symptom frequency assessed on a 5-point scale: 0=‘never’ to 4=‘7 or more times’. Symptom severity assessed on a 5-point scale: 0=‘no symptoms’ to 4=‘severe symptoms that interfered with normal daily activities’. *p<0.05, ***p<0.001 vs placebo. Correlation between E-RS and symptom questionnaires When all treatment groups were combined, there was significant correlation between the improvement in E-RS cough and sputum domain score and the improvement in the severity of morning cough symptoms assessed using the symptom questionnaire (r=0.684; p<0.001). Similar results were observed when the correlation between scores was assessed in each active treatment group (data not shown). Discussion This analysis is the first to specifically investigate the impact of a LAMA, or indeed any bronchodilator, on cough and sputum symptoms in patients with stable moderate-to-severe COPD. The results of the ATTAIN, ACCORD COPD I and active-comparator studies provide evidence that aclidinium is effective at reducing the severity and frequency of cough and sputum symptoms in patients with COPD, with improvements in E-RS total and E-RS cough and sputum scores as well as in evaluations of morning and night-time symptoms. These improvements were seen throughout the day and irrespective of the assessment tools used. Furthermore, aclidinium has previously been shown to be well tolerated in patients with COPD.17–19 23

h improvements in E-RS total and E-RS cough and sputum scores as well as in evaluations of morning and night-time symptoms. These improvements were seen throughout the day and irrespective of the assessment tools used. Furthermore, aclidinium has previously been shown to be well tolerated in patients with COPD.17–19 23 Chronic cough and mucus accumulation in the airways are strongly associated with disease progression, lung-function decline and risk of adverse outcomes in patients with COPD.2–4 However, most clinical trials designed to evaluate the efficacy of COPD treatments have focused on improvements in lung function and breathlessness and reductions in exacerbation risk as clinical outcomes. The few trials that have assessed the efficacy of a LAMA on cough and sputum symptoms to date have reported negative results. In phase III studies, there was no significant difference in physicians' assessment of cough symptoms between patients receiving tiotropium and those receiving placebo.32 33 Furthermore, no effect on mucociliary clearance was observed with ipratropium or tiotropium in patients with COPD.34 35 In contrast, patients with severe COPD treated with an inhaled corticosteroid (fluticasone) and a long-acting β2-agonist (salmeterol) have been shown to report significantly reduced cough symptoms versus placebo.36 Before the current analysis, only smoking cessation has consistently been shown to reduce cough and phlegm symptoms in patients with COPD.37 38

treated with an inhaled corticosteroid (fluticasone) and a long-acting β2-agonist (salmeterol) have been shown to report significantly reduced cough symptoms versus placebo.36 Before the current analysis, only smoking cessation has consistently been shown to reduce cough and phlegm symptoms in patients with COPD.37 38 Patients with COPD report variability in the frequency and severity of cough and sputum symptoms throughout the day, with greatest impact first thing in the morning and at night-time.10 11 39 The prevalence and severity of cough symptoms at the start of the day may relate to periods of increased activity associated with getting washed and dressed, whereas the night-time cough and sputum symptoms may be a consequence of mucus hypersecretion or reduced ciliary activity. In the analyses reported here, treatment with aclidinium significantly reduced the percentage of days and nights with symptoms of coughing and bringing up phlegm or mucus. The severity of morning and night-time cough and sputum symptoms was also found to be reduced. Treatment approaches that impact on cough and sputum symptoms throughout the whole 24-hour day may provide clinical benefits to some patients in terms of their overall well-being, particularly in the morning and night-time when patients report that these symptoms are most troublesome.

toms was also found to be reduced. Treatment approaches that impact on cough and sputum symptoms throughout the whole 24-hour day may provide clinical benefits to some patients in terms of their overall well-being, particularly in the morning and night-time when patients report that these symptoms are most troublesome. In the active-comparator study, while both LAMAs had an effect, the magnitude of improvement in overall symptoms and cough and sputum symptoms was greater with aclidinium compared with tiotropium. The reasons for this are unclear; however, the fact that both LAMAs improved symptoms of cough and sputum suggests that these may be class effects. Precisely how these compounds might exert an effect on cough and sputum is unclear, but there is an emerging body of preclinical evidence suggesting that multiple pathways may be involved. For example, there is evidence that muscarinic antagonists reduce experimental cough,40 and that tiotropium and ipratropium act on TRPV1 to reduce the cough response in preclinical models.41 In addition, a 2016 preclinical study in rabbits showed that, further to their anticholinergic activity and any action on TRPV1 receptors, aclidinium and tiotropium may also have antitussive actions involving mechanoreceptors and acid-sensing ion channels.42 These studies provide preclinical evidence of LAMA antitussive activity; however, it is not yet clear how this may translate into clinical practice.

inergic activity and any action on TRPV1 receptors, aclidinium and tiotropium may also have antitussive actions involving mechanoreceptors and acid-sensing ion channels.42 These studies provide preclinical evidence of LAMA antitussive activity; however, it is not yet clear how this may translate into clinical practice. Studies of capsaicin responsiveness suggested an increased cough reflex in patients with COPD;43 however, a recent study which evaluated predictors of cough frequency found no significant relationship between cough frequency and capsaicin cough reflex sensitivity.39 In contrast, cough frequency was independently associated with being a current smoker, smoking history, sputum production and neutrophilic inflammation.39 A recent study has demonstrated that M3 receptors may play a proinflammatory role in cigarette smoke-induced inflammation in animal models of COPD, suggesting another potential mechanism by which LAMAs may improve cough in patients with COPD.44 This is further supported by preclinical studies that have shown LAMAs can reduce neutrophils and inflammatory mediators, such as interleukin-6, tumour necrosis factor-α and interferon-γ, in cigarette smoke-exposed animal models.45 46 The efficacy of LAMAs to improve cough and sputum symptoms requires further investigation to determine if the effects observed with aclidinium are also seen with other drugs in this class.

mmatory mediators, such as interleukin-6, tumour necrosis factor-α and interferon-γ, in cigarette smoke-exposed animal models.45 46 The efficacy of LAMAs to improve cough and sputum symptoms requires further investigation to determine if the effects observed with aclidinium are also seen with other drugs in this class. The E-RS and night-time symptoms of COPD questionnaires are validated tools for assessing cough and sputum symptoms in patients with COPD.24 28 The observed improvements in E-RS cough and sputum symptoms with aclidinium 400 µg in ATTAIN (0.7 decrease from baseline in 24 weeks) and the active-comparator study (0.6 decrease from baseline in 6 weeks) compare well with the recently proposed minimum clinically important difference (MCID) of ≥0.7 decrease from baseline.47 The lack of a validated MCID in the other tools used in these studies may be considered to be a potential limitation of this analysis. However, the significant correlation between improvements from baseline in E-RS cough scores and the severity of morning cough symptoms assessed using symptom questionnaires in the active-comparator study supports the clinical utility of these tools to assess cough symptoms.

be a potential limitation of this analysis. However, the significant correlation between improvements from baseline in E-RS cough scores and the severity of morning cough symptoms assessed using symptom questionnaires in the active-comparator study supports the clinical utility of these tools to assess cough symptoms. This study has other potential limitations. It should be stated that none of the three phase III studies reported here was powered to detect differences in cough and sputum symptoms, and the studies were not specifically designed to assess these symptoms. Furthermore, there was no prespecified minimum level of symptoms in any of the studies, meaning the population was relatively heterogeneous in terms of symptoms. Clinical trials designed specifically to assess the effects of treatments on cough and sputum symptoms in patients with COPD, using a combination of patient-reported outcomes, cough-specific quality-of-life measures and objective measures of cough and sputum symptoms, are needed to fully understand the efficacy of novel treatments on these symptoms. Conclusions While few studies have investigated the effect of bronchodilators on cough and sputum symptoms, the results reported here suggest that in addition to improving lung function, LAMAs, such as aclidinium, can improve cough and sputum expectoration compared with placebo in patients with COPD. As cough and sputum symptoms impact negatively on overall patient well-being, controlling these symptoms may represent an important additional therapeutic benefit of this class of drugs.

ing lung function, LAMAs, such as aclidinium, can improve cough and sputum expectoration compared with placebo in patients with COPD. As cough and sputum symptoms impact negatively on overall patient well-being, controlling these symptoms may represent an important additional therapeutic benefit of this class of drugs. The authors would like to thank all of the patients and their families, the team of investigators, research nurses and operations staff involved in these studies. The authors would also like to thank Deborah McGregor, PhD, of Complete Medical Communications (Macclesfield, UK), who provided medical writing support under the direction of the authors, funded by AstraZeneca PLC. Contributors: LMG contributed to the interpretation of data, was involved in drafting the manuscript, revised it for intellectual content and provided final approval to submit. AHM contributed to the interpretation of the data, revised the manuscript for intellectual content and provided final approval to submit. JAS and SSB contributed to the interpretation of the data, revised the manuscript for intellectual content and provided final approval to submit. FC, BS and DJ contributed to the analyses and interpretation of the data, reviewed the manuscript for intellectual content and provided final approval to submit.

l to submit. JAS and SSB contributed to the interpretation of the data, revised the manuscript for intellectual content and provided final approval to submit. FC, BS and DJ contributed to the analyses and interpretation of the data, reviewed the manuscript for intellectual content and provided final approval to submit. Funding: These studies were funded by Almirall S.A., Barcelona, Spain, and Forest Laboratories LLC, a subsidiary of Actavis, New York, NY, USA. Almirall and Forest Laboratories were involved in the design of the study and analysis of the data. Almirall and AstraZeneca have reviewed the manuscript. The final decision to submit the manuscript remains with the authors. Competing interests: LMG has received honoraria and travel support from AstraZeneca. AHM has received honoraria, travel support and unrestricted grant aid from AstraZeneca. JAS is an inventor on a patent owned by the University Hospital South Manchester covering automated detection of cough from sound signals, and the patent is the subject of a license agreement with Vitalograph Ltd. (Buckinghamshire, UK); has received honoraria from Almirall and holds an MRC project grant part-funded by AstraZeneca. SSB received an honorarium for a scientific advisory board from Almirall. FC, BS and DJ are employees of AstraZeneca PLC and former employees of Almirall, Barcelona, Spain. Ethics approval: Institutional review boards/independent ethics committees at each site. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Competing interests: LMG has received honoraria and travel support from AstraZeneca. AHM has received honoraria, travel support and unrestricted grant aid from AstraZeneca. JAS is an inventor on a patent owned by the University Hospital South Manchester covering automated detection of cough from sound signals, and the patent is the subject of a license agreement with Vitalograph Ltd. (Buckinghamshire, UK); has received honoraria from Almirall and holds an MRC project grant part-funded by AstraZeneca. SSB received an honorarium for a scientific advisory board from Almirall. FC, BS and DJ are employees of AstraZeneca PLC and former employees of Almirall, Barcelona, Spain. Ethics approval: Institutional review boards/independent ethics committees at each site. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available. i Registered trademarks of AstraZeneca group of companies; for use within the USA as Pressair® and GenuairTM within all other licensed territories.

Key messages 39% of tuberculosis cases present to A+E in the preceding 6 months. Symptoms alone are insuffiently sensitive to rule out TB in an acute setting. Plain chest radiography in combination with risk factors and symptoms offers the best diagnostic yield. Introduction The WHO estimated that 9.6 million people developed tuberculosis (TB) in 2014.1 In England, the incidence of TB is persistently high. In 2014, 6520 cases of TB were reported.2 For the same year in the USA, which has a population nearly five times that of the UK's, 9421 cases of TB were reported3. Countries similar to the UK have made steady but intensive efforts to control TB and have seen a reduction in the number of new cases.4 Mycobacterium tuberculosis infection may have a prolonged subclinical phase during which the diagnosis and isolation of infectious cases can be difficult. In addition, current diagnostic methods have either a low specificity (eg, chest radiographs/Interferon Gamma Release Assays (IGRA)) or there is a long delay until results are available, especially when sputum smear results are negative and microbiological culture can take several weeks. These factors all contribute to making screening programmes for active TB more challenging to implement.5 Therefore, identification of factors that will lead to the consideration of a diagnosis of TB may prompt a more rapid initiation of investigations and sampling to allow for TB control measures.

take several weeks. These factors all contribute to making screening programmes for active TB more challenging to implement.5 Therefore, identification of factors that will lead to the consideration of a diagnosis of TB may prompt a more rapid initiation of investigations and sampling to allow for TB control measures. In the UK, the majority of new cases are concentrated in densely populated, urban areas. In 2014, London saw 2572 new cases of TB, nearly 40% of the national total, a rate of 30.1 per 100 000 which is nearly three times that of the UK average (12.0 per 100 000). The majority of these cases were in patients who were born outside the UK; the incidence was also higher in patients who were homeless, drug and alcohol misusers or immunosuppressed.2 The UK health authorities already recognise the difficulty in identifying and treating these patient groups and have released specific guidance to achieve early identification and successful treatment.6 The emergency department (ED) is an important facet of healthcare usage for these patient groups who do not regularly attend other healthcare services such as general practitioners (GPs) or may not be registered.7 8 9

groups and have released specific guidance to achieve early identification and successful treatment.6 The emergency department (ED) is an important facet of healthcare usage for these patient groups who do not regularly attend other healthcare services such as general practitioners (GPs) or may not be registered.7 8 9 There have been a small number of US-based studies looking at ED presentations of pulmonary TB (PTB), finding a high frequency of these patients born outside the USA, and with pulmonary symptoms and an abnormal chest radiograph increasing the likelihood of a TB diagnosis.10 11 In the UK, although one study has investigated TB symptoms and blood results in inpatients,12 and two short reports have examined cases at ED presentation,13 14 we have made a specific analysis of the usefulness of risk factors, symptoms and chest radiographs in the diagnosis of PTB and extrapulmonary TB (EPTB) in the ED. The aim of this study was to evaluate prediagnosis ED attendances in patients with a final diagnosis of confirmed active TB infection, to identify factors that are associated with delayed diagnosis, and investigate strategies to help identify such cases.

There have been a small number of US-based studies looking at ED presentations of pulmonary TB (PTB), finding a high frequency of these patients born outside the USA, and with pulmonary symptoms and an abnormal chest radiograph increasing the likelihood of a TB diagnosis.10 11 In the UK, although one study has investigated TB symptoms and blood results in inpatients,12 and two short reports have examined cases at ED presentation,13 14 we have made a specific analysis of the usefulness of risk factors, symptoms and chest radiographs in the diagnosis of PTB and extrapulmonary TB (EPTB) in the ED. The aim of this study was to evaluate prediagnosis ED attendances in patients with a final diagnosis of confirmed active TB infection, to identify factors that are associated with delayed diagnosis, and investigate strategies to help identify such cases. Methods Study population We performed a retrospective cohort analysis of patients that were diagnosed with active TB at two large central London hospitals between 1 January 2011 and 31 December 2012. ED discharge records were then used to identify patients with TB who had visited the ED in the previous 6 months. No age restriction was applied. Patients were only discounted from the study if there were no records available; this occurred in two cases. The study was approved as a service evaluation and ethics committee approval was not required.

n used to identify patients with TB who had visited the ED in the previous 6 months. No age restriction was applied. Patients were only discounted from the study if there were no records available; this occurred in two cases. The study was approved as a service evaluation and ethics committee approval was not required. Data collection The London TB Register (LTBR) is a database recording new TB cases and demographic, comorbidities, disease and outcome data; this along with ED notes, blood results and imaging databases were used in our data collection. We recorded the presence of a number of factors of interest, including demographic factors such as age, gender, country of birth, ethnicity, English speaking and number of years since entry into the UK. We also included known TB risk factors (eg, previous TB, TB contacts, alcohol misuse, homelessness and so on) as well as baseline physiological observations taken on presentation to ED and the presence of TB-related symptoms (cough, subjective fevers, night sweats, weight loss, sputum production, dyspnoea, haemoptysis and chest pain) and other symptoms. Any abnormal clinical findings related to TB on chest examination (crepitations, reduced air entry, dull percussion note or reduced expansion), lymphadenopathy or tenderness during an abdominal examination, were noted. We recorded whether the ED notes commented on the chest radiograph, the findings of the chest radiograph and any other imaging performed and the outcome of the encounter.

epitations, reduced air entry, dull percussion note or reduced expansion), lymphadenopathy or tenderness during an abdominal examination, were noted. We recorded whether the ED notes commented on the chest radiograph, the findings of the chest radiograph and any other imaging performed and the outcome of the encounter. Radiological evaluation The chest radiograph reports were reviewed. Any reports that referenced possible TB-related changes, for example, cavitation, infiltrates, effusion, lymphadenopathy or pleural thickening or if the report raised the possibility of TB as the cause were recorded as ‘abnormal’. If no reference to TB-related changes was made or the report found no abnormalities, it was recorded as ‘normal’. Statistical methods All data are presented as n (%) unless otherwise stated. The χ2 test or Fisher's exact test was used to compare categorical variables. The Mann-Whitney U test was used to compare continuous variables. Multivariable analysis was used to evaluate the association of variables recorded on presentation with suspicion of TB diagnosis in ED. The variables were included in the regression model are shown in online supplementary table S1. The threshold for statistical significance was p<0.05. 10.1136/bmjresp-2016-000154.supp1supplementary table Results Study cohort There were 397 patients included in the study, out of these 397 cases, 156 (39%) presented to the ED in the 6 months prior to their diagnosis. Two cases were excluded due to inadequate records being available. Baseline demographics of the study population are shown in table 1.

10.1136/bmjresp-2016-000154.supp1supplementary table Results Study cohort There were 397 patients included in the study, out of these 397 cases, 156 (39%) presented to the ED in the 6 months prior to their diagnosis. Two cases were excluded due to inadequate records being available. Baseline demographics of the study population are shown in table 1. Table 1 Baseline demographics of study cohort Demographics/patient factors Frequency (%) Population size 397 Male 231 (58%) Median age (IQR) 35 (25) Pulmonary TB 166 (42%) Non-pulmonary TB 231 (58%) Indian subcontinent 117 (29%) Black African 111 (28%) White 74 (19%) Black Caribbean 22 (6%) Arabic 26 (7%) Other ethnicity 47 (12%) Born outside UK 309 (78%) Previous BCG vaccination 223 (56%) Previous TB infection 16 (4%) Unemployed 67 (17%) Homelessness 29 (7%) Drug misuse 22 (6%) History of imprisonment 17 (4%) Not registered with GP 79 (20%) GP, general practitioner; TB, tuberculosis. Comparison of patients with TB who presented to ED prior to diagnosis with those who did not There were 154 (39%) of the study population who presented to ED in the 6-month period prior to their diagnosis of TB. Table 2 shows comparison of demographic and comorbid variables between patients who attended ED versus those who did not. Table 2 Baseline demographics and patients factors in subgroup who attended ED versus group who did not

Comparison of patients with TB who presented to ED prior to diagnosis with those who did not There were 154 (39%) of the study population who presented to ED in the 6-month period prior to their diagnosis of TB. Table 2 shows comparison of demographic and comorbid variables between patients who attended ED versus those who did not. Table 2 Baseline demographics and patients factors in subgroup who attended ED versus group who did not Attended ED (%) Did not attend ED (%) χ2 test (p value) Demographics Population size 154 243 – Male 100 (65%) 131 (54%) 0.030 Age (IQR) 34 (23) 35 (22) 0.408 Pulmonary TB 75 (49%) 91 (37%) 0.045 Non-pulmonary TB 79 (51%) 152 (63%) Ethnicity Indian subcontinent 40 (26%) 77 (32%) 0.224 Black African 44 (29%) 67 (28%) 0.829 White 31 (20%) 43 (18%) 0.544 Black Caribbean 12 (8%) 10 (4%) 0.119 Arabic 13 (8%) 13 (5%) 0.225 Other ethnicity 14 (9%) 33 (14%) 0.177 Comorbidities Born outside UK 120 (78%) 189 (78%) 0.973 Previous BCG vaccination 89 (58%) 134 (55%) 0.604 Previous TB infection 9 (6%) 7 (3%) 0.143 Unemployed 27 (18%) 40 (16%) 0.781 Homelessness 12 (8%) 17 (7%) 0.766 Drug misuse 11 (7%) 11 (5%) 0.267 Prison history 8 (5%) 9 (4%) 0.475 Not registered with GP 38 (25%) 41 (17%) 0.058 ED, emergency department; GP, general practitioners; TB, tuberculosis.

(58%) 134 (55%) 0.604 Previous TB infection 9 (6%) 7 (3%) 0.143 Unemployed 27 (18%) 40 (16%) 0.781 Homelessness 12 (8%) 17 (7%) 0.766 Drug misuse 11 (7%) 11 (5%) 0.267 Prison history 8 (5%) 9 (4%) 0.475 Not registered with GP 38 (25%) 41 (17%) 0.058 ED, emergency department; GP, general practitioners; TB, tuberculosis. Symptomatic presentation and baseline observations in patients who attended ED We next conducted an evaluation of the symptoms reported by patients who attended ED prior to TB diagnosis. Table 3 shows the most common symptoms reported and also the proportion of patients with baseline abnormalities in routine observations. Thirty-one per cent of cases presenting to the ED had no abnormal observations (48/154). Table 3 Symptoms and baseline observations of patients who attended ED prior to diagnosis of TB

Symptomatic presentation and baseline observations in patients who attended ED We next conducted an evaluation of the symptoms reported by patients who attended ED prior to TB diagnosis. Table 3 shows the most common symptoms reported and also the proportion of patients with baseline abnormalities in routine observations. Thirty-one per cent of cases presenting to the ED had no abnormal observations (48/154). Table 3 Symptoms and baseline observations of patients who attended ED prior to diagnosis of TB Risk factors Frequency (% of study population) Pulmonary TB symptoms Frequency (% of study population) Current smoker 27 (18%) Cough 62 (40%) Recent travel 19 (12%) Fever 53 (34%) Known TB contacts 13 (8%) Weight loss 53 (34%) Airways disease 12 (8%) Night sweats 33 (21%) Diabetes (type 1 or 2) 12 (8%) Chest pain 31 (20%) HIV positivity 6 (4%) Sputum production 28 (18%) Hepatitis B or C positivity 3 (2%) Dyspnoea 28 (18%) Alcohol misuse 13 (8%) Haemoptysis 11 (7%) Immunosuppressive treatment 3 (2%) Other symptoms Observations  Gastrointestinal 30 (19%) Tachycardia (<100/min) 53 (34%)  CNS/headache 16 (10%) Fever (>38°C) 30 (19%)  Back/joint pain 13 (8%) Tachypnoea (>20/min) 20 (13%)  Neck lump 8 (5%) Pulse oximetry <95% 8 (5%)  Generalised weakness and malaise 8 (5%) Hypotension (workup SBP<90 mm Hg) 3 (2%)  Other 9 (6%) CNS, central nervous system; ED, emergency department; SBP, systolic blood pressure; TB, tuberculosis.

ever (>38°C) 30 (19%)  Back/joint pain 13 (8%) Tachypnoea (>20/min) 20 (13%)  Neck lump 8 (5%) Pulse oximetry <95% 8 (5%)  Generalised weakness and malaise 8 (5%) Hypotension (workup SBP<90 mm Hg) 3 (2%)  Other 9 (6%) CNS, central nervous system; ED, emergency department; SBP, systolic blood pressure; TB, tuberculosis. Of the PTB cases, the three most prevalent symptoms recorded in the ED clerking were cough (64%), weight loss (49%) and subjective fevers (41%). EPTB cases were far less likely to have classical TB symptoms with the three most prevalent of these being subjective fevers (28%), weight loss (20%) and cough (18%). Fifty-eight per cent (90/154) had any of the three most prevalent symptoms (cough, weight loss and fevers). In patients with PTB the incidence was 72% (54/75), compared with the lower rate for EPTB 46% (36/79). Seventy-five per cent of the EPTB cases had other symptoms recorded, 38% (30/79) had gastrointestinal symptoms, 20% (16/79) had central nervous system (CNS)/headache symptoms and 16% (13/79) had back or joint pain symptoms. Comparison of patients who had TB suspected in ED versus those who did not TB was included in the differential diagnosis in the ED in 56 of 154 cases (36%). Table 4 shows a comparison between patients who had TB suspected in ED versus those who did not. Table 4 Comparison of patients where TB was suspected in the ED and those where it was not

Comparison of patients who had TB suspected in ED versus those who did not TB was included in the differential diagnosis in the ED in 56 of 154 cases (36%). Table 4 shows a comparison between patients who had TB suspected in ED versus those who did not. Table 4 Comparison of patients where TB was suspected in the ED and those where it was not TB suspected in ED Frequency (%) TB not suspected in ED Frequency (%) χ2 test (p value) Demographics Population size 56 98 – Male 37 (66%) 63 (64%) 0.823 Age (IQR) 30.5 (16.75) 36.5 (30.5) 0.276 (Mann-Whitney) Pulmonary TB 41 (73%) 34 (35%) <0.001 Non-pulmonary TB 15 (27%) 64 (65%) Indian subcontinent 14 (25%) 26 (27%) 0.835 Black African 19 (34%) 25 (26%) 0.266 White 10 (18%) 21 (21%) 0.595 Black Caribbean 6 (11%) 6 (6%) 0.306 Arabic 5 (9%) 8 (8%) 0.870 Other ethnicity 2 (4%) 12 (12%) 0.086 Born outside UK 47 (84%) 73 (74%) 0.174 Previous BCG vaccination 37 (66%) 52 (53%) 0.116 Unemployed 13 (23%) 14 (14%) 0.161 Homelessness 5 (9%) 7 (7%) 0.691 Prison history 2 (4%) 6 (6%) 0.711 Comorbidities Current smoker 12 (21%) 15 (15%) 0.336 Recent travel 14 (25%) 5 (5%) <0.001 Previous TB infection 6 (11%) 3 (3%) 0.0734 Known TB contacts 11 (20%) 2 (2%) <0.001 Airways disease 6 (11%) 6 (6%) 0.306 Diabetes (type 1 or 2) 4 (7%) 8 (8%) 1.000 HIV positivity 3 (5%) 3 (3%) 0.6686 Hepatitis B or C positivity 1 (0%) 2 (2%) 1.000 Drug misuse 5 (9%) 6 (6%) 0.515 Alcohol misuse 5 (9%) 8 (8%) 0.870 Immunosuppressive treatment 0 (0%) 3 (3%) 0.5541 Observations Tachycardia (<100/min) 22 (39%) 31 (32%) 0.336 Fever (<38°C) 13 (23%) 17 (17%) 0.376 Tachypnoea (>20/min) 9 (16%) 11 (11%) 0.389 Pulse oximetry <95% 3 (5%) 5 (5%) 1.000 Hypotension (SBP<90 mm Hg) 1 (2%) 2 (2%) 1.000 Symptoms Cough 41 (73%) 21 (21%) <0.001 Fever 30 (54%) 23 (23%) <0.001 Weight loss 33 (59%) 20 (20%) <0.001 Night sweats 25 (45%) 8 (8%) <0.001 Chest pain 14 (25%) 17 (17%) 0.255 Sputum production 18 (32%) 10 (10%) <0.001 Dyspnoea 14 (25%) 14 (14%) 0.097 Haemoptysis 7 (13%) 4 (4%) 0.0991 Chest radiograph Chest radiography performed 52 (93%) 65 (66%) <0.001 Normal radiograph (% of those who had radiography performed) 11 (21%) 30 (46%) 0.138 Abnormal radiograph (% of those who had radiography performed) 41 (79%) 35 (54%) <0.001 Bold indicates significant statistics.

.097 Haemoptysis 7 (13%) 4 (4%) 0.0991 Chest radiograph Chest radiography performed 52 (93%) 65 (66%) <0.001 Normal radiograph (% of those who had radiography performed) 11 (21%) 30 (46%) 0.138 Abnormal radiograph (% of those who had radiography performed) 41 (79%) 35 (54%) <0.001 Bold indicates significant statistics. ED, emergency department; SBP, systolic blood pressure; TB, tuberculosis. This rate of clinical suspicion of TB was significantly higher in patients with PTB than those with EPTB (55% vs 19%; p<0.01). There was also a significant difference between the median time from ED attendance to starting treatment. In patients where TB was suspected, median was 9.75 days (3–12.75), compared with 71.25 days (13–84.25) in those where it was not listed as a differential diagnosis (Mann-Whitney U test, p<0.001. Figure 1 Chest radiograph findings and type of TB in patients presenting to ED. ED, emergency department; TB, tuberculosis. Multivariable analysis On multivariable analysis, the following factors were independently associated with suspicion of TB in ED: cough (OR 6.30 (1.71–23.21), birth outside UK (OR 5.48 (1.29–23.30), known contact with TB infected person (OR 18.63 (2.49–139.27), history of night sweats (16.09 (3.39–76.42) and abnormal chest radiograph (OR 5.90 (1.70–20.53) (see online supplementary table S1).

re independently associated with suspicion of TB in ED: cough (OR 6.30 (1.71–23.21), birth outside UK (OR 5.48 (1.29–23.30), known contact with TB infected person (OR 18.63 (2.49–139.27), history of night sweats (16.09 (3.39–76.42) and abnormal chest radiograph (OR 5.90 (1.70–20.53) (see online supplementary table S1). Chest radiograph Of the 154 patients visiting the ED in the 6 months prior to their diagnosis, 24% (37/154) did not have a chest radiograph or other imaging performed on presentation to ED. For PTB cases, 86% (55/64) patients who had a chest radiograph had an abnormality. As would be expected, EPTB had a lower rate of abnormal radiographs, 40% (21/53), but it is notable that there is still a significant proportion with plain chest radiology abnormalities. Of all patients with an abnormal radiograph, many had multiple abnormal findings; the three most common abnormalities seen on radiograph were consolidation in 55% (42/76), effusions in 26% (20/76) and decreased volume or collapse in 22% (17/76). In the patients where TB was not suspected but who had an abnormal chest X-ray (CXR), the most common features were consolidation 46% (16/35), reduced volume 31% (11/35) and effusion 23% (8/35).

en on radiograph were consolidation in 55% (42/76), effusions in 26% (20/76) and decreased volume or collapse in 22% (17/76). In the patients where TB was not suspected but who had an abnormal chest X-ray (CXR), the most common features were consolidation 46% (16/35), reduced volume 31% (11/35) and effusion 23% (8/35). Of those 35 patients where the chest radiograph was abnormal and TB was not suspected, 63% (22/35) were men, the average age was 45 years and 71% (25/35) were born outside the UK. Thirty-seven per cent (13/35) of patients had other symptoms present. Fewer patients had any of the three most prevalent symptoms: cough (40%), weight loss (31%) and subjective fevers (29%). Thirty-one per cent (11/35) of these patients reattended the same ED compared with 26% (40/154) overall. Of patients attending the ED, 69% (107/154) had an abnormal chest radiograph report, presence of cough, subjective fevers or weight loss. There is no validated probability score in acute TB and we therefore used the features of night sweats and cavitation on chest radiograph as a proxy for cases where TB should have reasonably been suspected. Five per cent (7/154) patients had night sweats and cavitation and all of these patients had TB suspected in the ED. Of the 7% (11/154) patients who had cavitation but no night sweats on chest radiograph, all 11 patients had TB suspected. Eight out of 33 patients who had night sweats but not cavitation did not have TB suspected in the ED.

) patients had night sweats and cavitation and all of these patients had TB suspected in the ED. Of the 7% (11/154) patients who had cavitation but no night sweats on chest radiograph, all 11 patients had TB suspected. Eight out of 33 patients who had night sweats but not cavitation did not have TB suspected in the ED. Forty per cent (21/53) of patients with EPTB who had a radiography performed had an abnormality recorded in the report. Thirty-three per cent (26/79) had no radiography performed. Treatment and outcomes Fifty-four per cent (83/154) of patients were admitted following their accident and emergency (A&E) department attendance and 26% (40/154) of patients reattended A&E following their first presentation. Of those patients who were admitted, 35% (29/83) had TB suspected in the ED and 51% (42/83) had an abnormal CXR compared with 38% (27/71) and 48% (34/71), respectively, for those who were not admitted. The mean interval between presentation to the ED and the initiation of TB treatment was 39 days (range 0–185 days, median 17 days). Thirty-one per cent (48/154) patients were started on TB treatment within a week of their A&E visit. For those patients where TB was considered as part of the differential, the time to starting treatment was dramatically lower, average of 14.4 days compared with 53.6 days.

as 39 days (range 0–185 days, median 17 days). Thirty-one per cent (48/154) patients were started on TB treatment within a week of their A&E visit. For those patients where TB was considered as part of the differential, the time to starting treatment was dramatically lower, average of 14.4 days compared with 53.6 days. Discussion In our study population, a high proportion of patients, 39%, presented to the ED in the 6 months prior to their diagnosis, demonstrating the importance of this particular location as an opportunity for the detection of patients suffering from this disease. In those suspected to have TB during their attendance, the ED has allowed a timely diagnosis and it is possible that had the individual not accessed healthcare through the ED the case would have remained undiagnosed. In addition, 7% (11/154) had sputum smear performed within 24 hours of attending A&E and these patients may not have had sputum sent in other healthcare settings. It has also been established that large Western European cities have significantly higher rates of TB infection compared with national rates; London has experienced a continued rise in TB notification rates from 24 per 100 000 in 1990 to 45 per 100 000 in 2011.15 16 This high incidence of TB coupled with the high use of emergency care in this urban population emphasises the importance of case detection being initiated promptly by clinicians in the ED.

rates; London has experienced a continued rise in TB notification rates from 24 per 100 000 in 1990 to 45 per 100 000 in 2011.15 16 This high incidence of TB coupled with the high use of emergency care in this urban population emphasises the importance of case detection being initiated promptly by clinicians in the ED. Our cohort reflected the national picture of TB being more prevalent in patients born outside the UK, homeless, drug and alcohol misusers or immunosuppressed.2 This again emphasises the importance of considering TB in any cases presenting with these epidemiological risk factors. A quarter of patients who presented to the ED were not registered with a GP, a higher rate than those not attending the ED. Although the difference was not statistically significant, it highlights that many patients with TB underuse primary care services and are therefore more likely to present to emergency care services. Apart from male gender, there were no statistically significant differences between the demographics and comorbidities of those who presented to the ED compared with those who did not. Nationally, 50% of ED attenders are men, whereas 65% of the study patients with TB who attended ED were men.17 However, the study relied on the accuracy of documentation for these risk factors and comorbidities which will have varied between clinicians.

of those who presented to the ED compared with those who did not. Nationally, 50% of ED attenders are men, whereas 65% of the study patients with TB who attended ED were men.17 However, the study relied on the accuracy of documentation for these risk factors and comorbidities which will have varied between clinicians. Pulmonary TB cases represented approximately half of the study population; in this group tachycardia and fever were the most commonly abnormal observations and cough, weight loss and fevers were the most common classical TB symptoms that were present. Almost half of the EPTB patients had no classical TB symptoms and three-quarters presented with other symptoms of which gastrointestinal were the most common. The low incidence rate of classical symptoms, particularly in the EPTB group, makes diagnosis on symptoms alone unfeasible. As would be expected, cases of PTB more frequently displayed chest radiograph abnormalities than EPTB. In keeping with these findings, PTB was more often suspected as a diagnosis in the ED than EPTB (55% vs 19% of cases).

classical symptoms, particularly in the EPTB group, makes diagnosis on symptoms alone unfeasible. As would be expected, cases of PTB more frequently displayed chest radiograph abnormalities than EPTB. In keeping with these findings, PTB was more often suspected as a diagnosis in the ED than EPTB (55% vs 19% of cases). Overall, patients where TB was suspected in the ED had a much greater frequency of chest radiograph, 93% compared with 66%. However, patients where TB was suspected had a greater frequency of all pulmonary symptoms and were therefore more likely to have had a chest radiograph requested. On multivariable analysis, cough, country of birth outside the UK, a TB contact history, a history of night sweats and an abnormal chest radiograph were independently associated with suspicion of TB in ED. The significant proportion of patients where TB was not suspected in the ED but had abnormal radiograph findings (35 patients, 54%) highlights the importance of carefully reviewing all imaging undertaken in the ED and ensuring that any abnormal reports are acted on. Contemporary reporting of imaging while the patient is still in the ED could mean that clinical review can be correlated with radiograph findings. It is important to note that the plain chest radiograph had the best sensitivity when comparing this to symptoms. Importantly, our results indicate that even in cases of EPTB, there is an important role for performing a plain chest radiograph and that this has a broadly similar sensitivity compared with the incidence of classical symptoms. However, the most common chest radiograph abnormalities were consolidation, reduced volume and effusion; these radiographic characteristics can be present for a wide differential of respiratory infections.

hest radiograph and that this has a broadly similar sensitivity compared with the incidence of classical symptoms. However, the most common chest radiograph abnormalities were consolidation, reduced volume and effusion; these radiographic characteristics can be present for a wide differential of respiratory infections. This study has a number of strengths and is, to the best of our knowledge, the first study of its kind in the UK to look systematically at radiographs in combination with risk factors and symptoms. We collected and analysed data on a large study population across two sites and it was conducted in a moderately high-incidence urban area. This study also links the initial ED presentation to TB clinic visit events. Previous similar studies have been in lower prevalence areas and in insurance-led healthcare rather than free at the point of use health systems.10 11

population across two sites and it was conducted in a moderately high-incidence urban area. This study also links the initial ED presentation to TB clinic visit events. Previous similar studies have been in lower prevalence areas and in insurance-led healthcare rather than free at the point of use health systems.10 11 Other studies in the USA have previously investigated the use of chest radiographs to help TB diagnosis in the ED.10 11 These studies found that frequently there is a wide range of symptoms and the chest radiograph helps in assisting the clinician's diagnosis. Our study supports these findings and this is in keeping with recently published national guidelines that encourage a chest radiograph to be performed in all patients with suspected TB (even in cases of EPTB) and to ensure that those with relevant radiographic changes are referred directly to local TB services.18 Mobile chest radiographs have previously been shown to have a high sensitivity but in the context of imaging in select high-risk groups.19 Our study lends support to interventions such as the ‘Find and Treat’ initiative in London that uses such units to actively find PTB cases.20 21

red directly to local TB services.18 Mobile chest radiographs have previously been shown to have a high sensitivity but in the context of imaging in select high-risk groups.19 Our study lends support to interventions such as the ‘Find and Treat’ initiative in London that uses such units to actively find PTB cases.20 21 As the study was retrospective, collection of data relied on availability of scanned or electronic records. Importantly, we cannot assume that all patients had active infection with TB 6 months before their diagnosis, or that a diagnosis could have been made on their ED visit. The study only included two EDs and patients may have attended other EDs within the study period and the figure of 39% may underestimate the actual number of visits. The point of contact this group makes with healthcare will increasingly be urgent care centres which were not part of the UK healthcare infrastructure at the time of the study. Also, our study did not have a control group to allow comparisons with the wider population. Ideally, this would be a control group of patients with another respiratory condition, for example, pneumonia which may allow us to look more closely at symptoms and patient factors predicting TB.

ture at the time of the study. Also, our study did not have a control group to allow comparisons with the wider population. Ideally, this would be a control group of patients with another respiratory condition, for example, pneumonia which may allow us to look more closely at symptoms and patient factors predicting TB. The high proportion of patients subsequently diagnosed with TB, that present to the ED prior to their diagnosis, lends support to rapid molecular-based TB diagnostics such as GeneXpert as an adjunct to conventional microscopy, culture and drug sensitivity testing. This would allow timelier testing as part of the workup for patients with suspected TB. It also highlights that patients that are potentially infective are attending a clinical environment where there is a high density of patients and there should be a lower threshold for consideration of isolation especially in higher prevalence, urban areas.

testing as part of the workup for patients with suspected TB. It also highlights that patients that are potentially infective are attending a clinical environment where there is a high density of patients and there should be a lower threshold for consideration of isolation especially in higher prevalence, urban areas. Our study also shows that in cases of EPTB, there was an even greater range of presenting symptoms potentially making the diagnosis even more challenging. Approximately, half of TB patients presenting to the ED have EPTB and although these patients were less likely to have an abnormal chest radiograph (46% vs 86% in PTB), this yield is still important and a chest radiograph should be recommended in any case of potential EPTB. In addition, this study supports the development of a defined diagnostic pathway between ED and TB services when an individual has epidemiological risk factors for TB. These data also support the need for there to be a rapid and direct referral pathway from radiology to TB services should there be any abnormalities indicating potential TB when formally reported.

ment of a defined diagnostic pathway between ED and TB services when an individual has epidemiological risk factors for TB. These data also support the need for there to be a rapid and direct referral pathway from radiology to TB services should there be any abnormalities indicating potential TB when formally reported. Conclusions This study highlights the importance of the ED as a common key healthcare point of access for TB patients and demonstrates the value of TB departments working across healthcare services as the patient population may access healthcare in a less traditional way. To improve detection and control rates for TB, the diagnosis should be suspected in any patient with a demographic profile or symptoms associated with TB risk and the ED physician should have a low threshold for further investigations for TB. The results of this study suggest that the classical symptoms of TB alone are insufficient in assessment of patients with suspected disease as it frequently presents in a non-specific manner. Our results support the use of chest radiographs in helping alert a potential diagnosis in PTB and EPTB. Importantly, it also raises the difficulties of diagnosing EPTB given its wide varying presentations and highlights that a plain chest radiograph has the highest sensitivity of all the factors normally reviewed in an ED and should be requested in any possible case even when there are no respiratory symptoms. Kevin Kow, Imperial College London, contributed data on patient demographics.

Conclusions This study highlights the importance of the ED as a common key healthcare point of access for TB patients and demonstrates the value of TB departments working across healthcare services as the patient population may access healthcare in a less traditional way. To improve detection and control rates for TB, the diagnosis should be suspected in any patient with a demographic profile or symptoms associated with TB risk and the ED physician should have a low threshold for further investigations for TB. The results of this study suggest that the classical symptoms of TB alone are insufficient in assessment of patients with suspected disease as it frequently presents in a non-specific manner. Our results support the use of chest radiographs in helping alert a potential diagnosis in PTB and EPTB. Importantly, it also raises the difficulties of diagnosing EPTB given its wide varying presentations and highlights that a plain chest radiograph has the highest sensitivity of all the factors normally reviewed in an ED and should be requested in any possible case even when there are no respiratory symptoms. Kevin Kow, Imperial College London, contributed data on patient demographics. Contributors: OMK, SCA and DWC devised the study and SCA, PB and DP performed the data collection. SCA and PB performed the data analysis. SCA, DWC, AS, PB, DP, BC and OMK contributed to data interpretation. SA wrote the manuscript with assistance from DC, OMK and AS. All authors contributed to the review and editing of the manuscript.

, SCA and DWC devised the study and SCA, PB and DP performed the data collection. SCA and PB performed the data analysis. SCA, DWC, AS, PB, DP, BC and OMK contributed to data interpretation. SA wrote the manuscript with assistance from DC, OMK and AS. All authors contributed to the review and editing of the manuscript. Funding: The research was supported by the National Institute for Health Research (NIHR) Biomedical Research Centre based at Imperial College Healthcare NHS Trust and Imperial College London. 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: AS has received honoraria for speaking from GlaxoSmithKline. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Key messages HIV-associated chronic obstructive pulmonary disease (HIV-COPD) has a unique plasma metabolomic profile that includes sphingolipids and fatty acids. Tryptophan catabolism is increased in PLWH, but does not correlate with COPD status. Additional studies are needed to determine how such metabolic pathways contribute to HIV-COPD and whether therapeutic interventions to alter sphingolipid expression can reduce the risk of COPD in persons living with HIV.

Key messages HIV-associated chronic obstructive pulmonary disease (HIV-COPD) has a unique plasma metabolomic profile that includes sphingolipids and fatty acids. Tryptophan catabolism is increased in PLWH, but does not correlate with COPD status. Additional studies are needed to determine how such metabolic pathways contribute to HIV-COPD and whether therapeutic interventions to alter sphingolipid expression can reduce the risk of COPD in persons living with HIV. Introduction Combination antiretroviral therapy (ART) has markedly improved the survival of individuals living with HIV-1. However, improved survival has led to a higher prevalence of several chronic illnesses including chronic obstructive pulmonary disease (COPD), which affects an estimated 3–23% of persons living with HIV (PLWH).1–8 Although the prevalence of smoking is high in PLWH, observational studies suggest that HIV infection is an independent risk factor for accelerated lung function decline and subsequent COPD.1 2 5 In the era prior to the common use of highly active ART (HAART), pulmonary obstruction was mostly associated with advanced HIV/AIDS and frequent pulmonary infections.9 In the HAART era, COPD persists as a frequent comorbidity in PLWH even in the absence of AIDS or frequent pulmonary infections.1–5 8 10–14 It is highly likely that several factors are involved in HIV-associated COPD (HIV-COPD) and many investigators have implicated the HIV virus itself. Currently, the underlying mechanisms of HIV-COPD remain unclear and there is no biomarker to identify which patients with HIV will be at a higher risk for the subsequent development of COPD. For this study, we identified individuals in the Strategic Timing of Antiretroviral Treatment (START) study with HIV-COPD prior to the diagnosis of AIDS or initiation of HAART, offering a unique opportunity to characterise HIV-COPD.15

ch patients with HIV will be at a higher risk for the subsequent development of COPD. For this study, we identified individuals in the Strategic Timing of Antiretroviral Treatment (START) study with HIV-COPD prior to the diagnosis of AIDS or initiation of HAART, offering a unique opportunity to characterise HIV-COPD.15 Recent developments in metabolomics and metabolic profiling have enabled the measurements of thousands of metabolites simultaneously in biological samples leading to new discoveries of biomarkers of disease.16 Metabolomics, the study of small molecules, is a field that complements genomics and proteomics to provide a snapshot into the physiology of human disease. As such, global, untargeted metabolomic profiling is a method for analysing complex diseases to understand the underlying pathways involved in disease pathogenesis and to identify biomarker candidates for disease. In this study, the primary aim was to perform large-scale untargeted analysis with mass spectrometry of plasma to identify metabolites associated with HIV-COPD. Our results revealed that HIV-COPD has a unique metabolomic signature and among the metabolome are sphingolipids. Methods We performed a cross-sectional, matched case–control study using plasma samples from two large cohort studies.

Recent developments in metabolomics and metabolic profiling have enabled the measurements of thousands of metabolites simultaneously in biological samples leading to new discoveries of biomarkers of disease.16 Metabolomics, the study of small molecules, is a field that complements genomics and proteomics to provide a snapshot into the physiology of human disease. As such, global, untargeted metabolomic profiling is a method for analysing complex diseases to understand the underlying pathways involved in disease pathogenesis and to identify biomarker candidates for disease. In this study, the primary aim was to perform large-scale untargeted analysis with mass spectrometry of plasma to identify metabolites associated with HIV-COPD. Our results revealed that HIV-COPD has a unique metabolomic signature and among the metabolome are sphingolipids. Methods We performed a cross-sectional, matched case–control study using plasma samples from two large cohort studies. Study population Cases were HIV positive and were selected from a pulmonary substudy of the START trial.15 17 START enrolled HIV-positive, ART-naive persons with a CD4+ count >500 cells/mm3. From this cohort, we selected 38 participants with HIV-COPD (defined as forced expiratory volume in 1 s (FEV1)/forced vital capacity (FVC)< lower limits of normal and FEV1<80% predicted at baseline pulmonary function testing, table 1) for metabolomic profiling. HIV-positive controls consisted of 40 individuals with normal lung function (defined as FEV1/FVC>lower limits of normal and FEV1>80% predicted) matched on age, sex, region and smoking status. HIV-negative controls were identified in the Genetic Epidemiology of Chronic Obstructive Pulmonary Disease (COPDGene) study, an observational study that enrolled current or former smokers. COPDGene controls were matched on lung function, age, sex and race. These consisted of 17 individuals without COPD and 20 individuals with COPD (table 1). This study was approved by the University of Minnesota Institutional Review Board.

(COPDGene) study, an observational study that enrolled current or former smokers. COPDGene controls were matched on lung function, age, sex and race. These consisted of 17 individuals without COPD and 20 individuals with COPD (table 1). This study was approved by the University of Minnesota Institutional Review Board. Table 1 Characteristics of cases and controls HIV(±) COPD(±) (n=38) HIV(±) COPD(−) (n=40) HIV(−) COPD(±) (n=20) HIV(−) COPD(−) (n=17) Age, years 38.97±9.21* 38.93±7.78* 48.18±3.67* 55.91±5.48 Sex 27M/11F 29M/11F 18M/2F 15M/2F FEV1pp 70.95±11.6* 94.90±15.6* 73.08±12.8* 109.6±13.2 FEV1/FVC 0.663±0.0704* 0.815±0.0476 0.618±0.065* 0.818±0.0341 HIV RNA log10, copies/mL 3.95±1.04 3.86±0.994 NA NA CD4+ T-cell count, cells/mm3 707.03±198.89 716.26±182.84 NA NA CD8+ T-cell count, cells/mm3 1129.68±596.74 1142.86±912.3 NA NA CD4/CD8 ratio 0.793±0.395 0.936±0.722 NA NA Smoking status† 16/3/19 16/4/20 16/4/0 0/0/17 Race‡ 15/22/1 16/22/2 15/5/0 15/2/0 Region§ 6/13/17/2 8/13/17/2 20/0/0/0 17/0/0/0 *Significant from COPD(−) HIV(−) using Student's t-test. †Smoking status (current/former/never). ‡Race (Caucasian/black/Latino). §Region (North America/Europe/Africa/South America). COPD, chronic obstructive pulmonary disease; FEV1pp, forced expiratory volume in 1 s, per cent predicted; FVC, forced vital capacity; NA, not available.

HIV(±) COPD(±) (n=38) HIV(±) COPD(−) (n=40) HIV(−) COPD(±) (n=20) HIV(−) COPD(−) (n=17) Age, years 38.97±9.21* 38.93±7.78* 48.18±3.67* 55.91±5.48 Sex 27M/11F 29M/11F 18M/2F 15M/2F FEV1pp 70.95±11.6* 94.90±15.6* 73.08±12.8* 109.6±13.2 FEV1/FVC 0.663±0.0704* 0.815±0.0476 0.618±0.065* 0.818±0.0341 HIV RNA log10, copies/mL 3.95±1.04 3.86±0.994 NA NA CD4+ T-cell count, cells/mm3 707.03±198.89 716.26±182.84 NA NA CD8+ T-cell count, cells/mm3 1129.68±596.74 1142.86±912.3 NA NA CD4/CD8 ratio 0.793±0.395 0.936±0.722 NA NA Smoking status† 16/3/19 16/4/20 16/4/0 0/0/17 Race‡ 15/22/1 16/22/2 15/5/0 15/2/0 Region§ 6/13/17/2 8/13/17/2 20/0/0/0 17/0/0/0 *Significant from COPD(−) HIV(−) using Student's t-test. †Smoking status (current/former/never). ‡Race (Caucasian/black/Latino). §Region (North America/Europe/Africa/South America). COPD, chronic obstructive pulmonary disease; FEV1pp, forced expiratory volume in 1 s, per cent predicted; FVC, forced vital capacity; NA, not available. Solvents and reagents The acetonitrile (ACN) used in this study was of high pressure liquid chromatography (HPLC) grade and purchased from Fisher Scientific (Pittsburg, Pennsylvania, USA). HPLC grade methanol (MeOH) was acquired from Sigma-Aldrich (St Louis, Missouri, USA) and analytical grade formic acid from Fluka (Sigma, St Louis, Missouri, USA). The mass spectrometer lock-spray solution, leucine encephaline, was purchased from Waters (Milford, Massachusetts, USA). Ultra high purity bottled water was procured from Invitrogen (Grand Island, New York, USA).

ich (St Louis, Missouri, USA) and analytical grade formic acid from Fluka (Sigma, St Louis, Missouri, USA). The mass spectrometer lock-spray solution, leucine encephaline, was purchased from Waters (Milford, Massachusetts, USA). Ultra high purity bottled water was procured from Invitrogen (Grand Island, New York, USA). Sample preparation: Aliquots of 100 µL of plasma had a heavy standard of 3 µL of 100 µM of kynurenine (Kyn)-D6 and 3 µL of 1 mM of tryptophan (Trp)-C1113 (Cambridge Isotope Laboratories, Tewksbury, Massachusetts, USA) added in prior to any preparation. The aliquots of plasma were then mixed with 400 µL of ice-cold solvent (100% MeOH), vortexed and placed on ice for 10 min. Samples were then centrifuged at 13 000×g for 10 min at 4°C, and the supernatant was removed and transferred into a clean low-retention phial. This step was repeated once. Samples were concentrated using a vacuum centrifuge to ∼50 µL. Formic acid was used to acidify the plasma samples which were added to the starting buffer used in ultraperformance liquid chromatography (5% ACN, 95% water, 0.1% formic acid) to ∼100 µL.

ferred into a clean low-retention phial. This step was repeated once. Samples were concentrated using a vacuum centrifuge to ∼50 µL. Formic acid was used to acidify the plasma samples which were added to the starting buffer used in ultraperformance liquid chromatography (5% ACN, 95% water, 0.1% formic acid) to ∼100 µL. MS analysis For analyte profiling and quantitation, 10 µL of the undiluted sample was injected into Thermo Q-exactive LC-MS (ThermoFisher Scientific, Marietta, Ohio, USA) with Acquity ultra performance liquid chromatography (UPLC) C18 2.1×100 mm, 1.7 µm column (Waters, Milford, Massachusetts, USA) at 40˚C. All of the samples were subjected to a gradient going from buffer A to B over 15 min then flushing for 5 min. Buffer A consisted of 99.9% water with 0.1% formic acid, while buffer B was 99.9% ACN with 0.1% formic acid. To avoid batch effects, all samples were run in one continuous MS run. For quality control, internal standards consisted of blank samples spiked with four metabolites (phenylalanine-D6, hippuric acid-D5, Kyn-D6 and Trp-C11) and were placed at random intervals in the MS run. For quantitation, the individual samples were spiked with the same four metabolites for normalisation of the peak intensities. For metabolite identification, two samples from each group (±COPD, ±HIV) were run on the Thermo Q-exactive LC-MS in MS/MS data-independent acquisition mode with a 25 min of gradient A to B followed by 5 min of flushing.

vidual samples were spiked with the same four metabolites for normalisation of the peak intensities. For metabolite identification, two samples from each group (±COPD, ±HIV) were run on the Thermo Q-exactive LC-MS in MS/MS data-independent acquisition mode with a 25 min of gradient A to B followed by 5 min of flushing. MS data processing The .RAW files from the Q-Exactive were converted into mzXML using MSconvert (http://proteowizard.sourceforge.net/tools.shtml), and processed using the XCMS algorithm as implemented in the xcms package for R using the recommended pipeline (the symmetric family was used for retention time (RT) correction). Data from each sample were normalised to the mean of the heavy standards we introduced for Trp (mass/charge (m/z)=215.1231 and RT=1.841226) and Kyn (m/z=216.1318 and RT=1.941770), so that on the log scale the means of these two analytes were zero for all individuals. LC-MS/MS analysis for analyte identification The .RAW files from Q-exactive LC-MS/MS were analysed using Thermo Xcalibur Qual Browser for fragmentation data of the analytes of interest. These analytes were matched to fragmentation spectra based on m/z and RT matching between LC-MS and LC-MS/MS. The fragmentation data for all of these analytes were compared and matched to metabolites via Metlin's MS/MS metabolite database within a 12 ppm difference.

r for fragmentation data of the analytes of interest. These analytes were matched to fragmentation spectra based on m/z and RT matching between LC-MS and LC-MS/MS. The fragmentation data for all of these analytes were compared and matched to metabolites via Metlin's MS/MS metabolite database within a 12 ppm difference. LC-MS/MS selective reaction monitoring analysis of Trp and Kyn Diluted samples (20 µL, diluted 1:1000 for Trp and 1:100 for Kyn) were subjected to injection using an Agilent autosampler LC-MS/MS with an analytical Waters Symmetry C18, 3.5 µm column connected to the 5500 iontrap (Sciex, Framingham, Massachusetts, USA) fitted with a turbo V electrospray source. The samples were subjected to a linear gradient of 2% ACN, 0.1% formic acid to 98% ACN 0.1% formic acid for 10 min at a column flow rate of 250 µL/min. Transitions monitored are in online supplementary table 1S. The data were analysed using MultiQuant (Applied Biosystems, Foster City, California, USA) providing the peak area for the transitions. A standard curve was constructed using concentration ratios of Trp/Trp and Kyn /Kyn from fentomole to nanomole in 20 µL. 10.1136/bmjresp-2017-000180.supp1supplementary table 1

LC-MS/MS selective reaction monitoring analysis of Trp and Kyn Diluted samples (20 µL, diluted 1:1000 for Trp and 1:100 for Kyn) were subjected to injection using an Agilent autosampler LC-MS/MS with an analytical Waters Symmetry C18, 3.5 µm column connected to the 5500 iontrap (Sciex, Framingham, Massachusetts, USA) fitted with a turbo V electrospray source. The samples were subjected to a linear gradient of 2% ACN, 0.1% formic acid to 98% ACN 0.1% formic acid for 10 min at a column flow rate of 250 µL/min. Transitions monitored are in online supplementary table 1S. The data were analysed using MultiQuant (Applied Biosystems, Foster City, California, USA) providing the peak area for the transitions. A standard curve was constructed using concentration ratios of Trp/Trp and Kyn /Kyn from fentomole to nanomole in 20 µL. 10.1136/bmjresp-2017-000180.supp1supplementary table 1 Statistics To test for differences in metabolite concentrations, mixed-effects models were used that included fixed effects for HIV status, COPD status, the interaction between HIV and COPD status, FEV1 per cent predicted, race and smoking status and a random effect for match group. All matched groups were matched on age and sex. For analysis of individual metabolites, a logarithmic transformation was employed prior to model fitting, whereas the component of the targeted analysis which focused on the ratio of Kyn to Trp did not employ such a transformation. Otherwise, the same approach was employed for the analysis of the metabolites identified in the untargeted analysis as for the targeted analysis, except that we report an unadjusted p value for the targeted analysis, whereas we use the q-value approach to control the false discovery rate (FDR) at 10% for the untargeted analysis. V.3.2.3 of the statistical software package R was used for all statistical analyses.

untargeted analysis as for the targeted analysis, except that we report an unadjusted p value for the targeted analysis, whereas we use the q-value approach to control the false discovery rate (FDR) at 10% for the untargeted analysis. V.3.2.3 of the statistical software package R was used for all statistical analyses. Analyte bioinformatics MS/MS spectra .Raw files were observed using Thermo Xcalibur Qual Browser. MS/MS spectra for each analyte of interest were matched to metabolite spectra through the Metlin metabolite database from Scripps (https://metlin.scripps.edu/index.php). A degree of restraint, within 10% of the RT, was used when matching to feasible options based on RT of the analyte. Analytes illuted in the first 30s and after 15 min were eliminated. Results Study participant characteristics Cases consisted of 38 individuals with HIV-COPD (table 1). HIV(+) controls were 40 individuals with HIV, but without COPD, matched on age, sex, smoking status, region and race. Baseline plasma samples were drawn prior to ART initiation or any AIDS-defining illness. All had CD4+ T-cell counts >500 cells/mm3 and there were no significant differences in plasma HIV RNA, CD4+ and CD8+ T-cell counts between HIV-positive cases and controls. The HIV(−) controls were significantly older than the HIV-positive participants.

es were drawn prior to ART initiation or any AIDS-defining illness. All had CD4+ T-cell counts >500 cells/mm3 and there were no significant differences in plasma HIV RNA, CD4+ and CD8+ T-cell counts between HIV-positive cases and controls. The HIV(−) controls were significantly older than the HIV-positive participants. Metabolomics analyte profiles MS analysis detected 1689 plasma analyte signals, defined as a discreet m/z and RT, that correlated with a yet unknown metabolite. Since there is no universally accepted methodology to quantify analytes detected by mass spectrometry, we chose relative abundance as the sum of all peak intensities detected by the mass spectrometer that associated with the given analyte. Using a mixed-effects model with fixed effects for COPD status, HIV status, their interaction, FEV1 per cent predicted, race, sex and smoking status, we identified 880 analytes at 10% FDR that were attributable to the main effect of HIV status, that is, differences due to HIV averaging over COPD status; we identified no analytes that differed at 10% FDR for the main effect of COPD. Tests of the statistical interaction between HIV status and COPD status identified 263 analytes that displayed differences of interest (ie, a q-value <0.1). Of these 263 analytes, 232 (88%) also had significant main effects for HIV status. On examination of the estimated regression coefficients for HIV status and the interaction between HIV and COPD status of the 263 analytes, it was found that differences between HIV(+) and HIV(−) participants are, on average, smaller in COPD(+) participants compared with COPD(−) participants. This is reflected in figure 1 that displays the raw data for 25 analytes with the lowest q-values that met our FDR criterion for the interaction term. After eliminating isotopes and contaminants, 172 of the original 263 analytes differed based on HIV and COPD status and therefore qualified for further analysis described next.

reflected in figure 1 that displays the raw data for 25 analytes with the lowest q-values that met our FDR criterion for the interaction term. After eliminating isotopes and contaminants, 172 of the original 263 analytes differed based on HIV and COPD status and therefore qualified for further analysis described next. Figure 1 The top 25 analytes by relevant q-value differentially expressed in HIV with COPD. The top number is m/z and the bottom number is RT. Y-axis is intensity. (A) HIV− COPD−, (B) HIV+ COPD−, (C) HIV− COPD+, (D) HIV+ COPD+. COPD, chronic obstructive pulmonary disease; m/z, mass/charge; RT, retention time. Metabolite identification To identify some of the significant analytes of interest, we performed LC-MS/MS targeting the 172 analytes we identified that differed in HIV-COPD. We were able to attribute 17 of these analyte spectra to lipids, including sphingolipids, diacylglycerol and fatty acids (table 2, figure 2, online supplementary table 2S). Ceramide (m/z 560.50948) concentrations were higher in HIV-COPD compared with +HIV/normal lung function, but did not quite meet our FDR threshold (q-value=0.11). Diacylglycerol was lower in HIV-COPD compared with +HIV/normal lung function and similar to those without HIV. We identified three lipids as fatty acids and one as a phospholipid. Generally, the fatty acids were higher in the HIV-COPD persons compared with +HIV/normal lung function, whereas the different sphingolipids demonstrated both increased and decreased concentrations. Table 2 Metabolites identified by LC-MS/MS

Metabolite identification To identify some of the significant analytes of interest, we performed LC-MS/MS targeting the 172 analytes we identified that differed in HIV-COPD. We were able to attribute 17 of these analyte spectra to lipids, including sphingolipids, diacylglycerol and fatty acids (table 2, figure 2, online supplementary table 2S). Ceramide (m/z 560.50948) concentrations were higher in HIV-COPD compared with +HIV/normal lung function, but did not quite meet our FDR threshold (q-value=0.11). Diacylglycerol was lower in HIV-COPD compared with +HIV/normal lung function and similar to those without HIV. We identified three lipids as fatty acids and one as a phospholipid. Generally, the fatty acids were higher in the HIV-COPD persons compared with +HIV/normal lung function, whereas the different sphingolipids demonstrated both increased and decreased concentrations. Table 2 Metabolites identified by LC-MS/MS m/z RT (s) Adduct Accurate mass Formula Tentative identification q-Value 245.22577 571.05 M+H-H2O 262.2297 C18H30O Fatty acid 0.035141 261.22061 537.33 M+H-H2O 278.2246 C16H32O2 Fatty acid 0.002931 262.25198 568.04 M+H-H2O 279.2562 C18H33NO Linoleamide 0.031476 263.23627 571.16 M+H 262.2297 C18H32O2 Fatty acid 0.023163 268.26277 582.06 M+H-H2O 268.2640 C17H35NO2 Sphingolipid 0.014423 270.27842 641.51 M+H-H2O 287.2824 C17H37NO2 Sphingolipid 0.010165 280.26257 571.35 M+H-H2O 297.2668 C18H33NO Sphingolipid 0.046177 293.28267 697.97 M+Na 270.2923 C18H38O Octadecanol 0.024386 296.29396 656.67 M+H-H2O 313.2981 C19H39NO2 Sphingolipid 0.021558 306.27791 491.34 M+Na 283.2875 C18H37NO Stearamide 0.002926 310.30938 698.11 M+H-H2O 327.3137 C20H41NO2 Sphingolipid 0.021558 312.31609 698.29 M+H-H2O 329.3197 C20H43NO2 Sphingolipid 0.021623 338.34061 791.42 M+H 337.3345 C22H43NO Docosenamide 0.005362 340.35618 879.47 M+H 339.3501 C22H45NO Docosenamide 0.027448 560.50948 568.79 M+H-H2O 577.5070 C36H67NO4 Ceramide 0.113237 613.49004 493.11 M+H 612.4754 C39H65O5 Diacylglycerol 0.014779 854.57988 533.96 M+Na 831.5989 C45H86NO10P Phospholipid 0.006038 m/z, Mass to charge ratio of metabolite; RT, retention time (s) in mass spectrometry column.

9.3501 C22H45NO Docosenamide 0.027448 560.50948 568.79 M+H-H2O 577.5070 C36H67NO4 Ceramide 0.113237 613.49004 493.11 M+H 612.4754 C39H65O5 Diacylglycerol 0.014779 854.57988 533.96 M+Na 831.5989 C45H86NO10P Phospholipid 0.006038 m/z, Mass to charge ratio of metabolite; RT, retention time (s) in mass spectrometry column. Figure 2 Lipids differentially expressed in HIV. The top number is m/z and the bottom number is RT. Y-axis is intensity. (A) HIV− COPD−, (B) HIV+ COPD−, (C) HIV− COPD+, (D) HIV+ COPD+. COPD, chronic obstructive pulmonary disease; m/z, mass/charge; RT, retention time. 10.1136/bmjresp-2017-000180.supp2supplementary table 2

9.3501 C22H45NO Docosenamide 0.027448 560.50948 568.79 M+H-H2O 577.5070 C36H67NO4 Ceramide 0.113237 613.49004 493.11 M+H 612.4754 C39H65O5 Diacylglycerol 0.014779 854.57988 533.96 M+Na 831.5989 C45H86NO10P Phospholipid 0.006038 m/z, Mass to charge ratio of metabolite; RT, retention time (s) in mass spectrometry column. Figure 2 Lipids differentially expressed in HIV. The top number is m/z and the bottom number is RT. Y-axis is intensity. (A) HIV− COPD−, (B) HIV+ COPD−, (C) HIV− COPD+, (D) HIV+ COPD+. COPD, chronic obstructive pulmonary disease; m/z, mass/charge; RT, retention time. 10.1136/bmjresp-2017-000180.supp2supplementary table 2 Targeted Trp and Kyn concentrations One of the analytes differentially expressed in HIV+ participants was consistent with the essential amino acid Trp (m/z 205.097). Since Trp catabolism has been associated with both HIV infection and COPD,18–20 and standards are readily available, we sought to quantify Trp and its major metabolite, Kyn (m/z 209.094) by performing targeted selective reaction monitoring (SRM; figure 3). Indoleamine-2,3-dioxygenase (IDO) is the main inducible and rate-limiting enzyme involved in Trp catabolism, and IDO activity is expressed as the Kyn/Trp ratio. We found that Trp was lower in HIV(+) individuals compared with HIV(−) participants as previously reported,21–23 although this was not significant after adjustment for relevant covariates (figure 3A, p=0.0681). Trp was also lower in individuals with COPD, but this was not statistically significant (figure 3A, p=0.39). Kyn was higher in HIV-positive participants but was not statistically significant after adjustment for relevant covariates (figure 3B, p=0.5586), and was lower in COPD participants, albeit not significantly (figure 3B, p=0.1888). As a reflection of IDO activity, the Kyn/Trp ratio was significantly higher in HIV (figure 3C, p=0.022), but there was no difference due to COPD status (p=0.95).

ignificant after adjustment for relevant covariates (figure 3B, p=0.5586), and was lower in COPD participants, albeit not significantly (figure 3B, p=0.1888). As a reflection of IDO activity, the Kyn/Trp ratio was significantly higher in HIV (figure 3C, p=0.022), but there was no difference due to COPD status (p=0.95). Figure 3 (A) Plasma tryptophan concentrations. (B) Plasma kynurenine concentrations. (C) IDO activity as reflected by kynurenine/tryptophan ratio. COPD, chronic obstructive pulmonary disease; IDO, indoleamine-2,3-dioxygenase. Discussion A major gap in knowledge is the limited understanding of why PLWH are more susceptible to developing COPD independent of smoking status.1 2 5 10 Currently, there is no biomarker to identify risk or lend insight into the mechanisms of developing rapid decline in lung function and subsequent COPD in PLWH. The pulmonary substudy of START offers a unique opportunity to identify biomarkers of HIV-COPD prior to the development of an AIDS-defining illness. This is important since pulmonary infection itself can lead to a loss of lung function.24 25 In this exploratory study using an untargeted metabolomic approach, we identified 172 unique plasma analytes that associated with HIV-COPD compared with participants with HIV without COPD after controlling for multiple confounders such as smoking status, race and sex.

on itself can lead to a loss of lung function.24 25 In this exploratory study using an untargeted metabolomic approach, we identified 172 unique plasma analytes that associated with HIV-COPD compared with participants with HIV without COPD after controlling for multiple confounders such as smoking status, race and sex. We performed LC-MS/MS to identify metabolites of interest and found that 17 of these metabolites were lipids. Among these lipids were diacylglycerol and several sphingolipids. Sphingolipids serve as a structural component of the plasma membrane lipid bilayer and also participate in cell recognition and signalling. In individuals with COPD, increased concentrations of sphingolipids have been measured in sputum. Telenga et al26 identified 168 sphingolipids and 36 phosphatidylethanolamine lipids in the sputum that were significantly higher in smokers with COPD compared with smokers without COPD. These lipids correlated with reduced lung function, sputum inflammation and smoking history. In our study, we found a complex expression of sphingolipids with both elevated and decreased plasma concentrations of various sphingolipids. This complex plasma expression of sphingolipids was observed by Bowler et al, where they used both targeted and untargeted metabolomic platforms that identified 23 sphingolipids common to both platforms in COPD. Some of these lipids correlated to COPD phenotype with five associating with emphysema and seven associating with COPD exacerbations.26 27 In addition, sphingolipid expression was complex depending on lipid type and COPD phenotype. Our data support these findings in individuals with HIV, as we found that sphingolipids concentrations were altered in the setting of HIV-COPD when controlling for smoking status.

seven associating with COPD exacerbations.26 27 In addition, sphingolipid expression was complex depending on lipid type and COPD phenotype. Our data support these findings in individuals with HIV, as we found that sphingolipids concentrations were altered in the setting of HIV-COPD when controlling for smoking status. Among the lipids, we also identified ceramide. Although ceramide concentrations were higher in HIV-COPD, in this relatively small exploratory study, the difference did not quite meet statistical significance (q-value=0.11). Ceramide consists of a sphingolipid backbone (sphingosine) plus a fatty acid. It has been implicated in the pathogenesis of COPD. Petrache et al28 found that ceramide acts as a second messenger and a crucial mediator of apoptosis and thus alveolar destruction in a murine model of emphysema. In human studies, elevated concentrations of ceramide have been found in sputum and lung tissue.26 28 29 Bowler et al27 further characterised plasma lipids in various COPD phenotypes and found that ceramides had a negative association with emphysema as measured by quantitative high-resolution CT (HRCT); however, trihexosylceramides had a positive association with frequent exacerbators. We also found an elevation of diacylglycerol, a protein kinase C activator that cooperates with ceramide in inducing apoptosis,30 but we did not have additional studies, such as HRCT, for further phenotyping.

high-resolution CT (HRCT); however, trihexosylceramides had a positive association with frequent exacerbators. We also found an elevation of diacylglycerol, a protein kinase C activator that cooperates with ceramide in inducing apoptosis,30 but we did not have additional studies, such as HRCT, for further phenotyping. Lipid dysregulation is a known complication of AIDS and is a well-known side effect of protease inhibitors. Chetwynd et al31 found that ART resulted in reductions in certain urinary metabolites, including sphingamines. Our study is significant as all of the START participants were ART-naïve and prior to the onset of AIDS. Therefore, the differences in lipid profiles between those with and without COPD cannot be attributed to ART. In addition, lipid abnormalities have been linked to markers of inflammation in HIV compared with healthy controls; however, all of the participants were on HAART and it is unclear whether the lipid abnormalities are due to HIV infection versus ART.32 Pertinent to our study, Scarpelini et al33 identified metabolites that associated with rapid progression of HIV infection and these included sphingomyelin metabolism. It is possible that lipid dysregulation in HIV infection is a marker of worsened disease, including the development of COPD.

HIV infection versus ART.32 Pertinent to our study, Scarpelini et al33 identified metabolites that associated with rapid progression of HIV infection and these included sphingomyelin metabolism. It is possible that lipid dysregulation in HIV infection is a marker of worsened disease, including the development of COPD. In addition to lipids, we identified an analyte consistent with Trp that trended towards an association with HIV-COPD in our metabolite profiling. Trp and its main catabolic enzyme, IDO, have been associated with both COPD and HIV infection.18 20 Therefore, we used targeted SRM to measure Trp and its major metabolite, Kyn, to determine if the Trp catabolism is a biomarker for HIV-COPD. Consistent with previous reports, we found decreased Trp and increased Kyn concentrations in individuals with HIV. As a reflection of IDO activity, the Kyn/Trp ratio was significantly elevated in the setting of HIV, but not in COPD participants. Although not statistically significant, we did find that Kyn concentrations were higher than one would predict in those with both COPD and HIV (data not shown). One possibility is that the effects of HIV on IDO activation are much larger than those of COPD, therefore requiring a larger cohort to determine if IDO activation plays a role in HIV-COPD.

nificant, we did find that Kyn concentrations were higher than one would predict in those with both COPD and HIV (data not shown). One possibility is that the effects of HIV on IDO activation are much larger than those of COPD, therefore requiring a larger cohort to determine if IDO activation plays a role in HIV-COPD. There are several limitations to this study. While HIV-positive controls were matched on age, sex, race, region and smoking status to our cases, our non-HIV controls were difficult to match on age and race as our HIV-COPD cases were much younger and over half were black. In this study, we found that metabolites associated with HIV-COPD were not associated with COPD in HIV-negative controls. This may be due to confounders, such as age and race; however, COPD in HIV-negative individuals is most likely a very different phenotype and disease. Therefore, diagnosis and treatment may need to be very different for HIV-COPD and these metabolites have the potential to identify those with HIV who are at risk of developing COPD. This study was relatively small and may have been underpowered to determine all significant biomarkers and metabolomics pathways of HIV-COPD. Future studies would benefit from larger sample sizes and longitudinal study designs such as that recently completed and now possible in the START Pulmonary Substudy.

oping COPD. This study was relatively small and may have been underpowered to determine all significant biomarkers and metabolomics pathways of HIV-COPD. Future studies would benefit from larger sample sizes and longitudinal study designs such as that recently completed and now possible in the START Pulmonary Substudy. Conclusion HIV-COPD has a unique plasma metabolomic profile that includes sphingolipids and fatty acids. Additional studies are needed to determine how such metabolic pathways contribute to HIV-COPD and whether therapeutic interventions to alter sphingolipid expression can reduce the risk of COPD in PLWH. The authors wish to thank the START participants and investigators. A full list of START investigators can be found in N Engl J Med 2015;373:795–807 and a full list of START Pulmonary Substudy investigators can be found in Lancet Respir Med 2016;4(12):980–989. Contributors: SHo, TJG, CR, SHa, BAW, BJS, KMK and CHW provided substantial contributions to the conception and design of the work, and the acquisition, analysis and interpretation of data. SHo, CR, SHa, BAW, BJS, KMK and CHW were involved in drafting the work or revising it critically for important intellectual content. SHo, TJG, CR, SHa, BAW, BJS, KMK and CHW were involved in final approval of the version published. SHo, TJG, CR, SHa, BAW, BJS, KMK and CHW were involved in the agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

KMK and CHW were involved in final approval of the version published. SHo, TJG, CR, SHa, BAW, BJS, KMK and CHW were involved in the agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Funding: The project described was supported by the University of Minnesota Developmental Center for AIDS Research (Research Award). START and its Pulmonary Substudy were funded by the National Heart, Lung and Blood Institute (USA, R01 HL096453), the National Institute of Allergy and Infectious Diseases (USA, UM1AI068641 and UM1AI120197), Agence Nationale de Recherches sur le SIDA et les Hipatites Virales (France), National Health and Medical Research Council (Australia), National Research Foundation (Denmark), Bundes ministerium f|r Bildung und Forschung (Germany), European AIDS Treatment Network, Medical Research Council (UK), National Institute for Health Research, National Health Service (UK), and the University of Minnesota. Antiretroviral drugs were donated to the START central drug repository by AbbVie, Bristol-Myers Squibb, Gilead Sciences, GlaxoSmithKline/ViiV Healthcare, Janssen Scientific Affairs and Merck. COPDGene was supported by the National Heart, Lung and Blood Institute (USA, R01HL089897 and R01HL089856) and the COPD Foundation through contributions made to an Industry Advisory Board comprised of AstraZeneca, Boehringer Ingelheim, Novartis, Pfizer, Siemens, Sunovion, and GlaxoSmithKline.

ientific Affairs and Merck. COPDGene was supported by the National Heart, Lung and Blood Institute (USA, R01HL089897 and R01HL089856) and the COPD Foundation through contributions made to an Industry Advisory Board comprised of AstraZeneca, Boehringer Ingelheim, Novartis, Pfizer, Siemens, Sunovion, and GlaxoSmithKline. Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health, the Department of Veterans Affairs, the US Government, other funding agencies or any of the authors' affiliated institutions or organisations. Competing interests: None declared. Ethics approval: University of Minnesota Institutional Review Board. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

key messages Internal migration process could be related with the increase and differences in asthma prevalence between urban and rural areas of Latin-America. Few studies have evaluated the effect of internal migration on asthma in an urban area of low-income and middle-income countries. Our findings show how some migrant categories based on temporal and spatial characteristics area associated with asthma/wheeze prevalence. Introduction Over the past 40 years or so, there has been a progressive increase in the prevalence of asthma and other allergic diseases particularly in high-income countries (HICs) and in urban areas of HICs and low-income and middle-income countries (LMICs).1 However, in recent years the prevalence of allergic disorders may have reached a plateau in some HICs but continues to increase in LMICs.2 The reasons for these variations remain unexplained but are likely to be caused by a complex interplay of biological, environmental and social factors.3 4

income countries (LMICs).1 However, in recent years the prevalence of allergic disorders may have reached a plateau in some HICs but continues to increase in LMICs.2 The reasons for these variations remain unexplained but are likely to be caused by a complex interplay of biological, environmental and social factors.3 4 Urbanisation is a social process that has been causally implicated in trends of increasing asthma prevalence in LMICs,3–6 and the lower prevalence in rural compared with urban populations.3 7 Rural to urban differences in asthma prevalence have been attributed to the protective effects of environmental exposures such as farming that are typical of a rural way of life.8 However, recent studies have shown that allergic disorders may be increasing in rural areas thus reducing the prevalence gap between urban and rural settings.9 10 Furthermore, an array of environmental and social changes stemming from the urbanisation process have been identified as potential risk factors for asthma in urban and rural areas.4 8

ve shown that allergic disorders may be increasing in rural areas thus reducing the prevalence gap between urban and rural settings.9 10 Furthermore, an array of environmental and social changes stemming from the urbanisation process have been identified as potential risk factors for asthma in urban and rural areas.4 8 Rural to urban migration and migration between cities are important components of the urbanisation process in LMICs and are related to environmental, socioeconomic and behavioural changes.11 However, few epidemiological studies have investigated the influence of internal migration on asthma and other allergic diseases in LMICs.12 13 Most studies of the effects of migration on allergic diseases have investigated populations migrating from LMICs (presumed low risk for allergic diseases) to HICs (presumed high risk).14 15 These studies have shown that being born in an area of low risk provides protection against asthma,16 17 but this protection may decline with the length of residence in the new environment.18 Others studies have shown that age of migration and time since migration are associated with the risk of asthma and other allergic diseases,14 often leading to a higher risk of atopy and allergic diseases among migrants than the local population.

ection may decline with the length of residence in the new environment.18 Others studies have shown that age of migration and time since migration are associated with the risk of asthma and other allergic diseases,14 often leading to a higher risk of atopy and allergic diseases among migrants than the local population. Although internal migration can be defined as the movement of an individual between two geographical locations (rural to urban or urban to urban) of the same country or region,19 the social, economic and environmental conditions of migrant populations transform this simple movement into a more complex process with different effects on health.20 Temporal, spatial and social characteristics of the population produce different groups of migrants each with specific features with potentially differing effects on asthma risk. Further, the consequences of migration are relevant to individual migrants and to their families and communities both in place of origin and destination.21 A better understanding how these factors, relating to the migration process may alter risk of asthma and other allergic diseases, may contribute to our  comprehension of the causes of temporal increases in asthma prevalence and the differences in prevalence between urban and rural populations of Latin America (LA).22 23 The aim of the present study was to explore the effects of internal migration on the prevalence of wheeze in schoolchildren living in a coastal city in Ecuador.

sion of the causes of temporal increases in asthma prevalence and the differences in prevalence between urban and rural populations of Latin America (LA).22 23 The aim of the present study was to explore the effects of internal migration on the prevalence of wheeze in schoolchildren living in a coastal city in Ecuador. Methods Study population The study was conducted in the city of Esmeraldas, the provincial capital of the tropical coastal province of Esmeraldas in north-western Ecuador, located 140 km south of the Colombian border. With approximately 190 000 inhabitants, Esmeraldas is the principal northern port and is home to the country’s largest oil refinery. The main economic and industrial activities of the population are based on oil processing and export, commerce, agriculture (especially tropical fruit and palm oil), timber, fishing and tourism. Based on the last national census of 2010, the coverage of basic services in the city is deficient: 28% of the households have no access to running water, 22% lack a sewage system and 5% have no access to electricity. The educational level of the population is low compared with the national average: 6% of the population is illiterate while only 18% has higher education.24

e of basic services in the city is deficient: 28% of the households have no access to running water, 22% lack a sewage system and 5% have no access to electricity. The educational level of the population is low compared with the national average: 6% of the population is illiterate while only 18% has higher education.24 Study design and sample A cross-sectional study was conducted in schoolchildren aged 5–16 years to evaluate risk factors for allergy and asthma and history of migration.12 A convenience sample of 10 schools was selected from nine barrios or neighbourhoods in which there was a predominance of Afro-Ecuadorian migrants from the rural districts of the province. . All children attending the schools at the time of the survey were eligible for inclusion. The response rate was 90.8% (of the annually updated school lists). Data collection was done between November 2007 and January 2010. Data collection Detailed information on risk factors and symptoms of wheeze was collected using a questionnaire based on the International Study of Asthma and Allergies in Childhood (ISAAC) phase II.25 The questionnaire, translated into Spanish and adapted to local conditions, was administered to the parent or guardian of each child by trained field workers. Wheeze was defined as a positive response to the question ‘Has your child had wheeze in the chest in the last 12 months’. The questionnaire was also used to collect information on migration history of the children and their parents.

ons, was administered to the parent or guardian of each child by trained field workers. Wheeze was defined as a positive response to the question ‘Has your child had wheeze in the chest in the last 12 months’. The questionnaire was also used to collect information on migration history of the children and their parents. Migrant categories Migration data were analysed based on the recommendations of the United Nations Secretariat to measure internal migration.19 Detailed information about place of birth (community/city, parish, province and country) and temporal characteristics of migratory movements were collected for each child. Migration was defined as a change of residence from one civil division to another in which community was the minor division. Migration status was measured based on the place of birth of the children by the question ‘Where was the child born?’ Children who had been born in the city of Esmeraldas were treated as non-migrants and all others as migrants. Migrants were classified into several categories based on spatial and temporal characteristics of their migratory movements: (1) Direction of migration classified migrants by direction of migration between place of birth and current place of residence (rural to urban or urban to urban), (2) Locality of migration classified migrants by the population size at the place of birth. (Categories explained in table 1), (3). Age of migration classified migrants by age when they left their place of birth (≤3 years and >3 years), and (4) Time since migration classified migrants by time spent in the city of Esmeraldas (<3 years, 3–5 years and >5 years). Migrant status of parents was also included. Categories and definitions are provided in table 1. Parents were treated as non-migrants if they were born in the city of Esmeraldas and as migrants if they were born elsewhere. Due to the proximity of the city of Esmeraldas to the international border with Colombia (140 km), the variable ‘Colombian children’ was included also representing children with a history of migration from Colombia as first-generation or second-generation international migrants (ie, children born in Colombia or born to migrant Colombian parents).

city of Esmeraldas to the international border with Colombia (140 km), the variable ‘Colombian children’ was included also representing children with a history of migration from Colombia as first-generation or second-generation international migrants (ie, children born in Colombia or born to migrant Colombian parents). Additional data collected for each child included age, sex, parents living in the child’s home (both, one, none), farm environment (households characterised by farming activities and presence of peridomestic animals), type of house (urban house: residences with connection to running water, concrete building materials for walls, presence of a flushing toilet, ownership of a set of electric appliances and two to three urban services; transitional house: residences with an incomplete set of electric appliances, latrine for bathroom, use of mixed materials for house construction) and consumption of junk food (consumption of fizzy drinks). Variables to represent farm environment and type of housing were created using multiple correspondence analysis (MCA) and methodology is explained elsewhere.26 The variables used in MCA to define ‘farm environment’ were parental agricultural activities, contact with animals in farms and animal breeding in or around the household. Variables to define ‘type of house’ were basic services, disposal of faeces, electrical appliances, household construction materials and type of cooking fuel (table 1). Table 1 History of migration, demographic variables and socioeconomic variables

Additional data collected for each child included age, sex, parents living in the child’s home (both, one, none), farm environment (households characterised by farming activities and presence of peridomestic animals), type of house (urban house: residences with connection to running water, concrete building materials for walls, presence of a flushing toilet, ownership of a set of electric appliances and two to three urban services; transitional house: residences with an incomplete set of electric appliances, latrine for bathroom, use of mixed materials for house construction) and consumption of junk food (consumption of fizzy drinks). Variables to represent farm environment and type of housing were created using multiple correspondence analysis (MCA) and methodology is explained elsewhere.26 The variables used in MCA to define ‘farm environment’ were parental agricultural activities, contact with animals in farms and animal breeding in or around the household. Variables to define ‘type of house’ were basic services, disposal of faeces, electrical appliances, household construction materials and type of cooking fuel (table 1). Table 1 History of migration, demographic variables and socioeconomic variables Dimensions/ indicators Definitions Categories n Total population Migrant population Migrant status Classified children by difference between place of birth and place of residence NMa 1694 68.7% Migrant 772 31.3% 100% Direction of migration Classified migrants by migration movement between place of birth and place of residence based on the political division of Ecuador NM 1694 68.7% Rural to urban 286 11.6% 37% Urban to urban 486 19.7% 63% Locality of migration Classified migrants by size of population of place of birth: communities (rural village), small city (urban towns), medium city (provincial capitals) and large city (migrants from Quito and Guayaquil, the two largest cities in the country) NM 1694 68.7% Community 286 11.6% 37% Small city 170 6.9% 22% Medium city 124 5% 16% Large city 192 7.8% 25% Age at migration Classified migrants considering the age when children left their place of birth (years) NM 1694 68.7% ≤3 291 11.8% 38% >3 481 19.5% 62% Time since migration Classified migrants by time spent in current locality (years) NM 1694 68.7% <3 275 11.2% 36% 3–5 178 7.2% 23% >5 319 12.9% 41% Migration status of parents Parents were treated as non-migrants if they were born in the city of Esmeraldas and as migrants if they were born elsewhere NM 575 24% One migrant parent 877 36% Two migrant parents 981 40% Colombian children First-generartion or secondgeneration migrants from Colombia No 2332 92.9% Yes 103 4.1% Age of the children Age of the children (years) ≤9 1294 51.6% >10 1216 48.4% Sex Sex of the child Male 1321 52.7% Female 1189 47.4% Farm environment Classified houses of the children based on farm characteristics (eg, presence of animals and agricultural activities)26 No 2141 85.3% Yes* 369 14.7% Type of house Classified houses of the children based on construction materials, presence of urban services and electrical appliances26 Urban† 2190 87.3% Transitional‡ 320 12.7% Consumption of junk food Classified children by fizzy drink consumption Barely 738 33.7% Sometimes/week 1132 51.6% Daily 323 14.7% Parents living in the child’s house Classified children based on the presence of parents at home Both 1138 45% One 889 36% None 483 19% *Houses characte

† 2190 87.3% Transitional‡ 320 12.7% Consumption of junk food Classified children by fizzy drink consumption Barely 738 33.7% Sometimes/week 1132 51.6% Daily 323 14.7% Parents living in the child’s house Classified children based on the presence of parents at home Both 1138 45% One 889 36% None 483 19% *Houses characte rised by farm activities and peridomestic animal breeding. †Urban house: residences with connection to running water, concrete building materials for walls, use of a flushing toilet, ownership of a set of appliances and two to three urban services. ‡Transitional: residences with an incomplete set of electric appliances, latrine bathroom and use of mixed materials in house construction. a NM: no migrant Statistical analysis Logistic regression was used to explore associations between migrant categories and wheeze in the last 12 months. Each migrant category was adjusted for age, sex, farm environment, type of house and consumption of junk food. Associations with p<0.05 were considered statistically significant. All analyses were done using SPSS V.23. Results We evaluated a total of 2510 schoolchildren in the city of Esmeraldas of whom 44 were excluded due to lack of information on child’s birthplace. The prevalence of wheeze in the last 12 months was 9.4%.

Statistical analysis Logistic regression was used to explore associations between migrant categories and wheeze in the last 12 months. Each migrant category was adjusted for age, sex, farm environment, type of house and consumption of junk food. Associations with p<0.05 were considered statistically significant. All analyses were done using SPSS V.23. Results We evaluated a total of 2510 schoolchildren in the city of Esmeraldas of whom 44 were excluded due to lack of information on child’s birthplace. The prevalence of wheeze in the last 12 months was 9.4%. History of migration Almost a third (31.3%) of schoolchildren were migrants (table 1). Among migrants, 63% had a history of urban to urban migration and 37% of rural to urban migration, while 25% came from large cities (Quito or Guayaquil), 16% from medium cities (provincial capitals), 22% from small cities (city towns) and 37% from rural communities. Age at migration was most frequently reported after 3 years of age (62%) and most migrants (41%) had lived more than 5 years in the city of Esmeraldas. Of the total study population, 4% were first-generation or second-generation international migrants from Colombia, and history of parental migration was reported for 76%.

e at migration was most frequently reported after 3 years of age (62%) and most migrants (41%) had lived more than 5 years in the city of Esmeraldas. Of the total study population, 4% were first-generation or second-generation international migrants from Colombia, and history of parental migration was reported for 76%. Sociodemographic characteristics of the migrant population A higher proportion of migrants than non-migrants was female (45.1 vs 52.6%) while rural to urban migrants tended to be older than the other groups (table 2). Farm environment was more common in migrants (17.3%) than non-migrants (13.6%). Only 25% of the rural to urban migrant children were living with both parents compared with 52% for non-migrant children. Daily consumption of junk food was greater in non-migrants. Table 2 Demographic and socioeconomic variables by direction of migration Variables Categories Direction of migration of children Non-migrant Rural to urban Urban to urban χ2

Sociodemographic characteristics of the migrant population A higher proportion of migrants than non-migrants was female (45.1 vs 52.6%) while rural to urban migrants tended to be older than the other groups (table 2). Farm environment was more common in migrants (17.3%) than non-migrants (13.6%). Only 25% of the rural to urban migrant children were living with both parents compared with 52% for non-migrant children. Daily consumption of junk food was greater in non-migrants. Table 2 Demographic and socioeconomic variables by direction of migration Variables Categories Direction of migration of children Non-migrant Rural to urban Urban to urban χ2 n % n % n % p Value Migration status of parents NM 466 28.4% 20 7.2% 75 15.9% <0.001 One migrant parent 600 36.6% 83 30.1% 182 38.5% Two migrant parents 574 35.0% 173 62.7% 216 45.7% Colombian children No 1598 97.4% 270 97.8% 420 88.4% <0.001 Yes 42 2.6% 6 2.2% 55 11.6% Age of children (years) ≤9 901 53.2% 121 42.3% 246 50.6% 0.003 >10 793 46.8% 165 57.7% 240 49.4% Sex of children Male 930 54.9% 130 45.5% 239 49.2% 0.003 Female 764 45.1% 156 54.5% 247 50.8% Farm environment Non-farm 1464 86.4% 236 82.5% 403 82.9% 0.059 Farm 230 13.6% 50 17.5% 83 17.1% Type of house Basic urban 1491 88.0% 242 84.6% 416 85.6% 0.148 Transitional 203 12.0% 44 15.4% 70 14.4% Consumption of junk food Barely 525 35.3% 73 31.7% 129 29.9% 0.041 Sometimes/week 733 49.3% 130 56.5% 243 56.4% Daily 230 15.5% 27 11.7% 59 13.7% Parents living in child’s house Both 876 51.7% 70 24.5% 176 36.2% <0.001 One 580 34.2% 103 36.0% 183 37.7% None 238 14.0% 113 39.5% 127 26.1% Associations between history of migration and wheeze Positive associations were observed with direction of migration and locality of migration in unadjusted analyses (table 3). Rural to urban migrant children (or migrants from communities) had greater odds of wheeze than non-migrants (OR 1.66, 95% CI 1.15 to 2.41, p=0.007). Children with a history of migration from Colombia had a higher prevalence of wheeze (OR 1.83, 95% CI 1.05 to 3.18, p=0.032) compared with Ecuadorian children. Multivariable analyses were adjusted for sex, age, farm environment, type of housing and consumption of junk food. Adjusted analyses showed a higher risk of wheeze compared with non-migrant children for the following: (1) Direction and locality of migration—children who migrated from rural (or communities) to urban areas (OR 2.01, 95% CI 1.30 to 3.01, p=0.001) had two times more wheeze than non-migrant children; (2) History of migration of the parents—children for whom both parents had a history of migration had a greater risk of wheeze than children whose parents had not (OR 1.52, 95% CI 1.01 to 2.29, p=0.046) and (3) Colombian children had greater risk of wheeze (OR 2.15, 95% CI 1.22 to 3.78, p=0.008) than Ecuadorian children.

) History of migration of the parents—children for whom both parents had a history of migration had a greater risk of wheeze than children whose parents had not (OR 1.52, 95% CI 1.01 to 2.29, p=0.046) and (3) Colombian children had greater risk of wheeze (OR 2.15, 95% CI 1.22 to 3.78, p=0.008) than Ecuadorian children. Table 3 ORs and 95% CIs for associations between wheeze in the last 12 months and history of migration adjusted for age, sex and socioeconomic variables

) History of migration of the parents—children for whom both parents had a history of migration had a greater risk of wheeze than children whose parents had not (OR 1.52, 95% CI 1.01 to 2.29, p=0.046) and (3) Colombian children had greater risk of wheeze (OR 2.15, 95% CI 1.22 to 3.78, p=0.008) than Ecuadorian children. Table 3 ORs and 95% CIs for associations between wheeze in the last 12 months and history of migration adjusted for age, sex and socioeconomic variables Wheeze Univariable Multivariable Dimensions Categories Prevalence OR (95% CI) p OR* (95% CI) p Migration status Non-migrant 9.1% 1 1 Migrant 10.4% 1.15 (0.86 to 1.53) 0.342 1.25 (0.91 to 1.71) 0.175 Direction of migration Non-migrant 9.1% 1 1 Rural to urban 14.3% 1.66 (1.15 to 2.41) 0.007 2.01 (1.30 to 3.01) 0.001 Urban to urban 8% 0.87 (0.60 to 1.25) 0.443 0.91 (0.61 to 1.63) 0.660 Locality of migration Non-migrant 9.1% 1 1 Communities 14.3% 1.66 (1.15 to 2.41) 0.007 2.01 (1.30 to 3.01) 0.001 Small city 5.3% 0.56 (0.28 to 1.11) 0.095 0.59 (0.29 to 1.20) 0.145 Medium city 12.1% 1.37 (0.78 to 2.40) 0.278 1.68 (0.92 to 3.07) 0.089 Large city 7.8% 0.84 (0.48 to 1.46) 0.540 0.80 (0.42 to 1.52) 0.486 Age at migration Non-migrant 9.1% 1 1 ≤3 vs NM 10.3% 1.14 (0.75 to 1.73) 0.530 1.20 (0.76 to 1.89) 0.437 >3 vs NM 10.4% 1.15 (0.82 to 1.61) 0.410 1.28 (0.86 to 1.87) 0.204 Time since migration (years) Non-migrant 9.1% 1 1 <3 12.4% 1.40 (0.94 to 2.08) 0.095 1.36 (0.87 to 2.13) 0.185 3–5 9% 0.98 (0.57 to 1.68) 0.943 1.12 (0.63 to 1.98) 0.705 >5 9.4% 1.03 (0.68 to 1.55) 0.885 1.23 (0.77 to 1.94) 0.388 History of migration of the parents Non-migrant 7.8% 1 1 One migrant parent 9.5% 1.23 (0.84 to 1.80) 0.282 1.39 (0.92 to 2.11) 0.123 Two migrant parents 10.3% 1.35 (0.94 to 1.95) 0.108 1.52 (1.01 to 2.29) 0.046 Colombian children No 9.1% 1 1 Yes 15.5% 1.83 (1.05 to 3.18) 0.032 2.15 (1.22 to 3.78) 0.008 *OR adjusted by sex, age, farm environment, quality of the house and consumption of junk food.

5% 1.23 (0.84 to 1.80) 0.282 1.39 (0.92 to 2.11) 0.123 Two migrant parents 10.3% 1.35 (0.94 to 1.95) 0.108 1.52 (1.01 to 2.29) 0.046 Colombian children No 9.1% 1 1 Yes 15.5% 1.83 (1.05 to 3.18) 0.032 2.15 (1.22 to 3.78) 0.008 *OR adjusted by sex, age, farm environment, quality of the house and consumption of junk food. Table 4 shows the adjusted associations between wheeze and age at migration and time since migration, stratified by direction of migration. Positive and statistically significant associations were observed for rural to urban migrants but not for urban to urban migrants. Among rural to urban migrants, those who had migrated after 3 years of age had a greater risk of wheeze (OR 2.5, 95% CI 1.56 to 3.67, p<0.001) than non-migrant children and those who had spent less than 3 years and between 3–5 years in the new area of residence had a higher prevalence of wheeze than non-migrant children (OR 2.34, 95% CI 1.26 to 4.33, p<0.007 and OR 3.03, 95% CI 1.49 to 6.15, p=0.002, respectively). Table 4 ORs and 95% CIs for associations between wheeze in the last 12 months and age at migration and time since migration, stratified by direction of migration. ORs adjusted for age, sex and socioeconomic variables

Table 4 shows the adjusted associations between wheeze and age at migration and time since migration, stratified by direction of migration. Positive and statistically significant associations were observed for rural to urban migrants but not for urban to urban migrants. Among rural to urban migrants, those who had migrated after 3 years of age had a greater risk of wheeze (OR 2.5, 95% CI 1.56 to 3.67, p<0.001) than non-migrant children and those who had spent less than 3 years and between 3–5 years in the new area of residence had a higher prevalence of wheeze than non-migrant children (OR 2.34, 95% CI 1.26 to 4.33, p<0.007 and OR 3.03, 95% CI 1.49 to 6.15, p=0.002, respectively). Table 4 ORs and 95% CIs for associations between wheeze in the last 12 months and age at migration and time since migration, stratified by direction of migration. ORs adjusted for age, sex and socioeconomic variables Rural to urban Urban to urban Dimensions Categories OR* (95% CI) p OR* (95% CI) p Age at migration (years) Non-migrant 1 1 ≤3 vs NM 1.05 (0.44 to 2.49) 0.909 1.30 (0.76 to 2.19) 0.319 >3 vs NM 2.51 (1.56 to 3.97) <0.001 0.67 (0.38 to 1.18) 0.164 Time since migration (years) Non-migrant 1 1 <3 2.34 (1.26 to 4.33) 0.007 0.94 (0.51 to 1.73) 0.840 3–5 3.03 (1.49 to 6.15) 0.002 0.41 (0.15 to 1.13) 0.086 >5 1.15 (0.54 to 2.46) 0.714 1.28 (0.74 to 2.20) 0.374 *OR adjusted by sex, age, farm environment, quality of the house and consumption junk food.

38 to 1.18) 0.164 Time since migration (years) Non-migrant 1 1 <3 2.34 (1.26 to 4.33) 0.007 0.94 (0.51 to 1.73) 0.840 3–5 3.03 (1.49 to 6.15) 0.002 0.41 (0.15 to 1.13) 0.086 >5 1.15 (0.54 to 2.46) 0.714 1.28 (0.74 to 2.20) 0.374 *OR adjusted by sex, age, farm environment, quality of the house and consumption junk food. Discussion In the present analysis, we have explored how internal migration affects the prevalence of recent wheeze among schoolchildren living in a city in a coastal tropical area of LA. Clearly, the study of migration is complex because of the spatial and temporal dimensions of migratory movements and also due to different social backgrounds within migrant populations. These characteristics produce several types of migrants each with specific population features.20 21 Additionally, migration is a multistage process with effects not only at the individual level, but also at family and community levels.21 Considering this, we have taken a multidimensional approach to analyse migration, using categories based on temporary and spatial characteristics of the migratory movements of children and their parents. Our data provide evidence that internal migration, specifically rural to urban migration, was associated with a higher risk of wheeze. Further, international history of migration of the children, specifically migrants from Colombia, was associated with a greater prevalence of wheeze: Colombian children had a twofold greater risk of wheeze compared with Ecuadorian children. Our data extend our previous observations on the effects of migration on allergic diseases in Afro-Ecuadorian schoolchildren living in rural communities in Esmeraldas Province in Ecuador13 in which we have shown that migration before 1 year of age and international migration (from nearby areas of the border between Ecuador and Colombia) to a rural community were associated with a higher prevalence of recent wheeze and rhinitis in transitional rural communities. We also observed that the absence of the mother at home, due to temporary or permanent migration, was associated with an increase in the occurrence of wheeze, rhinitis and eczema in rural areas.13

a rural community were associated with a higher prevalence of recent wheeze and rhinitis in transitional rural communities. We also observed that the absence of the mother at home, due to temporary or permanent migration, was associated with an increase in the occurrence of wheeze, rhinitis and eczema in rural areas.13 Although internal migrants account for nearly four times as many individuals as international migrants,21 associations between migration status and asthma and other allergic diseases have generally been investigated in populations migrating between countries, mostly by comparing those that have migrated from LMICs to HICs. Two publications have reviewed studies of differences in prevalence of asthma and other allergic diseases between international migrants, the host population and the population of origin. The first, by Rottem and colleagues who reviewed available literature published before 2003,14 concluded from 14 published studies that international migrants from LMICs to HICs tended to develop more allergies and asthma compared with their populations of origin in a time-dependent fashion, had a greater risk if migration occurred before 2 years of age and were more prone to allergies than the host populations. The second (a systematic review by Cabieses and colleagues),15 evaluated 54 studies of which 41 were published in the last 10 years. The authors concluded that the prevalence of asthma but not ‘allergies’ was lower in migrants compared with the host population and that the prevalence of asthma tended to converge with that of the host population over time. Further, the study also concluded that asthma prevalence was generally higher in the second-generation compared with the first-generation of migrants. Although the overall conclusions of these reviews of published studies were consistent with the premise that migrants from LMICs suffer less asthma symptoms than host populations for a period following migration, not all studies supported such a conclusion. A recent analysis of data from the ISAAC phase III studies that included study centres from both LMICs and HICs indicated that being born outside the country of residence was associated with a lower prevalence of asthma, rhinoconjunctivitis and eczema but only for migrants to affluent countries.27 The results for non-affluent countries showed a higher prevalence of eczema symptoms in migrants and no associations for asthma and rhinoconjunctivitis.27

born outside the country of residence was associated with a lower prevalence of asthma, rhinoconjunctivitis and eczema but only for migrants to affluent countries.27 The results for non-affluent countries showed a higher prevalence of eczema symptoms in migrants and no associations for asthma and rhinoconjunctivitis.27 Studies conducted in Asia and LA have used history of rural residence to evaluate the effects of rural/farm environment on allergic diseases in urban populations.28 29 Although these studies did not focus on the study of migration as a risk factor, they provided a good starting point to evaluate the effects of internal migration on asthma in LMICs. The first study, conducted in Mongolia, showed that subjects aged 10–60 years who relocated from a small rural village into a town, were more likely to develop asthma than subjects who lived in a town from birth, although this trend was not statistically significant.28 Another study conducted in an urban area of Argentina showed that adolescents aged 13–14 years with a history of rural residence had the same prevalence of wheeze compared with those who had always lived in the urban area.29 Our study focused on the effects of internal migration on wheeze prevalence in an urban population of an LMIC, where migrants formed a diverse group including migrants from rural communities, migrants from other urban settings, and those crossing voluntarily or being displaced by civil conflict across the international border with Colombia (figure 1). In this setting, we found that rural to urban migration is an important determinant of a higher risk of wheeze in an urban population. Further, family history of migration was associated with an increase in wheeze prevalence that was especially marked among Colombian children. However, in contrast to previous international comparisons, our data showed that migrants from populations considered to be at low risk for allergic diseases (rural communities) actually had a slightly higher prevalence of wheeze than the host population in the urban area (14.3% vs 9.1%), challenging the assumption that rural residence protects against allergic diseases.4 At the same time, the prevalence of asthma in rural migrants was slightly higher (14.3% vs 10.1%) than that of the population of origin (rural communities located north of the city of Esmeraldas),26 indicating an increase in risk related to the migration process itself.

al residence protects against allergic diseases.4 At the same time, the prevalence of asthma in rural migrants was slightly higher (14.3% vs 10.1%) than that of the population of origin (rural communities located north of the city of Esmeraldas),26 indicating an increase in risk related to the migration process itself. Another important finding was the effect of time since migration and age of migration in rural migrants: the risk of wheeze increased with greater time since arriving in the city up to 5 years, after which the risk disappeared (table 4). Our data also indicate that age of migration was associated an increased risk of wheeze among children who migrated after 3 years of age. Figure 1 Study site. Map of Ecuador showing migrant categories by location. Red square represents Esmeraldas city (study area), green area represents internal migrations in the Esmeraldas Province and yellow area represents internal moves (migrants of other provinces of Ecuador). Yellow, green and black circles represent large, medium and small cities, respectively.

ant categories by location. Red square represents Esmeraldas city (study area), green area represents internal migrations in the Esmeraldas Province and yellow area represents internal moves (migrants of other provinces of Ecuador). Yellow, green and black circles represent large, medium and small cities, respectively. In an international context, the apparently lower prevalence of asthma and allergic diseases in migrants could be explained by the ‘healthy migrant effect’ in which recent migrants, including those migrating from LMICs to HICs, are on average healthier than the native population.30 International migrants are not a random sample of their country of origin but a highly selected group who are able or motivated to deal with the stress, cost and organisation that such a process entails. Individuals or families who migrate are in a relatively advantageous position, whether financial or social. However, migrant health may deteriorate with increasing length of residence in the new country.29 In the case of rural to urban migration, evidence for a healthy migrant effect is limited.31 In our study, socioeconomic variables of the migrant population were not statistically different than those of the non-migrant population. However, the tough social conditions that new rural migrants face in LMIC cities could explain partly the higher prevalence of wheeze/asthma in rural migrants. It is well known that rural migrants move to the cities in search of work or to improve their quality of life. Most rural migrants settle at the periphery of growing cities in areas that lack basic services and infrastructure. Such newly established neighbourhoods are characterised by low quality of life, poor housing and poverty.32 Several studies conducted in urban centres of LA have shown an increased risk of wheeze/asthma to be associated with factors indicative of poverty, dirt and infections.33 34 Thus, the adverse urban environment in which new rural migrants find themselves could increase asthma risk.

y of life, poor housing and poverty.32 Several studies conducted in urban centres of LA have shown an increased risk of wheeze/asthma to be associated with factors indicative of poverty, dirt and infections.33 34 Thus, the adverse urban environment in which new rural migrants find themselves could increase asthma risk. Further, psychosocial stressors arising from the adaptation process in the new environment and family dissolutions consequent to migration could contribute to an increase in wheeze/asthma in migrant populations.35–37 A high proportion of migrants in LMICs are women who provide the primary economic support for their families working in urban areas in unskilled service jobs.38 In our study, 75% of the children with history of rural migration lived in families with one parent or without parents (table 2). As we have seen previously, the absence of parents at home (especially the mother) is an important determinant in the increase of wheeze in children of migrant parents.13 Another factor that could explain the higher prevalence of wheeze in rural to urban migrants could be migration in search of medical attention for asthma (reverse causality), emphasising the importance of the migration process in the spatial and temporal distribution of asthma between urban and rural areas.

parents.13 Another factor that could explain the higher prevalence of wheeze in rural to urban migrants could be migration in search of medical attention for asthma (reverse causality), emphasising the importance of the migration process in the spatial and temporal distribution of asthma between urban and rural areas. The study of migration using various categories provides a better understanding of the possible factors and mechanisms affecting the development of asthma in urban areas of LMICs. For example, in addition to ‘direction of migration’ which classified migrants according to rural or urban birthplaces, we used ‘locality of migration’ to describe more precisely the possible social and physical environment of the previous residence of the children. In our study, for example, migrants from medium cities (capital cities of Ecuadorian provinces) had 68% more wheeze than non-migrants. We also identified an important migration flow from Colombia, especially from Nariño Department, a region that borders the province of Esmeraldas and which is located less than 150 km from the city of Esmeraldas (figure 1). Although this is a group of international migrants, they provided a useful comparison group because of proximity and a social and ecological environment similar to that found in Esmeraldas Province. Children born in Colombia or born of Colombian parents had a greater prevalence of wheeze compared with Ecuadorian children. Many of these migrants are involuntary migrants or refugees fleeing guerrilla and paramilitary violence in Colombia and the higher prevalence of wheeze in this population might be explained by psychosocial stressors related to displacement rather than changes in lifestyles.39–41 However, our study is subject to several limitations. First, the cross-sectional design does not permit assumptions of the direction of causality. Second, misreporting of birthplace related to recall bias and misclassification is possible, especially for boundary changes in the geographical units of study. However, the use of place of birth to define migration status is more precise than previous studies that have either not defined migration status or used other variables (eg, use of ethnic surnames as a surrogate marker for migrant status15). Further, specific information about the history of migration of the parents was limited.

e use of place of birth to define migration status is more precise than previous studies that have either not defined migration status or used other variables (eg, use of ethnic surnames as a surrogate marker for migrant status15). Further, specific information about the history of migration of the parents was limited. Third, not all wheezing is asthma, although this definition is probably more useful in rural populations where access to healthcare is limited and where alternative definitions such as doctor diagnosis may be subject to significant misclassification. Fourth, because selection of schools resulted in a sample of predominantly Afro-Ecuadorian schoolchildren (to represent the original source population of migrants from rural districts in the north of the province), our findings cannot necessarily be generalised to populations of differing ethnic compositions within the province. However, we believe that our findings provide novel insights into how social and demographic factors may affect asthma burden in LMICs.

tion of migrants from rural districts in the north of the province), our findings cannot necessarily be generalised to populations of differing ethnic compositions within the province. However, we believe that our findings provide novel insights into how social and demographic factors may affect asthma burden in LMICs. Clearly, internal migration is a major contributor to the urbanisation process in LMICs and has a direct effect on the prevalence of asthma in urban and rural populations.42 It is possible that the differences in asthma prevalence between urban and rural areas or between LA countries could be in part explained by different rates of urbanisation and migration within the region. As for international migrants, internal migrants also face important economic, social and environmental changes, especially those associated with changes in diet, physical activity, housing, family composition and air pollution, all factors related to asthma risk.43 Special consideration must be given to rural migrants living in poor conditions in their new urban environments, generally slums and informal settlements. Populations living in such environments are exposed to a number of factors that could increase the risk of asthma or exacerbate existing disease such as inadequate housing, increased risk of respiratory infections through overcrowding and high levels of violence.29 Finally, migration results in exposure to a new set of pollutants and allergens, and to new socioeconomic and cultural factors resulting in important changes to the social and family environment in which the new arrivals find themselves and which may contribute to asthma risk. A better understanding of the effects of migration on asthma will allow us to identify potential public health interventions that could be tested with the aim of alleviating the growing burden of asthma disability, particularly among the urban poor of LA.

rrivals find themselves and which may contribute to asthma risk. A better understanding of the effects of migration on asthma will allow us to identify potential public health interventions that could be tested with the aim of alleviating the growing burden of asthma disability, particularly among the urban poor of LA. Conclusion We evaluated the effects of different migrant categories on the prevalence of wheeze in an urban area of LA. Our study provides evidence that rural to urban migration is a risk factor for wheezing in urban schoolchildren. Age of migration and time since migration were associated with an increased risk of wheeze only for rural to urban migrants but not for urban to urban migrants. Temporal, spatial and socioeconomic dimensions of the migration process may have different effects on the prevalence of wheeze/asthma and other allergic diseases. Further studies in different populations living in rural and urban areas of LMICs, that are subject to migration processes, are required in which detailed information is collected at individual, household and community levels.

ss may have different effects on the prevalence of wheeze/asthma and other allergic diseases. Further studies in different populations living in rural and urban areas of LMICs, that are subject to migration processes, are required in which detailed information is collected at individual, household and community levels. The authors thank the Ecuadorian Elimination Programme for Onchocerciasis (Dr Eduardo Gomez, Lcda Raquel Lovato, Lcda Sra Margarita Padilla, Lcda Anabel Ponce, Ing Sandra Barreno, Sra Magdalena Cortez), CECOMET (Dr Gregorio Montalvo and Lcda Monica Marquez) and SNEM (Dr César Diaz) for support in visiting rural communities and urban neighbourhoods in the city of Esmeraldas. The authors also thank the health promoters, teachers, parents and children for their enthusiastic cooperation. The authors also thank the National Government of the Republic of Ecuador and the Ministry of Higher Education, Science, Technology and Innovation (SENESCYT).

d urban neighbourhoods in the city of Esmeraldas. The authors also thank the health promoters, teachers, parents and children for their enthusiastic cooperation. The authors also thank the National Government of the Republic of Ecuador and the Ministry of Higher Education, Science, Technology and Innovation (SENESCYT). The authors thank the Ecuadorian Elimination Programme for Onchocerciasis (Dr Eduardo Gomez, Lcda Raquel Lovato, Lcda Sra Margarita Padilla, Lcda Anabel Ponce, Ing Sandra Barreno, Sra Magdalena Cortez), CECOMET (Dr Gregorio Montalvo and Lcda Monica Marquez) and SNEM (Dr César Diaz) for support in visiting rural communities and urban neighborhoods in the city of Esmeraldas. The authors also thank the health promoters, teachers, parents and children for their enthusiastic cooperation. The authors also thank the National Government of the Republic of Ecuador and the Ministry of Higher Education, Science, Technology and Innovation (SENESCYT). Contributors: Study design: AR, PJC, MEC, LCR, MLB. Data collection: AR, MGV. Data analysis: AR. Draft manuscript: AR, PJC. Manuscript review: AR, PJC, LCR, MLB. Funding: The research was supported by grants from the Wellcome Trust (072405/Z/03/Z and088862/Z/09/Z). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. Competing interests: None declared. Patient consent: Parental consent obtained. Ethics approval: Ethics Committee of the Hospital Pedro Vicente Maldonado, Ecuador. Provenance and peer review: Not commissioned; externally peer reviewed.

Funding: The research was supported by grants from the Wellcome Trust (072405/Z/03/Z and088862/Z/09/Z). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. Competing interests: None declared. Patient consent: Parental consent obtained. Ethics approval: Ethics Committee of the Hospital Pedro Vicente Maldonado, Ecuador. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.

Introduction Lung cancer is the most commonly diagnosed cancer worldwide,1 and surgery offers the best chance of cure or long-term survival. There are variable early stage lung cancer surgical rates across developed countries of between 5% and 21%.2 Patients with resected early stage lung cancer have recurrence rates of between 30% and 75%,3 and it is likely that with increasing interest in and adoption of screening for lung cancer,4 rates of those having surgery for early stage disease may increase. In addition, patients with a previous lung cancer are at high risk of developing metachronous lung cancers at a rate of 1%–5% per annum.5 There is no consensus as to the duration and nature of follow-up following curative surgical resection of lung cancer with the most recent American guidelines citing the evidence base as ‘weak, low quality’.6 Typically, after postoperative follow-up, current practice involves CT of the chest at 6 months and then annually up to 5 years after resection. The American guidelines recommend every 6 months for 2 years,6 although there is a wide variation in practice in across countries with similar healthcare systems. This follow-up is normally hospital based, with respiratory specialists or thoracic surgeons. After lobectomy, local recurrence is most common in the first 2 years with the risk of recurrence nearly twofold higher (HR 1.86; 95% CI 1.01 to 3.41) in persistent smokers.7

n across countries with similar healthcare systems. This follow-up is normally hospital based, with respiratory specialists or thoracic surgeons. After lobectomy, local recurrence is most common in the first 2 years with the risk of recurrence nearly twofold higher (HR 1.86; 95% CI 1.01 to 3.41) in persistent smokers.7 Cost-effective healthcare relies on the provision of appropriate care in the right time at the right place; countries with a strong primary care component have been demonstrated to be more cost-effective than those that are over-reliant on hospital-led services.8 As cancer survival rates increase, the focus on managing patient flow back into primary care is a key area for future-effective and cost-effective cancer care.9 It is recognised that follow-up of patients undergoing palliative surgical interventions is very different; these patients are likely to benefit from regular secondary care contact.10 This article will concentrate on follow-up after surgery with curative intent for non-small cell lung cancer (NSCLC).

tive cancer care.9 It is recognised that follow-up of patients undergoing palliative surgical interventions is very different; these patients are likely to benefit from regular secondary care contact.10 This article will concentrate on follow-up after surgery with curative intent for non-small cell lung cancer (NSCLC). The role of primary care in follow-up of patients with cancer There is increasing acceptance of the role of primary care in the follow-up of cancer survivors,11–14 and in turn this is seen as increasingly critical for the long-term sustainability for healthcare systems in many developed countries.15 In many developed countries, there is an increasing financial demand on secondary and tertiary care services. There is thus a need to better define which clinical services require specialist oversight and which conditions could, perhaps with appropriate guidance and oversight,16 17 be acceptably managed within primary care. Evidence from randomised clinical studies in bowel and breast cancer follow-up suggests that integrated follow-up in primary care has no adverse outcomes, similar health-related quality of life and high patient satisfaction rates.18 19 The results of an Australian randomised control trial examining integrated care with hospitals and general practitioners (GPs) for patients with prostate cancer also demonstrated that this was a safe and acceptable model of care at lower cost to the healthcare system.20 GPs feel confident and enabled to contribute to long-term cancer care21 and cancer survivors are satisfied with care delivery in primary care.22

and general practitioners (GPs) for patients with prostate cancer also demonstrated that this was a safe and acceptable model of care at lower cost to the healthcare system.20 GPs feel confident and enabled to contribute to long-term cancer care21 and cancer survivors are satisfied with care delivery in primary care.22 There have been a number of studies examining follow-up of lung cancer in different settings. A retrospective cohort study in Japan reported the outcomes of postoperative patients with NSCLC dependent on follow-up with either thoracic surgeons or chest physicians.23 Survival in more advanced stage disease was better with the chest physician group; there was data to suggest that the use of regular CT chest scans may improve detection of recurrence (or new primary) to enable therapy with curative intent (regardless of doctor in charge of follow-up).

c surgeons or chest physicians.23 Survival in more advanced stage disease was better with the chest physician group; there was data to suggest that the use of regular CT chest scans may improve detection of recurrence (or new primary) to enable therapy with curative intent (regardless of doctor in charge of follow-up). Another retrospective cohort study from Canada describes outcomes of postsurgical patients with NSCLC, comparing thoracic surgeon followed by other health professionals.24 Despite hospital clinic follow-up, two-thirds of recurrences were detected by the patient’s GP, with no overall survival differences. It was postulated that GP-based care might be associated with 75% cost savings from follow-up. A randomised controlled trial from the UK comparing nurse-led care with conventional approaches for advanced NSCLC reported improved satisfaction, earlier recognition of deterioration, better emotional functioning, fewer consultations and no difference in survival with nurse-led care.25 Nurse-led care is reported to be acceptable to patients, carers, GPs and treating physicians.26 Such non-specialist follow-up may therefore be acceptable as long as there are clear protocols including the ability to refer to specialists easily and access to appropriate radiology.26

ce in survival with nurse-led care.25 Nurse-led care is reported to be acceptable to patients, carers, GPs and treating physicians.26 Such non-specialist follow-up may therefore be acceptable as long as there are clear protocols including the ability to refer to specialists easily and access to appropriate radiology.26 The potential shift in the care for patient with cancer with greater GP involvement also raises concerns that GP workloads may be increased.21 There is additional concern that perhaps GPs will see very few of these cases in their practice and thus may not be adequately confident or have sufficient experience to safely follow-up these patients.15 27 The proposed solution in this situation would be the use of a protocol in conjunction with clinical acumen to guide follow-up.28 Evidence suggests that certain core elements for successful models of shared cancer care are required: clear roles and responsibilities for GPs, timely effective communication about care, guidance on follow-up protocols and management of common treatment side effects and rapid routes to access specialist care.16 17 20 Therefore, a more collaborative and integrated approach would allow GPs to participate more with the follow-up care of patients with cancer, perhaps with integration with virtual or remote monitoring clinics from secondary care.

gement of common treatment side effects and rapid routes to access specialist care.16 17 20 Therefore, a more collaborative and integrated approach would allow GPs to participate more with the follow-up care of patients with cancer, perhaps with integration with virtual or remote monitoring clinics from secondary care. In some countries, access to specialist care varies, notably, specialist visits in populations that have a lower income or education levels are lower. The inequity in specialist access favours the populations with higher income and education in Organisation for Economic Co-operation and Development countries.22 23 29 30 Access to GPs, which is comparably better than access to specialists, would potentially improve this inequity and allow the same (or improved) standards of care.

in specialist access favours the populations with higher income and education in Organisation for Economic Co-operation and Development countries.22 23 29 30 Access to GPs, which is comparably better than access to specialists, would potentially improve this inequity and allow the same (or improved) standards of care. Unmet needs for patients with lung cancer As acknowledged in the National Institute for Health and Care Excellence lung cancer clinical guidelines, there are sparse data examining unmet needs for patients with early stage lung cancer.10 Interviews with patients with lung cancer (with more advanced stage) have identified at least moderate levels of anxiety and depression.31 A further study reported that hospital consultants were failing to recognise anxiety in many patients and largely overlooked the needs of informal carers (in whom it was reported had the greatest onus of care).32 Key areas of unmet need may be most apparent during periods away from acute care.32 Other in depth interviews with patients with lung cancer and carers identified feelings of isolation and identified a need for coordinated family-oriented care.33 Patients identified a reliance on their hospital-based consultant and found it difficult to transition back to primary (or palliative) care. Variations in perceptions of care may be associated with a patients’ educational level, with a higher level of education associated with more focus on the logistics of care, rather than the psychosocial aspects.34

their hospital-based consultant and found it difficult to transition back to primary (or palliative) care. Variations in perceptions of care may be associated with a patients’ educational level, with a higher level of education associated with more focus on the logistics of care, rather than the psychosocial aspects.34 There is a further unmet need to support smoking cessation in lung cancer survivors with a 5-year survival rate following resection for lung cancer of 77% in those able to quit and 33% in those unable to quit smoking.7 Between our two institutions, 33%–62% of newly diagnosed lung cancer cases between 2015 and 2016 were current smokers at the time of their diagnosis (unpublished data). Physicians and surgeons who treat lung cancer give limited smoking cessation advice and implementation of plans due to limited resources.35 Smoking cessation and support for other behaviour change are key roles of primary care with links to community-based resources to support patients to stop smoking and adopt a healthier lifestyle.

surgeons who treat lung cancer give limited smoking cessation advice and implementation of plans due to limited resources.35 Smoking cessation and support for other behaviour change are key roles of primary care with links to community-based resources to support patients to stop smoking and adopt a healthier lifestyle. Therefore, the present system of hospital-based care may not be adequately identifying all the needs of a patient with lung cancer, in particular, psychosocial aspects. There is further need for improved communication between hospital specialists and GPs33 36 37 and a more personalised explanation of treatment plan for patients.38 It is already accepted that primary care physicians are motivated and capable of providing follow-up to patients with cancer after their initial treatment, with good satisfaction among patients. Additionally, there is an identified economic need to decentralise care away from secondary and tertiary centres where possible.15