Browse the corpus

Translational perspective The immune system plays critical roles in myocardial injury and repair following acute myocardial infarction (AMI). Evidence from experimental models strongly implicates monocytes as critical to these processes and their specific targeting results in a significant reduction in infarct size and improved healing. It is currently unclear if monocytes play a similarly important role in humans. Examining changes in the patterns of gene expression can address this question. Here, we show that the peripheral blood monocyte response following AMI is conserved between species and that monocytes appear to be ‘programmed’ prior to their arrival at sites of myocardial injury. This investigation may translate to the future development of therapeutics to treat patients presenting with AMI that importantly would be effective in the hours after the onset of ischaemia.

is conserved between species and that monocytes appear to be ‘programmed’ prior to their arrival at sites of myocardial injury. This investigation may translate to the future development of therapeutics to treat patients presenting with AMI that importantly would be effective in the hours after the onset of ischaemia. Introduction In AMI, irreversible tissue injury occurs due to sustained ischaemia. In addition, recent pivotal studies have shown that the innate immune system is activated in AMI, sequentially mediating aspects of both injury and repair.1–3 Importantly, early phase monocytes appear to act as mediators of injury, since attenuating the response of inflammatory monocytes with siRNA,4,5 angiotensin converting enzyme inhibitors,6 or splenectomy6 significantly reduced infarct size in experimental models.4–6 These processes have been studied in great detail in mice5–8 but analogous data from humans are sparse9,10 and it is therefore not clear if monocytes in humans are equally important in comparison with experimental models.

enzyme inhibitors,6 or splenectomy6 significantly reduced infarct size in experimental models.4–6 These processes have been studied in great detail in mice5–8 but analogous data from humans are sparse9,10 and it is therefore not clear if monocytes in humans are equally important in comparison with experimental models. In fact, for both humans and mice, characterization of the monocyte response following AMI has largely relied on the analysis of a very limited repertoire of cell surface proteins.11,12 In mice, monocytes are functionally and phenotypically heterogeneous and can be divided into subsets by the presence of the cell surface protein Ly6C. Ly6Chi monocytes express high levels of CCR2 and accumulate preferentially in inflammatory sites.11 These can be distinguished from Ly6Clo monocytes, which patrol the vasculature13 and have roles in tissue repair and angiogenesis.7 Following AMI in mice, and possibly humans,10,14 monocytes from the spleen15 and the bone marrow16 are recruited to ischaemic myocardium via the blood in a coordinated manner.9 In mice, Ly6Chi monocytes are mobilized early, peaking in blood at ∼48 h. There is also monocyte heterogeneity in humans, with analogous subsets based on the cell surface protein expression of CD14 and CD16.12 CD14++CD16− resemble Ly6Chi monocytes, whilst CD14dimCD16+ monocytes resemble Ly6Clo monocytes in mice.

nner.9 In mice, Ly6Chi monocytes are mobilized early, peaking in blood at ∼48 h. There is also monocyte heterogeneity in humans, with analogous subsets based on the cell surface protein expression of CD14 and CD16.12 CD14++CD16− resemble Ly6Chi monocytes, whilst CD14dimCD16+ monocytes resemble Ly6Clo monocytes in mice. Whilst cell surface proteins distinguish monocyte subsets in mice and humans, are of descriptive value and are widely used, they provide little functional insight. In contrast, whole-genome expression profiling in leukocytes provides an opportunity for unbiased analysis of cellular function and therefore of the characterization of pathways and processes with far greater complexity than can be attained using only surface markers.17–19 Accordingly, we quantify changes in the transcriptome of monocytes isolated from peripheral blood early after AMI, as well as in absolute monocyte numbers (and their subsets).

nd therefore of the characterization of pathways and processes with far greater complexity than can be attained using only surface markers.17–19 Accordingly, we quantify changes in the transcriptome of monocytes isolated from peripheral blood early after AMI, as well as in absolute monocyte numbers (and their subsets). Mouse models can be highly informative and have enabled increased understanding in the pathogenesis of AMI. However, cautionary experience from other models argues that they cannot be assumed to be representative of human disease.20,21 By comparing changes in the monocyte transcriptome following AMI in mice and humans, we evaluate whether there is a conserved response between species, thereby contributing to the validation of mouse models for the study of disease mechanisms and therapies in this domain. Furthermore, we use gene set enrichment analysis (GSEA) to determine and understand the functional relevance of these responses, which we confirm at protein level, thereby identifying new possibilities for patient stratification and targeted therapies post AMI. Materials and methods Ethics statement Animal studies were undertaken with the approval of the University of Oxford Ethical Review Committee and procedures were conducted in accordance with the UK Home Office Animals (Science Procedures) Act 1986 (HMSO UK) incorporating European directive 2010/63/EU. The clinical study protocols were approved by the Oxfordshire Research Ethics Committee (references 08/H0603/41 and 11/SC/0397). All patients provided informed consent to participate.

ucted in accordance with the UK Home Office Animals (Science Procedures) Act 1986 (HMSO UK) incorporating European directive 2010/63/EU. The clinical study protocols were approved by the Oxfordshire Research Ethics Committee (references 08/H0603/41 and 11/SC/0397). All patients provided informed consent to participate. Mouse model of acute myocardial infarction All experiments were conducted on female C57BL/6J mice (Harlan, Blackthorn, UK) at 24 weeks of age (n = 6/group) due to the higher incidence of acute ventricular rupture in male mice.22 Surgical AMI was induced as previously described.23 Sham-procedure mice underwent the same protocol, but without ligation of the coronary artery. Transthoracic echocardiography (TTE) was carried out with a Vevo 2100 ultrasound system (Visualsonics, Amsterdam, The Netherlands) 48 h after AMI and fractional shortening was used to quantify infarct size of infarct. Human subjects Thirty consecutive patients presenting to the Oxford Heart Centre with STEMI were recruited between June 2012 and November 2012 as part of the Oxford acute myocardial infarction (OxAMI) study. Peripheral venous blood was obtained from patients at presentation (the ‘hyperacute’ time point—before primary angioplasty and during AMI) and at 48 h following presentation (48 h time point) and processed within 30 min. Additionally, 24 stable patients, with confirmed coronary atherosclerosis were recruited to act as controls (Table 1). Table 1 Patient demographic data

t presentation (the ‘hyperacute’ time point—before primary angioplasty and during AMI) and at 48 h following presentation (48 h time point) and processed within 30 min. Additionally, 24 stable patients, with confirmed coronary atherosclerosis were recruited to act as controls (Table 1). Table 1 Patient demographic data CONTROL STEMI P-value n 24 30 n/a M:F 18:6 27:3 P = 0.71 Age (mean, range) 63.4 (46–79) 60.1 (38–87) P = 0.64 CV risk Diabetes mellitus 3 (12.5%) 4 (13.3%) P = 0.81 Smoker 13 (54.2%) 19 (63.3%) P = 0.42 Hypertension 21 (87.5%) 10 (33.3%) P < 0.001 Hypercholesterolaemia 10 (41.7%) 4 (13.3%) P < 0.001 Family history 10 (41.7%) 12 (40%) P = 0.53 Previous stroke/AMI 11 (45.8%) 0 (0%) P < 0.001 Lipoproteins (mmol/L) (mean, SD) Total cholesterol 4.14 (1.09) 4.74 (0.87) P = 0.77 LDL-c 2.2 (0.82) 3.1 (1.23) P = 0.26 HDL-c 1.09 (0.29) 1.07 (0.3) P = 0.89 TG 1.78 (0.85) 1.5 (1.28) P = 0.19 Admission medication Aspirin n/a 1 (3.3%) n/a Clopidogrel/prasugrel n/a 0 (0%) n/a β-Blocker n/a 2 (6.7%) n/a ACE-I/ARB n/a 5 (16.7%) n/a Statin n/a 4 (13.3%) n/a Calcium blocker n/a 5 (16.7%) n/a Diuretic n/a 1 (3.3%) n/a Renal function (mean, SD) Creatinine (μmol/L) 85.54 (19) 82.23 (14.95) P = 0.24 Infarct-related artery Left main stem n/a 0 (0%) n/a Left anterior descending n/a 14 (46.7%) n/a Circumflex n/a 4 (13.3%) n/a Right coronary artery n/a 12 (40%) n/a Peri-procedural medication Bivalirudin n/a 30 (100%) n/a Heparin n/a 16 (53.3%) n/a Glycoprotein IIb/IIIa inhibitor n/a 0 (0%) n/a Revascularization time (min) Call-to-balloon n/a 119.4 (47.93) n/a Door-to-balloon n/a 35.48 (32.25) n/a Discharge medication Aspirin 24 (100%) 30 (100%) P = 1.0 Clopidogrel/prasugrel 10 (41.7%) 30 (100%) P < 0.001 β-Blocker 17 (70.8%) 29 (96.7%) P < 0.01 ACE-I/ARB 18 (75%) 29 (96.7%) P < 0.01 Statin 17 (70.8%) 30 (100%) P < 0.01 Calcium blocker 7 (29%) 0 (0%) P < 0.001 Diuretic 3 (12.5%) 4 (13.3%) P = 0.93 STEMI, ST-segment elevation myocardial infarction; n, number; M, male; F, female; CV, cardiovascular; LDL-c, low-density lipoprotein cholesterol; HDL-c, high-density lipoprotein cholesterol; TG, triglycerides; ACEI, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker.

3 (12.5%) 4 (13.3%) P = 0.93 STEMI, ST-segment elevation myocardial infarction; n, number; M, male; F, female; CV, cardiovascular; LDL-c, low-density lipoprotein cholesterol; HDL-c, high-density lipoprotein cholesterol; TG, triglycerides; ACEI, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker. All patients (n = 30) underwent 3 Tesla cardiac magnetic resonance (CMR; Verio, Siemens, Germany) within 48 h of presentation and 23 patients underwent a second CMR scan at 6 months. Both CMR scans assessed left ventricular function (LVF), oedema, and late gadolinium enhancement (LGE) as previously described24 (Supplementary material online, Table S2). Cardiac magnetic resonance analysis was carried out using cmr42® software (Circle Cardiovascular Imaging Inc., Calgary, Canada). Flow cytometry All stained samples were analysed using the flow cytometer (CyAN ADP Flow cytometer, Dako, Ely, UK). Data were analysed using FlowJo software, version 7.6.3 (Tree Star Inc., OR, USA). Mice Whole blood was obtained (n = 6/group) and cell suspensions were prepared as previously described25 and were incubated with the following antibodies: CD11b-APC, Ly6G-PE, Ly6C-FITC (all BD Pharmingen, Oxford, UK), and CD115-PerCP (eBioscience, San Diego, USA). Human Whole venous blood was stained with the following antibodies: CD14-APC, CD16-PE-Cy7, CD86-PE, CD42b-FITC, CD11b-Pacific Blue, CD66b-PE-Cy5, CD56-PE-Cy5, CD123-PE-Cy5 (all BD Pharmingen), and TLR2-FITC (all BD Pharmingen) or permeabilized and stained with Ki67-FITC (BD Pharmingen) for validation studies.

Mice Whole blood was obtained (n = 6/group) and cell suspensions were prepared as previously described25 and were incubated with the following antibodies: CD11b-APC, Ly6G-PE, Ly6C-FITC (all BD Pharmingen, Oxford, UK), and CD115-PerCP (eBioscience, San Diego, USA). Human Whole venous blood was stained with the following antibodies: CD14-APC, CD16-PE-Cy7, CD86-PE, CD42b-FITC, CD11b-Pacific Blue, CD66b-PE-Cy5, CD56-PE-Cy5, CD123-PE-Cy5 (all BD Pharmingen), and TLR2-FITC (all BD Pharmingen) or permeabilized and stained with Ki67-FITC (BD Pharmingen) for validation studies. Monocyte cell sorting Mouse cell suspensions were immediately sorted on an MOFLO cell sorter (Dako, Ely, UK). Human monocytes were isolated using the EasySep Human monocyte enrichment kit without CD16 depletion (StemCell Technologies, Grenoble, France) as per the manufacturer's instructions. Microarray Extracted and purified RNA samples from individual samples with a 260/280 > 2.0 and an RNA integrity number >7.0 (Agilent, Wokingham, UK) were deemed acceptable. Following amplification and biotin labelled, mouse RNA was hybridized to Mouse WG-6 BeadChips (Illumina, San Diego, CA, USA) and human RNA to Illumina Human HT12v3.0 BeadChips (Illumina). Immunohistochemistry Frozen sections of tissue harvested from mice were stained with CD11b (BD Pharmingen, Oxford, UK) and Ki67 (Abcam, Cambridge, UK). Five random FOVs at 20× magnification were recorded (Leica DM2500 microscope) and the number of cells positive for CD11b and Ki67 were assessed.

Microarray Extracted and purified RNA samples from individual samples with a 260/280 > 2.0 and an RNA integrity number >7.0 (Agilent, Wokingham, UK) were deemed acceptable. Following amplification and biotin labelled, mouse RNA was hybridized to Mouse WG-6 BeadChips (Illumina, San Diego, CA, USA) and human RNA to Illumina Human HT12v3.0 BeadChips (Illumina). Immunohistochemistry Frozen sections of tissue harvested from mice were stained with CD11b (BD Pharmingen, Oxford, UK) and Ki67 (Abcam, Cambridge, UK). Five random FOVs at 20× magnification were recorded (Leica DM2500 microscope) and the number of cells positive for CD11b and Ki67 were assessed. Analysis of gene expression profiles Microarray data were analysed using GenePattern (Broad Institute, Cambridge, MA, USA). Differentially expressed genes were identified using an unpaired t-test with a fold change of >1.5 and a P < 0.05 deemed necessary for significance. Gene-set enrichment analysis and leading-edge analysis using the C7 collection of the molecular signature database (MSIGDB; http://www.broadinstitute.org/gsea/msigdb) was performed as previously described,26,27 a false discovery rate (FDR) < 0.25 was deemed necessary for significance.

deemed necessary for significance. Gene-set enrichment analysis and leading-edge analysis using the C7 collection of the molecular signature database (MSIGDB; http://www.broadinstitute.org/gsea/msigdb) was performed as previously described,26,27 a false discovery rate (FDR) < 0.25 was deemed necessary for significance. Gene set enrichment analysis Gene set enrichment analysis yields a quantitative measure of the over-representation of a set of genes (e.g. genes encoding products in a same metabolic pathway) at the top or bottom of a ranked list of genes. Candidate genes are ranked by their differential expression between two phenotypes. The statistic is a weighted Kolmogorov–Smirnov-like statistic and significance is calculated using an empirical permutation test.28 Here we applied an extended version of conventional GSEA in order to produce an enrichment score in a single sample, as described previously.29 Such a score is necessary in order to make a predictive call on single samples without reference to a larger group of samples. In this approach, the genes are ordered based on either absolute expression or the relative changes with respect to the baseline level. Gene set enrichment analysis was performed as described previously.28 Gene sets enriching with FDR < 0.01 were used for leading-edge analysis using Pearson correlation.30 Gene clusters were functionally annotated by calculating overlap with GO terms using PANTHER algorithm31 on AmiGO 2 version 2.1.3.32,33

Gene set enrichment analysis Gene set enrichment analysis yields a quantitative measure of the over-representation of a set of genes (e.g. genes encoding products in a same metabolic pathway) at the top or bottom of a ranked list of genes. Candidate genes are ranked by their differential expression between two phenotypes. The statistic is a weighted Kolmogorov–Smirnov-like statistic and significance is calculated using an empirical permutation test.28 Here we applied an extended version of conventional GSEA in order to produce an enrichment score in a single sample, as described previously.29 Such a score is necessary in order to make a predictive call on single samples without reference to a larger group of samples. In this approach, the genes are ordered based on either absolute expression or the relative changes with respect to the baseline level. Gene set enrichment analysis was performed as described previously.28 Gene sets enriching with FDR < 0.01 were used for leading-edge analysis using Pearson correlation.30 Gene clusters were functionally annotated by calculating overlap with GO terms using PANTHER algorithm31 on AmiGO 2 version 2.1.3.32,33 Statistics Values are expressed as mean ± SD. The Gaussian distribution of all parameters was tested. Differences in continuous variables between groups were compared using the Student t-test. Categorical variables are presented as numeric values and percentages. Correlative analyses were determined by Spearman statistical test to determine the percentage of variance explained or a spline fit for non-linear relationships. All statistical tests were two-tailed and a P < 0.05 was considered statistically significant. Analyses were performed with GraphPad Prism Version 5 (GraphPad Software Inc., San Diego, CA, USA) and SPSS Version 21 (IBM Corporation, NY, USA).

percentage of variance explained or a spline fit for non-linear relationships. All statistical tests were two-tailed and a P < 0.05 was considered statistically significant. Analyses were performed with GraphPad Prism Version 5 (GraphPad Software Inc., San Diego, CA, USA) and SPSS Version 21 (IBM Corporation, NY, USA). Results Acute myocardial infarction results in a peripheral monocytosis at 48 h Following AMI in the mouse there was an increase (4-fold; P < 0.001) in total peripheral circulating monocytes at 48 h (Figure 1A and B). Similarly, in humans, flow cytometry of whole blood 48 h following onset of symptoms showed a 2-fold increase (P < 0.001) in total circulating monocytes in comparisons with both (i) the hyperacute time point and (ii) stable control subjects (Figure 1D–F). There was no significant difference in total peripheral circulating monocytes at presentation (hyperacute time point) compared with stable controls (Figure 1E). Figure 1 Acute myocardial infarction results in an increase in total peripheral circulating monocytes of an inflammatory phenotype in both mice and humans. (A) Fluorescence-activate cell sorting (FACS) gating strategy to identify monocytes in peripheral blood (mice, n = 6/group). (B) Acute myocardial infarction results in a significant 4-fold increase in total monocytes (P < 0.001), which were predominately of an inflammatory Ly6Chi phenotype (C). (D) Human monocyte (and monocyte subset) gating strategy (n = 24 control group, n = 30 acute myocardial infarction group). (E) Acute myocardial infarction results in a significant 2-fold increase in total monocytes (P < 0.001) at 48 h following injury but not at the hyperacute (at presentation) timepoint (F). (G) Monocytes at both the hyperacute and 48 h time points exhibited an inflammatory CD14++CD16− phenotype. Data are represented by mean ± standard deviation.

rction results in a significant 2-fold increase in total monocytes (P < 0.001) at 48 h following injury but not at the hyperacute (at presentation) timepoint (F). (G) Monocytes at both the hyperacute and 48 h time points exhibited an inflammatory CD14++CD16− phenotype. Data are represented by mean ± standard deviation. Monocytes display an inflammatory phenotype 48 h following acute myocardial infarction In the mouse model of AMI at 48 h, monocytes were predominantly of the Ly6Chi, inflammatory subset (P = 0.005, Figure 1C). Similarly, in human patients there was a small, but statistically significant, CD14++CD16− inflammatory subset preponderance at 48 h (82.9 ± 6.9% vs. 75.9 ± 5.1% (P < 0.001), compared with control patients with stable atherosclerosis (Figure 1G).

predominantly of the Ly6Chi, inflammatory subset (P = 0.005, Figure 1C). Similarly, in human patients there was a small, but statistically significant, CD14++CD16− inflammatory subset preponderance at 48 h (82.9 ± 6.9% vs. 75.9 ± 5.1% (P < 0.001), compared with control patients with stable atherosclerosis (Figure 1G). The magnitude of the monocyte response in both mice and humans correlates with the extent of myocardial injury Due to the anticipated marked variation in infarct size in mice that underwent AMI, mice were analysed by tertiles of infarct size, as characterized by TTE (Figure 2A). The total peripheral monocyte count was highly correlated with size of infarct (r2 = 0.72, P < 0.001, Figure 2C). Similarly, in humans, the magnitude of the monocyte response for each patient (the difference between the 48 h monocyte count and the hyperacute monocyte count) following AMI correlated with the extent of irreversible myocardial injury determined by LGE at 6 months (r2 = 0.42, P = 0.001, Figure 2D) and weakly with the volume of acute ischaemia, as measured by oedema on T2-weighted magnetic resonance imaging (r2 = 0.29, P = 0.021; Figure 2E). Figure 2 The extent of the monocyte response correlates with the extent of myocardial injury. (A) Transthoracic echocardiography of mouse hearts following acute myocardial infarction (n = 6/group). (B) Cardiac magnetic resonance imaging to quantify area of injury (T2-weighted sequence) and infarction (LGE sequence) in humans (n = 30 acute timepoint, n = 23 follow up). (C) In mice, the total peripheral monocyte count was highly correlated with size of infarct (r2 = 0.72, P < 0.001). (D) In humans, the magnitude of the monocyte response for each patient following acute myocardial infarction correlated with the extent of irreversible myocardial injury determined by LGE at 6 months (r2 = 0.42, P = 0.001) and weakly (non-linear relationship) with the area of risk, as measured by oedema (r2 = 0.29, P = 0.021, E).

ns, the magnitude of the monocyte response for each patient following acute myocardial infarction correlated with the extent of irreversible myocardial injury determined by LGE at 6 months (r2 = 0.42, P = 0.001) and weakly (non-linear relationship) with the area of risk, as measured by oedema (r2 = 0.29, P = 0.021, E). Having confirmed similarities in the elevation of monocyte subtypes acutely after myocardial infarction, we next sought to evaluate functional characteristics, as evidenced by alterations in patterns of gene expression. Gene expression profiling of peripheral circulating monocytes identifies differentially expressed genes in mice and humans Unsupervised principal components analysis demonstrated a clear difference in groups in mice (Supplementary material online, Figure S1A). A similar effect was present in human subjects, though, as expected, due to the variability in infarct size, ischaemic territory, ischaemia time, and genetic heterogeneity, the group effect was less pronounced (Supplementary material online, Figure S1B).

ce in groups in mice (Supplementary material online, Figure S1A). A similar effect was present in human subjects, though, as expected, due to the variability in infarct size, ischaemic territory, ischaemia time, and genetic heterogeneity, the group effect was less pronounced (Supplementary material online, Figure S1B). Supervised analysis of the transcriptome at baseline in comparison with 48 h following AMI in mouse monocytes revealed differential expression of 196 genes (fold change >2; P < 0.01) of which 168 genes were significantly up-regulated and 28 genes were significantly down-regulated (Figure 3A). A similar comparison in humans identified that 122 genes were differentially expressed (fold change >2; P < 0.01), of which 72 genes were significantly up-regulated and 50 genes were significantly down-regulated (Figure 3B). The change in gene expression was confirmed by PCR in selected genes in both mice and human studies (Supplementary material online, Figures S2 and S3). Figure 3 Acute myocardial infarction results in differential expression of genes in circulating monocytes in both humans and mice. (A) Analysis of the transciptome in mouse monocytes revealed differential expression of 196 genes (fold change >2; P < 0.01) of which 168 genes were significantly up-regulated and 28 genes were significantly down-regulated (n = 6/group). (B) In humans, 122 genes were differentially expressed (fold change > 2; P < 0.01) of which 72 genes were significantly up-regulated and 50 genes were significantly down-regulated (n = 24, control group, n = 30, AMI group).

es were significantly up-regulated and 28 genes were significantly down-regulated (n = 6/group). (B) In humans, 122 genes were differentially expressed (fold change > 2; P < 0.01) of which 72 genes were significantly up-regulated and 50 genes were significantly down-regulated (n = 24, control group, n = 30, AMI group). Monocyte transcriptional response in acute myocardial infarction is conserved in mice and humans Having identified individual genes that were significantly differentially expressed following AMI in peripheral circulating monocytes, we then investigated whether these observed changes conformed to similar patterns in mice and humans. In order to do this, we generated two ranked gene ‘signatures' (n = 200 genes) of the most up-regulated genes in circulating monocytes following AMI, one in the mouse model and one in humans. Gene set enrichment analysis revealed significant enrichment of the mouse signature in the ranked list of genes up-regulated in human monocytes after AMI (P < 0.001, FDR < 0.001, Figure 4A). Similarly, a signature of genes differentially expressed in humans revealed significant enrichment in the ranked list of genes up-regulated in mouse monocytes after AMI (P < 0.001, FDR < 0.01, Figure 4A). This comparison of the primary data sets shows that the pattern of gene expression induced in monocytes following myocardial infarction is significantly conserved between species. Figure 4 Monocytes from myocardial infarction in mouse and human are transcriptionally similar. (A) Mouse monocyte gene signature in acute myocardial infarction strongly enriches in the ranked gene list of human blood monocytes 48 h following infarct compared with the stable state (left) and vice versa (right). (B) Overlap of genes in same top enriching gene sets with the ranked gene list of the infarct vs. baseline state of monocytes in human (left) and mouse (right). (C) Number of gene sets in ImmuneSigDB v1.0 strongly enriching (FDR < 0.01) in infarct human and/or mouse monocytes. (D) Representation of genes in the leading edges of the 163 shared enriched gene sets. Genes were ranked from most to least represented in human dataset and shown is the number of times that gene appears in the leading edge of the enriched gene sets. Similarity of genes enriching in leading edges of enriched gene sets in both organisms was assessed by Spearman correlation.

163 shared enriched gene sets. Genes were ranked from most to least represented in human dataset and shown is the number of times that gene appears in the leading edge of the enriched gene sets. Similarity of genes enriching in leading edges of enriched gene sets in both organisms was assessed by Spearman correlation. We then identified the biological processes present in human and mouse monocytes following AMI by testing each monocyte expressed dataset for up-regulation of a curated compendium of ∼2000 immune-related signatures.28 We identified 163 gene sets that were enriched in both species following AMI, including gene sets related to myeloid lineage, bacterial infection, and cytokine stimulation of innate cells (Figure 4B and C). We next examined the representation of genes that lay in the leading edges of gene sets showing enrichment in both species. The leading-edge genes of an enriched gene set are those that contributed most significantly to the enrichment score, and reflect the major ‘drivers’ of enrichment. The leading-edge genes showed striking similarity between the two species (Figure 4D), again suggesting that the transcriptional profile of monocytes following AMI in human and mouse shared common biological characteristics. Several of the most commonly represented genes in both human and mouse included CD14, LGALS1, ITGAM, CD163, IFNGR1, CYBB, and TLR2 with roles in monocyte inflammation, cell-to-cell signalling, and cellular proliferation (Figure 4D).

of monocytes following AMI in human and mouse shared common biological characteristics. Several of the most commonly represented genes in both human and mouse included CD14, LGALS1, ITGAM, CD163, IFNGR1, CYBB, and TLR2 with roles in monocyte inflammation, cell-to-cell signalling, and cellular proliferation (Figure 4D). Gene set enrichment analysis identifies functional characteristics of circulating monocytes in acute myocardial infarction To gain further insights into the biological function revealed by these enrichments, we performed leading-edge analysis of the most significantly enriched gene sets (FDR < 0.01) corresponding to the most differentially expressed genes in mouse monocytes, following myocardial infarction. We found two clusters of genes (vertical clustering) that were represented in multiple immune gene sets (horizontal clustering) following AMI. This suggests that these genes form part of transcriptional modules of coordinately regulated genes that are up-regulated in mouse monocytes. We functionally annotated these clusters of up-regulated genes using Gene Ontology (Figure 5), and found significant over-representation of genes related to inflammation (including TLR2, integrins, and chemokines including CCR1; Supplementary material online, Tables S3 and S4) and cell cycle (including cyclins, annexins, and tetraspanins; Supplementary material online, Tables S5 and S6). Figure 5 Leading-edge analysis identifies inflammation and cell cycling as novel biological states underlying the monocyte response to acute myocardial infarction. Leading-edge analysis of most significantly enriching gene sets (FDR < 0.01) in mouse monocytes following myocardial infarction reveals metagenes involved in inflammatory and cell cycle processes. Representative genes in these metagenes are listed in Supplementary material online, Table.

to acute myocardial infarction. Leading-edge analysis of most significantly enriching gene sets (FDR < 0.01) in mouse monocytes following myocardial infarction reveals metagenes involved in inflammatory and cell cycle processes. Representative genes in these metagenes are listed in Supplementary material online, Table. Validation of transcriptional findings Up-regulation of proliferation-associated genes was a prominent feature of the transcriptional profile of monocytes following AMI in both humans and mice. To confirm this finding, we undertook fluorescence-activate cell sorting (FACS) analysis of peripheral blood mononuclear cells stored from the same patients at the 48 h time point for the proliferation markers (Ki67 and cyclin) and markers of inflammation (TLR-2 and TLR-4). As predicted from the leading-edge analysis, there was (i) a significant increase in TLR-2 at protein level (P < 0.001, Figure 6A), and (ii) no change in TLR4 (P = 0.14, Figure 6B). Ki67, a marker of cell proliferation, was increased by 70% (P < 0.001, Figure 6C) in monocytes 48 h after MI. Given the elevation in Ki67 in peripheral monocytes, we next analysed mouse hearts to establish whether monocyte-derived cells in the myocardium at this very early time point already showed evidence of proliferation. In sham-treated mice, there were small numbers of CD11b+ cells of which virtually none were Ki67+. Following AMI in mice, there was a massive (130-fold), increase in the number of CD11b+ cells within the myocardium, with abundant Ki67-expressing CD11b+ leukocytes (P < 0.001 vs. sham-operated hearts, Figure 6D and E). Figure 6 Inflammatory and mitosis pathways are up-regulated in circulating monocytes en route to inflamed myocardium. (A) Flow cytometry of peripheral circulating monocytes 48 h following acute myocardial infarction in humans identified a significant increase in the expression of TLR2 (P < 0.001) but not TLR4 (P = 0.14, B) (n = 12/group). (C) Ki67 expression was significantly up-regulated in monocytes at 48 h following injury but not hyperacutely (at presentation) in comparison with controls (P < 0.001) (n = 12/group). (D and E) Immunohistochemistry of mouse hearts (blue: DAPI, red: CD11b, green Ki67) following acute myocardial infarction confirmed a significant increase in the number of leukocytes expressing Ki67 (white arrows) (P < 0.001) indicating up-regulation of mitosis pathways in infarcted myocardium (n = 3/group). Data are represented by mean ± standard deviation.

hearts (blue: DAPI, red: CD11b, green Ki67) following acute myocardial infarction confirmed a significant increase in the number of leukocytes expressing Ki67 (white arrows) (P < 0.001) indicating up-regulation of mitosis pathways in infarcted myocardium (n = 3/group). Data are represented by mean ± standard deviation. Discussion Data from mouse models strongly implicate the innate immune system, and in particular monocytes as critical to initial inflammatory and latter reparative processes following AMI.2,4–7 There is an early influx of Ly6chi monocytes, which are inflammatory,15 with a later preponderance of Ly6clow, F4/80high macrophages, which contribute to myocardial repair.3 The relatively sparse data from human studies in the context of AMI are consistent with these findings,14 but analyses have been confined to a small number of cell surface markers that provide limited insight into function and are not biologically useful.9,10

/80high macrophages, which contribute to myocardial repair.3 The relatively sparse data from human studies in the context of AMI are consistent with these findings,14 but analyses have been confined to a small number of cell surface markers that provide limited insight into function and are not biologically useful.9,10 Whilst previous studies have investigated the monocyte transcriptome, these have been limited to cells in the resting state.15,34 Here we present analyses of the changes in the blood monocyte transcriptome in mice and humans early after AMI. Through analysis of the changes in patterns of gene expression, we have been able to gain further functional insights into monocyte biology and this work contributes two potentially important findings. Firstly, the transcriptional response following AMI in monocytes between mice and humans appears to be largely conserved. This provides a validation of mouse models to study innate immune responses following acute MI. This is important since conservation across species cannot be assumed. For instance, recent comparisons of changes in gene expression in models of sepsis showed differences in gene expression between humans and mouse models. Secondly, current thinking suggests that monocytes act in ‘response mode’ whereby monocytes enter the circulation and ‘patrol the vasculature…. before being recruited to sites of inflammation.’.35 Although the absolute number and proportion of Ly6chi monocytes in circulating blood is well known to increase in AMI, changes in their function, implying a form of ‘programming’ prior to arrival at the site of inflammation have not previously been demonstrated. Here we show that, in mice and human monocytes, patterns of gene expression associated with inflammation and proliferation are switched on prior to their infiltration of injured tissue.

n their function, implying a form of ‘programming’ prior to arrival at the site of inflammation have not previously been demonstrated. Here we show that, in mice and human monocytes, patterns of gene expression associated with inflammation and proliferation are switched on prior to their infiltration of injured tissue. Whilst monocytes might be programmed in the blood, the recent observation that a cardiosplenic axis exists in humans14 (supporting findings in experimental models15) suggests that these cells could also be programmed in a different pool and ‘released’ in response to injury. Of relevance to this concept, observations from platelet studies have demonstrated that the expression of MRP-14 was found to be significantly up-regulated in circulating platelets of patients prior to myocardial infarction.36 Our work therefore suggests that a ‘pre-programmed state’ may not just be limited to platelets but may also extend to monocytes. Whilst our data do not exclude the possibility that naïve monocytes are recruited and then activated in the myocardium and reintroduced into the blood, this is not likely given what is already known of (i) the timing and (ii) large net flux of monocytes from spleen to heart in AMI. Furthermore, recent studies in experimental models show that once in the infarcted myocardium, Ly6chi monocytes differentiate in situ into Ly6chi, F4/80+ macrophages.3,37

ed into the blood, this is not likely given what is already known of (i) the timing and (ii) large net flux of monocytes from spleen to heart in AMI. Furthermore, recent studies in experimental models show that once in the infarcted myocardium, Ly6chi monocytes differentiate in situ into Ly6chi, F4/80+ macrophages.3,37 In keeping with previous reports, we show that AMI leads to a small but statistically significant increase in the number of Ly6chigh monocytes. However, it is notable that in both the resting state and after AMI, Ly6chigh monocytes predominate. Consequently, analysis of Ly6chigh monocyte numbers provides virtually no functional insight. We therefore undertook whole transcriptome analyses, using a non-selective, unbiased approach that was not constrained by examination of one or other monocyte subsets. Better to understand changes in gene expression, we employed GSEA. Gene set enrichment analysis features a number of advantages when compared with single-gene methods. Firstly, it provides a structure for interpretation by identifying pathways and processes. Rather than focusing on highly regulated single genes (which can be difficult to interpret mechanistically), GSEA focuses on gene sets, which tend to be more reproducible and more interpretable. Secondly, when the members of a gene set exhibit strong cross-correlation, GSEA boosts signal-to-noise ratio making it possible to appreciate the contributions of even modest changes in individual genes. Thirdly, ‘leading-edge analysis' can help define gene subsets to elucidate key players and identify critical biological processes.28

he members of a gene set exhibit strong cross-correlation, GSEA boosts signal-to-noise ratio making it possible to appreciate the contributions of even modest changes in individual genes. Thirdly, ‘leading-edge analysis' can help define gene subsets to elucidate key players and identify critical biological processes.28 In the first instance, we show conservation between the mouse and human monocyte response to AMI. This is important since it provides surety for mechanistic studies in mice and for the development of new therapies. There were, however, also differences between species, with some biological processes (e.g. complement activation) up-regulated in the mouse but not in humans. Identification of inter-species differences is also important since they may anticipate discrepant outcomes associated with targeting particular process, including the complement cascade in clinical studies.38 Conversely, we also identified pathways, e.g. integrin-linked kinase pathway that has roles in cell migration, adhesion, and signalling,39 which have not been described previously in this pathology possibly due to a lack of function in this setting in mice.

ocess, including the complement cascade in clinical studies.38 Conversely, we also identified pathways, e.g. integrin-linked kinase pathway that has roles in cell migration, adhesion, and signalling,39 which have not been described previously in this pathology possibly due to a lack of function in this setting in mice. Leading-edge analysis identified single genes CD14, LGALS1, ITGAM, CD163, IFNGR1, CYBB, and TLR2 with roles in inflammation and cellular proliferation as central biological processes to the monocyte response to AMI in humans. Cellular proliferation within target tissue has recently, and unexpectedly, been shown to be critical to monocyte mediated reparative processes following AMI in experimental models3 and in atherosclerotic plaque progression40 and thus might represent drug-sensitive targets.

es to the monocyte response to AMI in humans. Cellular proliferation within target tissue has recently, and unexpectedly, been shown to be critical to monocyte mediated reparative processes following AMI in experimental models3 and in atherosclerotic plaque progression40 and thus might represent drug-sensitive targets. Gene expression profiling methods have provided a clearer view of the level of molecular heterogeneity that underlies pathologies including cancer,41 cardiomyopathy,42 cardiac transplantation,43 vaccine development44 and in the diagnosis of giant cell myocarditis.45 These technologies have also been used in the identification of novel targets for drug development46 and in predicting response to treatment.47 By applying these methodologies to the setting of AMI, we have identified a number of biological pathways that are up-regulated in monocytes following AMI in both mice and humans. The specific targeting of immune cells in AMI is an attractive therapeutic option; however, clinical trials that have adopted a broad strategy of inhibiting inflammation have been disappointing to date.48,49 Targeted therapies based on patient stratification on the basis of target cell transcriptome is appealing, identifying patients who have most to gain from specific biological therapies and those that would only be subjected to undue risk.

a broad strategy of inhibiting inflammation have been disappointing to date.48,49 Targeted therapies based on patient stratification on the basis of target cell transcriptome is appealing, identifying patients who have most to gain from specific biological therapies and those that would only be subjected to undue risk. Limitations Although the gene sets were obtained from large compendia of published data, they are inevitably incomplete in terms of the totality of possible biological states and also biased by the focus of experimental data submitted. Therefore, the interpretations of the transcriptome should be judged in that context and not considered definitive. It is also unlikely that the monocyte transcriptome is static, changes will occur over time, perhaps in response to new signals, and thus the peripheral monocyte transcriptome may not necessarily reflect the transcriptome following infiltration into ischaemic myocardium. This investigation was undertaken at a single time point, in part informed by the data for mice in the literature, although observed changes may vary between species at different timepoints. Subsequent studies will extend current evaluations to later timepoints.

Limitations Although the gene sets were obtained from large compendia of published data, they are inevitably incomplete in terms of the totality of possible biological states and also biased by the focus of experimental data submitted. Therefore, the interpretations of the transcriptome should be judged in that context and not considered definitive. It is also unlikely that the monocyte transcriptome is static, changes will occur over time, perhaps in response to new signals, and thus the peripheral monocyte transcriptome may not necessarily reflect the transcriptome following infiltration into ischaemic myocardium. This investigation was undertaken at a single time point, in part informed by the data for mice in the literature, although observed changes may vary between species at different timepoints. Subsequent studies will extend current evaluations to later timepoints. The magnitude of the monocyte response correlated with extent of LGE at 6 months but only weakly with early myocardial oedema. This may reflect a monocyte response that is driven by the extent of tissue that has undergone irreversible injury. However, a recent study has shown a biphasic accumulation of myocardial oedema that diminishes the utility of a single measurement to indicate extent of acute ischemic injury.50

h early myocardial oedema. This may reflect a monocyte response that is driven by the extent of tissue that has undergone irreversible injury. However, a recent study has shown a biphasic accumulation of myocardial oedema that diminishes the utility of a single measurement to indicate extent of acute ischemic injury.50 Conclusions This study demonstrates that the peripheral blood monocyte response following AMI is conserved between species both phenotypically but also at the level of the transcriptome. We show that circulating monocytes are ‘programmed’ with a number of biological processes up-regulated prior to their arrival at sites of myocardial injury and have identified mitosis a particularly important, but little recognized contributor to early inflammation and possibly latter reparative processes. Although monocyte proliferation has recently been shown to be important several days after AMI in mice,3 our data suggest both pre-activation and early proliferation at the site of infarction. Supplementary material Supplementary material is available at European Heart Journal online. Funding This work is supported by the Wellcome Trust, the British Heart Foundation, the National Institutes of Health (NIH), and the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre. There are no relationships with industry. Funding to pay the Open Access publication charges for this article was provided by The Wellcome Trust. Conflict of interest: none declared.

Funding This work is supported by the Wellcome Trust, the British Heart Foundation, the National Institutes of Health (NIH), and the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre. There are no relationships with industry. Funding to pay the Open Access publication charges for this article was provided by The Wellcome Trust. Conflict of interest: none declared. Acknowledgements The authors thank the clinical staff in the Oxford Heart Centre for the overall clinical care of patients recruited to the study. Dr Claire Fernandez, Melanie Jones, Juliet Semple, Carol Davey, and Lisa Gaughran in the Oxford Acute Vascular Imaging Centre are gratefully acknowledged for their expertise and work in the coordination of the OxAMI study. P.T. and J.B. are gratefully acknowledged for general laboratory management. The authors thank BMS staff for their expert care of mice used in this study. We thank Mr Drew Worth at the FACS core facility. We thank the High-Throughput Genomics Group at the Wellcome Trust Centre for Human Genetics (funded by Wellcome Trust grant reference 090532/Z/09/Z and MRC Hub grant G0900747 91070) for the general of the Gene Expression data. NR is a British Heart Foundation Centre of Research Excellence Clinical Research Fellow. R.P.C. is a Wellcome Trust Senior Research Fellow in Clinical Science.

See page 198 for the editorial comment on this article (doi:10.1093/eurheartj/ehv469) Introduction Dual antiplatelet therapy (DAPT) with acetylsalicylic acid and a P2Y12-receptor antagonist reduces the risk of thrombotic complications compared with treatment with only acetylsalicylic acid in patients with acute coronary syndrome (ACS).1 The risk of thrombotic complications is further reduced if one of the new more potent platelet inhibitors, ticagrelor or prasugrel, is used instead of clopidogrel,2,3 but the risk of both spontaneous and surgical bleeding complications may increase with the new inhibitors.3,4 Major bleeding complications impair outcome after cardiac surgery.5,6 Acute coronary syndrome patients on DAPT who need acute or urgent coronary artery bypass grafting (CABG) are at high risk of major bleeding.6–8 Current revascularization guidelines therefore recommend that clopidogrel and ticagrelor are discontinued 5 days before surgery and prasugrel 7 days before elective surgery,9,10 but the patient's condition may render this impossible. Most ACS patients are hospitalized while waiting for CABG. If a shorter discontinuation time of the platelet inhibitor would be safe from a bleeding perspective, it would reduce the risk of thrombotic complications during the waiting time, and save hospital resources. The risk of bleeding complications in relation to discontinuation time has not previously been investigated in real life in sufficiently large patient cohorts.

See page 1152 for the editorial comment on this article (doi:10.1093/eurheartj/ehv024) Introduction Cardiovascular disease (CVD) is a leading cause of mortality and morbidity, with an estimated 17 million deaths in 2008.1 Over 80% of CVD deaths now occur in low- and middle-income countries such as China,2 and the burden of disease from CVD is projected to increase further over the next few decades in these countries.3 Increased blood pressure is one of the most important modifiable causes of CVD,4–6 and is known to be associated with lifestyle and environmental factors. There is evidence that outdoor temperature affects blood pressure,7–9 with particularly large effects in China.8 Exposure to cold temperatures can lead to vasoconstriction and tachycardia, both of which contribute to increased blood pressure and cardiac load.10 Excess CVD mortality has been reported during the cold seasons,11–15 part of which may be driven by temperature-related increases in blood pressure. Individuals who are already suffering from CVD may be at, particularly, increased risk when exposed to cold temperatures. However, limited large-scale data exist about seasonal variation in blood pressure and CVD mortality among people with prior-CVD, especially in China where few people have access to central heating in winter and where few people with prior-CVD are properly managed.16

y, increased risk when exposed to cold temperatures. However, limited large-scale data exist about seasonal variation in blood pressure and CVD mortality among people with prior-CVD, especially in China where few people have access to central heating in winter and where few people with prior-CVD are properly managed.16 We report data from the China Kadoorie Biobank (CKB) study of over 500 000 adults aged 30–79 years who were recruited from 10 diverse regions in China during 2004–8 and followed up ever since for mortality and morbidity to 31 December 2013. The present study is of 23 000 participants who reported at the baseline visit having a physician diagnosed CVD and aims (a) to examine the association of outdoor temperature with blood pressure measured at recruitment, both overall and in different population subgroups, (b) to investigate the association between usual blood pressure and subsequent CVD mortality, and (c) to assess any seasonal variation in CVD mortality rates and then compare it with predicted risk based on seasonal changes in blood pressure.

measured at recruitment, both overall and in different population subgroups, (b) to investigate the association between usual blood pressure and subsequent CVD mortality, and (c) to assess any seasonal variation in CVD mortality rates and then compare it with predicted risk based on seasonal changes in blood pressure. Methods Baseline survey Detailed information about the study design and procedures has been reported previously.8,17 Briefly, the baseline survey took place between 2004 and 2008 in 10 geographically defined areas in China (Figure 1). At the baseline survey, detailed information about general demographic and socio-economic status, dietary and other lifestyle habits (e.g. smoking, alcohol drinking, and physical activity), indoor air pollution, medical history, and current medication were collected using an interviewer-administered laptop-based questionnaire. Participants were asked whether they had ever been diagnosed by a physician with a range of chronic diseases [e.g. diabetes, ischaemic heart disease (IHD), stroke or transient ischaemic attack (TIA), hypertension, chronic obstructive pulmonary disease, cancer, or other common diseases), and if so, the age at first diagnosis and whether they were still on treatment. Participants who had reported having a prior history of IHD, stroke/TIA, diabetes, or hypertension were additionally asked about current usage of specific drugs (aspirin, ACE-I, β-blocker, statins, diuretics, or calcium antagonist).16 A range of physical measurements were undertaken for each participant and a blood sample was collected. Figure 1 Locations of the recruitment centres in China Kadoorie Biobank. The solid circles denote urban areas, and the open circles denote rural areas.

ACE-I, β-blocker, statins, diuretics, or calcium antagonist).16 A range of physical measurements were undertaken for each participant and a blood sample was collected. Figure 1 Locations of the recruitment centres in China Kadoorie Biobank. The solid circles denote urban areas, and the open circles denote rural areas. Ethical approval was obtained from the Ethical Review Committee of the Chinese Centre for Disease Control and Prevention, Beijing, China and the Oxford Tropical Research Ethics Committee, University of Oxford, UK. All study participants provided written informed consent. Blood pressure measurement Blood pressure was measured at least twice using a UA-779 digital monitor after participants had remained at rest in a seated position for at least 5 min. If the difference between the two measurements was >10 mmHg for SBP, a third measurement was made and the last two measurements were recorded. The procedure for blood pressure measurement was standardized across the 10 study areas, and all measurements were made by trained study personnel. All devices were regularly calibrated to ensure consistency of measurements.8,17

>10 mmHg for SBP, a third measurement was made and the last two measurements were recorded. The procedure for blood pressure measurement was standardized across the 10 study areas, and all measurements were made by trained study personnel. All devices were regularly calibrated to ensure consistency of measurements.8,17 Meteorological data Daily meteorological measurements for the years 2004–2008 were obtained from China Meteorological Administration local offices at each region. Mean monthly, summer (June, July, August), and winter (December, January, February) outdoor temperatures were calculated as the average of the recorded mean daily outdoor temperatures (the average of four measurements taken at 0200, 0800, 1400, and 2000 h) for all the participants who were surveyed during that month or season.

thly, summer (June, July, August), and winter (December, January, February) outdoor temperatures were calculated as the average of the recorded mean daily outdoor temperatures (the average of four measurements taken at 0200, 0800, 1400, and 2000 h) for all the participants who were surveyed during that month or season. Follow-up for mortality and morbidity Since recruitment, participants have been followed up for cause-specific morbidity and mortality through linkage with regional disease and death registers and with the recently established national health insurance (HI) system. Causes of death are sought chiefly from official death certificates supplemented, if necessary, by reviewing medical records or undertaking verbal autopsy, a WHO standard tool to determine probable causes of death for those who died without any medical attention.18 A bespoke IT system developed by the project team for this study is used for entering the data from death and disease registers, with the death certificates or the disease reporting cards scanned. Data linkage with HI agencies is carried out every 6 months in each region, and all hospitalized events occurring in that last half-year are retrieved for matched study participants. So far, ∼98% of the study population is covered by the HI system. To minimize losses to follow-up, active follow-up (i.e. visiting local community or directly contacting participants) is also performed annually.17

and all hospitalized events occurring in that last half-year are retrieved for matched study participants. So far, ∼98% of the study population is covered by the HI system. To minimize losses to follow-up, active follow-up (i.e. visiting local community or directly contacting participants) is also performed annually.17 The main analyses of the present study only involved individuals who reported having IHD, stroke/TIA at the baseline survey, i.e. defined as ‘prior-CVD participants’. For prospective analyses, only deaths from IHD (ICD-10: I20–I25) or stroke (ICD-10: I60–I61, I63–I64) were considered, henceforth referred to as ‘CVD mortality’.

analyses of the present study only involved individuals who reported having IHD, stroke/TIA at the baseline survey, i.e. defined as ‘prior-CVD participants’. For prospective analyses, only deaths from IHD (ICD-10: I20–I25) or stroke (ICD-10: I60–I61, I63–I64) were considered, henceforth referred to as ‘CVD mortality’. Statistical methods Baseline characteristics—mean (SD) for continuous variables, N (%) for categorical variables—were calculated separately for people with or without CVD. All main analyses related to blood pressure and CVD mortality were only for prior-CVD participants. Mean systolic and diastolic blood pressure (SBP, DBP), adjusted for area, age, and sex were calculated separately for each calendar month of the study recruitment period. These blood pressure means were then plotted against the mean day of the year for each month of recruitment, i.e. regardless of the year, with January and February combined to allow for the fall in recruitment over the Chinese New Year. To obtain the estimated change in blood pressure per 10°C lower outdoor temperature, a multiple linear regression analysis was performed of individual SBP on the individual outdoor temperature, adjusted for age and sex. Observations <5°C were omitted from the regression because few participants experienced daily winter temperatures regularly well below that range, with exception of those from Harbin where the winter temperature usually drops well below −10°C and nearly all households have proper central heating from mid-October until March the following year.8 Subgroup analyses by area, sex, age, body mass index (BMI), education, use of anti-hypertensive treatment, medical history of hypertension or diabetes and, for men only, current smoking and alcohol drinking (there were too few female drinkers and smokers for reliable analyses) were performed to assess whether seasonal changes in blood pressure were modified by other known vascular risk factors. To help make allowance for the multiplicity of comparisons when many subgroup analyses are performed, the separate heterogeneity χ2 statistics for each were summed (as were their degrees of freedom) to yield a global test for heterogeneity.19 The individual heterogeneity tests are not shown where trend tests are more appropriate.

e allowance for the multiplicity of comparisons when many subgroup analyses are performed, the separate heterogeneity χ2 statistics for each were summed (as were their degrees of freedom) to yield a global test for heterogeneity.19 The individual heterogeneity tests are not shown where trend tests are more appropriate. Cox proportional hazards models were used to calculate hazard ratios (HRs), with blood pressure as the exposure variable and CVD mortality as the outcome, and stratified by sex, region (10 groups), age at risk (10 groups), and anti-hypertensive treatment status, and further adjusted for smoking, alcohol drinking, BMI, and education. The 95% confidence interval (CI) for each log HR was estimated using the ‘floating absolute risk’ method, which facilitates many different comparisons and tests for trend between different categories, rather than just pair-wise comparisons between one arbitrarily chosen reference group and each of the other categories.20 To correct for regression dilution bias,4,21 HRs in the groups determined at baseline were plotted against the usual blood pressure, i.e. the mean value of systolic blood pressure in that group at the subsequent resurvey, on average 2.6 years after the baseline survey. Finally, the number of CVD deaths and person-years for each calendar month were calculated and summed into groups by calendar month regardless of year. Poisson regression, adjusted for age, sex, and area, was used to calculate CVD mortality rates in each of the groups. Analyses were performed using SAS version 9.3 and R version 3.0.1.

ber of CVD deaths and person-years for each calendar month were calculated and summed into groups by calendar month regardless of year. Poisson regression, adjusted for age, sex, and area, was used to calculate CVD mortality rates in each of the groups. Analyses were performed using SAS version 9.3 and R version 3.0.1. Results A total of 506 673 people (99% of all participants) had data on both blood pressure and outdoor temperature on the day of baseline survey. Table 1 compares the baseline characteristics and the number of deaths from CVD during follow-up between individuals with and without prior-CVD. Overall, 23 040 (4.5%) participants reported having prior-CVD with mean years since diagnosis of 7.2 years. About one-third of those with prior-CVD was from Harbin, which is located in the far northeast of China (Tables 1 and 2). Those with prior-CVD tended to be older (mean age of 61 vs. 51 years) and more likely to have higher mean SBP (141 vs. 131 mmHg) and BMI (24.9 vs. 23.6 kg/m2). At any given age, men with prior-CVD were less likely to be current regular smokers or weekly alcohol drinkers, but more likely to be ex-smokers or ex-drinkers than men without prior-CVD. The lower proportions of current regular smokers or drinkers among men with prior-CVD were probably due to increased rates of quitting because of illness. People with prior-CVD were also more likely to report a prior diagnosis of hypertension (48 vs. 10%) or diabetes (12 vs. 3%), and to be taking anti-hypertensive treatment (31 vs. 4%). During an average of 7.1 (SD 1.3) years of follow-up, a total of 1484 CVD deaths were recorded among those with prior-CVD (748 from IHD and 736 from stroke) (Table 1). Table 1 Baseline characteristics of study participants and number of cardiovascular disease-related deaths by self-reported hospital diagnosis of prior-cardiovascular disease

.1 (SD 1.3) years of follow-up, a total of 1484 CVD deaths were recorded among those with prior-CVD (748 from IHD and 736 from stroke) (Table 1). Table 1 Baseline characteristics of study participants and number of cardiovascular disease-related deaths by self-reported hospital diagnosis of prior-cardiovascular disease Overall With prior-CVD No prior-CVD Number of participants 506 673 23 040 483 633 Age (SD, years) 52 (11) 61 (9) 51 (11) SBP (SD, mmHg) 131 (21) 141 (23) 131 (21) DBP (SD, mmHg) 78 (11) 80 (12) 78 (11) BMI (SD, kg/m2) 23.7 (3.4) 24.9 (3.6) 23.6 (3.4) Years since diagnosis (SD, years) 7.2 (7.1) – Men (%) 41.0 43.5 40.9 Current smokers (%) 61.0 43.0 61.9 Ex-smokers (%) 13.3 29.1 12.5 Weekly drinkers (%) 32.9 21.0 33.6 Ex-drinkers (%) 3.7 11.8 3.3 With central heating (%) 12.3 33.0 11.3 Self-reported hypertension (%) 11.7 48.2 10.0 Treated for hypertension (%) 4.8 30.6 3.6 Self-reported diabetes (%) 3.2 12.3 2.7 Number of IHD/stroke deathsa 7144 1484 5660 aIHD (ICD-10: I20–I25), stroke (ICD-10: I60–I61, I63–I64), follow-up duration: baseline to 31 December 2013. Table 2 Mean temperature (°C) and systolic blood pressure (mmHg) in summer and winter, by area, among participants with prior-cardiovascular disease

Overall With prior-CVD No prior-CVD Number of participants 506 673 23 040 483 633 Age (SD, years) 52 (11) 61 (9) 51 (11) SBP (SD, mmHg) 131 (21) 141 (23) 131 (21) DBP (SD, mmHg) 78 (11) 80 (12) 78 (11) BMI (SD, kg/m2) 23.7 (3.4) 24.9 (3.6) 23.6 (3.4) Years since diagnosis (SD, years) 7.2 (7.1) – Men (%) 41.0 43.5 40.9 Current smokers (%) 61.0 43.0 61.9 Ex-smokers (%) 13.3 29.1 12.5 Weekly drinkers (%) 32.9 21.0 33.6 Ex-drinkers (%) 3.7 11.8 3.3 With central heating (%) 12.3 33.0 11.3 Self-reported hypertension (%) 11.7 48.2 10.0 Treated for hypertension (%) 4.8 30.6 3.6 Self-reported diabetes (%) 3.2 12.3 2.7 Number of IHD/stroke deathsa 7144 1484 5660 aIHD (ICD-10: I20–I25), stroke (ICD-10: I60–I61, I63–I64), follow-up duration: baseline to 31 December 2013. Table 2 Mean temperature (°C) and systolic blood pressure (mmHg) in summer and winter, by area, among participants with prior-cardiovascular disease Area Latitude (°N) % with central heating Number with prior-CVDs Summer (June–August) Winter (December–February) Difference (Summer vs. Winter) Change (SE) in SBP per 10°C lower temperatureb (≥5°C only) Temp (°C) SBPa Temp (°C) SBPa Temp (°C) SBPa Harbin 46 94 7672 22.7 131 −14.1 138 36.8 −7 5.9 (0.52) Qingdao 36 15 2009 23.9 138 1.7 146 22.1 −9 8.3 (0.81) Henan 35 0 3000 26.9 139 1.5 150 25.4 −12 5.4 (0.62) Gansu 35 1 1342 22.3 139 −0.1 151 22.4 −12 9.0 (1.14) Zhejiang 31 0 817 28.1 138 6.0 153 22.2 −15 6.9 (1.05) Suzhou 31 2 1017 28.0 137 5.9 146 22.1 −9 5.8 (0.93) Sichuan 31 0 476 25.7 137 6.5 144 19.2 −7 4.3 (1.55) Hunan 28 1 2623 27.9 143 6.9 152 21.0 −9 4.8 (0.58) Liuzhou 24 0 3423 28.7 133 12.1 144 16.5 −11 7.1 (0.59) Haikou 20 0 661 28.6 129 18.5 140 10.2 −11 7.4 (1.87) Overall 29 23 040 25.5 136 3.8 145 21.7 −9 6.2 (0.24) aSBP, systolic blood pressure; adjusted for age, sex, and area (where appropriate).

4 19.2 −7 4.3 (1.55) Hunan 28 1 2623 27.9 143 6.9 152 21.0 −9 4.8 (0.58) Liuzhou 24 0 3423 28.7 133 12.1 144 16.5 −11 7.1 (0.59) Haikou 20 0 661 28.6 129 18.5 140 10.2 −11 7.4 (1.87) Overall 29 23 040 25.5 136 3.8 145 21.7 −9 6.2 (0.24) aSBP, systolic blood pressure; adjusted for age, sex, and area (where appropriate). bAdjusted for age, sex, and area (where appropriate).

4 19.2 −7 4.3 (1.55) Hunan 28 1 2623 27.9 143 6.9 152 21.0 −9 4.8 (0.58) Liuzhou 24 0 3423 28.7 133 12.1 144 16.5 −11 7.1 (0.59) Haikou 20 0 661 28.6 129 18.5 140 10.2 −11 7.4 (1.87) Overall 29 23 040 25.5 136 3.8 145 21.7 −9 6.2 (0.24) aSBP, systolic blood pressure; adjusted for age, sex, and area (where appropriate). bAdjusted for age, sex, and area (where appropriate). A seasonal cycle of mean blood pressure, especially SBP, was observed in people with prior-CVD, with highest levels in winter and lowest levels in summer in both men and women (Figure 2). Overall, SBP varied from 145 mmHg in winter to 136 mmHg in summer (P < 0.001), but the magnitude of this variation differed between regions (Table 2). Indeed, in Harbin, although the mean outdoor temperature was on average 36°C colder in winter than in summer, the mean SBP only differed by 7 mmHg between winter and summer seasons, with a small fall, rather than increase, in SBP in winter when the central heating was turned on (Figure 3). Figure 2 Monthly variation in blood pressure and outdoor temperature in people with prior-cardiovascular disease in (A) men and (B) women. The horizontal placement of each month combined over the full 4 years of recruitment represents the mean number of days since the first participant was recruited for participants recruited in average over years for that month. J/F = January and February combined (recruitment dropped in January and February due to the Chinese New Year). For both blood pressure and temperature, the mean monthly values are the mean for all participants whose baseline survey happened during that month (regardless of the year). Means of blood pressure were adjusted for sex and age. The winter months are placed centrally to display the winter peak in blood pressure.

Year). For both blood pressure and temperature, the mean monthly values are the mean for all participants whose baseline survey happened during that month (regardless of the year). Means of blood pressure were adjusted for sex and age. The winter months are placed centrally to display the winter peak in blood pressure. Figure 3 Monthly variation in systolic blood pressure and outdoor temperature among people with prior-cardiovascular disease, Other regions together vs. Harbin. Conventions as in Figure 1. There was an approximately linear inverse association between SBP and temperature >5°C, with a mean increase of 6.2 (SE 0.24) mmHg in SBP for each 10°C decrease in outdoor temperature. There was a 2-fold difference in the strength of this relationship between areas (phet = 0.003), ranging from 4.3 mmHg in Sichuan to 9.0 mmHg in Gansu (Table 2, Supplementary material online, eFigure S1). However, the association was largely consistent between most other subgroups studied (P for global test of heterogeneity = 0.02) (Figure 4). Figure 4 Seasonal variation in systolic blood pressure and outdoor temperature among people with prior-cardiovascular disease by various subgroups, with the temperature range at least >5°C. The analysis was adjusted for age, sex, and area (where appropriate). Each closed square represents a change in systolic blood pressure per 10°C lower outdoor temperatures. The dotted vertical line indicates the overall change in systolic blood pressure; the open diamond indicates it and its 95% CI.

ange at least >5°C. The analysis was adjusted for age, sex, and area (where appropriate). Each closed square represents a change in systolic blood pressure per 10°C lower outdoor temperatures. The dotted vertical line indicates the overall change in systolic blood pressure; the open diamond indicates it and its 95% CI. A positive association was observed between usual SBP and CVD mortality, with 21% (95% CI: 16–27%) higher risk of CVD mortality for each 10 mmHg higher usual SBP (Figure 5). The excess risk was similar between people treated with blood pressure lowering agents and those without (HR (95% CI) = 1.25 (1.18–1.33) vs. 1.18 (1.08–1.28), P for heterogeneity = 0.28). Overall, the adjusted CVD mortality rate was 2.35 per 1000 person-years, but fluctuated over the course of the year, with a winter peak observed (2.78 in winter vs. 1.98 per 1000 person-years in summer) (Figure 6). Among people with prior-CVD, both absolute-CVD mortality rates and the relative seasonal changes were weaker in Harbin than in the other regions (Harbin: 1.96 vs. 1.61 per 1000 person-years, rate ratio (95% CI) = 1.22 (1.07–1.39); all other combined: 4.10 vs. 2.73 per 1000 person-years, rate ratio = 1.50 (1.39–1.61)). Figure 5 Hazard ratios for cardiovascular disease mortality vs. usual blood pressure among people with prior-cardiovascular disease. Analyses were stratified by region, age, gender, and blood pressure lowering treatment status, and adjusted for education, smoking, alcohol drinking, and body mass index. The hazard ratios are plotted on a floating absolute scale. Each square has an area inversely proportional to the standard error of the log risk. Vertical lines indicate the corresponding 95% confidence intervals. Numbers above confidence intervals are of hazard ratios and those below are the numbers of cardiovascular disease deaths.

os are plotted on a floating absolute scale. Each square has an area inversely proportional to the standard error of the log risk. Vertical lines indicate the corresponding 95% confidence intervals. Numbers above confidence intervals are of hazard ratios and those below are the numbers of cardiovascular disease deaths. Figure 6 Seasonal variation in cardiovascular disease mortality rates among people with prior-cardiovascular disease, between 2004 and 2013. Deaths and person-days at risk for a given months are totalled across the follow-up period. The analysis is adjusted for age group, study site, and sex, but not for year of follow-up. To make the curve smoother, two calendar months are combined and winter months are again placed centrally, as in Figure 1. Vertical lines indicate the corresponding 95% confidence intervals. Numbers above confidence intervals are of mortality rates (per 1000 person-years) and those below are the numbers of deaths.

-up. To make the curve smoother, two calendar months are combined and winter months are again placed centrally, as in Figure 1. Vertical lines indicate the corresponding 95% confidence intervals. Numbers above confidence intervals are of mortality rates (per 1000 person-years) and those below are the numbers of deaths. Discussion In this large study of over 23 000 individuals with prior-CVD who were recruited from the general communities in 10 diverse regions of China, we observed substantial seasonal variation of blood pressure, especially, in areas where there is little use of central heating in the cold months. Overall, the mean SBP was 9 mmHg higher in winter than in summer, and, >5°C in outdoor temperature, SBP was 6.2 mmHg higher for each 10°C decrease in temperature. Among people with prior-CVD, blood pressure is a strong independent predictor of subsequent CVD mortality, with each 10 mmHg higher usual SBP associated with ∼21% higher risk of CVD death. Mirroring the seasonal variation of blood pressure, there was a 41% increase in CVD mortality during winter.

crease in temperature. Among people with prior-CVD, blood pressure is a strong independent predictor of subsequent CVD mortality, with each 10 mmHg higher usual SBP associated with ∼21% higher risk of CVD death. Mirroring the seasonal variation of blood pressure, there was a 41% increase in CVD mortality during winter. Several studies have provided evidence about seasonal variation of blood pressure, but most of them were in general populations, with only a few involving high-risk populations such as people in old age,7,22–24 those with low BMI,7,25 hypertension,7,24,26 or end-stage renal disease.27 The present large study in China provides large-scale evidence about climate-related changes in blood pressure in people with prior-CVD. In this Chinese population, the seasonal variation in blood pressure was generally more extreme than that observed in western populations, and was abolished by the use of home central heating. The large variation in blood pressure between seasons observed in the present study has strong implications for detection and clinical management of hypertension. People maybe more likely to be diagnosed with hypertension if examined in winter than in summer, and hypertensive patients may suffer even higher blood pressure in the cold season due to inadequate blood pressure control. In a prospective study of 184 elderly Israeli patients with essential hypertension, supplementary anti-hypertensive treatment was required during winter in 38% of them.28 We were not able to assess seasonal changes within individuals since it is not practicable to monitor changes in participants' blood pressure and treatment during follow-up among 0.5 million participants. In our study, the mean blood pressure was much higher in winter than in summer among both those taking and those not taking blood pressure lowering treatment. Blood pressure, especially SBP, is an independent predictor of CVD risk not only in general population but also in those with prior vascular diseases.4,29 The beneficial effects of blood pressure lowering treatments on the risks of major CVD diseases are well established.30 In this study population with prior-CVD, the strength of the association – i.e.

an independent predictor of CVD risk not only in general population but also in those with prior vascular diseases.4,29 The beneficial effects of blood pressure lowering treatments on the risks of major CVD diseases are well established.30 In this study population with prior-CVD, the strength of the association – i.e. 21% higher CVD mortality per 10 mmHg higher usual SBP—was weaker than previous estimates from general populations (both Chinese and Western), that have estimated ∼40% higher stroke mortality and 30% higher IHD death.4 This may be explained by the fact that a high proportion of the individuals with prior-CVD in the present study were on anti-hypertensive treatment that may well attenuate the effects of blood pressure on subsequent CVD risk.

nese and Western), that have estimated ∼40% higher stroke mortality and 30% higher IHD death.4 This may be explained by the fact that a high proportion of the individuals with prior-CVD in the present study were on anti-hypertensive treatment that may well attenuate the effects of blood pressure on subsequent CVD risk. A winter peak in CVD mortality has been consistently reported among general populations in many studies.11,12,31–35 However, only one previous study has simultaneously examined seasonal changes in blood pressure and mortality in people with prior-CVD.15 In that study of 19 000 male British civil servants aged 40–69 years with 25-year follow-up , the seasonal effect on all-cause mortality was more extreme among 3284 men with prevalent IHD than those without (rate ratio for winter vs. summer of 1.38 and 1.18, respectively, P = 0.03). Among those with prior-IHD participants, the rate ratios were 1.31 for IHD mortality and 1.48 for stroke.15 In our study, where mean SBP differed by ∼10 mmHg between winter and summer, we observed 41% higher CVD mortality in winter compared with summer, which is somewhat greater than predicted from our prospective analysis of prognostic effect of blood pressure on CVD mortality, although it is comparable with the risk estimates for a prolonged 10 mmHg SBP difference reported from the general population.4 It is unlikely that all the epidemiologically expected risk would be observed by the short- to medium-term rise in blood pressure during winter and it is likely that there are additional factors that contributed to the winter rise in mortality. Nevertheless, some of the observed winter peak in CVD mortality may well be driven by rises in blood pressure due to the cold temperature.

ld be observed by the short- to medium-term rise in blood pressure during winter and it is likely that there are additional factors that contributed to the winter rise in mortality. Nevertheless, some of the observed winter peak in CVD mortality may well be driven by rises in blood pressure due to the cold temperature. There are some limitations to our study. Although it was explicit that all reported prior disease history should be based on physician diagnosis, it was not possible with such a large cohort to adjudicate all the self-reported cases retrospectively. However, there is an evidence from other studies of similar nature in China that self-reported history of CVD showed a good consistency with hospital records.36 Moreover, in a separate analysis when comparing the CVD mortality rates observed in our prior-CVD population with rates in those without, a 3-fold higher risk was consistently appeared in each month [overall HR = 2.98 (95% CI: 2.80–3.18)] (Supplementary material online, eFigure S2). This provides indirect evidence that self-reports of prior-CVD in our study were generally reliable. Occlusive and haemorrhagic CVD may involve different pathophysiological mechanisms.37 Although the present study included a sufficiently large number of CVD deaths to assess the overall effect of season on CVD mortality, the statistical power to separately examine the seasonal patterns of subtype of CVD mortality is rather limited. The isolated clinic BP measurements used in our study are unable to provide as reliable an assessment of blood pressure as ambulatory blood pressure monitoring, and may lead us to underestimate any patterns or associations. However, the high comparability of baseline characteristics between participants enrolled at different seasons or months, as reported previously,8 suggests that seasonal variation in blood pressure is likely to be driven primarily by changes in outdoor temperature rather than by other lifestyle factors.

r associations. However, the high comparability of baseline characteristics between participants enrolled at different seasons or months, as reported previously,8 suggests that seasonal variation in blood pressure is likely to be driven primarily by changes in outdoor temperature rather than by other lifestyle factors. Our study findings suggest that seasonal variation in CVD mortality rates in people with prior-CVD result, at least partly, from seasonal variation of blood pressure. Given the high prevalence of CVD in the Chinese population, especially of stroke, even moderate increases in CVD mortality rates during the winter months, as shown in this study, must account for large excess numbers of CVD deaths. The results from the present analyses suggest that seasonal changes in blood pressure should be taken into account in the diagnosis and treatment of hypertension and CVD. For patients with prior-CVD or other high-risk individuals, more intensive anti-hypertensive treatment and more frequent blood pressure monitoring may be required in winter to achieve the same blood pressure control as in other seasons. Longer follow-up of this large cohort study and the use of well-characterized non-fatal CVD events may allow for more detailed analyses of seasonal variation separately in the main components of CVD, namely IHD, ischaemic, and haemorrhagic stroke. Supplementary material Supplementary material is available at European Heart Journal online.

Our study findings suggest that seasonal variation in CVD mortality rates in people with prior-CVD result, at least partly, from seasonal variation of blood pressure. Given the high prevalence of CVD in the Chinese population, especially of stroke, even moderate increases in CVD mortality rates during the winter months, as shown in this study, must account for large excess numbers of CVD deaths. The results from the present analyses suggest that seasonal changes in blood pressure should be taken into account in the diagnosis and treatment of hypertension and CVD. For patients with prior-CVD or other high-risk individuals, more intensive anti-hypertensive treatment and more frequent blood pressure monitoring may be required in winter to achieve the same blood pressure control as in other seasons. Longer follow-up of this large cohort study and the use of well-characterized non-fatal CVD events may allow for more detailed analyses of seasonal variation separately in the main components of CVD, namely IHD, ischaemic, and haemorrhagic stroke. Supplementary material Supplementary material is available at European Heart Journal online. Funding The baseline survey and the first re-survey were supported by a research grant from the Kadoorie Charitable Foundation in Hong Kong. The long-term continuation of the project during 2009–2014 is supported by program grants from the Wellcome Trust in the UK (088158/Z/09/Z) and the Chinese Ministry of Science and Technology. The UK Medical Research Council, the British Heart Foundation and Cancer Research UK also provide core funding to the Clinical Trial Service Unit and Epidemiological Studies Unit at Oxford University for the project. Funding to pay the Open Access publication charges for this article was provided by Welcome Trust.

ogy. The UK Medical Research Council, the British Heart Foundation and Cancer Research UK also provide core funding to the Clinical Trial Service Unit and Epidemiological Studies Unit at Oxford University for the project. Funding to pay the Open Access publication charges for this article was provided by Welcome Trust. Conflict of interest: none declared. Appendix Members of the China Kadoorie Biobank collaborative group (a) International steering committee Liming Li (PI), Junshi Chen, Rory Collins, Richard Peto, Zhengming Chen (PI). (b) Study Coordinating Centres International Co-ordinating Centre, Oxford: Zhengming Chen, Garry Lancaster, Xiaoming Yang, Alex Williams, Margaret Smith, Ling Yang, Yumei Chang, Iona Millwood, Yiping Chen, Sarah Lewington. National Co-ordinating Centre, Beijing: Yu Guo, Zheng Bian, Can Hou, Yunlong Tan, Huiyan Zhou. Regional Co-ordinating Centres, 10 areas in China: Qingdao Qingdao CDC: Zengchang Pang, Shaojie Wang, Yun Zhang, Kui Zhang. Licang CDC: Silu Liu, Wei Hou. Heilongjiang Provincial CDC: Zhonghou Zhao, Shumei Liu, Zhigang Pang. Nangang CDC: Weijia Feng, Shuling Wu, Liqiu Yang, Huili Han, Hui He, Bo Yu. Hainan Provincial CDC: Xianhai Pan, Shanqing Wang, Hongmei Wang. Meilan CDC: Xinhua Hao, Chunxing Chen, Shuxiong Lin, Xiangyang Zheng. Jiangsu Provincial CDC: Xiaoshu Hu, Minghao Zhou, Ming Wu, Ran Tao. Suzhou CDC: Yeyuan Wang, Yihe Hu, Liangcai Ma, Renxian Zhou, Guanqun Xu, Yan Lu. Guangxi Provincial CDC: Baiqing Dong, Naying Chen, Ying Huang. Liuzhou CDC: Mingqiang Li, Jinhuai Meng, Zhigao Gan, Jiujiu Xu, Yun Liu, Jingxin Qing. Sichuan

Provincial CDC: Xianhai Pan, Shanqing Wang, Hongmei Wang. Meilan CDC: Xinhua Hao, Chunxing Chen, Shuxiong Lin, Xiangyang Zheng. Jiangsu Provincial CDC: Xiaoshu Hu, Minghao Zhou, Ming Wu, Ran Tao. Suzhou CDC: Yeyuan Wang, Yihe Hu, Liangcai Ma, Renxian Zhou, Guanqun Xu, Yan Lu. Guangxi Provincial CDC: Baiqing Dong, Naying Chen, Ying Huang. Liuzhou CDC: Mingqiang Li, Jinhuai Meng, Zhigao Gan, Jiujiu Xu, Yun Liu, Jingxin Qing. Sichuan Provincial CDC: Xianping Wu, Yali Gao, Ningmei Zhang Pengzhou CDC: Guojin Luo, Xiangsan Que, Xiaofang Chen. Gansu Provincial CDC: Pengfei Ge, Jian He, Xiaolan Ren. Maiji CDC: Hui Zhang, Enke Mao, Guanzhong Li, Zhongxiao Li, Jun He, Yulong Lei, Xiaoping Wang. Henan Provincial CDC: Guohua Liu, Baoyu Zhu, Gang Zhou, Shixian Feng. Huixian CDC: Yulian Gao, Tianyou He, Li Jiang, Jianhua Qin, Huarong Sun. Zhejiang Provincial CDC: Liqun Liu, Min Yu, Yaping Chen, Ruying Hu. Tongxiang CDC: Zhixiang Hu, Jianjin Hu, Yijian Qian, Zhiying Wu, Chunmei Wang, Lingli Chen. Hunan Provincial CDC: Wen Liu, Guangchun Li, Huilin Liu. Liuyang CDC: Xiangquan Long, Xin Xu, Youping Xiong, Zhongwen Tan, Xuqiu Xie, Yunfang Peng, Weifang Jia.

inhibitor would be safe from a bleeding perspective, it would reduce the risk of thrombotic complications during the waiting time, and save hospital resources. The risk of bleeding complications in relation to discontinuation time has not previously been investigated in real life in sufficiently large patient cohorts. The newer and more potent platelet inhibitors may increase the risk of major CABG-related bleeding complications, but it is unclear whether the incidence differs between ticagrelor and clopidogrel in real life after adjustment for time since discontinuation and other factors that influence bleeding risk. The aims of the present real-life study were to investigate whether a shorter discontinuation time before surgery increases the risk of major bleeding with ticagrelor or clopidogrel and to compare the unadjusted and adjusted incidence of CABG-related bleeding complications between clopidogrel and ticagrelor.

The newer and more potent platelet inhibitors may increase the risk of major CABG-related bleeding complications, but it is unclear whether the incidence differs between ticagrelor and clopidogrel in real life after adjustment for time since discontinuation and other factors that influence bleeding risk. The aims of the present real-life study were to investigate whether a shorter discontinuation time before surgery increases the risk of major bleeding with ticagrelor or clopidogrel and to compare the unadjusted and adjusted incidence of CABG-related bleeding complications between clopidogrel and ticagrelor. Methods Study patients The study was a retrospective analysis of prospectively collected data. All 2244 ACS patients on DAPT who underwent acute or urgent CABG in Sweden from January 2012 to December 2013 were included in a retrospective study. The patients were treated preoperatively with acetylsalicylic acid (aspirin) and either ticagrelor (n = 1266) or clopidogrel (n = 978) within the last 14 days before surgery. In 2012–13, ticagrelor was introduced in the Swedish regional guidelines to replace clopidogrel as the first treatment option in ACS patients planned for interventional treatment. Prasugrel is also used in Sweden. Patients treated with prasugrel were included in the registry but not in this analysis, due to the small number, just 10 patients, over the study period. The patients underwent CABG at one of the eight cardiothoracic surgery centres in Sweden: Umeå University Hospital (n = 291), Uppsala University Hospital (n = 97), Karolinska University Hospital (n = 267), Örebro University Hospital (n = 90), Linköping University Hospital (n = 326), Sahlgrenska University Hospital (n = 473), Blekinge Hospital (n = 130), and Skåne University Hospital, Lund (n = 570). The study was conducted in accordance with the Declaration of Helsinki, and was approved by the Regional Research Ethics Committee in Gothenburg on 30 April 2014 (reference number 031-14), which waived the need for individual consent from the patients before inclusion in the registry. Preoperative patient characteristics are summarized in Table 1. Table 1 Baseline demographics and preoperative variables

ed by the Regional Research Ethics Committee in Gothenburg on 30 April 2014 (reference number 031-14), which waived the need for individual consent from the patients before inclusion in the registry. Preoperative patient characteristics are summarized in Table 1. Table 1 Baseline demographics and preoperative variables Clopidogrel (n = 978) Ticagrelor (n = 1266) P-value Female gender 203 (20.8%) 271 (21.4%) 0.75 Age (years) 68.4 ± 9.5 n = 978 67.8 ± 9.4 n = 1266 0.082 BMI (kg/m2) 27.3 ± 4.2 n = 976 27.3 ± 4.0 n = 1262 0.38 Diabetes 252 (25.8%) 347 (27.4%) 0.44 Preoperative haemoglobin (g/L) 137 ± 16 n = 978 136 ± 15 n = 1266 0.068 Preoperative platelet count (109/L) 246 ± 73 n = 965 250 ± 73 n = 1255 0.066 Preoperative creatinine (μmol/L) 95 ± 72 n = 975 91 ± 42 n = 1259 0.86 Preoperative prothrombin time (INR) 1.09 ± 0.30 n = 958 1.08 ± 0.16 n = 1242 0.97 Preoperative APTT (s) 36 ± 19 n = 878 36 ± 18 n = 1163 0.0056 EuroSCORE I (additive) Mean 5.62 ± 3.28 Median 5.0 (0.0–20) n = 974 Mean 5.50 ± 3.14 Median 5.0 (0.0–22) n = 1254 0.49 Ejection fraction (%) >50 607 (62.4%) 792 (63.4%) 0.14 31–50 290 (29.8%) 392 (31.4%) >20–30 66 (6.8%) 60 (4.8%) ≤20 10 (1.0%) 6 (0.5%) Warfarin treatment at any time before surgery 47 (4.8%) 26 (2.1%) 0.0005 Fondaparinux at any time before surgery 645 (66.2%) 919 (72.6%) 0.0011 LMWH at any time before surgery 221 (22.6%) 373 (29.6%) 0.0002 GPIIb/IIIa inhibitor before surgery 2 (0.2%) 3 (0.2%) 1.0 Discontinuation of clopidogrel/ticagrelor (days) 5.2 ± 3.6 Median 5 (0–14) 5.9 ± 3.5 Median 6 (0–14) <0.0001 Discontinuation (h) 0–24 65 (6.6%) 110 (8.7%) <0.0001 24–48 147 (15.0%) 62 (4.9%) 48–72 76 (7.8%) 54 (4.3%) 72–96 71 (7.3%) 89 (7.0%) 96–120 73 (7.5%) 104 (8.2%) >120 546 (55.8%) 847 (66.9%) Acute surgery 99 (10.1%) 159 (12.6%) 0.080 Values are given as mean ± SD, median (interval), or frequency (percent). P-values from Fisher's exact test for dichotomous variables, Mantel–Haenszel χ2 test for ordered categorical variables, and Mann–Whitney U-test for continuous variables.

.2%) >120 546 (55.8%) 847 (66.9%) Acute surgery 99 (10.1%) 159 (12.6%) 0.080 Values are given as mean ± SD, median (interval), or frequency (percent). P-values from Fisher's exact test for dichotomous variables, Mantel–Haenszel χ2 test for ordered categorical variables, and Mann–Whitney U-test for continuous variables. BMI, body mass index; INR, international normalized ratio; APTT, activated partial thromboplastin time; LMWH, low-molecular-weight heparin; GPIIb/IIIa, glycoprotein IIb/IIIa.

.2%) >120 546 (55.8%) 847 (66.9%) Acute surgery 99 (10.1%) 159 (12.6%) 0.080 Values are given as mean ± SD, median (interval), or frequency (percent). P-values from Fisher's exact test for dichotomous variables, Mantel–Haenszel χ2 test for ordered categorical variables, and Mann–Whitney U-test for continuous variables. BMI, body mass index; INR, international normalized ratio; APTT, activated partial thromboplastin time; LMWH, low-molecular-weight heparin; GPIIb/IIIa, glycoprotein IIb/IIIa. Study design The patients were identified in the SWEDEHEART registry11 and/or institutional databases. Data were obtained from SWEDEHEART, hospital records, and the participating hospitals' surgical databases, and were compiled in a nationwide registry. The patients were grouped according to the platelet inhibitor used. If a patient had been treated with both agents, the patient was classified according to the last medication before surgery. Outcome variables assessed were incidence of major bleeding complications in the clopidogrel and ticagrelor groups, overall and after adjustment. We also compared the incidence of major bleeding complications within and between the ticagrelor and clopidogrel groups when the platelet inhibitor was discontinued 0–72, 72–120, or >120 h before surgery, assessed postoperative bleeding volume during the first 12 postoperative hours in relation to the timing of discontinuation of the platelet inhibitor, and the incidence and number of allogeneic blood products [red blood cells (RBCs), plasma, and platelets] transfused during the index hospital stay in relation to the period of discontinuation. Thirty-day mortality and thrombotic events during hospital stay were only registered for safety reasons. No statistical testing or detailed analysis was performed for these variables, due to the lack of statistical power.

platelets] transfused during the index hospital stay in relation to the period of discontinuation. Thirty-day mortality and thrombotic events during hospital stay were only registered for safety reasons. No statistical testing or detailed analysis was performed for these variables, due to the lack of statistical power. Definitions Major bleeding was defined according to four published definitions: Bleeding Academic Research Consortium (BARC) type 4, CABG-related bleeding (bleeding resulting in death, or reoperation due to bleeding, or intracranial haemorrhage, or transfusion of 5 or more units of RBCs over 48 h, or chest tube drainage in excess of 2000 mL over 24 h),12 Blood Conservation Using Antifibrinolytics in a Randomized Trial (BART; postoperative blood loss >1500 mL/12 h, or re-exploration due to bleeding, or RBC transfusion of 10 units or more, or death because of bleeding),13 PLATelet inhibition and patient Outcomes (PLATO) life-threatening bleeding (fatal bleeding, or pericardial bleeding requiring repeat surgery, or drop in haemoglobin of ≥50 g/L, or transfusion of 4 or more units of RBCs), and PLATO major bleeding (pericardial bleeding requiring repeat surgery, or drop in haemoglobin of ≥30 g/L, or transfusion of 2 or more units of RBCs).2

-threatening bleeding (fatal bleeding, or pericardial bleeding requiring repeat surgery, or drop in haemoglobin of ≥50 g/L, or transfusion of 4 or more units of RBCs), and PLATO major bleeding (pericardial bleeding requiring repeat surgery, or drop in haemoglobin of ≥30 g/L, or transfusion of 2 or more units of RBCs).2 Major bleeding is reported for all four bleeding definitions while adjusted data are reported only for the BARC-CABG definition. Bleeding volume was defined as mediastinal drainage volume. A thrombotic event was defined as ischaemic stroke with duration exceeding 24 h and verified by CT or MRI, pulmonary embolism, or deep vein thrombosis evidenced during the index hospitalization. Acute surgery was defined as procedure starting within 24 h of acceptance. Urgent surgery was defined as within hospitalization for ACS. Clinical management Patients were treated in accordance with standard practice at the participating centres. All patients received 75 mg aspirin daily. Ticagrelor or clopidogrel was administered with a loading dose followed by 75 mg once daily for clopidogrel or 90 mg twice daily for ticagrelor. According to the current European guidelines, the platelet inhibitor should be discontinued 5 days prior to surgery if clinically feasible.9 Fondaparinux and low-molecular-weight heparin (LMWH) were discontinued at least 12 h before non-acute surgery. Aspirin was not discontinued before surgery.

e daily for ticagrelor. According to the current European guidelines, the platelet inhibitor should be discontinued 5 days prior to surgery if clinically feasible.9 Fondaparinux and low-molecular-weight heparin (LMWH) were discontinued at least 12 h before non-acute surgery. Aspirin was not discontinued before surgery. Statistical analysis The two groups were compared at baseline by Fisher's exact test for dichotomous variables, the Mantel–Haenszel χ2 test for ordered categorical variables, and the Mann–Whitney U-test for continuous variables. Logistic regression modelling was used to identify factors related to major bleeding and to compare incidence of bleeding between discontinuation groups. Factors that were significantly different between platelet inhibitors and associated with major bleeding with a P-value of <0.10 were included in a multivariable logistic regression model for adjustment. The sample size (at least 500 patients in each group) was chosen to achieve 80% power in finding a significant difference in the incidence of major bleeding complications between clopidogrel and ticagrelor after stratification by time from discontinuation of medication to surgery. The power estimation was based on a previous single-centre study.6 Data are presented as mean (± standard deviation), median (range) or frequency (percent). Statistical significance was assumed with a two-sided P-value of <0.05. No adjustment for multiplicity was performed. SAS software, version 9.4 (SAS Institute, Cary, NC, USA), was used for statistical analysis.

ingle-centre study.6 Data are presented as mean (± standard deviation), median (range) or frequency (percent). Statistical significance was assumed with a two-sided P-value of <0.05. No adjustment for multiplicity was performed. SAS software, version 9.4 (SAS Institute, Cary, NC, USA), was used for statistical analysis. Results Baseline variables Demographic data are presented in Table 1. More clopidogrel-treated patients were treated with warfarin at some time prior to surgery, 4.8 vs. 2.1% (P = 0.0005). Only 13 patients, 5 in the ticagrelor group and 8 in the clopidogrel group, were treated with warfarin <5 days before surgery. The preoperative prothrombin time did not differ between groups. Preoperative treatment with LMWH and fondaparinux was more common in the ticagrelor group. In 44.2% of the clopidogrel-treated patients, the platelet inhibitor was discontinued <5 days prior to surgery, compared with 33.1% in the ticagrelor group (P < 0.0001), and mean discontinuation was 5.2 ± 3.6 days for the clopidogrel-treated patients compared with 5.9 ± 3.5 days for the ticagrelor-treated patients (P < 0.0001).

of the clopidogrel-treated patients, the platelet inhibitor was discontinued <5 days prior to surgery, compared with 33.1% in the ticagrelor group (P < 0.0001), and mean discontinuation was 5.2 ± 3.6 days for the clopidogrel-treated patients compared with 5.9 ± 3.5 days for the ticagrelor-treated patients (P < 0.0001). Procedures All but 16 of the patients (99.3%) were operated with cardiopulmonary bypass, and mean cardiopulmonary bypass time was marginally longer in the clopidogrel group (81 ± 37 vs. 77 ± 31 min; P = 0.025). Total operation time was not significantly different between clopidogrel and ticagrelor (193 ± 67 vs. 189 ± 57 min; P = 0.66) and neither was duration of aortic cross clamp (47 ± 22 vs. 46 ± 20 min; P = 0.23). The number of distal anastomoses did not differ between the ticagrelor group and the clopidogrel group (3.2 ± 1.0 vs. 3.3 ± 1.0, P = 0.12), and concomitant valve repair was performed in 2.0% of cases for clopidogrel and 2.5% for ticagrelor (P = 0.54). Table 2 Postoperative outcome variables

vs. 46 ± 20 min; P = 0.23). The number of distal anastomoses did not differ between the ticagrelor group and the clopidogrel group (3.2 ± 1.0 vs. 3.3 ± 1.0, P = 0.12), and concomitant valve repair was performed in 2.0% of cases for clopidogrel and 2.5% for ticagrelor (P = 0.54). Table 2 Postoperative outcome variables Clopidogrel (n = 978) Ticagrelor (n = 1266) P-value Postoperative bleeding (mL) 12 h Mean 614 ± 393 Median 500 (0–3940) Mean 579 ± 411 Median 470 (100–5475) 0.0017 24 h Mean 830 ± 498 Median 700 (0–5398) Mean 813 ± 554 Median 671 (140–7501) 0.093 Incidence of transfusion Any 559 (57.2%) 645 (51.0%) 0.0038 RBC 517 (52.9%) 589 (46.6%) 0.0028 Plasma 238 (24.4%) 240 (19.0%) 0.0025 Platelets 226 (23.1%) 263 (20.8%) 0.20 Amount of transfusions (U) Any Mean 3.95 ± 7.25 Median 2 (0–72) Mean 3.92 ± 10.18 Median 1 (0–176) 0.0012 RBC Mean 2.42 ± 3.92 Median 1 (0–41) Mean 2.26 ± 4.93 Median 0 (0–73) 0.0012 Plasma Mean 1.01 ± 2.93 Median 0 (0–36) Mean 1.01 ± 4.11 Median 0 (0–79) 0.0036 Platelets Mean 0.525 ± 1.271 Median 0 (0–12) Mean 0.653 ± 1.869 Median 0 (0–24) 0.43 Reoperation due to bleeding 74 (7.6%) 77 (6.1%) 0.19 Lowest postoperative haemoglobin (g/L) Mean 92 ± 12 Median 91 (51–147) Mean 93 ± 12 Median 92 (52–150) 0.018 Highest postoperative creatinine (μmol/L) Mean 115 ± 81 Median 94 (42–980) Mean 112 ± 75 Median 91 (40–1055) 0.082 Time in ICU (days) Mean 2.26 ± 3.37 Median 1.0 (0–41) Mean 2.0 ± 3.04 Median 1.0 (0–38) 0.0008 Hospital length-of-stay after CABG (days) Mean 7.83 ± 5.11 Median 7.0 (1–62) Mean 7.58 ± 5.20 Median 6.0 (1–69) 0.0005 Values are given as mean ± SD, median (interval), or frequency (percent). P-values from Fisher's exact test for dichotomous variables, Mantel–Haenszel χ2 test for ordered categorical variables, and Mann–Whitney U-test for continuous variables.

Mean 7.83 ± 5.11 Median 7.0 (1–62) Mean 7.58 ± 5.20 Median 6.0 (1–69) 0.0005 Values are given as mean ± SD, median (interval), or frequency (percent). P-values from Fisher's exact test for dichotomous variables, Mantel–Haenszel χ2 test for ordered categorical variables, and Mann–Whitney U-test for continuous variables. RBC, red blood cells; ICU, intensive care unit; CABG, coronary artery bypass grafting. Major bleeding Overall, there were significantly less CABG-related major bleeding complications in ticagrelor-treated patients according to three of the four definitions (Figure 1): BARC-CABG 12.9 vs. 17.6% (P = 0.0024), BART major bleeding 8.8 vs. 11.6% (P = 0.041), and PLATO life-threatening major bleeding 46.8 vs. 54.0% (P = 0.0008) for ticagrelor- and clopidogrel-treated patients, respectively. PLATO major bleeding did not differ significantly (89.9 vs. 92.1%; P = 0.076). Incidence of BARC-CABG major bleeding by day of discontinuation is shown in Figure 2. Other postoperative outcome variables are summarized in Table 2. Figure 1 Incidence of major bleeding complications according to BARC-CABG, BART, PLATO life-threatening, and PLATO major bleeding (P-values from Fisher's exact test between ticagrelor and clopidogrel). BARC, Bleeding Academic Research Consortium; CABG, coronary artery bypass grafting; BART, Blood Conservation Using Antifibrinolytics in a Randomized Trial; PLATO, PLATelet inhibition and patient Outcomes.

life-threatening, and PLATO major bleeding (P-values from Fisher's exact test between ticagrelor and clopidogrel). BARC, Bleeding Academic Research Consortium; CABG, coronary artery bypass grafting; BART, Blood Conservation Using Antifibrinolytics in a Randomized Trial; PLATO, PLATelet inhibition and patient Outcomes. Figure 2 Incidence of BARC-CABG major bleeding by day of discontinuation of clopidogrel/ticagrelor to surgery (P-values from Fisher's exact test). BARC, Bleeding Academic Research Consortium; CABG, coronary artery bypass grafting. The use of ticagrelor was associated with a reduced risk of BARC-CABG major bleeding both before adjustment [odds ratio (OR) 0.69 (95% confidence interval, CI, 0.55–0.87), P = 0.002] and after adjustment for time since discontinuation and the other factors significantly influencing the risk of bleeding in the univariable analysis [adjusted OR 0.72 (95% CI 0.56–0.92), P = 0.012]. Other factors associated with major bleeding in univariable logistic regression are presented in Table 3. Table 3 Factors associated with major bleeding in univariable logistic regression

factors significantly influencing the risk of bleeding in the univariable analysis [adjusted OR 0.72 (95% CI 0.56–0.92), P = 0.012]. Other factors associated with major bleeding in univariable logistic regression are presented in Table 3. Table 3 Factors associated with major bleeding in univariable logistic regression No major bleeding (n = 1909) Major bleeding (n = 335) Unadjusted OR (95% CI) P-value Female gender 20% 27.8% 1.54 (1.18–2.01) 0.0013 Age (years, OR per 10 years) 67.8 ± 9.3 69.9 ± 9.9 1.28 (1.12–1.46) 0.0002 BMI (kg/m2) 27.4 ± 4.1 26.9 ± 4.1 0.97 (0.94–1.00) 0.0541 Diabetes 26.7% 26.9% 1.01 (0.78–1.32) 0.9225 Preoperative haemoglobin (g/L, OR per 10 units) 138 ± 15 132 ± 17 0.80 (0.74–0.86) <0.0001 Preoperative platelet count (109/L, OR per 10 units) 247 ± 71 257 ± 82 1.02 (1.00–1.03) 0.0156 Preoperative creatinine (μmol/L, OR per 10 units) 90 ± 52 104 ± 83 1.03 (1.01–1.05) 0.0004 Preoperative prothrombin time (INR) 1.1 ± 0.2 1.1 ± 0.1 1.10 (0.71–1.71) 0.6582 Preoperative APTT (s, OR per 10 units) 35 ± 16 41 ± 27 1.13 (1.07–1.19) <0.0001 EuroSCORE I (additive) 5.2 ± 2.9 7.3 ± 4.1 1.20 (1.16–1.24) <0.0001 Ejection fraction (%) >50 66.7% 49.2% 1.84 (1.56–2.17) <0.0001 31–50 30.1% 37.7% 20–30 2.6% 11.2% ≤20 0.5% 1.9% Platelet inhibitor (ticagrelor) 57.8% 48.7% 0.69 (0.55–0.87) 0.0020 Warfarin treatment 3.1% 4.5% 1.49 (0.83–2.66) 0.1796 Fondaparinux treatment 70.3% 67.2% 0.86 (0.67–1.11) 0.2509 Heparin treatment 14.8% 22.2% 1.65 (1.23–2.19) 0.0007 LMWH treatment 27.2% 22.8% 0.79 (0.60–1.04) 0.0889 Operation duration (min, OR per h) 183 ± 49 231 ± 99 1.92 (1.71–2.15) <0.0001 CPB duration (min, OR per h) 76 ± 30 95 ± 48 2.37 (1.95–2.88) <0.0001 Cross-clamp duration (min, OR per h) 46 ± 19 51 ± 29 1.88 (1.38–2.55) <0.0001 Concomitant valve proc. 1.6% 5.7% 3.84 (2.14–6.91) <0.0001 Acute surgery 8.8% 27.2% 3.89 (2.92–5.19) <0.0001 Discontinuation of platelet inhibitor (days) 5.9 ± 3.5 3.9 ± 3.5 0.84 (0.81–0.87) <0.0001 Odds ratio (OR) per increase of one unit of the predictor if not otherwise indicated.

19 51 ± 29 1.88 (1.38–2.55) <0.0001 Concomitant valve proc. 1.6% 5.7% 3.84 (2.14–6.91) <0.0001 Acute surgery 8.8% 27.2% 3.89 (2.92–5.19) <0.0001 Discontinuation of platelet inhibitor (days) 5.9 ± 3.5 3.9 ± 3.5 0.84 (0.81–0.87) <0.0001 Odds ratio (OR) per increase of one unit of the predictor if not otherwise indicated. BMI, body mass index; INR, international normalized ratio; APTT, activated partial thromboplastin time; LMWH, low-molecular-weight heparin; CPB, cardiopulmonary bypass. Bleeding volume and transfusions Overall, ticagrelor-treated patients bled less after surgery, and received fewer transfusions of blood products (Table 2). However, when medication was discontinued <24 h before surgery, ticagrelor-treated patients bled markedly more and received more transfusions than clopidogrel-treated patients (Table 4). Table 4 Bleeding and transfusions by time from discontinuation of platelet inhibitor

fewer transfusions of blood products (Table 2). However, when medication was discontinued <24 h before surgery, ticagrelor-treated patients bled markedly more and received more transfusions than clopidogrel-treated patients (Table 4). Table 4 Bleeding and transfusions by time from discontinuation of platelet inhibitor Discontinuation Clopidogrel Ticagrelor P-value Mean ± SD Median Mean ± SD Median Postoperative bleeding first 12 h (mL) 0–24 h 663 ± 627 488 (340–721) 813 ± 478 670 (498–1103) <0.001 24–48 h 714 ± 462 600 (415–890) 641 ± 337 585 (400–753) 0.514 48–72 h 659 ± 313 570 (440–815) 709 ± 707 510 (370–735) 0.207 72–96 h 682 ± 462 560 (400–790) 630 ± 541 450 (343–738) 0.118 96–120 h 701 ± 454 520 (405–800) 550 ± 296 450 (350–698) 0.036 Over 120 h 555 ± 313 480 (358–653) 534 ± 363 450 (349–610) 0.021 Any transfusion (units) 0–24 h 8.9 ± 13.4 4 (2–11) 14.7 ± 22.5 8.5 (4–17) 0.001 24–48 h 5.8 ± 8.6 3 (0–8) 7.9 ± 9.8 4 (2–11) 0.111 48–72 h 4.9 ± 6.4 3 (0–6) 7.6 ± 17.4 3 (0–7) 0.779 72–96 h 4.6 ± 8.7 2 (0–6) 2.6 ± 4.7 2 (0–3) 0.013 96–120 h 3.6 ± 4.7 2 (0–6) 2.0 ± 3.7 0 (0–3) 0.025 Over 120 h 2.7 ± 5.4 0 (0–3) 2.4 ± 6.3 0 (0–2) 0.044 RBC transfusion (units) 0–24 h 4.9 ± 6.8 2 (1–6) 6.9 ± 9.8 4.5 (2–9) 0.028 24–48 h 3.4 ± 4.5 2 (0–5) 4.4 ± 5.7 2 (0–6.3) 0.400 48–72 h 2.8 ± 3.5 2 (0–3.8) 4.0 ± 9.9 2 (0–4) 0.618 72–96 h 3.0 ± 5.3 2 (0–4) 1.7 ± 3.2 0 (0–2) 0.033 96–120 h 2.3 ± 2.9 1 (0–4) 1.3 ± 2.1 0 (0–2) 0.046 Over 120 h 1.7 ± 3.0 0 (0–3) 1.6 ± 3.2 0 (0–2) 0.096 Plasma transfusion (units) 0–24 h 2.5 ± 5.5 0 (0–3) 4.6 ± 10.1 2 (0–4.3) 0.022 24–48 h 1.5 ± 3.6 0 (0–2) 1.9 ± 3.2 0 (0–2.3) 0.154 48–72 h 1.3 ± 2.7 0 (0–1.8) 1.8 ± 4.9 0 (0–1) 0.480 72–96 h 0.96 ± 2.8 0 (0–0) 0.54 ± 1.3 0 (0–0) 0.490 96–120 h 0.84 ± 1.6 0 (0–1) 0.35 ± 1.2 0 (0–0) 0.001 Over 120 h 0.70 ± 2.4 0 (0–0) 0.56 ± 2.7 0 (0–0) 0.005 Platelet transfusion (units) 0–24 h 1.5 ± 2.3 0 (0–2) 3.2 ± 3.7 2 (0.8–4) <0.001 24–48 h 0.94 ± 1.5 0 (0–2) 1.6 ± 2.2 1 (0–2) 0.033 48–72 h 0.79 ± 1.4 0 (0–1) 1.8 ± 3.7 0 (0–2) 0.430 72–96 h 0.68 ± 1.4 0 (0–1) 0.44 ± 0.81 0 (0–0.5) 0.563 96–120 h 0.51 ± 1.0 0 (0–1) 0.32 ± 0.90 0 (0–0) 0.086 Over 120 h 0.25 ± 0.84 0 (0–0) 0.24 ± 0.95 0 (0–0) 0.357 Values are given as mean ± SD and median (25–75 percentiles). For comparison between groups, the Mann–Whitney U-test was used.

–1) 1.8 ± 3.7 0 (0–2) 0.430 72–96 h 0.68 ± 1.4 0 (0–1) 0.44 ± 0.81 0 (0–0.5) 0.563 96–120 h 0.51 ± 1.0 0 (0–1) 0.32 ± 0.90 0 (0–0) 0.086 Over 120 h 0.25 ± 0.84 0 (0–0) 0.24 ± 0.95 0 (0–0) 0.357 Values are given as mean ± SD and median (25–75 percentiles). For comparison between groups, the Mann–Whitney U-test was used. RBC, red blood cells; SD, standard deviation. Impact of time since discontinuation The difference in the incidence of major bleeding complications between ticagrelor and clopidogrel was mainly driven by a significant reduction in major bleeding complications in the ticagrelor group when clopidogrel/ticagrelor was discontinued 72–120 h before surgery [unadjusted OR 0.39 (95% CI 0.20–0.76), P = 0.006, Figure 3]. When either drug was discontinued according to the current guidelines (>120 h before surgery), there was no significant difference in the incidence of major bleeding complications between ticagrelor- and clopidogrel-treated patients [9 vs. 12%; unadjusted OR 0.72 (95% CI 0.51–1.02), P = 0.065].

), P = 0.006, Figure 3]. When either drug was discontinued according to the current guidelines (>120 h before surgery), there was no significant difference in the incidence of major bleeding complications between ticagrelor- and clopidogrel-treated patients [9 vs. 12%; unadjusted OR 0.72 (95% CI 0.51–1.02), P = 0.065]. Within the ticagrelor group, there was no significant difference in major bleeding complications between discontinuation 72–120 or >120 h before surgery [unadjusted OR 0.93 (95% CI 0.53–1.64), P = 0.80], whereas discontinuation 0–72 h was associated with a significantly higher rate of major bleeding compared with both 72–120 h [unadjusted OR 5.17 (95% CI 2.89–9.27), P < 0.0001] and >120 h [unadjusted OR 4.81 (95% CI 3.34–6.95), P < 0.0001]. In contrast, clopidogrel-treated patients had a higher incidence of major bleeding complications when discontinued 72–120 compared with >120 h before surgery [unadjusted OR 1.71 (95% CI 1.04–2.79), P = 0.033]. Likewise, in the clopidogrel group, discontinuation 0–72 h was associated with an increased incidence of major bleeding compared with 72–120 h [unadjusted OR 1.67 (95% CI 1.02–2.73), P = 0.042] and >120 h [unadjusted OR 2.85 (95% CI 1.98–4.10), P < 0.0001] (Figure 3).

[unadjusted OR 1.71 (95% CI 1.04–2.79), P = 0.033]. Likewise, in the clopidogrel group, discontinuation 0–72 h was associated with an increased incidence of major bleeding compared with 72–120 h [unadjusted OR 1.67 (95% CI 1.02–2.73), P = 0.042] and >120 h [unadjusted OR 2.85 (95% CI 1.98–4.10), P < 0.0001] (Figure 3). Mortality and thrombotic events Thirty-day mortality was 1.7% in the ticagrelor group and 2.7% in the clopidogrel group. The mortality was significantly higher in patients with major bleeding complications [9.9 vs. 0.7%, unadjusted OR 14.78 (95% CI 7.82–27.93), P < 0.0001]. Thrombotic events during the postoperative hospital stay occurred in 2.3% of the ticagrelor group and 2.8% of the clopidogrel group. Figure 3 Incidence of BARC-CABG major bleeding stratified by time from discontinuation of clopidogrel/ticagrelor to surgery (P-values denoting difference between the platelet inhibitors, and within clopidogrel/ticagrelor between discontinuation strata as indicated). BARC, Bleeding Academic Research Consortium; CABG, coronary artery bypass grafting. Discussion The main finding of this nationwide registry study was that discontinuation of the platelet inhibitor 3 days before surgery, as opposed to 5 days, did not increase the incidence of major bleeding complications in ticagrelor-treated patients, but increased the incidence in clopidogrel-treated patients. In addition, a lower incidence of major bleeding complications was observed in ticagrelor-treated patients, except when the platelet inhibitor was discontinued <72 h before surgery.

the incidence of major bleeding complications in ticagrelor-treated patients, but increased the incidence in clopidogrel-treated patients. In addition, a lower incidence of major bleeding complications was observed in ticagrelor-treated patients, except when the platelet inhibitor was discontinued <72 h before surgery. The timing of discontinuation of the platelet inhibitor is influenced by a number of factors, including institutional guidelines, the patient's condition, logistical reasons, and the individual surgeon's decision. This is illustrated in the present study where 45% of the clopidogrel patients and 33% of the ticagrelor patients were operated after a shorter discontinuation than the 5 days recommended in the guidelines. We used this variation to compare the incidence of major bleeding complications after different discontinuation times. The data in the present study were collected during 2012 and 2013. During this period, ticagrelor replaced clopidogrel as first-choice P2Y12-receptor antagonist for ACS patients in the Swedish guidelines. The introduction of ticagrelor did not occur simultaneously in all parts of Sweden; instead, there was a gradual increase in the use of ticagrelor over time. Consequently, during this time period, there were ACS patients with the same characteristics treated with either ticagrelor or clopidogrel. We deliberately used this time period for the study to obtain comparable groups.

y in all parts of Sweden; instead, there was a gradual increase in the use of ticagrelor over time. Consequently, during this time period, there were ACS patients with the same characteristics treated with either ticagrelor or clopidogrel. We deliberately used this time period for the study to obtain comparable groups. The most clinically important observation in this study was the lack of difference in major bleeding complications when ticagrelor was discontinued 3 days before surgery compared with 5 days. This suggests that it is safe to operate on ACS patients treated with ticagrelor earlier after discontinuation than is currently recommended in guidelines. A reduction in the waiting time from 5 to 3 days would reduce the risk for thrombotic events while waiting for CABG, and save hospital resources. The results of the study also confirm the detrimental effect of major bleeding complications on outcome after CABG. The unadjusted 30-day mortality was almost 15 times greater in patients with major bleeding compared with those without bleeding complications (9.9 vs. 0.7%). This corroborates previous registry studies about the effects of bleeding complications5,6,14,15 and emphasizes the importance of, if possible, avoiding excessive bleeding after CABG. Timely discontinuation of platelet inhibitors would help to achieve this.

ared with those without bleeding complications (9.9 vs. 0.7%). This corroborates previous registry studies about the effects of bleeding complications5,6,14,15 and emphasizes the importance of, if possible, avoiding excessive bleeding after CABG. Timely discontinuation of platelet inhibitors would help to achieve this. We also observed a lower overall incidence of major bleeding complications in patients treated with ticagrelor. It should be pointed out that the result in this regard is hypothesis-generating rather than conclusive, given the retrospective observational design of the study. The lower incidence of major bleeding complications in the ticagrelor group in the present study could not be explained by the longer discontinuation time in the ticagrelor group (Table 1), since the difference was maintained also after adjustment. Instead, it is mainly explained by the lower incidence of major bleeding complications in the ticagrelor group when ticagrelor/clopidogrel was discontinued 3–5 days before surgery. This is, in turn, explained by differences in the pharmacokinetic and pharmacodynamic profiles of ticagrelor and clopidogrel. Ticagrelor is a direct-acting P2Y12-receptor antagonist with greater antiplatelet effect and more consistent platelet inhibition than clopidogrel.16,17 Ticagrelor has also faster on-set of action (within 30 min of loading) and faster off-set, i.e. the antiplatelet effect of ticagrelor returns faster to baseline than with clopidogrel.16 Despite these known difference in off-set time of the antiplatelet effect, current guidelines recommend that both ticagrelor and clopidogrel are discontinued 5 days before surgery.9,10 The results of the present study instead support the use of differentiated discontinuation times for ticagrelor and clopidogrel, i.e. 3 days for ticagrelor and 5 days for clopidogrel.

f the antiplatelet effect, current guidelines recommend that both ticagrelor and clopidogrel are discontinued 5 days before surgery.9,10 The results of the present study instead support the use of differentiated discontinuation times for ticagrelor and clopidogrel, i.e. 3 days for ticagrelor and 5 days for clopidogrel. There was a higher incidence of transfusions and a larger bleeding volume in the ticagrelor group when the platelet inhibitor was discontinued within 24 h before surgery (Table 4), even if the difference in the incidence of major bleeding complications according to the BARC-CABG definition did not reach statistical significance (38 vs. 31%). This indicates, in accordance with previous studies,6,18 that the risk for severe bleeding is higher with ticagrelor than with clopidogrel if it cannot be discontinued before surgery, which also is consistent with the stronger antiplatelet effect of ticagrelor compared with clopidogrel.16,17 This increased bleeding risk with ticagrelor may be considered before loading if there is high risk for acute CABG. In contrast, there was no significant difference in bleeding complications in the present study when clopidogrel and ticagrelor were discontinued in accordance with the current guidelines, i.e. 5 days or more before CABG.

eeding risk with ticagrelor may be considered before loading if there is high risk for acute CABG. In contrast, there was no significant difference in bleeding complications in the present study when clopidogrel and ticagrelor were discontinued in accordance with the current guidelines, i.e. 5 days or more before CABG. The incidence of major bleeding complications was higher in women than in men (Table 3). This may at least partly be due to the higher risk of transfusions in women, since transfusions are included in all definitions of major bleeding utilized in the present study. Postoperative bleeding volume was not larger in women than in men in this material (data not shown). The higher risk of transfusion in women is mainly caused by lower preoperative haemoglobin levels and smaller blood volume, leading to a higher grade of haemodilution during cardiopulmonary bypass. Current guidelines recommend discontinuation of the platelet inhibitor at a fixed time interval before surgery. In the future, platelet function tests may be used to optimize timing of the procedure rather than the choice of drug and the corresponding interval. This strategy is supported by a recent study, where Ranucci et al.19 show a significant association between the grade of ADP-dependent platelet aggregability and CABG-related bleeding complications in clopidogrel-treated patients. So far, no data are available in ticagrelor-treated patients.

the corresponding interval. This strategy is supported by a recent study, where Ranucci et al.19 show a significant association between the grade of ADP-dependent platelet aggregability and CABG-related bleeding complications in clopidogrel-treated patients. So far, no data are available in ticagrelor-treated patients. The present study has all inherent limitations of an observational study, including selection bias and unregistered confounders. Unregistered confounders may include, e.g. history of bleeding, liver disease, heart failure, and renal failure. To prove the current findings, a randomized clinical trial with different discontinuation times would be required. The study also has strengths; it represents a complete nationwide cohort of ACS patients on DAPT with ticagrelor or clopidogrel operated with acute or urgent CABG in Sweden during a 2-year period, giving a relatively large cohort of similar patients. The gradual introduction of the new inhibitor over the country gave an opportunity for comparison, even though the groups are not randomized. In conclusion, discontinuation of ticagrelor 3 days before surgery did not increase the risk of major bleeding complications after CABG compared with 5 days. The overall risk of major CABG-related bleeding complications was lower with ticagrelor than with clopidogrel in this real-life observational study. The difference was driven by a lower incidence with ticagrelor than with clopidogrel when the platelet inhibitor was discontinued 72–120 h before surgery.

red with 5 days. The overall risk of major CABG-related bleeding complications was lower with ticagrelor than with clopidogrel in this real-life observational study. The difference was driven by a lower incidence with ticagrelor than with clopidogrel when the platelet inhibitor was discontinued 72–120 h before surgery. Funding This research was conducted with support from an Investigator Sponsored Study Programme of AstraZeneca and the Swedish Heart and Lung Foundation [Grant numbers: 20120372, 2014021]. The study sponsors had no influence on the analysis and interpretation of data, on the writing of the report, or on the decision to submit the paper for publication. Funding to pay the Open Access publication charges for this article was provided by AstraZeneca. Conflict of interest: E.C.H. and A.J. have received speaker's honorarium from AstraZeneca. A.J. has also received support from AstraZeneca for other investigator-initiated studies.

See page 1060 for the editorial comment on this article (doi:10.1093/eurheartj/ehv517) Clinical perspective Contrast-enhanced cardiac magnetic resonance (CMR) imaging is the established approach for imaging infarct pathology in survivors of ST-elevation myocardial infarction (STEMI). The pathophysiological and prognostic importance of infarct pathology disclosed by non-contrast native T1 CMR mapping in acute STEMI patients is unknown. We performed a prospective single centre cohort study in 300 reperfused STEMI patients who underwent CMR 2 days and 6 months (n = 267) post-myocardial infarction and clinical follow-up (median duration 2.5 years). Infarct core pathology revealed by native T1 mapping was feasible [n = 288 (96%) with evaluable data] and had superior prognostic value compared with infarct core T2 and myocardial haemorrhage, and similar prognostic value compared with microvascular obstruction revealed by late gadolinium enhancement CMR. Infarct core native T1 is a novel non-contrast CMR biomarker with potential for infarct characterization and prognostication in STEMI survivors. Introduction Myocardial infarct size1,2 and microvascular obstruction3–5 revealed by contrast-enhanced cardiac magnetic resonance (CMR) reflect the efficacy of reperfusion therapy and are prognostically important findings in survivors of ST-elevation myocardial infarction (STEMI).

Clinical perspective Contrast-enhanced cardiac magnetic resonance (CMR) imaging is the established approach for imaging infarct pathology in survivors of ST-elevation myocardial infarction (STEMI). The pathophysiological and prognostic importance of infarct pathology disclosed by non-contrast native T1 CMR mapping in acute STEMI patients is unknown. We performed a prospective single centre cohort study in 300 reperfused STEMI patients who underwent CMR 2 days and 6 months (n = 267) post-myocardial infarction and clinical follow-up (median duration 2.5 years). Infarct core pathology revealed by native T1 mapping was feasible [n = 288 (96%) with evaluable data] and had superior prognostic value compared with infarct core T2 and myocardial haemorrhage, and similar prognostic value compared with microvascular obstruction revealed by late gadolinium enhancement CMR. Infarct core native T1 is a novel non-contrast CMR biomarker with potential for infarct characterization and prognostication in STEMI survivors. Introduction Myocardial infarct size1,2 and microvascular obstruction3–5 revealed by contrast-enhanced cardiac magnetic resonance (CMR) reflect the efficacy of reperfusion therapy and are prognostically important findings in survivors of ST-elevation myocardial infarction (STEMI). Human tissue has fundamental magnetic properties, including the longitudinal (spin-lattice) relaxation time (native T1 in milliseconds). Native T1 is influenced by water content, binding with macromolecules (water mobility), and cell content.6,7 Native T1 CMR does not involve an intravenous contrast agent. Tissue water content increases as a result of ischaemia and longer T1 times may represent a biomarker of localized myocardial injury.8–16

liseconds). Native T1 is influenced by water content, binding with macromolecules (water mobility), and cell content.6,7 Native T1 CMR does not involve an intravenous contrast agent. Tissue water content increases as a result of ischaemia and longer T1 times may represent a biomarker of localized myocardial injury.8–16 The clinical significance of tissue changes within the infarct core in patients with acute reperfused STEMI has not been directly assessed. We hypothesized that baseline native T1 values would be (i) inversely associated with the severity of MI, including microvascular obstruction, (ii) independently associated with left ventricular (LV) remodelling, and (iii) independently associated with pre-defined health outcomes. Should these hypotheses be confirmed then infarct core native T1 mapping without an intravenous contrast agent might have potential as an alternative biomarker to microvascular obstruction revealed by contrast-enhanced CMR. To investigate these hypotheses, we measured native T1 in myocardial regions of interest in STEMI patients undergoing serial CMR imaging 2 days and 6 months post-MI. We assessed the clinical associates of native T1 within the hypo-intense infarct core and subsequent LV remodelling and examined its association with all-cause death and first hospitalization for heart failure.

n myocardial regions of interest in STEMI patients undergoing serial CMR imaging 2 days and 6 months post-MI. We assessed the clinical associates of native T1 within the hypo-intense infarct core and subsequent LV remodelling and examined its association with all-cause death and first hospitalization for heart failure. Methods Study population and ST-elevation myocardial infarction management We performed an observational prospective CMR cohort study in a single regional cardiac centre between 14 July 2011 and 22 November 2012. Three hundred and forty three STEMI patients provided written informed consent to undergo CMR 2 days and 6 months post-MI. Patients were eligible if they had an indication for primary percutaneous coronary intervention (PCI) or thrombolysis for acute STEMI due to a history of symptoms consistent with acute myocardial ischaemia and with supporting changes on the electrocardiogram (ECG) (i.e. ST-segment elevation or new left bundle-branch block).17 Exclusion criteria represented standard contra-indications to contrast CMR, including a pacemaker and estimated glomerular filtration rate <30 mL/min/1.73 m2. The study was approved by the National Research Ethics Service and all participants provided written informed consent. Acute STEMI management followed contemporary guidelines.17,18 Aspiration thrombectomy, direct stenting, anti-thrombotic drugs, and other therapies were administered according to clinical judgment (Supplementary material online, Methods). The ClinicalTrials.gov identifier is NCT02072850.

ovided written informed consent. Acute STEMI management followed contemporary guidelines.17,18 Aspiration thrombectomy, direct stenting, anti-thrombotic drugs, and other therapies were administered according to clinical judgment (Supplementary material online, Methods). The ClinicalTrials.gov identifier is NCT02072850. Cardiac magnetic resonance acquisition Cardiac magnetic resonance was performed on a Siemens MAGNETOM Avanto (Erlangen, Germany) 1.5-Tesla scanner with a 12-element phased array cardiac surface coil.19 The imaging protocol included cine magnetic resonance imaging with steady-state free precession (SSFP), native T1 mapping,15,20 T2 mapping,21,22 T2*-mapping, and delayed-enhancement phase-sensitive inversion-recovery pulse sequences.23 The scan acquisitions were spatially co-registered and also included different slice orientations to enhance diagnostic confidence. Cardiac magnetic resonance was also performed in 50 healthy volunteers of similar age and gender in order to obtain local reference values for myocardial native T1 (Supplementary material online). Patients and healthy volunteers underwent the same imaging protocol except that healthy volunteers <45 years did not receive gadolinium. The coefficients of variation for native T1 were also measured (Supplementary material online, Results).

cal reference values for myocardial native T1 (Supplementary material online). Patients and healthy volunteers underwent the same imaging protocol except that healthy volunteers <45 years did not receive gadolinium. The coefficients of variation for native T1 were also measured (Supplementary material online, Results). Native T1 maps were acquired in three short-axial slices (basal, mid, and apical), using an optimized modified look-locker inversion recovery (MOLLI) T1-mapping investigational prototype sequence15,20 before contrast administration (Supplementary material online, Methods; work-in-progress 448, Siemens Healthcare). The MOLLI T1 cardiac-gated acquisition involved three inversion recovery prepared Look-Locker experiments combined within one protocol.15 The CMR parameters were: bandwidth ∼1090 Hz/pixel, flip angle 35°, echo time (TE) 1.1 ms, T1 of first experiment 100 ms, TI increment 80 ms, matrix 192 × 124 pixels, spatial resolution 2.2 × 1.8 × 8.0 mm, slice thickness 8 mm, and scan time 17 heartbeats. The prototype pulse sequence did not involve motion correction. T2 maps were acquired in contiguous short-axis slices covering the whole ventricle, using an investigational prototype T2-prepared TrueFisp sequence21,22 (Supplementary material online, Methods). Typical imaging parameters were: bandwidth ∼947 Hz/pixel, flip angle 70°, T2 preparations: 0, 24, and 55 ms, respectively, matrix 160 × 105 pixels, spatial resolution 2.6 × 2.1 × 8.0 mm, and slice thickness 8 mm.

g an investigational prototype T2-prepared TrueFisp sequence21,22 (Supplementary material online, Methods). Typical imaging parameters were: bandwidth ∼947 Hz/pixel, flip angle 70°, T2 preparations: 0, 24, and 55 ms, respectively, matrix 160 × 105 pixels, spatial resolution 2.6 × 2.1 × 8.0 mm, and slice thickness 8 mm. T2*-maps were obtained using an investigational prototype T2* map sequence acquired in three short-axis slices (basal, mid, and apical). Typical imaging parameters were: bandwidth ∼814 (8×) Hz/pixel; flip angle 18°; matrix 256 × 115; spatial resolution 2.6 × 1.6 × 10 mm; slice thickness 8 mm. Early gadolinium enhancement (EGE) imaging was acquired 1, 3, 5, and 7 min post-contrast injection using a TrueFISP readout and fixed inversion time (TI) of 440 ms. Late gadolinium enhancement images covering the entire LV were acquired 10–15 min after IV injection of 0.15 mmol/kg of gadoterate meglumine (Gd2+-DOTA, Dotarem, Guebert S.A.) using segmented phase-sensitive inversion recovery turbo fast low-angle shot sequence.23 Typical imaging parameters were: matrix = 192 × 256, flip angle = 25°, TE = 3.36 ms, bandwidth = 130 Hz/pixel, echo spacing = 8.7 ms, and trigger pulse = 2. The voxel size was 1.8 × 1.3 × 8 mm3. Inversion times were individually adjusted to optimize nulling of apparently normal myocardium (typical values, 200–300 ms).

Typical imaging parameters were: matrix = 192 × 256, flip angle = 25°, TE = 3.36 ms, bandwidth = 130 Hz/pixel, echo spacing = 8.7 ms, and trigger pulse = 2. The voxel size was 1.8 × 1.3 × 8 mm3. Inversion times were individually adjusted to optimize nulling of apparently normal myocardium (typical values, 200–300 ms). Cardiac magnetic resonance analyses The images were analysed on a Siemens work-station by observers with at least 3 years CMR experience (N.A., D.C., I.M, and S.R.). All of the images were reviewed by experienced CMR cardiologists (C.B. and N.T.). Left ventricular dimensions, volumes, and ejection fraction were quantified using computer-assisted planimetry (syngo MR®, Siemens Healthcare, Erlangen, Germany). The late gadolinium enhancement images were analysed for infarct size and microvascular obstruction by observers (N.A. and I.M.) who were blinded to all of the other data. In healthy volunteers, the absence of late gadolinium enhancement was determined qualitatively by visual assessment.

Healthcare, Erlangen, Germany). The late gadolinium enhancement images were analysed for infarct size and microvascular obstruction by observers (N.A. and I.M.) who were blinded to all of the other data. In healthy volunteers, the absence of late gadolinium enhancement was determined qualitatively by visual assessment. Native T1 mapping: standardized measurements in myocardial regions of interest Native T1 mapping is a CMR method providing a parametric colour-encoded anatomical map in which the T1 value is encoded in each pixel24 (Figure 1). The native T1 map analyses were informed by contemporary CMR guidelines.24 Left ventricular contours were delineated with computer-assisted planimetry on the raw T1 image and copied onto the colour-encoded spatially co-registered map. Apical segments were not included because of partial volume effects. Particular care was taken to delineate regions of interest with adequate margins of separation from tissue interfaces prone to partial volume averaging such as between myocardium and blood.19,24,25 Each T1 map image was assessed for the presence of artefacts relating to susceptibility effects, or cardio-respiratory motion. Each colour map was evaluated against the original images. When artefacts occurred the affected segments were not included in the analysis. Figure 1 Three patients with acute ST-elevation myocardial infarction treated by primary PCI and with the same anti-thrombotic therapies, including aspirin, clopidogrel, heparin, and intravenous tirofiban. Each patient had normal thrombolysis in myocardial infarction Grade 3 flow at the end of PCI. Cardiac magnetic resonance imaging was performed for each patient 2 days later. (A) Patient with no T1 hypo-intense infarct core and no microvascular obstruction. Native T1 within the injury zone (middle) measured 1211 ms. Acute infarct size revealed by late gadolinium enhancement (right) was 22.2%. The left ventricular ejection fraction and left ventricular end-diastolic volume were 55.2% and 143.1 mL, respectively. Analysis of the repeat magnetic resonance imaging scan after 6 months follow-up indicated that the final infarct size was 15.6% of left ventricular mass and the left ventricular end-diastolic volume had reduced to 103.0 mL. This patient had an uncomplicated clinical course. (B) Patient with both T1 hypo-intense infarct core and microvascular obstruction.

nance imaging scan after 6 months follow-up indicated that the final infarct size was 15.6% of left ventricular mass and the left ventricular end-diastolic volume had reduced to 103.0 mL. This patient had an uncomplicated clinical course. (B) Patient with both T1 hypo-intense infarct core and microvascular obstruction. T1 mapping (middle) revealed a hypo-intense region within the infarct core, corresponding to the area of microvascular obstruction on contrast-enhanced magnetic resonance imaging (right). Native T1 within the infarct core measured 1036 ms, which was substantially lower than the T1 value measured at the periphery of the infarct zone (1193 ms). Acute infarct size revealed by late gadolinium enhancement (right) was 33.0%. Microvascular obstruction depicted as the central dark zone within the infarct territory was 3.6% of left ventricular mass. The left ventricular ejection fraction and end-diastolic volume were 45.8% and 199.3 mL, respectively. The final infarct size at 6 months was 22.6% of left ventricular mass and the left ventricular end-diastolic volume had increased to 221.8 mL. This patient was re-hospitalized for new onset heart failure during follow-up.

left ventricular ejection fraction and end-diastolic volume were 45.8% and 199.3 mL, respectively. The final infarct size at 6 months was 22.6% of left ventricular mass and the left ventricular end-diastolic volume had increased to 221.8 mL. This patient was re-hospitalized for new onset heart failure during follow-up. In STEMI patients, myocardial T1 values were segmented spatially and regions of interest were defined as (i) remote myocardium, (ii) injured myocardium, and (iii) infarct core. The regions of interest were planimetered to include the entire area of interest with distinct margins of separation from tissue interfaces to avoid partial volume averaging. The remote myocardium region of interest was defined as myocardium 180° from the affected zone with no visible evidence of infarction, oedema, or wall motion abnormalities (assessed by inspecting corresponding contrast-enhanced T1-weighted, T2-weighted, and cine images, respectively). The infarct zone region of interest was defined as myocardium with pixel values (T1 or T2) >2 SD from remote myocardium on T2-weighted CMR.21,22 The hypo-intense infarct core was defined as an area in the centre of the infarct territory having a mean T1 value of at least 2 standard deviations (SDs) below the T1 value of the periphery of the area at risk.21,22 The assessment of T1 maps and adjudication (present/absent) of a hypo-intense core was performed independently by D.C.

e infarct core was defined as an area in the centre of the infarct territory having a mean T1 value of at least 2 standard deviations (SDs) below the T1 value of the periphery of the area at risk.21,22 The assessment of T1 maps and adjudication (present/absent) of a hypo-intense core was performed independently by D.C. In healthy volunteers, the mid-ventricular T1 colour-encoded map was segmented into six equal segments, using the anterior right ventricular-LV insertion point as the reference point.26 T1 was measured in each of these segments, and regions of interest were planimetered distinct and separate from blood-pool and tissue interfaces. These segmental values were also averaged to provide one value per subject. Results are presented as average values for segments and slices.

n point as the reference point.26 T1 was measured in each of these segments, and regions of interest were planimetered distinct and separate from blood-pool and tissue interfaces. These segmental values were also averaged to provide one value per subject. Results are presented as average values for segments and slices. Infarct definition and size The presence of acute infarction was established based on abnormalities in cine wall motion, rest first-pass myocardial perfusion, and delayed-enhancement imaging in two imaging planes. In addition, supporting changes on the ECG and coronary angiogram were also required. Acute infarction was considered present only if late gadolinium enhancement was confirmed on both the axial and long-axis acquisitions. The myocardial mass of late gadolinium (grams) was quantified using computer-assisted planimetry and the territory of infarction was delineated using a signal-intensity threshold of >5 SDs above a remote reference region and expressed as a percentage of total LV mass.27 Infarct regions with evidence of microvascular obstruction were included within the infarct area and the extent of microvascular LV ventricular mass was also measured. The measurements of infarct size were performed by I.M. and N.A. Microvascular obstruction Microvascular obstruction was defined as a dark zone on EGE imaging 1, 3, 5, and 7 min post-contrast injection that remained present within an area of LGE at 15 min. Identification of microvascular obstruction was performed independently by I.M. and N.A.

Infarct definition and size The presence of acute infarction was established based on abnormalities in cine wall motion, rest first-pass myocardial perfusion, and delayed-enhancement imaging in two imaging planes. In addition, supporting changes on the ECG and coronary angiogram were also required. Acute infarction was considered present only if late gadolinium enhancement was confirmed on both the axial and long-axis acquisitions. The myocardial mass of late gadolinium (grams) was quantified using computer-assisted planimetry and the territory of infarction was delineated using a signal-intensity threshold of >5 SDs above a remote reference region and expressed as a percentage of total LV mass.27 Infarct regions with evidence of microvascular obstruction were included within the infarct area and the extent of microvascular LV ventricular mass was also measured. The measurements of infarct size were performed by I.M. and N.A. Microvascular obstruction Microvascular obstruction was defined as a dark zone on EGE imaging 1, 3, 5, and 7 min post-contrast injection that remained present within an area of LGE at 15 min. Identification of microvascular obstruction was performed independently by I.M. and N.A. Area-at-risk Area-at-risk was defined as LV myocardium with pixel values (T2) >2 SDs from remote myocardium.4,21,22,28–30 In order to assess the area-at-risk, the epi- and endocardial contours on the last corresponding T2-weighted raw image with an echo time of 55 ms were planimetered.21 Contours were then copied to the computed T2 map and corrected when necessary by consulting the SSFP cine images.

Ds from remote myocardium.4,21,22,28–30 In order to assess the area-at-risk, the epi- and endocardial contours on the last corresponding T2-weighted raw image with an echo time of 55 ms were planimetered.21 Contours were then copied to the computed T2 map and corrected when necessary by consulting the SSFP cine images. Myocardial salvage Myocardial salvage was calculated by subtraction of per cent infarct size from per cent area at risk.4,30,31 The myocardial salvage index was calculated by dividing the myocardial salvage area by the initial area at risk. Adverse remodelling Adverse remodelling was pre-defined as an increase in LV end-diastolic volume ≥20% at 6 months from baseline.3 Myocardial haemorrhage On the T2* maps, a region of reduced signal intensity within the infarcted area, with a T2* value of <20 ms32–35 was considered to confirm the presence of myocardial haemorrhage. Electrocardiogram A 12-lead ECG was obtained before coronary reperfusion and 60 min afterwards. The extent of ST-segment resolution on the ECG assessed 60 min after reperfusion compared with the baseline ECG before reperfusion17 was expressed as complete (≥70%), incomplete (30% to <70%), or none (≤30%). Laboratory analyses The acquisition of the ECGs and blood samples for biochemical and hematologic analyses are described in Supplementary material online, Methods.

Electrocardiogram A 12-lead ECG was obtained before coronary reperfusion and 60 min afterwards. The extent of ST-segment resolution on the ECG assessed 60 min after reperfusion compared with the baseline ECG before reperfusion17 was expressed as complete (≥70%), incomplete (30% to <70%), or none (≤30%). Laboratory analyses The acquisition of the ECGs and blood samples for biochemical and hematologic analyses are described in Supplementary material online, Methods. Pre-specified health outcomes We pre-specified adverse health outcomes that are pathophysiologically linked with STEMI. The primary composite outcome was (i) all-cause death or first heart failure hospitalization (Supplementary material online, Methods). Other health outcomes included major adverse cardiac events (MACEs) defined as cardiac death, non-fatal MI, or hospitalization for heart failure.

pathophysiologically linked with STEMI. The primary composite outcome was (i) all-cause death or first heart failure hospitalization (Supplementary material online, Methods). Other health outcomes included major adverse cardiac events (MACEs) defined as cardiac death, non-fatal MI, or hospitalization for heart failure. Research staff screened for events from enrolment by checking the medical records and by contacting patients and their primary and secondary care physicians, as appropriate with no loss to follow-up (Figure 2). Each serious adverse event (SAE) was reviewed by a cardiologist who was independent of the research team and blinded to all of the clinical and CMR data. The SAEs were defined according to standard guidelines36,37 (Supplementary material online, Methods) and categorized as having occurred either during the index admission or post-discharge. All study participants were followed up for a minimum of 18 months after discharge. The median duration of follow-up was of 845 days [post-discharge censor duration (range) 598–1098 days]. Figure 2 Flow diagram of the cohort study. Statistical analyses The sample size calculation is described in Supplementary material online, Methods. We estimated that at least 30 MACE events would occur based on a conservative estimate of the event rate (10–12%) at 18 months.

Research staff screened for events from enrolment by checking the medical records and by contacting patients and their primary and secondary care physicians, as appropriate with no loss to follow-up (Figure 2). Each serious adverse event (SAE) was reviewed by a cardiologist who was independent of the research team and blinded to all of the clinical and CMR data. The SAEs were defined according to standard guidelines36,37 (Supplementary material online, Methods) and categorized as having occurred either during the index admission or post-discharge. All study participants were followed up for a minimum of 18 months after discharge. The median duration of follow-up was of 845 days [post-discharge censor duration (range) 598–1098 days]. Figure 2 Flow diagram of the cohort study. Statistical analyses The sample size calculation is described in Supplementary material online, Methods. We estimated that at least 30 MACE events would occur based on a conservative estimate of the event rate (10–12%) at 18 months. Categorical variables are expressed as number and percentage of patients. Most continuous variables followed a normal distribution and are therefore presented as means together with SD. Those variables that did not follow a normal distribution are presented as medians with interquartile range. Differences in continuous variables between groups were assessed by the Student's t-test or analysis of variance (ANOVA) for continuous data with normal distribution, otherwise the non-parametric Wilcoxon rank sum test or Kruskal–Wallis test. Differences in categorical variables between groups were assessed using a χ2 test or Fisher's test, as appropriate. Correlation analyses were Pearson or Spearman tests, as indicated. Random effects models were used to compute inter- and intra-rater reliability measures [intra-class correlation coefficient (ICC)] for the reliability of infarct core native T1 values measured independently by 2 observers in 12 randomly selected patients from the cohort.

es were Pearson or Spearman tests, as indicated. Random effects models were used to compute inter- and intra-rater reliability measures [intra-class correlation coefficient (ICC)] for the reliability of infarct core native T1 values measured independently by 2 observers in 12 randomly selected patients from the cohort. Univariable and multivariable linear regression methods to identify associates of T1 values for (i) remote myocardium, (ii) injured myocardium within the area at risk, (iii) infarct core in all patients, and (iv) in patients without late microvascular obstruction are described in Supplementary material online, Methods. Receiver-operating curve (ROC), Kaplan–Meier, and Cox proportional hazards methods were used to identify potential clinical predictors of all-cause death/heart failure events and MACE, including patient characteristics, CMR findings, and native T1. The net reclassification improvement (NRI) was calculated as described by Pencina et al.38 All P-values are two sided. P-value of<0.05 should be interpreted exploratively. Statistical analyses were performed using R version 2.15.1 or SAS v 9.3, or higher versions of these programs.

Receiver-operating curve (ROC), Kaplan–Meier, and Cox proportional hazards methods were used to identify potential clinical predictors of all-cause death/heart failure events and MACE, including patient characteristics, CMR findings, and native T1. The net reclassification improvement (NRI) was calculated as described by Pencina et al.38 All P-values are two sided. P-value of<0.05 should be interpreted exploratively. Statistical analyses were performed using R version 2.15.1 or SAS v 9.3, or higher versions of these programs. Results Of 343 STEMI patients referred for emergency reperfusion therapy, 300 underwent serial CMR at 1.5 T 2.2 ± 1.9 days and 6 months after hospital admission (Figure 2). Two hundred and ninety-two STEMI patients had a T1-map acquisition and 288 (99%) had evaluable T1 data (Figure 2). Cardiac magnetic resonance follow-up at 6 months was achieved in 267 (93%) of the patients and the reasons for non-attendance are summarized in Figure 2. Information on vital status and SAEs were available in all (100%) of the 288 participants. Patient characteristics Table 1 shows the characteristics of the patients, including the patients with a hypo-intense infarct core revealed by native T1 mapping [n = 160 (56%), grouped by thirds of native T1]. The C-reactive protein and leucocyte results are described in Supplementary material online, Table S1. The characteristics of those patients with missing CMR data at 6 months are described in Supplementary material online, Table S2.

rct core revealed by native T1 mapping [n = 160 (56%), grouped by thirds of native T1]. The C-reactive protein and leucocyte results are described in Supplementary material online, Table S1. The characteristics of those patients with missing CMR data at 6 months are described in Supplementary material online, Table S2. Left ventricular function and pathology Initial cardiac magnetic resonance findings following hospital admission The CMR findings are summarized in Table 2 and case examples are shown in Figure 2. At baseline, the mean (SD) myocardial infarct size was 18 (14) % of LV mass. The average infarct core native T1 (997 (57)) was higher than native T1 in the remote myocardium 961 (25) ms; P < 0.01] but lower than native T1 in the area at risk (1097 (52) ms; P < 0.01). The ICC for T1 core is described in Supplementary material online, Results. Baseline associates of infarct core native T1 (Hypothesis 1) Native T1 in the infarct core was inversely associated with thrombolysis in myocardial infarction (TIMI) coronary flow grades at the end of emergency PCI, Killip class and neutrophil count at initial presentation (all P < 0.04), independent of left ventricular ejection fraction (LVEF), LV end-diastolic volume, or infarct size (Table 3). Table 3 Associates of infarct core native T1 time (for a 10 ms difference) in 160 ST-elevation myocardial infarction survivors with infarct core pathology revealed by native T1 mapping with cardiac magnetic resonance 2 days post-myocardial infarction

fraction (LVEF), LV end-diastolic volume, or infarct size (Table 3). Table 3 Associates of infarct core native T1 time (for a 10 ms difference) in 160 ST-elevation myocardial infarction survivors with infarct core pathology revealed by native T1 mapping with cardiac magnetic resonance 2 days post-myocardial infarction Multiple stepwise regression (for a 10 ms difference in infarct core T1) Coefficient (95% CI) P-value A. Including patient characteristics and angiographic data* Systolic blood pressure at initial angiography (mmHg) −0.05 (−0.09, −0.01) 0.007 Killip Class 3 or 4 −3.84 (−6.87, −0.80) 0.014 TIMI flow Grade 2 or 3 post-PCI −7.51 (−15.42, 0.40) 0.063 B. Including patient characteristics, angiographic data, and minimum neutrophil count* Systolic blood pressure at initial angiography (mmHg) −0.05 (−0.09, −0.01) 0.015 Killip Class 3 or 4 −3.39 (−6.45, −0.33) 0.030 TIMI flow Grade 2 or 3 at the end of PCI −9.77 (−17.67, −1.87) 0.005 Minimum neutrophil count (×109 L) −0.50 (−0.86, −0.15) 0.005 C. Including patient characteristics, angiographic data, minimum neutrophil count, and T2 core (1 ms)a T2 core (1 ms) 0.50 (0.32, 0.67) <0.001 Neutrophils −0.39 (−0.71, 0.07) 0.016 Gender (male) −2.32 (−4.25, 0.39) 0.019 SBP −0.03 (−0.07, 0.00) 0.059 TIMI 2/3 post-PCI −5.46 (−12.62, 1.70) 0.134 The coefficient (95% CIs) indicates the magnitude and direction of the effect of the patient characteristic (binary or continuous) on the infarct core T1 (ms). For example, in models A and B, on average, infarct core native T1 (10 ms difference) is 0.50 lower for each 1 mmHg increase in SBP.

−5.46 (−12.62, 1.70) 0.134 The coefficient (95% CIs) indicates the magnitude and direction of the effect of the patient characteristic (binary or continuous) on the infarct core T1 (ms). For example, in models A and B, on average, infarct core native T1 (10 ms difference) is 0.50 lower for each 1 mmHg increase in SBP. aThe clinical and angiographic characteristics that were assessed are listed in Table 1. The univariable associates with native T1 in the infarct core are described in Supplementary material online, Results. Separate multivariable analyses were performed for (A) patient characteristics and angiographic data and (B) CMR data. Cardiac magnetic resonance parameters, which were all highly correlated with one another, were included separately in multiple stepwise regression models with patient characteristics and angiographic data to reduce multi-collinearity. Similar results were obtained when area at risk, LV ejection fraction, LV end-systolic volume, and infarct size were included. Maximum leucocyte count (P = 0.053) and maximum monocyte count (P = 0.034) remained associates of infarct core native T1 after adjustment for LV end-diastolic volume. Similar results were also obtained in the multivariable model with LV end-diastolic volume for minimum leucocyte count (P = 0.011). Infarct core native T1 (ms) was univariably associated with infarct core T2 (ms) (r = 0.42; P < 0.001).

Similar results were obtained when area at risk, LV ejection fraction, LV end-systolic volume, and infarct size were included. Maximum leucocyte count (P = 0.053) and maximum monocyte count (P = 0.034) remained associates of infarct core native T1 after adjustment for LV end-diastolic volume. Similar results were also obtained in the multivariable model with LV end-diastolic volume for minimum leucocyte count (P = 0.011). Infarct core native T1 (ms) was univariably associated with infarct core T2 (ms) (r = 0.42; P < 0.001). Relationships for native T1 infarct core vs. infarct pathology, including infarct core T2, myocardial haemorrhage, and microvascular obstruction and 137 (86%) STEMI patients with a hypo-intense native T1 infarct core also had microvascular obstruction. In contrast, only 6.3% of those without hypo-intense infarct core had late microvascular obstruction. The negative- and positive predictive values of native T1 infarct core for T2 infarct core, myocardial haemorrhage, EGE, and microvascular obstruction are summarized in Supplementary material online, Table S2. Table 1 Clinical and angiographic characteristics of 288 ST-elevation myocardial infarction patients who had cardiac magnetic resonance with evaluable maps for myocardial native T1 magnetization, including the subset of patients with an infarct core revealed by native T1 (all and categorized by tertiles of native T1)

. Table 1 Clinical and angiographic characteristics of 288 ST-elevation myocardial infarction patients who had cardiac magnetic resonance with evaluable maps for myocardial native T1 magnetization, including the subset of patients with an infarct core revealed by native T1 (all and categorized by tertiles of native T1) Characteristicsa All patients (n = 288) Patients with a native T1 infarct core (n = 160) (56%) Patients with a native T1 infarct core grouped by tertile of infarct core zone native T1 (ms) at baseline P-value T1 core ≤973 ms (n = 54) (33%) 974 <T1 core ≤1010 ms (n = 53) (33%) T1 core >1010 ms (n = 53) (33%) Age (years) 59 (11) 59 (11) 59 (11) 57 (11) 61 (11) 0.238 Male sex, n (%) 211 (73) 123 (77) 46 (85) 37 (70) 40 (76) 0.144 BMI (kg/m2) 29 (5) 29 (5) 29 (4) 29 (5) 28 (5) 0.674 Medical history Hypertension, n (%) 93 (32) 57 (36) 17 (32) 21 (40) 19 (36) 0.684 Current smoking, n (%) 177 (62) 100 (62) 32 (59) 34 (64) 34 (64) 0.858 Hypercholesterolaemia, n (%) 82 (28) 44 (28) 12 (22) 17 (32) 15 (28) 0.527 Diabetes mellitus,b n (%) 32 (11) 20 (12) 7 (13) 7 (13) 6 (11) 1.000 Previous angina, n (%) 34 (12) 21 (13) 8 (15) 4 (8) 9 (17) 0.304 Previous myocardial infarction, n (%) 23 (8) 15 (9) 5 (9) 3 (6) 7 (13) 0.415 Previous PCI, n (%) 16 (6) 14 (9) 4 (7) 3 (6) 7 (13) 0.414 Presenting characteristics Heart rate (bpm) 78 (17) 78 (16) 80 (16) 79 (16) 76 (17) 0.401 Systolic blood pressure (mmHg) 136 (24) 136 (22) 137 (24) 140 (23) 131 (19) 0.095 Diastolic blood pressure (mmHg) 79 (14) 80 (14) 82 (14) 83 (13) 76 (13) 0.010 Time from symptom onset to reperfusion (min) 174 (120, 311)a 188 (125, 388) 223 (145, 406) 163 (113, 313) 198 (128, 257) 0.268 Ventricular fibrillationc, n (%) 20 (7) 150 (94) 3 (6) 2 (4) 5 (9) 0.518 Heart failure, Killip class at presentation, n (%) I 205 (71%) 101 (63) 29 (54%) 38 (72%) 34 (64%) II 64 (22%) 43 (27) 15 (28%) 14 (26%) 14 (26%) 0.059 III/IV 19 (7) 16 (10) 10 (18) 1 (2) 5 (9) ECG ST segment elevation resolution post-PCI, n (%) Complete, ≥70% 129 (45) 55 (35) 15 (28) 21 (40) 19 (36) Incomplete, 30% to <70% 115 (40) 74 (46) 27 (50) 23 (44) 24 (45) 0.715 None, ≤30% 43 (15) 30 (19) 12 (22) 8 (15) 10 (19) Reperfusion strategy, n (%) Primary PCI 268 (93) 148 (92) 49 (91) 49 (92) 50 (94) Rescue PCI (failed thrombolysis) 13 (4) 10 (6) 4 (7) 3 (6) 3 (6) 1.000 Successful thrombolysis 7 (2) 2 (1) 1 (2) 1 (2) 0 (0) Coronary angiography Number of diseased arteries,d n (%) 1 156 (54) 89 (56) 156 (54) 156 (54) 156 (54) 2 89 (29) 44 (28) 90 (31) 90 (31) 9

%) Primary PCI 268 (93) 148 (92) 49 (91) 49 (92) 50 (94) Rescue PCI (failed thrombolysis) 13 (4) 10 (6) 4 (7) 3 (6) 3 (6) 1.000 Successful thrombolysis 7 (2) 2 (1) 1 (2) 1 (2) 0 (0) Coronary angiography Number of diseased arteries,d n (%) 1 156 (54) 89 (56) 156 (54) 156 (54) 156 (54) 2 89 (29) 44 (28) 90 (31) 90 (31) 9 0 (31) 0.436 3 42 (15) 24 (15) 42 (15) 42 (15) 42 (15) LM 6 (2) 3 (2) 0 (0) 2 (4) 1 (2) Culprit artery, n (%) LAD 108 (38) 60 (38) 22 (41) 19 (36) 19 (36) LCX 51 (18) 31 (19) 10 (18) 12 (23) 9 (17) 0.915 RCA 129 (45) 69 (34) 22 (41) 22 (42) 25 (47) TIMI coronary flow grade pre-PCI, n (%) 0/1 208 (72) 135 (84) 49 (91) 39 (74) 47 (89) 2 52 (18) 27 (13) 5 (9) 11 (21) 5 (9) 0.085 3 28 (10) 4 (2) 0 (0) 3 (6) 1 (2) TIMI coronary flow grade post-PCI, n (%) 0/1 3 (1) 2 (1) 0 (0) 1 (2) 1 (2) 2 13 (4) 8 (5) 3 (6) 3 (6) 2 (4) 0.959 3 272 (94) 150 (94) 51 (94) 49 (92) 50 (94) Medical therapy ACE-I or ARB 285 (99) 159 (>99) 54 (100) 53 (100) 52 (98) 0.663 β-Blocker 278 (96) 158 (99) 53 (98) 52 (98) 53 (100) 1.000 Initial blood results on admission C-reactive protein, (mg/L), median (IQR), range 3.0 (2.0–7.0) 0–265.0 4.0 (2.0, 8.0) 1.0–265 3.5 (2.0–11.0) 1.0–125.0 3.0 (1.0–6.2) 1.0–92.0 4.0 (2.0–7.0) 1.0–265.0 0.696 Leucocyte cell count (×109 L) 12.4 (3.5) 12.8 (3.6) 12.9 (3.5) 13.3 (3.5) 12.3 (3.6) 0.310 Neutrophil count (×109 L) 9.6 (3.2) 10.1 (3.3) 10.0 (3.4) 10.6 (3.4) 9.6 (3.0) 0.244 Monocytes (×109 L) 0.4 (0.4) 0.9 (0.4) 1.0 (0.4) 0.9 (0.3) 0.9 (0.5) 0.485 NT-proBNP (pg/mL) 824 (350, 1642) 1103 (628, 1849) 1456 (702, 2455) 980 (565, 1637) 1021 (529, 1436) 0.354 Missing data: heart rate, n = 1; time from symptom onset to reperfusion, n = 20; ST-segment resolution, n = 1; CRP, n = 7; leucocyte count, n = 1.

s (×109 L) 0.4 (0.4) 0.9 (0.4) 1.0 (0.4) 0.9 (0.3) 0.9 (0.5) 0.485 NT-proBNP (pg/mL) 824 (350, 1642) 1103 (628, 1849) 1456 (702, 2455) 980 (565, 1637) 1021 (529, 1436) 0.354 Missing data: heart rate, n = 1; time from symptom onset to reperfusion, n = 20; ST-segment resolution, n = 1; CRP, n = 7; leucocyte count, n = 1. The patients are grouped according to tertile of T1 in hypo-intense core at baseline. ACE-I or ARB, angiotensin converting enzyme inhibitor or angiotensin receptor blocker; LAD, left anterior descending coronary artery; LCX, left circumflex coronary artery; LM, left main coronary artery; RCA, right coronary artery; TIMI, thrombolysis in myocardial infarction grade; PCI, percutaneous coronary intervention. Killip classification of heart failure after acute myocardial infarction: class I—no heart failure, class II—pulmonary rales or crepitations, a third heart sound, and elevated jugular venous pressure, class III—acute pulmonary oedema, and class IV—cardiogenic shock. aData are reported as mean (SD), median (IQR), or N (%) as appropriate. P-values have been obtained from a one-way ANOVA or Fisher test. Thrombolysis in myocardial infarction flow grades pre- and post-PCI were grouped 0/1 vs. 2/3 for this analysis. bSuccessfully electrically cardioverted ventricular fibrillation at presentation or during emergency PCI procedure. cDiabetes mellitus was defined as a history of diet-controlled or treated diabetes.

aData are reported as mean (SD), median (IQR), or N (%) as appropriate. P-values have been obtained from a one-way ANOVA or Fisher test. Thrombolysis in myocardial infarction flow grades pre- and post-PCI were grouped 0/1 vs. 2/3 for this analysis. bSuccessfully electrically cardioverted ventricular fibrillation at presentation or during emergency PCI procedure. cDiabetes mellitus was defined as a history of diet-controlled or treated diabetes. dMulti-vessel coronary artery disease was defined according to the number of stenoses of at least 50% of the reference vessel diameter, by visual assessment and whether or not there was left main stem involvement. The blood results on admission and their changes during the first 2 days after admission are described in Supplementary material online, Table S1. Table 2 Comparison of cardiac magnetic resonance findings at baseline in 288 ST-elevation myocardial infarction survivors and 6-month cardiac magnetic resonance findings in 278 ST-elevation myocardial infarction patients

dMulti-vessel coronary artery disease was defined according to the number of stenoses of at least 50% of the reference vessel diameter, by visual assessment and whether or not there was left main stem involvement. The blood results on admission and their changes during the first 2 days after admission are described in Supplementary material online, Table S1. Table 2 Comparison of cardiac magnetic resonance findings at baseline in 288 ST-elevation myocardial infarction survivors and 6-month cardiac magnetic resonance findings in 278 ST-elevation myocardial infarction patients Characteristics* All patients Patients with a native T1 infarct core Patients with a native T1 infarct core grouped by tertile of infarct core zone native T1 (ms) at baseline P-value All patients (n = 288) Hypo-intense core (n = 160) ≤973 ms (n = 54) 974 <T1 core ≤1014 ms (n = 53) >1014 ms (n = 53) CMR findings 2 days post-MI LV ejection fraction (%) 55 (10) 52 (9) 52 (10) 51 (8) 53 (10) 0.418 LV end-diastolic volume, ml Men 162 (33) 168 (147, 187) 168 (22) 169 (36) 166 (30) 0.900 Women 124 (25) 125 (113, 145) 122 (30) 134 (26) 126 (21) 0.497 LV end-systolic volume (mL) Men 73 (55, 94) 79 (64, 98) 75 (64, 94) 81 (74, 103) 76 (60, 100) 0.496 Women 53 (41, 66) 64 (50, 69) 64 (57, 71) 66 (57, 70) 56 (45, 65) 0.383 LV mass (g) Men 142 (124, 159) 145 (130, 166) 149 (135, 170) 143 (126, 159) 141 (130, 160) 0.526 Women 97 (84, 108) 101 (89, 124) 103 (92, 113) 109 (93, 132) 97 (83, 101) 0.113 Oedema and infarct characteristics Area at risk, % LV mass 32 (12) 40 (11) 37 (11) 35 (10) 36 (11) 0.482 Infarct size, % LV mass 16 (7, 27) 25 (16, 32) 25 (18, 34) 27 (18, 32) 22 (16, 32) 0.386 Myocardial salvage, % of LV mass 18 (12, 24) 17 (12, 23) 18 (12, 24) 17 (10, 22) 16 (13, 22) 0.546 Myocardial salvage index, % of LV mass 62 (44, 84) 49 (36, 62) 50 (40, 62) 46 (30, 62) 50 (40, 63) 0.590 Late microvascular obstruction present, n (%) 145 (50) 23 (14) 49 (91) 45 (85) 43 (81) 0.356 Late microvascular obstruction, % LV mass 0.1 (0.0, 3.5) 2.7 (0.8, 7.5) 5.2 (1.7, 10.5) 2.7 (0.9, 7.1) 1.7 (0.3, 4.7) 0.005 Myocardial haemorrhage, n (%)* 96 (40) 94 (67) 34 (76) 35 (70) 25 (54) 0.086 Myocardial native T1 values T1 remote myocardium (all subjects) (ms) 961 (25) 964 (26) 958 (28) 962 (20) 972 (28) 0.014 Men 959 (25) 962 (26) 955 (29) 959 (19) 973 (26) 0.004 Women 968 (25) 969 (26) 969 (22) 969 (22) 968 (36) 0.992 T1 infarct zone (ms) 1097 (52) 1093 (52) 1052 (37) 1088 (33) 1140 (22) <0.001 T1 hypo-intense infarct core (ms) 997 (57) 997 (57) 938 (30) 995 (12) 1060 (37) <0.001 Myocardial native T2 values T2 infarct core (n = 171, ms) 54 (5) 54 (5) 52 (4) 53 (4) 56 (5) <0.001 CMR findings 6 months post-MI (n = 267) LV ejection fraction at 6 months (%) 63 (57, 69) 60 (53, 65) 59 (53, 65)

3) 1140 (22) <0.001 T1 hypo-intense infarct core (ms) 997 (57) 997 (57) 938 (30) 995 (12) 1060 (37) <0.001 Myocardial native T2 values T2 infarct core (n = 171, ms) 54 (5) 54 (5) 52 (4) 53 (4) 56 (5) <0.001 CMR findings 6 months post-MI (n = 267) LV ejection fraction at 6 months (%) 63 (57, 69) 60 (53, 65) 59 (53, 65) 59 (54, 64) 61 (54, 68) 0.542 LV end-diastolic volume at 6 months (mL) Men 165 (140, 193) 176 (155, 204) 188 (160, 209) 169 (153, 197) 171 (156, 196) 0.367 Women 124 (110, 136) 120 (96, 139) 120 (96, 139) 130 (122, 153) 127 (118, 142) 0.338 LV end-systolic volume at 6 months (mL) Men 61 (43, 78) 69 (56, 95) 73 (58, 98) 69 (62, 84) 63 (53, 96) 0.667 Women 43 (34, 58) 55 (44, 61) 45 (41, 56) 60 (50, 65) 53 (40, 57) 0.213 LV, left ventricle; T1, myocardial longitudinal relaxation time. Area at risk was measured with T2-mapping. Data are given as n (%) or mean (SD). P-values were obtained from one-way ANOVA, Kruskal–Wallis test, or a Fisher test. *Data are reported as mean (SD), median (IQR), or n (%) as appropriate. Data on T2*-CMR for myocardial haemorrhage were not available in 48 patients.

59 (54, 64) 61 (54, 68) 0.542 LV end-diastolic volume at 6 months (mL) Men 165 (140, 193) 176 (155, 204) 188 (160, 209) 169 (153, 197) 171 (156, 196) 0.367 Women 124 (110, 136) 120 (96, 139) 120 (96, 139) 130 (122, 153) 127 (118, 142) 0.338 LV end-systolic volume at 6 months (mL) Men 61 (43, 78) 69 (56, 95) 73 (58, 98) 69 (62, 84) 63 (53, 96) 0.667 Women 43 (34, 58) 55 (44, 61) 45 (41, 56) 60 (50, 65) 53 (40, 57) 0.213 LV, left ventricle; T1, myocardial longitudinal relaxation time. Area at risk was measured with T2-mapping. Data are given as n (%) or mean (SD). P-values were obtained from one-way ANOVA, Kruskal–Wallis test, or a Fisher test. *Data are reported as mean (SD), median (IQR), or n (%) as appropriate. Data on T2*-CMR for myocardial haemorrhage were not available in 48 patients. Three T1 maps (basal-, mid-, and distal-ventricular levels) were measured in each patient (n = 876 T1-maps overall). Overall, 20 (6.8%) patients had poor quality T1 maps and 4 (1.3%) patients had no evaluable T1 maps (Figure 2). In all, 42 (4.8%) T1 maps were unsuitable for analysis because of SSFP off-resonance artefacts and 19 (2.2%) T1 maps were affected by motion artefacts. T1 values were higher in infarct tissue surrounding the infarct core than within the infarct core (P < 0.001) and remote myocardium (P < 0.001).

ble T1 maps (Figure 2). In all, 42 (4.8%) T1 maps were unsuitable for analysis because of SSFP off-resonance artefacts and 19 (2.2%) T1 maps were affected by motion artefacts. T1 values were higher in infarct tissue surrounding the infarct core than within the infarct core (P < 0.001) and remote myocardium (P < 0.001). Infarct core tissue characteristics as a marker of subsequent left ventricular remodelling (Hypothesis 2) At 6 months, LV end-diastolic volume increased on average (SD) by 5 (25) ml in 262 patients with evaluable data (Table 2). Adverse remodelling occurred in 30 (12%) patients and 23 (77%) of these patients had a hypo-intense native T1 core at baseline. Infarct core native T1 (ms) was not associated with change in LV end-diastolic volume at follow-up (P = 0.531). In multivariable regression, native T1 (ms, continuous) within the hypo-intense core was inversely associated with adverse remodelling (Table 4). Table 4 Multivariable associates of adverse LV remodelling revealed by cardiac magnetic resonance in ST-elevation myocardial infarction survivorsa after 6 months follow-up

). In multivariable regression, native T1 (ms, continuous) within the hypo-intense core was inversely associated with adverse remodelling (Table 4). Table 4 Multivariable associates of adverse LV remodelling revealed by cardiac magnetic resonance in ST-elevation myocardial infarction survivorsa after 6 months follow-up Multivariable associations Odds ratio (95% CI) P-value A. Patient and angiographic characteristics Native T1 infarct core, per 10 ms 0.91 (0.82, 1.00) 0.061 Current smoking 5.27 (1.07, 26.00) 0.041 Sustained ventricular arrhythmia 16.06 (1.67, 154.43) 0.016 Incomplete ST-segment resolution 3.29 (0.85, 12.78) 0.085 B. Patient and angiographic characteristics and infarct core native T2 Native T1 infarct core, per 10 ms 0.91 (0.81, 1.01) 0.073 Native T2 infarct core, per 10 ms 1.01 (0.28, 3.67) 0.987 Current smoking 4.99 (0.99, 25.06) 0.051 Sustained ventricular arrhythmia 15.26 (1.57, 148.71) 0.019 Incomplete ST-segment resolution 3.18 (0.81, 12.43) 0.097 C. Patient and angiographic characteristics and myocardial haemorrhage Native T1 infarct core, per 10 ms 0.90 (0.81, 1.01) 0.070 Myocardial haemorrhage 0.57 (0.14, 2.41) 0.449 Current smoking 4.78 (0.83, 27.52) 0.080 Sustained ventricular arrhythmia 11.70 (0.94, 144.88) 0.055 Incomplete ST-segment resolution 3.68 (0.90, 15.02) 0.069 The odds ratio (95% CIs) indicates the magnitude and direction for adverse LV remodelling. For a 10 ms increase in native T1, the odds ratio for adverse LV remodelling reduced (0.91 (0.82, 1.00); P = 0.061). For a 1 ms increase in native T1, the odds ratio for adverse LV remodelling reduced [0.99 (0.98, 1.00); P = 0.061].

odds ratio (95% CIs) indicates the magnitude and direction for adverse LV remodelling. For a 10 ms increase in native T1, the odds ratio for adverse LV remodelling reduced (0.91 (0.82, 1.00); P = 0.061). For a 1 ms increase in native T1, the odds ratio for adverse LV remodelling reduced [0.99 (0.98, 1.00); P = 0.061]. aTwenty clinical characteristics at baseline that were univariable associates of adverse LV remodelling at 6 months post-MI were included in the multivariable model and these univariable associates are described in Supplementary material online, Results. Two hundred and sixty-seven STEMI patients had CMR at 6 months and baseline and 23 of these patients had missing data of at least one of the univariable characteristics that were included in this multivariable model. C-statistic [area-under-the-curve (AUC)] for the multivariable model in 244 subjects but not including native T1 core: 0.95; C-statistic (AUC) for the model (above) including infarct core native T1 (n = 136): 0.81; net reclassification index for incremental addition of T1 core to the model: 0.31, P = 0.184. When the multivariable model for adverse remodelling included infarct size, the AUC without native T1 core (continuous, ms) was 0.823 and the AUC with T1 core values included was 0.857. Inclusion of native T1 core values neither increased nor reduced the predictive value of this model (net reclassification index P = 0.16). There was no threshold for native T1 core value in the infarct core in relation to its association with LV outcomes at baseline or during follow-up.

When the multivariable model for adverse remodelling included infarct size, the AUC without native T1 core (continuous, ms) was 0.823 and the AUC with T1 core values included was 0.857. Inclusion of native T1 core values neither increased nor reduced the predictive value of this model (net reclassification index P = 0.16). There was no threshold for native T1 core value in the infarct core in relation to its association with LV outcomes at baseline or during follow-up. In a sensitivity analysis, the occurrence of a hypo-intense core within the infarct zone on T1 mapping was associated with the odds ratio for being in the top quarter of an increase in LV end-diastolic volume at 6 months (native T1 core to predict Q4 (n = 66) vs. Q1–Q3 (n = 196) (n = 26 missing); odds ratio 0.994 (0.987, 0.999); P = 0.048).

sis, the occurrence of a hypo-intense core within the infarct zone on T1 mapping was associated with the odds ratio for being in the top quarter of an increase in LV end-diastolic volume at 6 months (native T1 core to predict Q4 (n = 66) vs. Q1–Q3 (n = 196) (n = 26 missing); odds ratio 0.994 (0.987, 0.999); P = 0.048). Native T1 infarct core, microvascular obstruction, and left ventricular outcomes at 6 months The relationships for infarct core native T1 (binary and continuous), T2 core (binary and continuous), microvascular obstruction (binary, % LV mass), and myocardial haemorrhage for LV outcomes, including LV end-diastolic volume and LV ejection fraction, are shown in Table 5. The presence of a hypo-intense infarct core disclosed by native T1 and native T2, the presence and amount of microvascular obstruction, and the occurrence of myocardial haemorrhage, were consistently associated with LV outcomes. Native T1 (ms) was not associated with LV volumes at follow-up, and there was no evidence of non-linearity between infarct core T1 (ms) and LV outcomes. Table 5 The univariable relationships for infarct core characteristics revealed by native T1 and microvascular obstruction for LV outcomes at baseline and during follow-up in 288 ST-elevation myocardial infarction patients

llow-up, and there was no evidence of non-linearity between infarct core T1 (ms) and LV outcomes. Table 5 The univariable relationships for infarct core characteristics revealed by native T1 and microvascular obstruction for LV outcomes at baseline and during follow-up in 288 ST-elevation myocardial infarction patients LVEDV at baseline LVEDV at 6 months LVEF at baseline LVEF at follow-up T1 core (per 10 ms) Standardized β −0.042 −0.035 0.151 0.055 P-value 0.596 0.520 0.057 0.485 T1 core (binary) β 16.410 13.80 −6.642 −4.652 P-value <0.0001 <0.0001 <0.0001 <0.0001 T2 core (per 10 ms) Standardized β 0.035 0.057 0.159 −0.033 P-value 0.653 0.282 0.037 0.586 T2 core (binary) β 15.538 12.875 −6.542 −4.494 P-value <0.001 <0.0001 <0.0001 <0.0001 Myocardial haemorrhage (T2* core, binary) β 17.205 16.811 −6.374 −5.769 P-value <0.0001 <0.0001 <0.0001 <0.0001 Microvascular obstruction (% of LV mass) Standardized β 0.186 0.209 −0.443 −0.283 P-value 0.002 <0.0001 <0.0001 0.004 Microvascular obstruction (binary) β 15.853 12.454 −6.620 −4.464 P-value <0.001 <0.0001 <0.0001 <0.0001 The relationships for infarct core native T1 relaxation time (per 10 ms), native T1 infarct core (binary), and the presence and the amount of microvascular obstruction (n = 145 STEMI patients) with LV outcomes are summarized by P-values and, for continuous predictors, standardized regression coefficients or odds ratios per SD increase in native T1 (ms) or extent of microvascular obstruction (% of LV mass). Models with follow-up data are adjusted for baseline. Binary predictors are summarized by P-values and unstandardized regression coefficients or odds ratios (listed beneath table).

tandardized regression coefficients or odds ratios per SD increase in native T1 (ms) or extent of microvascular obstruction (% of LV mass). Models with follow-up data are adjusted for baseline. Binary predictors are summarized by P-values and unstandardized regression coefficients or odds ratios (listed beneath table). The odds ratio (P-values) for adverse remodelling and infarct core characteristics are: native T1 core (per 10 ms): 0.939, P = 0.122; native T1 core (present/absent): 2.692, P = 0.016; T2 core (present/absent): 2.874, P = 0.026; myocardial haemorrhage (present/absent): 2.556, P = 0.025; microvascular obstruction (% LV mass): 1.112, P = 0.004; microvascular obstruction (present/absent): 1.883, P = 0.115. LVEDV, left ventricular end-diastolic volume; LVEF, left ventricular ejection fraction. Infarct core native T1 early post-MI was associated with the concentration of NT-proBNP, a biochemical measure of LV remodelling, at 6 months independent of LV end-diastolic volume at baseline (Supplementary material online, Results). Infarct core tissue characteristics and health outcomes (Hypothesis 3) All 288 patients had long-term follow-up data completed. Thirty (10.4%) patients died or experienced a heart failure event. These events included 5 cardiovascular deaths, 3 non-cardiovascular deaths, and 22 episodes of heart failure (Killip Class 3 or 4 heart failure (n = 20) or defibrillator implantation n = 2). Thirteen (4.5%) patients died or experienced a first heart failure hospitalization post-discharge, and 8 (61.5%) of these patients had a hypo-intense infarct core at baseline.

3 non-cardiovascular deaths, and 22 episodes of heart failure (Killip Class 3 or 4 heart failure (n = 20) or defibrillator implantation n = 2). Thirteen (4.5%) patients died or experienced a first heart failure hospitalization post-discharge, and 8 (61.5%) of these patients had a hypo-intense infarct core at baseline. Native T1 values (ms) within the hypo-intense infarct core (n = 160 STEMI patients) were inversely associated with the risk of all-cause death or first hospitalization for heart failure post-discharge (for a 10 ms increase in native T1: hazard ratio 0.730, 95% confidence interval (CI) 0.617, 0.863; P < 0.001) including after adjustment for LVEF at baseline, infarct core T2 (10 ms difference), and myocardial haemorrhage (Figure 3; Table 6). Infarct core T1 retained its prognostic significance over and above infarct core T2 and myocardial haemorrhage (Table 6, models C–F). The net reclassification index for the inclusion of infarct core native T1 (10 ms) in a multivariable prognostic model for all-cause death or heart failure post-discharge was 1.129 (95% CI 0.516, 1.742); P < 0.001) (Table 6). Using ROC analysis, the C-index for infarct core native T1 for all-cause death or heart failure was 0.806. The C-indexes for the prognostic model without and with infarct core native T1 (ms) were 0.715 and 0.931, respectively. Table 6 Relationships for infarct core T1 and T2 relaxation times (10 ms) revealed by cardiac magnetic resonance at baseline in 160 ST-elevation myocardial infarction patients with an infarct core and all-cause death or first hospitalization for heart failure post-discharge [median (range) follow-up duration of 841 (723–945) days]

tionships for infarct core T1 and T2 relaxation times (10 ms) revealed by cardiac magnetic resonance at baseline in 160 ST-elevation myocardial infarction patients with an infarct core and all-cause death or first hospitalization for heart failure post-discharge [median (range) follow-up duration of 841 (723–945) days] Associations Hazard ratio (95% CI) P-value Univariable associations Infarct core native T1 (for a 10 ms difference) 0.730 (0.617, 0.863) <0.001 Myocardial haemorrhage 2.488 (0.814, 7.609) 0.110 LVEF at baseline (for a 1% difference) 0.934 (0.885, 0.985) 0.013 Peak log eosinophil count (×109/L) 0.617 (0.432, 0.881) 0.008 Model A Infarct core native T1 (for a 10 ms difference) 0.744 (0.627, 0.883) <0.001 LVEF at baseline (1% difference) 0.938 (0.883, 0.996) 0.036 Model B Infarct core native T1 (for 10 ms difference) 0.737 (0.621, 0.875) <0.001 Peak log eosinophil count (1 ×109/L) 0.728 (0.476, 1.114) 0.144 Univariable associations Infarct core native T2 (for a 10 ms difference) 0.186 (0.032, 1.094) 0.063 Model C Infarct core native T2 (for a 10 ms difference) 0.244 (0.039, 1.528) 0.132 LVEF at baseline (for a 1% difference) 0.932 (0.870, 0.998) 0.044 Model D Infarct core native T2 (for 10 ms difference) 0.203 (0.034, 1.297) 0.093 Peak log eosinophil count (×109/L) 0.681 (0.460, 1.007) 0.054 Model E Infarct core T1 (for 10 ms difference) 0.738 (0.624, 0.873) <0.001 Myocardial haemorrhage 1.965 (0.229, 16.864) 0.538 Model F Infarct core T1 (for 10 ms difference) 0.752 (0.634, 0.893) 0.001 Infarct core T2 (for a 10 ms difference) 0.428 (0.068, 2.683) 0.365 Myocardial haemorrhage 1.485 (0.159, 13.879) 0.729 Thirteen (8.1%) patients experienced all-cause death or heart failure hospitalization post-discharge. Given the limited number of adverse events, the models were specified to assess the prognostic relationships of infarct core native T1 vs. circulating markers of systemic inflammation, LV function, LV volume, and infarct characteristics that were measured at approximately the same time 2 days after hospital admission.

Given the limited number of adverse events, the models were specified to assess the prognostic relationships of infarct core native T1 vs. circulating markers of systemic inflammation, LV function, LV volume, and infarct characteristics that were measured at approximately the same time 2 days after hospital admission. LVEF, left ventricular ejection fraction. Infarct core T1 (10 ms difference) is highlighted in bold. Figure 3 Kaplan–Meier survival curves for 160 ST-elevation myocardial infarction patients grouped according to the native T1 value in the infarct core with patients grouped by thirds (lowest T1 tertile vs. tertiles 2 and 3) and all-cause death or first heart failure hospitalization (n = 13) after discharge from hospital to the end of follow-up [censor time 839 (598–1099) days]. Infarct core native T1 values in the lowest tertile were associated with all-cause death or heart failure hospitalization.

(lowest T1 tertile vs. tertiles 2 and 3) and all-cause death or first heart failure hospitalization (n = 13) after discharge from hospital to the end of follow-up [censor time 839 (598–1099) days]. Infarct core native T1 values in the lowest tertile were associated with all-cause death or heart failure hospitalization. Prognostic importance of infarct core native T1: comparisons with microvascular obstruction and longer term health outcomes In univariate Cox models for infarct core native T1 (ms), native T1 core (binary), T2 core (ms), T2 core (binary), myocardial haemorrhage and the presence (binary) and amount of microvascular obstruction (% LV mass), only infarct core native T1 (ms) (P < 0.001) and the amount of microvascular obstruction (% LV mass) (P < 0.001) were associated with all-cause death or first heart failure hospitalisation after discharge. In a post-hoc analysis stimulated by peer review, the odds ratio for infarct core T1 (10 ms) at baseline for the occurrence of all-cause death, heart failure hospitalisation or adverse LV remodelling was 0.92 (95% CI 0.85, 0.99), P = 0.0312. The inverse relationships between infarct core native T1 (ms) and LV surrogate and adverse health outcomes were reasonably linear, and there was no cut-off value for infarct core T1 (ms) for these outcomes.

ll-cause death, heart failure hospitalisation or adverse LV remodelling was 0.92 (95% CI 0.85, 0.99), P = 0.0312. The inverse relationships between infarct core native T1 (ms) and LV surrogate and adverse health outcomes were reasonably linear, and there was no cut-off value for infarct core T1 (ms) for these outcomes. Discussion The main findings of our study are (i) native T1 mapping revealed without an intravenous contrast agent resulted in evaluable scans in 96% of STEMI survivors 2 days post-MI; (ii) acute culprit coronary artery blood flow and circulating measures of systemic inflammation at the time of the hospital admission were multivariable associates of native T1 within the hypo-intense infarct core revealed by T1 mapping 2 days later; (iii) native T1 values (ms) within the infarct core were clinically meaningful since they were independently associated with adverse remodelling, NT-proBNP concentrations at 6 months, and all-cause death or heart failure hospitalization post-discharge during longer term follow-up; (iv) compared with infarct core T2 or myocardial haemorrhage revealed by T2* mapping, infarct core T1 was more consistently associated with LV surrogate outcomes and all-cause death or heart failure hospitalization (Table 6), implying T1 core is more closely linked with infarct pathology; (v) compared with microvascular obstruction revealed by contrast-enhanced CMR, a hypo-intense infarct core revealed by T1 mapping had similar prognostic significance for LV outcomes at 6 months and for post-discharge cardiac events including all-cause mortality and heart failure hospitalization in the longer term (Tables 5 and 6). Finally, our paper adds to the emerging literature on the prognostic value of quantitative native T1 CMR39 and reaffirms the prognostic importance of MVO post-STEMI.3

es at 6 months and for post-discharge cardiac events including all-cause mortality and heart failure hospitalization in the longer term (Tables 5 and 6). Finally, our paper adds to the emerging literature on the prognostic value of quantitative native T1 CMR39 and reaffirms the prognostic importance of MVO post-STEMI.3 The results of this study extend what is known about infarct core pathology, and also provide a potential mechanistic explanation. Infarct size1,2 and pathology, including microvascular obstruction,3 haemorrhage,5 and salvage,4 predict cardiac morbidity and mortality post-MI. These pathologies are revealed by contrast-enhanced CMR, and until recently, the assessment of infarct tissue without an intravenous contrast agent has been limited to T2-weighted and T2* imaging of myocardial haemorrhage.5,11,29,40,41 T1-mapping methods, including MOLLI15,20 and shMOLLI,42,43 can now be integrated into clinical CMR protocols. Previous studies have assessed myocardial native T1 in experimental MI models ex vivo,8,10 in vivo,11 or in proof-of-concept clinical studies involving much smaller numbers of MI patients.9,12–15 Our study extends these findings in a much larger STEMI cohort and provides new evidence that native T1 core is more reflective of the severity of infarct injury and its prognostic importance than infarct core T2 and potentially also myocardial haemorrhage.

studies involving much smaller numbers of MI patients.9,12–15 Our study extends these findings in a much larger STEMI cohort and provides new evidence that native T1 core is more reflective of the severity of infarct injury and its prognostic importance than infarct core T2 and potentially also myocardial haemorrhage. We also compared infarct core pathology delineated by native T1 mapping with microvascular obstruction, which is an established prognostic CMR biomarker post-MI.3 Native T1 mapping is obtained without the use of an intravenous gadolinium-based contrast agent whereas microvascular obstruction is revealed by CMR imaging of EGE and LGE after intravenous contrast administration. We observed a high degree of concordance between the occurrence of a hypo-intense infarct core depicted by native T1 CMR (56%) and microvascular obstruction (50%) as revealed by contrast-enhanced CMR. Although both a native T1 core and microvascular obstruction are depicted as a hypo-intense core within the hyperintense infarct zone (Figure 2), the physics of the CMR techniques is entirely different. A hypo-intense infarct core depicted by non-contrast native T1 mapping is due to local destruction of the T1 magnetization signal. On the other hand, microvascular obstruction (Figure 2) is due to a failure of gadolinium contrast to penetrate within the infarct core. Both CMR methods are T1-weighted but contrast kinetics are not relevant for native T1 mapping since intravenous contrast is not administered. Accordingly, T1 mapping avoids the theoretical risks and actual restrictions associated with contrast-enhanced CMR. Furthermore, acquisition of the native T1 map does not prolong the CMR scan, in contrast to late gadolinium enhancement imaging for microvascular obstruction which is typically imaged 10–15 min after dosing.19

dingly, T1 mapping avoids the theoretical risks and actual restrictions associated with contrast-enhanced CMR. Furthermore, acquisition of the native T1 map does not prolong the CMR scan, in contrast to late gadolinium enhancement imaging for microvascular obstruction which is typically imaged 10–15 min after dosing.19 Culprit artery coronary flow at the end of emergency PCI reflects the efficacy of coronary reperfusion, and reduced coronary flow initially independently predicted native T1 relaxation time within infarct core as assessed by CMR 2 days later. Similar associations also exist for microvascular obstruction,44,45 and in our study, both infarct core native T1 and microvascular obstruction were independently associated with circulating biomarkers of acute systemic inflammation. The occurrence of an infarct core disclosed by native T1 mapping, and the nature of the core (i.e. the native T1 value), was associated with the initial severity of MI (i.e. Killip heart failure class), systemic inflammation (i.e. leucocyte counts), and LV remodelling and health outcomes in the longer term. We think that the prognostic significance of native T1 values within the hypo-intense core are a distinctive attribute compared with microvascular obstruction since signal-intensity values within microvascular obstruction are not clinically meaningful beyond binary categorization (i.e. present/absent).

the longer term. We think that the prognostic significance of native T1 values within the hypo-intense core are a distinctive attribute compared with microvascular obstruction since signal-intensity values within microvascular obstruction are not clinically meaningful beyond binary categorization (i.e. present/absent). Limitations We performed a single centre natural-history study involving near-consecutive STEMI admissions. The STEMI patients in our natural-history study were recruited 24/7 therefore flow cytometry and routine NT-proBNP testing in all participants was not pragmatically possible. T1 assessment is sensitive to motion artefacts and imperfect breath holding, which and may reduce image quality. A shortened version of this sequence (ShMOLLI) involving only nine heart beats has been developed. This method shortens breath hold time and may help to account for these limitations.42 Despite this, the MOLLI method has high precision reproducibility. Our T1 measurements are in good agreement with in vivo data published in the literature, including previous measurements using the ShMOLLI sequence.42,43

n developed. This method shortens breath hold time and may help to account for these limitations.42 Despite this, the MOLLI method has high precision reproducibility. Our T1 measurements are in good agreement with in vivo data published in the literature, including previous measurements using the ShMOLLI sequence.42,43 The limited number of adverse events constrained the number of variables that could be included in the multivariable models (e.g. Tables 4 and 6); however, the associations between infarct core native T1 and a range of surrogate and clinical outcomes including adverse remodelling revealed by CMR, NT-proBNP, and the primary health outcome (all-cause death/heart failure), supports the adverse prognostic importance of infarct core native T1. Our study does not permit inference on causality, and other interpretations of our data are possible and further studies are warranted. Conclusions We found that infarct core pathology revealed by native T1 maps had similar prognostic value compared with microvascular obstruction revealed by late gadolinium enhancement CMR. Native T1 mapping is potentially widely applicable in clinical practice, not limited by renal disease, and so potentially could represent an alternative non-contrast CMR option for the assessment of infarct pathology. Supplementary material Supplementary material is available at European Heart Journal online.

Conclusions We found that infarct core pathology revealed by native T1 maps had similar prognostic value compared with microvascular obstruction revealed by late gadolinium enhancement CMR. Native T1 mapping is potentially widely applicable in clinical practice, not limited by renal disease, and so potentially could represent an alternative non-contrast CMR option for the assessment of infarct pathology. Supplementary material Supplementary material is available at European Heart Journal online. Authors’ contributions C.H., I.F.: performed statistical analysis. C.B., K.O., I.F.: handled funding and supervision. D.C., M.M., M.P., H.E., M.L., S.W., S.H., A.M.: acquired the data. C.B., I.F.: conceived and designed the research. D.C., C.B., C.H.: drafted the manuscript. D.C., C.B., C.H., N.S., P.W.: made critical revision of the manuscript for key intellectual content. S.R., I.M., N.A., N.T., A.R.: have contributed other than the above listed. Funding This research was supported by the British Heart Foundation Grant (Project Grant PG/11/2/28474), the National Health Service, and the Chief Scientist Office. Professor Berry was supported by a Senior Fellowship from the Scottish Funding Council. Dr Welsh is supported by BHF Fellowship FS/12/62/29889. Funding to pay the Open Access publication charges for this article was provided by the University of Glasgow. Conflict of interest: This project was supported by research collaboration with Siemens Healthcare.

Funding This research was supported by the British Heart Foundation Grant (Project Grant PG/11/2/28474), the National Health Service, and the Chief Scientist Office. Professor Berry was supported by a Senior Fellowship from the Scottish Funding Council. Dr Welsh is supported by BHF Fellowship FS/12/62/29889. Funding to pay the Open Access publication charges for this article was provided by the University of Glasgow. Conflict of interest: This project was supported by research collaboration with Siemens Healthcare. Acknowledgements We thank the patients who participated in this study and the staff in the Cardiology and Radiology Departments. We thank Peter Weale and Patrick Revell (Siemens Healthcare, UK).

See page 1131 for the editorial comment on this article (doi:10.1093/eurheartj/ehv570) Introduction The association between bleeding and increased morbidity and mortality after percutaneous coronary intervention (PCI) has prompted the implementation of bleeding reduction strategies at the time of PCI.1,2 According to clinical trial data, compared with femoral access, the radial artery approach for PCI has been demonstrated to have similar efficacy with the benefit of reduced major bleeding and access site complications.3–6 Owing to a favourable safety profile, several guidelines endorse the radial artery as the preferred PCI access site.7,8 Cangrelor is an intravenous P2Y12 receptor antagonist with an immediate (within 2 min) onset of action and a half-life of 3–6 min, allowing platelet function to return to baseline within 60 min of infusion cessation. In the Cangrelor vs. Standard Therapy to Achieve Optimal Management of Platelet Inhibition (CHAMPION) PHOENIX trial, the intravenous P2Y12 receptor antagonist, cangrelor, reduced the rate of ischaemic events at 48 h in patients undergoing percutaneous revascularization without a significant increase in severe bleeding or transfusions.9 In this pre-specified secondary analysis, we explore the efficacy and safety of cangrelor according to access site (femoral vs. radial) in the CHAMPION PHOENIX trial.

ate of ischaemic events at 48 h in patients undergoing percutaneous revascularization without a significant increase in severe bleeding or transfusions.9 In this pre-specified secondary analysis, we explore the efficacy and safety of cangrelor according to access site (femoral vs. radial) in the CHAMPION PHOENIX trial. Methods Patient population CHAMPION PHOENIX was a double-blind, double-dummy, placebo-controlled trial of cangrelor in patients undergoing PCI. Both the design and primary findings have been published previously.9,10 Briefly, 11 145 patients undergoing either elective or urgent PCI and receiving guideline-recommended therapy were randomized after angiography to receive a bolus (30 µg/kg) and infusion (4 µg/kg/min for a minimum of 2 h or the duration of the procedure whichever was longer) of cangrelor or a 600 or 300 mg loading dose of clopidogrel. The timing (before or after PCI) and dose of clopidogrel were at the discretion of the site investigator. The access approach for PCI was determined by the site investigator and did not require institutional review board (IRB) approval. At the end of the infusion, patients then received either 600 mg of clopidogrel (cangrelor group) or matching placebo (clopidogrel group; Figure 1). Figure 1 Study design. SA, stable angina; NSTE-ACS, non-ST segment elevation acute coronary syndrome; STEMI, ST segment elevation myocardial infarction; PCI, percutaneous coronary intervention.

sion, patients then received either 600 mg of clopidogrel (cangrelor group) or matching placebo (clopidogrel group; Figure 1). Figure 1 Study design. SA, stable angina; NSTE-ACS, non-ST segment elevation acute coronary syndrome; STEMI, ST segment elevation myocardial infarction; PCI, percutaneous coronary intervention. Endpoints The primary efficacy endpoint was a composite of death (by any cause), myocardial infarction (MI), ischaemia-driven revascularization (IDR), or stent thrombosis in the first 48 h following randomization. Criteria for MI within 48 h post-PCI were defined as an elevation in creatine kinase-myocardial band (CK-MB) greater than three times the upper limit of normal or by a combination of CK-MB elevation in addition to ischaemic symptoms, angiographic evidence, and/or ECG changes. The key secondary endpoint was the incidence of stent thrombosis at 48 h, which was defined according to the Academic Research Consortium criteria or as intraprocedural stent thrombosis.11 All events of death, MI, IDR, and stent thrombosis were adjudicated. The primary safety endpoint was severe non-coronary artery bypass grafting (CABG) bleeding according to the Global Use of Strategies to Open Occluded Coronary Arteries (GUSTO) criteria. Thrombolysis In Myocardial Infarction (TIMI) and Acute Catheterization and Urgent Intervention Triage strategY (ACUITY) bleeding definitions were also evaluated. Bleeding endpoints, based on pre-specified criteria, were derived from investigator-reported data using a computer algorithm. Bleeding endpoints were not adjudicated.

ysis In Myocardial Infarction (TIMI) and Acute Catheterization and Urgent Intervention Triage strategY (ACUITY) bleeding definitions were also evaluated. Bleeding endpoints, based on pre-specified criteria, were derived from investigator-reported data using a computer algorithm. Bleeding endpoints were not adjudicated. Statistical analysis This was a pre-specified analysis outlined in the CHAMPION PHOENIX study protocol, which was approved by local ethics committees. Independent verification of these analyses by Harvard Clinical Research Institute did not require specific IRB approval. Efficacy analyses were performed from the modified intention-to-treat (mITT) population, defined as subjects undergoing PCI and receiving study drug. Bleeding analyses were performed from the safety population, defined as subjects receiving study drug. Baseline characteristics were summarized by vascular access (femoral vs. radial) and treatment (cangrelor vs. clopidogrel); and were compared using analysis of variance for continuous variables and the χ2 test for categorical variables. Treatment comparisons within access site (cangrelor vs. clopidogrel) were not adjusted for baseline characteristics, were based on the χ2 test, and are presented as odds ratios with 95% confidence intervals (CIs). The number needed to treat (NNT) was derived within the femoral and radial subgroups as the inverse of the difference between the cangrelor and clopidogrel event rates. The interaction between treatment and PCI access site was tested using the Breslow–Day method. Time-to-event curves for the primary efficacy endpoint at 48 h were constructed using the Kaplan–Meier method and compared using the log-rank test.

inverse of the difference between the cangrelor and clopidogrel event rates. The interaction between treatment and PCI access site was tested using the Breslow–Day method. Time-to-event curves for the primary efficacy endpoint at 48 h were constructed using the Kaplan–Meier method and compared using the log-rank test. Multivariable logistic regression modelling was performed using propensity scores to assess the effect of PCI access site (femoral or radial), using femoral approach as the reference, with study treatment in the model and the propensity score based on the following potential variables: diagnosis at presentation, region (US/non-US), smoking status, hyperlipidaemia, previous MI, prior PCI, prior CABG, peripheral artery disease, planned clopidogrel loading dose (300 or 600 mg), and anticoagulant (bivalirudin or heparin). Statistical analyses were conducted using the SAS software, version 9.3 (SAS Institute, Cary, North Carolina, USA). Results Patients Out of 11 145 patients randomized in CHAMPION PHOENIX, 10 942 patients comprised the mITT population who received study treatment and underwent PCI. Of these patients, 8064 (74%) underwent PCI via the femoral artery and 2855 (26%) via the radial approach. Three patients in the radial cohort did not complete the 48-h post-PCI follow-up and were excluded from the efficacy endpoint analyses; 23 patients underwent brachial PCI and were not included in this analysis.

I. Of these patients, 8064 (74%) underwent PCI via the femoral artery and 2855 (26%) via the radial approach. Three patients in the radial cohort did not complete the 48-h post-PCI follow-up and were excluded from the efficacy endpoint analyses; 23 patients underwent brachial PCI and were not included in this analysis. Cangrelor vs. clopidogrel Baseline characteristics according to access site (femoral vs. radial) and randomized treatment (cangrelor vs. clopidogrel) are depicted in Table 1. In the femoral cohort, subjects randomized to cangrelor had higher rates of peripheral artery disease (8.2 vs. 6.7%, P = 0.01). In the radial cohort, subjects randomized to cangrelor had a lower median weight (84 vs. 85 kg, P = 0.008). Table 1 Baseline characteristics

cangrelor vs. clopidogrel) are depicted in Table 1. In the femoral cohort, subjects randomized to cangrelor had higher rates of peripheral artery disease (8.2 vs. 6.7%, P = 0.01). In the radial cohort, subjects randomized to cangrelor had a lower median weight (84 vs. 85 kg, P = 0.008). Table 1 Baseline characteristics Femoral Radial Cangrelor Clopidogrel P-value Cangrelor Clopidogrel P-value Characteristic Demographic n 4053 4011 1410 1445 Age, years Median 64 64 0.51 64.5 64 0.11 Interquartile range 56, 72 56, 72 57, 72 56, 71 Female sex, n (%) 1144 (28.2) 1107 (27.6) 0.53 413 (29.3) 382 (26.4) 0.09 Weight, kg Median 84 84 0.86 84 85 0.008 Interquartile range 73, 96 73, 96 73, 95 75, 96 Medical history, n (%) Diabetes mellitus 1106/4048 (27.3) 1094/4005 (27.3) 1.00 409/1407 (29.1) 438/1444 (30.3) 0.46 Current smoker 1142/3950 (28.9) 1148/3918 (29.3) 0.70 357/1380 (25.9) 398/1408 (28.3) 0.15 Hypertension 3239/4040 (80.2) 3156/3998 (78.9) 0.17 1127/1410 (79.9) 1164/1442 (80.7) 0.59 Hyperlipidaemia 2434/3507 (69.4) 2366/3447 (68.6) 0.49 923/1336 (69.1) 963/1376 (70.0) 0.61 Prior stroke or TIA 210/4039 (5.2) 180/3996 (4.5) 0.15 61/1407 (4.3) 61/1442 (4.2) 0.89 Prior myocardial infarction 844/4030 (20.9) 914/3982 (23.0) 0.03 245/1402 (17.5) 258/1435 (18.0) 0.73 Prior PTCA or PCI 913/4045 (22.6) 947/4003 (23.7) 0.25 353/1408 (25.1) 383/1444 (26.5) 0.38 CABG 495/4048 (12.2) 436/4006 (10.9) 0.06 81/1409 (5.7) 63/1444 (4.4) 0.09 Heart failure 412/4044 (10.2) 443/4002 (11.1) 0.20 137/1407 (9.7) 140/1440 (9.7) 0.99 Peripheral artery disease 328/4015 (8.2) 268/3978 (6.7) 0.01 119/1384 (8.6) 112/1427 (7.8) 0.47 TIA, transient ischaemic attack; PTCA, percutaneous transluminal coronary angioplasty; PCI, percutaneous coronary intervention; CABG, coronary artery bypass graft.

0.2) 443/4002 (11.1) 0.20 137/1407 (9.7) 140/1440 (9.7) 0.99 Peripheral artery disease 328/4015 (8.2) 268/3978 (6.7) 0.01 119/1384 (8.6) 112/1427 (7.8) 0.47 TIA, transient ischaemic attack; PTCA, percutaneous transluminal coronary angioplasty; PCI, percutaneous coronary intervention; CABG, coronary artery bypass graft. Radial vs. femoral Baseline characteristics according to access site only (femoral vs. radial) are depicted in the Supplementary material online, Table S1. Compared with the femoral cohort, subjects in the radial cohort had higher rates of diabetes mellitus (27.3 vs. 29.7%, P = 0.01) and prior PCI (23.1 vs. 25.8%, P = 0.004), but lower rates of current smoking (29.1 vs. 27.1%, P = 0.04), prior MI (21.9 vs. 17.7%, P < 0.0001), and prior CABG (11.6 vs. 5.0%, P < 0.0001) (Supplementary material online, Table S1). Procedure characteristics Cangrelor vs. clopidogrel Procedure characteristics according to access site (femoral vs. radial) and randomized treatment (cangrelor vs. clopidogrel) are depicted in Table 2. In the femoral cohort, subjects randomized to cangrelor had lower rates of glycoprotein IIb/IIIa inhibitor use (2.7 vs. 4.3%, P = 0.0001). Table 2 Procedure characteristics

ocedure characteristics according to access site (femoral vs. radial) and randomized treatment (cangrelor vs. clopidogrel) are depicted in Table 2. In the femoral cohort, subjects randomized to cangrelor had lower rates of glycoprotein IIb/IIIa inhibitor use (2.7 vs. 4.3%, P = 0.0001). Table 2 Procedure characteristics Femoral Radial Cangrelor Clopidogrel P-value Cangrelor Clopidogrel P-value Indication, n (%) 0.95 0.06 Stable angina 2290/4053 (56.5) 2258/4011 (56.3) 888/1410 (63.0) 905/1445 (62.6) NSTE-ACS 1099/4053 (27.1) 1085/4011 (27.1) 365/1410 (25.9) 341/1445 (23.6) STEMI 664/4053 (16.4) 668/4011 (16.7) 157/1410 (11.1) 199/1445 (13.8) Antithrombotic, n (%) Aspirin 3851/4050 (95.1) 3796/4007 (94.7) 0.47 1304/1410 (92.5) 1338/1444 (92.7) 0.86 Clopidogrel, 300 mg loading dose (planned) 1322/4053 (32.6) 1332/4011 (33.2) 0.57 79/1410 (5.6) 62/1445 (4.3) 0.11 Clopidogrel, 600 mg loading dose (planned) 2731/4053 (67.4) 2679/4011 (66.8) 0.57 1331/1410 (94.4) 1383/1445 (95.7) 0.11 Low molecular weight heparin 580/4053 (14.3) 579/4011 (14.4) 0.87 150/1410 (10.6) 174/1443 (12.1) 0.23 Unfractionated heparin 3045/4053 (75.1) 3020/4010 (75.3) 0.85 1220/1410 (86.5) 1243/1445 (86.0) 0.70 Fondaparinux 117/4053 (2.9) 92/4011 (2.3) 0.09 39/1409 (2.8) 43/1445 (3.0) 0.74 Bivalirudin 944/4053 (23.3) 940/4009 (23.4) 0.87 307/1410 (21.8) 326/1445 (22.6) 0.61 Glycoprotein IIb/IIIa inhibitor 111/4053 (2.7) 173/4011 (4.3) 0.0001 41/1410 (2.9) 54/1445 (3.7) 0.22 NSTE-ACS, non-ST-elevation acute coronary syndrome; STE ACS, ST-elevation acute coronary syndrome; PCI, percutaneous coronary intervention.

udin 944/4053 (23.3) 940/4009 (23.4) 0.87 307/1410 (21.8) 326/1445 (22.6) 0.61 Glycoprotein IIb/IIIa inhibitor 111/4053 (2.7) 173/4011 (4.3) 0.0001 41/1410 (2.9) 54/1445 (3.7) 0.22 NSTE-ACS, non-ST-elevation acute coronary syndrome; STE ACS, ST-elevation acute coronary syndrome; PCI, percutaneous coronary intervention. Femoral vs. radial Procedure characteristics according to access site only (femoral vs. radial) are depicted in the Supplementary material online, Table S2. Patients in the radial cohort had lower rates of aspirin (94.9 vs. 92.6%, P < 0.0001) and low molecular weight heparin (14.4 vs. 11.4%, P < 0.0001) use, but had higher rates of unfractionated heparin (75.2 vs. 86.3%, P < 0.0001) use at the time of PCI compared with patients in the femoral cohort. The clopidogrel 600 mg loading dose was used less often in the femoral (67.1%) cohort compared with the radial (95.1%) cohort, P < 0.0001. Percutaneous coronary intervention duration (17 min femoral vs. 18 min radial, P = 0.34) and success rate (98.2% femoral and 98.3% radial, P = 0.68) were similar in both groups. Of the femoral cohort, 53.5% of patients received a drug eluting stent, compared with 61.8% of patients in the radial cohort, P < 0.001.

0001. Percutaneous coronary intervention duration (17 min femoral vs. 18 min radial, P = 0.34) and success rate (98.2% femoral and 98.3% radial, P = 0.68) were similar in both groups. Of the femoral cohort, 53.5% of patients received a drug eluting stent, compared with 61.8% of patients in the radial cohort, P < 0.001. Outcomes Cangrelor vs. clopidogrel In the femoral cohort, the rate of the primary efficacy endpoint of death, MI, IDR, or stent thrombosis at 48 h was 4.8% with cangrelor vs. 6.0% with clopidogrel (odds ratio [OR] 95% CI = 0.79 [0.65–0.96]), NNT 84; in the radial cohort, the primary efficacy endpoint rate was 4.4% with cangrelor vs. 5.7% with clopidogrel (OR [95% CI] = 0.76 [0.54–1.06]), P-interaction = 0.83; NNT 74. Figure 2A and B depicts the Kaplan–Meier estimates for the time-to-event for the primary endpoint in both the femoral and radial cohorts. Among the femoral cohorts, the key secondary endpoint of stent thrombosis at 48 h was 0.8% with cangrelor vs. 1.5% with clopidogrel (OR [95% CI] = 0.52 [0.34–0.80]); in the radial cohort, the rate of stent thrombosis at 48 h was 0.9% with cangrelor vs. 0.8% with clopidogrel (OR [95% CI] = 1.11 [0.51–2.45]), P-interaction 0.09. The effects of cangrelor on the primary and secondary endpoints in the overall study population and according to access site are shown in Figure 3A. Figure 2 (A) Kaplan–Meier curves for the primary efficacy endpoint in the subgroup undergoing femoral access (cangrelor vs. clopidogrel). (B) Kaplan–Meier curves for the primary efficacy endpoint in the subgroup undergoing radial access (cangrelor vs. clopidogrel). HR, hazard ratio; CI, confidence interval.

Figure 3A. Figure 2 (A) Kaplan–Meier curves for the primary efficacy endpoint in the subgroup undergoing femoral access (cangrelor vs. clopidogrel). (B) Kaplan–Meier curves for the primary efficacy endpoint in the subgroup undergoing radial access (cangrelor vs. clopidogrel). HR, hazard ratio; CI, confidence interval. Figure 3 (A) Efficacy of cangrelor vs. clopidogrel in the femoral and radial subgroups. MI, myocardial infarction; IDR, ischaemia-driven revascularization; ST, stent thrombosis; OR, odds ratio; CI, confidence interval. (B) Bleeding with cangrelor vs. clopidogrel in the femoral and radial subgroups. GUSTO, Global Use of Strategies to Open Occluded Arteries; TIMI, Thrombolysis In Myocardial Infarction; ACUITY, Acute Catheterization and Urgent Intervention Triage strategY; OR, odds ratio; CI, confidence interval.

I, confidence interval. (B) Bleeding with cangrelor vs. clopidogrel in the femoral and radial subgroups. GUSTO, Global Use of Strategies to Open Occluded Arteries; TIMI, Thrombolysis In Myocardial Infarction; ACUITY, Acute Catheterization and Urgent Intervention Triage strategY; OR, odds ratio; CI, confidence interval. In both access cohorts, there were no significant differences in the rates of GUSTO severe bleeding, TIMI major bleeding, or blood transfusions in patients treated with cangrelor compared with clopidogrel. In the femoral cohort, the rate of ACUITY major bleeding was 5.2% with cangrelor vs. 3.1% with clopidogrel (OR [95% CI] = 1.69 [1.35–2.12), P < 0.0001; in the radial cohort, the rate of ACUITY major bleeding was 1.5% with cangrelor vs. 0.7% with clopidogrel (OR [95% CI] = 2.17 [1.02–4.62]), P = 0.04 and P-interaction 0.54. The effects of cangrelor on bleeding according to access site and categorized by the various bleeding criteria are listed in Table 3. The effects of cangrelor on bleeding in the overall study population and according to access site are shown in Figure 3B. Table 3 Safety endpoints at 48 h after randomization

ction 0.54. The effects of cangrelor on bleeding according to access site and categorized by the various bleeding criteria are listed in Table 3. The effects of cangrelor on bleeding in the overall study population and according to access site are shown in Figure 3B. Table 3 Safety endpoints at 48 h after randomization Endpoint Femoral Radial P-interaction Cangrelor Clopidogrel OR (95% CI) P-value Cangrelor Clopidogrel OR (95% CI) P-value GUSTO bleeding Severe or life threatening 7/4052 (0.2) 4/4012 (0.1) 1.73 (0.51–5.93) 0.37 2/1411 (0.1) 2/1444 (0.1) 1.02 (0.14–7.28) 0.98 0.65 Moderate 20/4052 (0.5) 12/4012 (0.3) 1.65 (0.81–3.39) 0.16 2/1411 (0.1) 1/1444 (0.1) 2.05 (0.19–22.61) 0.55 0.87 Severe/moderate 27/4052 (0.7) 16/4012 (0.4) 1.68 (0.90–3.11) 0.10 4/1411 (0.3) 3/1444 (0.2) 1.37 (0.31–6.11) 0.68 0.80 TIMI bleeding Major 3/4052 (0.1) 3/4012 (0.1) 0.99 (0.20–4.91) 0.99 2/1411 (0.1) 2/ 1444 (0.1) 1.02 (0.14–7.28) 0.98 0.98 Minor 9/4052 (0.2) 3/4012 (0.1) 2.97 (0.80–11.00) 0.09 0/1411 (0.0) 0/1444 (0.0) Major/minor 12/4052 (0.3) 6/4012 (0.1) 1.98 (0.74–5.29) 0.16 2/1411 (0.1) 2/1444 (0.1) 1.02 (0.14–7.28) 0.98 0.55 ACUITY bleeding Major 209/4052 (5.2) 125/4012 (3.1) 1.69 (1.35–2.12) <0.0001 21/1411 (1.5) 10/1444 (0.7) 2.17 (1.02–4.62) 0.04 0.54 Minor 510/4052 (12.6) 369/4012 (9.2) 1.42 (1.23–1.64) <0.0001 139/1411 (9.9) 94/1444 (6.5) 1.57 (1.19–2.06) 0.001 0.53 Major/minor 690/4052 (17.0) 483/4012 (12.0) 1.50 (1.32–1.70) <0.0001 158/1411 (11.2) 104/1444 (7.2) 1.62 (1.25–2.11) 0.0002 0.59 Blood transfusion 22/4052 (0.5) 14/4012 (0.3) 1.56 (0.80–3.05) 0.19 3/1411 (0.2) 2/1444 (0.1) 1.54 (0.26–9.21) 0.63 0.99 GUSTO, Global Use of Strategies to Open Occluded Arteries; Thrombolysis In Myocardial Infarction; ACUITY, Acute Catheterization and Urgent Intervention Triage strategY; OR, odds ratio; CI, confidence interval.

sfusion 22/4052 (0.5) 14/4012 (0.3) 1.56 (0.80–3.05) 0.19 3/1411 (0.2) 2/1444 (0.1) 1.54 (0.26–9.21) 0.63 0.99 GUSTO, Global Use of Strategies to Open Occluded Arteries; Thrombolysis In Myocardial Infarction; ACUITY, Acute Catheterization and Urgent Intervention Triage strategY; OR, odds ratio; CI, confidence interval. Femoral vs. radial Among patients undergoing PCI via the femoral artery, the rate of the primary efficacy endpoint at 48 h was 5.4 vs. 5.1% in the radial group (unadjusted OR 95% CI = 0.95 [0.78–1.15]), P = 0.58. After multivariable analysis, compared with patients undergoing PCI via the femoral artery, there was no difference in the adjusted rate of primary efficacy endpoint in the radial access cohort (OR [95% CI] = 1.03 [0.81–1.29]), P = 0.83. Efficacy endpoints (unadjusted and adjusted) according to access site are displayed in Supplementary material online, Table S3.

atients undergoing PCI via the femoral artery, there was no difference in the adjusted rate of primary efficacy endpoint in the radial access cohort (OR [95% CI] = 1.03 [0.81–1.29]), P = 0.83. Efficacy endpoints (unadjusted and adjusted) according to access site are displayed in Supplementary material online, Table S3. In the femoral cohort, the rate of GUSTO severe/moderate bleeding was 0.5% compared with 0.2% in the radial cohort (unadjusted OR [95% CI] = 0.46 [0.21–1.02]; P = 0.05). The rate of TIMI major/minor bleeding in the femoral cohort was 0.2% compared with 0.1% in the radial cohort (unadjusted OR [95% CI] = 0.63 [0.21–1.85]; P = 0.39). Lastly, the rate of ACUITY major/minor bleeding was 14.5% in the femoral cohort and 9.2% in the radial group (unadjusted OR [95% CI] = 0.60 [0.52–0.68]; P < 0.0001). After multivariable analysis, compared with PCI via the femoral artery, using the radial approach was associated with a decreased odds of GUSTO severe/moderate (OR [95% CI] = 0.35 [0.12–1.01]; P = 0.05), TIMI major/minor bleeding (OR [95% CI] = 0.34 [0.07–1.54]; P = 0.16), and ACUITY major/minor bleeding (OR [95% CI] = 0.70 [0.59–0.83]; P < 0.0001 (Table 4). Table 4 Safety endpoints at 48 h: radial vs. femoral

oach was associated with a decreased odds of GUSTO severe/moderate (OR [95% CI] = 0.35 [0.12–1.01]; P = 0.05), TIMI major/minor bleeding (OR [95% CI] = 0.34 [0.07–1.54]; P = 0.16), and ACUITY major/minor bleeding (OR [95% CI] = 0.70 [0.59–0.83]; P < 0.0001 (Table 4). Table 4 Safety endpoints at 48 h: radial vs. femoral Endpoint Femoral Radial OR (95% CI) unadjusted P-value OR (95% CI) adjusted P-value GUSTO bleeding Severe or life threatening 11/8064 (0.1) 4/2855 (0.1) 1.03 (0.33, 3.23) 0.96 Moderate 32/8064 (0.4) 3/2855 (0.1) 0.26 (0.08, 0.86) 0.02 Severe/moderate 43/8064 (0.5) 7/2855 (0.2) 0.46 (0.21, 1.02) 0.05 0.35 (0.12–1.01) 0.05 TIMI bleeding Major 6/8064 (0.1) 4/2855 (0.1) 1.89 (0.53, 6.67) 0.32 Minor 12/8064 (0.1) 0/2855 (0.0) – 0.04 Major/minor 18/8064 (0.2) 4/2855 (0.1) 0.63 (0.21, 1.85) 0.39 0.34 (0.07–1.54) 0.16 ACUITY bleeding Major 334/8064 (4.1) 31/2855 (1.1) 0.25 (0.18, 0.37) <0.0001 Minor 879/8064 (10.9) 233/2855 (8.2) 0.72 (0.63, 0.85) <0.0001 Major/minor 1173/8064 (14.5) 262/2855 (9.2) 0.60 (0.52, 0.68) <0.0001 0.70 (0.59–0.83) <0.0001 All non-CABG bleeding 1173/8064 (14.5) 262/2855 (9.2) 0.60 (0.52, 0.68) <0.0001 GUSTO, Global Use of Strategies to Open Occluded Arteries; Thrombolysis In Myocardial Infarction; ACUITY, Acute Catheterization and Urgent Intervention Triage strategY; CABG, coronary artery bypass graft; OR, odds ratio; CI, confidence interval.

on-CABG bleeding 1173/8064 (14.5) 262/2855 (9.2) 0.60 (0.52, 0.68) <0.0001 GUSTO, Global Use of Strategies to Open Occluded Arteries; Thrombolysis In Myocardial Infarction; ACUITY, Acute Catheterization and Urgent Intervention Triage strategY; CABG, coronary artery bypass graft; OR, odds ratio; CI, confidence interval. Discussion Intravenous adenosine diphosphate (ADP) receptor blockade with cangrelor, when compared with clopidogrel, reduced the primary composite outcome of death, MI, IDR, or stent thrombosis at 48 h after randomization regardless of PCI access site. Among the femoral access subjects, cangrelor compared with clopidogrel reduced the odds of the primary composite outcome by 21%; within the radial cohort, there was a consistent 24% reduction in the odds of the primary composite outcome. Although the interaction tests for access site did not reach statistical significance, when the primary composite endpoint was evaluated according to access site, cangrelor demonstrated a reduction in odds of ischaemic events among patients undergoing femoral PCI. In patients undergoing PCI via the radial artery, cangrelor demonstrated a non-significant trend towards fewer ischaemic events, likely due to lack of statistical power because of a smaller sample size. In the femoral cohort, cangrelor's benefit with respect to ischaemic events was driven by a reduction in MI and stent thrombosis, consistent with the overall CHAMPION PHOENIX results. The radial cohort, however, only experienced a reduction in MI. The lack of benefit regarding stent thrombosis in the radial cohort is conceivably due to a low event frequency within this particular subgroup: 95 total (1.2%) in the femoral vs. 25 total (0.9%) in the radial.

nt with the overall CHAMPION PHOENIX results. The radial cohort, however, only experienced a reduction in MI. The lack of benefit regarding stent thrombosis in the radial cohort is conceivably due to a low event frequency within this particular subgroup: 95 total (1.2%) in the femoral vs. 25 total (0.9%) in the radial. In both the femoral and radial groups, cangrelor compared with clopidogrel was not associated with a significant increase in the pre-specified GUSTO-defined severe bleeding, the primary safety endpoint, or in blood transfusions; though more sensitive definitions such as ACUITY-defined bleeding did show increased rates of bleeding with cangrelor in both the femoral and radial cohorts. It is important to note that the absolute rates of GUSTO severe/moderate bleeding and blood transfusions were approximately two to three times higher with the femoral approach compared with the radial, in both the cangrelor and clopidogrel arms of the study. These findings are consistent with prior studies. The analysis of all femoral vs. radial randomized PCI clinical trials has found radial access, when compared with femoral, to be associated with a 42% reduction in non-CABG bleeding.5 Contemporary studies have demonstrated that up to 70% of bleeding events occurring in the PCI setting can be attributed to complications arising at the site of vascular access.12 The present study suggests that in a large contemporary international trial, the radial approach for PCI has the potential to play a key role in reducing periprocedural bleeding in a wide variety of PCI patients.

s occurring in the PCI setting can be attributed to complications arising at the site of vascular access.12 The present study suggests that in a large contemporary international trial, the radial approach for PCI has the potential to play a key role in reducing periprocedural bleeding in a wide variety of PCI patients. It is postulated that a reduction in access site bleeding may in turn lead to fewer subsequent adverse events.13 For example, the aforementioned 42% reduction in major non-CABG bleeding associated with the radial approach for PCI was paralleled with an aggregate reduction of major adverse cardiac events (death, MI, or stroke) of 14%, P = 0.005.5 After multivariable analysis the present study finds the odds of periprocedural bleeding 30–66% lower, depending on bleeding definition, when PCI was performed via the radial artery compared with the femoral approach. However, the favourable bleeding profile associated with the radial artery approach to PCI in CHAMPION PHOENIX did not translate into a reduction in the primary efficacy endpoint at 48 h.

ing 30–66% lower, depending on bleeding definition, when PCI was performed via the radial artery compared with the femoral approach. However, the favourable bleeding profile associated with the radial artery approach to PCI in CHAMPION PHOENIX did not translate into a reduction in the primary efficacy endpoint at 48 h. There are certain limitations to this analysis. First, similar to the overall CHAMPION PHOENIX trial, there is the potential for the benefit of cangrelor to be attenuated in the setting of more prolonged pretreatment with clopidogrel or with the use or ticagrelor or prasugrel. Secondly, bleeding endpoints were not adjudicated. Thirdly, the treatment by radial vs. femoral access was not randomized and even with the adjusted analyses, there may be residual confounding. Lastly, CHAMPION PHOENIX was not powered to test the interaction between treatment and PCI access site; therefore, all interaction terms should be interpreted with caution. In CHAMPION PHOENIX, intravenous ADP-receptor inhibition with cangrelor reduced ischaemic events with no significant increase in severe bleeding or blood transfusions regardless of PCI access site. Compared with the femoral approach, rates of bleeding complications appeared to be lower with radial access for PCI in both randomized arms of the study.

s ADP-receptor inhibition with cangrelor reduced ischaemic events with no significant increase in severe bleeding or blood transfusions regardless of PCI access site. Compared with the femoral approach, rates of bleeding complications appeared to be lower with radial access for PCI in both randomized arms of the study. Authors' contributions J.P.: performed statistical analysis. D.L.B.: handled funding and supervision, acquired the data, and conceived and designed the research. J.A.G.: drafted the manuscript. R.A.H.: made critical revision of the manuscript for key intellectual content. All authors analyzed the data and interpreted the results. All authors contributed to writing revisions and approved the final manuscript. Supplementary material Supplementary material is available at European Heart Journal online. Funding The CHAMPION PHOENIX trial was funded by The Medicines Company.

Authors' contributions J.P.: performed statistical analysis. D.L.B.: handled funding and supervision, acquired the data, and conceived and designed the research. J.A.G.: drafted the manuscript. R.A.H.: made critical revision of the manuscript for key intellectual content. All authors analyzed the data and interpreted the results. All authors contributed to writing revisions and approved the final manuscript. Supplementary material Supplementary material is available at European Heart Journal online. Funding The CHAMPION PHOENIX trial was funded by The Medicines Company. Conflict of interest: J.A.G. discloses the following relationships—Honoraria: Boehringer Ingelheim. R.A.H. discloses the following relationships—Advisory Board: Evidint, Regado, Scanadu; Honoraria: Amgen, Daiichi-Lilly, Gilead Sciences, Gilead Sciences, Inc., Janssen R and D, Medtronic, Merck, Novartis Corporation, The Medicines Company, Vida Health, Vox Media, WebMD; Other: American Heart Association; Research Funding: AstraZeneca, BMS, CSL Behring, GSK, Merck, Portola, Sanofi-Aventis, The Medicines Company; Ownership Interest: Element Science, MyoKardia. J.C.B. has no conflict of interests to declare. G.W.S. discloses the following relationships—Honoraria: Boston Scientific, InspireMD, Atriaum, Eli Lilly—Daiichi Sankyo partnership, AstraZeneca. Ph.G.S. discloses the following relationships—Honoraria: Amarin, AstraZeneca, Bayer; Other: The Medicines Company. C.M.G. discloses the following relationships—Honoraria: The Medicines Company. C.W.H. discloses the following relationships—Honoraria: AstraZeneca, Sanofi-Aventis, Lilly; Research Funding: Astra Zeneca, The Medicines Company. M.J.P. discloses the following relationships—Honoraria: AstraZeneca, Merck & Co., Accriva Diagnostics, The Medicines Company. P.G. discloses the following relationships—Honoraria: Abbott Vascular. J.P. discloses the following relationships—Employment: The Medicines Company. E.N.D. discloses the following relationships—Employment: The Medicines Company. K.W.M. discloses the following relationships—Honoraria: Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Cubist, Eli Lilly, Epson, Forest, Glaxo Smith Kline, Johnson & Johnson, Medtronic, Merck, Mt. Sinai, Myokardia, Omthera, Portola, Purdue Pharma, Spring Publishing, Vindico, WebMD; Research Funding: Daiichi, Johnson & Johnson, Medtronic, St. Jude, Tenax. H.D.W. discloses the following relationships—Honoraria: AstraZeneca; Research Funding: Sanofi-Aventis, Eli Lilly, National Health Institute, Glaxo Smith Kline, Merck Sharpe & Dohme, AstraZeneca. D.L.B.

e Pharma, Spring Publishing, Vindico, WebMD; Research Funding: Daiichi, Johnson & Johnson, Medtronic, St. Jude, Tenax. H.D.W. discloses the following relationships—Honoraria: AstraZeneca; Research Funding: Sanofi-Aventis, Eli Lilly, National Health Institute, Glaxo Smith Kline, Merck Sharpe & Dohme, AstraZeneca. D.L.B. discloses the following relationships—Advisory Board: Cardax, Elsevier Practice Update Cardiology, Medscape Cardiology, Regado Biosciences; Board of Directors: Boston VA Research Institute, Society of Cardiovascular Patient Care; Chair: American Heart Association Get With The Guidelines Steering Committee; Data Monitoring Committees: Duke Clinical Research Institute, Harvard Clinical Research Institute, Mayo Clinic, Population Health Research Institute; Honoraria: American College of Cardiology (Senior Associate Editor, Clinical Trials and News, ACC.org), Belvoir Publications (Editor in Chief, Harvard Heart Letter), Duke Clinical Research Institute (clinical trial steering committees), Harvard Clinical Research Institute (clinical trial steering committee), HMP Communications (Editor in Chief, Journal of Invasive Cardiology), Journal of the American College of Cardiology (Guest Editor; Associate Editor), Population Health Research Institute (clinical trial steering committee), Slack Publications (Chief Medical Editor, Cardiology Today's Intervention), WebMD (CME steering committees); Other: Clinical Cardiology (Deputy Editor); Research Funding: Amarin, AstraZeneca, Bristol-Myers Squibb, Eisai, Ethicon, Forest Laboratories, Ischemix, Medtronic, Pfizer, Roche, Sanofi-Aventis, The Medicines Company (including for his role as Co-Chair of CHAMPION PHOENIX); Site Co-Investigator: Biotronik, St. Jude Medical; Trustee: American College of Cardiology; Unfunded Research: FlowCo, PLx Pharma, Takeda.

yers Squibb, Eisai, Ethicon, Forest Laboratories, Ischemix, Medtronic, Pfizer, Roche, Sanofi-Aventis, The Medicines Company (including for his role as Co-Chair of CHAMPION PHOENIX); Site Co-Investigator: Biotronik, St. Jude Medical; Trustee: American College of Cardiology; Unfunded Research: FlowCo, PLx Pharma, Takeda. Acknowledgements We thank Steven E. Elkin MS and Debra Bernstein PhD of The Medicines Company for their statistical support, along with Yuyin Liu MS and Lanyu Lei MS of the Harvard Clinical Research Institute for their independent verification of the analyses. Harvard Clinical Research Institute received funding from The Medicines Company for these analyses.

See page 1920 for the editorial comment on this article (doi:10.1093/eurheartj/ehw186) Introduction Primary percutaneous coronary intervention (PPCI) is the default reperfusion therapy for ST-segment elevation myocardial infarction (STEMI).1 However, residual mortality (up to 12% at 6 months1) and morbidity may be partially attributable to sub-optimal microvascular perfusion.2,3 Microvascular obstruction (MVO) can occur in up to 70% patients when detected with cardiovascular magnetic resonance (CMR) imaging.4,5 Microvascular obstruction may contribute to infarct expansion, adverse left ventricular (LV) remodelling and adverse clinical outcomes, independent of infarct size.4,5

lar perfusion.2,3 Microvascular obstruction (MVO) can occur in up to 70% patients when detected with cardiovascular magnetic resonance (CMR) imaging.4,5 Microvascular obstruction may contribute to infarct expansion, adverse left ventricular (LV) remodelling and adverse clinical outcomes, independent of infarct size.4,5 Various strategies have been proposed to attenuate MVO6 and have included calcium channel blockers, nicorandil, atrial natriuretic peptide, glycoprotein IIb/IIIa (GPIIb/IIIa) inhibitors, thrombolytics, and, perhaps most studied, vaso-active agents such as sodium nitroprusside (SNP)2,7,8 and adenosine.8–12 Adenosine inhibits neutrophil-related processes central to the evolution of MVO and in a canine model, reduced ischaemia-reperfusion injury (IRI), limited infarct size, and improved ventricular function.13 Sodium nitroprusside is a direct nitric oxide donor that mediates potent arteriolar vasodilation, inhibits platelet adhesion and promotes anti-inflammatory processes,14 which effectively reduce MVO in animal IRI models.15 Previous clinical trials with these agents varied in their design and they mostly lacked a sensitive method to detect MVO, which may have contributed to the multiple conflicting results seen.7–9,12,16,17 Uncertainty remains regarding the potential therapeutic impact of these agents.1

reduce MVO in animal IRI models.15 Previous clinical trials with these agents varied in their design and they mostly lacked a sensitive method to detect MVO, which may have contributed to the multiple conflicting results seen.7–9,12,16,17 Uncertainty remains regarding the potential therapeutic impact of these agents.1 We hypothesized that, in patients undergoing PPCI for STEMI, appropriate doses of intra-coronary (IC) adenosine or SNP delivered distally to the coronary bed, and again after stent deployment, was the optimal administration that would reduce infarct size and MVO as determined by the sensitive measure of in-patient CMR. Methods Study design The trial design and methods have been published previously.16 Briefly, REFLO-STEMI was a multi-centre, prospective, randomized, open-label, controlled trial, with blinded endpoint analysis, conducted in accordance with the Helsinki Declaration. Ethical approval for the study (reference 11/H0405/10) was obtained from the National Research Ethics Service (UK). ST-segment elevation myocardial infarction patients with single vessel disease and thrombolysis in myocardial infarction (TIMI) flow grades 0–1 in the IRA were enrolled between October 2011 and April 2014 in four regional cardiac centres in the UK. All patients >18 years age, presenting within 6 h of symptom onset, with ST-segment elevation ≥2 mm in ≥2 contiguous leads and with a baseline corrected QT interval (QTc) <450 ms on admission electrocardiogram (ECG) were eligible and assented.16 Full eligibility criteria are shown in Supplementary material online, Table S1.

atients >18 years age, presenting within 6 h of symptom onset, with ST-segment elevation ≥2 mm in ≥2 contiguous leads and with a baseline corrected QT interval (QTc) <450 ms on admission electrocardiogram (ECG) were eligible and assented.16 Full eligibility criteria are shown in Supplementary material online, Table S1. Randomization and treatment Patients were randomized using a 24-h interactive voice recognition service with concealed allocation. All patients were pre-treated with aspirin 300 mg and prasugrel 60 mg or ticagrelor 180 mg1 as well as bivalirudin 0.75 mg/kg bolus plus infusion of 1.75 mg/kg/h until completion of PPCI or continued for 4 h if clinically indicated. Bivalirudin, recommended in all patients, was standard of care in the enrolling centres at the time of trial initiation. Allocation was 1:1:1 to adenosine, SNP, or control (standard PPCI alone) with stratification by ‘symptoms to balloon either less than or ≥3 h’ and ‘anterior infarction or not’. Manual thrombectomy was mandated in all patients and the thrombectomy catheter then used to deliver the first drug bolus (adenosine 1 mg or SNP 250 μg) IC distal in the coronary bed, after thorough flushing of the catheter. Immediately following stent deployment, and providing a repeat QTc was <450 ms (or <60 ms increased over baseline value), the second dose (adenosine 1 mg if IRA was the right coronary artery otherwise 2 mg or SNP 250 μg) was delivered over 1 min via the guide catheter (see Figure 2).

thorough flushing of the catheter. Immediately following stent deployment, and providing a repeat QTc was <450 ms (or <60 ms increased over baseline value), the second dose (adenosine 1 mg if IRA was the right coronary artery otherwise 2 mg or SNP 250 μg) was delivered over 1 min via the guide catheter (see Figure 2). Study outcomes Cardiac magnetic resonance was performed at 24–96 h on a 3.0T scanner (Figure 1) to determine infarct size (primary endpoint).16 Secondary outcomes included: MVO extent assessed on early- and late-gadolinium enhanced (EGE, LGE) images and the presence of intra-myocardial haemorrhage (IMH) on CMR; ST-segment resolution (STR) on ECG performed 90 min post-PPCI; major adverse cardiovascular events (death, MI, heart failure, target lesion revascularization, and stroke; major adverse cardiac events, MACE) within 6 months. Important adverse events and study outcome measures are defined online (see Supplementary material online, Tables S5 and S6, respectively). A clinical events committee, blinded to patient details and treatment allocation, reviewed and adjudicated key trial adverse events using original source documents. An independent DSMB, with an independent statistician, periodically reviewed trial conduct and outcome data. Figure 1 Cardiac magnetic resonance Protocol. 4C, 2C, 3C = 4,2,3-chamber long-axis views; SPAMM, spatial magnetization modulation; ETL, echo train length; FOV, field of view; LV, left ventricle; SAX, short axis; SSFP, steady-state-free precession; ST, slice thickness; TI, inversion time; TE, echo time; TR, repetition time.

re 1 Cardiac magnetic resonance Protocol. 4C, 2C, 3C = 4,2,3-chamber long-axis views; SPAMM, spatial magnetization modulation; ETL, echo train length; FOV, field of view; LV, left ventricle; SAX, short axis; SSFP, steady-state-free precession; ST, slice thickness; TI, inversion time; TE, echo time; TR, repetition time. Cardiac magnetic resonance analysis Cardiac magnetic resonance was performed as previously described16 and analysis, blinded to all patient details and treatment allocation, was undertaken offline in a central core lab (University of Leicester) using cmr42 (Circle Cardiovascular Imaging, Calgary, Canada). Early-gadolinium enhanced and LGE images were acquired after 1–3 and 10–15 min, respectively, following 0.15 mmol/kg gadolinium-DTPA (Magnevist, Bayer, Germany) administration using a single-shot or segmented inversion-recovery gradient-echo sequence. Early- and late-MVO (E-MVO, L-MVO) were defined as hypoenhancement within the infarct territory on EGE and LGE imaging, respectively. Left ventricular volumes, infarct size, myocardial oedema (area at risk, AAR), intra-myocardial haemorrhage (IMH), and MSI were determined as described previously.16 Sample size To detect a reduction in infarct size from 20 to 15% of LV mass, assuming a standard deviation of 10%,18–22 α of 0.05 and two tailed, with 80% power, a total of 192 patients were required. Allowing for a drop-out rate of 20% between PPCI and CMR resulted in a final sample size requirement of 240 patients.

Cardiac magnetic resonance analysis Cardiac magnetic resonance was performed as previously described16 and analysis, blinded to all patient details and treatment allocation, was undertaken offline in a central core lab (University of Leicester) using cmr42 (Circle Cardiovascular Imaging, Calgary, Canada). Early-gadolinium enhanced and LGE images were acquired after 1–3 and 10–15 min, respectively, following 0.15 mmol/kg gadolinium-DTPA (Magnevist, Bayer, Germany) administration using a single-shot or segmented inversion-recovery gradient-echo sequence. Early- and late-MVO (E-MVO, L-MVO) were defined as hypoenhancement within the infarct territory on EGE and LGE imaging, respectively. Left ventricular volumes, infarct size, myocardial oedema (area at risk, AAR), intra-myocardial haemorrhage (IMH), and MSI were determined as described previously.16 Sample size To detect a reduction in infarct size from 20 to 15% of LV mass, assuming a standard deviation of 10%,18–22 α of 0.05 and two tailed, with 80% power, a total of 192 patients were required. Allowing for a drop-out rate of 20% between PPCI and CMR resulted in a final sample size requirement of 240 patients. Statistical methods Primary analysis was by intention-to-treat with a secondary per-protocol analysis in those patients who received both doses of the investigational drugs. Continuous variables, including infarct size, were investigated for normality and were log-transformed when found to be non-normally distributed. Normally distributed continuous variables were expressed as mean ± standard deviation and compared using t-tests and analysis of variance (ANOVA). Non-normally distributed data were presented as median (25th–75th quartiles) and compared using non-parametric methods (Mann–Whitney or Kruskal–Wallis). Comparison between groups for categorical outcomes was undertaken using χ2 tests. Potential significant confounders of infarct size (age, sex, diabetes, anterior MI, ischaemia time and collateral blood flow to the infarct territory determined by the Rentrop score23) were assessed using a forward selection procedure (statistical significance level of 5%). Multivariable analysis using linear regression was used to adjust for significant confounders. Time-to-event Cox proportional hazards regression models were used to assess predictors of first MACE (both at 30 days and at the study end) and to obtain unadjusted and adjusted hazard ratios (HRs) with 95% confidence intervals (CIs) for adenosine/SNP vs. control. In addition, actuarial event rates of MACE by 1 and 6 months for each intervention were reported with 95% CIs. The assumption of proportional hazards was assessed by a global hypothesis test.24 No formal adjustment for multiple hypothesis testing was used, but P-values were interpreted cautiously. All statistical analyses were performed using SPSS (Version 20.0. Armonk, NY: IBM Corporation) and R (Version 3.1.2. The R Foundation for Statistical Computing, Vienna).

s was assessed by a global hypothesis test.24 No formal adjustment for multiple hypothesis testing was used, but P-values were interpreted cautiously. All statistical analyses were performed using SPSS (Version 20.0. Armonk, NY: IBM Corporation) and R (Version 3.1.2. The R Foundation for Statistical Computing, Vienna). Results Patient recruitment is outlined in Figure 2. Two hundred and forty-seven patients were randomized with 207 (84%) undergoing CMR. Ten patients did not complete the CMR (due to claustrophobia or musculoskeletal discomfort) so that the primary outcome of infarct size was assessed in 197 patients (80% of those randomized). Figure 2 Study recruitment flowchart (CONSORT). CABG, coronary artery bypass graft; CTO, chronic total occlusion; IRA, infarct-related artery; LCA, left coronary artery; POBA, plain old balloon angioplasty; PPCI, primary percutaneous coronary intervention; RCA, right coronary artery; SNP, sodium nitroprusside; TIMI, thombolysis in myocardial infarction. Baseline characteristics Baseline characteristics for randomized patients are presented in Table 1 (see Supplementary material online, Table S7 for patients completing CMR by treatment allocation). There were no significant differences in characteristics of those randomized and those who completed CMR. Table 1 Demography of the total trial population

e characteristics for randomized patients are presented in Table 1 (see Supplementary material online, Table S7 for patients completing CMR by treatment allocation). There were no significant differences in characteristics of those randomized and those who completed CMR. Table 1 Demography of the total trial population Characteristics Adenosine (n = 82) SNP (n = 79) Control (n = 86) P-value Clinical Age (years) 57.9 ± 12.8 60.5 ± 13.0 59.5 ± 11.2 0.406 Male 65/82 (79.3) 66/79 (83.5) 64/86 (74.4) 0.355 Hypertension 23/82 (28.0) 26/79 (32.9) 22/86 (25.6) 0.574 Current smoking 47/82 (57.3) 41/79 (51.9) 45/86 (52.3) 0.332 Diabetes 6/82 (7.3) 12/79 (15.2) 9/86 (10.5) 0.274 Hypercholesterolaemia 17/82 (20.7) 23/79 (29.1) 11/86 (12.8) 0.035 Previous MI 0/81 (0) 3/79 (3.8) 3/86 (3.5) 0.219 Killip class >1 3/82 (3.7) 4/79 (5.1) 4/86 (4.7) 0.905 Total ischaemia time (min) 159 (124–221) 150 (122–201) 145 (105–196) 0.169 BMI (kg/m2) 27.5 (25.0–30.1) 26.1 (24.3–30.8) 27.3 (24.4–30.6) 0.659 SBP (mmHg) 137.5 ± 25.1 133.0 ± 23.4 135.1 ± 23.3 0.505 DBP (mmHg) 86.6 ± 19.2 81.8 ± 16.4 80.5 ± 17.0 0.067 HR (beats/min) 74.0 ± 17.3 71.9 ± 14.8 71.2 ± 14.6 0.487 Cr clearance (mL/min/1.73 m2) 98.1 ± 28.1 93.4 ± 29.1 92.7 ± 25.6 0.401 Anti-platelet use Aspirin 82/82 (100.0) 79/79 (100.0) 86/86 (100.0) 1.000 Clopidogrel 16/82 (19.5) 13/79 (16.5) 12/86 (14.0) 0.625 Prasugrel 38/82 (46.3) 33/79 (41.8) 40/86 (46.5) 0.790 Ticagrelor 28/82 (34.1) 33/79 (41.8) 34/86 (39.5) 0.591 Medication on discharge β-Blocker 72/78 (92.3) 69/75 (92.0) 73/81 (90.1) 0.867 ACE-inhibitor/A2RB 74/78 (94.9) 75/75 (100.0) 79/81 (97.5) 0.133 Statin 77/78 (98.7) 74/75 (98.7) 80/81 (98.8) 0.999 Infarct-related artery LAD—proximal 19/82 (23.2) 18/79 (22.8) 20/86 (23.3) 0.997 LAD—other 13/82 (15.9) 15/79 (19.0) 14/86 (16.3) 0.848 LCX 10/82 (12.2) 13/79 (16.5) 18/86 (20.9) 0.314 RCA 40/82 (48.8) 33/79 (41.8) 34/86 (39.5) 0.455 TIMI flow pre-PPCI 0–1 80/81 (98.8) 72/79 (91.1) 83/86 (96.5) 0.057 2 1/81 (1.2) 5/79 (6.3) 1/86 (1.2) 0.078 3 0/81 (0.0) 2/79 (2.5) 2/86 (2.3) 0.367 Thrombus score 4–5 75/81 (92.6) 72/79 (91.1) 81/86 (94.2) 0.754 Values are mean ± SD or n (%).

(16.5) 18/86 (20.9) 0.314 RCA 40/82 (48.8) 33/79 (41.8) 34/86 (39.5) 0.455 TIMI flow pre-PPCI 0–1 80/81 (98.8) 72/79 (91.1) 83/86 (96.5) 0.057 2 1/81 (1.2) 5/79 (6.3) 1/86 (1.2) 0.078 3 0/81 (0.0) 2/79 (2.5) 2/86 (2.3) 0.367 Thrombus score 4–5 75/81 (92.6) 72/79 (91.1) 81/86 (94.2) 0.754 Values are mean ± SD or n (%). ACE-inhibitors/A2RB, angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers; BMI, body mass index; CABG, coronary artery bypass graft; CAD, coronary artery disease; Cr, creatinine; DBP, diastolic blood pressure; HR, heart rate; LAD, left anterior descending coronary artery; LCX, left circumflex coronary artery; PCI, percutaneous coronary intervention; RCA, right coronary artery; SBP, systolic blood pressure; TA, thrombus aspiration; TIMI, thrombolysis in myocardial infarction.

se; Cr, creatinine; DBP, diastolic blood pressure; HR, heart rate; LAD, left anterior descending coronary artery; LCX, left circumflex coronary artery; PCI, percutaneous coronary intervention; RCA, right coronary artery; SBP, systolic blood pressure; TA, thrombus aspiration; TIMI, thrombolysis in myocardial infarction. Angiography and primary percutaneous coronary intervention procedure details Radial access, thrombectomy, and drug-eluting stent use were uniformly high (Table 2). Intra-procedural complications were similar across all groups. However, the incidence of transient atrio-ventricular (AV) block not requiring pacing was greater in the control arm. There was a low incidence of AV block requiring pacing with the study drugs (adenosine 2.4% vs. SNP 1.3% vs. 0% in control arm). The rate of transient hypotension (not requiring vasopressor or intra-aortic balloon-pump) was almost three-fold greater (16.5 vs. 5.8%, P = 0.028) in the SNP group compared with control. Table 2 Procedural (angiographic, electrocardiographic, and enzymatic) data and intra-procedural complications according to treatment group

e of transient hypotension (not requiring vasopressor or intra-aortic balloon-pump) was almost three-fold greater (16.5 vs. 5.8%, P = 0.028) in the SNP group compared with control. Table 2 Procedural (angiographic, electrocardiographic, and enzymatic) data and intra-procedural complications according to treatment group Characteristics Adenosine (n = 82) SNP (n = 79) Control (n = 86) P-valuea P-valueb Procedural data Radial approach 70 (85.4) 70 (88.6) 79 (91.9) 0.184 0.481 Thrombectomy 81 (98.8) 75 (98.7) 80 (93.0) 0.118 0.122 DES implantation 73 (89.0) 72 (91.1) 81 (94.2) 0.226 0.452 Number of stents 1.0 (1.0–2.0) 1.0 (1.0–2.0) 1.0 (1.0–2.0) 0.613 0.790 Diameter of stented segment (mm) 3.5 (3.0–3.5) 3.5 (3.0–3.5) 3.0 (3.0–3.5) 0.465 0.649 Length of stented segment (mm) 26.0 (18.0–39.0) 23.0 (18.0–38.0) 24.0 (18.0–34.5) 0.585 0.833 Electrocardiographic Baseline maximal sum of ST-segment elevation 9.0 (6.0–12.0) 9.0 (5.0–13.0) 9.0 (5.0–13.0) 0.931 0.686 Post-PCI maximal sum of ST-segment elevation 2.0 (0.0–5.0) 2.0 (0.0–5.0) 2.0 (0.0–4.0) 0.684 0.472 STR >70% 56 (68.3) 48 (60.8) 56 (65.1) 0.662 0.562 Enzymatic Peak CK (mg/dL) 1559 (601–2804) 1171 (430–2259) 1336 (511–2632) 0.601 0.393 Intra-procedural complications Transient AV block not requiring pacing 7 (8.5) 2 (2.5) 10 (11.6) 0.507 0.034 AV block requiring pacing 2 (2.4) 1 (1.3) 0 (0.0) 0.237 0.479 Transient hypotension not requiring vasopressor drugs or IABP 5 (6.1) 13 (16.5) 5 (5.8) 0.938 0.028 Hypotension requiring vasopressor drugs or IABP 5 (6.1) 3 (3.8) 6 (7.0) 0.818 0.499 Ventricular tachycardia/fibrillation 5 (6.1) 3 (3.8) 5 (5.8) 0.938 0.722 Values are mean ± SD, median (interquartile range), or n (%).

9 Transient hypotension not requiring vasopressor drugs or IABP 5 (6.1) 13 (16.5) 5 (5.8) 0.938 0.028 Hypotension requiring vasopressor drugs or IABP 5 (6.1) 3 (3.8) 6 (7.0) 0.818 0.499 Ventricular tachycardia/fibrillation 5 (6.1) 3 (3.8) 5 (5.8) 0.938 0.722 Values are mean ± SD, median (interquartile range), or n (%). AV, atrio-ventricular; CK, creatine kinase; DES, drug-eluting stent; IABP, intra-aortic balloon-pump; STR, ST-segment resolution; TMPG, tissue myocardial perfusion grade. aAdenosine vs. Control. bSNP vs. Control. Drug delivery Both doses of investigational drugs were administered in 66/82 (80%) and 53/79 (67%) patients (adenosine and SNP, respectively). The commonest reasons for withholding the second dose were QTc prolongation in 30 (adenosine: 9; SNP: 21) and hypotension in 3 (adenosine: 2; SNP: 1) patients. Failure to cross the lesion or deploy a stent and prohibiting intra-procedural complications accounted for the remainder. Angiographic, electrocardiogram, and enzymatic assessment of myocardial injury There was no difference in the occurrence of post-PPCI TIMI flow grade <3 between groups (Table 2). Likewise ECG (STR>70%) assessment of microvascular tissue perfusion was similar across the groups. There was no statistically significant difference between groups in peak creatine kinase levels.

ent of myocardial injury There was no difference in the occurrence of post-PPCI TIMI flow grade <3 between groups (Table 2). Likewise ECG (STR>70%) assessment of microvascular tissue perfusion was similar across the groups. There was no statistically significant difference between groups in peak creatine kinase levels. Cardiac magnetic resonance assessment of myocardial injury: primary end point There was no statistically significant difference in the primary outcome measure of unadjusted infarct size between groups (Table 3). On multivariable regression analysis, adjusting for significant confounders, mean infarct size was increased in the adenosine group (mean difference 2.73, 95% CI: −0.18 to 5.64, P = 0.066) when compared with controls. This was not seen in the SNP group. Table 3 Cardiac magnetic resonance data according to treatment group

multivariable regression analysis, adjusting for significant confounders, mean infarct size was increased in the adenosine group (mean difference 2.73, 95% CI: −0.18 to 5.64, P = 0.066) when compared with controls. This was not seen in the SNP group. Table 3 Cardiac magnetic resonance data according to treatment group Characteristic Adenosine, n = 63 SNP, n = 69 Control, n = 65 P-valuea P-valueb P-valuec Time from MI to CMR (h) 49.0 (28.4–75.0) 49.7 (26.2–76.1) 49.0 (38.0–74.8) 0.773 0.843 0.881 Primary endpointd Infarct size (%LVM) 10.1 (4.7–16.2) 10.0 (4.2–15.8) 8.3 (1.9–14.0) 0.062 0.160 0.133 Microvascular injury Presence of E-MVO (n, %) 41/60 (68.3) 42/59 (71.2) 38/63 (60.3) 0.452 0.254 0.416 E-MVO (%LVM) 1.2 (0.0–5.2), n = 60/82 1.0 (0.0–5.0), n = 59/79 1.4 (0.0–4.3), n = 63/86 0.637 0.770 0.891 Presence of L-MVO (n, %) 43/63 (68.3) 52/69 (75.4) 37/65 (56.9) 0.205 0.029 0.074 L-MVO (%LVM) 1.0 (0.0–3.7) 0.6 (0.0–2.4) 0.3 (0.0–2.8) 0.205 0.244 0.368 Presence of IMH (n, %) 20/38 (52.6) 19/43 (44.2) 16/38 (42.1) 0.358 0.850 0.619 n = 34 n = 38 n = 37 Salvage AAR (%LVM) 30.6 ± 12.2 34.7 ± 14.4 30.3 ± 11.5 0.907 0.152 0.266 MSI (%) 60.2 ± 23.3 63.6 ± 24.7 67.5 ± 23.3 0.188 0.477 0.429 n = 63 n = 71 n = 68 Function and volumes LVEDVI (mL/m2) 91.3 ± 16.1 87.0 ± 16.9 84.4 ± 14.6 0.011 0.336 0.044 LVESVI (mL/m2) 52.3 ± 13.8 49.1 ± 12.7 46.1 ± 11.6 0.006 0.155 0.023 LVMI (g/m2) 59.1 ± 11.5 56.0 ± 11.2 53.6 ± 9.2 0.003 0.174 0.014 EF (%) 43.2 ± 7.9 43.9 ± 6.5 45.7 ± 8.0 0.080 0.165 0.155 Values are mean ± SD or median (interquartile range) unless otherwise stated.

3 ± 16.1 87.0 ± 16.9 84.4 ± 14.6 0.011 0.336 0.044 LVESVI (mL/m2) 52.3 ± 13.8 49.1 ± 12.7 46.1 ± 11.6 0.006 0.155 0.023 LVMI (g/m2) 59.1 ± 11.5 56.0 ± 11.2 53.6 ± 9.2 0.003 0.174 0.014 EF (%) 43.2 ± 7.9 43.9 ± 6.5 45.7 ± 8.0 0.080 0.165 0.155 Values are mean ± SD or median (interquartile range) unless otherwise stated. AAR, area at risk; BSA, body surface area; EF, ejection fraction; IMH, intra-myocardial haemorrhage; LV, left ventricular; LVEDVI, LV end-diastolic volume indexed to BSA; LVESVI, LV end-systolic volume indexed to BSA; LVM, LV mass; LVMI, LVM indexed to BSA; MI, myocardial infarction; MSI, myocardial salvage index; E- or L-MVO, early- or late-microvascular obstruction. aAdenosine vs. Control. bSNP vs. Control. cAll groups. dPrimary endpoint—comparison via independent t-test on log-transformed scale. Late-MVO was present in 67% patients and was significantly more prevalent in the SNP arm vs. control (75.4 vs. 56.9%, P = 0.029). However, the extent of L-MVO was small and not significantly different between any groups. Other CMR parameters of microvascular injury (E-MVO and IMH) were also similar between groups and none of the potential confounders were identified as being of statistical importance by the forward selection procedure. Diagnostic quality T2-weighted (oedema) imaging was only obtainable in 109 patients (55%). There were no significant differences in AAR or MSI between groups.

MH) were also similar between groups and none of the potential confounders were identified as being of statistical importance by the forward selection procedure. Diagnostic quality T2-weighted (oedema) imaging was only obtainable in 109 patients (55%). There were no significant differences in AAR or MSI between groups. Left ventricular end-diastolic and end-systolic volumes indexed to BSA (LVEDVI, LVESVI) were significantly increased with adenosine compared with control and this was accompanied by a non-significant reduction in ejection fraction (EF). Left ventricular volumes and function were similar in the SNP and standard PPCI arms (Table 3). Clinical outcomes Two hundred and thirty-three (94%) patients completed follow-up (12 patients withdrew consent/refused follow-up and 2 patients were lost to follow-up). There was a significant increase in MACE at 6 months (15.6 vs. 2.5%, P = 0.004) in the adenosine-treated group compared with control (see Figure 3) driven by incidence of heart failure (10.4 vs. 1.2%, P = 0.016); HR 6.53 (95% CIs 1.46–29.2), P = 0.01. When adjusted for potential confounders, the observed MACE difference remained with similar hazard ratios with P-values: P = 0.018 at 30 days and P = 0.01 at 6 months (see Supplementary material online, Table S9). There was no statistically significant difference in bleeding between groups (see Table 4). Table 4 Clinical events (first event) to 6 months according to treatment group

rence remained with similar hazard ratios with P-values: P = 0.018 at 30 days and P = 0.01 at 6 months (see Supplementary material online, Table S9). There was no statistically significant difference in bleeding between groups (see Table 4). Table 4 Clinical events (first event) to 6 months according to treatment group Characteristics Adenosine (all subjects) (n = 82) SNP (all subjects) (n = 79) Control (all subjects) (n = 86) P-valuea P-valueb MACE 12 (15.0, 7.0–22.0)c 5 (6.0, 1.0–15.0)c 2 (2.0, 0.0–5.0)c 0.004 0.261 Death 1 (1.2) 1 (1.3) 0 (0.0) 0.488 0.479 TIA/stroke 1 (1.2) 1 (1.3) 0 (0.0) 0.488 0.479 MI 2 (2.4) 1 (1.3) 1 (1.2) 0.614 1.000 HF 8 (9.8) 2 (2.5) 1 (1.2) 0.016 0.607 TLR 0 (0.0) 0 (0.0) 0 (0.0) 1.000 1.000 Composite of death, MI, and HF 11 (13.4) 4 (5.1) 2 (2.3) 0.009 0.428 No. of patients with >1 event 3 (3.7) 1 (1.3) 1 (1.2) 0.359 1.000 Bleeding All bleeding 4 (4.9) 2 (2.5) 5 (5.8) 1.000 0.446 Fatal bleeding 0 (0.0) 0 (0.0) 0 (0.0) 1.000 1.000 HF, heart failure; MACE, major adverse cardiac events; MI, myocardial infarction; TIA, transient ischaemic attack; TIMI, thrombolysis in MI; TLR, target lesion revascularization. aAdenosine vs. Control. bSNP vs. Control. cValues are n (%), except for MACE (pre-defined secondary outcome) for which %s are actuarial. Figure 3 Kaplan–Meier graphs showing clinical outcome. Clinical outcomes at 30 days (top) and 6 months (bottom). HR, hazard ratio; ITT, intention-to-treat; MACE, major adverse cardiac events; PH, proportional hazards.

bSNP vs. Control. cValues are n (%), except for MACE (pre-defined secondary outcome) for which %s are actuarial. Figure 3 Kaplan–Meier graphs showing clinical outcome. Clinical outcomes at 30 days (top) and 6 months (bottom). HR, hazard ratio; ITT, intention-to-treat; MACE, major adverse cardiac events; PH, proportional hazards. Per-protocol analysis In patients who actually received both doses of adenosine there was, when compared with controls, a significant increase in infarct size (%LVM, 12.0 vs. 8.3, P = 0.031) and LV volumes (LVEDVI [mL/m2], 91.4 ± 14.1 vs. 84.4 ± 14.6, P = 0.009) and EF (%) was reduced (42.5 ± 7.2 vs. 45.7 ± 8.0, P = 0.027). Major adverse cardiac events was increased in per-protocol adenosine-treated patients compared with control at 30 days (HR 5.91 [95% CIs 1.28–27.25], P = 0.036) and 6 months (HR 7.31 [95% CIs 1.62–33.0], P = 0.008). There was no increase in MACE in those who actually received SNP compared with controls (see Supplementary material online, Table S9 and Figure S1). Sub-group analyses In patients with anterior STEMI, those who received adenosine had significantly greater IS %LVM (16.2, 9.5–25.8 vs. 9.1, 1.8–13.4, P = 0.028) and extent of L-MVO %LVM (3.7, 0.5–6.1 vs. 0.3, 0.0–2.0, P = 0.008) compared with controls (see Supplementary material online, Table S8). Furthermore, the composite secondary clinical endpoint of death, MI, and heart failure was significantly greater in these patients (21.2 vs. 3.1%, P = 0.025) compared with controls.

and extent of L-MVO %LVM (3.7, 0.5–6.1 vs. 0.3, 0.0–2.0, P = 0.008) compared with controls (see Supplementary material online, Table S8). Furthermore, the composite secondary clinical endpoint of death, MI, and heart failure was significantly greater in these patients (21.2 vs. 3.1%, P = 0.025) compared with controls. Discussion Sub-optimal microvascular perfusion in STEMI confers additional morbidity and mortality despite TIMI 3 flow in the epicardial vessel.25 Adenosine and SNP are powerful vasodilators with pleiotropic effects, including anti-inflammatory, anti-platelet, and immune-modulatory actions, that have been shown to reduce IRI, particularly when delivered IC targeting the microvascular bed.16 However, our results show that neither adenosine nor SNP significantly reduced infarct size or MVO. Furthermore, IC adenosine appeared to be associated with detrimental cardiac effects and worse clinical outcome.

ions, that have been shown to reduce IRI, particularly when delivered IC targeting the microvascular bed.16 However, our results show that neither adenosine nor SNP significantly reduced infarct size or MVO. Furthermore, IC adenosine appeared to be associated with detrimental cardiac effects and worse clinical outcome. Infarct size and markers of reperfusion injury There was no significant difference in infarct size between the groups, although there was a trend to significant increase in infarct size in the adenosine group compared with controls when the results were adjusted for confounders. There was no reduction in MSI, early- or late-MVO measured by CMR with either drug. These results are consistent with the only other randomized trial using CMR, which found that high-dose IC adenosine did not increase myocardial salvage and did not reduce either infarct size or MVO.9 The per-protocol analysis showed that patients who received both doses of adenosine had significantly increased infarct size, LV volumes, and reduced EF on CMR compared with the control group, which suggests not only a lack of efficacy but potential cardiac toxicity and partial negation of any beneficial effects of reperfusion. This effect was particularly apparent in patients with anterior MI. Reasons for these adverse findings remain unclear but are considered below.

on CMR compared with the control group, which suggests not only a lack of efficacy but potential cardiac toxicity and partial negation of any beneficial effects of reperfusion. This effect was particularly apparent in patients with anterior MI. Reasons for these adverse findings remain unclear but are considered below. The proportion of patients achieving STR >70%, as an ECG surrogate marker of coronary vascular flow, was relatively high and similar across groups compared with previous studies.8–11,17 This may reflect our recruitment of only patients who presented within 6 h rather than within 12 h of symptom onset as in some previous studies.7–9,17 In REOPEN-AMI,8 STR >70% occurred in 71% of patients treated by adenosine, in 54% of patients treated by SNP, and in 51% of patients treated by saline (risk ratio: 1.39 [1.07–1.79]; P = 0.009, and risk ratio: 1.04 [0.78–1.40]; P = 0.75, adenosine or SNP vs. saline, respectively). Furthermore, in REOPEN-AMI, the enzymatic infarct size was 30% lower in the adenosine compared with the saline group but this is a greater effect size than would be expected by the attenuation of MVO alone, which makes interpretation difficult. However, in the absence of corroborative AAR or infarct size data from robust imaging such as CMR, this may imply that the adenosine group were destined to have smaller infarcts compared with control (saline), potentially over-estimating the benefit of adenosine in that study.

one, which makes interpretation difficult. However, in the absence of corroborative AAR or infarct size data from robust imaging such as CMR, this may imply that the adenosine group were destined to have smaller infarcts compared with control (saline), potentially over-estimating the benefit of adenosine in that study. In our REFLO-STEMI trial, a positive impact of IC SNP on MVO assessed by ECG or CMR was not seen. This is consistent with the REOPEN-AMI study8 and an earlier double-blind, placebo-controlled RCT.7 The demonstration in our study, and in those of others,3,26 of failure of reduction in MVO or infarct size using adjunctive therapies, despite an extensive body of positive experimental data, raises the question of trial design or even whether IRI is a distinct entity clinically and whether it represents exclusively ischaemic injury.26,27

our study, and in those of others,3,26 of failure of reduction in MVO or infarct size using adjunctive therapies, despite an extensive body of positive experimental data, raises the question of trial design or even whether IRI is a distinct entity clinically and whether it represents exclusively ischaemic injury.26,27 Clinical outcome There were significantly worse outcomes for the adenosine group compared with control, largely driven by increased early heart failure events. This is a novel finding. Whilst this may have occurred by chance, the hazard ratio is high and cannot be discounted. Our results, including definitive and significantly worse clinical outcome particularly in those with anterior STEMI receiving adenosine, contradicts the post hoc analysis of AMISTAD-II, which reported that a 3 h adenosine infusion decreased 1-month (5.2 vs. 9.2%, respectively, P = 0.014) and 6-month mortality (7.3 vs. 11.2%, P = 0.033) compared with placebo, and reduced the composite clinical endpoint of death or heart failure at 6 months (12.0 vs. 17.2%, P = 0.022).28 However, there were several methodological weaknesses with that study including: (i) adenosine was infused intravenously after the PCI, (ii) infarct size was measured in only 11% of patients and by technetium-99m sestamibi single-photon emission computed tomography, which may underestimate infarct size compared with CMR; and (iii) no measure of myocardial salvage was obtained. Furthermore, the adverse MACE signal seen with adenosine in our study is consistent whether assessed as intention-to-treat or per-protocol analysis, at 1 and 6 months follow-up and remains despite adjustment for potential confounding variables. This suggests that high-dose IC adenosine, delivered as we have indicated in this study, may lead to significantly worse clinical outcomes.

study is consistent whether assessed as intention-to-treat or per-protocol analysis, at 1 and 6 months follow-up and remains despite adjustment for potential confounding variables. This suggests that high-dose IC adenosine, delivered as we have indicated in this study, may lead to significantly worse clinical outcomes. The finding of an adverse effect of high-dose adenosine remains difficult to explain. It could be that high-dose adenosine tends to activate other receptors, which lead to adverse events; adenosine has been speculated to be responsible for excess dyspnoea observed with ticagrelor in the PLATO study. However, any effect of ticagrelor on circulating adenosine levels is likely to be small compared with the relatively high doses of adenosine administered in our study. Furthermore, the distribution of ticagrelor and prasugrel across the treatment arms was comparable (see Table 1) and there were no significant differences in the primary endpoint or MVO for the patients receiving ticagrelor (see Supplementary material online, Table S8). Additionally, adenosine may mediate diuretic resistance; it acts on renal A1-AR on afferent arterioles to reduce glomerular flow and filtration, stimulate renin release, and enhance proximal tubular sodium reabsorption.29 It could be postulated that cross-activation of renal AR promoted fluid retention, increased LV volumes and advanced heart failure pathophysiology in our adenosine-treated cohort. Even though our finding is hypothesis generating, our data strongly suggest high-dose IC adenosine may lead to adverse events (possibly through cross-activation of other receptors) and probably should not be used to prevent MVO, in the circumstances of this trial.

ailure pathophysiology in our adenosine-treated cohort. Even though our finding is hypothesis generating, our data strongly suggest high-dose IC adenosine may lead to adverse events (possibly through cross-activation of other receptors) and probably should not be used to prevent MVO, in the circumstances of this trial. Limitations The study was open-label, which may have influenced management of patients. However, the primary outcome was assessed on blinded CMR scans. The finding of an increased hazard signal with use of high-dose IC adenosine in our study must be interpreted with caution given our relatively small sample size. The power of the study may have been reduced due to the use of the FWHM technique, which has better observer variability but results in lower estimation of infarct size compared with previous studies using more conservative thresholds. Whilst we did not formally adjust for multiple hypothesis tests, which might not be considered ideal, this was a Phase II trial, and we did not want to exclude identifying potential efficacy of either pharmacological intervention. However, even if we had used a more stringent level of statistical significance, to account for testing both of our two active treatments relative to our control group, the statistical significance of our main results would remain unaltered. Similarly, although there was some evidence against the assumption of proportional hazards for the primary clinical outcome analysis (MACE by 6 months) this was not overwhelming, and our results were further supported by actuarial estimates of the absolute MACE event rate by 6 months.

in results would remain unaltered. Similarly, although there was some evidence against the assumption of proportional hazards for the primary clinical outcome analysis (MACE by 6 months) this was not overwhelming, and our results were further supported by actuarial estimates of the absolute MACE event rate by 6 months. It is possible that the dose of adenosine used in our study was insufficient to achieve sustained activation of A2-AR; 30 We utilized a slow-bolus dosing regimen and it could be that, owing to the short-half life of adenosine, concentrations of adenosine were insufficient in the microvascular bed to augment the benefits of reperfusion. A1-AR activation at lower adenosine doses may exacerbate MVO by promoting neutrophil chemotaxis.30 As such, delivering adenosine as a continuous infusion, to ensure more prolonged availability of adenosine in the microvasculature, may be superior to boli injections in antagonizing MVO-related processes. It should be noted that one might still consider low-dose (50–100 μg) adenosine, as is common in clinical practice, to reverse established slow-flow or no-reflow when it occurs. Conclusions Intra-coronary adenosine and SNP did not reduce infarct size nor MVO during PPCI for STEMI. Furthermore, high-dose adenosine appeared to be associated with adverse clinical outcomes, increased infarct size, and reduced EF compared with control. These data suggest that neither agent is effective for sub-clinical no-reflow and should not be used routinely and prophylactically in the setting of PPCI to prevent reperfusion injury.

dose adenosine appeared to be associated with adverse clinical outcomes, increased infarct size, and reduced EF compared with control. These data suggest that neither agent is effective for sub-clinical no-reflow and should not be used routinely and prophylactically in the setting of PPCI to prevent reperfusion injury. Supplementary material Supplementary material is available at European Heart Journal online. Authors’ contributions S.N., K.A. performed statistical analysis; A.G., G.M., J.G., D.B., I.M., L.S., K.A., S.N. handled funding and supervision; S.N., J.K., G.M., J.G., D.B., V.K., M.B., A.G. acquired the data; A.G., G.M., J. G., D.B., I.M., L.S., S.N. conceived and designed the research; S.N., A.G. drafted the manuscript; G.M., J.G., D.B., V.K., J.K., I.M., M.B., R.W., A.A.J.A. made critical revision of the manuscript for key intellectual content. Funding REFLO-STEMI was funded by the Medical Research Council (MRC) through the Efficacy and Mechanism Evaluation (EME) Board (project number 09/150/28) and managed by the NIHR on behalf of the MRC-NIHR partnership. Funding to pay the Open Access publication charges for this article was provided by The University of Leicester. Conflict of interest: none declared.

Funding REFLO-STEMI was funded by the Medical Research Council (MRC) through the Efficacy and Mechanism Evaluation (EME) Board (project number 09/150/28) and managed by the NIHR on behalf of the MRC-NIHR partnership. Funding to pay the Open Access publication charges for this article was provided by The University of Leicester. Conflict of interest: none declared. Acknowledgements The study was sponsored by the University Hospitals of Leicester NHS Trust. We also acknowledge the contribution of the following as TSC members: Dr Peter Ludman, Prof Jim Nolan, Mr Gerry Thompson and Dr David Hetmanski, and DSMB members: Dr Ian BA Menown, Dr Mazhar Khan and Mr Cathal Walsh. We thank the Newcastle Angiographic Core Lab (Ross Fowkes, Jin Howe Tee, and Murugapathy Veerasamy) for undertaking angiographic analyses. The support nurses and particularly the patients warrant our gratitude. G.P.M. is funded by a National Institute for Health (NIHR) Research Fellowship.

Introduction Given the high prevalence of coronary artery disease (CAD) and associated mortality, prevention of fatal and non-fatal myocardial infarctions (MI) in CAD patients remains an ongoing clinical challenge. Mortality rates among stable CAD patients range between 1% and 3%, while rates of non-fatal events are 1–2% annually.1 In patients with acute coronary syndromes (ACS) who survive the acute event, the rate of MI and death is markedly higher, particularly during the first year.2 However, at the individual level, the event risk may vary considerably, which makes risk estimation tools necessary to improve patient management. Expedient risk stratification should identify individuals at risk requiring more intensive therapy. Conversely, patients with a favorable prognosis should be identified to avoid drug overuse and associated side effects.3

k may vary considerably, which makes risk estimation tools necessary to improve patient management. Expedient risk stratification should identify individuals at risk requiring more intensive therapy. Conversely, patients with a favorable prognosis should be identified to avoid drug overuse and associated side effects.3 Hypothesis free lipidomic analyses have revealed a handful of lipids potentially qualifying as useful prognostic markers for CAD.4–6 In our initial lipidomic study, distinct ceramide species were significantly associated with CVD among CAD patients.4 Molecular lipid species, particularly ceramide(d18:1/16:0), were also associated with necrotic core tissue type and lipid core burden in coronary angiography, and were predictive for 1-year clinical outcome in 581 ACS and stable CAD patients.7 In these studies, plasma CVD risk-related ceramide molecules (Cer(d18:1/16:0), Cer(d18:1/18:0), and Cer(d18:1/24:1)), and their ratios with Cer(d18:1/24:0), emerged as potential risk stratifiers for CAD patients.4 Ceramides are known to associate with many central processes of atherosclerosis development including lipoprotein uptake, inflammation, and apoptosis (Supplementary material online, Figure S1).8 Ceramide species are produced by six fatty acyl selective ceramide synthases (CerSs; Supplementary material online, Figure S2), and it is becoming evident that individual ceramide species have specific physiological functions.9–12 Thus, monitoring ratios of ceramides species may provide insight into the metabolic regulation of atherosclerotic events. In this study, we establish the suggested role of ceramides and their distinct ratios as risk predictors for CV death in patients with stable CAD and ACS.

have specific physiological functions.9–12 Thus, monitoring ratios of ceramides species may provide insight into the metabolic regulation of atherosclerotic events. In this study, we establish the suggested role of ceramides and their distinct ratios as risk predictors for CV death in patients with stable CAD and ACS. Methods More detailed method descriptions are available in Supplementary material online. Study subjects Corogene study: stable coronary artery disease patients Corogene is a prospective, consecutive cohort study of Finnish patients referred for coronary angiography to the Helsinki University Central Hospital between 2006 and 2008. A nested case control study was designed using the Corogene database and including data from the national death certificate registry. As cases, all patients who experienced coronary death within an average follow-up of 2½ years were selected. Matched control patients had established CAD (>50% stenosis at least in one epicardial coronary artery), but remained alive during the follow-up period. Baseline characteristics of the Corogene subjects are shown in Table 1 and Supplementary material online, Table S1. Table 1 Baseline characteristics of the subjects in Corogene, SPUM-ACS and BECAC studies

hed CAD (>50% stenosis at least in one epicardial coronary artery), but remained alive during the follow-up period. Baseline characteristics of the Corogene subjects are shown in Table 1 and Supplementary material online, Table S1. Table 1 Baseline characteristics of the subjects in Corogene, SPUM-ACS and BECAC studies Characteristic COROGENE SPUM-ACS BECAC Cases Controls Cases Controls Cases Controls No of subjects 80 80 51 1586 81 1506 Gender Male, n (%) 60 (75%) 60 (75%) 42 (82%) 1223 (77%) 55 (68%) 889 (59%) Age (years) 70.2 (62.6–77.1) 70 (63.4–76.9) 77.1 (69–83) 62.8 (53.9–72.9) 71 (64–78) 61 (54–70) Body mass index 27.5 (23.6–30.9) 26 (24.1–29.5) 25.1 (23.1–28.3) 26.6 (24.3–29.4) 24 (22–28) 26 (23–28) Time to death/follow-up time (days) 528 (134–739) 1955 (1669–2173) 25 (7–216) 365 (359–365) 626 (196–1332) 1720 (1368–2111) Creatinine (µmol/L) 98 (84–134) 79 (69–90) 100 (79–134) 75 (65–88) 98 (87–115) 87 (79–97) Current smoker Yes, n (%) 37 (46%) 37 (46%) 16 (31%) 663 (42%) 29 (36%) 355 (24%) No, n (%) 43 (54%) 43 (54%) 33 (65%) 897 (57%) 50 (62%) 1146 (76%) NA 2 (4%) 26 (2%) 2 (2%) 5 (0%) Diabetes Yes, n (%) 32 (40%) 32 (40%) 11 (22%) 266 (17%) 16 (20%) 159 (11%) No, n (%) 48 (60%) 48 (60%) 40 (78%) 1320 (83%) 64 (79%) 1333 (89%) NA 1 (1%) 14 (1%) Hypertension Yes, n (%) 60 (75%) 60 (75%) 36 (71%) 912 (58%) 55 (68%) 674 (45%) No, n (%) 20 (25%) 20 (25%) 15 (29%) 674 (42%) 26 (32%) 832 (55%) Lipid-lowering treatment Yes, n (%) 61 (76%) 61 (76%) 14 (27%) 432 (27%) 61 (75%) 933 (62%) No, n (%) 19 (24%) 19 (24%) 34 (67%) 1146 (72%) 20 (25%) 573 (38%) NA 3 (6%) Previous AMI Yes, n (%) 67 (84%) 0 (0%) 8 (16%) 210 (13%) 54 (67%) 482 (32%) No, n (%) 13 (16%) 80 (100%) 43 (84%) 1374 (87%) 27 (33%) 1024 (68%) NA 2 (0%) Previous stroke Yes, n (%) 17 (21%) 10 (12%) 2 (4%) 37 (2%) 15 (19%) 108 (7%) No, n (%) 63 (79%) 70 (88%) 49 (96%) 1549 (98%) 66 (81%) 1398 (93%)

73 (38%) NA 3 (6%) Previous AMI Yes, n (%) 67 (84%) 0 (0%) 8 (16%) 210 (13%) 54 (67%) 482 (32%) No, n (%) 13 (16%) 80 (100%) 43 (84%) 1374 (87%) 27 (33%) 1024 (68%) NA 2 (0%) Previous stroke Yes, n (%) 17 (21%) 10 (12%) 2 (4%) 37 (2%) 15 (19%) 108 (7%) No, n (%) 63 (79%) 70 (88%) 49 (96%) 1549 (98%) 66 (81%) 1398 (93%) Bergen Coronary Angiography Cohort cohort: patients with stable coronary artery disease The Bergen Coronary Angiography Cohort (BECAC) includes 1580 adults referred to elective coronary angiography because of suspected stable angina pectoris recruited at the Haukeland University Hospital in Bergen, Norway between 2000 and 2004. Information on cardiovascular deaths was collected from the Cause of Death Registry at the Norwegian Institute of Public Health, and verified against hospital medical records whenever available. During a median follow-up of 4.6 years, a total of 81 patients died from cardiovascular disease. Baseline characteristics of the BECAC participants are reported in Table 1 and Supplementary material online, Table S1.

Norwegian Institute of Public Health, and verified against hospital medical records whenever available. During a median follow-up of 4.6 years, a total of 81 patients died from cardiovascular disease. Baseline characteristics of the BECAC participants are reported in Table 1 and Supplementary material online, Table S1. SPUM-ACS cohort: patients with acute coronary syndromes Special Program University Medicine—Inflammation in Acute Coronary Syndromes (SPUM-ACS) is a prospective, multi-centre (Bern, Geneva, Lausanne, and Zürich) cohort study. Patients with a primary diagnosis of ACS and referred for invasive management were enrolled at four Swiss university hospitals. Baseline characteristics of the SPUM-ACS patients are summarized in Table 1 and Supplementary material online, Table S1. At one-year follow-up, a total of 51 patients died from cardiac reasons. Clinical laboratory analyses Standard lipids measurements were determined using standard methods available at each of the three study sites. In Corogene subjects, apolipoproteins (AI, AII, and B), lipoprotein (a), lipoprotein-associated phospholipase A2 activity, and HDL and LDL particle numbers and sizes were measured as described in Supplementary material online, Methods. Quantification of ceramides The plasma levels of Cer(d18:1/16:0), Cer(d18:1/18:0), Cer(d18:1/24:0), and Cer(d18:1/24:1) were quantified on a 5500 QTRAP (SCIEX, Framingham, MA) mass spectrometer equipped with an Eksigent 100-XL UHPLC system as described recently.13

Clinical laboratory analyses Standard lipids measurements were determined using standard methods available at each of the three study sites. In Corogene subjects, apolipoproteins (AI, AII, and B), lipoprotein (a), lipoprotein-associated phospholipase A2 activity, and HDL and LDL particle numbers and sizes were measured as described in Supplementary material online, Methods. Quantification of ceramides The plasma levels of Cer(d18:1/16:0), Cer(d18:1/18:0), Cer(d18:1/24:0), and Cer(d18:1/24:1) were quantified on a 5500 QTRAP (SCIEX, Framingham, MA) mass spectrometer equipped with an Eksigent 100-XL UHPLC system as described recently.13 Statistical analyses Wilcoxon's rank sum test was applied for group comparisons. Odds ratios (ORs) per standard deviation were estimated using logistic regression. Hazard ratios were calculated using the Cox proportional hazard model. The GRACE14 risk score, consisting of Killip class, systolic blood pressure, heart rate, age, creatinine, cardiac arrest at admission, ST-segment deviation, and elevated cardiac enzyme levels (troponin, CK-MB), was used to calculate the risk of long-term mortality for ACS patients. The following Marschner score15 variables were used in the modeling of stable CAD patient data: total cholesterol, HDL-C, age, gender, smoking status, previous acute MI, diabetes, hypertension, and prior stroke.

iac enzyme levels (troponin, CK-MB), was used to calculate the risk of long-term mortality for ACS patients. The following Marschner score15 variables were used in the modeling of stable CAD patient data: total cholesterol, HDL-C, age, gender, smoking status, previous acute MI, diabetes, hypertension, and prior stroke. Net reclassification improvement was estimated as described by Pencina et al.16 For the 1-year event risk of the secondary prevention population in the SPUM-ACS study we categorized subjects to low risk (<1% event probability), intermediate (1–5%) risk or high-risk (>5%) groups. For the BECAC study the same categorization was used for 3-year risk. The ceramide risk score was calculated as follows: For each individual, all three ceramide ratios and each concentration (apart from Cer(d18:1/24:0)) were compared with the whole study population. If the variable belonged to the 3rd quartile, the individual received +1 point, and if to the 4th quartile, +2 points (Supplementary material online, Table S11). Thus, the score ranges from 0 to 12 and based on the score, the subjects were split into four risk categories (0–2, 3–6, 7–9, and 10–12). More details on statistical methods can be found in Supplementary material online.

The ceramide risk score was calculated as follows: For each individual, all three ceramide ratios and each concentration (apart from Cer(d18:1/24:0)) were compared with the whole study population. If the variable belonged to the 3rd quartile, the individual received +1 point, and if to the 4th quartile, +2 points (Supplementary material online, Table S11). Thus, the score ranges from 0 to 12 and based on the score, the subjects were split into four risk categories (0–2, 3–6, 7–9, and 10–12). More details on statistical methods can be found in Supplementary material online. Results Ceramide concentrations in high- and low-risk patients with coronary artery disease In stable CAD patients of the Corogene study LDL-based markers such as LDL-C, LDL particle number (LDL-P), small dense LDL (sdLDL), and apoB did not differ significantly between cases who experienced coronary death and controls who remained alive, and neither did Lp(a) nor Lp-PLA2. However, the HDL-related markers HDL-C, HDL particle number (HDL-P), small dense HDL (sdHDL), and ApoA1 were all significantly (P < 0.001) different between the groups with the medians being −17.1, −14.3, −19.0, and −12.9% lower in cases, respectively. The differences between cases and controls in plasma ceramides and established lipid markers are provided in Table 2 (the percentage of observations for each marker is provided in Supplementary material online, Table S2). Table 2 Medians and inter quartile ranges of established lipid markers and ceramides in case and control groupsa

s between cases and controls in plasma ceramides and established lipid markers are provided in Table 2 (the percentage of observations for each marker is provided in Supplementary material online, Table S2). Table 2 Medians and inter quartile ranges of established lipid markers and ceramides in case and control groupsa BECAC SPUM-ACS Cases (n = 81) Controls (n = 1499) P-value Cases (n = 51) Controls (n = 1586) P-value Cer(d18:1/16:0)/Cer(d18:1/24:0) 0.121 (0.101–0.145) 0.100 (0.085–0.119) <0.001 0.116 (0.099–0.170) 0.093 (0.079–0.113) <0.001 Cer(d18:1/18:0)/Cer(d18:1/24:0) 0.046 (0.036–0.059) 0.038 (0.031–0.049) <0.001 0.064 (0.044–0.084) 0.047 (0.037–0.060) <0.001 Cer(d18:1/24:1)/Cer(d18:1/24:0) 0.498 (0.408–0.624) 0.413 (0.337–0.508) <0.001 0.489 (0.415–0.675) 0.394 (0.337–0.474) <0.001 Cer(d18:1/16:0) (µmol/L) 0.271 (0.235–0.326) 0.253 (0.213–0.300) 0.010 0.313 (0.255–0.385) 0.292 (0.247–0.346) 0.090 Cer(d18:1/18:0) (µmol/L) 0.108 (0.077–0.143) 0.096 (0.076–0.123) 0.097 0.161 (0.109–0.234) 0.146 (0.112–0.189) 0.163 Cer(d18:1/24:0) (µmol/L) 2.335 (1.843–2.866) 2.548 (2.030–3.098) 0.035 2.366 (2.112–3.084) 3.107 (2.490–3.826) <0.001 Cer(d18:1/24:1) (µmol/L) 1.056 (0.927–1.344) 1.028 (0.844–1.257) 0.026 1.421 (1.012–1.628) 1.229 (1.004–1.484) 0.175 LDL-C (mg/dL) 110 (89–133) 116 (93–147) 0.087 101 (81–128) 121 (93–150) 0.001 HDL-C (mg/dL) 46 (35–58) 50 (41–62) 0.036 48 (36–58) 44 (36–53) 0.266 TC (mg/dL) 185 (158–212) 193 (166–224) 0.081 159 (147–189) 189 (161–221) <0.001 TG (mg/dL) 135 (100–169) 126 (92–182) 0.738 76 (54–108) 92 (61–142) 0.014 COROGENE Cases (n = 80) Controls (n = 80) P-value Cer(d18:1/16:0)/Cer(d18:1/24:0) 0.132 (0.105–0.175) 0.105 (0.090–0.128) <0.001 Cer(d18:1/18:0)/Cer(d18:1/24:0) 0.062 (0.047–0.077) 0.046 (0.037–0.062) <0.001 Cer(d18:1/24:1)/Cer(d18:1/24:0) 0.703 (0.582–0.846) 0.556 (0.483–0.665) <0.001 Cer(d18:1/16:0) (µmol/L) 0.275 (0.222–0.326) 0.235 (0.212–0.282) 0.007 Cer(d18:1/18:0) (µmol/L) 0.118 (0.094–0.152) 0.107 (0.092–0.137) 0.195 Cer(d18:1/24:0) (µmol/L) 1.923 (1.475–2.511) 2.235 (1.993–2.672) 0.008 Cer(d18:1/24:1) (µmol/L) 1.385 (1.189–1.620) 1.245 (1.091–1.427) 0.017 TC (mg/dL) 128 (111–165) 139 (122–163) 0.064 TG (mg/dL) 108 (86–140) 92 (75–139) 0.110 LDL-C (mg/dL) 69 (55–99) 75 (65–92) 0.251 LDL-P (nmol/L) 830 (694–1110) 928 (712–1175) 0.395 sdLDL (nmol/L) 533 (304–659) 548 (376–737) 0.265 ApoB (mg/dL) 67 (55–82) 68.5 (57–84) 0.997 HDL-C (mg/dL) 34 (29–40) 41 (33–51) <0.001 HDL-P (µmol/L) 24 (21–27) 28 (24–31) <0.001 sdHDL (µmol/L) 12.8

(86–140) 92 (75–139) 0.110 LDL-C (mg/dL) 69 (55–99) 75 (65–92) 0.251 LDL-P (nmol/L) 830 (694–1110) 928 (712–1175) 0.395 sdLDL (nmol/L) 533 (304–659) 548 (376–737) 0.265 ApoB (mg/dL) 67 (55–82) 68.5 (57–84) 0.997 HDL-C (mg/dL) 34 (29–40) 41 (33–51) <0.001 HDL-P (µmol/L) 24 (21–27) 28 (24–31) <0.001 sdHDL (µmol/L) 12.8 (9.3–15.6) 15.8 (13.1–18.2) <0.001 ApoA1 (mg/dL) 115 (101–131) 132 (115–150) <0.001 Lp(a) (mg/dL) 7.2 (2–35) 3.6 (1–28) 0.319 Lp-PLA2 (nmol/min/ml) 138 (119–166) 130 (115–163) 0.354 C-reactive protein (mg/L) 3.1 (1.6–8.7) 1.1 (0.7–2.8) <0.001 aCer, ceramide; TC, total cholesterol; TG, triacylglycerols, LDL-C low-density lipoprotein cholesterol, HDL-C high-density lipoprotein cholesterol, sdLDL small dense low-density lipoprotein cholesterol, LDL-P low-density lipoprotein particle number, sdHDL small dense high-density lipoprotein cholesterol, HDL-P high-density lipoprotein particle number, ApoB apolipoprotein B, ApoA1 apolipoprotein A1, Lp(a) lipoprotein (a), Lp-PLA2 lipoprotein-associated phospholipase A2. SI conversion factors: To convert cholesterol to mmol/L, multiply values by 0.0259; to convert triacylglycerols to mmol/L, multiply values by 0.01129.

(9.3–15.6) 15.8 (13.1–18.2) <0.001 ApoA1 (mg/dL) 115 (101–131) 132 (115–150) <0.001 Lp(a) (mg/dL) 7.2 (2–35) 3.6 (1–28) 0.319 Lp-PLA2 (nmol/min/ml) 138 (119–166) 130 (115–163) 0.354 C-reactive protein (mg/L) 3.1 (1.6–8.7) 1.1 (0.7–2.8) <0.001 aCer, ceramide; TC, total cholesterol; TG, triacylglycerols, LDL-C low-density lipoprotein cholesterol, HDL-C high-density lipoprotein cholesterol, sdLDL small dense low-density lipoprotein cholesterol, LDL-P low-density lipoprotein particle number, sdHDL small dense high-density lipoprotein cholesterol, HDL-P high-density lipoprotein particle number, ApoB apolipoprotein B, ApoA1 apolipoprotein A1, Lp(a) lipoprotein (a), Lp-PLA2 lipoprotein-associated phospholipase A2. SI conversion factors: To convert cholesterol to mmol/L, multiply values by 0.0259; to convert triacylglycerols to mmol/L, multiply values by 0.01129. In the Corogene study, the concentrations of Cer(d18:1/16:0), Cer(d18:1/18:0), and Cer(d18:1/24:1) were significantly different (P < 0.001 for all) between cases who had a fatal MI during the follow-up period and controls (medians in cases +17.0%, +10.3, and +11.2% higher than in controls, respectively). In contrast, the Cer(d18:1/24:0) behaved differently, with the median in cases being −14.9% lower than in controls (P < 0.001). Similarly to our earlier observations,4 highly significant differences, were observed for the three predefined ceramide ratios, with the medians of cases ranging between +25.7 and +34.8% (P < 0.001) relative to controls. The difference between stable CAD patients and controls is illustrated in Supplementary material online, Figure S3.

earlier observations,4 highly significant differences, were observed for the three predefined ceramide ratios, with the medians of cases ranging between +25.7 and +34.8% (P < 0.001) relative to controls. The difference between stable CAD patients and controls is illustrated in Supplementary material online, Figure S3. Ceramide ratios in coronary artery disease patients In the Corogene study, the most significant ORs for coronary death were found for HDL markers and ceramide ratios. Ceramide ORs remained predictive after adjustment for conventional lipid markers LDL-C, HDL-C, total cholesterol, triacylglycerols, and C-reactive protein. Lp-PLA2 and Lp(a) had no significant association with CV mortality. LDL-C and LDL particle number were inversely associated with risk (LDL-C unadjusted upper quartile OR 0.89 95% CI 0.37–2.14; LDL-P upper quartile OR 0.67, 95%CI 0.29–1.59). For comparison, the unadjusted upper quartile OR for the Cer(d18:1/16:0)/Cer(d18:1/24:0) ratio was 10.33 (95% CI 3.69–28.97). Figure 1 shows non-adjusted and adjusted ORs for different lipid markers and ceramides in the Corogene study. Figure 1 Cardiovascular death odds ratios for per standard deviation and 4th quartile for different lipid markers and ceramides in Corogene study. Adjustment is made for total cholesterol, triacylglycerols, LDL-C, HDL-C, and C-reactive protein.

justed ORs for different lipid markers and ceramides in the Corogene study. Figure 1 Cardiovascular death odds ratios for per standard deviation and 4th quartile for different lipid markers and ceramides in Corogene study. Adjustment is made for total cholesterol, triacylglycerols, LDL-C, HDL-C, and C-reactive protein. Ceramides and risk in stable coronary artery disease patients An independent assessment of ceramides was performed in a cohort (BECAC) of stable patients. We found that the predefined ceramide ratios were significantly higher in 81 patients who died following a CV event within 4.6-year follow-up compared with those who did not die during follow-up (Table 2; Supplementary material online, Table S2). For comparability with the Corogene results, non-adjusted and adjusted ORs for standard lipid markers and ceramides are given in Supplementary material online, Table S3. The incremental improvement of discrimination for CV death was further demonstrated by calculating hazard ratios adjusted for standard lipids and the Marschner score variables (Table 3). Table 3 Association between ceramides and cardiovascular death in BECACa

amides are given in Supplementary material online, Table S3. The incremental improvement of discrimination for CV death was further demonstrated by calculating hazard ratios adjusted for standard lipids and the Marschner score variables (Table 3). Table 3 Association between ceramides and cardiovascular death in BECACa Univariate model Multivariableb model 1b Multivariablec model 2c Hazard ratiod (95% CI) P-value Hazard ratiod (95% CI) P-value Hazard ratiod (95% CI) P-value Cer(d18:1/16:0)/Cer(d18:1/24:0)e 1.77 (1.46–2.16) <0.001 1.79 (1.45–2.20) <0.001 1.52 (1.21–1.92) <0.001 Cer(d18:1/18:0)/Cer(d18:1/24:0)e 1.63 (1.31–2.04) <0.001 1.58 (1.25–2.00) <0.001 1.29 (1.01–1.65) 0.039 Cer(d18:1/24:1)/Cer(d18:1/24:0)e 1.61 (1.30–1.98) <0.001 1.58 (1.27–1.97) <0.001 1.31 (1.03–1.66) 0.028 Cer(d18:1/16:0)e 1.44 (1.17–1.77) <0.001 2.09 (1.61–2.73) <0.001 1.75 (1.30–2.35) <0.001 Cer(d18:1/18:0)e 1.33 (1.07–1.65) 0.011 1.54 (1.19–2.01) 0.001 1.27 (0.98–1.66) 0.076 Cer(d18:1/24:0)e 0.83 (0.67–1.02) 0.081 0.82 (0.62–1.10) 0.182 0.91 (0.69–1.21) 0.510 Cer(d18:1/24:1)e 1.39 (1.13–1.72) 0.002 1.74 (1.34–2.25) <0.001 1.38 (1.04–1.82) 0.023 Cer denotes ceramide. aCV death denotes death from MI, stroke, and heart failure. bThe model was adjusted for TC, TG, HDL-C, and LDL-C. cThe model was adjusted as for model 1 with additional adjustment for the following Marschner score variables: age, gender, smoking status, previous acute MI, diabetes, hypertension, and prior stroke. dHazard ratios are for 1 SD increase. eNatural logarithm of the ceramides and ceramide ratio.

bThe model was adjusted for TC, TG, HDL-C, and LDL-C. cThe model was adjusted as for model 1 with additional adjustment for the following Marschner score variables: age, gender, smoking status, previous acute MI, diabetes, hypertension, and prior stroke. dHazard ratios are for 1 SD increase. eNatural logarithm of the ceramides and ceramide ratio. Adjustment for statin treatment did not have a major impact on the results (Supplementary material online, Table S4). The odds ratios were calculated for ceramides and LDL-C also in patients that were on or not on statin treatment both at baseline and after 1-year of follow-up (Supplementary material online, Table S5). Ceramides were predictive in both instances, although in patients without statin treatment the odds ratios were better. LDL-C did not show significant predictive value, confirming that the lack of a direct association between LDL-C and CV death was not caused by interference with statin treatment. The incremental prognostic value of ceramides was tested by comparing the base model composed of the Marschner score variables to a new model with the Marschner score variables combined with the Cer(d18:1/16:0)/Cer(d18:1/24:0) ratio. The ceramides increased the cross-validated c-statistics from 0.78 (0.75–0.80) to 0.80 (0.77–0.82). Further, the predicted probabilities for a 1-year event risk by logistic regression yielded an NRI of 0.15 (95% CI 0.06–0.25; 9.6% improvement for events and 5.8% improvement for non-events).

16:0)/Cer(d18:1/24:0) ratio. The ceramides increased the cross-validated c-statistics from 0.78 (0.75–0.80) to 0.80 (0.77–0.82). Further, the predicted probabilities for a 1-year event risk by logistic regression yielded an NRI of 0.15 (95% CI 0.06–0.25; 9.6% improvement for events and 5.8% improvement for non-events). Ceramides and risk prediction in acute coronary syndromes patients Another independent assessment of ceramides was performed in the SPUM-ACS cohort enrolling ACS patients. In 51 patients who died following a cardiac event within one-year-follow-up the ceramide ratios were significantly higher compared with those who survived during follow-up (Table 2; Supplementary material online, Table S2). For comparability with the Corogene results, non-adjusted and adjusted ORs for standard lipid markers and ceramides are given in Supplementary material online, Table S3, and the effects of statin treatment are accounted for the results in Supplementary material online, Table S4. The incremental improvement of discrimination for cardiac death was further demonstrated by adjusting for standard lipids and the GRACE score (Table 4), and also by taking into account diabetes mellitus and smoking status (Supplementary material online, Table S6). Table 4 Association between ceramides and cardiovascular death in SPUM-ACSa

ovement of discrimination for cardiac death was further demonstrated by adjusting for standard lipids and the GRACE score (Table 4), and also by taking into account diabetes mellitus and smoking status (Supplementary material online, Table S6). Table 4 Association between ceramides and cardiovascular death in SPUM-ACSa Univariate model Multivariableb model 1b Multivariablec model 2c Hazard ratiod (95% CI) P-value Hazard ratiod (95% CI) P-value Hazard ratiod (95% CI) P-value Cer(d18:1/16:0)/Cer(d18:1/24:0) 1.81 (1.52–2.14) <0.001 1.82 (1.51–2.21) <0.001 1.69 (1.39–2.06) <0.001 Cer(d18:1/18:0)/Cer(d18:1/24:0) 1.66 (1.43–1.96) <0.001 1.65 (1.39–1.97) <0.001 1.48 (1.24–1.76) <0.001 Cer(d18:1/24:1)/Cer(d18:1/24:0) 1.74 (1.45–2.08) <0.001 1.77 (1.44–2.17) <0.001 1.64 (1.32–2.03) <0.001 Cer(d18:1/16:0)e 1.45 (1.10–1.93) 0.010 1.96 (1.45–2.66) <0.001 1.98 (1.49–2.62) <0.001 Cer(d18:1/18:0)e 1.43 (1.07–1.90) 0.015 1.77 (1.31–2.38) <0.001 1.66 (1.26–2.20) <0.001 Cer(d18:1/24:0)e 0.66 (0.51–0.87) 0.003 0.74 (0.52–1.05) 0.090 0.91 (0.65–1.29) 0.609 Cer(d18:1/24:1)e 1.23 (0.93–1.63) 0.154 1.74 (1.25–2.42) 0.001 1.73 (1.27–2.36) <0.001 Cer denotes ceramide. aCV death denotes death from MI, stroke, and heart failure. bThe model was adjusted for TC, TG, HDL-C, and LDL-C. cThe model was adjusted as for model 1 with additional adjustment for the Grace score (Killip class, systolic blood pressure, heart rate, age, creatinine, cardiac arrest at admission, ST-segment deviation, and elevated cardiac enzyme levels). dHazard ratios are for one standard deviation increase. eNatural logarithm of the ceramides.

cThe model was adjusted as for model 1 with additional adjustment for the Grace score (Killip class, systolic blood pressure, heart rate, age, creatinine, cardiac arrest at admission, ST-segment deviation, and elevated cardiac enzyme levels). dHazard ratios are for one standard deviation increase. eNatural logarithm of the ceramides. The incremental prognostic value of ceramides was tested by comparing the base model composed of the GRACE score to a new model with the GRACE score and the Cer(d18:1/16:0)/Cer(d18:1/24:0) ratio on top. The ceramide ratio increased the cross-validated c-statistics from 0.73 (0.70–0.77) to 0.82 (0.79–0.85). Furthermore, the predicted probabilities for a 1-year event risk obtained by logistic regression yielded an NRI of 0.17 (95% CI 0.07–0.27; 8.2% improvement for events and 9.1% improvement for non-events). The performance of the ceramide ratio Cer(d18:1/16:0)/Cer(d18:1/24:0) in predicting non-fatal MI was also investigated by calculating the hazard ratios both for Q-wave and non-Q wave MIs. The ratio showed a significant result for Q-wave MI, while no significant results were obtained for non-Q wave infarctions (Supplementary material online, Table S7).

atio Cer(d18:1/16:0)/Cer(d18:1/24:0) in predicting non-fatal MI was also investigated by calculating the hazard ratios both for Q-wave and non-Q wave MIs. The ratio showed a significant result for Q-wave MI, while no significant results were obtained for non-Q wave infarctions (Supplementary material online, Table S7). Ceramides and C-reactive protein In the Corogene and SPUM-ACS studies, ceramides associated significantly with LDL-C and C-reactive protein. Particularly, the CV mortality-related Cer(d18:1/16:0) and Cer(d18:1/18:0) were positively correlated with C-reactive protein, while small negative correlations were seen in both studies between C-reactive protein and Cer(d18:1/24:0). Furthermore, the ‘protective’ Cer(d18:1/24:0) had the strongest associations with LDL-C. Correlation coefficients for associations of ceramides and ceramide ratios with LDL-C and C-reactive protein are presented in Supplementary material online, Tables S8 and S9. Finally, the synergy of C-reactive protein and Cer(d18:1/16:0)/Cer(d18:1/24:0) in risk stratification was investigated by calculating event rates in different quartiles for both BECAC and SPUM-ACS. Especially in the SPUM-ACS study the highest enrichment of events (11.4% 1-year mortality) was observed if both the ceramide ratio and C-reactive protein were in the highest quartile of the whole population (Supplementary material online, Table S10).

ating event rates in different quartiles for both BECAC and SPUM-ACS. Especially in the SPUM-ACS study the highest enrichment of events (11.4% 1-year mortality) was observed if both the ceramide ratio and C-reactive protein were in the highest quartile of the whole population (Supplementary material online, Table S10). Ceramide score and risk for cardiovascular death We have developed a tentative risk score (Supplementary material online, Table S11) based on ceramide concentrations and their ratios to model the clinical use of these risk predictors. Based on the score, the patients were placed into four risk categories (low–moderate–increased–high) and both in the BECAC and SPUM-ACS studies the risk increased along with the increasing score (Table 5). In the stable CAD and ACS patients 4.2- and 6.0-fold relative risk increase was observed when comparing the high- to low-risk category, respectively. When subjects were sorted according to their LDL-C concentrations and split into four categories in the same proportion as for the ceramide risk score the enrichment of high-risk patients was not observed along with increasing LDL-C concentration. Table 5 Ceramide score and risk for cardiovascular death

, respectively. When subjects were sorted according to their LDL-C concentrations and split into four categories in the same proportion as for the ceramide risk score the enrichment of high-risk patients was not observed along with increasing LDL-C concentration. Table 5 Ceramide score and risk for cardiovascular death BECAC (5-year risk) SPUM-ACS (1-year risk) Score No death Death % Relative risk Score No death Death % Relative risk 0–2 534 15 2.7% 1.0 0–2 566 9 1.6% 1.0 3–6 572 29 4.8% 1.8 3–6 595 16 2.6% 1.7 7–9 268 20 6.9% 2.5 7–9 261 9 3.3% 2.1 10–12 132 17 11.4% 4.2 10–12 164 17 9.4% 6.0 LDL-C (mg/dl) No death Death % Relative risk LDL-C (mg/dl) No death Death % Relative risk ≤100 513 36 6.6% 1.0 ≤106 532 27 4.8% 1.0 100–143 572 29 4.8% 0.7 106–145 576 17 2.9% 0.6 143–175 278 10 3.5% 0.5 145–174 260 3 1.1% 0.2 ≥175 142 6 4.1% 0.6 ≥174 174 2 1.1% 0.2 See Supplementary material online, Table S11 for information on Ceramide Score calculation. To compare Ceramide Score performance with that of LDL-C study, subjects were sorted according to their LDL-C levels and split into four categories in the same proportion as for the ceramide risk score. Discussion The present results provide evidence that distinct ceramide species serve as significant predictors for cardiovascular death beyond currently used lipid markers in two patient groups—patients with stable CAD and higher risk ACS patients . Importantly, the prediction also works in patients who are already statin treated and is therefore a potential indicator of residual risk.

de species serve as significant predictors for cardiovascular death beyond currently used lipid markers in two patient groups—patients with stable CAD and higher risk ACS patients . Importantly, the prediction also works in patients who are already statin treated and is therefore a potential indicator of residual risk. Battes et al.17 recently performed a systematic review of models predicting outcome in patients with stable CAD. The authors concluded that risk stratification should be improved to predict recurrent coronary events and to optimize secondary prevention strategies. Our data using ceramides address this unmet need and show robust performance for predicting coronary death both in stable CAD and ACS patients. The present results do not prove causality. However, it is tempting to speculate that ceramides are associated with plaque vulnerability as they are known to fuel many central atherosclerosis processes including lipoprotein aggregation and uptake, inflammation, superoxide anion production, and apoptosis8,18–21 (Supplementary material online, Figure S1). Several enzymes of the sphingolipid synthesis have already been tested as potential drug targets as inhibition of glycosphingolipid biosynthesis has been shown to decrease atherosclerosis in mice.10,22 Evidence is also accumulating on ceramide chain-length-specific functions. In a recent study, the relative increase in long-chain species (C16) but not in very-long-chain (C24-24:1) species was shown to mediate insulin resistance in mice.11,12 In Caenorhabditis elegans, long-chain ceramides were pro-apoptotic, and very-long-chain ceramides were anti-apoptotic.9 Consistently in the present study, long-chain species (d18:1/16:0 and d18:1/18:0) were more harmful than very-long-chain (d18:1/24:0) species. Altered ceramide compositions may partially be explained by CerS isoforms, providing a putative biological explanation for the use of ceramide ratios, and possibilities for medical intervention (Supplementary material online, Figure S2). Interestingly, Cer(d18:1/24:1) behaved differently compared with Cer(d18:1/24:0), emphasizing that additional regulation also takes place. While the current study reveals an association between ceramides and CV events, it will be a highly interesting topic for future investigations to establish if ceramide composition can be influenced and how it might translate to cardiovascular benefit.

8:1/24:0), emphasizing that additional regulation also takes place. While the current study reveals an association between ceramides and CV events, it will be a highly interesting topic for future investigations to establish if ceramide composition can be influenced and how it might translate to cardiovascular benefit. The response of ceramides to lipid-lowering treatments such as statins has been documented in our previous study.4 We have observed that PCSK9 knock-out mice have significantly reduced plasma ceramide concentrations and that human PCSK9 loss-of-function mutations are associated with lower plasma ceramide concentrations compared with individuals carrying the major alleles.4,23 Study limitations in addition to the lack of causality data include the limited number of events both in BECAC and SPUM-ACS. Thus, the ceramide risk score derived from these studies should be further validated in sizeable cohorts in order to fine-tune the relative risk estimates for different risk categories. Finally, it is likely that the careful one-to-one case–control matching in the Corogene study is leading to somewhat optimistic biomarker results compared with a real-life patient care situation where controlling for confounding factors is more difficult.

-tune the relative risk estimates for different risk categories. Finally, it is likely that the careful one-to-one case–control matching in the Corogene study is leading to somewhat optimistic biomarker results compared with a real-life patient care situation where controlling for confounding factors is more difficult. The lack of a discernible relationship between LDL-related parameters and CV risk across the studies included here, even after statin stratification, is thought-provoking but, in line with previous reports.24–26 For example, Sachdeva et al.24 analysed admission lipid levels in a broad population of 136,995 patients hospitalized for CAD in 541 hospitals and observed that nearly half of the admission LDL-C concentrations were <100 mg/dl although before admission only 21.1% patients were receiving lipid-lowering medications. Furthermore, in the MIRACL trial, the plasma HDL-C, but not LDL-C, measured in the initial stage of ACS predicted the risk of recurrent cardiovascular events.25 The lower LDL-C in higher risk patients may not be a phenomenon of ACS solely as in the Saturn trial investigators observed that C-reactive protein, but not LDL-C levels, were associated with coronary atheroma regression and cardiovascular events after intensive statin therapy.26 Taken together, it appears that the LDL-C concentrations may be similar or even lower in CAD patients at high risk for future CV events compared with patients with more favorable prognosis. This may be a phenomenon related to disease progression and culmination, and should not detract from the value of applying LDL-C to gauge the lifetime risk to develop atherosclerotic plaques. A potential explanation for this is provided in Gierens et al.27 who demonstrated that interleukin-6 (IL-6) activates LDL-receptor (LDLr) transcription and subsequently enhances LDLr activity in the liver leading to an increased elimination of LDL-C from the circulation. Thus, chronic, and in particular acute bursts of, inflammation in CAD patients may enhance LDL-C clearance, resulting in lowered blood LDL-C concentrations and impaired risk prediction.

transcription and subsequently enhances LDLr activity in the liver leading to an increased elimination of LDL-C from the circulation. Thus, chronic, and in particular acute bursts of, inflammation in CAD patients may enhance LDL-C clearance, resulting in lowered blood LDL-C concentrations and impaired risk prediction. Ceramide measurement in high-throughput quality controlled environments is straightforward and cost-efficient. Isotope labelled standards enable precise quantification and analytical stability. Most clinical laboratories are equipped with robotized sample handling systems and also house mass spectrometry equipment.

transcription and subsequently enhances LDLr activity in the liver leading to an increased elimination of LDL-C from the circulation. Thus, chronic, and in particular acute bursts of, inflammation in CAD patients may enhance LDL-C clearance, resulting in lowered blood LDL-C concentrations and impaired risk prediction. Ceramide measurement in high-throughput quality controlled environments is straightforward and cost-efficient. Isotope labelled standards enable precise quantification and analytical stability. Most clinical laboratories are equipped with robotized sample handling systems and also house mass spectrometry equipment. Thus, ceramide-based identification of coronary patients at high cardiovascular risk will soon be possible. These high-risk patients should then benefit from treatments that extend beyond standard care. The suggested actions could include more frequent follow-up visits and efficient life-style counselling as well as the consideration for higher statin doses, ezetimibe combinations, or novel therapies such as PCSK9 inhibitors. In future, additional therapies may include other options, for example, ongoing randomized clinical trials are looking into the effect of methotrexate and interleukin-1β inhibition for treating cardiovascular risk.28 Indeed, ceramides are closely linked to inflammatory processes and recently CERS6 has been identified as a target for methotrexate.29 It is hence plausible to think that ceramide testing could become increasingly relevant, especially if the trials with anti-inflammatory compounds turn out positive. A health economic dimension of ceramide testing is its ability to target more intense, potentially more expensive treatments such as PCSK9 inhibitors to those at the highest risk. Another aspect of ceramide testing is its potential for motivating patient's adherence, whether for medication or life-style changes, due to its rather direct linkage with CV mortality. It has been shown that over 40% of the patients prescribed statins are non-adherent, which may translate to many avoidable additional events and hospitalizations.30

ting is its potential for motivating patient's adherence, whether for medication or life-style changes, due to its rather direct linkage with CV mortality. It has been shown that over 40% of the patients prescribed statins are non-adherent, which may translate to many avoidable additional events and hospitalizations.30 While the ceramide-based risk stratification extends beyond the current lipid-based diagnostics and addresses the unmet need for improved identification of high-risk CAD patients there are further scientific and clinical issues that need attention. There are two major lines of future research concerning the present ceramide correlation and cardiovascular risk. On one hand it is of interest to pursue the biology of these molecules and work out their molecular mechanism of action in cardiovascular disease. This will involve multi-disciplinary efforts of cell biologists, biochemists, geneticists, and clinicians developing appropriate cell and animal models. Efforts in this regard are already being made for example in the scope of the EU funded ‘EUFP7-Atheroflux’ consortium. The other line to follow is to establish the utility of these markers in clinical practice. In the USA, the ceramide testing is entering the clinic this year and only this real-life evaluation will allow for a better judgement of the ceramide utility and will establish them as a new armament in the clinical diagnostic tool-kit. Supplementary material Supplementary material is available at European Heart Journal online.

While the ceramide-based risk stratification extends beyond the current lipid-based diagnostics and addresses the unmet need for improved identification of high-risk CAD patients there are further scientific and clinical issues that need attention. There are two major lines of future research concerning the present ceramide correlation and cardiovascular risk. On one hand it is of interest to pursue the biology of these molecules and work out their molecular mechanism of action in cardiovascular disease. This will involve multi-disciplinary efforts of cell biologists, biochemists, geneticists, and clinicians developing appropriate cell and animal models. Efforts in this regard are already being made for example in the scope of the EU funded ‘EUFP7-Atheroflux’ consortium. The other line to follow is to establish the utility of these markers in clinical practice. In the USA, the ceramide testing is entering the clinic this year and only this real-life evaluation will allow for a better judgement of the ceramide utility and will establish them as a new armament in the clinical diagnostic tool-kit. Supplementary material Supplementary material is available at European Heart Journal online. Authors’ contributions M. S.-A., M. H., M. S. performed statistical analysis; R. L., R. H., M. N., J. S., O. N., T. L., W. M. handled funding and supervision; K. E., D. K., H. S., T. S., E. V., M.-L. L., R. K., C. M., T. H., P. J., N. R., L. R., S. W., B. G., E. R. P., G. S. T., F. M. acquired the data; R. L., R. H., J. S., T. L., F. M., O. N. conceived and designed the research; R. L., T. V. drafted the manuscript; K. E., M. S.-A., M. H., R. H., D. K., W. M., H. S., T. S., E. V., M. L. L., M. N., R. K., C. M., T. H., P. J., N. R., L. R., S. W., E. P., G. T., B. G., F. M., J. S., O. N., T. L. made critical revision of the manuscript for key intellectual content.

eived and designed the research; R. L., T. V. drafted the manuscript; K. E., M. S.-A., M. H., R. H., D. K., W. M., H. S., T. S., E. V., M. L. L., M. N., R. K., C. M., T. H., P. J., N. R., L. R., S. W., E. P., G. T., B. G., F. M., J. S., O. N., T. L. made critical revision of the manuscript for key intellectual content. Funding This work was supported by the European Union's Seventh Framework Programme FP7/2007-2013 RiskyCAD Project (3057392) and further by research grants of the Swiss National Research Foundation (SPUM 33CM30-124112), the Swiss Heart Foundation, both Bern Switzerland, the Foundation for Cardiovascular Research—Zürich Heart House, Zürich, Switzerland as well as AstraZeneca, Zug; Eli Lilly Indianapolis, USA; and Vernier, Medtronic, Tolochenaz; Merck Sharpe and Dohme, Glattbrugg; Sanofi, Vernier; and St. Jude Medical, Zürich (all Switzerland). The Corogene study was supported by grants from Aarno Koskelo Foundation, Helsinki University Central Hospital Special Government Funds (EVO #TYH7215, #TKK2012005, #TYH2012209, and #TYH2014312), and Finnish Foundation for Cardiovascular research. The BECAC study was supported by a grant from the Western Norway Regional Health Authority (911570). The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. Funding to pay the Open Access publication charges for this article was provided by Zora Biosciences.

role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. Funding to pay the Open Access publication charges for this article was provided by Zora Biosciences. Conflict of interest: F.M. has received research grants to the institution from Amgen, AstraZeneca, Boston Scientific, Biotronik, Medtronic, MSD, Eli Lilly, and St. Jude Medical including speaker or consultant fees. S.W. has received research grants to the institution from Abbott, AstraZeneca, Boston Scientific, Biosensors, Biotronik, Cordis, Eli Lilly, Medtronic, and St. Jude Medical. T.F.L. received research grants to the institution from AstraZeneca, Bayer Health Care, Biosensors, Biotronik, Boston Scientific, Medtronic, Merck, Sharpe and Dohme, Merck, Inc., Roche, and Servier, including lecture fees. C.M.M. received research grants to the institution from Eli Lilly, AstraZeneca, Roche and MSD and speaker or consultant fees from Eli Lilly, Daiichi Sankyo, AstraZeneca, Roche and MSD. W.M. is employed with Synlab Holding Germany GmbH, has ownership interest in Synlab Holding International GmbH and has received research grants to the institution from Aegerion Pharmaceuticals, Amgen, Astrazeneca, Danone Research, Sanofi/Genzyme, Hoffmann LaRoche, Numares, Unilever, and BASF and speaker and consultancy fees from Aegerion Pharmaceuticals, Amgen, Astrazeneca, Danone Research, Sanofi/Genzyme, Hoffmann LaRoche, Merck Sharp and Dohme, Pfizer, Sanofi, Synageva, Numares, Unilever, and BASF. Zora Biosciences holds patents for the diagnostic use of ceramides and R.H. and R.L. are shareholders of Zora Biosciences.

consultancy fees from Aegerion Pharmaceuticals, Amgen, Astrazeneca, Danone Research, Sanofi/Genzyme, Hoffmann LaRoche, Merck Sharp and Dohme, Pfizer, Sanofi, Synageva, Numares, Unilever, and BASF. Zora Biosciences holds patents for the diagnostic use of ceramides and R.H. and R.L. are shareholders of Zora Biosciences. Acknowledgements We thank Mrs. Sirpa Sutela-Tuominen and Mrs. Ritva Huuhilo for expert technical support and sample preparation.

See page 2081 for the editorial comment on this article (doi:10.1093/eurheartj/ehv688) Introduction The physiological behaviour of human coronary stenoses has been inferred from animal experiments that studied changes of flow velocity and pressure in the presence of artificially created stenoses.1–3 These experiments determined that stenoses created by external constriction or ligation had a non-linear relationship between the degree of coronary narrowing and trans-stenotic flow velocity and pressure gradient.3 Early attempts to replicate this across patients with coronary artery disease were unsuccessful, presumably due to the effect of atherosclerosis and cardiovascular risk factors on other domains of the coronary circulation.4,5 Nevertheless, animal models continue to be used to describe human physiology.

ssure gradient.3 Early attempts to replicate this across patients with coronary artery disease were unsuccessful, presumably due to the effect of atherosclerosis and cardiovascular risk factors on other domains of the coronary circulation.4,5 Nevertheless, animal models continue to be used to describe human physiology. Since plotting pressure–flow relationships can be challenging in clinical practice, indices have been formulated to describe the importance of stenoses.6,7 Those involving both pressure and flow velocity measurement have predominantly been used in a research setting.7–10 Pressure-only measurements, being easier to perform, have gained more common clinical application. Fractional flow reserve (FFR), a pressure-only hyperaemic measure is treated as a simplified surrogate for flow based upon assessments in animals. It has compelling outcome data and is widely advocated to guide coronary assessment.11 Another newer pressure-only index, instantaneous wave-free ratio (iFR)12,13 is measured under resting conditions, obviating the need for hyperaemic vasodilators such as adenosine. While the rationale for the hyperaemic physiological assessment of coronary artery stenosis has been extensively validated, resting stenosis assessment has been less extensively explored.

s wave-free ratio (iFR)12,13 is measured under resting conditions, obviating the need for hyperaemic vasodilators such as adenosine. While the rationale for the hyperaemic physiological assessment of coronary artery stenosis has been extensively validated, resting stenosis assessment has been less extensively explored. The aim of this study is to investigate the coronary pressure–flow relationship in patients with and without angiographic evidence of obstructive atherosclerosis under resting and hyperaemic conditions. The IDEAL dataset is used to analyse 567 human coronary artery intracoronary pressure and flow velocity recordings to revisit pressure–flow relationships in a large clinical cohort of patients with stable coronary artery disease, representative of the clinical ‘real world’. Methods Study population This study incorporates prospectively collected data from a total of 567 combined pressure and Doppler flow velocity measurements in 301 patients at the Amsterdam Medical Center Amsterdam, The Netherlands (n = 161), Imperial College London, UK (n = 160), Hospital Clinico San Carlos, Madrid, Spain (n = 21), and VU University Medical Center, Amsterdam, The Netherlands (n = 225). All patients recruited were scheduled for elective coronary angiography with physiological stenosis assessment by FFR and gave written informed consent for acquisition of additional physiological data for study purposes.

os, Madrid, Spain (n = 21), and VU University Medical Center, Amsterdam, The Netherlands (n = 225). All patients recruited were scheduled for elective coronary angiography with physiological stenosis assessment by FFR and gave written informed consent for acquisition of additional physiological data for study purposes. While acquisition methodology of physiological data was similar for all participating centres, the study protocol was different for each centre. Individual centre recruitment criteria are shown in Supplementary material online. Composite exclusion criteria were severe valvular heart disease, weight >200 kg (determined by the catheter laboratory table capacity), previous coronary artery bypass surgery, vessels with angiographically identifiable myocardial bridging or collateral arteries and vessels with a previous myocardial infarction. Patients with an acute myocardial infarction within 48 h were not included.

ht >200 kg (determined by the catheter laboratory table capacity), previous coronary artery bypass surgery, vessels with angiographically identifiable myocardial bridging or collateral arteries and vessels with a previous myocardial infarction. Patients with an acute myocardial infarction within 48 h were not included. Coronary catheterization Coronary angiography and pressure–flow assessments of coronary stenoses were performed using conventional approaches.14 Intracoronary nitrates (200–300 μg) were administered in all cases. Contemporary combined pressure and Doppler flow velocity wires (ComboWire XT, Volcano Corporation, San Diego, CA, USA) were used and the distal pressure sensor was equalized with the aortic guiding pressure at the coronary ostium before distal passing of the wire. Measurements were made distal to the stenosis at least three vessel diameters from the stenosis. Adenosine was administered by intravenous infusion in 234 measurements (140 μg/kg/min) and by intracoronary bolus injection in 333 measurements (60–150 μg).

pressure at the coronary ostium before distal passing of the wire. Measurements were made distal to the stenosis at least three vessel diameters from the stenosis. Adenosine was administered by intravenous infusion in 234 measurements (140 μg/kg/min) and by intracoronary bolus injection in 333 measurements (60–150 μg). Doppler signals were optimized carefully to ensure adequate tracking profiles were observed. Electrocardiogram (ECG), pressures, and flow velocity signals were directly extracted from the device console (ComboMap, Volcano Corporation, San Diego, CA, USA). At the end of each recording, the pressure sensor was returned to the catheter tip to assess pressure drift. If pressure drift was identified (>2 mmHg) measurements were repeated or corrected for upon analysis. Data were analysed off-line, using a custom software package designed with MATLAB (Mathworks, Inc, Natick, MA, USA). A total of 653 cases were originally acquired, but 86 vessels (13.2%) were excluded because of poor Doppler flow velocity or uncontrollable pressure drift leaving 567 vessels for final analysis. Resting indices were calculated at a time of stability, without any preceding injection of contrast or saline. Hyperaemic indices were calculated during stable hyperaemia, excluding ectopy and conduction delay.

ecause of poor Doppler flow velocity or uncontrollable pressure drift leaving 567 vessels for final analysis. Resting indices were calculated at a time of stability, without any preceding injection of contrast or saline. Hyperaemic indices were calculated during stable hyperaemia, excluding ectopy and conduction delay. Stenosed and reference vessels Five hundred and sixty-seven coronary assessments were made. Three hundred and sixty-six vessels had an angiographically visible stenosis. Two hundred and one vessels had no angiographic obstruction as determined by the physician performing the procedure and confirmed by two observers (S.N. and G.d.W.). Stenosis stratification Both FFR and diameter stenosis assessed by quantitative angiography (QCA) were used to stratify stenosis severity. Myocardial FFR measurements were performed, using the ratio of distal coronary pressure to proximal pressure during stable hyperaemia. Quantitative angiography parameters (diameter stenosis % (DS%), minimal lumen diameter, minimal lumen area, area stenosis, and lesion length) were calculated for stenoses using dedicated workstations (CAAS II, Pie Medical, Maastricht, The Netherlands or McKesson, San Francisco, USA), which performed automated contour analysis with manual correction limited to situations causing artefacts or very tight stenoses.

l lumen area, area stenosis, and lesion length) were calculated for stenoses using dedicated workstations (CAAS II, Pie Medical, Maastricht, The Netherlands or McKesson, San Francisco, USA), which performed automated contour analysis with manual correction limited to situations causing artefacts or very tight stenoses. Calculation of hemodynamic parameters Flow velocity was assessed over four periods: first, flow velocity at rest over the entire cardiac cycle and secondly over the specific diastolic wave-free period (during which iFR is calculated), which was detected using the ECG signals. Flow velocity was also assessed during adenosine-mediated hyperaemia over the whole cardiac cycle and the wave-free period. The same two time periods in the cardiac cycle, both at rest and hyperaemia, were used to derive measures of microvascular resistance (MVR) and trans-stenotic pressure gradient (TG). Figure 1 shows an example of simultaneous pressure and flow velocity measurements, together with the cardiac phases over which the study parameters were calculated. Figure 1 Top panel: example of simultaneous coronary pressure and flow velocity measurement obtained distal to an left anterior descending stenosis of 69% diameter stenosis by quantitative angiography with FFR of 0.79. Bottom panel: the analysed phases during the resting and hyperaemic state are shown.

were calculated. Figure 1 Top panel: example of simultaneous coronary pressure and flow velocity measurement obtained distal to an left anterior descending stenosis of 69% diameter stenosis by quantitative angiography with FFR of 0.79. Bottom panel: the analysed phases during the resting and hyperaemic state are shown. Microvascular resistance (mmHg/cm/s) is calculated by dividing distal pressure (Pd, mmHg) by flow velocity (cm/s). When calculated for the whole cardiac cycle, values of pressure and flow velocity were averaged over an entire heart beat; typically measurements were made as an average over five beats. When calculated over the wave-free period, pressure and flow velocity data were constrained to that averaged over the diastolic wave-free period. For all measurements, computation of the parameters was performed by a single analyst blinded to the coronary angiograms or patient specific factors, using an automated MATLAB script (MathWorks, Natick, MA, USA) with built-in wave-free algorithm (developed at Imperial College, London and licensed to Volcano Corp, San Diego, CA, USA), as previously described.12

arameters was performed by a single analyst blinded to the coronary angiograms or patient specific factors, using an automated MATLAB script (MathWorks, Natick, MA, USA) with built-in wave-free algorithm (developed at Imperial College, London and licensed to Volcano Corp, San Diego, CA, USA), as previously described.12 Statistical analysis Categorical data are presented as numbers and percentages, while continuous data are presented as mean ± standard deviation. Regression analysis was performed between quantitative values to determine the coefficient of determination. Curve fitting was achieved by applying 2nd and 3rd order and fractional polynomials. Association between flow velocity (dependent variable) and strata of stenosis severity were assessed by analysis of variance (ANOVA) with correction for repeated measures followed by post hoc pairwise methods, including Bonferroni, Sidak, and Scheffe; this was followed by Tukey HSD testing where appropriate. Findings were confirmed using Kruskal–Wallis testing to avoid assumptions of normality. Trends across strata were assessed with regression and a non-parametric extension of the Wilcoxon signed rank test (nptrend) and also with generalized estimating equations; in all analyses, the findings were the same suggesting a robust analysis. The analysis was repeated for TG or MVR as dependent variables. A P-value of <0.05 was considered statistically significant. All statistical analyses were performed using Stata 11.2 (StataCorp, TX, USA).

lso with generalized estimating equations; in all analyses, the findings were the same suggesting a robust analysis. The analysis was repeated for TG or MVR as dependent variables. A P-value of <0.05 was considered statistically significant. All statistical analyses were performed using Stata 11.2 (StataCorp, TX, USA). Results Patient and vessel characteristics Five hundred and sixty-seven coronary assessments were derived from 301 patients (age 53.5 ± 22.3 years old, 59% male, Table 1). No patients had hypertrophic cardiomyopathy or gross hypertrophy secondary to hypertension. Characteristics of patients and vessels is shown per participating centre in Supplementary material online, Table S1. Three hundred and sixty-six (65%) were from vessels with a visible stenosis, with 85 measurements taken post-percutaneous coronary intervention (PCI). Two hundred and one (35%) measurements were from vessels free from angiographic disease, which served as reference vessels. For the 366 vessels with a stenosis, the FFR ranged from 0.28 to 1.07. Table 1 Demographics and stenosis characteristics

, with 85 measurements taken post-percutaneous coronary intervention (PCI). Two hundred and one (35%) measurements were from vessels free from angiographic disease, which served as reference vessels. For the 366 vessels with a stenosis, the FFR ranged from 0.28 to 1.07. Table 1 Demographics and stenosis characteristics N or mean % or standard deviation Patients 301 Age (years) 60.6 9.6 Male 209 69% Hypertension 157 52% Hyperlipidemia 172 57% Current or ex-smoker 128 43% Diabetes Mellitus 67 22% Chronic renal impairment 5 2% Family history of CAD 129 43% Previous myocardial infarction 34 11% Impaired LV function EF < 30% 2 0.7% Stable angina 290 96% Unstable angina 11 4% Vessels 567 Angiographic stenosis 366 65% Angiographically unobstructed 201 35% Coronary artery Left Anterior Descending 277 49% Left Circumflex 172 30% Right Coronary Artery 118 21% Adenosine administration Central intravenous 234 41% Intracoronary bolus 333 59% Coronary stenoses % Diameter stenosis 46.0 21.3 % Area stenosis 68.9 22.8 Minimal lumen diameter (mm) 1.47 0.75 Minimal lumen area (mm2) 2.09 2.21 Stenosis length (mm) 17.0 12.5

277 49% Left Circumflex 172 30% Right Coronary Artery 118 21% Adenosine administration Central intravenous 234 41% Intracoronary bolus 333 59% Coronary stenoses % Diameter stenosis 46.0 21.3 % Area stenosis 68.9 22.8 Minimal lumen diameter (mm) 1.47 0.75 Minimal lumen area (mm2) 2.09 2.21 Stenosis length (mm) 17.0 12.5 The distribution of stenoses was consistent with that typically found in clinical practice with mean FFR 0.81 ± 0.16 and mean QCA diameter stenosis of 48.5 ± 25.2%, indicating that the majority was of intermediate severity (Figure 2). In the 201 reference vessels, FFR ranged from 0.85 to 1.08 with a mean FFR of 0.96 ± 0.04. Overall, these findings confirmed the absence of obstructive epicardial disease. In 23 (11%) of the reference cases, an FFR of >0.80 and ≤0.90 was documented, suggesting the existence of abnormal epicardial conductance. A value of FFR > 1.0 was noted in 30 vessels despite careful drift assessment. This predominantly occurred in the LCx (18 cases, 60%) and in reference vessels (26 cases, 87%) and was likely due to hydrostatic consequences of the wire being in a distal vessel below the position of the transducer. For the stenosed vessels, the relationship between FFR and anatomical DS% demonstrated significant scatter (R2 = 0.23, P < 0.01; Figure 3). Similar findings were noted when plotting FFR and anatomical minimal luminal diameter (R2 = 0.19, P < 0.001; Supplementary material online, Figure S1). Figure 2 Distribution of the coronary arteries measured stratified according to percentage of diameter stenosis % (upper panel) and fractional flow reserve (lower panel).

milar findings were noted when plotting FFR and anatomical minimal luminal diameter (R2 = 0.19, P < 0.001; Supplementary material online, Figure S1). Figure 2 Distribution of the coronary arteries measured stratified according to percentage of diameter stenosis % (upper panel) and fractional flow reserve (lower panel). Figure 3 Distribution of percentage of diameter stenosis and fractional flow reserve in stenoses. Despite a significant inverse correlation between percentage of diameter stenosis and fractional flow reserve, a substantial variability between the two parameters is noted (R2 = 0.23, P < 0.01). The curve is fitted by second-order polynomial.

of percentage of diameter stenosis and fractional flow reserve in stenoses. Despite a significant inverse correlation between percentage of diameter stenosis and fractional flow reserve, a substantial variability between the two parameters is noted (R2 = 0.23, P < 0.01). The curve is fitted by second-order polynomial. Whole cycle pressure–flow velocity relationships A given stenosis has a unique curvilinear relationship between the flow velocity and the TG across the stenosis. Pressure–flow velocity relationships over the whole cardiac cycle were calculated both during resting and hyperaemic conditions, and averaged for each stratum of stenosis severity. Mean pressure–flow velocity relationships were stratified according to DS% (Figure 4, left panel) and according to FFR classification (Figure 4, right panel). Figure 4 Relationships between trans-stenotic pressure gradient and flow velocity for coronaries grouped by stenosis severity (left panel according to anatomical severity by percentage of diameter stenosis, and right panel according to physiological severity by fractional flow reserve). Relationships are described by trans-stenotic pressure gradient = A*flow + B*flow2 and can be fitted by three points: the zero TG—zero flow crossing, the mean trans-stenotic pressure gradient and flow during whole cycle at rest, and during hyperaemia. Trans-stenotic pressure gradient and flow during the wave-free period closely follow these relationships, both at rest and during hyperaemic conditions. Curves are fitted by second-order polynomials.

ero flow crossing, the mean trans-stenotic pressure gradient and flow during whole cycle at rest, and during hyperaemia. Trans-stenotic pressure gradient and flow during the wave-free period closely follow these relationships, both at rest and during hyperaemic conditions. Curves are fitted by second-order polynomials. Diastolic pressure and flow velocity relationships Pressure–flow velocity relationships were calculated over the wave-free period specifically, both at rest and hyperaemia. For each stratification, these relationships closely fit the pressure–flow velocity curves derived from whole-cycle physiology (Figure 4). Resting wave-free period flow velocity was significantly higher than whole cycle resting conditions (P < 0.05 for each FFR or QCA stratum), and consistently produced a higher TG, both at rest and during hyperaemia (P < 0.05 for each FFR or QCA stratum). The only exception was TG in reference vessels and FFR > 0.91 under resting conditions, where TG was equivalent for whole cycle and wave-free period (1.5 ± 0.2 vs. 1.6 ± 0.2 mmHg, respectively; P = 0.53, and 1.1 ± 0.1 vs. 1.3 ± 0.1 mmHg; P = 0.08) as there was no stenosis sufficient to cause diastolic pressure separation. Under hyperaemic conditions, both a consistently higher flow velocity and TG were found during the wave-free period than during whole cycle (P < 0.05 for each FFR or QCA stratum).

pectively; P = 0.53, and 1.1 ± 0.1 vs. 1.3 ± 0.1 mmHg; P = 0.08) as there was no stenosis sufficient to cause diastolic pressure separation. Under hyperaemic conditions, both a consistently higher flow velocity and TG were found during the wave-free period than during whole cycle (P < 0.05 for each FFR or QCA stratum). Influence of stenosis severity on coronary flow velocity Resting flow velocity stratified according to angiographic and FFR strata is depicted in Figure 5, upper panel. Numerical relationships between stenosis severity and the analysed parameters, as well as the physiological indices, are shown in Tables 2 and 3. Resting flow velocity has no significant relationship with stenosis severity whether assessed by FFR or anatomical severity (Ptrend = 0.16). Hyperaemic flow velocity, over the whole cycle and the wave-free period, shows a strong statistical association and trend to decline with incremental stenosis severity (P < 0.001 for all assessments). Table 2 Flow velocity, TG, MVR and physiological indices according to lesion severity defined by FFR

y (Ptrend = 0.16). Hyperaemic flow velocity, over the whole cycle and the wave-free period, shows a strong statistical association and trend to decline with incremental stenosis severity (P < 0.001 for all assessments). Table 2 Flow velocity, TG, MVR and physiological indices according to lesion severity defined by FFR FFR ≤ 0.50 SD FFR 0.51–0.60 SD FFR 0.61–0.70 SD FFR 0.71–0.80 SD FFR 0.81–0.90 SD FFR >0.91 SD Flow velocity (cm/s) Resting whole cycle 14.8 6.3 15.5 7.1 18.6 6.7 21.9 12.1 18.9 8.7 19.1 8.4 Resting wave-free period 18.7 7.4 20.3 10.4 24.6 10.1 29.1 18.3 25.4 13.2 25.1 11.4 Hyperemic whole cycle 18.5 10.3 22.8 13.6 28.3 13.7 36.9 18.9 38.9 18.0 48.2 23.7 Hyperemic wave-free period 21.0 10.8 25.7 14.6 34.1 16.9 45.0 24.6 50.4 24.8 61.4 28.0 Trans-stenotic gradient (mmHg) Resting whole cycle 45.6 19.7 26.0 11.8 15.5 8.6 9.81 5.0 5.19 3.06 1.17 2.32 Resting wave-free period 55.5 20.3 35.1 15.8 20.7 12.3 13.2 7.56 6.88 4.01 1.48 2.43 Hyperemic whole cycle 54.6 12.1 42.6 7.86 30.17 6.14 22.2 4.63 13.0 4.71 4.71 3.19 Hyperemic wave-free period 61.1 11.0 51.2 8.49 37.9 8.40 30.4 7.37 17.6 5.65 6.34 4.25 Microvascular Resistance (mmHg/cm/s) Resting whole cycle 4.20 1.57 5.55 2.40 5.04 2.03 5.61 4.34 5.80 2.40 6.09 2.39 Resting wave-free period 2.04 0.83 3.22 1.62 3.19 1.52 3.76 3.01 3.93 1.82 4.27 1.79 Hyperemic whole cycle 2.65 1.19 2.97 1.55 2.37 0.88 2.36 1.37 2.32 1.10 2.20 1.12 Hyperemic wave-free period 1.22 0.56 1.55 0.90 1.37 0.62 1.41 0.88 1.45 0.81 1.49 0.86 Indices Pd/Pa 0.56 0.15 0.74 0.11 0.84 0.07 0.91 0.05 0.95 0.03 0.99 0.02 iFR 0.39 0.18 0.60 0.17 0.76 0.12 0.84 0.10 0.92 0.05 0.98 0.03 FFR 0.42 0.06 0.54 0.03 0.66 0.03 0.75 0.03 0.85 0.03 0.95 0.03 HSR (mmHg/cm/s) 3.92 2.36 2.49 1.30 1.23 0.46 0.78 0.51 0.40 0.21 0.12 0.11 BSR (mmHg/cm/s) 3.83 2.58 2.02 1.25 0.88 0.44 0.55 0.44 0.32 0.23 0.07 0.16 CFR 1.26 0.44 1.51 0.69 1.61 0.71 1.81 0.54 2.16 0.74 2.62 0.88 Anatomical parameters % Diameter stenosis 71.3 12.4 65.0 13.3 60.0 18.6 51.8 18.0 43.3 18.7 33.4 18.6 % Area stenosis 90.1 8.36 87.8 9.26 78.0 19.0 77.2 17.7 66.7 22.6 57.7 22.6 Minimal lumen diameter (mm) 0.80 0.39 0.93 0.29 1.01 0.42 1.20 0.46 1.53 0.72 1.87 0.79 Minimal lumen area (mm2) 0.63 0.55 0.76 0.46 0.94 0.83 1.26 0.99 2.23 2.27 3.13 2.57 Stenosis length (mm) 26.7 16.1 32.4 18.1 20.9 18.2 19.5 13.3 15.0 9.05 12.9 8.79

enosis 90.1 8.36 87.8 9.26 78.0 19.0 77.2 17.7 66.7 22.6 57.7 22.6 Minimal lumen diameter (mm) 0.80 0.39 0.93 0.29 1.01 0.42 1.20 0.46 1.53 0.72 1.87 0.79 Minimal lumen area (mm2) 0.63 0.55 0.76 0.46 0.94 0.83 1.26 0.99 2.23 2.27 3.13 2.57 Stenosis length (mm) 26.7 16.1 32.4 18.1 20.9 18.2 19.5 13.3 15.0 9.05 12.9 8.79 Table 3 Flow velocity, TG, MVR and physiological indices according to lesion severity defined by anatomical stenosis severity (% diameter stenosis)

enosis 90.1 8.36 87.8 9.26 78.0 19.0 77.2 17.7 66.7 22.6 57.7 22.6 Minimal lumen diameter (mm) 0.80 0.39 0.93 0.29 1.01 0.42 1.20 0.46 1.53 0.72 1.87 0.79 Minimal lumen area (mm2) 0.63 0.55 0.76 0.46 0.94 0.83 1.26 0.99 2.23 2.27 3.13 2.57 Stenosis length (mm) 26.7 16.1 32.4 18.1 20.9 18.2 19.5 13.3 15.0 9.05 12.9 8.79 Table 3 Flow velocity, TG, MVR and physiological indices according to lesion severity defined by anatomical stenosis severity (% diameter stenosis) QCA ≥ 90% SD QCA 80–89% SD QCA 70–79% SD QCA 60–69% SD QCA 50–59% SD QCA ≤ 49% SD Reference SD Flow Velocity (cm/s) Resting whole cycle 16.4 8.02 17.5 6.52 14.9 7.60 19.4 9.63 17.7 9.0 20.0 8.9 17.8 6.9 Resting wave-free period 23.7 12.6 21.4 7.63 19.5 10.9 25.3 14.4 23.1 12.1 26.5 12.8 23.3 10.2 Hyperemic whole cycle 24.7 17.9 24.1 11.0 23.5 12.2 31.1 14.2 35.0 19.0 45.4 22.3 44.9 16.0 Hyperemic wave-free period 28.8 19.2 27.7 11.5 28.8 16.9 38.6 19.5 44.4 25.0 57.8 27.9 58.1 21.6 Trans-stenotic gradient (mmHg) Resting whole cycle 34.7 26.1 29.5 27.2 22.8 19.1 13.4 16.1 8.10 9.36 4.42 4.97 1.53 2.51 Resting wave-free period 47.4 31.9 35.9 29.8 28.8 22.4 17.9 19.3 10.8 12.7 5.64 6.72 1.58 2.82 Hyperemic whole cycle 41.9 18.5 37.5 23.2 32.4 18.2 24.9 17.9 18.0 12.4 10.8 7.72 3.55 4.02 Hyperemic wave-free period 52.1 18.0 43.1 23.5 39.3 19.3 31.3 19.4 23.5 15.2 14.2 10.1 4.04 5.28 Microvascular Resistance (mmHg/cm/s) Resting whole cycle 4.9 2.37 4.22 1.60 5.50 2.29 5.30 2.13 6.27 3.55 5.73 2.46 6.16 2.33 Resting wave-free period 1.93 0.31 2.45 0.97 3.38 1.91 3.41 1.61 4.25 2.51 3.92 1.85 4.38 1.83 Hyperemic whole cycle 2.84 1.91 2.65 1.28 2.69 0.99 2.52 1.10 2.56 1.50 2.14 1.01 2.18 0.80 Hyperemic wave-free period 1.07 0.37 1.44 0.67 1.49 0.71 1.47 0.71 1.58 1.08 1.38 0.75 1.48 0.62 Indices Pd/Pa 0.68 0.21 0.72 0.21 0.77 0.19 0.87 0.15 0.92 0.10 0.96 0.05 0.98 0.03 iFR 0.51 0.28 0.60 0.27 0.67 0.24 0.80 0.21 0.87 0.15 0.93 0.08 0.98 0.03 FFR 0.55 0.16 0.60 0.17 0.64 0.18 0.73 0.17 0.80 0.14 0.88 0.09 0.96 0.04 HSR (mmHg/cm/s) 3.33 3.52 2.34 2.60 1.22 1.50 0.78 1.03 0.78 1.03 0.32 0.32 0.09 0.10 BSR (mmHg/cm/s) 3.26 3.32 2.22 2.68 0.91 1.31 0.55 0.79 0.55 0.79 0.25 0.27 0.09 0.17 CFR 1.37 0.38 1.36 0.36 1.65 0.67 1.67 0.59 2.12 0.85 2.39 0.86 2.64 0.76 Hyperemic stenosis resistance (HSR) is defined as the ratio between the TG and flow velocity under hyperemic conditions, while baseline stenosis resistance (BSR) is calculated the same way but under resting conditions instead. Coronary flow reserve (CFR) is defined as the ratio between hyperemic and resting flow velocity.

eremic stenosis resistance (HSR) is defined as the ratio between the TG and flow velocity under hyperemic conditions, while baseline stenosis resistance (BSR) is calculated the same way but under resting conditions instead. Coronary flow reserve (CFR) is defined as the ratio between hyperemic and resting flow velocity. Figure 5 Behaviour of phasic coronary flow velocity, microvascular resistance (MVR) and (TG) according to stenosis severity (left panel diameter stenosis by QCA, and right panel by FFR). Parameters are shown for resting and hyperemic conditions, both during whole cycle and wave-free period only. Curves are fitted by second-, third-order, and fractional polynomials.

city, microvascular resistance (MVR) and (TG) according to stenosis severity (left panel diameter stenosis by QCA, and right panel by FFR). Parameters are shown for resting and hyperemic conditions, both during whole cycle and wave-free period only. Curves are fitted by second-, third-order, and fractional polynomials. Microvascular resistance (Figure 5, middle panel) measured at rest showed highly significant trends to fall with increasing stenosis severity, during both whole cycle and wave-free period specifically (Ptrend < 0.001). Hyperaemic MVR over the whole cycle was low for all stenosis severities, but showed a trend to increase in the most severe strata of stenosis severity (Ptrend = 0.01). This trend was not observed during the wave-free period under hyperaemic conditions, which remained consistent across all strata of stenosis severity (Ptrend = 0.89 for FFR and Ptrend = 0.82 for anatomical stratification). The trend toward higher values of hyperaemic MVR over the whole cycle but not during the wave-free period, arises from an increasing MVR under the systole specifically (Supplementary material online, Figure S2). Findings shown in Figure 5 are maintained when post-PCI measurements are excluded or when only the post-PCI measurements are analysed (Supplementary material online, Figure S3).

but not during the wave-free period, arises from an increasing MVR under the systole specifically (Supplementary material online, Figure S2). Findings shown in Figure 5 are maintained when post-PCI measurements are excluded or when only the post-PCI measurements are analysed (Supplementary material online, Figure S3). In contrast, TG (Figure 5, lower panel) whether measured at rest or hyperaemia, had strong and significant relationships with stenosis severity and followed identical trends, also for the wave-free period (Ptrend < 0.001 for all assessments). Hyperaemic TG was strongly related to FFR (R2 = 0.95, P < 0.001). The behaviour of flow, MVR and TG over the whole cycle under resting conditions for the entire study cohort is summarized in Figure 6. Figure 6 Mean TG, flow velocity and MVR data under resting conditions over the whole cardiac cycle, stratified according to angiographic stenosis severity. With progressive stenosis severity, TG increases, while flow velocity is maintained at a stable level by progressive compensatory reduction of MVR. Curves are fitted by 2nd, 3rd order and fractional polynomials.

der resting conditions over the whole cardiac cycle, stratified according to angiographic stenosis severity. With progressive stenosis severity, TG increases, while flow velocity is maintained at a stable level by progressive compensatory reduction of MVR. Curves are fitted by 2nd, 3rd order and fractional polynomials. A natural incremental hierarchy exists between the physiological states assessed: resting whole cycle, resting wave-free period, hyperaemic whole cycle and hyperaemic wave-free period physiology. This was true for flow velocity and TG (P < 0.001 for both), while for MVR the same hierarchy exists in reverse (P < 0.001). When stenoses had diameter stenosis >90% or FFR ≤ 0.50, the hierarchy was no longer valid with resting flow velocity exceeding hyperaemic and MVR being lower at rest than during hyperaemia.

logy. This was true for flow velocity and TG (P < 0.001 for both), while for MVR the same hierarchy exists in reverse (P < 0.001). When stenoses had diameter stenosis >90% or FFR ≤ 0.50, the hierarchy was no longer valid with resting flow velocity exceeding hyperaemic and MVR being lower at rest than during hyperaemia. Anatomical stenosis severity When analysed according to area stenosis or minimal lumen diameter the same physiological outcomes for flow velocity, MVR and TG were observed (Supplementary material online, Table S1) as when analysed according to FFR or DS%. When data were analysed according to lesion length, resting flow velocity decreased numerically slightly but statistically significantly with increasing stenosis length (Supplementary material online, Figure S4). Very long lesion lengths were the main contributor to this trend. In stenoses up to 40 mm long, resting wave-free flow velocity was preserved at 24.1 ± 12.6 cm/s; in stenoses over 40 mm, flow velocity was 19.1 ± 7.9, creating a significant trend (P = 0.01) although not a significant difference in mean values by ANOVA (P = 0.62) or T-test (P = 0.15). The same was true for whole cycle resting flow velocity (18.3 ± 8.7 vs. 15.4 ± 1.9 cm/s, P = 0.23). Hyperaemic flow velocity similarly diminished from 39.4 ± 22.4 to 26.7 ± 18.8 cm/s, demonstrating a strong trend. Microvascular resistance appeared to be unrelated to lesion length, with no significant trends noted for rest or hyperaemia. Trans-stenotic pressure gradient was strongly related to length for all four physiological states (Ptrend < 0.001 for all). Minimal lumen diameter and area were in keeping with results as stratified according to FFR or DS% (Supplementary material online, Table S2).

th no significant trends noted for rest or hyperaemia. Trans-stenotic pressure gradient was strongly related to length for all four physiological states (Ptrend < 0.001 for all). Minimal lumen diameter and area were in keeping with results as stratified according to FFR or DS% (Supplementary material online, Table S2). Overall trends and relationships observed remain unchanged when data are stratified according to the presence of diffuse and focal disease (Supplementary material online, Figure S5) or according to singular and serial stenoses (Supplementary material online, Figure S6). Adenosine administration route Stratification of data by adenosine administration route according to FFR showed findings in keeping with the overall dataset for flow velocity and TG (Ptrend < 0.001 for all phases) (Supplementary material online, Figure S7). Hyperaemic whole cycle MVR significantly increased with progressive stenosis severity for intracoronary administration (Ptrend = 0.04), but remained consistent for intravenous administration (Ptrend = 0.35). During the wave-free period, hyperaemic MVR was consistent for the intracoronary route (Ptrend = 0.17), but showed a trend to being lower with progressive stenosis severity with intravenous adenosine (Ptrend = 0.03). Discussion In this study, we describe the relationship between coronary flow velocity, TG, and MVR, estimated from measurements obtained over the whole cardiac cycle or selectively within the wave-free period, under resting and hyperaemic conditions.

Adenosine administration route Stratification of data by adenosine administration route according to FFR showed findings in keeping with the overall dataset for flow velocity and TG (Ptrend < 0.001 for all phases) (Supplementary material online, Figure S7). Hyperaemic whole cycle MVR significantly increased with progressive stenosis severity for intracoronary administration (Ptrend = 0.04), but remained consistent for intravenous administration (Ptrend = 0.35). During the wave-free period, hyperaemic MVR was consistent for the intracoronary route (Ptrend = 0.17), but showed a trend to being lower with progressive stenosis severity with intravenous adenosine (Ptrend = 0.03). Discussion In this study, we describe the relationship between coronary flow velocity, TG, and MVR, estimated from measurements obtained over the whole cardiac cycle or selectively within the wave-free period, under resting and hyperaemic conditions. Firstly, we show that non-hyperaemic flow velocity remains constant across the full spectrum of stenosis severities. Secondly, this preservation of flow velocity is mediated by a reduction in resting MVR in response to increasing stenosis resistance. Thirdly, the maintenance of resting flow velocity occurs at the expense of distal coronary pressure, which falls with widening TG as stenosis severity increases. The capacity for resting gradients to increase while preserving flow velocity lends support to clinical use of invasive resting coronary pressure assessment to determine functional stenosis significance. Finally, we provide reference values of parameters used in physiological assessment of the coronary circulation stratified according to stenosis severity (Tables 2 and 3).

eserving flow velocity lends support to clinical use of invasive resting coronary pressure assessment to determine functional stenosis significance. Finally, we provide reference values of parameters used in physiological assessment of the coronary circulation stratified according to stenosis severity (Tables 2 and 3). Auto-regulation ensures that resting blood flow remains stable Maintenance of resting coronary flow is regulated by endogenous adenosine release, changes in intrinsic myogenic tone, endothelial cell signalling and neurohumoral control, which combine to produce continuous auto-regulatory adaption of arteriolar vessel diameter.15 In this study, we use invasively measured resting flow velocity and found that this was stable in human coronary arteries across a wide range of stenosis severities. While the resistance imposed by a stenosis rises according to the Hagen-Poiseuille equation,16 we observe a reduction in MVR to compensate (Figure 6 summarizes the results for whole cycle resting conditions). This reduction is closely related to stenosis severity keeping resting flow stable and therefore, for the majority of moderate stenoses, a vasodilator reserve should be expected. This is observed by a reduction in MVR in the presence of a hyperaemic agent, such as adenosine. For stenoses with very little physiological impact, a large vasodilatory reserve is present meaning a large potential increase in flow during hyperaemia. In more significant stenoses, however, vasodilatory reserve will become progressively exhausted, with limited increase in flow response to an exogenous vasodilator. When a critical stenosis severity is reached (likely to be exceeding 85–90% diameter by formal QCA measurement, or FFR values <0.60), coronary auto-regulation becomes saturated with limited vasodilatory response to exogenous agents. When a stenosis is beyond this critical point, resting flow velocity is expected to fall. Clinically, this may manifest as angina on increasingly lower levels of exertion.

by formal QCA measurement, or FFR values <0.60), coronary auto-regulation becomes saturated with limited vasodilatory response to exogenous agents. When a stenosis is beyond this critical point, resting flow velocity is expected to fall. Clinically, this may manifest as angina on increasingly lower levels of exertion. Anatomical and physiological markers of stenosis severity Cursory assessment of Figure 4 may suggest that FFR and anatomical classification of stenoses are equivalent as the same patterns of change in flow and MVR are observed. However, as shown by many authors, there is a limited relationship between FFR and anatomical severity, which is confirmed in this cohort (Figure 2). The assessment is presented, not to state that anatomy and physiology are equivalent, rather because the overall trends are so strong that they are preserved even when the random scatter of the FFR-DS% relationship limits the potential relationship. Since anatomical assessment of stenoses remains mainstay and is readily understood by clinicians, it is appropriate to consider the underlying physiological response to anatomical parameters, despite the crude limitations of diameter stenosis. When study outcomes are analysed according to other parameters that describe lesion tightness (minimal lumen diameter, minimal lumen area, and area stenosis) similar findings are noted. This is also true for the presence of diffuse compared with focal disease and singular compared with serial stenoses. Resting flow velocity showed a trend to falling with very long lesion lengths (over 40 mm) but only showed a small change, while hyperaemic flow diminished significantly. Figures 4 and 5 visually show that when stratified according to DS%, a remarkable overlap with respect to the TG for the 50–59 and 60–69% groups as well as the 70–79 and 80–89% groups is present. This observation reinforces that in stenosis of intermediate severity, physiological assessment is required to inform on haemodynamic significance.

y show that when stratified according to DS%, a remarkable overlap with respect to the TG for the 50–59 and 60–69% groups as well as the 70–79 and 80–89% groups is present. This observation reinforces that in stenosis of intermediate severity, physiological assessment is required to inform on haemodynamic significance. The use of resting parameters to assess stenoses The stability of resting flow velocity for the majority of stenoses means that resting flow alone cannot distinguish between stenosis severities. However, since distal coronary pressure falls with increasing stenosis severity, a combined pressure and flow velocity measurement such as baseline stenosis resistance or a resting pressure-only index such as iFR, can distinguish stenosis severity. This implies that non-invasive imaging modalities such as positron emission tomography, that measure myocardial perfusion without knowledge of distal coronary pressure, require induction of the hyperaemic state to yield satisfactory diagnostic accuracy.17

ssure-only index such as iFR, can distinguish stenosis severity. This implies that non-invasive imaging modalities such as positron emission tomography, that measure myocardial perfusion without knowledge of distal coronary pressure, require induction of the hyperaemic state to yield satisfactory diagnostic accuracy.17 The change in resting TG is predominantly driven by a change in MVR and resistance imposed by the stenosis. Since the impact of physiological vasodilation at rest on the proximal driving pressure is negligible, changes in distal pressure represent the true physiological impact of the stenosis on the distal coronary bed. Small gradients at rest suggest little compensatory vasodilatation is required, while large gradients indicate substantial compensation. For a stenosis to have a physiological impact upon the vessel, it should therefore have a gradient that is detectible at rest and the induction of hyperaemia will only exacerbate this gradient. Stenoses without a resting gradient which manifests only upon vasodilator administration more likely represents a situation in which the microcirculatory bed retains the capacity to dilate significantly and high flow velocities can be generated across a trivial stenosis with subsequent turbulence and pressure loss by the Bernoulli phenomenon.18 These changes may manifest as a high coronary flow reserve (CFR), but a low FFR value. Human data with 10-year follow-up confirm that when such patients are deferred from PCI, the clinical event rate remains low, demonstrating a clear paradox between hyperaemic measurements of pressure and flow.19

rnoulli phenomenon.18 These changes may manifest as a high coronary flow reserve (CFR), but a low FFR value. Human data with 10-year follow-up confirm that when such patients are deferred from PCI, the clinical event rate remains low, demonstrating a clear paradox between hyperaemic measurements of pressure and flow.19 Microvascular remodelling Resting MVR reduces with increasing stenosis severity. Theoretically, microcirculatory angiogenesis and arteriogenesis could explain this.20 However, if this phenomenon applies, it would not be restricted to the resting situation and a substantial reduction in MVR would remain during hyperaemia in severe stenoses. Our results indicate that this is not the case and instead we confirm observations from smaller studies, that hyperaemic MVR increases in critical stenoses.8,10,21 The increased hyperaemic MVR in severe stenoses is primarily a systolic phenomenon (Supplementary material online, Figure S2). We presume this observed rise in hyperaemic MVR, can be attributed to the contribution of collateral circulation. Because collateral arteries connect with the receiving vessel distal to the position of the pressure–flow wire, flow supplied by the collateral arteries will not be detected, while collateral pressure is transmitted through the vessels and can be detected by the wire. Microvascular resistance is calculated as the ratio of distal pressure (elevated by collateral supply) and flow, and the calculated MVR will falsely rise accordingly.22 Naturally, when using whole cycle pressure values, the contribution of the elevated pressures is higher than when using the lower diastolic pressure values as collateral pressure is elevated mainly during systole and much less so during diastole.23 Moreover, in the intracoronary adenosine subgroup, a trend was observed towards higher MVR during hyperaemia whole cycle, while in the intravenous adenosine subgroup, hyperaemic whole cycle MVR was consistent across stenosis severities. In the intravenous subgroup, collateral supply may be diminished due to the coronary steal phenomenon during hyperaemia and thereby the MVR in these severe stenoses remains at normal values. However, this analysis has the limitation that coronary steal phenomenon might still apply in the intracoronary subgroup for the left coronary artery. Further work to assess the collateral flow or pressure during diastole is required to understand this in detail.

e MVR in these severe stenoses remains at normal values. However, this analysis has the limitation that coronary steal phenomenon might still apply in the intracoronary subgroup for the left coronary artery. Further work to assess the collateral flow or pressure during diastole is required to understand this in detail. Clinical implications In this study, we provide flow velocity and resistance data from a wide spectrum of coronary stenoses and reference vessels. These data are valuable for accurate development and improvement of computer flow dynamics models. For current flow models, such as CT-FFR, data were derived from animals and small human studies without significant disease.24,25 Our data demonstrate that unobstructed vessels have a mean CFR of 2.64 ± 0.76 in contrast to older data informing CT-FFR, which assumes flow rises of 3.5-fold.30 Similarly, early CT systems assume resistance falls by 4.5-fold with adenosine, while our data show whole cycle MVR is reduced by 2.8-fold. Secondly, the data presented here provide reference values stratified according to stenosis severity for the most commonly used physiological indices. Exploration of less commonly used physiological parameters such as the instantaneous hyperaemic diastolic velocity–pressure slope (IHDVPS) and zero-flow pressure (ZFP) may be of future interest to better indicate their clinical applicability.

ording to stenosis severity for the most commonly used physiological indices. Exploration of less commonly used physiological parameters such as the instantaneous hyperaemic diastolic velocity–pressure slope (IHDVPS) and zero-flow pressure (ZFP) may be of future interest to better indicate their clinical applicability. Finally, the data support the concept that stenosis interrogation under resting conditions, as suggested by iFR, BSR, or baseline Pd/Pa10,12,13 has clinical utility beyond comparisons of classification match with hyperaemic measures. Furthermore, our findings demonstrate that the wave-free period consistently provides a higher flow velocity and a lower MVR than assessment over the whole cardiac cycle at rest. This means that wave-free period gradients are consistently larger than over the whole cycle and iFR may provide greater sensitivity in moderate stenoses when compared with baseline Pd/Pa. To provide a definitive answer to which physiological index is preferable, randomized clinical outcome data are needed.

le at rest. This means that wave-free period gradients are consistently larger than over the whole cycle and iFR may provide greater sensitivity in moderate stenoses when compared with baseline Pd/Pa. To provide a definitive answer to which physiological index is preferable, randomized clinical outcome data are needed. Conclusion This large multicentre study of coronary pressure–velocity measurements shows that with progressive stenosis severity, TG rises, while resting coronary flow is maintained by compensatory reduction of MVR. This suggests that resting pressure indices can be used to detect the haemodynamic significance of coronary artery stenoses. Our results confirm the applicability of the general principles of coronary physiology determined in animals to patients with atherosclerotic lesions. The main difference observed is a relatively blunted response to hyperaemia as flow velocity rose to half what has been observed in animal models in vessels with <50% diameter stenosis. Limitations This study has a number of limitations. Volumetric flow was not assessed because of the limitations of accurate stenosis and vessel dimension calculation, as well as determining the mass of the subtended myocardium which can only be estimated from angiographic parameters. Since vessels taper, flow velocity will fall less than volumetric flow and without knowledge of subtended mass, flow velocity might be preferable to volumetric flow.26–28

d vessel dimension calculation, as well as determining the mass of the subtended myocardium which can only be estimated from angiographic parameters. Since vessels taper, flow velocity will fall less than volumetric flow and without knowledge of subtended mass, flow velocity might be preferable to volumetric flow.26–28 Wedge pressure was not routinely measured and therefore definitive assessment of the impact of collaterals on the results cannot be made. However, measurements were not made in vessels with visible collaterals. While reference vessels were free of angiographic disease, intravascular ultrasound studies demonstrated significant burden of atherosclerosis in apparently unobstructed coronary arteries.29 Diffuse atherosclerosis can cause pressure loss and this may account for the wide-range of FFR values observed in reference vessels (lowest obtained 0.85). It remains uncommon to routinely perform intravascular imaging in unobstructed vessels and therefore, together with the large number of unobstructed vessels, our findings should be applicable to patients undergoing coronary angiography. Although we stratified data according to FFR and QCA DS%, both are imperfect measures. In the presence of microcirculatory dysfunction, FFR may underestimate true haemodynamic stenosis significance.30 Quantitative angiography provides limited information of the physiological impact of a given stenosis. However, both measures are easy to comprehend and familiar to clinicians providing a familiar conceptual framework to interpret the data.

y dysfunction, FFR may underestimate true haemodynamic stenosis significance.30 Quantitative angiography provides limited information of the physiological impact of a given stenosis. However, both measures are easy to comprehend and familiar to clinicians providing a familiar conceptual framework to interpret the data. Finally, it must be borne in mind, however, that our results are inferred on group basis and heterogeneous factors such as microvascular dysfunction and diffuse epicardial disease could obscure these findings on a patient-specific level.31,32 Theoretically, however, any factor that impairs auto-regulatory responses to a stenosis could also impact upon microcirculatory responses to vasodilators such as adenosine. When there are discrepancies between resting and hyperaemic factors, it remains unclear which parameters provide prognostic information. Randomized clinical outcome studies are currently being undertaken to assess the safety and performance of resting parameters to guide revascularization.33,34 Supplementary material Supplementary material is available at European Heart Journal online.

Finally, it must be borne in mind, however, that our results are inferred on group basis and heterogeneous factors such as microvascular dysfunction and diffuse epicardial disease could obscure these findings on a patient-specific level.31,32 Theoretically, however, any factor that impairs auto-regulatory responses to a stenosis could also impact upon microcirculatory responses to vasodilators such as adenosine. When there are discrepancies between resting and hyperaemic factors, it remains unclear which parameters provide prognostic information. Randomized clinical outcome studies are currently being undertaken to assess the safety and performance of resting parameters to guide revascularization.33,34 Supplementary material Supplementary material is available at European Heart Journal online. Authors’ contributions S.N. performed statistical analysis; J.D., N.v.R. handled funding and supervision. S.N., G.d.W., S.S., T.v.d.H., R.P., M.E.-P., M.v.L., M.M., I.D., P.K., J.E., J.P., J.D., and N.v.R. acquired the data. S.N., G.d.W., J.D., and N.v.R. conceived and designed the research. S.N., G.d.W., J.D., and N.v.R. drafted the manuscript. S.S., T.v.d.H., R.P., M.E.-P., M.v.L., M.M., I.D., P.K., J.E., and J.P. made critical revision of the manuscript for key intellectual content.

P.K., J.E., J.P., J.D., and N.v.R. acquired the data. S.N., G.d.W., J.D., and N.v.R. conceived and designed the research. S.N., G.d.W., J.D., and N.v.R. drafted the manuscript. S.S., T.v.d.H., R.P., M.E.-P., M.v.L., M.M., I.D., P.K., J.E., and J.P. made critical revision of the manuscript for key intellectual content. Funding This work was supported by the Medical Research Council (UK), British Heart Foundation and the National Institute for Health Research Imperial Biomedical Research Centre (to S.N., S.S., and R.P.) and the Institute for Cardiovascular Research of the VU University of Amsterdam (ICaR-VU) (to G.d.W. and N.v.R.). Funding to pay the Open Access publication charges for this article was provided by the VU University Medical Center. Conflict of interest: J.E.D. and J.J.P. report consultancy work for Volcano Corporation. J.E. reports consultancy work for Volcano Corporation and St Jude Medical. J.E.D. holds intellectual property which is under license.

Introduction Early and accurate identification of individuals with increased risk of coronary heart disease (CHD) is critical for effective implementation of preventative lifestyle modifications and medical interventions, such as statin treatment.1,2 To this end, risk scores such as the Framingham Risk Score (FRS)3 and the American College of Cardiology/American Heart Association 2013 risk score (ACC/AHA13),1 based on clinical factors and lipid measurements, have been developed and are widely used. Although the scores can identify individuals at very high risk, a large proportion of individuals developing CHD during the next 10 years remain unidentified. In particular, they do not provide sufficient discrimination at a younger age when implementation of preventative measures is likely to provide the greatest long-term benefit.

cores can identify individuals at very high risk, a large proportion of individuals developing CHD during the next 10 years remain unidentified. In particular, they do not provide sufficient discrimination at a younger age when implementation of preventative measures is likely to provide the greatest long-term benefit. Genetic factors have long been recognized to make a substantial contribution to CHD risk.4 Although a positive family history is an independent risk factor for CHD, it may not completely and solely capture genetic risk. Recently, genome-wide association studies (GWAS) have identified 56 genetic loci associated with CHD at genome-wide significance.5–9 Studies of the predictive power of the top single nucleotide polymorphisms (SNPs) at some of these loci either individually or in combination have typically shown small improvements in CHD risk prediction,10–17 probably because together these variants only explain less than 20% of CHD heritability.8 As demonstrated recently for other traits such as height and BMI,18,19 the majority of unexplained heritability is likely hidden amongst the thousands of SNPs that did not reach genome-wide significance. Indeed, recent advances have shown that genomic prediction models that consider all available genetic variants can more efficiently stratify those at increased risk of complex disease.20–24 To leverage the maximum amount of information, we examined whether a genomic risk score (GRS) comprising a large number of SNPs, including those with less than genome-wide significance, could produce clinically relevant predictive power for CHD risk.

efficiently stratify those at increased risk of complex disease.20–24 To leverage the maximum amount of information, we examined whether a genomic risk score (GRS) comprising a large number of SNPs, including those with less than genome-wide significance, could produce clinically relevant predictive power for CHD risk. Methods A summary of the key methods for the study is given here. The study design is given in Figure 1. Additional details are provided in the see Supplementary material online, Supplementary Appendix. Figure 1 Study workflow. (A) The procedure for deriving the GRS of incident CHD. The analysis workflow for evaluating the GRS within (B) ARGOS, (C) FINRISK, and (D) FHS.

s given here. The study design is given in Figure 1. Additional details are provided in the see Supplementary material online, Supplementary Appendix. Figure 1 Study workflow. (A) The procedure for deriving the GRS of incident CHD. The analysis workflow for evaluating the GRS within (B) ARGOS, (C) FINRISK, and (D) FHS. Prospective study cohorts We utilized two sets of prospective cohorts: (i) FINRISK, consisting of three prospective cohorts from Finland with 10–20 years of follow-up, from collections 1992, 1997, and 2002 (FR92, FR97, and FR02, respectively)25 and (ii) the Framingham Heart Study (FHS),26–28 with individuals of Western and Southern European ancestry taken from the Original and Offspring cohorts with 40–48 years of follow-up. In total, the FINRISK consisted of n = 12 676 individuals and the FHS of n = 3406 individuals, all of whom had the requisite data and were independent of the CARDIoGRAMplusC4D stage-2 meta-analysis utilized to generate the GRS (Table 1). The cohorts have been genome-wide SNP genotyped and further imputed to the 1000 Genomes reference panel (see Supplementary material online, Supplementary Methods). After genotype imputation and quality control, 69 044 autosomal SNPs of the 79 128 CARDIoGRAMplusC4D SNPs were available for subsequent analyses in the FINRISK, and 78 058 autosomal SNPs available in FHS. Table 1 Characteristics of the FINRISK and FHS cohorts

(see Supplementary material online, Supplementary Methods). After genotype imputation and quality control, 69 044 autosomal SNPs of the 79 128 CARDIoGRAMplusC4D SNPs were available for subsequent analyses in the FINRISK, and 78 058 autosomal SNPs available in FHS. Table 1 Characteristics of the FINRISK and FHS cohorts Study FINRISK Framingham Heart Study Cohort FR92 (n=3547) FR97 (n=4761) FR02 (n=4368) Total FINRISK (n=12,676) FHS Original (n=950) FHS Offspring (n=2456) Total FHS (n=3406) Men 1578 (44%) 2316 (49%) 1919 (44%) 5813 (46%) 370 (39%) 1179 (48%) 1549 (45%) Women 1969 (56%) 2445 (51%) 2449 (56%) 6863 (54%) 580 (61%) 1277 (52%) 1857 (55%) Baseline age, years 43.59 (11.31) 46.68 (13.15) 47.12 (13.01) 45.97 (12.7) 53.7 (6.09) 40.66 (7.47) 44.3 (9.21) Current smoker 1027 (29%) 1148 (24%) 1162 (27%) 3337 (26%) 422 (44%) 948 (39%) 1370 (40%) Blood pressure, systolic, mm Hg 134.79 (19.13) 135.02 (19.62) 134.94 (20.24) 134.93 (19.7) 131.54 (19.35) 122.64 (15.98) 125.12 (17.45) Cholesterol, total, mmol/L 5.6 (1.12) 5.54 (1.06) 5.62 (1.14) 5.58 (1.11) 6.14 (1.08) 5.21 (0.98) 5.47 (1.09) Cholesterol, HDL, mmol/L 1.41 (0.35) 1.42 (0.35) 1.52 (0.43) 1.45 (0.38) 1.3 (0.37) 1.33 (0.39) 1.32 (0.39) Prevalent type 2 diabetes 119 (3%) 299 (6%) 278 (6%) 696 (5%) 19 (2%) 39 (2%) 58 (2%) Lipid lowering treatment 43 (1%) 117 (2%) 231 (5%) 391 (3%) – – – Anti-hypertensive treatment 302 (9%) 569 (12%) 582 (13%) 1453 (11%) 57 (6%) 75 (3%) 132 (4%) Follow up, years 18.49 (3.77) 13.82 (2.88) 9.47 (1.51) 13.63 (4.53) 29.91 (11.32) 31.95 (8.44) 31.38 (9.38) Incident CHD event (before age 75) 261 (7%) 324 (7%) 172 (4%) 757 (6%) 173 (18%) 414 (17%) 587 (17%) Categorical variables are shown as counts and percentages, continuous variables (age, follow-up time, cholesterol, and blood pressure) as means and standard deviations. Sample sizes are for participants with GWAS data after quality control and all other exclusions. Lipid lowering treatments were not assessed in FHS due to an insufficient number of exams with this information.

continuous variables (age, follow-up time, cholesterol, and blood pressure) as means and standard deviations. Sample sizes are for participants with GWAS data after quality control and all other exclusions. Lipid lowering treatments were not assessed in FHS due to an insufficient number of exams with this information. The outcome of interest in FINRISK was primary incident CHD event, defined as myocardial infarction (MI), a coronary revascularization procedure, or death from CHD, before age 75 years (see Supplementary material online, Supplementary Methods). Individuals with prevalent cardiovascular disease (CVD) at baseline were excluded from the analysis. We censored events for individuals with an attained age of >75 years, as not all FINRISK cohorts had sufficient numbers of CHD events beyond that age. In FHS, we used the FHS definition of CHD, which included recognized/unrecognized MI or death from CHD as well as angina pectoris or coronary insufficiency (see Supplementary material online, Supplementary Methods). FHS individuals with prevalent CHD or <30 years of age at baseline were excluded, and for consistency with the FINRISK analysis, a censoring age of 75 years was also applied to the FHS analyses.

ath from CHD as well as angina pectoris or coronary insufficiency (see Supplementary material online, Supplementary Methods). FHS individuals with prevalent CHD or <30 years of age at baseline were excluded, and for consistency with the FINRISK analysis, a censoring age of 75 years was also applied to the FHS analyses. Secondary external validation of the GRS was also performed in the ARGOS study, a Dutch case/control dataset where all individuals had familial hypercholesterolemia (248 young cases with early CHD, 216 elderly controls without CHD), imputed to 1000 Genomes reference panel (74 135 SNPs of the 79 128 CARDIoGRAMplusC4D SNPs were available; see Supplementary material online, Supplementary Methods). Statistical analysis GRSs were generated via thinning the CARDIoGRAMplusC4D SNPs by linkage disequilibrium (LD) thresholds and evaluated using logistic regression and area under receiver-operating characteristic curve (AUC) for each threshold (see Supplementary material online, Figure S1). To avoid overfitting we only used weights (log odds) from the CARDIoGRAMplusC4D stage-2 meta-analysis, which were not based on the WTCCC-CAD or MIGen studies (see Supplementary material online, Supplementary Methods). We combined the estimates for WTCCC and MIGen-Harps using fixed-effects inverse-variance weighted meta-analysis.

verfitting we only used weights (log odds) from the CARDIoGRAMplusC4D stage-2 meta-analysis, which were not based on the WTCCC-CAD or MIGen studies (see Supplementary material online, Supplementary Methods). We combined the estimates for WTCCC and MIGen-Harps using fixed-effects inverse-variance weighted meta-analysis. Subsequent performance of the GRS was evaluated in external, independent validation data. For analysis of FINRISK, we used Cox proportional hazard models to evaluate the association of the GRS with time to incident CHD events, stratifying by sex and adjusting for geographic location and cohort, using age as the time scale. Secondary analyses adjusted for one of the clinical risk scores (FRS or ACC/AHA13), or individual baseline variables and known risk factors (cohort, geographical location, prevalent type-2 diabetes, log total cholesterol, log HDL, log systolic BP, smoking status, lipid treatment, and family history). Family history in FINRISK was self-reported and was defined as having a 1st-degree relative who had experienced MI before age 60. For FHS, we evaluated the association of the GRS with incident CHD using Cox proportional hazard models, stratifying by sex and adjusting for cohort (Original or Offspring), using age as the time scale. Family history was not available for both FHS cohorts and thus not considered in FHS analyses. Survival analyses allowing for competing risks were performed using the Aalen-Johansen estimator of survival and cause-specific Cox models (see Supplementary material online, Supplementary Methods). Model discrimination of incident CHD event was evaluated in three groups of individuals: (i) all individuals (n = 12 676 in FINRISK, n = 3406 in FHS), (ii) individuals aged <60 years at baseline (n = 10 606 in FINRISK, n = 3218 in FHS), and (iii) individuals aged ≥60 years at baseline (n = 2070 in FINRISK, n = 188 in FHS).

l discrimination of incident CHD event was evaluated in three groups of individuals: (i) all individuals (n = 12 676 in FINRISK, n = 3406 in FHS), (ii) individuals aged <60 years at baseline (n = 10 606 in FINRISK, n = 3218 in FHS), and (iii) individuals aged ≥60 years at baseline (n = 2070 in FINRISK, n = 188 in FHS). Discrimination of incident CHD events within 10 years was assessed using Harrell’s C-index, and the difference in C-index between two models was assessed using the correlated jackknife test. Competing risk analyses were performed using the Aalen-Johansen empirical estimator of cumulative incidence and cause-specific Cox proportional hazard models. Risk reclassification was evaluated using continuous Net Reclassification Improvement (NRI), categorical NRI, and Integrated Discrimination Improvement. Meta-analysis of the discrimination statistics was performed using fixed-effect inverse-variance weighting. Additional details on the statistical methods are provided in the see Supplementary material online, Supplementary Methods.

ification Improvement (NRI), categorical NRI, and Integrated Discrimination Improvement. Meta-analysis of the discrimination statistics was performed using fixed-effect inverse-variance weighting. Additional details on the statistical methods are provided in the see Supplementary material online, Supplementary Methods. Results To construct an optimized GRS using the WTCCC and MIGen-Harps datasets, we first generated a series of GRSs, starting with the 79 128 CARDIoGRAMplusC4D SNPs then progressively lowering the r2 threshold for LD to reduce the redundancy of predictive information and corresponding number of SNPs in the score (Methods and Figure 1). An r2 threshold of 0.7 provided optimal discrimination of CHD cases and controls (WTCCC and MIGen-Harps meta-analysis odds ratio(OR) = 1.70 per S.D. of GRS, 95% confidence interval (CI 1.61–1.80; meta-analysis AUC = 0.64, 95% CI 0.63–0.66), corresponding to 49 310 SNPs in WTCCC (see Supplementary material online, Figure S1). Of these 49 310 SNPs, 85.9% (42 364 SNPs) and 95% (46 773 SNPs) were available in the FINRISK and FHS, respectively.

ratio(OR) = 1.70 per S.D. of GRS, 95% confidence interval (CI 1.61–1.80; meta-analysis AUC = 0.64, 95% CI 0.63–0.66), corresponding to 49 310 SNPs in WTCCC (see Supplementary material online, Figure S1). Of these 49 310 SNPs, 85.9% (42 364 SNPs) and 95% (46 773 SNPs) were available in the FINRISK and FHS, respectively. The 49K GRS showed similar odds ratios for incident CHD as a binary outcome in FINRISK (OR = 1.74, 95% CI 1.61–1.89, per S.D.), WTCCC (OR = 1.74, 95% CI 1.63–1.86, per S.D.), and MIGen-Harps (OR = 1.57, 95% CI 1.37–1.81, per S.D.) (Table 2). However in the FHS, the association was weaker, OR = 1.30 (95% CI 1.19–1.43, per S.D.) (Table 2). Density plots of the GRS in FINRISK and FHS for those with and without CHD <75 years are shown in see Supplementary material online, Figure S2. Table 2 Association of the 49K GRS with incident CHD (binary outcome in logistic regression) in the five studies, per standard deviation of the GRS Dataset # Incident CHD/Non-CHD Odds Ratio (95% CI) WTCCC-CAD1 1926/2938 1.74 (1.63–1.86) MIGen-Harps 488/531 1.57 (1.37–1.81) ARGOS FH 248/216 1.49 (1.21–1.84) FINRISK 757/11919 1.74 (1.61–1.89) FHS 587/2819 1.28 (1.17–1.41) WTCCC-CAD1: adjusted for sex and 5 PCs of the genotypes; MIGen-Harps: adjusted for sex and 5 PCs; ARGOS: adjusted for sex and 5 PCs; FINRISK: adjusted for sex, cohort, east/west, and 5 PCs; FHS: adjusted for sex, cohort, and 5 PCs.

48/216 1.49 (1.21–1.84) FINRISK 757/11919 1.74 (1.61–1.89) FHS 587/2819 1.28 (1.17–1.41) WTCCC-CAD1: adjusted for sex and 5 PCs of the genotypes; MIGen-Harps: adjusted for sex and 5 PCs; ARGOS: adjusted for sex and 5 PCs; FINRISK: adjusted for sex, cohort, east/west, and 5 PCs; FHS: adjusted for sex, cohort, and 5 PCs. Using survival analyses of time to incident CHD, within FINRISK the GRS had stronger association with CHD (HR = 1.74, 95% CI 1.61–1.86, per S.D.) than the 28 SNP score studied by Tikkanen et al.11 (HR = 1.21, 95% CI 1.13–1.30, per S.D.), the 27 SNP score used by Mega et al.29 (HR = 1.21, 95% CI 1.12–1.30 per S.D.), or the 153 SNPs found at FDR <0.05 by the CARDIoGRAMplusC4D consortium8 (HR = 1.25, 95% CI 1.16–1.39 per S.D.) (see Supplementary material online, Supplementary Results). In FHS, the GRS showed weaker but statistically significant association with CHD (HR = 1.28 per S.D. of the GRS, 95% CI 1.18–1.38). The fixed-effect meta-analysis estimate for the GRS combining FINRISK and FHS was HR = 1.66 (95% CI 1.55–1.78), however, heterogeneity was high (I2 = 89.2%, Cochran’s Q P = 0.0023). The top vs. bottom quintiles of the GRS showed significantly different incident CHD risk overall (FINRISK HR = 4.51, 95% CI 3.47–5.85; FHS HR = 1.84 95% CI 1.43–2.37). For both FINRISK and FHS, the GRS showed improved prediction for incident CHD over the other risk scores composed of smaller numbers of SNPs (see Supplementary material online, Supplementary Results and Table S3).

different incident CHD risk overall (FINRISK HR = 4.51, 95% CI 3.47–5.85; FHS HR = 1.84 95% CI 1.43–2.37). For both FINRISK and FHS, the GRS showed improved prediction for incident CHD over the other risk scores composed of smaller numbers of SNPs (see Supplementary material online, Supplementary Results and Table S3). In both FINRISK and FHS, the hazard ratios for GRS were not substantially attenuated by adjusting for FRS or ACC/AHA13 clinical risk scores, lipid treatment at baseline, other established risk factors (including family history in FINRISK), or 5 principal components of the genotypes (see Supplementary material online, Figures S3 and S4). The correlation between GRS and either FRS or ACC/AHA13 scores was close to zero with almost none of the variation in GRS explained by either clinical risk score (in both FINRISK and FHS, r2 < 0.004 between GRS and either FRS and ACC/AHA13; see Supplementary material online, Figure S5). To further test that the CHD risk conferred by the GRS was largely independent of the effects of cholesterol, we further validated the GRS in the ARGOS familial hypercholesterolemia study, with comparable results to those obtained in WTCCC/MIGen (OR = 1.49, 95% CI 1.21–1.84 per S.D. of the GRS, adjusted for sex and five principal components) (see Supplementary material online, Supplementary Methods).

he effects of cholesterol, we further validated the GRS in the ARGOS familial hypercholesterolemia study, with comparable results to those obtained in WTCCC/MIGen (OR = 1.49, 95% CI 1.21–1.84 per S.D. of the GRS, adjusted for sex and five principal components) (see Supplementary material online, Supplementary Methods). To assess the predictive power of the GRS, we compared its performance in discrimination of time to CHD event (C-index) with that of family history and the FRS and ACC/AHA13 clinical risk scores. We also assessed the incremental value of the GRS on top of the clinical risk scores. In both FINRISK and FHS, addition of GRS to either FRS or ACC/AHA13 scores provided statistically significant improvements in C-index, in FINRISK: +1.7% (P < 10 − 6) and +1.6% (P < 10 − 6) for FRS and ACC/AHA13, respectively; in FHS: +1.1% (P < 0.0443) and +1.1% (P < 0.0344) for FRS and ACC/AHA13, respectively (Figure 2). Overall, fixed-effects meta-analysis of the two studies showed that GRS improved the C-index by +1.6% (95% CI 0.01–0.02, P < 10 − 6; heterogeneity: I2 = 2.2%, Q = 1.02, P = 0.312) for FRS and GRS combined (FRS + GRS) over FRS alone and, similarly, +1.5% (95% CI 0.009–0.02, P < 10 − 6; heterogeneity: I2 = 0%, Q = 0.78, P = 0.378) for ACC/AHA13 + GRS over ACC/AHA13 alone (Figure 2). Larger increases in C-index were observed among older individuals, with the C-index of FRS + GRS compared with FRS alone increasing by 5.1% in individuals aged ≥60 years at baseline, while individuals aged <60 years at baseline showed C-index gains of 1.4% (see Supplementary material online, Figure S6). Within FINRISK, the GRS had higher C-index than family history (+1.9%, P < 1.3 × 10 − 6). Figure 2 Difference in C-index (95% CI) for time to incident CHD event within 10 years, relative to the reference model in the FINRISK and FHS cohorts. Reference models used age as the time scale, stratified by sex (FINRISK: adjusted for cohort and geographic location; FHS: adjusted for cohort). Family history was not available for all of the FHS cohorts and thus not considered here. P-values are from the correlated jackknife test.

odel in the FINRISK and FHS cohorts. Reference models used age as the time scale, stratified by sex (FINRISK: adjusted for cohort and geographic location; FHS: adjusted for cohort). Family history was not available for all of the FHS cohorts and thus not considered here. P-values are from the correlated jackknife test. We assessed if the GRS improved the individual 10 years risk reclassification when added to clinical risk scores. Analyses within FINRISK and FHS are given in Table 3 for FRS and in Table 4 for ACC/AHA13. Overall, meta-analysis of the two datasets showed that the categorical Net Reclassification Improvement was 0.1 for both FRS + GRS and ACC/AHA13 + GRS, respectively (P < 0.0001; see Supplementary material online, Figure S7). Meta-analysis of continuous NRI was 0.344 (P < 0.001) and 0.334 (P < 0.001) for the FRS + GRS and ACC/AHA13 + GRS, respectively (see Supplementary material online, Figure S8). Meta-analysis of IDI scores showed gains of 0.01 (P < 0.001) and 0.009 (P < 0.001) for FRS + GRS and ACC/AHA13 + GRS, respectively, however IDI scores showed high heterogeneity across FINRISK and FHS (I2 > 97%, Cochran’s Q P < 0.0001, see Supplementary material online, Figure S9). Table 3 Reclassification of incident CHD event risk within 10 years for combined FRS + GRS compared with FRS only, in the FINRISK and FHS cohorts

A13 + GRS, respectively, however IDI scores showed high heterogeneity across FINRISK and FHS (I2 > 97%, Cochran’s Q P < 0.0001, see Supplementary material online, Figure S9). Table 3 Reclassification of incident CHD event risk within 10 years for combined FRS + GRS compared with FRS only, in the FINRISK and FHS cohorts FINRISK FHS FRS+GRS FRS+GRS 0–7.5% 7.5–10% 10–20% 20–100% Total Reclass % 0–7.5% 7.5–10% 10–20% 20–100% Total Reclass % All individuals FRS 0–7.5% 9566 218 138 6 9928 3.6 2482 88 4 0 2574 3.6 7.5–10% 368 190 223 21 802 76.3 122 165 83 1 371 55.5 10–20% 299 290 767 298 1,654 53.6 11 74 339 19 443 23.5 20–100% 1 14 114 156 285 15.7 0 0 5 13 18 27.8 Total 10,234 712 1,242 481 12669 15.7 2,615 327 431 33 3,406 11.9 Incident CHD present FRS 0–7.5% 110 21 19 2 152 27.6 67 6 0 0 73 8.2 7.5–10% 22 12 28 4 66 81.8 5 11 6 0 22 50.0 10–20% 22 24 108 78 232 53.4 2 5 43 4 54 20.4 20–100% 0 2 17 48 67 28.4 0 0 0 1 1 0 Total 154 59 172 132 517 46.2 74 22 49 5 150 18.7 Incident CHD absent FRS 0–7.5% 9456 197 119 4 9776 3.3 2415 82 4 0 2501 3.4 7.5–10% 346 178 195 17 736 75.8 117 154 77 1 349 55.9 10–20% 277 266 659 220 1422 53.7 9 69 296 15 389 23.9 20–100% 1 12 97 108 218 50.5 0 0 5 12 17 29.4 Total 10 080 653 1070 349 12 152 14.4 2541 305 382 28 3256 11.6 All individuals FINRISK FHS FRS+GRS FRS+GRS NRI (categorical) [95% CI] Total: 0.146 [0.088–0.20]; P < 1 × 10−6 NRI for events: 0.126 [0.068–0.183]; P = 1.9 × 10−5 NRI for non-events: 0.020 [0.014–0.027]; P < 1 × 10−6 Total: 0.033 [−0.037–0.103]; P = 0.35 NRI for events: 0.27 [−0.042–0.096]; P = 0.449 NRI for non-events: 0.006 [−0.005–0.018]; P = 0.281

All individuals FINRISK FHS FRS+GRS FRS+GRS NRI (categorical) [95% CI] Total: 0.146 [0.088–0.20]; P < 1 × 10−6 NRI for events: 0.126 [0.068–0.183]; P = 1.9 × 10−5 NRI for non-events: 0.020 [0.014–0.027]; P < 1 × 10−6 Total: 0.033 [−0.037–0.103]; P = 0.35 NRI for events: 0.27 [−0.042–0.096]; P = 0.449 NRI for non-events: 0.006 [−0.005–0.018]; P = 0.281 NRI (continuous) [95% CI] Total: 0.371 [0.285–0.457]; P < 1 × 10−6 NRI for events: 0.195 [0.111–0.280]; P < 6 × 10−6 NRI for non-events: 0.175 [0.159–0.192]; P < 1 × 10−6 Total: 0.249 [0.087–0.411]; P < 0.0026 NRI for events: 0.147 [−0.012–0.305]; P = 0.069 NRI for non-events: 0.102 [0.069–0.136]; P < 1 × 10−6 IDI (continuous) [95% CI] 0.028 [0.026–0.034]; P < 1 × 10−6 0.005 [0.002–0.008]; P < 0.00098 In FINRISK, 7 individuals of the 12 676 were excluded in this analysis due to missing clinical measurements. Table 4 Reclassification of incident CHD event risk within 10 years for combined ACC/AHA13 + GRS compared with ACC/AHA13 only, in the FINRISK and FHS cohorts FINRISK FHS ACC/AHA13+GRS ACC/AHA13+GRS 0–7.5% 7.5–10% 10–20% 20–100% Total Reclass % 0–7.5% 7.5–10% 10–20% 20–100% Total Reclass % All individuals ACC/AHA13 0–7.5% 9,588 211 144 7 9,950 3.6 2,513 78 7 0 2,598 3.3 7.5–10% 381 176 199 14 770 77.1 112 159 66 1 338 53.0 10–20% 279 275 755 271 1,580 52.2 7 67 308 32 414 25.6 20–100% 2 10 127 230 369 37.7 0 0 16 40 56 28.6 Total 10,250 672 1,225 522 12,699 15.2 2,632 304 397 73 3,406 11.3

otal Reclass % All individuals ACC/AHA13 0–7.5% 9,588 211 144 7 9,950 3.6 2,513 78 7 0 2,598 3.3 7.5–10% 381 176 199 14 770 77.1 112 159 66 1 338 53.0 10–20% 279 275 755 271 1,580 52.2 7 67 308 32 414 25.6 20–100% 2 10 127 230 369 37.7 0 0 16 40 56 28.6 Total 10,250 672 1,225 522 12,699 15.2 2,632 304 397 73 3,406 11.3 Incident CHD present ACC/AHA13 0–7.5% 118 16 17 1 152 22.4 67 8 0 75 75 10.7 7.5–10% 20 14 29 6 69 79.7 6 6 11 23 23 73.9 10–20% 15 29 104 60 208 50.0 1 6 34 46 46 26.1 20–100% 0 0 15 73 88 17.0 0 0 2 6 6 33.3 Total 153 59 165 140 517 40.2 74 20 47 150 150 26.0 Incident CHD absent ACC/AHA13 0–7.5% 9,470 195 127 6 9,798 3.3 2,446 70 7 0 2,523 3.1 7.5–10% 361 162 170 8 701 76.9 106 153 55 1 315 51.4 10–20% 264 246 651 211 1,372 62.6 6 61 274 27 368 25.5 20–100% 2 10 112 157 281 44.1 0 0 14 36 50 28.0 Total 10,097 613 1,060 382 12,152 14.1 2,558 284 350 64 3,256 10.7 All individuals FINRISK FHS ACC/AHA13+GRS ACC/AHA13+GRS NRI (categorical) [95% CI] Total: 0.120 [0.065–0.174]; P = 1.7 × 10−5 NRI for events: 0.097 [0.043–0.151]; P = 4.52 × 10−4 NRI for non-events: 0.023 [0.016–0.030]; P < 1 × 10−6 Total: 0.068 [−0.014–0.150]; P = 0.1 NRI for events: 0.060 [−0.021–0.141]; P = 0.147 NRI for non-events: 0.008 [−0.003–0.020]; P = 0.147 NRI (continuous) [95% CI] Total: 0.356 [0.270–0.442]; P < 1 × 10−6 NRI for events: 0.176 [0.091–0.261]; P = 4.79 × 10−5 NRI for non-events: 0.180 [0.164–0.196]; P < 1 × 10−6 Total: 0.255 [0.093–0.416]; P = 0.00197 NRI for events: 0.160 [0.002–0.318]; P = 0.047 NRI for non-events: 0.095 [0.061–0.128]; P < 1 × 10−6

NRI for non-events: 0.008 [−0.003–0.020]; P = 0.147 NRI (continuous) [95% CI] Total: 0.356 [0.270–0.442]; P < 1 × 10−6 NRI for events: 0.176 [0.091–0.261]; P = 4.79 × 10−5 NRI for non-events: 0.180 [0.164–0.196]; P < 1 × 10−6 Total: 0.255 [0.093–0.416]; P = 0.00197 NRI for events: 0.160 [0.002–0.318]; P = 0.047 NRI for non-events: 0.095 [0.061–0.128]; P < 1 × 10−6 IDI (continuous) [95% CI] 0.028 [0.021–0.034]; P < 1 × 10−6 0.005 [0.002–0.008]; P = 0.00184 In FINRISK, 7 individuals of the 12,676 were excluded in this analysis due to missing clinical measurements.

Total: 0.255 [0.093–0.416]; P = 0.00197 NRI for events: 0.160 [0.002–0.318]; P = 0.047 NRI for non-events: 0.095 [0.061–0.128]; P < 1 × 10−6 IDI (continuous) [95% CI] 0.028 [0.021–0.034]; P < 1 × 10−6 0.005 [0.002–0.008]; P = 0.00184 In FINRISK, 7 individuals of the 12,676 were excluded in this analysis due to missing clinical measurements. We next examined how variation in genomic risk translated into differences in cumulative lifetime risk of CHD, using Kaplan-Meier estimates stratified by GRS quintiles for men and women separately (Figure 3). As expected, cumulative risk increased with age for both sexes, with men displaying higher absolute risk than women. In both sexes there were substantial differences in cumulative risk between GRS groups with 1.7-fold (in FHS) to 3.2-fold (in FINRISK) higher cumulative risk by age 75 in those in the top quintile of GRS vs. bottom quintile. When considering clinically relevant levels of risk, FINRISK men in the top quintile of genomic risk achieved 10% cumulative risk 18 years earlier than those in the bottom quintile (ages 52 and 70, respectively), with a comparable difference of 12 years in FHS (ages 51 and 64). Women in the top quintile of genomic risk achieved 10% cumulative risk by age 69 (FINRISK) and 64 (FHS), whereas women in the bottom quintile did not achieve 10% risk by age 75 in FINRISK, or by age 73 in FHS. Estimated lifetime CHD risk in FINRISK showed no evidence of being affected by competing risks (incident CHD vs. non-CHD death) (see Supplementary material online, Supplementary Methods and Supplementary Figure S10). Similarly, a cause-specific competing-risk Cox analysis of the GRS in FINRISK, adjusting for geographical location and cohort, resulted in a similar hazard ratio as standard Cox analysis (HR = 1.70, 95% CI 1.61–1.86). Figure 3 Kaplan-Meier cumulative risk of incident CHD event by genomic risk group for men and women in the FINRISK and FHS cohorts. Showing the cumulative risk in quintiles 0–20%, 40–60%, 80–100%. The vertical bars along the x-axis indicate the age at which each risk group attains a cumulative CHD risk of 10%. Dashed lines indicate 95% CI.

cumulative risk of incident CHD event by genomic risk group for men and women in the FINRISK and FHS cohorts. Showing the cumulative risk in quintiles 0–20%, 40–60%, 80–100%. The vertical bars along the x-axis indicate the age at which each risk group attains a cumulative CHD risk of 10%. Dashed lines indicate 95% CI. We next sought to investigate to what degree high genomic risk for CHD could be compensated for by low levels of clinical risk factors at baseline, and vice-versa. When considering baseline smoking status in both FINRISK and FHS, Kaplan-Meier analysis showed a substantial increase in cumulative risk of CHD in men who smoked and were also in the top quintile of genomic risk, relative to either non-smokers or smokers at low genomic risk (Figure 4 for FINRISK and see Supplementary material online, Figure S11 for FHS). Similar but weaker trends were observed for women in the top vs. bottom quintiles of genomic risk. To test whether there was evidence for smoking affecting CHD hazard differently based on an individual’s genomic background, we used a Cox model allowing for an interaction term between the GRS and smoking; the interaction was not statistically significant in FINRISK (P = 0.91) and FHS (P = 0.49). Figure 4 Kaplan-Meier curves for incident CHD event risk stratified by GRS quintiles and smoking status at baseline, for men and women in the FINRISK cohorts.

x model allowing for an interaction term between the GRS and smoking; the interaction was not statistically significant in FINRISK (P = 0.91) and FHS (P = 0.49). Figure 4 Kaplan-Meier curves for incident CHD event risk stratified by GRS quintiles and smoking status at baseline, for men and women in the FINRISK cohorts. We also examined the potential compensatory effects of baseline systolic blood pressure and total cholesterol, divided as tertiles of high, medium, and low levels (see Supplementary material online, Figures S12 and S13). For both systolic blood pressure and total cholesterol, we observed the expected trends in CHD risk for high, medium and low levels. However, males with high vs. low levels of systolic blood pressure or total cholesterol showed greater absolute CHD risk if they were in the top vs. bottom quintiles of genomic risk. Notably, in both FINRISK and FHS, women in the bottom quintile of genomic risk showed smaller differences in cumulative CHD risk when stratified by smoking. For tertiles of systolic blood pressure or total cholesterol, low genomic risk women in FINRISK showed similarly small differences in risk, but the effects in FHS for this subgroup were not consistent. Cox models allowing for interactions between the GRS and systolic blood pressure or total cholesterol did not show statistically significant interactions in either FINRISK or FHS (P > 0.2 for all).

women in FINRISK showed similarly small differences in risk, but the effects in FHS for this subgroup were not consistent. Cox models allowing for interactions between the GRS and systolic blood pressure or total cholesterol did not show statistically significant interactions in either FINRISK or FHS (P > 0.2 for all). Discussion We have generated a GRS for CHD based on 49 310 SNPs and, using three prospective FINRISK and two FHS prospective cohorts, demonstrated that the GRS is associated with incident CHD events independently of established and widely-used clinical risk scores or individual CHD risk factors, including family history. Secondary validation in a familial hypercholesterolemia study (ARGOS) showed that GRS was also associated with CHD in this group of high-risk individuals. Subsequently, combining the GRS with established risk scores improved 10-year CHD risk prediction in FINRISK and FHS. We have also shown that the GRS can be leveraged to achieve meaningful lifetime CHD risk stratification, and that the impact of traditional CHD risk factors such as smoking, blood pressure, and cholesterol, vary substantially depending on the underlying genetic risk, thus offering the potential for both earlier and more targeted preventative efforts.

can be leveraged to achieve meaningful lifetime CHD risk stratification, and that the impact of traditional CHD risk factors such as smoking, blood pressure, and cholesterol, vary substantially depending on the underlying genetic risk, thus offering the potential for both earlier and more targeted preventative efforts. A distinctive feature of our analysis compared with several previous prospective studies11,29,30 examining the predictive utility of GRS for incident CHD is that the best predictive model was achieved here with SNPs that did not necessarily reach genome-wide or even statistical significance in previous GWA studies. The GRS outperformed other smaller SNP models, and shows greater promise in CHD prediction between top and bottom GRS quintiles than a recently published study testing a genetic risk score of 50 SNPs in Scandinavians30 (GRS50 HR = 1.92 vs. GRS49K HR = 4.51). Genome-wide SNP models have been applied successfully to other heritable human traits which seem to follow an “infinitesimal” genetic architecture, such as height.18 These results highlight the differing goals of GWAS and of genomic prediction: the stringent detection of causal genetic variants involved in the disease process vs. the construction of a model that robustly and maximally predicts future disease. While stringent procedures for minimizing the false positive rate of associated loci in GWAS are appropriate, these concerns are less relevant in construction of GRSs, especially when there are a large number of weakly correlated SNPs20 and when rigorous internal and external validation is performed.

ts future disease. While stringent procedures for minimizing the false positive rate of associated loci in GWAS are appropriate, these concerns are less relevant in construction of GRSs, especially when there are a large number of weakly correlated SNPs20 and when rigorous internal and external validation is performed. While population stratification is a potential confounder of genomic prediction studies, our use of a large worldwide multi-ethnic meta-analysis to develop the GRS together with two fully independent prospective validation datasets and three independent case/control datasets minimizes this potential. Our GRS was constructed from the CARDIoGRAMplusC4D stage-2 meta-analysis and the FINRISK and FHS individuals are both independent of that study and of broadly European ancestry; thus it is unlikely that the GRS is substantially confounded by fine-scale population structure within these cohorts. Further, the LD-thinning threshold to maximize prediction was determined in the WTCCC and MIGen datasets prior to applying the GRS to ARGOS, FINRISK, or FHS. Nevertheless, for some measures, GRS gains were less pronounced in FHS than in FINRISK. This may partly be due to the different definitions of CHD in these studies, to differences in environmental exposures, or to differences in genetic effects.31 In addition, the FRS was developed in the FHS, leading to potential over-estimation of its association with CHD in the current analysis. Hence, there may be benefit from future development of population-specific GRSs, which may yield greater predictive power within each population.

to differences in genetic effects.31 In addition, the FRS was developed in the FHS, leading to potential over-estimation of its association with CHD in the current analysis. Hence, there may be benefit from future development of population-specific GRSs, which may yield greater predictive power within each population. The association of the GRS with incident CHD was not substantially attenuated by traditional risk factors or clinical risk scores derived from these risk factors. Furthermore, the GRS was strongly associated with CHD in a study consisting purely of individuals with familial hypercholesterolemia. These results suggest that genomic risk exerts its effect on CHD risk through molecular pathways that are largely independent of the effects of cholesterol, systolic blood pressure, and smoking. A hitherto unresolved question has been the extent to which a family history would capture any information that may be provided through genetic analysis. Here, we clearly demonstrate the superior performance of direct genetic information over self-reported family history of CHD, which is often incomplete and imprecise in practice and is influenced by family size and competing causes of death.

y would capture any information that may be provided through genetic analysis. Here, we clearly demonstrate the superior performance of direct genetic information over self-reported family history of CHD, which is often incomplete and imprecise in practice and is influenced by family size and competing causes of death. While we observed improvements in discrimination (C-index) resulting from adding the GRS to the clinical risk scores when considering adults of all ages, the improvements were substantially higher in older individuals (>60 years old). Rather than being driven by age-related differences in the effect of the GRS, these results are likely driven by differences in the clinical risk scores between the younger and older adults. Unlike the GRSs, the clinical risk scores showed substantial differences across ages, driven by temporal changes in the underlying risk factors as well as age itself. Beyond the aims of identifying older adults with high CHD risk, the invariance of genomic risk makes it particularly useful for CHD risk prediction earlier in life, in young adulthood or before, when traditional risk factors are typically not measured and less likely to be informative of risk later in life.

age itself. Beyond the aims of identifying older adults with high CHD risk, the invariance of genomic risk makes it particularly useful for CHD risk prediction earlier in life, in young adulthood or before, when traditional risk factors are typically not measured and less likely to be informative of risk later in life. Our analyses focused on two clinical scores, the FRS and ACC/AHA13. While other scores exist, for example the SCORE system,32 we elected to use the FRS and ACC/AHA13 due to their widespread use and the fact that the FINRISK cohorts were a major contributor to the SCORE analysis, potentially biasing the analysis in FINRISK, in the same way that FRS seems to be biased towards the FHS, inflating its predictive power of the clinical risk scores there relative to the reference model.

13 due to their widespread use and the fact that the FINRISK cohorts were a major contributor to the SCORE analysis, potentially biasing the analysis in FINRISK, in the same way that FRS seems to be biased towards the FHS, inflating its predictive power of the clinical risk scores there relative to the reference model. Stratifying individual baseline smoking, systolic blood pressure, and total cholesterol levels measures into genomic risk groups revealed substantial differences in cumulative risk patterns. Importantly, this demonstrates that improved lifestyle may compensate for the innate increased CHD risk captured by the GRS. For men with high genomic risk, modifiable risk factors showed large effects on cumulative CHD risk. For women, the observed impacts of smoking, systolic blood pressure, and total cholesterol were low or not detectable in the low genomic risk group, particularly in FINRISK, however, we could not determine whether this was due to inadequate statistical power or other biological effects and further studies in larger cohorts of women are necessary to determine any clinical implications.

, and total cholesterol were low or not detectable in the low genomic risk group, particularly in FINRISK, however, we could not determine whether this was due to inadequate statistical power or other biological effects and further studies in larger cohorts of women are necessary to determine any clinical implications. Our results, if validated in further studies and across different populations, suggest a potential paradigmatic shift in the current CHD screening strategy which has existed for over 40 years—namely determination of genomic risk at an early stage with screening later in life through traditional clinical risk scores to complement background genomic risk. Based on early genomic risk stratification, individuals at higher risk may benefit from earlier engagement with nutritionists, exercise regimes, smoking cessation programs or be initiated early on medical interventions such as statin therapy or blood-pressure lowering medications to minimize future CHD risk. In this context it is notable that Mega et al.29 recently demonstrated that the GRS of 27 CHD-associated SNPs better predicted which individuals would benefit most, both in relative and absolute terms, from statin treatment. In a study of type 2 diabetes, Florez et al.32 has shown that the effects of increased genetic susceptibility to disease can be ameliorated by lifestyle (diet and exercise) and therapeutic (metformin) interventions. Similar possibilities exist for CHD, whereby early targeted prevention strategies based on genomic CHD risk may be implemented well in advance of clinical risk scores attaining predictive capacity at later ages.33 Such early risk stratification will offer increased efficiency in allocating both therapeutic resources and lifestyle modifications with the potential for subsequent delay of onset of traditional risk factors and incident CHD risk.

d well in advance of clinical risk scores attaining predictive capacity at later ages.33 Such early risk stratification will offer increased efficiency in allocating both therapeutic resources and lifestyle modifications with the potential for subsequent delay of onset of traditional risk factors and incident CHD risk. While our study demonstrates both the independent and incremental predictive power provided by our GRS, it is important to note that even when combined with such scores, the overall positive predictive value still remains modest for an acceptable negative predictive value (see Supplementary material online, Figure S14a). Furthermore, despite overall improved reclassification of 10 years risk, some individuals who went on to develop an incident event were reclassified at a lower risk by the addition of the GRS compared with their initial classification using a clinical score (Tables 3 and 4), emphasizing the limitations of the current GRS. The magnitude of the GRS effect was weaker in FHS than in the other datasets examined (FINRISK, WTCCC-CAD, MIGen-Harps, and ARGOS; Table 2). In addition to potential technical and clinical FHS differences discussed above, these results suggest that the benefit and clinical utility of the GRS may vary between populations; further evaluation in large prospective studies of varying ancestry will be required in order to assess these differences and how best to account for them in risk prediction. In this context, it should be noted that our GRS based on a starting list of 79 128 common SNPs tested by the CARDIOGRAMplusC4D consortium could be further improved. Future studies that construct GRSs using increased sample sizes and capturing the full spectrum of common and rare variants9,34 will likely provide additional gains in prediction and risk stratification.

based on a starting list of 79 128 common SNPs tested by the CARDIOGRAMplusC4D consortium could be further improved. Future studies that construct GRSs using increased sample sizes and capturing the full spectrum of common and rare variants9,34 will likely provide additional gains in prediction and risk stratification. In summary, this study has demonstrated the potential clinical utility of genome-scale GRS for CHD, both for early identification of individuals at increased CHD risk and for complementing existing clinical risk scores. Given recent advances and reduced cost of genotyping microarrays and sequencing-based technologies and their cost efficiency, determination of genome-wide SNP variants (including the 49 310 SNPs used here) is no longer beyond the realm of clinical application. In terms of technical feasibility, genome-wide genotyping of hundreds of thousands of SNPs is now both reliable and cost effective (<US$70 in bulk), and clinically certified genotyping services are now becoming available. Statistical SNP imputation will further expand the number of SNPs to an order of several million. Additionally, germline genotyping is a one-time cost for each individual. Further validation and cost-benefit analyses will be required in order to establish how this technology is deployed in clinical settings. Supplementary material Supplementary material is available at European Heart Journal online.

In summary, this study has demonstrated the potential clinical utility of genome-scale GRS for CHD, both for early identification of individuals at increased CHD risk and for complementing existing clinical risk scores. Given recent advances and reduced cost of genotyping microarrays and sequencing-based technologies and their cost efficiency, determination of genome-wide SNP variants (including the 49 310 SNPs used here) is no longer beyond the realm of clinical application. In terms of technical feasibility, genome-wide genotyping of hundreds of thousands of SNPs is now both reliable and cost effective (<US$70 in bulk), and clinically certified genotyping services are now becoming available. Statistical SNP imputation will further expand the number of SNPs to an order of several million. Additionally, germline genotyping is a one-time cost for each individual. Further validation and cost-benefit analyses will be required in order to establish how this technology is deployed in clinical settings. Supplementary material Supplementary material is available at European Heart Journal online. Supplementary Data Acknowledgements We thank Julie Simpson, Melbourne School of Population Health and Global Health (University of Melbourne), for advice regarding survival analyses.

In summary, this study has demonstrated the potential clinical utility of genome-scale GRS for CHD, both for early identification of individuals at increased CHD risk and for complementing existing clinical risk scores. Given recent advances and reduced cost of genotyping microarrays and sequencing-based technologies and their cost efficiency, determination of genome-wide SNP variants (including the 49 310 SNPs used here) is no longer beyond the realm of clinical application. In terms of technical feasibility, genome-wide genotyping of hundreds of thousands of SNPs is now both reliable and cost effective (<US$70 in bulk), and clinically certified genotyping services are now becoming available. Statistical SNP imputation will further expand the number of SNPs to an order of several million. Additionally, germline genotyping is a one-time cost for each individual. Further validation and cost-benefit analyses will be required in order to establish how this technology is deployed in clinical settings. Supplementary material Supplementary material is available at European Heart Journal online. Supplementary Data Acknowledgements We thank Julie Simpson, Melbourne School of Population Health and Global Health (University of Melbourne), for advice regarding survival analyses. Funding National Health and Medical Research Council Early Career Fellowship (1090462 to G.A.); National Health and Medical Research Council and the National Heart Foundation of Australia (1061435 and 1062227 to M.I.); Finnish Foundation for Cardiovascular Research to V.S; British Heart Foundation and NIHR to N.J.S.; AP and SR are supported by the Academy of Finland (grant no. 251704, 286500, 293404 to AP, and 251217, 285380 to SR), Juselius Foundation, Finnish Foundation for Cardiovascular Research, NordForsk e-Science NIASC (grant no 62721) and Biocentrum Helsinki (to SR). The MI Genetics (MIGen) Consortium Study was funded by the National Heart, Lung, and Blood Institute of the United States National Institutes of Health (R01 HL087676). Genotyping was partially funded by The Broad Institute Center for Genotyping and Analysis, which was supported by grant U54 RR02027 from the National Center for Research Resources. This study makes use of data generated by the Wellcome Trust Case-Control Consortium. A full list of the investigators who contributed to the generation of the data is available from www.wtccc.org.uk. Funding for the project was provided by the Wellcome Trust under award 076113 and 085475. The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 261433 (Biobank Standardisation and Harmonisation for Research Excellence in the European Union—BioSHaRE-EU). We are grateful to the CARDIoGRAMplusC4D consortium for making their large-scale genetic data available. A list of members of the consortium and the contributing studies is available at www.cardiogramplusc4d.org.

1433 (Biobank Standardisation and Harmonisation for Research Excellence in the European Union—BioSHaRE-EU). We are grateful to the CARDIoGRAMplusC4D consortium for making their large-scale genetic data available. A list of members of the consortium and the contributing studies is available at www.cardiogramplusc4d.org. Ethics statements The FINRISK data and samples are part of the THL Biobank (https://www.thl.fi/en/web/thlfi-en/topics/information-packages/thl-biobank), which has been approved by the Coordinating Ethical Committee of The Helsinki and Uusimaa Hospital District (decision # 238/13/03/00/2014). The FHS dataset was obtained from dbGaP (phs000007), approved by the University of Melbourne Health Sciences Human Ethics Sub-Committee (HREC 1442186). The ARGOS study consisted of cases and controls recruited from a large study based on the Dutch nationwide screening program for familial hypercholesterolemia. All patients gave informed consent and the ethics committee of the Academic Medical Center of Amsterdam approved the protocol (MEC 00/41#00.17.628). Conflict of interest: none declared.

Introduction Recent prevalence estimates suggest that at least 33.5 million persons are affected by atrial fibrillation (AF).1 Catheter ablation is increasingly offered to relieve AF-related symptoms,2–4 based on evidence illustrating its efficacy compared with antiarrhythmic drug therapy.5–9 There is less evidence supporting AF ablation in persistent AF, although small studies suggest better maintenance of sinus rhythm. Two years ago, the first multicentre trial comparing catheter ablation with cardioversion and antiarrhythmic drugs as first-line therapy for persistent AF has been reported and demonstrated more effective maintenance of sinus rhythm, as well as better quality of life, in patients randomized to catheter ablation.10 In recently published surveys, approximately one-third of AF ablation procedures were performed in patients with persistent or long-standing persistent AF.4 Here, we discuss recent data suggesting that our clinical ability to discriminate paroxysmal and persistent AF is limited, review the evidence supporting the use of catheter ablation in persistent AF, illustrate approaches to improve sinus rhythm maintenance by comprehensive cardiovascular risk reduction,11 discuss the value of different ablation strategies, and highlight the need for adequate validation of novel approaches to catheter ablation for persistent AF.

orting the use of catheter ablation in persistent AF, illustrate approaches to improve sinus rhythm maintenance by comprehensive cardiovascular risk reduction,11 discuss the value of different ablation strategies, and highlight the need for adequate validation of novel approaches to catheter ablation for persistent AF. What is persistent atrial fibrillation Persistent AF is defined as AF that persists without interruption for 7 days or longer.5,6 Whether patients who have been cardioverted during the first 7 days of an AF episode should be classified as persistent or paroxysmal has been defined differently in the USA and in Europe,5,6 but this only pertains to a small number of patients who receive early cardioversion. Patients who are in AF for >1 year are classed as long-standing persistent AF.5,6 Most patients who initially present with paroxysmal, self-terminating AF will progress to chronic forms of the arrhythmia (including permanent AF) or switch back and forth between paroxysmal and persistent or long-standing persistent AF.12,13 This simple observation already suggests that paroxysmal and persistent AF are not biologically distinct entities, but rather constitute different presentations of the same arrhythmia, loosely associated with different stages of the disease. Consistent with this general concept, patients diagnosed with persistent AF are generally older than those in paroxysmal AF, and present with more comorbidities.14,15 Over the last decade, the ECG-monitoring technology available to health care professionals and to the general public has tremendously advanced, thus profoundly improving our ability to differentiate AF patterns, reflected in consensus statements that informed regulatory bodies.7,16 Importantly, a recent analysis of continued atrial rhythm monitoring using implanted devices suggests that the AF pattern (or ‘AF burden’) does not differ too much between patients who have been clinically diagnosed with ‘paroxysmal’ or ‘persistent’ AF.17 Hence, the mere clinical classification of the AF pattern as ‘persistent’ may neither be sufficient justification to decide on a specific ablation strategy nor too powerful to predict the effectiveness of catheter ablation. It rather seems reasonable to assume that some patients with persistent AF respond to pulmonary vein isolation (PVI) as well as patients in paroxysmal AF, while others (with either AF pattern) are likely to develop recurrent AF.

cific ablation strategy nor too powerful to predict the effectiveness of catheter ablation. It rather seems reasonable to assume that some patients with persistent AF respond to pulmonary vein isolation (PVI) as well as patients in paroxysmal AF, while others (with either AF pattern) are likely to develop recurrent AF. Prevention of recurrent atrial fibrillation in patients scheduled for ablation of persistent atrial fibrillation It is well established that many patients already present with substantial atrial damage at the first episode of AF, and concomitant conditions such as hypertension, obesity, heart failure, chronic kidney disease, and others will cause ‘remodelling’ of the atria prior to occurrence of AF.11,18 On top of this ‘remodelling’, AF induces electrical and structural changes in the atria that occur within days to weeks (Figure 1). Nonetheless, some patients develop AF without detectable concomitant conditions late in life.19 Figure 1 Major health modifiers promoting recurrent atrial fibrillation (orange boxes) and the likely intermediary mechanisms causing atrial damage and leading to atrial fibrillation (open boxes, top part, health modifiers taken from Fabritz et al.89). The green boxes at the bottom illustrate interventions that can mitigate or reverse these effects. These ancillary interventions should be an integral part of the management of patients undergoing catheter ablation of persistent atrial fibrillation.

(open boxes, top part, health modifiers taken from Fabritz et al.89). The green boxes at the bottom illustrate interventions that can mitigate or reverse these effects. These ancillary interventions should be an integral part of the management of patients undergoing catheter ablation of persistent atrial fibrillation. A number of factors promote recurrent AF in patients, including those that undergo ablation of AF. Interestingly, time since the first diagnosis of AF is only a weak factor, while the time spent in continuous AF prior to ablation is a better predictor of outcome, where maintenance of sinus rhythm after ablation is less likely in patients spending >3 years continuously in AF prior to ablation (‘long-standing persistent’ AF).20 Other drivers of recurrent AF are related to clinical conditions that promote structural remodelling. Left atrial remodelling, which can be assessed invasively by quantifying areas of low-voltage left atrial signals, or with MR imaging,21–24 is largely determined by concomitant cardiovascular conditions and possibly by duration of AF, while other factors were derived from clinical information.25,26 It is worthwhile to consider that the effect of any AF ablation procedure is to induce further areas of delayed gadolinium enhancement.23,27

th MR imaging,21–24 is largely determined by concomitant cardiovascular conditions and possibly by duration of AF, while other factors were derived from clinical information.25,26 It is worthwhile to consider that the effect of any AF ablation procedure is to induce further areas of delayed gadolinium enhancement.23,27 While the natural ageing process is presently difficult to modulate, several factors associated with recurrent AF after catheter ablation can be modified by medical therapy or life style interventions (Table 1). Initial results of such life style interventions, e.g. regular exercise28 or systematic weight reduction in obese patients scheduled for AF ablation,29,30 are promising. Initial results also suggest that antihypertensive therapies that modulate central sympathetic tone (moxonidine) reduce recurrent AF after catheter ablation.31 Angiotensin converting enzyme inhibitors or sartans, while not effective in preventing recurrent AF in patients without structural heart disease in the short term,32,33 may have long-term beneficial effects for the primary prevention of AF.34 It has also been suggested that evidence-based therapy of heart failure with reduced ejection fraction can help to prevent AF.35 Likewise, adequate revascularization and treatment of mitral valve disease are likely to help stabilize sinus rhythm and restore atrial function (Figure 1). While it will be difficult to evaluate each of these interventions separately in outcome trials, it seems reasonable to integrate adequate treatment of modifiable cardiovascular conditions into the management of patients undergoing ablation of AF based on their known general cardiovascular benefits.36 In addition, it is common practice to use antiarrhythmic drugs for 3 months after catheter ablation of AF, including for persistent AF. Such treatment probably suppresses short-term recurrences of AF, but does not alter the mid-term recurrence rate.37 This practice seems reasonable. Table 1 Clinical factors that contribute to recurrent atrial fibrillation after catheter ablation and potential interventions that could reduce their impact on recurrent atrial fibrillation

obably suppresses short-term recurrences of AF, but does not alter the mid-term recurrence rate.37 This practice seems reasonable. Table 1 Clinical factors that contribute to recurrent atrial fibrillation after catheter ablation and potential interventions that could reduce their impact on recurrent atrial fibrillation Factor associated with recurrent AF Possible intervention Age None available Chronic kidney disease ? Diabetes Weight reduction, regular exercise (?) Obesity Weight reduction, regular exercise Hypertension Antihypertensive therapy, possibly including monoxidine and RAAS inhibition Heart failure Therapy of HFrEF with ACE inhibitors, β blockers, mineralocorticoid antagonists, etc. High ventricular rate Rate control therapy (?) Left atrial diameter None available Duration of continuous AF prior to ablation Early rhythm control therapy (?) ACE, angiotensin converting enzyme; AF, atrial fibrillation; HFrEF, heart failure with reduced ejection fraction; RAAS, renin–angiotensin aldosterone.

sts, etc. High ventricular rate Rate control therapy (?) Left atrial diameter None available Duration of continuous AF prior to ablation Early rhythm control therapy (?) ACE, angiotensin converting enzyme; AF, atrial fibrillation; HFrEF, heart failure with reduced ejection fraction; RAAS, renin–angiotensin aldosterone. Figure 2 Reconstruction of the left atrium (posterior view) showing the pulmonary veins and the left atrial appendage. Red dots illustrate the current approach of isolation of the pulmonary veins, in this case including a line between the two superior and inferior veins. Orange lines indicate additional linear ablation lesions that have been proposed to enhance the success rate of atrial fibrillation ablation (roof line, mitral isthmus line, ‘box’ lesions consisting of a roof/superior and inferior connection between the pulmonary vein isolation circles, and left atrial appendage isolation). The effectiveness of these additional ablation interventions will require evaluation in adequately sized and powered controlled trials.

(roof line, mitral isthmus line, ‘box’ lesions consisting of a roof/superior and inferior connection between the pulmonary vein isolation circles, and left atrial appendage isolation). The effectiveness of these additional ablation interventions will require evaluation in adequately sized and powered controlled trials. Pulmonary vein isolation prevents recurrent atrial fibrillation in some, but not all patients with persistent atrial fibrillation The first description that triggers in the pulmonary veins initiate AF and that their elimination by radio frequency ablation prevents AF has been a key disruptive discovery shaping AF ablation.38 Pulmonary vein isolation remains the cornerstone of AF ablation until today.7 Two-thirds of surveyed European centres perform PVI without additional ablation targets as a first-line therapy for persistent AF.39 Pulmonary vein isolation conveys a 60–80% rate of maintaining sinus rhythm after 1 year in patients who predominantly present with paroxysmal forms of AF, not different between centres mainly using cryoballoon and irrigated radio frequency ablation.40–42 The rhythm outcome in patients with persistent AF is more variable, but not much worse in some series (Table 2). The maintenance of sinus rhythm is not dramatically different between ‘persistent’ AF patients undergoing PVI alone compared with patients undergoing more extensive ablation approaches, including a risk to develop left atrial tachycardia with either approach (Table 2).40 Table 2 Controlled trials and selected observational data sets reporting sinus rhythm rates after catheter ablation of persistent atrial fibrillation

oing PVI alone compared with patients undergoing more extensive ablation approaches, including a risk to develop left atrial tachycardia with either approach (Table 2).40 Table 2 Controlled trials and selected observational data sets reporting sinus rhythm rates after catheter ablation of persistent atrial fibrillation Patients Intervention Control Sinus rhythm outcome Ablation Control Controlled trials Wazni47 70 (paroxysmal and persistent) CA AAD + CV 87%a 46%a Oral48 146 PVI + Amiodarone Amiodarone + CV 74%a 58%a Stabile49 137 (paroxysmal or persistent) CA: PVI + mitral line + CTI AAD 56% 10% Forleo50 70 (41 persistent) CA AAD + CV 80%a 43%a Jones51 52 CA Rate control 88%a n.a. Mont10 146 CA AAD + CV 70% 44% Verma43 589 PVI PVI + lines, PVI + CFAE 59% 46%; 49% Dong52 146 CA + lines (fix) CA (stepwise) 67% 60% Observational data sets Hunter (multi centre)53 586 (persistent) CA (1.8 mean procedures, mainly PVI) n.a. 60%a n.a. Scherr (single centre)54 150 CA (AF termination outcome) n.a. 65% n.a. Schreiber (single centre)55 549 CA (stepwise approach) n.a. 56% n.a. Haissaguerre56 103 CA (driver domains) 65%a n.a. AAD, antiarrhythmic drugs; CA, catheter ablation; CV, cardioversion; PVI, pulmonary vein isolation. aNumbers in italic indicate success rates without intensive ECG monitoring.

Patients Intervention Control Sinus rhythm outcome Ablation Control Controlled trials Wazni47 70 (paroxysmal and persistent) CA AAD + CV 87%a 46%a Oral48 146 PVI + Amiodarone Amiodarone + CV 74%a 58%a Stabile49 137 (paroxysmal or persistent) CA: PVI + mitral line + CTI AAD 56% 10% Forleo50 70 (41 persistent) CA AAD + CV 80%a 43%a Jones51 52 CA Rate control 88%a n.a. Mont10 146 CA AAD + CV 70% 44% Verma43 589 PVI PVI + lines, PVI + CFAE 59% 46%; 49% Dong52 146 CA + lines (fix) CA (stepwise) 67% 60% Observational data sets Hunter (multi centre)53 586 (persistent) CA (1.8 mean procedures, mainly PVI) n.a. 60%a n.a. Scherr (single centre)54 150 CA (AF termination outcome) n.a. 65% n.a. Schreiber (single centre)55 549 CA (stepwise approach) n.a. 56% n.a. Haissaguerre56 103 CA (driver domains) 65%a n.a. AAD, antiarrhythmic drugs; CA, catheter ablation; CV, cardioversion; PVI, pulmonary vein isolation. aNumbers in italic indicate success rates without intensive ECG monitoring. One of the most important recent studies in field of ablation of persistent AF was the Star AF 2 trial.43 This landmark study randomized 589 patients with persistent AF to PV isolation alone (N = 67 patients), to PVI plus linear ablation (N = 259 patients), or to PVI and ablation of continuous fractionated electrograms ablation (CFAE, N = 263 patients). The results of this study revealed no difference in outcomes of these three ablation strategies. After 18 months of follow-up, 59% of patients assigned to PVI alone were AF free, when compared with 49% of patients assigned to PVI plus CFAE ablation and 46% of patients assigned to PVI plus linear ablation. The lack of additional effects of CFAE ablation possibly came as less of a surprise as the lack of effects of linear lesions.44,45 The longer procedure duration of extended ablation procedures, associated with higher radiation exposure and possibly higher complication rates, should be considered in this context. Star AF 2 clearly supports the use of PVI without further ablation as the first-line therapy in patients with persistent AF, opening the possibility of catheter ablation of persistent AF using cryothermy balloons in the future.46 We propose that a group of patients with persistent AF respond as well to PVI as patients with paroxysmal AF.

upports the use of PVI without further ablation as the first-line therapy in patients with persistent AF, opening the possibility of catheter ablation of persistent AF using cryothermy balloons in the future.46 We propose that a group of patients with persistent AF respond as well to PVI as patients with paroxysmal AF. Targets for catheter ablation beyond pulmonary vein isolation The most recent AF ablation consensus document considered PVI the ‘cornerstone’ of AF ablation.7 The document also stated that additional ablation strategies should be considered when ablating persistent AF, and expressed a need for sufficiently powered multicentre trials comparing different AF ablation strategies. At that time there was no consensus as to which of these ablation strategies was optimal.

ablation.7 The document also stated that additional ablation strategies should be considered when ablating persistent AF, and expressed a need for sufficiently powered multicentre trials comparing different AF ablation strategies. At that time there was no consensus as to which of these ablation strategies was optimal. Prior to the seminal description of triggers in the pulmonary veins initiating AF,38 several skilled groups developed different sets of linear left and right atrial lesions in an attempt to prevent AF.57–61 Several linear lesions, e.g. around the mitral isthmus or a ‘roof line’ connecting the ablation lesions encircling the pulmonary veins, have been re-used as relevant adjuncts to PVI in persistent AF (Figure 2).54,56,62,63 Additional lesions that have been proposed are a ‘box’ encircling the posterior left atrium including all four pulmonary veins, and a line isolating the left atrial appendage (Figure 2). Initial reports, often comprising procedures done in a single centre and relying on a few dozen of patients, were promising, while larger, confirmatory studies often yielded higher recurrence rates. One commonly proposed ablation strategy at the time was the ‘stepwise’ approach to ablation of persistent AF proposed and championed by the Bordeaux group.64,65 Like other developments in the field that were mainly developed and evaluated in a small number of centres, this approach has never gained wide-spread use. A recent publication describing the long-term outcomes of 150 patients who underwent the stepwise approach to ablation of persistent AF using the stepwise approach was somewhat sobering. Arrhythmia-free survival rates after a single procedure were 35.3 ± 3.9, 28.0 ± 3.7, and 16.8 ± 3.2% at 1, 2, and 5 years, respectively.54

ent publication describing the long-term outcomes of 150 patients who underwent the stepwise approach to ablation of persistent AF using the stepwise approach was somewhat sobering. Arrhythmia-free survival rates after a single procedure were 35.3 ± 3.9, 28.0 ± 3.7, and 16.8 ± 3.2% at 1, 2, and 5 years, respectively.54 New ablation strategies for ablation of persistent AF have started to emerge. One of these new strategies involves the use of a multi-electrode basket catheter to map ‘rotors’, i.e. areas that are critical for maintenance of AF.66,67 Some, but not all, of the published outcomes using this basket-based rotor mapping-based approach have reported encouraging results.67,68 Non-invasive body surface potential mapping has been used by another group to identify such critical areas (described as ‘drivers’ or ‘focal sources’ by these researchers).69,70 Another strategy that has been developed is to homogenize areas of scar, using MRI or voltage mapping to identify areas of scar (see Figure 3 for illustrative examples of abnormal areas of low voltage in the left atrium). Once identified these areas of scar are ablated in an attempt to eliminate any potential re-entrant circuits.71 Experimental data suggest that the core of a rotor may often co-localize with areas of conduction block, in line with the behaviour of voltage vectors constructed from filtered electrograms. Hence, these two approaches may result in somewhat overlapping ablation lesions. The concept of targeting fractionated electrograms (CFAE) has been abandoned by many centres after disappointing results of controlled trials. These lesions are placed based on electrogram characteristics and do not follow a defined anatomical pattern. Figure 3 Examples of left atrial voltage maps (view onto the posterior left atrium) showing normal left atrial voltage (upper panel), confined areas of low left atrial voltage (lower left panel), and homogeneous reduction of left atrial electrogram voltage (lower right panel). Purple colour indicates areas with normal (>0.5 mV) amplitude of bipolar electrograms, red areas with low (≤0.2 mV) left atrial voltage.

rmal left atrial voltage (upper panel), confined areas of low left atrial voltage (lower left panel), and homogeneous reduction of left atrial electrogram voltage (lower right panel). Purple colour indicates areas with normal (>0.5 mV) amplitude of bipolar electrograms, red areas with low (≤0.2 mV) left atrial voltage. Whether any of the novel strategies listed above proves to be superior to PVI alone for ablation of persistent AF remains to be determined. Currently, a large variety of ablation strategies are being employed with a goal of obtaining preliminary data concerning whether these new ablation strategies are more effective than PV isolation alone. When interpreting results from studies evaluating novel ablation strategies, it seems important to recognise a major limitation of current catheter ablation interventions: Even when the PVI is performed in selected, highly experienced centres with a clear aim to achieve complete isolation, this is only achieved in a minority of patients.72 Hence, better technology is needed to achieve transmural lesions. This has implications not only for the evaluation of linear lesions but also for other ablation concepts. At some point these new ablation strategies will need to be compared head to head with PV isolation alone in sufficiently powered multicentre trials similar to the design of the Star-AF 2 Trial described above. Until that point, these new approaches must be considered experimental and their value unproven.

Whether any of the novel strategies listed above proves to be superior to PVI alone for ablation of persistent AF remains to be determined. Currently, a large variety of ablation strategies are being employed with a goal of obtaining preliminary data concerning whether these new ablation strategies are more effective than PV isolation alone. When interpreting results from studies evaluating novel ablation strategies, it seems important to recognise a major limitation of current catheter ablation interventions: Even when the PVI is performed in selected, highly experienced centres with a clear aim to achieve complete isolation, this is only achieved in a minority of patients.72 Hence, better technology is needed to achieve transmural lesions. This has implications not only for the evaluation of linear lesions but also for other ablation concepts. At some point these new ablation strategies will need to be compared head to head with PV isolation alone in sufficiently powered multicentre trials similar to the design of the Star-AF 2 Trial described above. Until that point, these new approaches must be considered experimental and their value unproven. Early rhythm control therapy More and more electrophysiologists are opting to perform AF ablation early in the course of the AF journey in an attempt to reduce AF burden and limit AF-induced atrial damage. Consistent with this trend are the recommendations by both the ESC AF Guidelines and also the AHA/ACC/HRS AF consensus document that catheter ablation of paroxysmal AF may be considered as first-line therapy11,73 based on patient preference and when performed in experienced centres.74 While published trials, relying on antiarrhythmic drug therapy and often accepting interruption of oral anticoagulation after sinus rhythm restoration, have not shown a prognostic benefit of rhythm control therapy over usual care,75 there is biological reason to believe that sinus rhythm maintenance could help to prevent these cardiovascular events.76–79 Intermittent periods of sinus rhythm may reverse some of the underlying adaptive processes (‘electrical’ and ‘structural’ remodelling).78,80 Activation patterns in the fibrillating atria are heterogeneous and highly variable over time.81 The complexity of electrical activity in AF, described as drivers,56 epicardial break through,82–84 rotor,85 or AF cycle length,86 increases with longer duration of continuous AF. A novelty of the rhythm control approach in In EAST—AFNET 4 is the mandate for an early rhythm control therapy intervention,87,88 informed by the observation that AF-induced atrial remodelling may facilitate recurrent AF during later stages of the disease. It remains to be seen if this early intervention approach bear can also help to prevent strokes and other major cardiovascular outcomes in AF patients.

dose/effort stress prior to deteriorating at higher workloads.72 Low dose stress TDI allows quantification of systolic function: Doppler tissue velocities increase more in hibernating myocardium than in dysfunctional tissue that does not show improvement in systolic function following revascularization (Figure 8).73,74 Cardiac magnetic resonance examination allows for accurate assessment of diastolic wall thickness and regional wall motion abnormalities. Late gadolinium enhancement CMR contributes only indirectly to the diagnosis of hibernation. On LGE CMR, hibernating myocardium has the same signal characteristics as normal myocardium, with signal uniformly nulled. Late gadolinium enhancement CMR therefore only excludes the presence of myocardial infarction as the cause of contractile dysfunction, suggesting hibernation as one of several potential causes. Absence of LGE in dysfunctional segments in ischaemic heart disease may predict functional recovery following revascularization (Figure 9).75 Cardiac magnetic resonance prediction of recovery can be further enhanced by combined use of functional, LGE, and dobutamine stress imaging. Hibernating myocardium has reduced resting function, the absence of LGE and a biphasic response to dobutamine stress.72 Finally, myocardial perfusion can be assessed and quantified by first pass CMR. Resting MBF (mL/min/g) is normal in hibernating myocardium, and hyperaemic blood flow is reduced with subsequent improvement following successful revascularization.76Figure 9 Increasing transmurality of late gadolinium enhancement predicts lack of response to revascularization in chronically ischaemic dysfunctional myocardium: Progressive transmurality of scar predicts lack of improvement in systolic function following revascularization. (A) 25% late gadolinium enhancement of the inferior wall and infer-septum (B) 50% late gadolinium enhancement in the infero-lateral wall becoming increasingly transmural inferiorly. (C) 100% late gadolinium enhancement of the inferior wall.

rhythm control therapy intervention,87,88 informed by the observation that AF-induced atrial remodelling may facilitate recurrent AF during later stages of the disease. It remains to be seen if this early intervention approach bear can also help to prevent strokes and other major cardiovascular outcomes in AF patients. Summary Catheter ablation is a reasonably effective intervention to achieve restoration and maintenance of sinus rhythm in patients with persistent AF. The evidence underpinning the use of catheter ablation is less strong for persistent and long-standing persistent AF than for paroxysmal AF. Acknowledging the need for further data, we suggest the following principles to guide catheter ablation of persistent AF (Figure 4): Many patients who are clinically classified as ‘persistent AF’ will have a similar AF patterns to other patients who are classified as ‘paroxysmal AF’. Catheter ablation should be considered for symptom relief in patients with persistent AF, especially after failed antiarrhythmic drug therapy. Pulmonary vein isolation is a reasonable and often sufficiently effective ablation strategy in patients undergoing a first catheter ablation of persistent AF. Additional ablation targets should in our view not routinely be pursued in the first procedure. Optimal management of concomitant cardiovascular conditions should be an integral part of rhythm control therapy in patients with persistent AF undergoing catheter ablation.

Pulmonary vein isolation is a reasonable and often sufficiently effective ablation strategy in patients undergoing a first catheter ablation of persistent AF. Additional ablation targets should in our view not routinely be pursued in the first procedure. Optimal management of concomitant cardiovascular conditions should be an integral part of rhythm control therapy in patients with persistent AF undergoing catheter ablation. Early rhythm control intervention has conceptual benefits, but needs evaluation in controlled trials. At present, it seems reasonable to not delay catheter ablation unduly after the decision for rhythm control therapy has been taken. Figure 4 Proposed stepwise approach to catheter ablation of patients with persistent atrial fibrillation emphasizing the need to isolate the pulmonary veins before applying further ablation techniques, and illustrating the integration of medical and life style interventions underpinning the effect of catheter ablation. This proposal integrates available evidence. We recognize the need to evaluate the best AF ablation strategy in different populations of patients with persistent atrial fibrillation. What next? There are development needs for this exciting and understudied area in clinical electrophysiology. Among them, the following appear of special relevance to us: There is a need to develop better technology to achieve safe and reliable transmural lesions in the left atrium. We need to identify clinical markers for different types of AF that allow to identify persistent AF patients who will benefit from catheter ablation.89

What next? There are development needs for this exciting and understudied area in clinical electrophysiology. Among them, the following appear of special relevance to us: There is a need to develop better technology to achieve safe and reliable transmural lesions in the left atrium. We need to identify clinical markers for different types of AF that allow to identify persistent AF patients who will benefit from catheter ablation.89 Research into reduction of ablation complications and development of simple techniques to achieve PVI46 is needed. Several additional ablation strategies have been evaluated and abandoned, while others (including promising novel ideas) are in need of adequate evaluation in sufficiently powered controlled clinical trials, e.g. in patients with recurrent AF after successful PVI. This requires international cooperation of major ablation centres to allow the conduct of properly powered controlled trials. Authors' contributions P.K. handled funding and supervision, drafted the manuscript. P.K., H.C. (literature review) acquired the data, conceived and designed the research. H.C. made critical revision of the manuscript for key intellectual content. The concept of this review article is based on a presentation given during the 2015 ESC congress in London. Funding The writing of this review was partially supported by European Union [grant agreement no 633193 (CATCH ME, Characterizing Atrial fibrillation by Translating its Causes into Health Modifiers in the Elderly) to P.K.], Fondation Leducq (to P.K.), and British Heart Foundation (FS/13/43/30324 to P.K.).

Authors' contributions P.K. handled funding and supervision, drafted the manuscript. P.K., H.C. (literature review) acquired the data, conceived and designed the research. H.C. made critical revision of the manuscript for key intellectual content. The concept of this review article is based on a presentation given during the 2015 ESC congress in London. Funding The writing of this review was partially supported by European Union [grant agreement no 633193 (CATCH ME, Characterizing Atrial fibrillation by Translating its Causes into Health Modifiers in the Elderly) to P.K.], Fondation Leducq (to P.K.), and British Heart Foundation (FS/13/43/30324 to P.K.). Conflict of interest: H.C. reports personal fees from Medtronic, grants and personal fees from St Jude Medical, personal fees from Abbott Medical, outside the submitted work. P.K reports grants from British Heart Foundation and European Union, during the conduct of the study; personal fees from several industry partners, grants from several industry partners (institutional grants), grants from several public funders and charities (DZHK/BMBF, EU, BHF, and Leducq Foundation), outside the submitted work; In addition, P.K. has a patent AF therapy pending to University of Birmingham, and a patent Markers for AF pending to University of Birmingham.

Introduction Studies have demonstrated the benefits of pharmacological1–8 and device interventions9 for patients with left ventricular systolic dysfunction (LVSD) of ischaemic origin. While there are clear theoretical benefits in improving blood supply to large areas of dysfunctional ‘viable’ myocardium, several recent clinical trials have failed to demonstrate improved outcomes following revascularization.10–12 One explanation for these results may be the terminology employed to describe myocardial health. In particular the terms ‘viable’ and ‘hibernating’ are often used interchangeably and sometimes incorrectly. ‘Viable’ is a summative term used to describe a range of myocardial states including both normal and diseased segments that nonetheless contain a substantial number of cardiac myocytes. Viable myocardial segments can thus include partial thickness scar with fairly normal cardiac myocyte function in the residual portion or myocytes that exhibit contractile dysfunction or failure. On the other hand, ‘hibernating’ myocardium refers only to the latter subgroup and can be narrowly defined as chronically dysfunctional, ischaemic myocardium that recovers contractile function following improvement in myocardial perfusion limitation. It follows that assessment of ‘viability’ can be made, to a large degree, prior to revascularization, but identification of ‘hibernation’ may only be retrospective. Furthermore, identifying myocardium as ‘viable’ does not necessarily imply ‘hibernation’ or functional recovery following revascularization, as is sometimes assumed.

that assessment of ‘viability’ can be made, to a large degree, prior to revascularization, but identification of ‘hibernation’ may only be retrospective. Furthermore, identifying myocardium as ‘viable’ does not necessarily imply ‘hibernation’ or functional recovery following revascularization, as is sometimes assumed. A second challenge in the taxonomy of myocardial dysfunction is that it may be assessed with a range of non-invasive imaging techniques. The available techniques differ fundamentally in methodology, the properties of acquired images and the tests' limitations. Therefore, discrepancies between the modalities are inevitable, making consistency in the definition of myocardial states across imaging fields challenging. Understanding these differences is imperative, especially in the design and interpretation of studies to determine the role of revascularization in patients with long-standing LVSD. This article creates a clinical taxonomy for the different states of myocardium that may exist and coexist in ischaemic heart disease and cardiomyopathy with reference to histology and non-invasive imaging techniques. Available imaging techniques Non-invasive imaging techniques assess different aspects of myocardial health: function and morphology, perfusion, metabolism, and tissue characterization (Table 1). Table 1 Summary of key imaging modalities, and aspects assessed in defining myocardial Morphology and resting function Stress function/contractile reserve Perfusion Metabolism Tissue characterization Echo Strengths Readily assessed in a range of situations.

Available imaging techniques Non-invasive imaging techniques assess different aspects of myocardial health: function and morphology, perfusion, metabolism, and tissue characterization (Table 1). Table 1 Summary of key imaging modalities, and aspects assessed in defining myocardial Morphology and resting function Stress function/contractile reserve Perfusion Metabolism Tissue characterization Echo Strengths Readily assessed in a range of situations. Widely accessible Doppler, tissue Doppler, and GLS provide additional important indices Contrast improves accuracy* Physiological or pharmacological stressors may be employed Crude visual assessment of scar, Maybe quantified with integrated background scatter (not in routine clinical practice) Weaknesses Limited by habitus and lung disease Greater inter-observer variation and less reproducible than CMR** Affected by tethering in the presence of multiple wall motion abnormalities Microbubble perfusion remains predominantly a research tool Potential lack of local expertise Not assessed CMR Strengths Multiplanar imaging with excellent reproducibility# Myocardial mechanics may be assessed as per tissue Doppler Physiological or pharmacological stressors may be employed Highly reproducible Visual assessment Quantification of perfusion reserve possible Excellent spatial resolution Focal scar identified with LGE Quantification of fibrosis (T1 mapping) and oedema (T2 mapping) Weaknesses Less accessible than echo Unsuitable in critical illness Limited by presence of some implantable cardiac devices & extreme obesity Stress limitations similar to echo Exercise stressless practical Availability Absolute quantification of perfusion Standard perfusion not ‘whole heart’ coverage Gadolinium CI if eGFR < 30 mL/min/1.73 m2 Not assessed in clinical routine practice, MRS available in some centres SPECT Strengths Assessment of systolic function possible during gated perfusion examination Whole heart coverage Perfusion defect size may be quantified Weakness Other aspects of cardiac function not assessed Not assessed Quantitative assessment not possible Lower spatial resolution than PET and CMR Potential imitations in balanced ischaemia‡‡‡ Ionising radiation Not assessed Inferred only—scar over estimated‡‡ PET Strengths Gold standard for perfusion quantification Excellent spatial resolution Whole heart coverage Non-invasive assessment of carbohydrate and lipid metabolism possible Weaknesses Not usually assessed Not assessed Exposure to ionising radiation Exposure to ionising radiation Tissue composition inferred from metabolism/perfusion findings◊ Symbols: *136, **137, #16,17, ‡‡139,

t spatial resolution Whole heart coverage Non-invasive assessment of carbohydrate and lipid metabolism possible Weaknesses Not usually assessed Not assessed Exposure to ionising radiation Exposure to ionising radiation Tissue composition inferred from metabolism/perfusion findings◊ Symbols: *136, **137, #16,17, ‡‡139, ‡‡‡140, ◊66.

t spatial resolution Whole heart coverage Non-invasive assessment of carbohydrate and lipid metabolism possible Weaknesses Not usually assessed Not assessed Exposure to ionising radiation Exposure to ionising radiation Tissue composition inferred from metabolism/perfusion findings◊ Symbols: *136, **137, #16,17, ‡‡139, ‡‡‡140, ◊66. Function and morphology Echocardiography provides real-time cine assessment of cardiac structure, myocardial wall thickness, and contractile function. Tissue Doppler imaging (TDI) adds quantification of myocardial motion and global longitudinal strain can be used to detect subtle systolic and diastolic abnormalities that may have greater prognostic value compared with LVEF.13,14 Cardiac magnetic resonance (CMR) produces high-resolution morphological and cine images of the heart in unrestricted imaging planes and with high accuracy and reproducibility.15,16 In addition, cine CMR with tissue tagging or feature tracking analysis allows quantitative assessment of segmental contractile function. Both echocardiography and CMR can be combined with pharmacological (dobutamine) or physiological (treadmill/static bike) stress to detect ischaemia and viable, dysfunctional myocardium17,18 through demonstration of contractile reserve. Single photon emission computed tomography (SPECT) and positron emission tomography (PET) can provide information on global and regional contractile function, but at a lower resolution than echocardiography and CMR. Nuclear techniques are less suited than echocardiography and CMR to measure wall thickness. With all imaging techniques, assessment of segmental contractile reserve is significantly limited by adjacent wall motion abnormalities due to ‘tethering’.19

ile function, but at a lower resolution than echocardiography and CMR. Nuclear techniques are less suited than echocardiography and CMR to measure wall thickness. With all imaging techniques, assessment of segmental contractile reserve is significantly limited by adjacent wall motion abnormalities due to ‘tethering’.19 Perfusion Myocardial contrast echocardiography is used to measure perfusion by observing the distribution of intravascular microbubbles in the myocardium and provides a surrogate of myocardial blood flow.20 Myocardial perfusion CMR tracks the myocardial passage of predominantly extracellular paramagnetic contrast agents21 and allows estimation of absolute myocardial flow (MBF) and myocardial perfusion reserve (MPR).22 Single photon emission computed tomography and PET detect perfusion defects through a reduction of signal from radionucleotide bound perfusion tracers in the region of interest23; PET is considered the ‘gold-standard’ for quantitation of MBF and MPR (Figures 1–3). Figure 1 Intra-venous contrast agent use in echocardiography. Left, rest images have poorly defined endocardial border. Right, following administration of intra-venous contrast agent the apical wall motion defect is evident. Taken from Plana et al.141 Figure 2 Patterns of late gadolinium enhancement by CMR seen in different pathologies. (A) Sub-endocardial infero-lateral enhancement in myocardial infarction. (B) Anterior mid wall enhancement in dilated cardiomyopathy. (C) Lateral epicardial and pericardial enhancement in myo-pericarditis.

Perfusion Myocardial contrast echocardiography is used to measure perfusion by observing the distribution of intravascular microbubbles in the myocardium and provides a surrogate of myocardial blood flow.20 Myocardial perfusion CMR tracks the myocardial passage of predominantly extracellular paramagnetic contrast agents21 and allows estimation of absolute myocardial flow (MBF) and myocardial perfusion reserve (MPR).22 Single photon emission computed tomography and PET detect perfusion defects through a reduction of signal from radionucleotide bound perfusion tracers in the region of interest23; PET is considered the ‘gold-standard’ for quantitation of MBF and MPR (Figures 1–3). Figure 1 Intra-venous contrast agent use in echocardiography. Left, rest images have poorly defined endocardial border. Right, following administration of intra-venous contrast agent the apical wall motion defect is evident. Taken from Plana et al.141 Figure 2 Patterns of late gadolinium enhancement by CMR seen in different pathologies. (A) Sub-endocardial infero-lateral enhancement in myocardial infarction. (B) Anterior mid wall enhancement in dilated cardiomyopathy. (C) Lateral epicardial and pericardial enhancement in myo-pericarditis. Figure 3 Radionuclide perfusion imaging: (A) Hyperaemic stress single photon emission computed tomography and (B) positron emission tomography imaging (B) of the same patient. (A) shows an apparent inferior perfusion defect not present in (B) suggesting single photon emission computed tomography attenuation artefact.50

Figure 3 Radionuclide perfusion imaging: (A) Hyperaemic stress single photon emission computed tomography and (B) positron emission tomography imaging (B) of the same patient. (A) shows an apparent inferior perfusion defect not present in (B) suggesting single photon emission computed tomography attenuation artefact.50 Metabolism The principal method for metabolic myocardial imaging is currently PET, which allows myocardial metabolic substrate utilization to be characterized and quantified. Magnetic resonance spectroscopy is a less widely used method to interrogate myocardial energetics. Tissue characterization Non-invasive tissue characterization is predominantly performed with CMR. Late gadolinium enhancement (LGE) CMR allows identification of scar and focal fibrosis. T1 and extracellular volume (ECV) measurement appears a useful and reproducible surrogate for diffuse myocardial fibrosis,24,25 while T2-weighted CMR provides assessment of myocardial oedema. Integrated back-scatter echocardiography, which provides assessment of tissue fibrosis might also be applied in this context. Future directions Hybrid systems that combine anatomical with perfusion or metabolic imaging (e.g. PET/MR) are becoming available whilst molecular imaging may soon allow identification of specific biological processes. Taxonomy for myocardial segments Normal Ischaemic Reversible ischaemia (a) Acute prolonged ischaemia (b) Chronic intermittent ischaemia Stunning Hibernation Infarction Myopathic Normal

Future directions Hybrid systems that combine anatomical with perfusion or metabolic imaging (e.g. PET/MR) are becoming available whilst molecular imaging may soon allow identification of specific biological processes. Taxonomy for myocardial segments Normal Ischaemic Reversible ischaemia (a) Acute prolonged ischaemia (b) Chronic intermittent ischaemia Stunning Hibernation Infarction Myopathic Normal Definition: Normal myocardium is by definition viable and metabolically active with normal contractile function and exhibiting contractile reserve in response to increased demand.

Taxonomy for myocardial segments Normal Ischaemic Reversible ischaemia (a) Acute prolonged ischaemia (b) Chronic intermittent ischaemia Stunning Hibernation Infarction Myopathic Normal Definition: Normal myocardium is by definition viable and metabolically active with normal contractile function and exhibiting contractile reserve in response to increased demand. Metabolism: Normal cardiac function, including contraction, relaxation, and ionic regulation, is dependent upon adenosine triphosphate (ATP) metabolism. The majority of ATP consumption occurs in myo-fibrils throughout the cardiac cycle. Further ATP is used to regulate sarcolemmal calcium and, at the membrane, Na+/K+ ATPase transporter.26,27 The heart consumes ATP rapidly and is dependent upon constant renewal of ATP, which in turn is dependent upon creatine phosphate levels. Were ATP production to cease and consumption to continue unchecked, cardiac stores would be depleted in ∼10–15 s.26 There are three main pathways by which ATP is synthesised: fatty acid oxidation, ketone body, and carbohydrate metabolism. Fatty acid oxidation yields the most ATP, though all pathways share a common endpoint of mitochondrial ATP synthesis.28 In situations of increased metabolic demands, the proportion of ATP derived from carbohydrate metabolism increases (Figure 4).27Figure 4 Normal myocyte metabolism: taken from Myocardial Substrate Metabolism in the Normal and Failing Heart.28

though all pathways share a common endpoint of mitochondrial ATP synthesis.28 In situations of increased metabolic demands, the proportion of ATP derived from carbohydrate metabolism increases (Figure 4).27Figure 4 Normal myocyte metabolism: taken from Myocardial Substrate Metabolism in the Normal and Failing Heart.28 Histology: More than 70% of left ventricle (LV) myocardial tissue volume is cardiac myocytes; the rest is vasculature and extra cellular matrix (ECM).29 Predominantly containing collagen types I and III, ECM composition is regulated by a number of factors, including circulating neuro-hormones and mechanical strain.30,31Figure 5 Normal single photon emission computed tomography examination: Left ventricle seen in short axis, vertical long axis, and horizontal long axis. Adapted from Hasegawa et al.142 Imaging, morphology, and function: The LV is a conical structure that tapers from base to apex. The normal LV wall has a thickness of between 6 and 10 mm in men, and 6 and 9 mm in women,32 in diastole and thickens uniformly by at least 50% in systole. In response to increasing demand, contraction becomes increasingly dynamic.

, morphology, and function: The LV is a conical structure that tapers from base to apex. The normal LV wall has a thickness of between 6 and 10 mm in men, and 6 and 9 mm in women,32 in diastole and thickens uniformly by at least 50% in systole. In response to increasing demand, contraction becomes increasingly dynamic. As well as function, it is also possible to assess the constituents of myocardium and tissue homogeneity. By CMR the signal in healthy myocardium is uniform on T1 and T2 weighted and contrast enhanced images. Semi-quantitative assessment of the ratio of T2 signal in healthy heart to skeletal muscle is <1.9.33 Quantitative measurement of the T1 signal, by T1 mapping, shows normal values in narrowly defined ranges,34 but depend on scanner field strength and pulse-sequence used.34Extracellular volume, calculated using pre- and post-contrast T1 mapping, is less method dependent and is 26 ± 3% in healthy myocardium.24 Nuclear techniques of perfusion and metabolism display uniform signal throughout the myocardium (Figure 5). Myocardial perfusion may be quantified both at rest and hyperaemia, allowing calculation of MPR. By PET normal resting MBF is ∼0.7 mL/min/g, increasing to 2.75 mL/min/g on stress, with a flow reserve of >4.35 ‘Gold-standard’ PET flow quantification is not without limitation, especially when the flow reduction is only mild.36 Estimates of absolute MBF with CMR show limited agreement with PET,37 though MPR correlates well with PET in health and disease,37 with normal MPR by CMR being ∼2.2.38 Ischaemic Reversible ischaemia

Nuclear techniques of perfusion and metabolism display uniform signal throughout the myocardium (Figure 5). Myocardial perfusion may be quantified both at rest and hyperaemia, allowing calculation of MPR. By PET normal resting MBF is ∼0.7 mL/min/g, increasing to 2.75 mL/min/g on stress, with a flow reserve of >4.35 ‘Gold-standard’ PET flow quantification is not without limitation, especially when the flow reduction is only mild.36 Estimates of absolute MBF with CMR show limited agreement with PET,37 though MPR correlates well with PET in health and disease,37 with normal MPR by CMR being ∼2.2.38 Ischaemic Reversible ischaemia Definition: Myocardial ischaemia is a mismatch of oxygen supply and demand that precipitates change from aerobic to anaerobic cellular respiration. Metabolism: Ischaemia may either be complete, due to coronary occlusion, or limited due to epicardial coronary artery stenosis or abnormalities of the myocardial microvascular circulation. The degree of ischaemia is also determined by the presence and extent of a collateral circulation that can develop in humans with established coronary artery disease.

complete, due to coronary occlusion, or limited due to epicardial coronary artery stenosis or abnormalities of the myocardial microvascular circulation. The degree of ischaemia is also determined by the presence and extent of a collateral circulation that can develop in humans with established coronary artery disease. Changes in cell metabolism begin within one minute of onset of severe ischaemia. Sub-endocardial tissue becomes ischaemic first followed by sub-epicardial tissue.39 Shortly after the onset of severe ischaemia, oxygen present in myocardium is consumed and normal oxidative metabolism ceases. At the same time, electron transport across cell membranes decreases and myocyte contraction becomes impaired. During this initial phase, anaerobic respiration replaces aerobic respiration as the dominant source of ATP, and glycogen replaces fatty acids and glucose as the substrate for energy production.26 Anaerobic respiration in this setting provides approximately one-quarter of the amount of ATP as aerobic metabolism. Due to adverse intra-cellular conditions, including falling pH, ATP production at this rate is only sustained for ∼1 min before continuing at a much lower rate for up to 1 h.26 In non-severe ischaemia, a degree of aerobic respiration continues, consequently more ATP is produced compared with anaerobic glycolysis. Furthermore, hydrogen ions and lactate that accumulate in severe ischaemia are produced less quickly, and ‘washed out’ of still perfused tissue, preserving a more physiological environment.

severe ischaemia, a degree of aerobic respiration continues, consequently more ATP is produced compared with anaerobic glycolysis. Furthermore, hydrogen ions and lactate that accumulate in severe ischaemia are produced less quickly, and ‘washed out’ of still perfused tissue, preserving a more physiological environment. Once ischaemia has resolved, recovery of normal function is variable. Abnormalities of systolic function may persist for several days, myocardium that fails to recover normal systolic function immediately is said to be ‘stunned’.40 Histology: Abnormal function of cell membrane channels leads to myocyte oedema shortly after onset of ischaemia. In addition following short duration of severe ischaemia, depletion of glycogen stores, and the presence of ‘I-bands’ in myo-fibrils are seen on electron microscopy.41 Imaging, morphology, and function

Once ischaemia has resolved, recovery of normal function is variable. Abnormalities of systolic function may persist for several days, myocardium that fails to recover normal systolic function immediately is said to be ‘stunned’.40 Histology: Abnormal function of cell membrane channels leads to myocyte oedema shortly after onset of ischaemia. In addition following short duration of severe ischaemia, depletion of glycogen stores, and the presence of ‘I-bands’ in myo-fibrils are seen on electron microscopy.41 Imaging, morphology, and function Acute prolonged ischaemia Shortly following the onset of ischaemia, regional systolic function becomes impaired and may remain so for days after the ischaemic insult.40 Echocardiography is most commonly used to identify wall motion abnormalities associated with acute ischaemia. Single photon emission computed tomography and PET are rarely used clinically in the setting of acute ischaemia. On CMR, wall motion and thickness are assessed in a similar fashion to echocardiography. In addition, oedema is readily detectable and quantifiable as areas of high signal on T2-weighted images.42 T1 and T2 mapping may provide similar, quantitative information. The oedematous zone may be quantified to determine the extent of myocardial salvage following intervention, and delineate the ‘area at risk’.

aphy. In addition, oedema is readily detectable and quantifiable as areas of high signal on T2-weighted images.42 T1 and T2 mapping may provide similar, quantitative information. The oedematous zone may be quantified to determine the extent of myocardial salvage following intervention, and delineate the ‘area at risk’. Chronic intermittent ischaemia Exercise or dobutamine stress echocardiography allows detection of ischaemia as well as determining its location and extent.43 Ischaemic myocardium shows reduced contractile reserve with regional wall motion abnormalities developing at increasing levels of stress (Figure 6). Figure 6 Schematic of wall motion abnormalities in ischaemia, hibernation, sub-endocardial, and infarction. Cardiac magnetic resonance also allows detection of ischaemia through assessment of regional systolic function. Tissue perfusion can be assessed using first pass adenosine stress CMR (Figure 7). Quantitative assessment of perfusion with CMR has limited agreement with PET imaging44 and is less commonly used in clinical practice than qualitative assessment.

s detection of ischaemia through assessment of regional systolic function. Tissue perfusion can be assessed using first pass adenosine stress CMR (Figure 7). Quantitative assessment of perfusion with CMR has limited agreement with PET imaging44 and is less commonly used in clinical practice than qualitative assessment. Single photon emission computed tomography is commonly used in the investigation of chronic intermittent ischaemia, and while image quality on PET is superior to SPECT, availability of PET limits utility. The sensitivity and specificity of SPECT compare well with other non-invasive imaging techniques45; whole heart acquisition allows accurate quantification of the extent of ischaemia, a measure that may have prognostic value.46 Image interpretation may be limited by attenuation artefact in the inferior wall and anterior wall, especially in women.47 In addition to perfusion imaging and LV function, transient LV dilation (transient ischaemic dilation—TID) may be appreciated on stress SPECT imaging, marking adverse prognosis.48 Transient ischaemic dilation may either represent true dilation as a result of severe coronary disease and stunning, or rather may reflect sub-endocardial defects not appreciated on perfusion imaging.49Figure 7 Imaging in chronic ischaemia: (A) cardiac magnetic resonance first pass perfusion demonstrating mid-ventricular anterior and septal abnormality. (B) Single photon emission computed tomography demonstrating the anterior and septal stress (top row) abnormality that resolves at rest. (C) Angiography in the same patient showing severe left anterior descending stenosis. Adapted from Greenwood et al. Cardiovascular magnetic resonance and single photon emission computed tomography for diagnosis of coronary heart disease: a prospective trial.143

septal stress (top row) abnormality that resolves at rest. (C) Angiography in the same patient showing severe left anterior descending stenosis. Adapted from Greenwood et al. Cardiovascular magnetic resonance and single photon emission computed tomography for diagnosis of coronary heart disease: a prospective trial.143 In head-to-head studies, the theoretical advantages of PET over SPECT have been demonstrated.36,50 As well as detection of ischaemia it is also possible to measure myocardial blood flow with PET,51 enabling quantitative assessment of myocardial blood flow reserve, which correlates strongly with coronary artery stenosis severity.52 Stunning Definition: Myocardium is ‘stunned’ when contractile function is depressed following transient ischaemia, prior to a full recovery, and having sustained no irreversible myocyte damage. The mechanism of sustained systolic dysfunction in stunning is incompletely understood. However, it is believed that oxygen free radical formation and elevated myocardial calcium levels may lead to damage of myocardial proteins or sarcoplasmic reticulum.53

recovery, and having sustained no irreversible myocyte damage. The mechanism of sustained systolic dysfunction in stunning is incompletely understood. However, it is believed that oxygen free radical formation and elevated myocardial calcium levels may lead to damage of myocardial proteins or sarcoplasmic reticulum.53 Metabolism: Sub-epicardial and sub-endocardial myocardial blood flow normalises quickly following restoration of normal coronary flow, however normal myocardial metabolism takes time to recover. Metabolic changes in transient ischaemia, including fall in myocyte ATP, phosphocreatine and pH, take several hours to reverse.54 Post-ischaemic myocardial oxidative and glucose metabolism remain depressed by ∼20% of normal levels for several hours after an ischaemic insult, and take up to 1 week to recover to near normal levels.55 Histology: Histological changes reflect sustained ischaemia. In common with metabolic changes, resolving myocardial oedema, myocardial glycogen, and ATP depletion40 may be detected several days later. Imaging, morphology, and function: Systolic function of affected segments is impaired in stunning. The speed of recovery of systolic function is variable and may be related to the duration and severity of the ischaemic insult,40,56,57 Abnormalities of diastolic function, whether assessed by CMR, conventional or tissue Doppler persist beyond systolic abnormalities.58,59 It is likely that stunning is an under-appreciated phenomenon as functional abnormalities associated with acute ischaemia will often have recovered.

y of the ischaemic insult,40,56,57 Abnormalities of diastolic function, whether assessed by CMR, conventional or tissue Doppler persist beyond systolic abnormalities.58,59 It is likely that stunning is an under-appreciated phenomenon as functional abnormalities associated with acute ischaemia will often have recovered. Dobutamine stress echocardiography in reperfused acute MI has been shown to predict recovery of dysfunctional segments with sensitivity and specificity of 86 and 90%, respectively.60 Cardiac magnetic resonance demonstrates changes in keeping with ischaemia including high signal on T2-weighted images indicative of oedema, as well as regional systolic and diastolic wall motion abnormalities. Positron emission tomography FDG metabolism assessment may demonstrate depressed levels of glucose metabolism, but no significant metabolism perfusion mismatch, as in hibernation, is seen.58 Hibernation Definition: Chronically dysfunctional viable myocardium of ischaemic origin that recovers systolic function following revascularization.

Positron emission tomography FDG metabolism assessment may demonstrate depressed levels of glucose metabolism, but no significant metabolism perfusion mismatch, as in hibernation, is seen.58 Hibernation Definition: Chronically dysfunctional viable myocardium of ischaemic origin that recovers systolic function following revascularization. The processes underlying the development of hibernation remain unclear, although several mechanisms have been proposed. It is thought that although resting blood flow is normal, coronary flow reserves are low.61 This leads to repeated episodes of ischaemia and repetitive myocardial stunning, causing a complex series of physiological and structural changes characteristic of hibernation.62 The abnormalities seen in hibernating myocardium become more pronounced as the duration of hibernation increases.63,64 The time course of recovery of LV systolic function following revascularization is dependent upon the severity of myocardial change, with some studies suggesting that irreversible remodelling may occur with extended hibernation despite successful revascularization.64,65 However, it is not known if delayed recovery may occur beyond study duration.66 Metabolism: There remains debate regarding the metabolic changes present in hibernation. However, there is evidence to suggest that glucose uptake and utilization are increased and fatty acid metabolism is decreased in hibernating myocardium.67,68

The processes underlying the development of hibernation remain unclear, although several mechanisms have been proposed. It is thought that although resting blood flow is normal, coronary flow reserves are low.61 This leads to repeated episodes of ischaemia and repetitive myocardial stunning, causing a complex series of physiological and structural changes characteristic of hibernation.62 The abnormalities seen in hibernating myocardium become more pronounced as the duration of hibernation increases.63,64 The time course of recovery of LV systolic function following revascularization is dependent upon the severity of myocardial change, with some studies suggesting that irreversible remodelling may occur with extended hibernation despite successful revascularization.64,65 However, it is not known if delayed recovery may occur beyond study duration.66 Metabolism: There remains debate regarding the metabolic changes present in hibernation. However, there is evidence to suggest that glucose uptake and utilization are increased and fatty acid metabolism is decreased in hibernating myocardium.67,68 Histology: Hibernating myocardium is macroscopically similar to normal myocardium. However, at a microscopic level, there are diffuse changes within the myocyte and extracellular ultrastructure. Figure 8 Tissue Doppler in normal (A) and chronically ischaemic dysfunctional myocardium (B): E’ is related to the degree of interstitial fibrosis adapted from Shan et al.144

lly similar to normal myocardium. However, at a microscopic level, there are diffuse changes within the myocyte and extracellular ultrastructure. Figure 8 Tissue Doppler in normal (A) and chronically ischaemic dysfunctional myocardium (B): E’ is related to the degree of interstitial fibrosis adapted from Shan et al.144 All types of collagen increase in the ECM of hibernating segments, and are more than twice that found in normal myocardium when de-differentiation is severe.62 Structural changes in the ECM become more pronounced as duration of hibernation increases.64,69 Furthermore, there is down-regulation of myocyte mitochondria and increased glycogen storage when compared with both normal and remote myocardium.62,68 These changes reflect alteration of mRNA expression and disorganization of cytoskeletal proteins as a result of cellular de-differentiation.

hibernation increases.64,69 Furthermore, there is down-regulation of myocyte mitochondria and increased glycogen storage when compared with both normal and remote myocardium.62,68 These changes reflect alteration of mRNA expression and disorganization of cytoskeletal proteins as a result of cellular de-differentiation. Imaging, morphology, and function: On functional imaging, hibernating myocardium has impaired resting systolic function, and will typically be hypo- or akinetic. Diastolic wall thickness is >6 mm by CMR70 or 7 mm on trans-thoracic echocardiography,65 though recent studies have demonstrated that thinning below these thresholds, in the absence of extensive scar, does not preclude recovery.71 With inotropic stimulation, hibernating myocardium shows ‘contractile reserve’ or a ‘biphasic response’, with an improvement in contractile function on low-dose/effort stress prior to deteriorating at higher workloads.72 Low dose stress TDI allows quantification of systolic function: Doppler tissue velocities increase more in hibernating myocardium than in dysfunctional tissue that does not show improvement in systolic function following revascularization (Figure 8).73,74

predicts lack of improvement in systolic function following revascularization. (A) 25% late gadolinium enhancement of the inferior wall and infer-septum (B) 50% late gadolinium enhancement in the infero-lateral wall becoming increasingly transmural inferiorly. (C) 100% late gadolinium enhancement of the inferior wall. In SPECT imaging cellular uptake of Thallium is dependent upon a functional Na+/K+ ATPase and preserved sarcolemmal membrane function.66 In hibernating myocardium, early acquisition following tracer administration identifies a defect, reflecting impaired blood flow on stress, though delayed early uptake may also be evident on rest in severe LV dysfunction.77 On delayed acquisitions, the isotope has been taken up by metabolically active myocytes in the hibernating region, so the defect is reduced. Stress/redistribution SPECT thus allows quantification of ischaemia and the extent of potential recovery.78 Technetium-based tracers, which bind within myocytes to mitochondria, identify viable myocardium similarly to Thallium SPECT. Figure 10 Perfusion-metabolism mismatching on PET in hibernating myocardium: (A) Short axis. (B) Vertical long axis. (C) Horizontal long axis. Top row demonstrates a stress perfusion defect in the infero-lateral wall. Matched metabolism imaging shows preservation of metabolism in the same territory indicating potential hibernation. Adapted from Bengel, Cardiac Positron Emission Tomography.50

m: (A) Short axis. (B) Vertical long axis. (C) Horizontal long axis. Top row demonstrates a stress perfusion defect in the infero-lateral wall. Matched metabolism imaging shows preservation of metabolism in the same territory indicating potential hibernation. Adapted from Bengel, Cardiac Positron Emission Tomography.50 Positron emission tomography assessment is most commonly based on the assessment of myocardial glucose uptake with FDG. Tracer signal is proportional to metabolically active myocardium and likelihood of recovery may be predicted by the glucose metabolic rate (Figure 10).79,80,81 In addition, PET allows the detection and quantification of MPR, which is reduced in hibernation.82,83 Hibernation with non-trans-mural scar Hibernation commonly co-exists in the presence of sub-endocardial infarction and as a consequence overlap of imaging findings may be seen, e.g. sub-endocardial enhancement on LGE CMR, or both fixed and inducible perfusion defects on nuclear study. In the setting of partial infarction response to dobutamine stress on functional imaging is reduced. Decreased circumferential strain measured by echocardiography further facilitates differentiation of normal myocardium, sub-endocardial, and transmural infarction.84,85

Hibernation with non-trans-mural scar Hibernation commonly co-exists in the presence of sub-endocardial infarction and as a consequence overlap of imaging findings may be seen, e.g. sub-endocardial enhancement on LGE CMR, or both fixed and inducible perfusion defects on nuclear study. In the setting of partial infarction response to dobutamine stress on functional imaging is reduced. Decreased circumferential strain measured by echocardiography further facilitates differentiation of normal myocardium, sub-endocardial, and transmural infarction.84,85 The spatial resolution of CMR LGE enables the detection of small volumes of infarction that may be missed by SPECT or PET. Transmural extent of LGE predicts likelihood of systolic recovery: 60% with 1–25% LGE, 42% with 26–50% LGE, and only 7% of segments with >50% enhancement recovered at 3 months in a cohort with predominantly chronic, mild LVSD (LVEF 43%).75 In a second mixed cohort of acute and chronic patients with moderate LVSD (LVEF 38%) recovery rates were similar (0 LGE:73%, 1–25:56%, 26–50:45%, >50%:5%).86 Where hibernation co-exists with acute abnormalities a ‘mixed picture’ of imaging findings may result. Definitive classification of abnormalities in this setting is difficult and may require repeated imaging to classify myocardium. Infarction Definition: Myocardial infarction follows sustained ischaemia leading to myocyte necrosis and subsequent remodelling and fibrosis.

Where hibernation co-exists with acute abnormalities a ‘mixed picture’ of imaging findings may result. Definitive classification of abnormalities in this setting is difficult and may require repeated imaging to classify myocardium. Infarction Definition: Myocardial infarction follows sustained ischaemia leading to myocyte necrosis and subsequent remodelling and fibrosis. Metabolism: Necrosis occurs when sustained severe ischaemia leads to irreversible structural changes within the myocyte, including mitochondrial swelling and disruption of the sacrolemma.41 In the course of ischaemic injury, necrosis begins in the sub-endocardium where tissue perfusion is lowest and energy consumption is highest, leading to ATP supply exhaustion and accumulation of the by-products of glycolysis. Sub-endocardial infarction commences ∼20 min after ischaemia onset. Prolonged ischaemia leads to increasingly transmural necrosis, which moves as a ‘wave front’ from the endocardium to the epicardium,39 though sometimes with sparing of a thin rim of sub-endocardium.

on and accumulation of the by-products of glycolysis. Sub-endocardial infarction commences ∼20 min after ischaemia onset. Prolonged ischaemia leads to increasingly transmural necrosis, which moves as a ‘wave front’ from the endocardium to the epicardium,39 though sometimes with sparing of a thin rim of sub-endocardium. Histology: Characteristic histological changes occur during myocardial infarction, and evolve until the infarcted region undergoes scar replacement. On macroscopic examination there are few detectable changes over the first 4 h. From 4 to 12 h myocardium becomes mottled. Over the next week, the infarct centre becomes pale and yellows while developing red margins, followed by replacement of necrotic infarct by fibrous scar tissue.87,88Figure 11 Histological change following myocardial infarction: pathology of myocardial infarction, diagnostic histopathology 2013. (A) Light microscopy of wavy mitochondrial fibres and interstitial oedema 4 h postmyocardial infarction. (B) Myocardial fibres 24 h post-myocardial infarction, myocardial thinning, and interstitial infiltration by polymorphonuclear leukocytes. (C) Granulation tissue 7–10 days post infarction. Near complete removal of myocytes. (D) Light microscopy 2–3 weeks after infarction. Subendocardial fibrosis marked by arrowhead, with a small area of myocardial fibre preservation between layers of collagen deposition. Adapted from Chang et al.145

phonuclear leukocytes. (C) Granulation tissue 7–10 days post infarction. Near complete removal of myocytes. (D) Light microscopy 2–3 weeks after infarction. Subendocardial fibrosis marked by arrowhead, with a small area of myocardial fibre preservation between layers of collagen deposition. Adapted from Chang et al.145 Microscopically, myocardium undergoes a series of changes responsible for the macroscopically appreciable abnormalities. During the initial phase, glycogen depletion and oedema, in keeping with severe ischaemia is seen on electron microscopy.89 Between 4 and 12 h, oedema, necrosis, and intra-myocardial haemorrhage are seen. From 12 to 24 h neutrophil infiltration and ongoing necrosis develop. This is followed within 24–48 h by the disappearance of nuclei and striations, and macrophages remove dead cells at the infarct border. Following this granulation tissue and collagen deposition begins, eventually leading to the formation of collagen scar (Figure 11).90,91 Imaging, morphology, and function: Wall motion abnormalities develop before ECG changes or symptoms,92 while TDI systolic velocities decrease rapidly following onset of ischaemia.93 Following transmural infarction, myocardium is akinetic and thins with time. On stress or exercise there is no improvement in wall thickening with the absence of contractile reserve.94 Sub-endocardial infarction results in wall motion abnormalities of varying severities, and may be differentiated from transmural infarction using peak systolic circumferential strain and strain rate on stress echocardiography.95

rcise there is no improvement in wall thickening with the absence of contractile reserve.94 Sub-endocardial infarction results in wall motion abnormalities of varying severities, and may be differentiated from transmural infarction using peak systolic circumferential strain and strain rate on stress echocardiography.95 Cardiac magnetic resonance demonstrates gross anatomical changes associated with infarction including remodelling, wall thinning, and motion abnormalities.42 In addition, CMR imaging can be used to delineate infarct extent, indicate the area at risk, and detect microvascular and reperfusion injury.96,97 Late gadolinium enhancement CMR after acute MI displays high signal in the infarcted area due to persistence of gadolinium within areas of increased extracellular volume and abnormal washout kinetics.42 In animal models, infarct enhancement with LGE CMR has been observed within 90 min of the onset of ischaemia.98 In the acute phase, infarct extent may be overestimated due to increased signal and contrast in the oedematous peri-infarct zone.99 In acute myocardial infarction, LGE CMR may show an area of low signal within the otherwise high-signal infarct zone. This reflects microvascular obstruction (MO) leading to an absence of gadolinium delivery to the centre of the infarct. Extensive MO is associated with no-flow on invasive coronary angiography and adverse LV remodelling.100 Within an area of MO, myocardial haemorrhage may occur, causing low signal on T2 and T2* CMR.101 Haemorrhage represents more severe structural change in the setting of acute MI and is also associated with adverse LV remodelling following infarction.102 The combination of the information gained from T2 weighted, early gadolinium, and late gadolinium images allows measurement of the area at risk, infarct size, and myocardial salvage (difference between area at risk and infarct size) (Figure 12).97Figure 12 Acute myocardial infarction on cardiac magnetic resonance: Top row: (A) High signal on T2-weighted image demonstrating Infero-lateral oedema and ‘area at risk’. (B) Subendocardial infarction in the same patient on late gadolinium enhancement. Bottom row: (C) early gadolinium enhancement image in a second patient with extensive lateral wall hypoenhancement. (D) Late gadolinium enhancement confirms infarction and microvascular obstruction.

fero-lateral oedema and ‘area at risk’. (B) Subendocardial infarction in the same patient on late gadolinium enhancement. Bottom row: (C) early gadolinium enhancement image in a second patient with extensive lateral wall hypoenhancement. (D) Late gadolinium enhancement confirms infarction and microvascular obstruction. At the time of acute infarction SPECT imaging is not usually performed. However in research settings, infarct area as delineated by absent tracer uptake on SPECT correlates well with infarct size on pathological specimens.103 In chronic infarction, metabolically active myocytes are replaced by scar and the absence of an intact Na+/K+ ATPase leads to lack of uptake of the SPECT tracer within the region of infarction. The limited spatial resolution of SPECT can lead to an under-appreciation of the extent of infarction because small sub-endocardial infarction may not be detected.104 In acute myocardial infarction, FDG PET allows detection of infarction and the presence of viable tissue in or adjacent to the infarcted territory. The absence of detectable glucose metabolism is associated with irreversible myocardial injury.105 Myocardial perfusion by PET in the acute MI zone is depressed, significantly improves following coronary intervention, and may continue to improve up to 2 weeks later.106

tissue in or adjacent to the infarcted territory. The absence of detectable glucose metabolism is associated with irreversible myocardial injury.105 Myocardial perfusion by PET in the acute MI zone is depressed, significantly improves following coronary intervention, and may continue to improve up to 2 weeks later.106 In the setting of chronic myocardial infarction, concordant reduction in signal from both perfusion and metabolism (NH3 and FDG tracers, respectively) PET is seen and readily appreciated in trans-mural infarction; however, in sub-endocardial infarction PET may fail to detect small areas of sub-endocardial scar when compared with CMR.107 Myopathic myocardium Definition: Dysfunctional myocardium of non-ischaemic origin is considered to be myopathic. Myopathy covers a wide range of pathologies. This review will focus on dilated cardiomyopathy, in which pronounced isolated segmental systolic dysfunction is uncommon, with only passing reference to other aetiologies.

Incident stroke Incident strokes were centrally adjudicated by an independent endpoint committee in each trial using similar definitions and stroke was part of the primary or secondary composite cardiovascular outcomes in both trials.5,6,8,9 Stroke in both trials was defined as a persistent (≥24 h) disturbance of focal neurological function resulting in symptoms thought to be due to cerebral infarction, evidence of haemorrhage or for which there is no certain aetiology.5,6,8,9 Incident atrial fibrillation The occurrence of AF was retrospectively collected in CHARM-Preserved during the trial close-out using a specifically designed case-report form. Incident AF was recorded prospectively in I-Preserve, using a specific case-report form. Statistical methods We included only patients with an LVEF of ≥45% (all 4128 patients in I-Preserve and 2573 of the 3023 in CHARM-Preserved). Patients with AF were defined as those with either AF confirmed on their baseline ECG or a history of AF. The remaining patients were defined as those ‘without AF’. Descriptive statistics were used to describe the pooled patient population from both trials and to compare these two subgroups, using means (standard deviation) or medians [inter-quartile range (IQR)] for continuous variables and count (percentage) for categorical variables.

Myopathic myocardium Definition: Dysfunctional myocardium of non-ischaemic origin is considered to be myopathic. Myopathy covers a wide range of pathologies. This review will focus on dilated cardiomyopathy, in which pronounced isolated segmental systolic dysfunction is uncommon, with only passing reference to other aetiologies. Metabolism: Normal cardiac function relies upon matching of energy demand and consumption. This requires an appropriate oxygen supply to myocytes, mitochondrial function, ATP transport to the site of energy consumption, and a reliable feedback system to maintain appropriate metabolic rates. Energy depletion, perhaps due to mitochondrial dysfunction, could be a primary reason for impaired myocardial contraction. However, the primary defect may usually be in the contractile apparatus or in calcium handling, which increase metabolic demand. Importantly, myopathy may be an acquired phenomenon in un-infarcted myocardium that has undergone remodelling and may co-exist alongside segments affected by scar, ischaemia, stunning or hibernation. In patients with mild heart failure secondary to idiopathic dilated cardiomyopathy any change myocardial substrate utilization is subtle. In severe heart failure, cellular metabolism changes substantially, with greater glucose and less free fatty acid utilization,108,109 in most but not all studies.110Figure 13 Histological abnormalities in cardiomyopathy: (A) Dilated cardiomyopathy, increased interstitial fibrosis at the blue arrow. (B) Hypertrophic cardiomyopathy, increased interstitial fibrosis (blue), and myocyte disarray. (C) Fibro-fatty replacement in arrhythmogenic right ventricular cardiomyopathy. Taken from Hughes, Asimaki and Saffitz, Gulati.146–148

iomyopathy: (A) Dilated cardiomyopathy, increased interstitial fibrosis at the blue arrow. (B) Hypertrophic cardiomyopathy, increased interstitial fibrosis (blue), and myocyte disarray. (C) Fibro-fatty replacement in arrhythmogenic right ventricular cardiomyopathy. Taken from Hughes, Asimaki and Saffitz, Gulati.146–148 Histology: Histological abnormalities differ depending on the underlying aetiology including myocyte disarray and interstitial fibrosis111 in hypertrophic cardiomyopathy (HCM) and fibro-fatty replacement in arrhythmogenic right ventricular cardiomyopathy. A reduction of mitochondria in the failing heart is common.28 Differing degrees of myocyte hypertrophy are seen depending upon the aetiology of heart failure. However irrespective of aetiology, a common finding is the expansion of the ECM and fibrosis due to local factors and activation112 of the renin–aldosterone–angiotensin system.113 Collagen types I and III are the major structural components in the cardiac ECM, providing both tensile strength and elasticity.114 In the failing heart, collagen synthesis increases (especially type III collagen), leading to accumulation of intercellular collagen, limiting ventricular compliance, myocyte function, and contributing to both systolic and diastolic dysfunction (Figure 13).115,116 Imaging, morphology, and function: Functional imaging by echocardiography and CMR supplies important information for prognosis and risk stratification including ejection fraction and left ventricular end diastolic dimensions.117,118

Histology: Histological abnormalities differ depending on the underlying aetiology including myocyte disarray and interstitial fibrosis111 in hypertrophic cardiomyopathy (HCM) and fibro-fatty replacement in arrhythmogenic right ventricular cardiomyopathy. A reduction of mitochondria in the failing heart is common.28 Differing degrees of myocyte hypertrophy are seen depending upon the aetiology of heart failure. However irrespective of aetiology, a common finding is the expansion of the ECM and fibrosis due to local factors and activation112 of the renin–aldosterone–angiotensin system.113 Collagen types I and III are the major structural components in the cardiac ECM, providing both tensile strength and elasticity.114 In the failing heart, collagen synthesis increases (especially type III collagen), leading to accumulation of intercellular collagen, limiting ventricular compliance, myocyte function, and contributing to both systolic and diastolic dysfunction (Figure 13).115,116 Imaging, morphology, and function: Functional imaging by echocardiography and CMR supplies important information for prognosis and risk stratification including ejection fraction and left ventricular end diastolic dimensions.117,118 Comprehensive structural assessment by echocardiography and CMR may guide management and allow diagnosis of the underlying cardiomyopathic process. For example, it is possible to differentiate between morphologically similar cardiomyopathies on echocardiography: Improvement in long-axis function on stress TDI allows differentiation between ischaemic and non-ischaemic cardiomyopathy.119 In the case of left ventricular hypertrophy, echocardiographic 2D-strain assessment allows differentiation of HCM and hypertensive LV hypertrophy, while TDI enables differentiation of HCM and athlete's heart.120,121

xis function on stress TDI allows differentiation between ischaemic and non-ischaemic cardiomyopathy.119 In the case of left ventricular hypertrophy, echocardiographic 2D-strain assessment allows differentiation of HCM and hypertensive LV hypertrophy, while TDI enables differentiation of HCM and athlete's heart.120,121 Tissue Doppler imaging of the mitral annulus and mitral inflow velocity provides a non-invasive estimate of left atrial pressure.122 Cardiac magnetic resonance assessment of left atrial transit time also correlates strongly with LV early diastolic pressure.123 In many cardiomyopathies, LGE CMR shows characteristics patterns of enhancement. The presence and extent of LGE predicts outcomes in a range of cardiac diseases, including DCM,124 HCM,125 and ischaemic heart disease.126 Extracellular volume calculation allows measurement of both diffuse fibrotic and infiltrative processes that are unreliably assessed with visual analysis alone.127,128 T2* mapping allows detection and tracking of iron overload cardiomyopathy.129 The ability of SPECT alone to differentiate ischaemic from DCM is uncertain as mild stress perfusion defects are commonly seen with both aetiologies.130,131 The defects seen have mild stress defect severity ratios (>45%); however, similar abnormalities may be seen in multi-vessel coronary disease precluding the use of SPECT as the sole diagnostic tool in this situation.

patients were defined as those ‘without AF’. Descriptive statistics were used to describe the pooled patient population from both trials and to compare these two subgroups, using means (standard deviation) or medians [inter-quartile range (IQR)] for continuous variables and count (percentage) for categorical variables. The incidence rate of stroke (per 100 patient-years) was calculated over the trial follow-up period and was compared among the AF and no AF subgroups. We plotted Kaplan–Meier (KM) curves for the occurrence of stroke, according to AF status. To satisfy the assumption of the independence of stroke events, recurrent stroke events in a patient after randomization were not included in the analysis.

c from DCM is uncertain as mild stress perfusion defects are commonly seen with both aetiologies.130,131 The defects seen have mild stress defect severity ratios (>45%); however, similar abnormalities may be seen in multi-vessel coronary disease precluding the use of SPECT as the sole diagnostic tool in this situation. Decreased free fatty acid metabolism with increased glucose metabolism is found on PET examination in DCM.108 Hyperaemic blood flow by PET is lower in DCM than in healthy controls, 2.23 mL/min/mL vs. 4.33 mL/min/mL in one report132 and perfusion abnormalities in DCM are progressively worse in more severe heart failure133 and carry adverse prognosis.134 Clinical implications Different imaging modalities assess different facets of myocardial health and disease and are often complementary. An awareness of the principles underlying acquisition, and the aspect of myocardial health and pathology assessed is crucial. The detection of ‘normal’ myocardium is straightforward, and can be accomplished using any test capable of delivering good quality anatomical images. The most appropriate method will vary depending upon patient and institute: In terms of practicality, availability, and economy this will frequently be echocardiography. Atrial as well as ventricular volumes and function should be carefully assessed to detect more subtle disease. Minor deviations from normal, such as small areas of infarction, may be undetected unless high-resolution imaging methods such as LGE CMR are used.

bility, and economy this will frequently be echocardiography. Atrial as well as ventricular volumes and function should be carefully assessed to detect more subtle disease. Minor deviations from normal, such as small areas of infarction, may be undetected unless high-resolution imaging methods such as LGE CMR are used. Ischaemia detection or perfusion assessment may be performed using either stress/exercise echo/CMR, CMR first pass perfusion or nuclear imaging, and all are included in current practice guidelines.135 Single photon emission computed tomography is the mostly widely used modality worldwide, and PET remains the gold standard. In some patients, it may be desirable to exclude valvular disease or cardiomyopathy making DSE or CMR the most appropriate choice of test.

, and all are included in current practice guidelines.135 Single photon emission computed tomography is the mostly widely used modality worldwide, and PET remains the gold standard. In some patients, it may be desirable to exclude valvular disease or cardiomyopathy making DSE or CMR the most appropriate choice of test. Hibernation may only ever be identified retrospectively. However, in clinical practice the question most often posed relates to the likelihood of contractile recovery in a given coronary territory, or the potential for LV reverse remodelling following revascularization. For this purpose, any of the non-invasive imaging techniques covered may be selected, as long as the limitations of the chosen technique are recognized in decision making. However, randomized controlled trials have generally not shown a greater improvement in either ventricular function or prognosis with coronary revascularization compared to pharmacological treatment in patients with heart failure and a reduced LV ejection fraction who have greater myocardial viability,10,11,149 although extended follow of the STICH population recently showed modest 16% reduction in mortality after a median follow-up of 10 years (p = 0.019).150 This may reflect a failure of previous taxonomies rather than a failure of concept; further randomized controlled trials are required. Partial infarction is best appreciated by CMR examination, with the added benefit that systolic abnormalities not associated with coronary disease may be explained by characteristic abnormalities of cardiomyopathy on LGE CMR.

Hibernation may only ever be identified retrospectively. However, in clinical practice the question most often posed relates to the likelihood of contractile recovery in a given coronary territory, or the potential for LV reverse remodelling following revascularization. For this purpose, any of the non-invasive imaging techniques covered may be selected, as long as the limitations of the chosen technique are recognized in decision making. However, randomized controlled trials have generally not shown a greater improvement in either ventricular function or prognosis with coronary revascularization compared to pharmacological treatment in patients with heart failure and a reduced LV ejection fraction who have greater myocardial viability,10,11,149 although extended follow of the STICH population recently showed modest 16% reduction in mortality after a median follow-up of 10 years (p = 0.019).150 This may reflect a failure of previous taxonomies rather than a failure of concept; further randomized controlled trials are required. Partial infarction is best appreciated by CMR examination, with the added benefit that systolic abnormalities not associated with coronary disease may be explained by characteristic abnormalities of cardiomyopathy on LGE CMR. Where infarction and cardiomyopathy co-exist, multi-modality imaging may be necessary, often including coronary imaging to better understand both the dominant aetiology of LVSD and potential for recovery. Tissue Doppler imaging velocities have been shown to differ in DCM and ischaemic heart disease, potentially allowing discrimination of causes of LVSD. Late gadolinium enhancement CMR allows the extent of scarring due to either pathology to be determined, but not the benefit of any specific therapy.

ential for recovery. Tissue Doppler imaging velocities have been shown to differ in DCM and ischaemic heart disease, potentially allowing discrimination of causes of LVSD. Late gadolinium enhancement CMR allows the extent of scarring due to either pathology to be determined, but not the benefit of any specific therapy. Mild perfusion defects, with stress defect severity ratios of >45%, are common in DCM as discussed above. SPECT, PET, or CMR first pass perfusion in combination with coronary angiography may facilitate targeted revascularization, if indicated on conventional grounds, and avoiding unnecessary revascularization.

ential for recovery. Tissue Doppler imaging velocities have been shown to differ in DCM and ischaemic heart disease, potentially allowing discrimination of causes of LVSD. Late gadolinium enhancement CMR allows the extent of scarring due to either pathology to be determined, but not the benefit of any specific therapy. Mild perfusion defects, with stress defect severity ratios of >45%, are common in DCM as discussed above. SPECT, PET, or CMR first pass perfusion in combination with coronary angiography may facilitate targeted revascularization, if indicated on conventional grounds, and avoiding unnecessary revascularization. Conclusion Consistent adoption of standard nomenclature in clinical practice will facilitate thinking and hopefully decision making regardless of local access to different imaging modalities. Figure 14 summarizes key imaging findings covered and aims to provide a reference for future studies. A clear taxonomy for myocardial viability and dysfunction provides the basis for randomized controlled trials that will provide the scientific evidence upon which to base clinical decisions. Figure 14 Taxonomy of myocardial segments in left ventricular systolic dysfunction: This should be viewed as an aid to classification rather than a decision tree. Function: thickness compared with ‘normal’ denotes resting state, with subsequent contractile reserve displayed with increase in segmental thickness. Stunned myocardium may display increased thickness at rest due to oedema though may not be readily appreciable. Perfusion: ‘horseshoe’ displays typical Single photon emission computed tomography/positron emission tomography finding whilst ‘donut’ displays cardiac magnetic resonance findings: Single photon emission computed tomography/positron emission tomography: Red = normal; pale blue = minimally decreased or normal; dark blue = decreased. Cardiac magnetic resonance: grey = normal; black=hypoperfusion/ischaemia. Oedema & Scar: ‘donut’ displays typical T2 weighted (blue) and late gadolinium enhancement findings (black is normal, shades of grey represent late enhancement).

phy: Red = normal; pale blue = minimally decreased or normal; dark blue = decreased. Cardiac magnetic resonance: grey = normal; black=hypoperfusion/ischaemia. Oedema & Scar: ‘donut’ displays typical T2 weighted (blue) and late gadolinium enhancement findings (black is normal, shades of grey represent late enhancement). Funding A.K.M. is funded by a British Heart Foundation (BHF) Project Grant (PG/14/10/30641). S.P. is funded by a BHF Clinical Fellowship (FS/10/62/28409). Funding to pay the Open Access publication charges for this article was provided by The British Heart Foundation (BHF). Conflict of interest: none declared.

Introduction Pathologists have known for centuries that calcific deposits formed in vascular tissues such as the aortic valve. Calcific deposits are hard like bone and have often been described as such. In his ‘History of Animals’, Aristotle remarked that often in oxen and particularly in horse, the heart ‘has a bone inside it’.1 And in 1646, the French professor of medicine Lazare Riviére described his post-mortem examination of a patient's diseased aortic valve by remarking that it resembled ‘a cluster of hazelnuts’.2 Today, aortic valve calcification remains the most common cause of valve stenosis3,4 and is present in some 26% of the population over the age of 65, and as many as half of those over 85.5 Well into the 20th century, the formation of calcific lesions was believed to be a normal, passive process associated with aging. In the early 1990s, however, this attitude began to change when Boström et al. identified bone morphogenetic protein-2a in calcified human carotid arteries.6 Later, in addition to descriptions of standard ‘dystrophic calcification’, histological studies identified endochondral bone formation in heavily calcified aortic valves7 and gene expression analyses demonstrated up-regulation of several bone-specific genes.8 The biological case for bone formation in the vascular system had been established, but how closely do calcific lesions resemble bone as a material?

studies identified endochondral bone formation in heavily calcified aortic valves7 and gene expression analyses demonstrated up-regulation of several bone-specific genes.8 The biological case for bone formation in the vascular system had been established, but how closely do calcific lesions resemble bone as a material? To understand bone, biomineralization researchers have long utilized non-quantitative physical science techniques, including analytical electron microscopy, to examine tissues ex vivo. Such analyses can sometimes reveal more than standard histological stains that often only detect the presence of calcium and/or phosphate. In particular, standard histology does not definitively identify the specific type of mineral present, as calcium phosphate can exist in >10 different phases.9 The mollusc-smashing club that forms part of the exoskeleton of the variegated peacock mantis shrimp, for example, is partially formed of calcium phosphate,10 but fundamentally differs from bone. By applying such characterization techniques to bone, we know a great deal about its structure, composition, and mechanism of formation. However, only recently have similar techniques been applied to calcific lesions on the aortic valve. Here, we examine this new evidence and compare it with similar studies conducted on bone. These analyses lead us to believe that ubiquitous features of calcific lesions cannot be explained either by classic descriptions of dystrophic calcification or simply by ectopic bone formation. Consequently, we advocate complementing current biologically focussed research endeavours with interdisciplinary approaches to help disentangle the interwoven biological and physicochemical processes that drive aortic valve calcification with the aim of eventually developing strategies to eliminate and/or prevent it.

. Consequently, we advocate complementing current biologically focussed research endeavours with interdisciplinary approaches to help disentangle the interwoven biological and physicochemical processes that drive aortic valve calcification with the aim of eventually developing strategies to eliminate and/or prevent it. Close to the bone: the structure and composition of bone tissue Because of its inorganic nature and function as a structural material in the body, biologists and clinicians are not the only scientists who study bone. Quite uniquely, it has also received significant attention from the physical sciences communities. Chemists and engineers have long studied bone samples ex vivo using electron microscopy (nano-meter resolution imaging), energy dispersive X-ray spectroscopy (elemental analysis), and selected area electron diffraction (SAED, crystallographic structure) to precisely describe its hierarchical structure and composition (Figure. 1). As a result, the basic structure of bone, whether it be lamellar or woven, mature or immature, is well-established. In short, bone is composed of apatite crystals, which closely associate with collagen fibres11 (Figure. 1A–C), its elemental composition is predominantly calcium, phosphorus, carbon and oxygen (Figure. 1E), and its mineral is poorly crystalline (a measure of how regularly atoms align, Figure. 1D). Analyses of bones from different species show that bone has a similar structure, composition, and crystallinity across different sites in the body, and more widely, that it is comparable in nearly all vertebrates that have thus far been studied.12Figure 1 The nano-structure and composition of bone. (A and B) Representative density-dependent colour scanning electron micrographs of the surface of bone tissue from a human femoral head showing a homogenous density distribution of mineralized collagen fibres. The orange colour identifies the mineral, while green highlights the organic components of the tissue. Scale bar in (A): 10 µm and in (B) 1 µm. (C) Transmission electron microscopy image of bone. Green arrows and box indicate banded collagen fibres and black arrows highlight dark plate-like crystals. Notice how the dark crystals directly interact with the light collagen fibrils. Scale bar = 0.2 µm. (D) Selected area electron diffraction pattern of bone showing broad diffraction rings/halos (white arrows) indicative of disorder in the material's atomic structure.

and black arrows highlight dark plate-like crystals. Notice how the dark crystals directly interact with the light collagen fibrils. Scale bar = 0.2 µm. (D) Selected area electron diffraction pattern of bone showing broad diffraction rings/halos (white arrows) indicative of disorder in the material's atomic structure. The central bright spot represents the source electron beam (black arrow). The broad diffraction rings/halos are created by reflections of disordered atoms within the material. (E) Energy dispersion X-ray spectroscopy elemental analysis of bone highlighting its composition: calcium, phosphorus, carbon, and oxygen.

and black arrows highlight dark plate-like crystals. Notice how the dark crystals directly interact with the light collagen fibrils. Scale bar = 0.2 µm. (D) Selected area electron diffraction pattern of bone showing broad diffraction rings/halos (white arrows) indicative of disorder in the material's atomic structure. The central bright spot represents the source electron beam (black arrow). The broad diffraction rings/halos are created by reflections of disordered atoms within the material. (E) Energy dispersion X-ray spectroscopy elemental analysis of bone highlighting its composition: calcium, phosphorus, carbon, and oxygen. Heart of stone: the structure and composition of aortic valve calcific lesions As the structure and composition of bone has been well described, what then do the same techniques reveal about calcific lesions on the aortic valve? There are numerous electron micrographs of human calcific lesions available in the literature13–20 and we examined14 >30 recovered human aortic valves ex vivo using this non-quantitative technique. These images reveal that the inorganic component of calcific lesions takes one of three structures (Figure. 2A–C). The most common mineralized structure are spherical particles with sizes ranging from 100 nm to 5 µm (Figure. 2A). These ‘billiard ball-like’ particles have been observed in diseased samples, independent of disease severity. They are composed of calcium, phosphorus, oxygen, and magnesium (Figure. 2H), and their inner structure is formed from concentric layers of mineral distributed radially from the centre (Figure 2D).14 Although visible in SEM images published over the last 30 years,16,21 their presence was dismissed until recently, likely because standard TEM sample preparation fractures the spherical particles and histology techniques can miss them, as they are transparent like a quartz crystal. Figure 2 The nano-structure and composition of human aortic valve calcification. Representative density-dependent colour scanning electron micrographs of (A) calcified particles (arrow, scale bar = 1 µm), (B) calcified fibres (arrow, scale bar = 10 µm), and (C) compact calcification (white arrow) with a calcified particle (black arrow) in human aortic valve tissue (scale bar = 1 µm). Calcified areas appear orange, while the unmineralized extracellular matrix is shown in green. (D) Transmission electron microscopy image of a section through a calcified spherical particle. Notice that the particle is composed of a dense material all the way through. Scale bar = 0.2 µm. (E) Transmission electron microscopy image of compact calcification showing a trapped sphere (yellow arrow), compact amorphous calcium phosphate (white arrow), and collagen fibrils (black arrow). Notice how each component maintains its own structural identity.

f a dense material all the way through. Scale bar = 0.2 µm. (E) Transmission electron microscopy image of compact calcification showing a trapped sphere (yellow arrow), compact amorphous calcium phosphate (white arrow), and collagen fibrils (black arrow). Notice how each component maintains its own structural identity. Scale bar = 0.2 µm. (F) Selected area electron diffraction of a calcified spherical particle. The diffraction pattern shows regularly spaced bright spots or reflections (white arrow) of the diffracted source electron beam (black arrow). Regularly spaced reflections are created by materials that have highly ordered atomic structure. (G) Selected area electron diffraction of compact calcification with a typical amorphous pattern, marked by broad diffraction rings/halos (white arrow) which reflect from the source electron beam (black arrow). Broad diffraction rings/halos are generated by the irregular arrangement of atoms within a material's structure. (H) Energy dispersion X-ray spectroscopy of a calcified particle showing that it is composed of calcium, phosphorus, and magnesium. (I) Energy dispersion X-ray spectroscopy of compact calcification presenting calcium and phosphorus.

alos are generated by the irregular arrangement of atoms within a material's structure. (H) Energy dispersion X-ray spectroscopy of a calcified particle showing that it is composed of calcium, phosphorus, and magnesium. (I) Energy dispersion X-ray spectroscopy of compact calcification presenting calcium and phosphorus. ‘Matrix vesicles’, membrane-bound spherical structures that contain clusters of needle-like mineral, play an important role in mediating bone formation. The dense, spherical mineral particles in aortic valve calcifications, however, fundamentally differ from matrix vesicles in composition and structure.22–25 Spherical particles tend to be larger than matrix vesicles, which are only ∼100 nm in diameter,22,26 and are solid completely through. In contrast, matrix vesicles mineralize along their membranes and often have hollow cores. Moreover, while matrix vesicles contain amorphous mineral,22 the atoms that comprise spherical particles are arranged in a highly ordered fashion, like a single crystal14 (Figure 2F). In fact, the spots (or reflections) in their SAED patterns demonstrate that their mineral is different from any other material found in the body. They are unquestionably far more crystalline than bone.

atoms that comprise spherical particles are arranged in a highly ordered fashion, like a single crystal14 (Figure 2F). In fact, the spots (or reflections) in their SAED patterns demonstrate that their mineral is different from any other material found in the body. They are unquestionably far more crystalline than bone. The other two structures commonly found in calcific lesions are calcified fibres and compact calcification (Figure 2B and C).14 Most calcific lesions with these structures are likely formed by dystrophic mineralization, although 13% are reportedly true bone and possess features of mature lamellar bone, including mineral that templates on a collagen matrix.7 Most calcific lesions lack the nano-level organization typical of bone (Figure 2E), and although they are composed of calcium, phosphorus and oxygen, they do not contain detectable levels of magnesium (Figure 2I) and their mineral is poorly crystalline14 (Figure 2G). Interestingly, most mineralized structures in calcific lesions do not interact with collagen fibres. Instead, the mineral structures and organic components remain as isolated elements, each maintaining their own structural identity (Figure 3). Figure 3 The hierarchical and ordered structure of bone compared with inhomogenous aortic valve calcification. Diagram shows the structure of each tissue at the macroscopic, micron (visible by scanning electron microscopy), and nano (visible by transmission electron microscopy) scales. I. Macroscopic schematic bone. II. Collagen fibrils covered by calcium phosphate mineral. III. Collagen fibrils associated with calcium phosphate mineral crystals at the nano-scale. IV. Macroscopic schematic of cardiac tissue. V. Micron-level structures observed in aortic valve calcific lesions. VI. Organization of calcific lesions which contain fibrous structures, calcified particles, and compact calcification. Table summarizes the characteristics of bone as compared with aortic valve calcific lesions.

croscopic schematic of cardiac tissue. V. Micron-level structures observed in aortic valve calcific lesions. VI. Organization of calcific lesions which contain fibrous structures, calcified particles, and compact calcification. Table summarizes the characteristics of bone as compared with aortic valve calcific lesions. A change of heart: do mineralized particles play a role in aortic valve calcification? As conventional and emerging biological techniques are already providing exciting new insights into its pathogenesis, what additionally can we learn about aortic valve calcification from studying how bone forms? First, observations of ubiquitous spherical particles in diseased valves examined ex vivo suggest that calcific lesion formation does not follow a similar pattern to bone mineralization. One of the hallmarks of bone formation is that mineral crystals template on a collagen matrix.11,27–28 This is not the case in most calcific lesions in which the organic matrix does not associate with the mineral. Moreover, on the aortic valve, dense spherical particles are the first mineralized structure observed, and they are present even prior to the formation of calcific lesions.14 This observation offers the possibility that particles may play a role in mediating further mineralization. Indeed, their presence poses a number of questions: What is the role (if any) of these particles in calcification? Could they be initiators of calcification or a consequence?

the formation of calcific lesions.14 This observation offers the possibility that particles may play a role in mediating further mineralization. Indeed, their presence poses a number of questions: What is the role (if any) of these particles in calcification? Could they be initiators of calcification or a consequence? Analyses of calcific lesions with physical science techniques also provide complementary evidence to that gained from traditional biological techniques. For example, it has confirmed histological studies that show that the composition and structure of most calcific lesions are not bone like. True bone formation reportedly only occurs in a small fraction of calcific lesions, and therefore may be a consequence of calcification rather than a cause.29 The biomaterials community has long known that implanting some forms of calcium phosphate intramuscularly can prompt bone formation.30,31 Moreover, the presence of hydroxyapatite nano-crystals has been shown to upregulate osteogenic gene expression in vascular smooth muscle cells.32 If a similar mechanism manifested in the vascular system, the localization of calcium phosphate to valves could trigger cell transdifferentiation, bone-specific protein formation and gene expression, and eventually actual bone formation. Although a mechanism by which this process might proceed remains unclear, interstitial valve cells have transdifferentiation and mineralization potential.33 Moreover, cell biologists have recently recognized that the stiffness of the substrate on which cells attach can affect their phenotype.34,35 Stiff surfaces that mimic the stiffness of the developing osteon, for example, direct mesenchymal stem cells to differentiate to osteoblasts. If the same were true in the aortic valve, stiff calcium phosphate might drive the transdifferentiation of local cells or prompt circulating cells to adhere and adopt an osteoblastic phenotype.

that mimic the stiffness of the developing osteon, for example, direct mesenchymal stem cells to differentiate to osteoblasts. If the same were true in the aortic valve, stiff calcium phosphate might drive the transdifferentiation of local cells or prompt circulating cells to adhere and adopt an osteoblastic phenotype. The best of both worlds: the potential of an interdisciplinary approach As vascular calcification appears to be more complex than either simple dystrophic mineral formation or a cell-mediated process of bone formation ‘in the wrong place’,36 how then can the field move forward to develop treatments for calcific diseases? The methods now available include biological, biochemical, materials, and clinical techniques. Combining methods and taking an interdisciplinary approach has already yielded promising new avenues of research, as detailed here, and their continued use could be similarly efficacious. For example, optical and fluorescence microscopy are powerful techniques for visualizing cells and identifying labelled proteins. However, they can only reveal insights at the micro-scale because of the inherent limitations of the wavelength of light. On the other hand, analytical electron microscopy is non-quantitative and often cannot provide biological information about cells and proteins, but it can produce ultra-high-resolution images of ex vivo samples and provide clear information about tissue morphology, mineral crystallinity, and elemental composition. Combining the two is possible and the techniques are not inaccessible to standard laboratories. Moreover, the two methods can be used sequentially on the same sample.37 To study aortic valve calcification, sections of recovered valves labelled for specific proteins could be correlated with electron micrographs highlighting the location, size and composition of calcified areas. Elemental analysis applied directly to cells and their labelled compartments might also identify if and how specific cellular organelles might contribute to mineral nucleation and/or subsequent propagation.22

uld be correlated with electron micrographs highlighting the location, size and composition of calcified areas. Elemental analysis applied directly to cells and their labelled compartments might also identify if and how specific cellular organelles might contribute to mineral nucleation and/or subsequent propagation.22 Although an expectation that advances will come from applying physical science techniques to ex vivo aortic valve calcific lesions might seem like wishful thinking, methods borrowed from physics and chemistry are pervasive in clinical medicine. Ultrasound was first pioneered as a geological technique before it found a plethora of uses in medical imaging, and chemists utilized nuclear magnetic resonance (NMR) to verify the molecular structure of chemicals before the technique was applied in diagnostic radiology. Even electron microscopy techniques have been used for years to diagnose chronic renal allograft rejection and various viral infections. Indeed, in some circumstances electron microscopy remains the best diagnostic method available.38–45

r structure of chemicals before the technique was applied in diagnostic radiology. Even electron microscopy techniques have been used for years to diagnose chronic renal allograft rejection and various viral infections. Indeed, in some circumstances electron microscopy remains the best diagnostic method available.38–45 Conclusions Interdisciplinary research approaches have revealed new insights into the pathogenesis of aortic valve calcification. In particular, they have led to the discovery that calcific lesions contain unique, highly crystalline spherical particles. Applying such techniques can complement existing approaches, providing additional insights into aortic valve calcification, and perhaps also pathways for clinical treatments. For example, although our understanding of spherical particles in the vascular system is limited, their highly crystalline nature suggests that they would be quite difficult to dissolve. A consequence of this might be that research efforts that aim to prevent calcification or limit its progression, rather than trying to remove them, may be more clinically effective. Such new ways of looking at long-standing research problems may foster advances in cardiovascular and biomedical research. Authors' contributions E.G., S.B. handled funding and supervision; E.G., S.B. conceived and designed the research; E.G., S.B. drafted the manuscript; E.G., S.B. made critical revision of the manuscript for key intellectual content. Funding Funding to pay the Open Access publication charges for this article was provided by the Wellcome Trust.

Authors' contributions E.G., S.B. handled funding and supervision; E.G., S.B. conceived and designed the research; E.G., S.B. drafted the manuscript; E.G., S.B. made critical revision of the manuscript for key intellectual content. Funding Funding to pay the Open Access publication charges for this article was provided by the Wellcome Trust. Conflict of interest: none declared. Acknowledgements E.G. is supported by a Research Career Development Fellowship from the Wellcome Trust and a Philip Leverhulme Prize from the Leverhulme Trust.

Introduction Up to half of patients with heart failure have a preserved ejection fraction (HF-PEF).1–3 These patients differ from heart failure patients with reduced ejection fraction (HF-REF) in several aspects—they tend to be older, are more often women, and are more likely to have a history of hypertension and atrial fibrillation (AF); they are less likely to have coronary artery disease. Although, mortality rates may not be as high as in patients with HF-REF, the prognosis of HF-PEF patients is considerably worse than that of patients with hypertension, angina pectoris, AF, or diabetes in the same age range and gender distribution.4 The single most common cause of hospital admission in these patients is worsening heart failure and this, along with death, has been the focus of therapeutic interventions in HF-PEF.4 However, given the demographic profile and co-morbidity cluster characterizing these patients, stroke may also be a clinically important outcome in HF-PEF. Little is known about the incidence of stroke in HF-PEF, particularly in the absence of AF.

with death, has been the focus of therapeutic interventions in HF-PEF.4 However, given the demographic profile and co-morbidity cluster characterizing these patients, stroke may also be a clinically important outcome in HF-PEF. Little is known about the incidence of stroke in HF-PEF, particularly in the absence of AF. To investigate this further, we therefore combined and analysed patient-level data from two large HF-PEF trials, the Candesartan in Heart failure Assessment of Reduction in Mortality and Morbidity-Preserved trial (CHARM-Preserved, ClinicalTrials.gov NCT0 0634712)5 and the Irbesartan in Heart Failure with Preserved Systolic Function trial (I-Preserve, NCT00095238),6 to provide a robust estimate of the current incidence of stroke in patients with HF-PEF, with and without AF. We also tested a simple clinical model, developed in HF-REF,7 for predicting the risk of stroke in patients without AF in this pooled dataset. Easy identification of those at highest risk of stroke coupled with the availability of new oral anticoagulants with a low risk of bleeding might allow for a stroke prevention strategy which has an acceptable benefit/risk balance in patients with HF without AF.

k of stroke in patients without AF in this pooled dataset. Easy identification of those at highest risk of stroke coupled with the availability of new oral anticoagulants with a low risk of bleeding might allow for a stroke prevention strategy which has an acceptable benefit/risk balance in patients with HF without AF. Methods Trial patients In order to have a sufficiently large number of HF-PEF patients without AF for analysis, we pooled data from the CHARM-Preserved (NCT00634712) and I-Preserve (NCT00095238) trials. Each was a randomized, double-blind, placebo-controlled, multicentre trial and was approved by the appropriate institutional review boards. CHARM-Preserved and I-Preserve enrolled 3023 and 4128 patients, respectively.5,6 Together, these trials included a broad spectrum of patients with chronic HF-PEF.

00095238) trials. Each was a randomized, double-blind, placebo-controlled, multicentre trial and was approved by the appropriate institutional review boards. CHARM-Preserved and I-Preserve enrolled 3023 and 4128 patients, respectively.5,6 Together, these trials included a broad spectrum of patients with chronic HF-PEF. CHARM-Preserved enrolled patients aged ≥18 years in New York Heart Association (NYHA) functional class II–IV with a left ventricular ejection fraction (LVEF) >40% (although for the purposes of this study we included only patients with an LVEF ≥45%). I-Preserve enrolled patients aged ≥60 years in NYHA functional class II–IV with an LVEF ≥45% and corroborating ECG, echocardiographic or radiologic evidence. In addition, patients must have been hospitalized for heart failure in the preceding 6 months or, if not, had to be in NYHA functional class III or IV. N-terminal pro B-type natriuretic peptide (NT-proBNP) was measured at baseline in I-Preserve but not in CHARM-Preserved. In CHARM-Preserved, patients were randomly assigned to candesartan (target dose of 32 mg once daily) or matching placebo.5 In I-Preserve, patients were randomized to irbesartan (target dose 300 mg once daily) or matching placebo.6 The primary outcome in CHARM-Preserved was the composite of cardiovascular death or HF hospitalization5,8 and in I-Preserve it was the composite of all-cause mortality or cardiovascular hospitalization.6,9 The median follow-up in CHARM-Preserved was 3.1 years and in I-Preserve it was 4.1 years. Study treatment did not reduce the risk of the primary outcome or the risk of stroke in the either trial.5,6

HF hospitalization5,8 and in I-Preserve it was the composite of all-cause mortality or cardiovascular hospitalization.6,9 The median follow-up in CHARM-Preserved was 3.1 years and in I-Preserve it was 4.1 years. Study treatment did not reduce the risk of the primary outcome or the risk of stroke in the either trial.5,6 Incident stroke Incident strokes were centrally adjudicated by an independent endpoint committee in each trial using similar definitions and stroke was part of the primary or secondary composite cardiovascular outcomes in both trials.5,6,8,9 Stroke in both trials was defined as a persistent (≥24 h) disturbance of focal neurological function resulting in symptoms thought to be due to cerebral infarction, evidence of haemorrhage or for which there is no certain aetiology.5,6,8,9 Incident atrial fibrillation The occurrence of AF was retrospectively collected in CHARM-Preserved during the trial close-out using a specifically designed case-report form. Incident AF was recorded prospectively in I-Preserve, using a specific case-report form.

trial follow-up period and was compared among the AF and no AF subgroups. We plotted Kaplan–Meier (KM) curves for the occurrence of stroke, according to AF status. To satisfy the assumption of the independence of stroke events, recurrent stroke events in a patient after randomization were not included in the analysis. Continuous variables [e.g. body mass index (BMI), ejection fraction, and creatinine level] were assessed by visual inspection of restricted cubic splines to identify potential non-linear effects. Uni- and multivariable predictors of the risk for stroke were evaluated using Cox proportional hazards regression analysis in patients without AF. Two separate multivariable analyses for stroke were created. First, an ‘HF-PEF stroke model’ was created using established predictors of ischaemic stroke10–15 with the addition of variables that were significant (P < 0.05) in univariable analysis of our dataset. The final list of variables included was age, sex, LVEF, NYHA class III/IV, BMI, creatinine level, systolic blood pressure, history of stroke, hypertension, and diabetes treated with insulin. Second, we applied a recently published multivariable predictive model for stroke in patients with HF-REF (HF-REF stroke model) in our HF-PEF cohort.7 The five variables included in this model were age, BMI, NYHA class, history of stroke, and diabetes treated with insulin. There were no data missing for the baseline variables used either model. We calculated the hazard ratio and corresponding 95% confidence intervals (95% CI) to express the hazard rate of stroke. The statistical contribution of each variable to the predicted risk of stroke was assessed by the χ2 statistic. In order to be consistent with our previous publication,7 we compared each model’s discrimination ability using estimates of overall c-index for the Cox regression models according to the method of Pencina and D’Agostino,16 as outlined by Liu et al.17 We pre-determined that we would proceed using only the HF-REF stroke model if the overall c-indexes for the two models were not meaningfully different.

discrimination ability using estimates of overall c-index for the Cox regression models according to the method of Pencina and D’Agostino,16 as outlined by Liu et al.17 We pre-determined that we would proceed using only the HF-REF stroke model if the overall c-indexes for the two models were not meaningfully different. The coefficients from statistically significant variables in the final multivariable model were used to calculate an individual patient’s risk score for stroke. The KM curves for occurrence of stroke according to tertiles of risk score were plotted. Final model calibration and the ability to separate patients into risk groups were assessed by observing predicted compared with observed outcomes in terciles, and by using the Hosmer–Lemeshow goodness-of-fit test. The model’s discrimination abilities were evaluated by the overall c-index.16,17 We also conducted a sensitivity analysis to compare the cumulative incidence function for stroke estimated using competing risk technique (to account for the competing risk of death)19,20 with the rates of stroke described from the traditional KM curves above. We also compared the overall c-index of the model with the traditional Harrell’s c-statistic.16–18

s to compare the cumulative incidence function for stroke estimated using competing risk technique (to account for the competing risk of death)19,20 with the rates of stroke described from the traditional KM curves above. We also compared the overall c-index of the model with the traditional Harrell’s c-statistic.16–18 Finally, we validated the preferred risk model in a third HF-PEF trial: the Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist (TOPCAT) (NCT00094302).21 TOPCAT included patients aged ≥50 years with at least one symptom and sign of heart failure, an LVEF ≥45% and either a hospitalization with heart failure in the preceding 12 months or an elevated NT-proBNP or BNP. All analyses were undertaken using SAS version 9.3 (SAS Institute, Inc., Cary, NC, USA). Results Of the 6701 patients with an LVEF ≥45%, 2025 (30%) had a history of AF or AF on their baseline ECG and 4676 patients (70%) had no AF. Baseline characteristics The baseline characteristics of patients with and without AF are shown in the Supplementary material online, TableS1. The baseline characteristics of patients without AF, according whether or not they experienced a subsequent stroke, are shown in Table 1. Table 1 Baseline characteristics according to stroke outcome in patients without AF

teristics of patients with and without AF are shown in the Supplementary material online, TableS1. The baseline characteristics of patients without AF, according whether or not they experienced a subsequent stroke, are shown in Table 1. Table 1 Baseline characteristics according to stroke outcome in patients without AF Patients without AF  (N=4676) Non-stroke (n=4505) Stroke (n=171) Demographics, n  (%) Age, year 69  ± 9 69  ± 9 71  ± 8 <65 1400  (30) 1366  (30) 34  (20) 65 to < 75 2032  (43) 1956  (43) 76  (44) ≥75 1244  (27) 1183  (26) 61  (36) Race Caucasians 4273  (91) 4116  (91) 157  (92) Afro-American/Afro-Caribbean 155  (3) 148  (30) 4  (7) Other 248  (5) 241  (5) 7  (4) Female sex 2542  (54) 2459  (55) 83  (49) NYHA class II 1657  (35) 1612  (36) 42  (26) III 2918  (62) 2799  (62) 119  (70) IV 101  (2) 94  (2) 7  (4) Duration of heart failure, year <2 2778  (59) 2673  (59) 105  (61) 2–5 1110  (24) 1076  (24) 34  (20) >5 764  (16) 734  (16) 30  (18) LV ejection fraction, % 58  ± 9 58  ± 9 57  ± 8 Baseline vital signs BMI, kg/m2 30  ± 6 30  ± 6 29  ± 5 BP, mmHg Systolic 137  ± 16 137  ± 16 140  ± 15 Diastolic 79  ± 10 79  ± 10 79  ± 9 Pulse pressure 58  ± 14 58  ± 14 61  ± 14 Heart rate, b.p.m.

(61) 2–5 1110  (24) 1076  (24) 34  (20) >5 764  (16) 734  (16) 30  (18) LV ejection fraction, % 58  ± 9 58  ± 9 57  ± 8 Baseline vital signs BMI, kg/m2 30  ± 6 30  ± 6 29  ± 5 BP, mmHg Systolic 137  ± 16 137  ± 16 140  ± 15 Diastolic 79  ± 10 79  ± 10 79  ± 9 Pulse pressure 58  ± 14 58  ± 14 61  ± 14 Heart rate, b.p.m. 71  ± 11 71  ± 11 71  ± 10 Laboratory measurements Serum creatinine, µmol/L 88  ± 29 88  ± 29 96  ± 33 Haemoglobin, g/dL 14  ± 2 14  ± 2 14  ± 1 NT-proBNPa, pg/mL(median  ± IQR) 230  (104–537) 225  (104–525) 426  (170–1121) Medical history, n  (%) Coronary heart disease 2960  (63) 2855  (63) 105  (61) Myocardial infarction 1599  (34) 1534  (34) 65  (38) Angina pectoris 2517  (54) 2429  (54) 88  (51) CABG or PCI 1078  (23) 1044  (23) 34  (20) Hypertension 3779  (81) 3632  (81) 147  (86) Diabetes mellitus 1313  (28) 1245  (28) 68  (40) Stroke 379  (8) 343  (8) 36  (21) ICD 11  (0.2) 11  (0.2) 0  (0) Current smoker 2597  (56) 2502  (56) 95  (56) Medication, n  (%) Diuretic  (loop or thiazide) 3392  (73) 3266  (73) 126  (74) Loop diuretic 2278  (49) 2195  (49) 83  (49) Thiazide diuretic 1481  (32) 1430  (32) 51  (30) ACE inhibitor 1020  (22) 982  (22) 38  (22) Aldosterone antagonist 788  (17) 753  (17) 35  (20) Beta-blocker 2761  (59) 2663  (59) 98  (57) Digitalis glycoside 405  (9) 391  (9) 14  (8) Calcium channel blocker 1809  (39) 1747  (39) 62  (36) Antiarrhythmic drug 179  (4) 173  (4) 6  (4) Long-acting nitrate 1476  (32) 1410  (31) 66  (39) Lipid lowering therapy 1786  (38) 1734  (38) 52  (30) Antiplatelet therapy 3204  (69) 3080  (68) 124  (73) Anticoagulant therapy 263  (6) 255  (6) 8  (5) Any antihthrombotic (antiplatelet or anticoagulant therapy) 3408  (73) 3278  (73) 130  (76) Antidiabetic therapy 1096  (23) 1040  (23) 56  (33) Insulin therapy 438  (9) 409  (9) 29  (17) Placebo arm in the original trial 2322  (50) 2231  (50) 91  (53) All continuous values are given in mean ±  standard deviation unless stated otherwise.

c (antiplatelet or anticoagulant therapy) 3408  (73) 3278  (73) 130  (76) Antidiabetic therapy 1096  (23) 1040  (23) 56  (33) Insulin therapy 438  (9) 409  (9) 29  (17) Placebo arm in the original trial 2322  (50) 2231  (50) 91  (53) All continuous values are given in mean ±  standard deviation unless stated otherwise. AF, atrial fibrillation; n  (%), number of observations  (percentage of observations within the group); BMI, body mass index; BP, blood pressure; NYHA, New York Heart Association; NT-proBNP, N-terminal pro B-type natriuretic peptide; CABG, coronary artery bypass graft; PCI, percutaneous coronary intervention; ICD, implantable cardioverter defibrillator; ACE, angiotensin converting enzyme. a Available in 2452 patients. Table 2 List of variables from the ‘HF-REF model for stroke’ Variables Coefficients from HF-REF stroke model7 Previous Stroke 0.591 Diabetes treated with insulin 0.626 Age (per 10 years increase) 0.331 BMI (per 5 kg/m2 increase up to 30) −0.301 NYHA class (NYHA III and IV) 0.472 See the Supplementary material for explanation of how to use coefficients of the variables to calculate individual patient’s risk score of stroke. HF-REF, heart failure with reduced ejection fraction; BMI, body mass index; NYHA, New York Heart Association.

Variables Coefficients from HF-REF stroke model7 Previous Stroke 0.591 Diabetes treated with insulin 0.626 Age (per 10 years increase) 0.331 BMI (per 5 kg/m2 increase up to 30) −0.301 NYHA class (NYHA III and IV) 0.472 See the Supplementary material for explanation of how to use coefficients of the variables to calculate individual patient’s risk score of stroke. HF-REF, heart failure with reduced ejection fraction; BMI, body mass index; NYHA, New York Heart Association. Patients with and without atrial fibrillation Patients without AF were younger and were more likely to have a history of coronary artery disease and hypertension, compared with patients with AF. Patients without AF also had a slightly higher systolic blood pressure but had a lower mean serum creatinine and much lower median NT-proBNP level than patients with AF. There were also notable differences in medical therapy, particularly in use of antiplatelet therapy (69% of patients without AF vs. 39% of those with AF) and anticoagulant treatment (6% vs. 57%, respectively), but also in relation to diuretics, mineralocorticoid receptor antagonists, antiarrhythmic agents, and digoxin.

There were also notable differences in medical therapy, particularly in use of antiplatelet therapy (69% of patients without AF vs. 39% of those with AF) and anticoagulant treatment (6% vs. 57%, respectively), but also in relation to diuretics, mineralocorticoid receptor antagonists, antiarrhythmic agents, and digoxin. Patients without atrial fibrillation—with and without incident stroke during follow-up Among patients without AF, those who experienced a stroke (compared with those who did not) were older, more likely to have a history of diabetes, hypertension, and stroke and had worse NYHA functional class. Patients experiencing stroke also had a higher systolic blood pressure, creatinine, and NT-proBNP level. Compared with those not experiencing stroke, those who did were less likely to be treated with lipid lowering therapy but more likely to be taking nitrates, anti-platelet therapy, and insulin. Very few patients in either group were treated with an oral anticoagulant (263 in total, 6%). LVEF did not differ between patients with and without stroke. Rates of stroke

Patients without atrial fibrillation—with and without incident stroke during follow-up Among patients without AF, those who experienced a stroke (compared with those who did not) were older, more likely to have a history of diabetes, hypertension, and stroke and had worse NYHA functional class. Patients experiencing stroke also had a higher systolic blood pressure, creatinine, and NT-proBNP level. Compared with those not experiencing stroke, those who did were less likely to be treated with lipid lowering therapy but more likely to be taking nitrates, anti-platelet therapy, and insulin. Very few patients in either group were treated with an oral anticoagulant (263 in total, 6%). LVEF did not differ between patients with and without stroke. Rates of stroke Patients with atrial fibrillation The median follow-up time in patients with AF was 3.4 (IQR: 2.8–4.4) years and 124 of these 2025 patients (6.1%) experienced a stroke (1.80 per 100 patient-years). The 1, 2, and 3 year KM rates for stroke were 1.5 (95% CI: 1.0–2.1)%, 3.5 (95% CI: 2.7–4.4)%, and 5.5 (95% CI: 4.5–6.6)%, respectively (Figure 1). The stroke rate in patients treated with an anticoagulant was 1.51 per 100 patient-years; and in those not treated with an anticoagulant it was 2.19 per 100 patient-years (yearly rates shown in the Supplementary material online, FigureS1). Figure 1 Kaplan–Meier plot stroke for chronic heart failure patients with preserved ejection fraction according to atrial fibrillation status at baseline. AF, atrial fibrillation.

ose not treated with an anticoagulant it was 2.19 per 100 patient-years (yearly rates shown in the Supplementary material online, FigureS1). Figure 1 Kaplan–Meier plot stroke for chronic heart failure patients with preserved ejection fraction according to atrial fibrillation status at baseline. AF, atrial fibrillation. Patients without atrial fibrillation The median follow-up time in patients without AF was 3.5 (IQR: 3.0–4.6) years and 171 of these 4676 patients (3.7%) experienced a stroke (1.00 per 100 patient-years). The 1, 2, and 3 year KM rates of stroke were 1.0 (95% CI: 0.8–1.4)%, 2.0 (95% CI: 1.7–2.5)%, and 3.0 (95% CI: 2.5–3.5)%, respectively (Figure 1). Incident AF and risk of stroke In CHARM-Preserved, 1781 patients did not have AF at baseline. Out of 1781, 59 patients (3.3%) experienced a stroke. Of these 59 patients, 10 (17%) developed new AF before the occurrence of their stroke; the number of patients with a stroke without preceding AF was 49 (83%). Development of AF reported was not reported in any patient following a stroke. In I-Preserve, 2895 patients did not have AF at baseline. Out of 2895, 112 patients (4%) experienced a stroke. Of these 112 patients, 18 (16%) developed new AF before the occurrence of their stroke; the number of patients with a stroke without preceding AF was 94 (84%). Twenty patients (18%) with an incident stroke had new AF reported before or after their stroke. Predictors of stroke in patients with heart failure and preserved ejection fraction without AF Figure

In I-Preserve, 2895 patients did not have AF at baseline. Out of 2895, 112 patients (4%) experienced a stroke. Of these 112 patients, 18 (16%) developed new AF before the occurrence of their stroke; the number of patients with a stroke without preceding AF was 94 (84%). Twenty patients (18%) with an incident stroke had new AF reported before or after their stroke. Predictors of stroke in patients with heart failure and preserved ejection fraction without AF Figure 2 and Supplementary material online, TableS2 show the relationship between baseline variables and risk of stroke (univariable analysis). Supplementary material online, TableS3 shows an adjusted analysis using the four independent predictors identified in a multivariable stroke model developed in the present HF-PEF cohort (previous stroke, age, diabetes treated with insulin, and male sex). These overlapped with the five independent predictors in the HF-REF stroke model (previous stroke, age, diabetes treated with insulin, BMI, and NYHA class) (Table 2). The overall c-index for the HF-PEF model was 0.71 (95% CI: 0.57–0.84) compared with 0.73 (0.59–0.85) using the HF-REF model (P-value for difference = 0.415). Thus, we proceeded using the previously validated HF-REF model. This model can be used to calculate an individual’s risk of stroke as described in the Supplementary material online, Supplementary material. Figure 2 The relationship between baseline variables and risk of stroke in patients with heart failure and preserved ejection fraction without atrial fibrillation. Variables are divided by quintiles. BMI, body mass index; BP, blood pressure; LV, left ventricular; NT-proBNP, N-terminal pro-B-type natrieretic peptide.

l. Figure 2 The relationship between baseline variables and risk of stroke in patients with heart failure and preserved ejection fraction without atrial fibrillation. Variables are divided by quintiles. BMI, body mass index; BP, blood pressure; LV, left ventricular; NT-proBNP, N-terminal pro-B-type natrieretic peptide. Figure 3 shows the distribution of the risk score for stroke and illustrates the risk of stroke for a given score. A score of approximately 12 predicts a risk of stroke similar to that which was seen among patients with AF in the current cohort. Figure 4 shows KM curves for stroke with patients classified into three equal-sized groups according to risk score. The number of strokes in tertiles 1, 2 and 3 were 37, 45 and 89, respectively. The 1, 2 and 3 year KM rates of stroke in the two higher risk tertiles were tertile 2: 1.1 (95% CI: 0.7–1.7)%, 1.6 (95% CI: 1.1–2.4)%, and 2.4 (95% CI: 1.7–3.3)%, respectively; and tertile 3: 1.4 (95% CI: 1.0–2.2)%, 3.2 (95% CI: 2.4–4.2)%, and 4.9 (95% CI: 3.9–6.2)%, respectively (Figure 4). Patients in risk-tertile 3 had an overall stroke rate of 1.60 per 100 patient-years. Figure 3 Distribution of risk score for stroke and its relation to predicted risk of stroke within the follow-up period. Figure 4 Kaplan–Meier plot for stroke according to tertile of risk score in patients without atrial fibrillation. Figure

3 shows the distribution of the risk score for stroke and illustrates the risk of stroke for a given score. A score of approximately 12 predicts a risk of stroke similar to that which was seen among patients with AF in the current cohort. Figure 4 shows KM curves for stroke with patients classified into three equal-sized groups according to risk score. The number of strokes in tertiles 1, 2 and 3 were 37, 45 and 89, respectively. The 1, 2 and 3 year KM rates of stroke in the two higher risk tertiles were tertile 2: 1.1 (95% CI: 0.7–1.7)%, 1.6 (95% CI: 1.1–2.4)%, and 2.4 (95% CI: 1.7–3.3)%, respectively; and tertile 3: 1.4 (95% CI: 1.0–2.2)%, 3.2 (95% CI: 2.4–4.2)%, and 4.9 (95% CI: 3.9–6.2)%, respectively (Figure 4). Patients in risk-tertile 3 had an overall stroke rate of 1.60 per 100 patient-years. Figure 3 Distribution of risk score for stroke and its relation to predicted risk of stroke within the follow-up period. Figure 4 Kaplan–Meier plot for stroke according to tertile of risk score in patients without atrial fibrillation. Figure 5 shows the model’s goodness-of-fit by comparing observed and expected probabilities of stroke at 3 years with the patients divided into tertiles. The calibration was also assessed using the Hosmer–Lemeshow test, which was P = 0.761. Figure 5 Comparison of observed and expected stroke rates after 3 years for patients categorized by tertile of risk-score derived from the heart failure with reduced ejection fraction stroke model.7 Observed shows the 3 year Kaplan–Meier rate for each tertile; expected shows estimate from the Cox model for each tertile.

igure 5 Comparison of observed and expected stroke rates after 3 years for patients categorized by tertile of risk-score derived from the heart failure with reduced ejection fraction stroke model.7 Observed shows the 3 year Kaplan–Meier rate for each tertile; expected shows estimate from the Cox model for each tertile. Validation of stroke risk model We tested the predictive model in TOPCAT, which included 1240 patients with and 2205 patients without AF. The mean follow-up was 3.5 years. There were 65 strokes in the patients with AF and 52 strokes in those without AF, giving stroke rates in patients with and without AF 1.64 and 0.71 per 100 patients-years, respectively. Using the same analytical approach (see Supplementary material online, TableS4, FiguresS2andS3), the 1, 2 and 3 year KM rates of stroke in patients without AF, in the two higher risk tertiles were: tertile 2: 1.3 (95% CI: 0.6–2.4)%, 1.3 (95% CI: 0.6–2.3)% and 1.7 (95% CI: 0.9–2.9)%, respectively; and tertile 3: 1.5 (95% CI: 0.8–2.7)%, 1.7 (95% CI: 0.9–2.9)% and 2.6 (95% CI: 1.6–4.1)%, respectively. Patients in risk-tertile 3 of the validation model derived from TOPCAT cohort had an overall stroke rate of 1.06 per 100 patient-years. The overall c-index for the model was 0.86 (95% CI: 0.62–0.99).

%, respectively; and tertile 3: 1.5 (95% CI: 0.8–2.7)%, 1.7 (95% CI: 0.9–2.9)% and 2.6 (95% CI: 1.6–4.1)%, respectively. Patients in risk-tertile 3 of the validation model derived from TOPCAT cohort had an overall stroke rate of 1.06 per 100 patient-years. The overall c-index for the model was 0.86 (95% CI: 0.62–0.99). Sensitivity analysis that evaluated the cumulative incidence functions of stroke for the corresponding KM curves reported above is available in the Supplementary material online, FiguresS4–S7. There is little difference between the two types of curves. The comparison for the ‘stroke in HF-REF’ model’s discrimination ability within the HF-PEF cohort using the overall c-index and the traditional Harrell’s c-method is available in the Supplementary material online, TableS5. Discussion In this analysis, HF-PEF patients with AF were at a high risk of stroke, with an average incidence rate of 1.8% per year which is similar to that recently reported in HF-REF patients with AF (1.6% per year).7

Sensitivity analysis that evaluated the cumulative incidence functions of stroke for the corresponding KM curves reported above is available in the Supplementary material online, FiguresS4–S7. There is little difference between the two types of curves. The comparison for the ‘stroke in HF-REF’ model’s discrimination ability within the HF-PEF cohort using the overall c-index and the traditional Harrell’s c-method is available in the Supplementary material online, TableS5. Discussion In this analysis, HF-PEF patients with AF were at a high risk of stroke, with an average incidence rate of 1.8% per year which is similar to that recently reported in HF-REF patients with AF (1.6% per year).7 HF-PEF patients without AF in this study had a lower risk of stroke compared with those with AF. However, the overall rate of stroke in HF-PEF patients without AF (1.0% per year) was similar to the rate we recently reported in HF-REF patients without AF (1.2% per year). Moreover, as in HF-REF, a small number of demographic and clinical variables identified a subset of HF-PEF patients without AF who were at greater risk of stroke than the remainder. Specifically, in our pooled analysis, patients in the upper third of the risk score had a rate of stroke (1.6% per year) which was higher than in HF-PEF patients with AF receiving an anticoagulant (1.5% per year), although not as high as in similar patients not treated with an anticoagulant (2.2% per year).

remainder. Specifically, in our pooled analysis, patients in the upper third of the risk score had a rate of stroke (1.6% per year) which was higher than in HF-PEF patients with AF receiving an anticoagulant (1.5% per year), although not as high as in similar patients not treated with an anticoagulant (2.2% per year). We have been unable to find other reports of the risk of stroke in HF-PEF patients without AF although in patients in the same age range in clinical trials for hypertension (i.e. with a similar co-morbid phenotype to HF-PEF) have a stroke risk of around 1% per year or less.22–26 In HF-PEF patients with AF randomized to warfarin in ARISTOTLE27 the rate of stroke was 1.4% per year which was similar to the rate in anticoagulant-treated AF patients in our study (1.5% per year). In AF patients with HF and an LVEF >40% in RELY-AF28 the rate of stroke or systemic embolism was 2.07% per year in the warfarin group; in ROCKET-AF29 the rate of the same outcome in similarly defined patients was 2.06% per year. The higher event rates in the latter two trials are due to broader composite outcome (which included non-cerebral systemic embolism) and the requirement for patients in these trials to have additional risk factors for stroke.

in ROCKET-AF29 the rate of the same outcome in similarly defined patients was 2.06% per year. The higher event rates in the latter two trials are due to broader composite outcome (which included non-cerebral systemic embolism) and the requirement for patients in these trials to have additional risk factors for stroke. The similar risk of stroke in patients with HF-PEF and HF-REF, without AF, is also of interest. We previously reported that LVEF was not predictive of stroke in HF-REF patients without AF. Neither was LVEF an independent predictor of stroke risk in this study although we examined only patients with an LVEF ≥45%. This finding is consistent with observations in three recent trials comparing non-Vitamin K antagonist oral anticoagulants with warfarin in patients with AF. In those trials, the risk of stroke and systemic embolism was similar, irrespective of LVEF category, in patients with AF and concomitant HF. A similar conclusion was reached by the Atrial Fibrillation Clopidogrel Trial with Irbesartan for Prevention of Vascular Events (ACTIVE) in AF patients not treated with an oral anticoagulant where the risk of stroke was similar in patients with concomitant HF-REF or HF-PEF.30

ry, in patients with AF and concomitant HF. A similar conclusion was reached by the Atrial Fibrillation Clopidogrel Trial with Irbesartan for Prevention of Vascular Events (ACTIVE) in AF patients not treated with an oral anticoagulant where the risk of stroke was similar in patients with concomitant HF-REF or HF-PEF.30 As in HF-REF, we found that neither systolic blood pressure nor history of hypertension was independent predictor of stroke. Although this contrasts with the findings in other patient cohorts, it is consistent with the ‘reverse epidemiology’ of heart failure and the known association between higher blood pressure and better outcomes in this condition.31–33 Likewise, we saw an association between lower BMI and higher risk of stroke, another feature of the ‘reverse epidemiology’ in heart failure.31–33

cohorts, it is consistent with the ‘reverse epidemiology’ of heart failure and the known association between higher blood pressure and better outcomes in this condition.31–33 Likewise, we saw an association between lower BMI and higher risk of stroke, another feature of the ‘reverse epidemiology’ in heart failure.31–33 A particular strength of this study is the validation of our predictive model in another dataset (TOPCAT). Consequently, our findings have clear clinical implications. With a small number of routinely collected clinical variables it is possible to identify patients with HF-PEF, but without AF, who may be at sufficiently high risk of stroke potentially to justify anticoagulation. Clearly, there is as yet no trial evidence to justify such treatment but our findings suggest a means of identifying patients for such a trial. Consistent with this hypothesis, prior trials in patients with heart failure and reduced ejection fraction collectively suggest that anticoagulation can reduce the risk of stroke in patients in sinus rhythm. However, in the largest of these, the Warfarin vs. Aspirin in Reduced Cardiac Ejection Fraction trial (WARCEF), although warfarin was effective in reducing ischaemic stroke this benefit was offset by major bleeding. With non-vitamin K oral anticoagulants, the risk-to-benefit balance might be more favourable, especially as the target International Normalized Ratio (INR) in WARCEF was 2.75 (range 2.0–3.5).34–37

ion trial (WARCEF), although warfarin was effective in reducing ischaemic stroke this benefit was offset by major bleeding. With non-vitamin K oral anticoagulants, the risk-to-benefit balance might be more favourable, especially as the target International Normalized Ratio (INR) in WARCEF was 2.75 (range 2.0–3.5).34–37 Limitations Each of the two trials included had specific inclusion and exclusion criteria and, hence, our findings may not be generalizable to all patients with HF-PEF. Notably, few patients were in NYHA class IV and worse functional class was a predictor of higher risk of stroke. Hence, the risk of stroke may be higher in ‘real world’ patients than in the cohort studied. Although our data suggest that only the minority of strokes is related to incident AF, systematic detection of new onset AF was insensitive, e.g. continuous ambulatory monitoring was not performed. It is widely recognized that silent AF is frequent in heart failure and undetected AF may have accounted for more strokes than realized. However, waiting for the development of clinically recognized AF before employing anticoagulant therapy may not be the ideal preventive strategy and the best and most cost-effective way to screen for silent AF in HF-PEF is unknown. In addition, these patients may have other reasons to develop thromboembolic and other types of ischaemic stroke, e.g. endothelial dysfunction and blood stasis. Therefore, we believe that our findings support a potential preventive role for anticoagulant therapy in HF-PEF patients in sinus rhythm, particularly as new agents with a lower risk of bleeding are available. Of course, this hypothesis needs to be tested prospectively in a randomized trial and it may be too simplistic to assume that an anticoagulant can substantially reduce the risk of stroke in those with HF-PEF at highest risk.

tients in sinus rhythm, particularly as new agents with a lower risk of bleeding are available. Of course, this hypothesis needs to be tested prospectively in a randomized trial and it may be too simplistic to assume that an anticoagulant can substantially reduce the risk of stroke in those with HF-PEF at highest risk. CHARM-Preserved and I-Preserve were randomized controlled heart failure trials, rather than stroke trials, and used definitions of stroke consistent with other heart failure trials conducted during the same period. Although the definition may not be identical to that used in contemporary stroke trials, it was applied consistently by adjudicators blind to treatment allocation and thus gave an unbiased estimate of treatment effect. Unfortunately, classification of stroke subtype was not carried out in both trials. When the trials were conducted, neuroimaging was not standard in patients with suspected stroke in many, if not most countries, involved. Therefore, we are unable to distinguish between ischaemic and haemorrhagic strokes.

atment effect. Unfortunately, classification of stroke subtype was not carried out in both trials. When the trials were conducted, neuroimaging was not standard in patients with suspected stroke in many, if not most countries, involved. Therefore, we are unable to distinguish between ischaemic and haemorrhagic strokes. In conclusion, we found that a relatively high-risk subset of a third of HF-PEF patients without AF have a risk of stroke similar to that in HF-PEF patients with AF. This higher-risk subset can be identified using five simple clinical variables. The risk of stroke is similar in HF-PEF and HF-REF patients without AF and is predicted by the same variables. The risk of stroke in these patients might be reduced by treatment with an oral anticoagulant but this hypothesis needs to be tested in a clinical trial. The rate of stroke in the highest risk tertile was not quite as high as in patients with AF not treated with an anticoagulant so it is uncertain what the benefit/risk ratio of such treatment might be. Supplementary material Supplementary material is available at European Heart Journal online. Authors’ contributions A.H.A.R., A.C.P., R.L.M., and B.L.C. performed the statistical analysis; J.J.V.M. and K.R.L. handled funding and supervision; J.J.V.M. and A.H.A.R. conceived and designed the research, and drafted the manuscript; all co-authors made critical revision of the manuscript for key intellectual content. Funding A.H.A.R. received departmental funding to perform the analysis. A-CP was supported by the Medical Research Council award MC_UU_12017/10.

Authors’ contributions A.H.A.R., A.C.P., R.L.M., and B.L.C. performed the statistical analysis; J.J.V.M. and K.R.L. handled funding and supervision; J.J.V.M. and A.H.A.R. conceived and designed the research, and drafted the manuscript; all co-authors made critical revision of the manuscript for key intellectual content. Funding A.H.A.R. received departmental funding to perform the analysis. A-CP was supported by the Medical Research Council award MC_UU_12017/10. Conflict of interest: K.R.L. chairs the Data and Safety Monitoring Board for the RESPECT-ESUS Trial, sponsored by Boehringer Ingelheim. G.Y.H.L. is a consultant for Bayer/Janssen, Astellas, Merck, Sanofi, BMS/Pfizer, Biotronik, Medtronic, Portola, Boehringer Ingelheim, Microlife and Daiichi-Sankyo, and he is a speaker for Bayer, BMS/Pfizer, Medtronic, Boehringer Ingelheim, Microlife, Roche and Daiichi-Sankyo. The other co-authors declared no conflict of interest. Supplementary Material Supplementary Data Click here for additional data file.

Introduction Implantable cardiac defibrillators (ICD) are established as preventing death in patients with left ventricular dysfunction and ischaemic heart disease (IHD).1 In patients without IHD, however, ICDs are already considered controversial,2 and recent trial data have been interpreted as indicating that they are not beneficial.3 We set out to analyse the totality of RCT data of ICD vs. no ICD therapy in primary prevention of mortality in patients with left ventricular dysfunction. Methods Eligibility and search strategy We identified all reports of studies of the use of ICD therapy against no ICD therapy for primary prevention in patients with left ventricular systolic dysfunction, in which outcome data was available stratified by the presence of IHD, or recruited only one of these two groups. We included cardiac resynchronization therapy (CRT) RCTs that included a defibrillator arm (CRT-D) and a cardiac resynchronization pacing only arm (CRT-P). We did not include comparisons between CRT-D and no device. Pubmed (1st January 1946 to 18th December 2016), EMBASE (1st January 1974 to 18th December 2016), and the Cochrane Central register for randomized controlled trials using the search strategy detailed in Supplementary material online, Appendix S1. Only articles in English were considered. Reference lists and relevant systematic reviews were hand-searched for additional publications. No published protocol exists for this systematic review and meta-analysis.

domized controlled trials using the search strategy detailed in Supplementary material online, Appendix S1. Only articles in English were considered. Reference lists and relevant systematic reviews were hand-searched for additional publications. No published protocol exists for this systematic review and meta-analysis. Data abstraction Data was independently extracted by two authors (SZ, MJS), including year, participants, intervention, and outcomes. Disagreements were resolved by discussion with a third reviewer (DPF). The risk of bias was independently assessed by two authors (SZ, MJS). We sought data on the primary outcome measure of all-cause mortality. Secondary outcome measures included cardiovascular mortality and sudden cardiac death. We also collected data on specific ICD associated complications including inappropriate shocks and device-related infections. We abstracted reported hazard ratios with confidence intervals, and appropriately transformed them for meta-analysis. If hazard ratios or their confidence intervals were not available, but Kaplan-Meier plots were available, we extracted the underlying data using Digitizer4 and converted to hazard ratios and their standard errors.5 If a trial6 randomized patients to control, CRT-Defibrillator, and CRT-Pacemaker; and only presented data stratified by aetiology for the CRT-Defibrillator vs. control, and CRT-Pacemaker vs. control comparisons; the effect of the defibrillator component was determined by indirect comparison of the CRT-Defibrillator vs. the CRT-Pacemaker arms. The steps used to calculate the hazard ratio effect of the defibrillator component, and derive its confidence interval, for the groups with and without IHD separately, are shown in Supplementary material online, Appendix S2, and are based on formulae from Tierney et al.7

he CRT-Defibrillator vs. the CRT-Pacemaker arms. The steps used to calculate the hazard ratio effect of the defibrillator component, and derive its confidence interval, for the groups with and without IHD separately, are shown in Supplementary material online, Appendix S2, and are based on formulae from Tierney et al.7 If hazard ratio data were unavailable8 we extracted risk ratios. Risk of bias assessment We used the Cochrane Risk of Bias Tool9 to assess all trials for bias across six domains (selection, performance, detection, attrition, reporting, and other). Data analysis Where appropriate, we quantitatively synthesised the extracted hazard ratios and risk ratios using a random-effects meta-analyses with the Restricted Maximum Likelihood (REML) estimator. We calculated the annualized mortality rate across for each aetiology by dividing the overall mortality rate in the control group by the mean follow-up time, and weighting by study size. The I2 statistic was used to measure heterogeneity of trial results.10 We carried out a sensitivity analysis for patients without IHD by omitting each of the trials in turn and repeating the meta-analysis. Publication bias was graphically assessed using Funnel plots, with Egger’s test for asymmetry.11 Data were analysed using “R”,12 and the package “metafor”.13 The PRISMA checklist is included as Supplementary Data.14

nalysis for patients without IHD by omitting each of the trials in turn and repeating the meta-analysis. Publication bias was graphically assessed using Funnel plots, with Egger’s test for asymmetry.11 Data were analysed using “R”,12 and the package “metafor”.13 The PRISMA checklist is included as Supplementary Data.14 Results The primary search yielded 2698 records, which were processed as shown in the study flow chart (Figure 1). Full-text was independently reviewed for 219 articles and 11 trials of ICD therapy for primary prevention were included. Three additional articles reported secondary outcomes for included trials.15–17 Two trials enrolled patients with left ventricular dysfunction regardless of aetiology,6,18 four trials enrolled patients exclusively without IHD,8,19–21 three exclusively with chronic IHD,22–24 and two trials exclusively after an acute myocardial infarction.25,26 One trial8 used amiodarone as the comparator, all other trials continued prescribed therapy. Figure 1 Study flow chart. Three trials27–29 were excluded as they recruited patients resuscitated from an arrhythmic cardiac arrest, with an ICD inserted as secondary prevention. One trial30 was excluded as, whilst it was a randomized controlled trial, allocation to insertion of an ICD was not randomized. A total of 8567 participants were enrolled (4371 ICD therapy, 4196 control), 3128 without IHD and 5439 with IHD (Table 1, study characteristics). Table 1 Study characteristics

Three trials27–29 were excluded as they recruited patients resuscitated from an arrhythmic cardiac arrest, with an ICD inserted as secondary prevention. One trial30 was excluded as, whilst it was a randomized controlled trial, allocation to insertion of an ICD was not randomized. A total of 8567 participants were enrolled (4371 ICD therapy, 4196 control), 3128 without IHD and 5439 with IHD (Table 1, study characteristics). Table 1 Study characteristics Trial CABG-Patch MADIT I MADIT II CAT AMIOVIRT DEFINITE DINAMIT COMPANION SCD-HeFT IRIS DANISH Year 1996 1996 2002 2002 2003 2004 2004 2004 2005 2009 2016 Author Bigger Moss Moss Bänsch Strickberger Kadish Hohnloser Bristow Bardy Steinbeck Kober Intervention ICD ICD ICD ICD ICD ICD ICD CRT-D ICD ICD ICD Control SMT SMT SMT SMT Amiodarone SMT SMT CRT-P SMT SMT SMT LVEF cut-off <36% ≤35% ≤30% ≤30% ≤35% <36% ≤35% ≤35% ≤35% ≤40% ≤35% Randomized (N) 900 196 1232 104 103 458 674 1520 1676 898 1116 Without IHD – – – 100% (n = 104) 100% (n = 103) 100% (n = 458) – 44% (n = 669) 47% (n = 792) – 100% (n = 1116) With IHD 100% (n = 900) 100% (n = 196) 100% (n = 1232) – – – 100% (n = 674) 56% (n = 851) 53% (n = 884) 100% (n = 898) – ICD group N 446 95 742 50 51 229 332 595 829 445 556 Follow-up (months) 32 27 20 66 24 29 30 15.8 45.5 37 67.6 Primary outcome ACM ACM ACM ACM ACM ACM ACM ACM and hospitalization ACM ACM ACM Inclusion criteria Undergoing CABG, abnormal ECG MI, NSVT, NYHA 1-3 MI, NYHA 1–3 Recent DCM diagnosis, NYHA 2-3 NYHA 1-3, asymptomatic Symptomatic DCM, ambient arrhythmias Recent MI NYHA 3–4, recent HF hospitalization NYHA class 2–3, OMT Recent MI NYHA 2-4, raised NT- proBNP Exclusion criteria Sustained VT or VF Cardiac arrest, syncopal VT, MI within 1month Valvular, HCM or restrictive, prior MI Syncope NYHA 4, familial cardiomyopathy NYHA 4 NYHA 4, ventricular arrhythmia before or ≥ 48 h after EP inclusion criteria QRS ≥ 114 or other signal averaged ECG abnormalities NSVT (3–30 beats at rate >120) VE Excluded VT, VF, symptomatic brady Asymptomatic NSVT (>3 beats, HR > 100, lasting <30s) NSVT (3-15 beats, HR < 120) or  < 10 PVC/h None QRS≥120 ms, PR ≥ 150 ms, SR None HR ≥ 90, or NSVT (≥3 beats, HR ≥ 150) None IHD definition Undergoing CABG Q wave or cardiac enzyme positive MI Q wave, cardiac enzymes, fixed defect nuclear scan, akinesis ventriculography, CAD on angio No stenosis > 70% at coronary angiography Absent CAD or out of proportion to CAD Clinically significant CAD on angio or negative stress imaging Recent MI Not specified ≥75% narrowing of major artery, prior MI STEMI or NSTEMI No significant CAD on invasive or CT angiogram, or normal MPS.

s ventriculography, CAD on angio No stenosis > 70% at coronary angiography Absent CAD or out of proportion to CAD Clinically significant CAD on angio or negative stress imaging Recent MI Not specified ≥75% narrowing of major artery, prior MI STEMI or NSTEMI No significant CAD on invasive or CT angiogram, or normal MPS. Allowed 2 stenosed coronaries if felt not significant. Time after MI – >3 weeks >1 month NA NA NA 6–40 days – – 5–31 days NA ICD type Epicardial Epicardial 47% Transvenous Transvenous Transvenous Transvenous Transvenous Transvenous Transvenous Transvenous Transvenous Transvenous 53% CRT implantation permitted – – – – – Yes – Yes – – Yes Age (mean±sd) 64±9 63 65 ± 10 52 ± 11 59 ± 12 58 (range 20–84) 62 ± 11 67 60 63 ± 11 64 Male 84% 92% 85% 80% 70% 71% 76% 68% 77% 77% 73% ACEi/ARB 54% 62% 70% 96% 86% 97% 95% 89% 96% 82% 97% BB 21% 23% 70% 4% 52% 85% 87% 68% 69% 98% 92% CRT 0% 0% NR NR NR 2% NR 100% NR NR 58% LVEF 27% (Mean) 26% (Mean) 23% (Mean) 24% (Mean) 23% (Mean) 21% (Mean) 28% (Mean) 21% (Median) 25% (Median) 35% (Mean) 25% (Median) QRS width (ms) NR (73% >100 ms) NR NR (51% > 120 ms) 108 NR 115 106 160 NR NR CRT 160No CRT 108 QRS normal NR NR 49% 64% NR NR NR NR NR NR NR QRS abnormal LBBB 11% LBBB 8% LBBB 19% LBBB 30% LBBB 48% LBBB 20% NR LBBB 71% NR LBBB 8% CRT LBBB 94%, RBBB 3% No CRT LBBB 17%, RBBB 5% RBBB 8% RBBB 1% RBBB 12% RBBB 3% RBBB 11% NYHA I 37% 0% 16% 22% 13% 0% Excluded Recruited 0% NYHA II 73% (II and III) 65% (II and III) 35% 65% 64% 57% 60% 0% Recruited Recruited 53% NYHA III 24% 35% 20% 21% 27% 86% Recruited Recruited 45% NYHA IV 4% 0% 0% 0% 0% 14% Excluded Excluded 1% Hypertension NR 42% 53% NR 63% NR 46% NR 56% 66% 31% Diabetes 38% 6% (IDDM) 36% NR 34% 23% 30% NR 31% 34% 19% Atrial fibrillation NR NR NR NR NR 25% NR NR 15% 14% 22% ICD, implantable cardioverter defibrillator; CRT-D/P, cardiac resynchronization therapy-defibrillator/pacemaker; SMT, standard medical therapy; ACM, all-cause mortality; MI, myocardial infarction; NSVT, non-sustained ventricular tachycardia; NYHA, New York Heart Association Functional Classification; DCM, dilated cardiomyopathy; HF, heart failure; OMT, optimal medical therapy; VE, ventricular ectopics; VT, ventricular tachycardia; VF, ventricular fibrillation; CAD, coronary artery disease; MPS, myocardial perfusion scintigraphy; CABG, coronary artery bypass graft; NA, not applicable.

tion Functional Classification; DCM, dilated cardiomyopathy; HF, heart failure; OMT, optimal medical therapy; VE, ventricular ectopics; VT, ventricular tachycardia; VF, ventricular fibrillation; CAD, coronary artery disease; MPS, myocardial perfusion scintigraphy; CABG, coronary artery bypass graft; NA, not applicable. Risk of bias assessment Trial quality was assessed using Cochrane risk of bias tool (Table 2). There was no effective blinding of therapy in any of the trials. We assessed our primary end-point of all-cause mortality as having a low risk of bias. End-points requiring clinical judgement, such as sudden cardiac death and cardiovascular death, are at risk of bias if assessors are not blinded. Only five6,8,20,21,26 of the eleven trials reported on procedures to blind end-point assessment. Secondary outcomes were poorly reported, and often used different statistical measures to the primary outcome. Table 2 Risk of bias

n cardiac death and cardiovascular death, are at risk of bias if assessors are not blinded. Only five6,8,20,21,26 of the eleven trials reported on procedures to blind end-point assessment. Secondary outcomes were poorly reported, and often used different statistical measures to the primary outcome. Table 2 Risk of bias Trial CABG-Patch MADIT I MADIT II CAT AMIOVIRT DEFINITE DINAMIT COMPANION SCD-HeFT IRIS DANISH Year 1996 1996 2002 2002 2003 2004 2004 2004 2005 2009 2016 Author Bigger Moss Moss Bansch Strickberger Kadish Hohnloser Bristow Brady Steinbeck Køber Random sequence generation (selection bias) Low risk Unclear–not reported Unclear–not reported Low risk–central randomization Unclear–not reported Unclear-not reported Low risk–central randomization with stratification Unclear–not reported Low risk Low risk Low risk–Web-based randomization with stratification Allocation concealment (selection bias) Unclear Unclear–not reported Unclear–not reported Low risk– “closed envelopes with the assigned study group were sent to each centre … envelopes were opened when a patient was enrolled” Unclear–not reported Unclear-not reported Unclear–not reported Unclear–not reported Low risk Unclear–not reported Low risk Blinding of participants and personnel (performance bias) High–“nature of the intervention precluded the blinding of investigators or patients” High risk High risk High risk High risk High risk High risk High risk–“patients, physicians… were not blinded to the treatment assignments” High risk High risk High risk Blinding of outcome assessment (performance bias) Unclear–“accumulating data were reviewed by an independent Data and Safety Monitoring Board”, but no report of whether outcomes were blindly assessed Unclear–“two member end-point subcommittee reviewed information on the causes and circumstances of deaths”, but no report on whether blinded Unclear–not reported Unclear–not reported Low–“events committee determined the cause of death” … “independently evalutated all information available” and “to assure a blinded review, all references to amiodarone or ICD therapy was removed from the reviewed documents” Low - “cause of death was determined by an events committee… unaware of patients’ treatment assignment” High – “ascertainment of the cause of death was the responsibility of the local investigators”, but a “blinded central validation committee independently reviewed information on all deaths” Low–“steering committee and endpoint committee were unaware of the treatment

ommittee… unaware of patients’ treatment assignment” High – “ascertainment of the cause of death was the responsibility of the local investigators”, but a “blinded central validation committee independently reviewed information on all deaths” Low–“steering committee and endpoint committee were unaware of the treatment assignments” Unclear–not reported Low–“adverse-event committee that was unaware of the treatment assignments classified” the causes of death Low–“endpoint classification committee, the members of which were unaware of the treatment assignments, used prespecified criteria to adjudicate all prespecified cinical outcomes” Incomplete outcome data (attrition bias) Low risk Low risk Low risk Low risk Low risk Low risk Low risk Low risk Low risk Low risk Low risk Selective reporting (reporting bias) Low risk Low risk Low risk Low risk Low risk Low risk Low risk Low risk Low risk Low risk Low risk Other bias Trial funded by CPI/Guidant who supplied devices, but had no role in design, analysis, interpretation or writing.

risk Low risk Low risk Low risk Low risk Low risk Low risk Selective reporting (reporting bias) Low risk Low risk Low risk Low risk Low risk Low risk Low risk Low risk Low risk Low risk Low risk Other bias Trial funded by CPI/Guidant who supplied devices, but had no role in design, analysis, interpretation or writing. Trial funded by CPI/Guidant who supplied devices, but had no role in design, analysis, interpretation or writing. Trial funded by CPI/Guidant who supplied devices, but had no role in design, analysis, interpretation or writing. Trial funded by CPI/Guidant who supplied devices, but had no role in design, analysis, interpretation or writing. Supported in part by an unrestricted research grant from the Guidant Corporation Trial funded by St Jude who supplied devices, but had no role in design, analysis, interpretation or writing. Trial funded by St Jude who supplied devices, but had no role in design, analysis, interpretation or writing. Trial funded by Guidant who supplied devices, but had no role in design, analysis, interpretation or writing. Trial funded by Medtronic who supplied devices, but had no role in design, analysis, interpretation or writing. Trial funded by Medtronic who supplied devices and had access to the final pre-submission manuscript Trial funded by Medtronic, St Jude, TrygFonden, but had no role in design, analysis, interpretation or writing.

l funded by Medtronic who supplied devices, but had no role in design, analysis, interpretation or writing. Trial funded by Medtronic who supplied devices and had access to the final pre-submission manuscript Trial funded by Medtronic, St Jude, TrygFonden, but had no role in design, analysis, interpretation or writing. Populations studied Across the 11 trials, the mean age was 63.1 years. Most trials enrolled patients with an EF ≤ 35%; two trials enrolled those with an EF ≤ 30%,19,23 and one enrolled those with an LVEF ≤ 40%.26 All trials included patients with NYHA Class III symptoms. In 5 trials only patients who were NYHA Class II and III were included. Three trials included patients with NYHA Class IV symptoms, but these accounted for only a small proportion of patients (14%, 4%, 1%). One trial6 did not recruit NYHA Class II patients. 5 trials8,21,23,25,26 included NYHA Class I patients. The electrophysiology inclusion criteria varied between the trials with 6 trials enrolling based on previous NSVT or ectopics, and 5 having no specific electrophysiological inclusion criteria.

All trials included patients with NYHA Class III symptoms. In 5 trials only patients who were NYHA Class II and III were included. Three trials included patients with NYHA Class IV symptoms, but these accounted for only a small proportion of patients (14%, 4%, 1%). One trial6 did not recruit NYHA Class II patients. 5 trials8,21,23,25,26 included NYHA Class I patients. The electrophysiology inclusion criteria varied between the trials with 6 trials enrolling based on previous NSVT or ectopics, and 5 having no specific electrophysiological inclusion criteria. In one trial,22 ICDs were placed with epicardial leads during coronary artery bypass grafting (CABG) surgery. In one trial,24 47% were placed with epicardial leads and 53% placed with transvenous leads. In all other studies transvenous leads were used. The studies enrolling patients with chronic IHD recruited patients at least 3 weeks after previous MI; those enrolling patients with acute MI within 31 days26 or 40 days of an MI.25 Baseline characteristics, inclusion, and exclusion criteria are detailed in Table 1. Effect on all-cause mortality Left ventricular dysfunction without ischaemic heart disease Across the 3128 patients without ischaemic heart disease, there was a significant reduction in all-cause mortality with minor heterogeneity (HR 0.76, 95% CI 0.64 to 0.90, P = 0.001, I2 = 3%, Figure 2). The annualized mortality rate in control patients was 5.4%. Figure 2 Title: Left ventricular dysfunction without ischaemic heart disease: impact of primary prevention ICD on all-cause mortality.

Effect on all-cause mortality Left ventricular dysfunction without ischaemic heart disease Across the 3128 patients without ischaemic heart disease, there was a significant reduction in all-cause mortality with minor heterogeneity (HR 0.76, 95% CI 0.64 to 0.90, P = 0.001, I2 = 3%, Figure 2). The annualized mortality rate in control patients was 5.4%. Figure 2 Title: Left ventricular dysfunction without ischaemic heart disease: impact of primary prevention ICD on all-cause mortality. A sensitivity analyses, carried out by omitting each of the trials in turn, in each case shows a statistically significant consensus reduction in mortality (see Supplementary material online, Appendix S4). A funnel plot did not show any significant asymmetry (Egger’s test P = 0.5, Supplementary material online, Appendix S5).

itivity analyses, carried out by omitting each of the trials in turn, in each case shows a statistically significant consensus reduction in mortality (see Supplementary material online, Appendix S4). A funnel plot did not show any significant asymmetry (Egger’s test P = 0.5, Supplementary material online, Appendix S5). Left ventricular dysfunction with ischaemic heart disease Across the 3867 patients in all trials of primary prevention ICD therapy with ischaemic heart disease and no recent MI, there was a non-significant reduction in all-cause mortality (pooled HR 0.81, 95% CI 0.65 to 1.03, P = 0.08, Figure 3A). However, there was substantial heterogeneity (I2 = 62%). One trial22 was unique in inserting the ICD at the time of CABG surgery. There was a 16% higher infection rate in the ICD group, with 4.3% requiring removal. Current practice is to minimize infection risk by implanting the cardiac device separately from any open surgery. Running the analysis for the trials that tested this approach showed a significant reduction in mortality (HR 0.76, 95% CI 0.60 to 0.96, P = 0.02, I2 52%, Figure 3B). The annualized mortality rate in the control patients was 11.3%. A funnel plot did not show any significant asymmetry (Egger’s test P = 0.2, Supplementary material online, Appendix S5). Figure 3 (A) Title: Left ventricular dysfunction with ischaemic heart disease: impact of primary prevention ICD on all-cause mortality. (B). Title: Left ventricular dysfunction with ischaemic heart disease: impact of primary prevention ICD implanted during a dedicated procedure on all-cause mortality.

Left ventricular dysfunction with ischaemic heart disease Across the 3867 patients in all trials of primary prevention ICD therapy with ischaemic heart disease and no recent MI, there was a non-significant reduction in all-cause mortality (pooled HR 0.81, 95% CI 0.65 to 1.03, P = 0.08, Figure 3A). However, there was substantial heterogeneity (I2 = 62%). One trial22 was unique in inserting the ICD at the time of CABG surgery. There was a 16% higher infection rate in the ICD group, with 4.3% requiring removal. Current practice is to minimize infection risk by implanting the cardiac device separately from any open surgery. Running the analysis for the trials that tested this approach showed a significant reduction in mortality (HR 0.76, 95% CI 0.60 to 0.96, P = 0.02, I2 52%, Figure 3B). The annualized mortality rate in the control patients was 11.3%. A funnel plot did not show any significant asymmetry (Egger’s test P = 0.2, Supplementary material online, Appendix S5). Figure 3 (A) Title: Left ventricular dysfunction with ischaemic heart disease: impact of primary prevention ICD on all-cause mortality. (B). Title: Left ventricular dysfunction with ischaemic heart disease: impact of primary prevention ICD implanted during a dedicated procedure on all-cause mortality. Left ventricular dysfunction with acute myocardial infarction In the 2 trials that enrolled 1572 patients after an acute MI, ICD therapy did not cause a significant reduction in mortality (HR 1.05, 95% CI 0.86 to 1.30, P = 0.6, I2 = 0%, Figure 4). The annualized event rate in the control patients was 7.6%. Supplementary material online, Appendix S5 contains the funnel plot.

ction In the 2 trials that enrolled 1572 patients after an acute MI, ICD therapy did not cause a significant reduction in mortality (HR 1.05, 95% CI 0.86 to 1.30, P = 0.6, I2 = 0%, Figure 4). The annualized event rate in the control patients was 7.6%. Supplementary material online, Appendix S5 contains the funnel plot. Figure 4 Left ventricular dysfunction with acute myocardial infarction: impact of primary prevention ICD on all-cause mortality. Effect on secondary outcomes Secondary outcomes were inconsistently reported with not all trials presenting data. Some data were presented as raw counts from which risk ratios could be derived, and some as hazard ratios. ICD therapy was consistently associated with a statistically significant reduction in hazard ratio and risk ratios for all three groups (without IHD, with IHD, and acute MI) for sudden cardiac death (without IHD HR 0.4 RR 0.29; with IHD HR 0.38 RR 0.41; acute MI HR 0.49, RR 0.57, Supplementary material online, Appendix S3). Discussion Based on high-quality data from RCTs, this meta-analysis finds that primary prevention ICDs reduce all-cause mortality in patients with left ventricular dysfunction both with and without IHD. No benefit from ICDs is seen in the setting of acute myocardial infarction. These findings are consistent with the current ESC guideline recommended management.31,32

this meta-analysis finds that primary prevention ICDs reduce all-cause mortality in patients with left ventricular dysfunction both with and without IHD. No benefit from ICDs is seen in the setting of acute myocardial infarction. These findings are consistent with the current ESC guideline recommended management.31,32 Patients without ischaemic heart disease There has been controversy over the utility of ICDs in patients without IHD. Many of the published guidelines make a distinction between the aetiologies with respect to the level of evidence on which their recommendations are made. The 2015 European Society of Cardiology (ESC) ventricular arrhythmia guidelines,31 and the 2016 ESC heart failure guideline32 give ICDs for primary prevention a 1A recommendation for an ischaemic aetiology, and 1B for a non-ischaemic aetiology. Indeed, this uncertainty was the stimulus for conducting the recent DANISH study. Subsequent commentary3 has added to the uncertainty. Part of this uncertainty may have arisen as mortality rate in patients without IHD is lower than those with IHD (5.4%/year vs. 11.3%/year, respectively), and consequently the confidence intervals are wider for individual trials. However, all the point estimates lie in the range 0.55 to 0.87, and the trials showed minimal heterogeneity (I2 = 3%). The group without IHD in COMPANION was, even on its own, statistically significant for a reduction of all-cause mortality with ICD (see Supplementary material online, Appendix S2), although this was not the chosen central message of the COMPANION primary publication.

d the trials showed minimal heterogeneity (I2 = 3%). The group without IHD in COMPANION was, even on its own, statistically significant for a reduction of all-cause mortality with ICD (see Supplementary material online, Appendix S2), although this was not the chosen central message of the COMPANION primary publication. Our meta-analysis confirms a statistically significant reduction in all-cause mortality by primary prevention ICD in patients without IHD. Whilst only one trial was individually significant, the point estimates from all 6 trials were in the same direction, suggestive of benefit. Furthermore, even omitting both COMPANION and the recent DANISH trial from the meta-analysis still produces a statistically significant consensus reduction in mortality (see Supplementary material online, Appendix S4). Patients with ischaemic heart disease This meta-analysis supports the current consensus that ICDs reduce all-cause mortality in left ventricular dysfunction with IHD, in the trials that use the current clinical convention of a dedicated device implant procedure. Interestingly, the reduction in hazard ratio is numerically the same (24%) in patients with and without IHD. Consequently, when considering ICD therapy, distinctions between the two groups may be unnecessary. In acute myocardial infarction, however, there is no indication of a reduction in all-cause mortality.

Patients with ischaemic heart disease This meta-analysis supports the current consensus that ICDs reduce all-cause mortality in left ventricular dysfunction with IHD, in the trials that use the current clinical convention of a dedicated device implant procedure. Interestingly, the reduction in hazard ratio is numerically the same (24%) in patients with and without IHD. Consequently, when considering ICD therapy, distinctions between the two groups may be unnecessary. In acute myocardial infarction, however, there is no indication of a reduction in all-cause mortality. Difference between this meta-analysis and previous meta-analyses Our meta-analysis is the first to include the results of the patients without IHD from the COMPANION and DANISH trials. Other meta-analyses33 have omitted COMPANION, presumably because the paper did not display the hazard ratio explicitly. However, the hazard ratio and its confidence interval can be calculated from the steps shown in Supplementary material online, Appendix S2. The current meta-analysis therefore provides important new information regarding the role of ICD therapy in patients with left ventricular dysfunction without IHD. Study limitations Any meta-analysis can only examine studies that have actually been carried out. Different studies took different approaches to recruitment. However, it is notable that all six non-ischaemic trial results were concordant not only in the direction of effect, but also the approximate magnitude, with the I2 statistical test showing minor heterogeneity.

mine studies that have actually been carried out. Different studies took different approaches to recruitment. However, it is notable that all six non-ischaemic trial results were concordant not only in the direction of effect, but also the approximate magnitude, with the I2 statistical test showing minor heterogeneity. In the case of the COMPANION trial, the hazard ratio was calculated using the information published in the primary publication by steps shown in Supplementary material online, Appendix S2. The original publication did not comment on this hazard ratio. It is wise to be cautious of results of sub-group analyses, because many such analyses are possible and some will be positive by chance alone. However, the single most important dichotomy in current guidelines31,32 for primary prevention ICDs in left ventricular systolic dysfunction is the presence vs. absence of ischaemic heart disease. Therefore, this sub-group analysis need not be assumed to be a random result selected from many possible sub-groups analyses. Moreover, all six groups of patients without ischaemic heart disease showed the same direction of effect. Furthermore, the finding is stable to the removal of any one trial (see Supplementary material online, Appendix S4).

alysis need not be assumed to be a random result selected from many possible sub-groups analyses. Moreover, all six groups of patients without ischaemic heart disease showed the same direction of effect. Furthermore, the finding is stable to the removal of any one trial (see Supplementary material online, Appendix S4). Background medical therapy has improved over the time-course of these trials, with only 4% treated with beta-blockers in the CAT (2002), but 92% in DANISH (2016). Whilst the relative mortality-reduction effect size has remained remarkably consistent over time this will reduce the absolute effect size (when analysed over a fixed time window) of ICDs for primary prevention.

course of these trials, with only 4% treated with beta-blockers in the CAT (2002), but 92% in DANISH (2016). Whilst the relative mortality-reduction effect size has remained remarkably consistent over time this will reduce the absolute effect size (when analysed over a fixed time window) of ICDs for primary prevention. Our study could not consider the degree to which comorbidities might affect results. It has been noted that patients recruited into trials often have fewer comorbidities than those in the general population. The external validity of RCTs is always challenged by this, particularly in conditions such as heart failure where comorbidities may be frequent and severe.34 Furthermore, whilst this meta-analysis finds that stratifying by the presence or absence of ischaemic heart disease does not influence the mortality benefit of ICDs in primary prevention, other factors might. Supplementary material online, Appendix S4 includes data stratified by the presence or absence of CRT, but this analysis is hindered by the limited data in CRT group which is derived from COMPANION6 and a sub-group of DANISH.20 The 2013 ESC guidelines on cardiac pacing and cardiac resynchronization therapy35 similarly recognize that limited RCT data is available for the comparison between CRT-P and CRT-D. The guidelines suggest clinical conditions such as advanced or end-stage cardiac or renal disease may favour CRT-P over CRT-D.

of DANISH.20 The 2013 ESC guidelines on cardiac pacing and cardiac resynchronization therapy35 similarly recognize that limited RCT data is available for the comparison between CRT-P and CRT-D. The guidelines suggest clinical conditions such as advanced or end-stage cardiac or renal disease may favour CRT-P over CRT-D. Clinical implications The challenge facing clinical trials, as highlighted by McMurray,3 is that skilful modern treatment algorithms have reduced event rates down to low levels in the types of patients who would be eligible for, and willing to enter, randomized controlled trials; the annualized rate is 5.4% in patients without IHD. In light of this perhaps, we should pay maximal attention to information that RCTs give us. The low event rate in the trials is why viewing multiple trials is necessary to see the survival benefit. However, the 24% risk reduction is as sizable as one might realistically hope for, for any intervention. This meta-analysis provides strong support for the role of primary prevention ICDs in patients with left ventricular dysfunction. A 24% risk reduction in all-cause mortality is comparable with other therapies which we recommend in heart-failure such as candesartan36 or an angiotensin-neprilysin inhibitor (HR 0.77, 0.84, respectively).37 Conclusions In patients with left ventricular dysfunction, primary prevention ICDs reduce mortality. ICDs reduce mortality by 24% in both patients with (P = 0.03) and without IHD (P = 0.0023). When deciding on ICD therapy, classification of heart failure by aetiology may therefore not be useful.

The low event rate in the trials is why viewing multiple trials is necessary to see the survival benefit. However, the 24% risk reduction is as sizable as one might realistically hope for, for any intervention. This meta-analysis provides strong support for the role of primary prevention ICDs in patients with left ventricular dysfunction. A 24% risk reduction in all-cause mortality is comparable with other therapies which we recommend in heart-failure such as candesartan36 or an angiotensin-neprilysin inhibitor (HR 0.77, 0.84, respectively).37 Conclusions In patients with left ventricular dysfunction, primary prevention ICDs reduce mortality. ICDs reduce mortality by 24% in both patients with (P = 0.03) and without IHD (P = 0.0023). When deciding on ICD therapy, classification of heart failure by aetiology may therefore not be useful. Supplementary material Supplementary material is available at European Heart Journal online. Funding This work was supported by the British Heart Foundation [grant numbers FS/14/27/30752 (MJSS), FS/12/12/29294 (GC), FS/13/44/30291 (ZW), FS/10/038 (DPF)]. Conflict of interest: M.J.S.S., S.Z., J.P.H., G.C., and D.P.F. declare no conflict of interest. ZW has received speaker fees from St. Jude, and a research grant unrelated to this work from Medtronic. Supplementary Material Supplementary Data Click here for additional data file.

Introduction Between and within European country variation in the delivery and outcomes from acute myocardial infarction [AMI] suggest that the potential to reduce the burden of cardiovascular disease has not been realized.1–3 Measuring recognized standards of care is a mechanism by which geographic variation in the use of guideline-indicated treatments may be addressed and, therefore, cardiovascular outcomes improved. The 2016 European Society of Cardiology [ESC] Acute Cardiovascular Care Association [ACCA] quality indicators [QI] for the management of AMI4 are based upon the ESC guidelines for the management of AMI in patients presenting with ST-segment elevation5 and acute coronary syndrome in patients presenting without persistent ST-segment elevation.6 They comprise 7 domains across 20 QIs, including the evaluation of: (1) centre organization, (2) the reperfusion/invasive strategy, (3) in hospital risk assessment, (4) antithrombotic treatment during hospitalization, (5) secondary prevention discharge treatments, (6) patient satisfaction, and (7) composite QIs and Global Registry of Acute Coronary Events (GRACE) risk score adjusted 30-day mortality.

nization, (2) the reperfusion/invasive strategy, (3) in hospital risk assessment, (4) antithrombotic treatment during hospitalization, (5) secondary prevention discharge treatments, (6) patient satisfaction, and (7) composite QIs and Global Registry of Acute Coronary Events (GRACE) risk score adjusted 30-day mortality. To date, there has been no investigation of within country provider variation according to the ESC ACCA QIs or the relationship between the QIs and 30-day mortality. To address this knowledge gap, providing an external validation of the ESC ACCA QIs for AMI, we used data from the United Kingdom national heart attack register (Myocardial Ischaemia National Audit Project [MINAP]) which collects data from one health system, the National Health Service of England and Wales.

mortality. To address this knowledge gap, providing an external validation of the ESC ACCA QIs for AMI, we used data from the United Kingdom national heart attack register (Myocardial Ischaemia National Audit Project [MINAP]) which collects data from one health system, the National Health Service of England and Wales. Methods Setting and design The analyses were based on data from MINAP, a comprehensive registry of ACS hospitalizations mandated by the United Kingdom Department of Health.7 Each MINAP entry provides patient demographic data and clinical details of the patient journey across 122 data items; details of MINAP data collection and management have been described previously.7 The analytical cohort (n = 118 075) was drawn from patients (n = 118 168) with a discharge diagnosis of AMI admitted to one of 220 hospitals between 1st January 2012 and 30th June 2013 ( see Supplementary material online, Figure S1). Patients were eligible for the study if they were ≥18 years of age. For patients with multiple admissions the earliest record was used (to reduce potential bias from previous treatments). We excluded nine hospitals that treated ≤30 patients within the 18-month period of study. Patient-level data concerning demographics, cardiovascular risk factors, medical history, and clinical characteristics at the time of hospitalization were extracted from the registry. Unique patient identifiers were used to link patients with the Office for National Statistics such that vital status or date of death at 30 days could be ascertained. Data used within the study were fully anonymized and, as such, ethical approval was not required under NHS research governance arrangements. The National Institute for Cardiovascular Outcomes Research (NICOR) which includes the MINAP database (Ref: NIGB: ECC 1-06 (d)/2011) had support, under section 251 of the National Health Service Act 2006, to use patient information for medical research without consent. The study was conducted in compliance with the Declaration of Helsinki.

e for Cardiovascular Outcomes Research (NICOR) which includes the MINAP database (Ref: NIGB: ECC 1-06 (d)/2011) had support, under section 251 of the National Health Service Act 2006, to use patient information for medical research without consent. The study was conducted in compliance with the Declaration of Helsinki. ESC quality indicators The ESC ACCA position statement defined 7 domains of care encompassing 12 main and 8 secondary QIs (see Supplementary material online, Figure S2). All 20 QIs were mapped to each patient’s MINAP data to identify data fields that would enable their calculation. For each QI, we included patients who were eligible for the treatment or intervention and whose record had no missing data. As such, patients were classified as ineligible if a treatment was contra-indicated, not indicated, not applicable, if the patient declined treatment or treatment was deemed inappropriate due to co-morbidity. Domain seven specifies the use of an opportunity-based composite score and an all-or-none score (see Supplementary material online, Appendix S1). For this study, we calculated the composite score for each patient and subsequently each hospital, based on the number of times particular care processes were performed (numerator) divided by the number of chances a patient had to receive/hospital had to provide that care (denominator). The composite score was calculated using an equal weight method and included 9 measures (see Supplementary material online, Appendix S1).

on the number of times particular care processes were performed (numerator) divided by the number of chances a patient had to receive/hospital had to provide that care (denominator). The composite score was calculated using an equal weight method and included 9 measures (see Supplementary material online, Appendix S1). Statistical analyses Baseline characteristics for the study population were described using numbers and percentages for categorical data, and medians and IQR or means and standard deviations (SD) for continuous non-normally and normally distributed data. For the QIs, the proportion presented is of those eligible for treatment. We used a validated method for use with MINAP data to calculate each patient’s GRACE risk score. This used the scoring system and coefficients described by the GRACE investigators, given that not all records had information about Killip class and chronic renal failure, the ‘use of loop diuretic’ (as a surrogate for Killip class II-IV), and creatinine concentration, respectively, were added to each patient’s score.8,9

s used the scoring system and coefficients described by the GRACE investigators, given that not all records had information about Killip class and chronic renal failure, the ‘use of loop diuretic’ (as a surrogate for Killip class II-IV), and creatinine concentration, respectively, were added to each patient’s score.8,9 To estimate the GRACE risk score adjusted 30-day mortality, we used the predicted probabilities derived from a logistic regression model where the dependent variable was 30-day mortality and the independent variable was each patient’s calculated GRACE risk score. Data were summarized overall and at the hospital level. We used Spearman’s rank test to investigate the relationship between all combinations of QIs, except for the composite scores because they incorporated several single QIs. We fitted a logistic regression model to assess the strength of association between QI measures and 30-day mortality. For the composite opportunity measure, the performance was split into 4 categories: (1) no interventions received, (2) <40% of eligible interventions received, (3) ≥40% to <80% of eligible intervention received, and (4) ≥80% of interventions received.10,11 We excluded measures that had ≤30 patients with complete data for either aspect of the QI. All analyses were conducted using Stata MP Version 14.0 (StataCorp LP, TX, USA), with statistical significance determined at 5%.

ed, (3) ≥40% to <80% of eligible intervention received, and (4) ≥80% of interventions received.10,11 We excluded measures that had ≤30 patients with complete data for either aspect of the QI. All analyses were conducted using Stata MP Version 14.0 (StataCorp LP, TX, USA), with statistical significance determined at 5%. Results Patient characteristics Across 211 hospitals in England and Wales, (47 341 [40.1%] STEMI and 70 734 [59.9%] NSTEMI; mean age 68.5 (SD 14.0) years; 33.2% female), there were 37 487 (34.2%) patients with a history of ischaemic heart disease, 24 068 (21.2%) with diabetes, 5,579 (5.1%) with a history of heart failure, and 6678 (6.2%) with chronic renal failure (Table 1). The mean GRACE score was 119.8 (SD 34.1). Following hospitalization, 83 740 (78.2%) received coronary angiography and 2605 (2.5%) coronary artery bypass grafting [CABG]. Of STEMI, 21 567 (56.7%) received primary percutaneous coronary intervention [PCI] and of NSTEMI, 23 172 (40.2%) received sub-acute or elective PCI. Hospital variation in patients’ characteristics was small, compared with wider variation in QI attainment (Tables 1 and 2, Figure 1). At 30 days, there were 7063 (7.1%) deaths. Figure 1 Distribution of hospitals’ performance according to the European Society Cardiology; Acute Cardiovascular Care Association quality indicators for acute myocardial infarction. Table 1 Baseline characteristics and their hospital variation of patients with acute myocardial infarction, MINAP 2012–13

Results Patient characteristics Across 211 hospitals in England and Wales, (47 341 [40.1%] STEMI and 70 734 [59.9%] NSTEMI; mean age 68.5 (SD 14.0) years; 33.2% female), there were 37 487 (34.2%) patients with a history of ischaemic heart disease, 24 068 (21.2%) with diabetes, 5,579 (5.1%) with a history of heart failure, and 6678 (6.2%) with chronic renal failure (Table 1). The mean GRACE score was 119.8 (SD 34.1). Following hospitalization, 83 740 (78.2%) received coronary angiography and 2605 (2.5%) coronary artery bypass grafting [CABG]. Of STEMI, 21 567 (56.7%) received primary percutaneous coronary intervention [PCI] and of NSTEMI, 23 172 (40.2%) received sub-acute or elective PCI. Hospital variation in patients’ characteristics was small, compared with wider variation in QI attainment (Tables 1 and 2, Figure 1). At 30 days, there were 7063 (7.1%) deaths. Figure 1 Distribution of hospitals’ performance according to the European Society Cardiology; Acute Cardiovascular Care Association quality indicators for acute myocardial infarction. Table 1 Baseline characteristics and their hospital variation of patients with acute myocardial infarction, MINAP 2012–13 Total cohort (n = 118 075) Missing data, n,  (%) Hospital variation, mean  (SD) or median  (IQR) (n = 211 hospitals) Demographics Age in years, mean  (SD) 68.5  (14.0) 63  (0.05) 69.5  (3.4) Female 39 088  (33.2) 352  (0.30) 34.9  (30.4–38.6) Medical history History of ischaemic heart disease 37 487  (34.2) 8491  (7.2) 36.1  (29.4–40.9) Hypertension 55 397  (50.6) 8522  (7.2) 49.7  (44.1–56.0) Diabetes 24 068  (21.2) 4730  (4.01) 21.6  (17.7–24.4) Dyslipidaemia 36 890  (34.2) 10 296  (8.72) 30.4  (22.6–40.6) Family history of ischaemic heart disease 28 936  (30.5) 23 281  (19.7) 22.1  (16.3–30.8) Smoker  (current or previous) 67 670  (61.4) 7933  (6.7) 57.7  (50.4–63.5) Peripheral vascular disease 4699  (4.3) 9659  (8.2) 3.9  (2.6–5.3) Congestive cardiac failure 5579  (5.1) 9247  (7.8) 5.3  (2.93–8) COPD 16 326  (15.0) 9323  (7.9) 15.1  (12.1–17.6) Chronic kidney disease 6678  (6.2) 9854  (8.4) 6.1  (3.4–8.7) Cerebrovascular disease 9070  (8.4) 9489  (8.0) 8.2  (5.7–10.9) Clinical Presentation Heart rate at hospitalization, median  (IQR), b.p.m.

5.3) Congestive cardiac failure 5579  (5.1) 9247  (7.8) 5.3  (2.93–8) COPD 16 326  (15.0) 9323  (7.9) 15.1  (12.1–17.6) Chronic kidney disease 6678  (6.2) 9854  (8.4) 6.1  (3.4–8.7) Cerebrovascular disease 9070  (8.4) 9489  (8.0) 8.2  (5.7–10.9) Clinical Presentation Heart rate at hospitalization, median  (IQR), b.p.m. 78  (66–91) 18 887  (16.0) 78  (76–80) Systolic blood pressure at hospitalization, mean  (SD), mmHg 136  (27.8) 18 794  (15.9) 139  (5.2) Out of hospital cardiac arrest 3287  (2.9) 3737  (3.2) 1.9  (0.6–3.3) Initial creatinine, median  (IQR), µmol/L 86  (72–107) 11 622  (9.8) 87  (83–90) ST-segment deviation on admission 61 439  (53.5) 3140  (2.7) 38.8  (25.8–57.3) Killip classc 36 310  (30.8) I 64 254  (78.6) 78.1  (70.6–86.6) II 11 697  (14.3) 14.1%  (6.2–21.0) III 4424  (5.4) 5.0  (3.1–8.0) IV 1390  (1.7) 0.6  (0–1.7) GRACE risk score, mean  (SD) 119.8  (34.1) 33 536  (28.4) 119.6  (114.3–123.7) In hospital revascularization,  (of those eligible) Angiogramb 83 740  (78.2) 4210  (3.6) 67.5  (52.7–85.0) PCI 67 740  (65.6) 15 388  (13.0) 45.9  (28.0–73.7) CABG 2605  (2.5) 15 388  (13.0) 1.2  (0.2–3.5) Medications at discharge,d Aspirin 101 591  (98.1) 1374  (1.2) 98.7  (96.7–99.6) P2Y12 inhibitor 92 501  (87.1) 1434  (1.2) 89.3  (83.9–93.5) β-blocker 86 543  (95.6) 1412  (1.3) 96.5  (92.8–98.8) Statin 84 421  (96.5) 1275  (1.14) 96.9  (93.1–98.9) ACEi/ARB 84 681  (93.9) 1480  (1.33) 94.6  (89.5–98.2) Lifestyle advice, Cardiac rehabilitation 88 302  (81.7) 4340  (3.7) 85.5  (70.5–94.9) Smoking cessation advice 27 848  (74.4) 2222  (3.3) 78.0  (54.7–90.5) Dietary advice 81 745  (77.4) 7484  (6.3) 86.5  (55.3–95.9) Values presented are given as number  (percentage) unless stated.

(93.9) 1480  (1.33) 94.6  (89.5–98.2) Lifestyle advice, Cardiac rehabilitation 88 302  (81.7) 4340  (3.7) 85.5  (70.5–94.9) Smoking cessation advice 27 848  (74.4) 2222  (3.3) 78.0  (54.7–90.5) Dietary advice 81 745  (77.4) 7484  (6.3) 86.5  (55.3–95.9) Values presented are given as number  (percentage) unless stated. SD, standard deviation; IQR, interquartile range; IHD, ischaemic heart disease; COPD, chronic obstructive pulmonary disease; GRACE, Global Registry Acute Coronary Events; ACEi/ARB, angiotensin converting enzyme inhibitor/angiotensin II receptor blocker; PCI, percutaneous coronary intervention; CABG, coronary artery bypass grafting. a History of ischaemic heart disease refers to a history of CABG, MI, PCI, or angina. b Received angiography or PCI. c Killip class; 1: No clinical signs of heart failure, 2: Rales/crackles within the lungs, present S3, elevated JVP, 3: frank pulmonary oedema, 4: cardiogenic shock. d Includes eligible patients who started mediations during admission. Table 2 Overall and hospital variation in performance according to the European Society Cardiology; Acute Cardiovascular Care Association quality indicators for acute myocardial infarction

c Killip class; 1: No clinical signs of heart failure, 2: Rales/crackles within the lungs, present S3, elevated JVP, 3: frank pulmonary oedema, 4: cardiogenic shock. d Includes eligible patients who started mediations during admission. Table 2 Overall and hospital variation in performance according to the European Society Cardiology; Acute Cardiovascular Care Association quality indicators for acute myocardial infarction Domain Quality indicator Type of quality indicator Total patients eligible  (n) Proportion receiving care  (%) Hospital variation median % (IQR)  (n = 211 hospitals) 1: Centre organization 1.1 Centre organization: part of network Main 76 099 77.8 74.9  (48.6–85.3) 1.1a: Single emergency phone number to allow medical triage 118 075 100 100  (100–100) 1.1b: Pre hospital interpretation of the ECG 76 099 77.8 74.9  (48.6–85.3) 1.1c: Pre hospital activation of the catheter lab 118 075 100 100  (100–100) 1.2: Routine assessment of times to reperfusion Secondary 118 075 100 100  (100–100) 1.3: Participate in regular registry Secondary 100 100  (100–100) 2: Reperfusion/invasive strategy 2.1: Reperfusion within 12 h of presentation  (STEMI) Main 33 151 89.3 76.7  (33.3–91.4) 2.2 Timely reperfusion  (STEMI) Main 27 892 74.6 66.4  (35.8–78.8) 2.2a: Fibrinolysis received within 30 min  (PPCI centres and STEMI patients only) 547 55.0 33.3  (0–60.6) 2.2b: Primary PCI received within 60 min  (PPCI centres and STEMI patients only) 26 358 75.0 69.9  (54.6–80.8) 2.2c: Primary PCI received within 120 min  (non-PPCI centres and STEMI patients only) 672 93.9 40.0  (0–53.3) 2.2d: Door-in door-out time within 30 min  (non-PPCI centres and STEMI patients only) 538 23.8 5.7  (0–49.6) 2.3: Coronary angiography received within 72 h  (NSTEMI patients only).

26 358 75.0 69.9  (54.6–80.8) 2.2c: Primary PCI received within 120 min  (non-PPCI centres and STEMI patients only) 672 93.9 40.0  (0–53.3) 2.2d: Door-in door-out time within 30 min  (non-PPCI centres and STEMI patients only) 538 23.8 5.7  (0–49.6) 2.3: Coronary angiography received within 72 h  (NSTEMI patients only). Main 29 199 61.3 52.0  (39.1–73.2) 2.4: Time from diagnosis to wire passage  (STEMI), minutes  (median, IQR) Secondary 27 029 185  (135–284) 187.8  (169.8–210) 3: In Hospital risk assessment 3.1: GRACE risk score recorded in notes Main N/A 0 N/A 3.2: CRUSADE risk score recorded in notes Main N/A 0 N/A 3.3: LV function recorded in notes Main 104 004 54.5 54.7  (32.7–73.2) 4: Anti thrombotics during hospital 4.1: Adequate P2Y12 inhibition on discharge Main 106 157 87.1 89.3  (83.8–93.5) 4.2: Fondaparinux received  (NSTEMI patients only) Main 61 152 50.3 66.8  (0.4–84.7) Fondaparinux or LMWH received  (NSTEMI patients only) 61 185 85.2 90.3  (84.2–94.6) 4.3: DAPT received on discharge Secondary 101 582 88.1 90.5  (85.4–94.1) 5: Secondary prevention 5.1: High intensity statins on discharge Main N/A 0 N/A 5.2: ACEi/ARB on discharge for those with HF or LVEF ≤40 Secondary 33 531 94.2 95.5  (89.1–98.4) 5.3: β-blocker on discharge for those with HF or LVEF ≤40 Secondary 34 150 95.8 96.8  (92.0–99.0) 6: Patient satisfaction 6.1 Not applicable Main N/A N/A N/A 7: Composite QI 7.1 Composite QI  (opportunity-based) Main 118 071 83.3  (75.0–100) 82.8  (77.6–89.1) 7.2 Composite QI  (all-or-none, overall score) Secondary 118 075 83.1 84.8  (76.7–90.5) 7.2a Composite QI  (all-or-none, 3 measures)  (%) patients with no HF or LVEF≤40 Secondary 72 648 84.8 88.8  (81.8–93.3) 7.2b Composite QI  (all-or-none, 5 measures)  (%) patients with HF or LVEF≤40 Secondary 45 427 80.2 79.9  (69.5–87.0) 7.3 Mortality rate adjusted for GRACE risk 84 539 6.9 S.D 10.4 6.7  (5.4–7.9) Abbreviations: IQR, interquartile range; ECG, electrocardiogram; STEMI, ST-elevation myocardial infarction; PPCI, primary percutaneous coronary intervention; PCI, percutaneous coronary intervention; NSTEMI, non-ST elevation myocardial infarction; GRACE, Global Registry Acute Coronary Events; CRUSADE, Can Rapid Risk Stratification of Unstable Angina Patients Suppress ADverse Outcomes with Early Implementation of the ACC/AHA Guidelines); LV, left ventricular; LMWH, low molecular weight heparin; eGFR, estimated glomerular filtration rate; DAPT, dual antiplatelet therapy; ACEi/ARB, angiotensin converting enzyme inhibitor/a

Rapid Risk Stratification of Unstable Angina Patients Suppress ADverse Outcomes with Early Implementation of the ACC/AHA Guidelines); LV, left ventricular; LMWH, low molecular weight heparin; eGFR, estimated glomerular filtration rate; DAPT, dual antiplatelet therapy; ACEi/ARB, angiotensin converting enzyme inhibitor/a ngiotensin II receptor blocker; QI, quality indicator; HF, heart failure; LVEF, left ventricular ejection fraction; N/A, not applicable. Domains and quality indicators Of the 7 QI domains, MINAP contained data fields for the assessment of all, except the evaluation of patient satisfaction (Table 2). MINAP allowed the assessment of care according to 16 of the 20 QIs; 12 derived directly from corresponding data fields and 4 ascertained indirectly. The remaining 4 quality indicators including, the prescription of high intensity statins at hospital discharge, the recording of the GRACE and CRUSADE risk scores, and patient satisfaction could not be evaluated because MINAP did not collect this information. Figure 1 demonstrates the attainment and variation at a centre level for those QIs measured. Domain 1: Centre organization . Overall, 77.8% (n = 76 099) of eligible patients had pre-hospital interpretation of an ECG, higher than the median value for hospitals 74.8% (IQR 48.6–85.3%). For the remaining components of the main QI and both of the secondary QIs, the level of attainment for patients with AMI was 100% (n = 118 075) (Table 2).

tion . Overall, 77.8% (n = 76 099) of eligible patients had pre-hospital interpretation of an ECG, higher than the median value for hospitals 74.8% (IQR 48.6–85.3%). For the remaining components of the main QI and both of the secondary QIs, the level of attainment for patients with AMI was 100% (n = 118 075) (Table 2). Domain 2: Reperfusion/invasive strategy. For STEMI, 89.3% (n = 33 151) received reperfusion <12 h of onset of symptoms, and 74.6% (n = 20 815) received timely reperfusion. (Table 2) The median time from first medical contact with STEMI to wire passage for the whole cohort was 185 (IQR 135–284) min which was similar to the median time for the hospitals, although variation was less (188 min, IQR 170–210 min). For NSTEMI, the performance of coronary angiography was low and varied between hospitals (median, 52.0%, IQR 39.1–73.2%), and 61.3% (n = 17 895) received coronary angiography <72 h of hospitalization. Domain 3: In hospital risk assessment. Only one of the main QIs could be assessed; the assessment of LVEF was recorded in 54.5% (n = 56 680) of eligible patients, and demonstrated suboptimal attainment which varied by hospital (median 54.5%, IQR 32.7–73.2%).

Domain 2: Reperfusion/invasive strategy. For STEMI, 89.3% (n = 33 151) received reperfusion <12 h of onset of symptoms, and 74.6% (n = 20 815) received timely reperfusion. (Table 2) The median time from first medical contact with STEMI to wire passage for the whole cohort was 185 (IQR 135–284) min which was similar to the median time for the hospitals, although variation was less (188 min, IQR 170–210 min). For NSTEMI, the performance of coronary angiography was low and varied between hospitals (median, 52.0%, IQR 39.1–73.2%), and 61.3% (n = 17 895) received coronary angiography <72 h of hospitalization. Domain 3: In hospital risk assessment. Only one of the main QIs could be assessed; the assessment of LVEF was recorded in 54.5% (n = 56 680) of eligible patients, and demonstrated suboptimal attainment which varied by hospital (median 54.5%, IQR 32.7–73.2%). Domain 4: Antithrombotic treatment during hospitalization . The prescription of adequate P2Y12 inhibition at discharge was achieved in 87.1% (n = 92 501), but varied across hospitals (median 89.3%, IQR 83.8–93.5). Fondaparinux use was low (50.3%, n = 30 737) and exhibited the greatest hospital variation (median 66.8%, IQR 0.4–84.7%). However, when fondaparinux or low molecular weight heparin was considered, performance improved and variation reduced (median 90.3%, IQR 84.2–94.6%). The secondary QI found that 88.1% (n = 89 488) of eligible patients with AMI were discharged on dual antiplatelet therapy.

spital variation (median 66.8%, IQR 0.4–84.7%). However, when fondaparinux or low molecular weight heparin was considered, performance improved and variation reduced (median 90.3%, IQR 84.2–94.6%). The secondary QI found that 88.1% (n = 89 488) of eligible patients with AMI were discharged on dual antiplatelet therapy. Domain 5: Secondary prevention discharge treatments. In total, 96.5% (n = 84 421) of patients eligible for lipid lowering therapy were prescribed a statin at time of discharge from hospital. For the two secondary QIs, 94.2% (n = 31 569) of patients with AMI and heart failure or a LVEF ≤0.40 received an ACEi or ARB, and 95.8% (n = 32 728) with AMI and heart failure or a LVEF ≤0.40 received a β-blocker. Hospital attainment was high, but varied between hospitals (IQR 89.1–98.4 and 92.0–99.0, respectively). Domain 6: Patient Satisfaction. The MINAP registry recorded no data about patient satisfaction during the period of study. However, 81.7% (n = 108 110) of patients were referred for cardiac rehabilitation, 77.4% (n = 105 603) received dietary advice and 74.4% (n = 37 443) of current smokers received cessation advice. Across hospitals 85.5% of patients were referred for cardiac rehabilitation (IQR 70.5–94.9), 86.5% of smokers received cessation advice (IQR 54.7–90.5%), and 86.5% were offered dietary advice (IQR 55.3–95.9%).

.4% (n = 105 603) received dietary advice and 74.4% (n = 37 443) of current smokers received cessation advice. Across hospitals 85.5% of patients were referred for cardiac rehabilitation (IQR 70.5–94.9), 86.5% of smokers received cessation advice (IQR 54.7–90.5%), and 86.5% were offered dietary advice (IQR 55.3–95.9%). Domain 7: Composite quality indicators and GRACE risk score adjusted 30-day mortality. According to the opportunity-based composite score, patients received 83.3% (n = 118 071) of the interventions for which they were eligible. Hospital attainment was high, but varied between hospitals (median of 82.8%, IQR 77.5–88.7%). For the all-or-none composite score, 83.1% of patients received all of the interventions for which they were eligible which varied more among patients with heart failure or an ejection fraction ≤0.40 than those without (IQR 69.5–87.0 vs. 81.8–93.3). For the cohort, the median GRACE risk score adjusted 30-day mortality was 2.7% (IQR 0.9–8.1%). At the hospital level, variation was limited (median 6.7%, IQR 5.4–7.9%) (Figure 2). Figure 2 Caterpillar plot of hospital rank of hospital mean unadjusted and mean GRACE risk score adjusted hospital 30-day mortality rates.

. For the cohort, the median GRACE risk score adjusted 30-day mortality was 2.7% (IQR 0.9–8.1%). At the hospital level, variation was limited (median 6.7%, IQR 5.4–7.9%) (Figure 2). Figure 2 Caterpillar plot of hospital rank of hospital mean unadjusted and mean GRACE risk score adjusted hospital 30-day mortality rates. Correlation of quality indicators Overall 39 of the 45 QI to QI combinations demonstrated a weak correlation (Spearman correlation coefficient <0.3), 4 had a significant moderate correlation (0.4–0.7, all P < 0.001) and 2 had a significant strong correlation (≥0.7, all P < 0.001) (Figure 3). Figure 3 Scatter matrix of European Society Cardiology; Acute Cardiovascular Care Association quality indicators for acute myocardial infarction showing pairwise correlations of all quality indicator pairs, presented alongside Spearman’s rank correlation coefficient (where * indicates P < 0.05, ** P < 0.01, ***P < 0.001).

ure 3 Scatter matrix of European Society Cardiology; Acute Cardiovascular Care Association quality indicators for acute myocardial infarction showing pairwise correlations of all quality indicator pairs, presented alongside Spearman’s rank correlation coefficient (where * indicates P < 0.05, ** P < 0.01, ***P < 0.001). Quality indicators and mortality Eleven QIs had a significant inverse association with 30-day mortality (all P < 0.005) (Figure 4). The association with the greatest magnitude was for high attainment vs. zero attainment of the composite opportunity-based QI (odds ratio [OR] 0.04, 95% confidence interval [CI] 0.04–0.05, P < 0.001). Increasing patient attainment of this indicator from low (OR 0.42 CI 0.37–0.49, P < 0.001) to intermediate (0.15, 0.13–0.16, P < 0.001) to high (0.05, 0.04–0.06, P < 0.001) was significantly associated with a lower risk of 30-day mortality. On average, a unit increase in percentage attainment was significantly associated with a 3% decrease in 30-day mortality (0.97, 0.97–0.97, P < 0.001). Figure 4 Association between the European Society Cardiology; Acute Cardiovascular Care Association quality indicators for acute myocardial infarction and crude 30-day mortality. The composite opportunity QI was divided into the following categories: zero–received no interventions out of those eligible for, low–received <40% of interventions eligible for, intermediate–received ≥40 to <80% of interventions eligible for and high–received ≥80% of interventions eligible for.

crude 30-day mortality. The composite opportunity QI was divided into the following categories: zero–received no interventions out of those eligible for, low–received <40% of interventions eligible for, intermediate–received ≥40 to <80% of interventions eligible for and high–received ≥80% of interventions eligible for. Discussion Using a nationwide clinical database, comprising an analytical cohort of nearly 120 000 patients and over 200 hospitals between 2012–13, we found that the ESC ACCA QIs for AMI allowed the thorough evaluation of AMI care against international standards. The majority of QIs assessed were significantly inversely associated with 30-day mortality, the strongest being a composite indicator which, with increasing attainment, was associated with decreasing rates of death in a dose-response manner. Whilst we found high levels of performance with associated low levels of mortality, there was evidence for between hospital variation in key metrics, which mapped to class 1 guideline-indicated care. As such, the ESC QIs for AMI are applicable and valid, highlighting where in health systems there is potential to improve care and that high levels of performance according to the QIs is likely to reduce unwarranted variation and premature death from AMI.

n key metrics, which mapped to class 1 guideline-indicated care. As such, the ESC QIs for AMI are applicable and valid, highlighting where in health systems there is potential to improve care and that high levels of performance according to the QIs is likely to reduce unwarranted variation and premature death from AMI. Data from the EUROASPIRE studies show that the use of evidence-based treatments for AMI and associated outcomes vary widely between European countries.12–14 Other international comparisons provide evidence for variation both between and within countries.1,3,15–17 When there are data to show that adherence to guidelines improves clinical outcomes,3,18 variation in healthcare performance against set standards serves as indirect evidence for the potential to modify mortality. Results from our study show that whilst there was variation between hospitals in baseline patient characteristics, qualitatively this was less that that derived for the QIs, suggesting that the provision of treatments may have a greater role in determining outcomes than case mix.19

or the potential to modify mortality. Results from our study show that whilst there was variation between hospitals in baseline patient characteristics, qualitatively this was less that that derived for the QIs, suggesting that the provision of treatments may have a greater role in determining outcomes than case mix.19 The association between risk adjusted mortality and the ESC ACCA QIs is consistent with previous findings.20,21 However, this study evaluates a wider spectrum of QIs which map to guidelines and transcend the pathway of AMI care as well as organization of services. Four of the QIs were not associated with a significant reduction in mortality. Coronary angiography <72 h for patients with NSTEMI demonstrated a positive association with mortality. When further investigated, however, the provision of PCI was inversely associated with mortality (OR 0.58, 95% CI 0.51–0.65), in keeping with other evidence.19 We noted many QIs did not correlate with each other, implying they cover the spectrum of the AMI care pathway. The greatest correlation was between the prescription of P2Y12 inhibitors and dual antiplatelet therapy–given that the former is essential for the latter, this is not unexpected. The weakest association was for centre organization and timely angiography in NSTEMI; given that in the United Kingdom centre organization was primarily arranged to treat STEMI, it is not surprising that these two measures did not correlate.

apy–given that the former is essential for the latter, this is not unexpected. The weakest association was for centre organization and timely angiography in NSTEMI; given that in the United Kingdom centre organization was primarily arranged to treat STEMI, it is not surprising that these two measures did not correlate. Variation in the delivery of treatments is dependent upon many factors, including the availability of sufficient hospital facilities,16,22 a skilled workforce,15,22 high levels of knowledge transfer from scientific studies between healthcare professionals,23 the volume of cases admitted to the hospital,24 differences in the extent to which care is felt to be appropriate,25 as well as uniformity of recommendations from guidelines from different countries.17 Regarding the latter, we noted that the QI with the widest hospital variation was that for fondaparinux. We speculate this may be because the United Kingdom (2010) guideline for the management of patients with AMI,26 recommended fondaparinux only for cases of AMI who were not going for angiography <24 h of hospitalization–therein differing from current ESC recommendations. The QI with the least variation was centre organization. This is because, in the United Kingdom, emergency care for STEMI is institutionally operationalized as a result of the implementation of a national primary PCI service.22

for angiography <24 h of hospitalization–therein differing from current ESC recommendations. The QI with the least variation was centre organization. This is because, in the United Kingdom, emergency care for STEMI is institutionally operationalized as a result of the implementation of a national primary PCI service.22 In North America, there is a well-established program of quality improvement that, for AMI is based upon the ACTION registry27 and allows benchmarking of performance comparisons of providers. For the European community, the ESC AMI QIs offer an opportunity to study and consequently address deficits in care for cardiovascular disease. We demonstrate that this is possible only through a comprehensive clinical registry, as have others,27 but which for several European countries is lacking.1

isons of providers. For the European community, the ESC AMI QIs offer an opportunity to study and consequently address deficits in care for cardiovascular disease. We demonstrate that this is possible only through a comprehensive clinical registry, as have others,27 but which for several European countries is lacking.1 Although this study has many strengths, we recognize its limitations. The findings are summary measures of performance grounded on patient-level data and described at a hospital level. We followed the ESC AMI QI specification for the calculation of adjusted mortality,4 being mindful that hospital-specific influences were not accounted for in the modelling. Although we excluded hospitals with ≤30 patients recorded during the study period, for some hospitals in the separate QI analytical cohorts had ≤30 patients. For the GRACE score, we used surrogates for both Killip class and creatinine in keeping with previous validation work.8,9 MINAP does not record the specific type of statins, so ‘statin prescription’ was used as a surrogate for high intensity statin. Similarly, because there was imperfect recording of Ticagrelor, we used instead receipt of P2Y12 inhibitor.

es for both Killip class and creatinine in keeping with previous validation work.8,9 MINAP does not record the specific type of statins, so ‘statin prescription’ was used as a surrogate for high intensity statin. Similarly, because there was imperfect recording of Ticagrelor, we used instead receipt of P2Y12 inhibitor. The recoding, measurement, and reporting of within and between country performance against validated QIs representing class 1 indicated care is the critical next step in the international effort to reduce the burden and variation in premature deaths due to cardiovascular disease across Europe. This study provides good evidence for the application of the ESC ACCA QIs for AMI to clinical registries for the evaluation of cardiovascular care and outcomes; demonstrating their significant inverse association with mortality. Furthermore, this study identified potentially modifiable variation within a high performing health system and sets a road map for the development of standardized data collection in other ESC member countries. Greater and more uniform adherence to guideline-indicated care will result in improved and less varied mortality from AMI. Supplementary material Supplementary material is available at European Heart Journal online. Supplementary Material Supplementary Data Click here for additional data file.

The recoding, measurement, and reporting of within and between country performance against validated QIs representing class 1 indicated care is the critical next step in the international effort to reduce the burden and variation in premature deaths due to cardiovascular disease across Europe. This study provides good evidence for the application of the ESC ACCA QIs for AMI to clinical registries for the evaluation of cardiovascular care and outcomes; demonstrating their significant inverse association with mortality. Furthermore, this study identified potentially modifiable variation within a high performing health system and sets a road map for the development of standardized data collection in other ESC member countries. Greater and more uniform adherence to guideline-indicated care will result in improved and less varied mortality from AMI. Supplementary material Supplementary material is available at European Heart Journal online. Supplementary Material Supplementary Data Click here for additional data file. Acknowledgements TBD and MH were funded by the British Heart Foundation (Project Grant PG/13/81/30474). as a research assistant and research fellow, respectively. The Myocardial Ischaemia National Audit Project (MINAP) is commissioned by the Health Quality Improvement Partnership (HQIP) as part of the National Clinical Audit and Patient Outcomes Programme We gratefully acknowledge the contribution of all hospitals and healthcare professions who participate in the MINAP registry.

The Myocardial Ischaemia National Audit Project (MINAP) is commissioned by the Health Quality Improvement Partnership (HQIP) as part of the National Clinical Audit and Patient Outcomes Programme We gratefully acknowledge the contribution of all hospitals and healthcare professions who participate in the MINAP registry. Funding TBD and MH were funded by the British Heart Foundation (Project Grant PG/13/81/30474) as a research assistant and research fellow, respectively. Conflict of interest: none declared.

Introduction The incidence of atrial fibrillation (AF), a type of cardiac arrhythmia which produces rapid and irregular heartbeat, rises steeply with age,1 and has major health consequences due to increased likelihood of thrombo-embolic complications and heart failure. The risk of stroke is increased as much as five-fold in patients with AF.2 Stroke is a risk factor for dementia,3,4 and there is consistent evidence that AF increases the risk of dementia, particularly in stroke patients.5 A recent meta-analysis reported associations of AF with cognitive impairment and dementia in those without stroke, although the associations were weaker than in stroke patients.6 Much of the evidence on AF and cognitive impairment, including studies in the recent systematic reviews,5,6 comes from data on the elderly among whom both conditions are more common, and their association might simply reflect manifestations of underlying systemic vascular disease. Disentangling the direction of association in studies on the elderly is complicated as bidirectional effects are common. For example, in older adults dementia is a risk factor for stroke and vice-versa.7,8 As the pathophysiological processes underlying dementia unfold over a very long period prior to clinical onset,9 further evidence of the importance of AF for cognitive decline and dementia needs to examine whether AF at younger ages increases risk of accelerated cognitive decline and dementia.

stroke and vice-versa.7,8 As the pathophysiological processes underlying dementia unfold over a very long period prior to clinical onset,9 further evidence of the importance of AF for cognitive decline and dementia needs to examine whether AF at younger ages increases risk of accelerated cognitive decline and dementia. To address some of these complexities, we sought to examine associations of incident AF with subsequent cognitive decline over the 45 to 85 year age span. Our aim was to determine the potential dose-response association of duration of exposure to AF on cognitive decline, and whether stroke and coronary heart disease (CHD) subsequent to AF mediated this association. We also examined whether AF increased dementia risk, and the extent to which it was mediated by stroke and CHD.

aim was to determine the potential dose-response association of duration of exposure to AF on cognitive decline, and whether stroke and coronary heart disease (CHD) subsequent to AF mediated this association. We also examined whether AF increased dementia risk, and the extent to which it was mediated by stroke and CHD. Methods Study population The Whitehall II study is an ongoing cohort study of persons originally employed by the British civil service. A total of 10 308 persons (33% women), aged 35–55 years, were recruited to the study over the years 1985–1988.10 All participants responded to a comprehensive questionnaire and underwent a uniform, structured clinical evaluation, consisting of measures of anthropometry, cardiovascular and metabolic risk factors and disease. Since the baseline medical examination, follow-up examinations have taken place approximately every 5 years. Participant consent and research ethics approvals (University College London (UCL) ethics committee) are renewed at each contact; the latest approval was by the Joint UCL/UCLH Committee on the Ethics of Human Research (Committee Alpha), reference number 85/0938.

xaminations have taken place approximately every 5 years. Participant consent and research ethics approvals (University College London (UCL) ethics committee) are renewed at each contact; the latest approval was by the Joint UCL/UCLH Committee on the Ethics of Human Research (Committee Alpha), reference number 85/0938. Assessment of AF Two sources were used: a twelve-lead resting ECG (Mingorec, Siemens Healthcare, Erlangen, Germany) at all 6 waves of data collection between 1985 and 2013, and interpreted for the presence of AF/flutter at the University of Glasgow (Prof. Macfarlane). Coding was carried out independently in duplicate by trained technicians; Minnesota codes 83x were used to identify cases of AF.11 The second source was the national hospital episode statistics (HES) database on hospital day cases and inpatients, using ICD9: 427.3 and ICD10: I48. The National Health Service (NHS) in the United Kingdom provides most of the health care, and record linkage is based on encrypted values of the NHS number, unique for each person.

source was the national hospital episode statistics (HES) database on hospital day cases and inpatients, using ICD9: 427.3 and ICD10: I48. The National Health Service (NHS) in the United Kingdom provides most of the health care, and record linkage is based on encrypted values of the NHS number, unique for each person. Cognitive function The cognitive test battery was introduced to the study in 1997 and data are available at four assessments until 2013. The tests, described below, provide a comprehensive assessment of cognitive function and are appropriate for this population composed of individuals younger than in most studies on cognitive ageing.12Memory was assessed using a 20-word free recall test. Participants were presented a list of one or two syllable words at two second intervals and were then asked to recall in writing as many of the words in any order with two minutes to do so. Reasoning, assessed via the Alice Heim 4-I test which is composed of a series of 65 verbal and mathematical reasoning items of increasing difficulty.13 It tests inductive reasoning, measuring the ability to identify patterns and infer principles and rules. Participants had 10 minutes to do this section. Verbal fluency, measures of phonemic and semantic fluency were used.14 Participants were asked to recall in writing as many words beginning with ‘s’ (phonemic fluency) and as many animal names (semantic fluency) as they could. One minute was allowed for each test and the scores combined for the analysis.

Reasoning, assessed via the Alice Heim 4-I test which is composed of a series of 65 verbal and mathematical reasoning items of increasing difficulty.13 It tests inductive reasoning, measuring the ability to identify patterns and infer principles and rules. Participants had 10 minutes to do this section. Verbal fluency, measures of phonemic and semantic fluency were used.14 Participants were asked to recall in writing as many words beginning with ‘s’ (phonemic fluency) and as many animal names (semantic fluency) as they could. One minute was allowed for each test and the scores combined for the analysis. A global cognitive score was created using all tests by first standardizing the raw scores to z-scores [mean = 0; standard deviation (SD) = 1] and then averaging and standardizing them.

Verbal fluency, measures of phonemic and semantic fluency were used.14 Participants were asked to recall in writing as many words beginning with ‘s’ (phonemic fluency) and as many animal names (semantic fluency) as they could. One minute was allowed for each test and the scores combined for the analysis. A global cognitive score was created using all tests by first standardizing the raw scores to z-scores [mean = 0; standard deviation (SD) = 1] and then averaging and standardizing them. Dementia We used comprehensive tracing of electronic health records using three databases: HES, Mental Health Services Data Set (MHSDS) and the mortality register. Record linkage until 3Ist of March 2015, using ICD-10 codes F00, F01, F02, F03, F05.1, G30, G31.0, G31.1, and G31.8 identified cases of dementia. MHSDS is a national database which contains information for persons in contact with mental health services in hospitals, outpatient clinics, and the community. Mortality data were drawn from the British national mortality register. The validity of dementia ascertainment in our study is supported by findings on changes in the global cognitive score, showing accelerated decline in global cognitive score in the 8–10 years before dementia diagnosis (Supplementary material online, Figure S1). This is in agreement with previous studies that used a ‘gold-standard’ dementia ascertainment procedure.15

tudy is supported by findings on changes in the global cognitive score, showing accelerated decline in global cognitive score in the 8–10 years before dementia diagnosis (Supplementary material online, Figure S1). This is in agreement with previous studies that used a ‘gold-standard’ dementia ascertainment procedure.15 Covariates Socio-demographic measures included age, sex, ethnicity (white, non-white), and education (lower secondary school or less, higher secondary school (usually achieved at age 18), and university or higher degree). Analyses were also adjusted for health behaviours assessed by questionnaire, including smoking (current-, ex-, never-smoker); alcohol consumption (units of alcohol consumed in a week: no/occasional alcohol consumption, moderate alcohol consumption (1–14 (21)units/week in women (men), and heavy alcohol consumption (≥14 (21) units in women (men)); physical activity categorized as active (≥2.5 h/week of moderate physical activity or ≥1 h/week of vigorous physical activity), inactive (<1 h/week of moderate and vigorous activity), and intermediate level of activity for all others; and dietary behaviour self-reported frequency of fruit and vegetable consumption (<once a day, once a day, >once a day).

tive (≥2.5 h/week of moderate physical activity or ≥1 h/week of vigorous physical activity), inactive (<1 h/week of moderate and vigorous activity), and intermediate level of activity for all others; and dietary behaviour self-reported frequency of fruit and vegetable consumption (<once a day, once a day, >once a day). Chronic diseases included were hypertension (systolic/diastolic 140/90 mmHg or antihypertensive medication), prevalent diabetes mellitus (fasting glucose ≥ 7.0 mmol/L, a 2-h post-load glucose ≥ 11.1 mmol/L, doctor-diagnosed diabetes, diabetes medication), heart failure (ICD codes: I50), coronary heart disease (CHD, ICD codes: I20–I25), stroke (ICD codes: I60–I64), and self-reported use of medication for cardiovascular disease. Statistical analysis We examined associations between AF and participant characteristics in 1997–1999. Flow chart of persons included in the analyses described below is shown in Supplementary material online, Figure S2.

Chronic diseases included were hypertension (systolic/diastolic 140/90 mmHg or antihypertensive medication), prevalent diabetes mellitus (fasting glucose ≥ 7.0 mmol/L, a 2-h post-load glucose ≥ 11.1 mmol/L, doctor-diagnosed diabetes, diabetes medication), heart failure (ICD codes: I50), coronary heart disease (CHD, ICD codes: I20–I25), stroke (ICD codes: I60–I64), and self-reported use of medication for cardiovascular disease. Statistical analysis We examined associations between AF and participant characteristics in 1997–1999. Flow chart of persons included in the analyses described below is shown in Supplementary material online, Figure S2. Analysis of cognitive decline Mixed-effects models,16 with AF duration as a time varying covariate was used to estimate differences in cognitive decline between those without AF to those with AF exposure for 5, 10, and 15 years. The basic model (Model 1) contained time (here age, centred at 65 years), time squared, years since incident AF (0 = no AF, in years for those with AF), sex, ethnicity, education, and year of birth. Subsequent analyses (Model 2) were adjusted for health behaviours and chronic diseases in 1997–1999. Further analyses examined the mediating roles of stroke (Model 3), CHD (Model 4), and both stroke and CHD (Model 5) by including them as time varying covariates (1997 to 2013) in the analysis. The model yielded mean effects for AF duration; subsequently an interaction term with age and age2 was used to assess whether these effects varied as a function of age.

oles of stroke (Model 3), CHD (Model 4), and both stroke and CHD (Model 5) by including them as time varying covariates (1997 to 2013) in the analysis. The model yielded mean effects for AF duration; subsequently an interaction term with age and age2 was used to assess whether these effects varied as a function of age. Analysis of incident dementia We first used Cox regression to analyse the association of AF with incidence of dementia. Participants were followed from 1985 until the record of dementia, death, or March 31st 2015, whichever came first. Age was the time scale and analyses adjusted for sex, ethnicity, and education (Model 1) and then for health behaviours and chronic diseases at baseline (Model 2). In subsequent analyses, we examined the mediating role of stroke and CHD over the follow-up in the association between AF and incidence of dementia. These analyses were carried out using multistate models with a Weibull distribution. These models are an extension of competing risks survival analysis, allowing simultaneous estimation of the risk associated with AF in a) the incidence of stroke, b) the risk of dementia in those with stroke, and c) the risk of dementia in those free of stroke. Age was used as the timescale, and models were adjusted for sociodemographic factors, health behaviours, and chronic diseases at baseline. We repeated the analyses replacing stroke with CHD and then with CVD (stroke and CHD). These analyses were undertaken using R (SmoothHazard); all other analyses used Stata version 14. A two-sided P-value < 0.05 was considered statistically significant.

aphic factors, health behaviours, and chronic diseases at baseline. We repeated the analyses replacing stroke with CHD and then with CVD (stroke and CHD). These analyses were undertaken using R (SmoothHazard); all other analyses used Stata version 14. A two-sided P-value < 0.05 was considered statistically significant. Results Analysis of cognitive decline A total of 7428 participants were included in the analysis, characteristics in 1997–1999 presented in Table 1. A total of 43% of participants had data at all four assessments, 29% at three, 15% at two and 13% at only one assessment; a total of 414 cases of AF were used in the analysis. Table 1 Sample characteristics at the start of cognitive testing (1997–1999) No AF AF P* N 7014 414 Male, % 69.7 84.3 <0.001 Age (years), M (SD) 55.5 (6.0) 58.8 (5.9) <0.001 Education (<secondary school), % 44 45.7 0.64 Ethnicity (white), % 91.3 95.4 0.003 Current smoker, % 10.5 8.7 0.29 Heavy alcohol consumption,a % 25 32.1 0.001 Poor diet,b % 27.9 24.9 0.42 Physically inactive,c % 18.2 13.5 0.004 Diabetes, % 4.3 5.6 0.21 Hypertension, % 28 43.7 <0.001 CVD, % 5.1 12.8 <0.001 CVD medication,d % 6.1 14.7 <0.001 Heart failure, % 0.04 0 0.68 M, mean; SD, standard deviation; AF, atrial fibrillation; CVD, cardiovascular disease. * P for heterogeneity. a Heavy alcohol consumption was defined as 14+ units/week in women and 21+ units/week in men. b Corresponds to fruit and vegetable consumption <once a day. c Corresponds to <1 h/week of moderate and <1 h/week of vigorous physical activity.

No AF AF P* N 7014 414 Male, % 69.7 84.3 <0.001 Age (years), M (SD) 55.5 (6.0) 58.8 (5.9) <0.001 Education (<secondary school), % 44 45.7 0.64 Ethnicity (white), % 91.3 95.4 0.003 Current smoker, % 10.5 8.7 0.29 Heavy alcohol consumption,a % 25 32.1 0.001 Poor diet,b % 27.9 24.9 0.42 Physically inactive,c % 18.2 13.5 0.004 Diabetes, % 4.3 5.6 0.21 Hypertension, % 28 43.7 <0.001 CVD, % 5.1 12.8 <0.001 CVD medication,d % 6.1 14.7 <0.001 Heart failure, % 0.04 0 0.68 M, mean; SD, standard deviation; AF, atrial fibrillation; CVD, cardiovascular disease. * P for heterogeneity. a Heavy alcohol consumption was defined as 14+ units/week in women and 21+ units/week in men. b Corresponds to fruit and vegetable consumption <once a day. c Corresponds to <1 h/week of moderate and <1 h/week of vigorous physical activity. d CVD medication includes antihypertensives, lipid lowering drugs, nitrates, antiplatelets, and anticoagulants.

a Heavy alcohol consumption was defined as 14+ units/week in women and 21+ units/week in men. b Corresponds to fruit and vegetable consumption <once a day. c Corresponds to <1 h/week of moderate and <1 h/week of vigorous physical activity. d CVD medication includes antihypertensives, lipid lowering drugs, nitrates, antiplatelets, and anticoagulants. The age range of participants at the beginning and end of follow-up was 45–69 years and 61–83 years; mean follow-up was 14.7 years and annual decline in global cognitive function was −0.050 SD (−0.054, −0.046). Excess cognitive decline, averaged across all age-groups, in analysis adjusted for all confounders (Model 2) was greater in those with AF for longer, P = 0.01 (Table 2 and Figure 1). Incident stroke (Model 3), or CHD (Model 4) over the follow-up did not explain the excess decline. However, when both stroke and CHD (Model 5) were taken into account, the estimate for excess cognitive decline in those with AF was no longer statistically significant. The interaction term to test whether the effect of AF duration on cognitive decline changed with age did not suggest differences (P > 0.05), despite stronger associations in the youngest and oldest participants (Table 2). Furthermore, consideration of the incidence of stroke and CHD (model 5) attenuated associations in all but the youngest group (15 years difference in decline = −0.27; 95% CI: −0.51, −0.03). Table 2 Estimates of decline in the global cognitive score over 15 years in those with atrial fibrillation (AF) compared with those without AF

consideration of the incidence of stroke and CHD (model 5) attenuated associations in all but the youngest group (15 years difference in decline = −0.27; 95% CI: −0.51, −0.03). Table 2 Estimates of decline in the global cognitive score over 15 years in those with atrial fibrillation (AF) compared with those without AF Covariates at baseline Stroke/CHD over the follow-up Model 1 Model 2 Model 1+all covariates Model 3 Model 2 + Stroke Model 4 Model 2 + CHD Model 5 Model 2 + CVD Beta (95% CI) Beta (95% CI) Beta (95% CI) Beta (95% CI) beta (95% CI) Mean (across all age-groups) 15 year Cognitive Decline NO AF Ref. Ref. Ref. Ref. Ref.

consideration of the incidence of stroke and CHD (model 5) attenuated associations in all but the youngest group (15 years difference in decline = −0.27; 95% CI: −0.51, −0.03). Table 2 Estimates of decline in the global cognitive score over 15 years in those with atrial fibrillation (AF) compared with those without AF Covariates at baseline Stroke/CHD over the follow-up Model 1 Model 2 Model 1+all covariates Model 3 Model 2 + Stroke Model 4 Model 2 + CHD Model 5 Model 2 + CVD Beta (95% CI) Beta (95% CI) Beta (95% CI) Beta (95% CI) beta (95% CI) Mean (across all age-groups) 15 year Cognitive Decline NO AF Ref. Ref. Ref. Ref. Ref. additional decline when AF for 5 years −0.05 (−0.09, −0.02)* −0.05 (−0.09, −0.01)* −0.04 (−0.08, −0.01)* −0.04 (−0.08, −0.002)* −0.03 (−0.07, 0.005) additional decline when AF for 10 years −0.11 (−0.18, −0.03)* −0.10 (−0.17, −0.02)* −0.09 (−0.16, −0.01)* −0.08 (−0.15, −0.004)* −0.07 (−0.14, 0.01) additional decline when AF for 15 years −0.16 (−0.27, −0.05)* −0.15 (−0.26, −0.04)* −0.13 (−0.24, −0.02)* −0.12 (−0.23, −0.01)* −0.10 (−0.21, 0.01) P for trend 0.005 0.01 0.02 0.04 0.09 15 year Cognitive Decline as a function of age CURRENT AGE 60 years Decline between 45 and 60 year, NO AF −0.43 (−0.49, −0.37)* additional decline when AF at 45 years −0.31 (−0.56, −0.07)* −0.30 (−0.54, −0.06)* −0.27 (−0.51, −0.04)* −0.29 (−0.53, −0.06)* −0.27 (−0.51, −0.03)* 65 years Decline between 50 and 65 year, NO AF −0.53 (−0.57, −0.49)* additional decline when AF at 50 years −0.18 (−0.33, −0.03)* −0.16 (−0.32, −0.01)* −0.14 (−0.29, 0.01) −0.17 (−0.32, −0.02)* −0.14 (−0.30, 0.01) 70 years Decline between 55 and 70 year, NO AF −0.63 (−0.66, −0.59)* additional decline when AF at 55 years −0.12 (−0.24, 0.01) −0.10 (−0.23, 0.02) −0.08 (−0.21, 0.04) −0.10 (−0.22, 0.03) −0.07 (−0.20, 0.05) 75 years Decline between 60 and 75 years, NO AF −0.73 (−0.77, −0.68)* additional decline when AF at 60 years −0.13 (−0.25, −0.01)* −0.12 (−0.24, 0.001) −0.10 (−0.22, 0.01) −0.08 (−0.20, 0.04) −0.06 (−0.18, 0.06) 80 years Decline between 65 and 80 years, NO AF −0.83 (−0.90, −0.76)* additional decline when AF at 65 years −0.22 (−0.40, −0.04)* −0.20 (−0.38, −0.02)* −0.20 (−0.38, −0.02)* −0.12 (−0.31, 0.06) −0.11 (−0.30, 0.08) 85 years Decline between 70 and 85 years, NO AF −0.93 (−1.02, −0.83)* additional decline when AF at 70 years −0.38 (−0.73, −0.03)* −0.36 (−0.71, −0.01)* −0.38 (−0.73, −0.03)* −0.22 (−0.58, 0.14) −0.21 (−0.57, 0.15) Interaction between AF duration and age, P 0.16 0.17 0.14 0.22 0.23 Participants aged 45–69 years in 1997–1999 were followed until 2012–2013, mean follow-up 14.7 years. Estimates are for decline over 15 years.

ears −0.38 (−0.73, −0.03)* −0.36 (−0.71, −0.01)* −0.38 (−0.73, −0.03)* −0.22 (−0.58, 0.14) −0.21 (−0.57, 0.15) Interaction between AF duration and age, P 0.16 0.17 0.14 0.22 0.23 Participants aged 45–69 years in 1997–1999 were followed until 2012–2013, mean follow-up 14.7 years. Estimates are for decline over 15 years. Total N = 7428, atrial fibrillation, N = 414. *P < 0.05; CHD, coronary heart disease; CVD, cardiovascular disease (Stroke or CHD). Model 1: Analysis uses age as the time-scale, adjusted for sex, education, and ethnicity. Model 2: Model 1 + alcohol consumption, smoking, physical activity, diet, diabetes, hypertension, heart failure, CVD (stroke or CHD) and CVD medication at baseline (1997–1999). Model 3: Model 2 + time-dependent Stroke (1997–2013), N = 109. Model 4: Model 2 + time-dependent CHD (1997–2013), N = 1120. Model 5: Model 2 + time-dependent CVD (1997–2013), N = 1182. Figure 1 Decline in the global cognitive score as a function of atrial fibrillation (AF). *Analysis uses age as the time scale, adjusted for sex, education, and ethnicity. P-values represent the test for trend for greater effects on cognitive decline in longer exposure to AF. Supplementary material online, Tables S1 and S2 and Figure S3 show these results separately for the tests of memory, reasoning, and fluency; there were no associations with memory, even in the minimally adjusted model (Supplementary material online, Table S1).

Figure 1 Decline in the global cognitive score as a function of atrial fibrillation (AF). *Analysis uses age as the time scale, adjusted for sex, education, and ethnicity. P-values represent the test for trend for greater effects on cognitive decline in longer exposure to AF. Supplementary material online, Tables S1 and S2 and Figure S3 show these results separately for the tests of memory, reasoning, and fluency; there were no associations with memory, even in the minimally adjusted model (Supplementary material online, Table S1). Analysis of dementia Of a total of 10 214 persons, followed over a mean 26.6 years, there were 912 cases of incident AF and 324 cases of dementia; 73% of the latter recorded in the last 5 years of follow-up. The mean (SD) age at incident AF and dementia diagnosis was 68.5 (7.7) and 74.9 (5.4) years, respectively. Table 3 shows that in analysis adjusted for all confounders (Model 2), those with AF had 87% excess risk (95% CI: 1.37, 2.55) of dementia. The association was somewhat stronger, albeit not statistically significant, in those with incident AF before 70 years (Table 3). Table 3 Association of atrial fibrillation (AF) with incidence of dementia

nalysis adjusted for all confounders (Model 2), those with AF had 87% excess risk (95% CI: 1.37, 2.55) of dementia. The association was somewhat stronger, albeit not statistically significant, in those with incident AF before 70 years (Table 3). Table 3 Association of atrial fibrillation (AF) with incidence of dementia Model 1 Model 2 Atrial fibrillation N total N cases HR (95% CI) HR (95% CI) No 9302 274 1.00 1.00 Yes 912 50 1.93 (1.42, 2.63) 1.87 (1.37, 2.55) Analysis stratified by age of onset of AF No 9302 274 1.00 1.00 AF before age 70 500 21 2.15 (1.37, 3.37) 2.11 (1.35, 3.32) AF after age 70 412 29 1.79 (1.20, 2.65) 1.72 (1.15, 2.55) P for interaction 0.50 0.45 Model 1: Analysis adjusted for age, sex, education, and ethnicity. Model 2: Model 1 + alcohol consumption, smoking, physical activity, diet, diabetes, hypertension, heart failure, CVD (stroke or CHD), and CVD medication at baseline. In multistate models, AF was associated with a 6.22 times increased risk of stroke (95% CI: 4.74, 8.16) and its association with dementia was not fully explained by stroke as demonstrated by the increased risk of dementia in those free of stroke (HR = 1.67; 95% CI: 1.17, 2.38), Figure 2A. Further analysis showed AF to increase risk of CHD (HR = 5.29; 95% CI: 4.50, 6.22; Figure 2B) and CVD (HR = 5.74; 95% CI: 4.95, 6.65; Figure 2C). The analysis of CVD (Figure 2C) shows that the association between AF and dementia was present in those with CVD (HR = 1.79; 95% CI: 1.04, 3.08) but not in those free of CVD (HR = 1.29; 95% CI: 0.74, 2.24).

crease risk of CHD (HR = 5.29; 95% CI: 4.50, 6.22; Figure 2B) and CVD (HR = 5.74; 95% CI: 4.95, 6.65; Figure 2C). The analysis of CVD (Figure 2C) shows that the association between AF and dementia was present in those with CVD (HR = 1.79; 95% CI: 1.04, 3.08) but not in those free of CVD (HR = 1.29; 95% CI: 0.74, 2.24). Figure 2 Multistate models for the role of atrial fibrillation in transitions* to stroke (A), CHD (B), stroke or CHD (C), and dementia. *Role of AF (time varying) in the risk of transitions from: (a) Healthy to stroke (A), CHD (B), and CVD (stroke or CHD, C); (b) Stroke (A, N (AF/Stroke) = 78/352), CHD (B, N (AF/CHD) = 179/1790), and CVD (Stroke or CHD, C, N (AF/CVD) = 217/2024) to dementia; (c) Healthy to dementia in those free of Stroke (A), CHD (B), and CVD (Stroke or CHD, C). Analyses with age as timescale and adjusted for sex, education, ethnicity, alcohol consumption, smoking, physical activity, diet, diabetes, hypertension, heart failure, CVD, and CVD medication at baseline.

= 217/2024) to dementia; (c) Healthy to dementia in those free of Stroke (A), CHD (B), and CVD (Stroke or CHD, C). Analyses with age as timescale and adjusted for sex, education, ethnicity, alcohol consumption, smoking, physical activity, diet, diabetes, hypertension, heart failure, CVD, and CVD medication at baseline. Discussion Our study of cognitive decline over 15 years in adults aged 45–69 years at the start of follow-up shows greater cognitive decline in those with longer exposure to AF. Stroke occurring after the onset on AF did not explain this excess decline but when both stroke and CHD over the follow-up were taken into account the association between duration of exposure to AF and cognitive decline was no longer statistically significant. This finding was replicated in analysis of dementia where stroke explained only part of the association of AF with dementia. Importantly, these findings relate to incident AF in relatively young adults as more than two thirds of AF cases in our analysis occurred before 75 years of age.

longer statistically significant. This finding was replicated in analysis of dementia where stroke explained only part of the association of AF with dementia. Importantly, these findings relate to incident AF in relatively young adults as more than two thirds of AF cases in our analysis occurred before 75 years of age. AF is a common disorder in the elderly,17 its prevalence increases with age, doubling every decade of life after the age of 50 years to reach 10–20% after the age of 80 years. As the prevalence of dementia also increases with age, its association with AF has sometimes been attributed to common age-related mechanisms. However, increasing evidence of associations with cognitive decline suggests that AF may indeed be a risk factor for cognitive dysfunction.18 To the best of our knowledge, few studies have been able to use serial cognitive testing, starting in mid-life to assess the impact of AF. We show that even in adults aged 60 years, those with incident AF at age 50 and 55 years had accelerated cognitive decline.

may indeed be a risk factor for cognitive dysfunction.18 To the best of our knowledge, few studies have been able to use serial cognitive testing, starting in mid-life to assess the impact of AF. We show that even in adults aged 60 years, those with incident AF at age 50 and 55 years had accelerated cognitive decline. The effect of duration of AF has been examined in relation to dementia where the associations appear to be stronger in those with younger age of AF onset.19–21 Stronger effects associated with longer duration of AF have also been seen for pre-stroke cognitive impairment,8 and total brain and grey matter volume.22 AF is associated with cerebral hypoperfusion and is a known cause of embolic stroke, from thrombus originating in the atrial appendage. It is possible that longer duration of exposure allows a greater time window for damage from chronic hypoperfusion and development of emboli and cardiac failure. AF alters atrial size, substrate, and cardiac function, which develop very early after diagnosis, and increases the risk of macro- and micro-cerebral ischaemic events.

sible that longer duration of exposure allows a greater time window for damage from chronic hypoperfusion and development of emboli and cardiac failure. AF alters atrial size, substrate, and cardiac function, which develop very early after diagnosis, and increases the risk of macro- and micro-cerebral ischaemic events. The association of AF with cognitive impairment,6,23cognitive decline,24 hippocampal atrophy,25and dementia,26 appears to be independent of stroke history. We extended these findings by considering stroke occurring after AF to show that it does not explain associations with cognitive decline and dementia. A third of those with AF have silent infarction27 and cerebrovascular thrombo-embolism together with global brain hypoperfusion due to impaired cardiac haemodynamics, may account for the increased risk of developing dementia in AF patients. Our findings highlight the importance of CVD in the association of AF with cognitive outcomes. The association of AF with dementia in our study is similar to that in previous studies,20,26 particularly studies that examined AF onset at younger ages.19 It remains unclear whether it affects risk of all types of dementia; one study found evidence of stronger associations with Alzheimer’s disease,20 another showed the type of dementia not to matter,19 while MRI data associations with lower total grey matter but not neurodegenerative changes characteristic of Alzheimer’s disease.27

unclear whether it affects risk of all types of dementia; one study found evidence of stronger associations with Alzheimer’s disease,20 another showed the type of dementia not to matter,19 while MRI data associations with lower total grey matter but not neurodegenerative changes characteristic of Alzheimer’s disease.27 The main strengths of our study include the large sample size, consideration of both cognitive decline and dementia, modelling of time lived with AF, and explicit analysis of the role played by stroke and CHD after AF onset—an important consideration given the strong association of AF with these conditions. Our results on the interaction effects with age suggest that that the effect of AF on cognitive decline is not confined to older ages. We present results using the global cognitive score, allowing replication across studies in the future. We have previously shown the risk factor-CHD associations in our study to be similar to that in a general population study,28 suggesting that the present findings are generalizable.

ine is not confined to older ages. We present results using the global cognitive score, allowing replication across studies in the future. We have previously shown the risk factor-CHD associations in our study to be similar to that in a general population study,28 suggesting that the present findings are generalizable. Limitations of the study include the method of dementia ascertainment, which has high specificity but is unlikely to be sensitive. The fact that dementia tracing was available on all participants in the study allows the study results to be free from attrition and selection biases that are common in studies on older adults. Furthermore, the AF-CVD association in our study was similar to that in other studies,2 suggesting that case definitions for both conditions are valid. We were not able to distinguish paroxysmal from persistent AF, in previous studies the latter was more strongly associated with silent cerebral ischaemia and worse cognitive function,23 and higher incidence of adverse cardiovascular outcomes even in patients treated by anticoagulated therapy.29 We were also limited by the smaller number of patients with AF, particularly in subgroup analysis (e.g. those with incident CVD) that resulted in broad confidence limits.

chaemia and worse cognitive function,23 and higher incidence of adverse cardiovascular outcomes even in patients treated by anticoagulated therapy.29 We were also limited by the smaller number of patients with AF, particularly in subgroup analysis (e.g. those with incident CVD) that resulted in broad confidence limits. In sum, AF is the most common arrhythmia and its most well established consequence is stroke. With ageing of populations its impact on cognitive impairment and dementia has become increasingly important and understanding this association has relevance for the development of preventive and therapeutic strategies.30 AF can be treated with antiarrhythmic medication, cardioversion, or catheter ablation. Furthermore, thrombo-prophylaxis with oral anticoagulation is effective in reducing stroke risk, whether the same is true for dementia is unknown. The present longitudinal study shows that early onset AF and its duration matters for cognitive decline and dementia and highlight the importance of effectively treating cardiovascular disease in AF patients.1 In those with early age of AF onset, a longer exposure period might lead to changes that produce greater neuronal injury and loss, possibly due to the interaction of degenerative and vascular changes. Supplementary material Supplementary material is available at European Heart Journal online.

In sum, AF is the most common arrhythmia and its most well established consequence is stroke. With ageing of populations its impact on cognitive impairment and dementia has become increasingly important and understanding this association has relevance for the development of preventive and therapeutic strategies.30 AF can be treated with antiarrhythmic medication, cardioversion, or catheter ablation. Furthermore, thrombo-prophylaxis with oral anticoagulation is effective in reducing stroke risk, whether the same is true for dementia is unknown. The present longitudinal study shows that early onset AF and its duration matters for cognitive decline and dementia and highlight the importance of effectively treating cardiovascular disease in AF patients.1 In those with early age of AF onset, a longer exposure period might lead to changes that produce greater neuronal injury and loss, possibly due to the interaction of degenerative and vascular changes. Supplementary material Supplementary material is available at European Heart Journal online. Funding The Whitehall II study is supported by grants from the US National Institute on Aging (R01AG013196; R01AG034454); the UK Medical Research Council (MRC K013351). M.K. is supported by the MRC and NordForsk. MRC (MR/K013351/1) via a prepayment account with UCL. Conflict of interest: none declared. Supplementary Material Supplementary Figures and Tables Click here for additional data file.

Introduction Statins are first choice lipid-modifying medications for prevention and management of cardiovascular diseases (CVD).1,2 The UK is one of the largest users of statins worldwide,3 and with revised NICE guidelines approximately 12 million UK individuals will be prescribed statins.4,5 While statins are generally well tolerated, neurological,6 gastro-intestinal, or muscle-based7,8 adverse drug reactions are reported. Adverse reactions to statins are likely to manifest as muscle aches (myalgia) along with elevated creatine phosphokinase (CK). Adherence to statin treatment is often negatively impacted in response to adverse reactions.9,10 The inability to adhere to statin treatment, whether due to statin-induced myalgia or more general forms of statin intolerance result in poor on-statin outcomes.11 Therefore, examining risk factors predisposing to statin intolerance is crucial from a public health perspective.

cted in response to adverse reactions.9,10 The inability to adhere to statin treatment, whether due to statin-induced myalgia or more general forms of statin intolerance result in poor on-statin outcomes.11 Therefore, examining risk factors predisposing to statin intolerance is crucial from a public health perspective. A genome-wide association study (GWAS) by Dubé et al.,12 reported a missense variant Asp247Gly in the leukocyte immunoglobulin-like receptor subfamily B member 5 gene, LILRB5 on chromosome 19, base position 54759361 (Human Genome Build GRCh37), was associated with circulating serum CK levels. The mean CK levels of Asp247 homozygotes (T/T) were significantly higher. This association was found to be independent of statin use, however there is no known biological mechanism for the variant in determining CK levels. A GWAS by Kristjansson et al.13 of over 60 000 Icelanders replicated the association of the variant and CK levels. The same study also reported the association of the variant with serum lactate dehydrogenase (LDH) levels in a population of over 90 000 Icelanders.13 The LILRB5 variant showed the same direction of effect, i.e. Asp247 homozygotes had higher LDH and CK levels. LDH is often used in conjunction with CK as a marker of tissue damage. The findings suggest the variant might impart a statin independent susceptibility to muscle-based events. This makes LILRB5 a potential marker for susceptibility to the commonly noted muscle-based symptoms attributed to statin intolerance.

K levels. LDH is often used in conjunction with CK as a marker of tissue damage. The findings suggest the variant might impart a statin independent susceptibility to muscle-based events. This makes LILRB5 a potential marker for susceptibility to the commonly noted muscle-based symptoms attributed to statin intolerance. These discoveries warrant an investigation into the role of the LILRB5 variant in statin intolerance. Population-based studies use surrogate markers of intolerance, such as elevations in CK, trends in statin treatment, dose reductions, switching or the discontinuation of therapy. Therefore, a priori, we considered two definitions of statin intolerance, one dependent on and one independent of elevated CK levels. We hypothesize that carriers of variant associated with higher muscle enzyme levels (CK and LDH) will also be predisposed to forms of statin intolerance independent of CK levels.

herapy. Therefore, a priori, we considered two definitions of statin intolerance, one dependent on and one independent of elevated CK levels. We hypothesize that carriers of variant associated with higher muscle enzyme levels (CK and LDH) will also be predisposed to forms of statin intolerance independent of CK levels. The principal cohort used was the Genetics of Diabetes Audit and Research, Tayside Scotland (GoDARTS). GoDARTS has been previously used to establish pharmacogenetic associations of genes such as the hepatic influx transporter SLCO1B1 and statin intolerance.14 At present, GoDARTS contains 11 912 statin users and provides approximately 98 000 person-years of statin exposure, providing an ideal cohort to examine the association of this genetic variant with statin intolerance. Replication was examined in the Clinical Practice Research Datalink (CPRD) STAGE study15 and clinically adjudicated cases of statin-induced myopathy (SIM) in the European PREDICTION-ADR consortium study.9 The interaction of this effect with statin use was then studied among participants who developed myalgia in the JUPITER (Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin) randomized clinical trial (RCT) where individuals were allocated rosuvastatin or placebo to assess the relative reduction in vascular events.16,17

was then studied among participants who developed myalgia in the JUPITER (Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin) randomized clinical trial (RCT) where individuals were allocated rosuvastatin or placebo to assess the relative reduction in vascular events.16,17 Methods GoDARTS cohort Study population Tayside Medical Ethics Committee approved the GoDARTS study and informed consent was obtained for all participants. The dataset contains complete electronic medical records (EMR), prescription information and laboratory results from 18 306 Scottish Caucasian individuals. Prescribing data were available from 1 January 1990 to 31 July 2013. In all, 10 149 study participants had type 2 diabetes (T2D) at recruitment and the remainder (8157) were recruited as non-diabetic controls. We performed a case-control study for statin intolerance in this population. We found 11 947 statin users from GoDARTS who had at least two prescriptions of statins. The prescription patterns indicating intolerance used in this study are similar to those used by Donnelly et al.14 to establish the association between statin intolerance and SLCO1B1 genotypes in the GoDARTS study. For features used to define statin intolerance and tolerance, see Supplementary Material online, Methods S1.

prescription patterns indicating intolerance used in this study are similar to those used by Donnelly et al.14 to establish the association between statin intolerance and SLCO1B1 genotypes in the GoDARTS study. For features used to define statin intolerance and tolerance, see Supplementary Material online, Methods S1. Statin intolerance in GoDARTS General statin intolerance Cases of general statin intolerance (GSI) were defined as users with CK raised above the ULN (Upper Limit of Normal), after start of statin therapy, who either switched statin therapies two or more times (not including systemic shifts to atorvastatin after patent expiration in the UK) or discontinued therapy (n = 588). Controls (statin tolerant individuals—ST1) had over 90% coverage with statin prescriptions, for a minimum of 5 years, at a minimum average daily dose of 40 mg of simvastatin (or equivalent dose of another statin), had consistently normal CKs while on statins, had never switched their statin therapy (except the systematic switch to atorvastatin) and had not discontinued therapy (n = 356).

statin prescriptions, for a minimum of 5 years, at a minimum average daily dose of 40 mg of simvastatin (or equivalent dose of another statin), had consistently normal CKs while on statins, had never switched their statin therapy (except the systematic switch to atorvastatin) and had not discontinued therapy (n = 356). Lowest approved daily starting dose statin intolerance Since the LILRB5 Asp247Gly variant was known to be associated with CK levels, a phenotype independent of CK elevations was created in order to determine if the association of the variant with statin intolerance was confounded by the variant‘s association with CK levels. The European Society of Cardiology and European Atherosclerosis Society joint recommendations on the management of dyslipidaemias suggest cessation of statin treatment if the user presents with normal CK, but persistent symptoms of intolerance.18 This intolerance definition was derived from the GAUSS-2 trial, the consensus definition based on recommendations by Banach et al. and recommendations of the National Lipid Association (NLA) in 2014.19–22

uggest cessation of statin treatment if the user presents with normal CK, but persistent symptoms of intolerance.18 This intolerance definition was derived from the GAUSS-2 trial, the consensus definition based on recommendations by Banach et al. and recommendations of the National Lipid Association (NLA) in 2014.19–22 Cases of low dose intolerance (LDI) had used two or more different statins, and at least one statin that was discontinued would have to be at the lowest approved daily starting dose (NLA, 2014) before discontinuation, irrespective of their CK levels (n = 591). Controls (statin tolerant individuals—ST2) met all the criteria of the previous statin tolerant (ST1) group, except the definition was independent of the CK elevation criteria (n = 443). A higher proportion of controls had CK levels in the normal range, specifically, 354 of the 443 (80%) controls compared to 335 of 591 (57%) cases. Therefore, association tests for this phenotype were adjusted for (log-transformed) CK levels, in addition to other covariates. For genotyping methods, see Supplementary material online, Methods S1. Validation of statin intolerance phenotypes in GoDARTS These phenotypic definitions of statin intolerance were validated against known SLCO1B1 genotype risk score14 and the outcome of major adverse cardiovascular event (see Supplementary material online, Methods S2). They were significantly associated with both.

For genotyping methods, see Supplementary material online, Methods S1. Validation of statin intolerance phenotypes in GoDARTS These phenotypic definitions of statin intolerance were validated against known SLCO1B1 genotype risk score14 and the outcome of major adverse cardiovascular event (see Supplementary material online, Methods S2). They were significantly associated with both. Replication studies CPRD-STAGE study: statin-induced myopathy Replication was sought from the CPRD-STAGE study.15 Data were available for 129 cases of SIM and 2501 population controls from the Wellcome Trust Case Control Consortium (WTCCC1).15 Cases of statin myopathy were identified using CPRD and tertiary muscle clinics and conformed to SIM classification standards.9 Analysis presented is unadjusted for covariates due lack of available data for the WTCCC population controls. For additional cohort information, see Supplementary material online, Methods S3.

Cases of statin myopathy were identified using CPRD and tertiary muscle clinics and conformed to SIM classification standards.9 Analysis presented is unadjusted for covariates due lack of available data for the WTCCC population controls. For additional cohort information, see Supplementary material online, Methods S3. PREDICTION-ADR: statin-induced myopathy Cases and controls for SIM were contributed by the consortium‘s study centres in Uppsala (Sweden), Dundee and Liverpool (UK). Cases met criteria for classification of SIM.9 Identification of SIM from population cohorts: GoDARTS, Genetics of Scottish Health Registry (GoSHARE)23 and CPRD using EMRs was based on on-statin CK levels raised ≥4 times ULN. Subsequently, clinical adjudication was undertaken by physicians and specialists. Factors considered were resolution of CK after de-challenge, post-event prescribing changes (e.g. switching or total discontinuation), medical history of kidney disease, trauma, falls, myocardial infarction, thyroid disease and tests for HMGCR (3-Hydroxy-3-Methylglutaryl-CoA Reductase) antibodies, muscle biopsy and physical activity, if available. Cases were also identified from CVD clinics, general practitioners (GP) practices, and muscle disease clinics where adjudication was performed directly by physicians. Swedish cases were selected from Swedegene Biobank, which recruits patients reported to the adverse drug reaction registry at Medical Products Agency, Uppsala. Statin tolerant controls were on therapy for a minimum of 1 year with no recorded adverse events. Analysis was performed on 229 cases of SIM and 432 adjudicated controls of Caucasian ethnicity. Whole-exome sequencing was undertaken in laboratories in Liverpool, Dundee and Uppsala. For details of sequencing methods, see Supplementary material online, Methods S4. There were no sample overlaps from GoDARTS in the discovery and replication cohorts.

cases of SIM and 432 adjudicated controls of Caucasian ethnicity. Whole-exome sequencing was undertaken in laboratories in Liverpool, Dundee and Uppsala. For details of sequencing methods, see Supplementary material online, Methods S4. There were no sample overlaps from GoDARTS in the discovery and replication cohorts. JUPITER trial: myalgia The replication in JUPITER focused on 8749 study participants of verified European ancestry with available genetic data among whom 4381 were randomized to receive statin treatment and 4368 were randomized to placebo. The population demographics of the genotyped sub-population of the JUPITER trial has been previously described.16 The median follow-up period in the trial was 1.9 years, during which traits such as CK, therapy compliance, and myalgia were recorded.17 Myalgia was ascertained by physicians blinded to trial-allocation arm24 and 837 participants in the study sample were recorded as developing myalgia. Due to possible association between CK measures and diagnosis of myalgia, log-transformed final CK levels were included in analyses as a potential confounder. For genotyping methods, see Supplementary material online, Methods S1.

cation arm24 and 837 participants in the study sample were recorded as developing myalgia. Due to possible association between CK measures and diagnosis of myalgia, log-transformed final CK levels were included in analyses as a potential confounder. For genotyping methods, see Supplementary material online, Methods S1. Statistical analyses All statistical analyses on GoDARTS, CPRD-STAGE, and PREDICTION-ADR data were performed in SAS 9.3 (SAS Institute, Cary, NC, USA). Statistical analyses in JUPITER were performed using R.25 Binary logistic regression was used to test the association between the variant and each phenotype of intolerance. Covariates associated with intolerance such as gender, age, co-medication usage, type of statin, dose of statin, and CK levels were added to models where appropriate and available. A backwards step-wise approach was used to eliminate covariates that were not significant predictors in adjusted models. Finally, a fixed-effects meta-analysis on results from the discovery and replication cohorts was performed. Only one phenotype from GoDARTS could be selected since the two groups contained overlapping individuals. LDI phenotype was selected as the phenotype definition did not include the CK levels and the model was adjusted for CK measures. The analysis was performed using the metafor package in R26 and results are presented in a Forest plot (Figure 1).

TS could be selected since the two groups contained overlapping individuals. LDI phenotype was selected as the phenotype definition did not include the CK levels and the model was adjusted for CK measures. The analysis was performed using the metafor package in R26 and results are presented in a Forest plot (Figure 1). Figure 1 Forest plot representing meta-analysis of the association between LILRB5 Asp247Gly and outcomes observed across GoDARTS, CPRD-STAGE, PREDICTION-ADR, and JUPITER studies. Study sample size is in parentheses. LDI, low-dose intolerance; SIM, statin-induced myopathy. The minor allele frequency in GoDARTS, CPRD, PREDICTION-ADR, and JUPITER study populations were 0.37, 0.39, 0.37, and 0.40, respectively. The effect of the LILRB5 variant was considered dominant based on large-scale analyses with serum CK in GoDARTS (see Supplementary material online, Results S1). Therefore all association tests compared those homozygous for Asp247 (T/T) with carriers of 247GlyX (T/C or C/C).

ere 0.37, 0.39, 0.37, and 0.40, respectively. The effect of the LILRB5 variant was considered dominant based on large-scale analyses with serum CK in GoDARTS (see Supplementary material online, Results S1). Therefore all association tests compared those homozygous for Asp247 (T/T) with carriers of 247GlyX (T/C or C/C). Results Baseline characteristics of general statin intolerance and low dose intolerance Covariates associated with statin use or with the development of adverse drug reactions (ADR) were tested. Specifically, mean age at start and end of statin therapy, sex, diagnosis of T2D, first and last statin used, starting and ending doses, use of interacting co-medications, statin use for the secondary prevention of CVD, CK levels, and LDL levels prior to statin use. The Comparison of GSI with statin tolerance (ST1) and of LDI with statin tolerance (ST2) is presented (Table 1). Table 1 Contrasting general statin intolerance with raised CK and statin tolerance (ST1) and low dose intolerance with statin tolerance (ST2)

prevention of CVD, CK levels, and LDL levels prior to statin use. The Comparison of GSI with statin tolerance (ST1) and of LDI with statin tolerance (ST2) is presented (Table 1). Table 1 Contrasting general statin intolerance with raised CK and statin tolerance (ST1) and low dose intolerance with statin tolerance (ST2) Variables GSI P-value LDI P-value Cases (n = 588) Controls (n = 356) Cases (n = 591) Controls (n = 443) Mean age start therapy (SD) 60 (10) 62 (10) 0.005 60 (10) 60 (10) 0.9 Years on statin therapy (SD) 10.4 (5) 9.3 (3) <0.0001 10 (5) 9.5 (3) 0.007 Sex (% females) 50 43 0.007 48 46 0.16 Type 2 diabetics (%) 78 79 0.68 92 90 0.33 First statin as simvastatin (%) 64 71 <0.0001 59 65 <0.0001 Last statin as simvastatin (%) 31 41 <0.0001 31 36 <0.0001 Starting dose as ‘low’ (<20 mg/day) (%) 85 53 <0.0001 94 37 <0.0001 Ending dose as ‘high’ (≥80 mg/day) (%) 22 36 <0.0001 23 50 <0.0001 Interacting co-medications (yes %) 52 44 0.0025 51% 42% <0.0001 Statin use for secondary prevention of CVD (%) 27 23 0.18 28% 25% 0.3 CK levels (IU/L)a Median 200 76 <0.0001 98 85 <0.0001 Mean (minimum, maximum) 306 (120, 12, 700) 81 (17, 179) 170 (13, 12 735) 107 (19, 1369) LDL levels (mmol/L)a Median 3.5 3.2 0.07 3.2 3.1 0.38 Mean (minimum, maximum) (3.5) (1.1, 5.5) 3.2 (0.4, 8.7) 3.1 (1.1, 6.4) 3.2 (0.5, 8.7) a Indicates associations were tested using log 10 transformed values. SD, standard deviation; GSI, general statin intolerance; LDI, Low dose intolerance.