Chromosome 9p21 Variant Predicts Mortality After Coronary Artery Bypass Graft Surgery
Background—Recent genome-wide association studies have identified several chromosome 9p21 single nucleotide polymorphisms associated with coronary artery disease and myocardial infarction in nonsurgical populations. We have recently demonstrated an independent association between these 9p21 variants and perioperative myocardial injury after isolated primary coronary artery bypass graft (CABG) surgery. This study investigated the association of a 9p21 variant with mortality in patients after CABG surgery and its prognostic value to improve the EuroSCORE.
Methods and Results—In a 2-center, prospective, observational study of 846 white primary CABG surgery patients, we genotyped rs10116277, the 9p21 variant with the strongest association to perioperative myocardial injury in our cohort. To estimate the utility of rs10116277 for predicting all-cause mortality within 5 years after surgery, a Cox proportional hazard model was constructed to estimate the hazard ratios (HR) and 95% confidence intervals (CI) while adjusting for demographics and clinical covariates.
The homozygote minor allele of rs10116277 was associated with significantly increased risk of all-cause mortality even after adjusting for other clinical predictors of mortality in a Cox proportional hazards model (HR, 1.7; 95% CI, 1.1–2.7; P=0.026). Addition of rs10116277 to the logistic EuroSCORE also significantly improved model prediction for mortality (HR, 1.82; 95% CI, 1.15–2.88; P=0.01).
Conclusion—The 9p21 variant rs10116277 is independently associated with all-cause mortality after primary CABG surgery in whites and significantly improves the predictive value of the logistic EuroSCORE.
Clinical Trial Registration—URL: http://www.clinicaltrials.gov. Unique identifier: NCT00281164.
Coronary artery disease (CAD) and myocardial infarction have been shown to be heritable diseases in ambulatory population based-cohorts.1,2 Recently, genome-wide association studies have identified variants in a chromosome 9p21 locus associated with cardiovascular disease outcomes including CAD and myocardial infarction.3–5 These 9p21 variants are located within a newly annotated noncoding RNA gene called ANRIL (antisense noncoding RNA in the INK4 locus), expressed in cell types integral to atherosclerosis,6,7 and are also adjacent to genes for the cyclin-dependent kinases CDKN2A and CDKN2B, which play a critical role in regulating cell aging, cell proliferation, and apoptosis.3,8 Nonetheless, the biological mechanisms responsible for this association remain to be elucidated.
We recently identified an association between the same variants in the 9p21 locus and perioperative myocardial injury (PMI) after isolated primary coronary artery bypass graft (CABG) surgery with cardiopulmonary bypass (CPB), independent of coronary disease severity.9 This association is surprising given the perceived different etiologies of myocardial injury between ambulatory myocardial infarction and PMI resulting from aortic occlusion, cardiotomy, and an obligatory acute inflammatory response associated with CPB.10,11 Yet, the commonality of genetic association implies similar biological mechanisms for both disease processes, irrespective of etiologies.
Several studies have demonstrated associations between the degree of PMI after CABG surgery, measured by cardiac biomarker or ECG evidence of myocardial injury, and mortality.12–16 However, no study to date has related the 9p21 variants associated with cardiovascular disease to mortality after CABG surgery. Therefore, we hypothesized that a specific 9p21 variant is also associated with an increased incidence of mortality in patients undergoing CABG surgery with CPB. We further hypothesized that this association would provide additional predictive value to the logistic EuroSCORE that is commonly used for predicting postoperative mortality after cardiac surgery.
Materials and Methods
Two institutions (Brigham and Women’s Hospital and Texas Heart Institute) recruited patients within a single study structure known as the CABG Genomics Program (http://clinicaltrials.gov/show/NCT00281164). Since August 2001, we have prospectively enrolled patients aged 20 to 90 years undergoing nonemergent primary CABG surgery utilizing CPB without other concurrent surgery. Patients with a preoperative hematocrit <25% or transfusion of leukocyte-rich blood products within 30 days before surgery were not enrolled. To avoid potential population stratification, analysis was restricted to subjects who self-reported 4 generations of white grandparental ancestry. Study protocols were approved by respective Institutional Review Boards, and participants were enrolled after informed written consent. At each site, patient demographics, perioperative risk factors, medications, and postoperative outcomes using study-specific case report forms were recorded. Mortality was assessed by accessing hospital records and the Social Security Death Index (http://ssdi.rootsweb.ancestry.com/). Cardiovascular mortality was ascertained using the National Death Index service of the National Center for Health Statistics. Death status is queried at 5 years of follow-up or earlier in patients who are lost to follow-up at the time this analysis was performed.
DNA was extracted from white blood cells using standard protocols. We genotyped rs10116277, the 9p21 variant with the highest association to PMI in our cohort,9 using the Golden Gate assay with an Illumina Bead Station 500G system (Illumina) in accordance with the manufacturer’s standard recommendations. The genotyping call rate was >98% and genotypes were in Hardy-Weinberg equilibrium. We addressed population stratification by only including patients in the analysis who self-reported 4 generations of white grandparental ancestry. In addition, we tested genomic controls and tested for differences between Northern and Southern Europeans.
Blood samples were drawn from all patients at all 7 perioperative time points: before induction of anesthesia, after administration of protamine (≈20 minutes after aortic cross-clamp release), and on the mornings of postoperative days 1 to 5. Citrated plasma was stored in vapor-phase liquid nitrogen at −130°C until analysis. Cardiac troponin I, creatine kinase MB, and B-type natriuretic peptide were analyzed with a sandwich immunoassay on a Triage platform (Biosite) using monoclonal and polyclonal antibodies at a single core facility. All samples were analyzed at the conclusion of the study and the patient care providers had no access to the results. The root mean square coefficient of variation over the reportable range for B-type natriuretic peptide, creatine kinase MB, and cardiac troponin I assays were 11.3%, 14.3%, and 13%, respectively.
First, a simple univariate survival analysis using Cox proportional hazard model was performed to test for association between rs10116277 and mortality based on a 2-degrees-of-freedom genotypic model. No significant difference was observed between carriers with 0 (null) and 1 copy (heterozygote) of the minor allele (G), whereas carriers with 2 copies of the G allele showed significantly increased risk of death. Because there was essentially no difference for mortality between the nulls and heterozygotes of the rs10116277 G allele, only the recessive model, ie, 2 copies vs 0 or 1 copy, was considered for all further analyses.
Cumulative mortality was estimated by Kaplan–Meier analysis and the effect of each genotype was compared with a Wilcoxon rank test. A multivariable logistic model was used to derive perioperative and demographic variables associated with 5-year all-cause mortality. Patients who are still alive at the end of 5-year follow-up period are censored; patients who are still alive when they are lost to follow-up are censored as well. Factors previously associated with mortality after cardiac surgery in previous studies were forced into the logistic regression model along with clinically relevant variables using stepwise selection. Age was nonlinear and partitioned into quintiles. To estimate the independent prognostic utility of rs10116277 on 5-year all-cause mortality, a Cox proportional hazard model was constructed to estimate the hazard ratios (HR) and 95% confidence intervals (CI) while adjusting for demographics and clinical covariates identified by logistic regression.
To demonstrate the model improvement of adding rs10116277 to the prediction of mortality, we computed the c-index for the censored data. The c-index is a generalization of the area under the receiver-operator characteristic curve that represents the probability of concordance between predicted and observed survival, with c=0.5 for random prediction and c=1 for a perfectly discriminating model. We then computed the U statistic to test whether the model with the genetic variant was more concordant than the model without the variant. Additionally, we used 4 global measures of model fit (the likelihood ratio test [−2LL], Akaike information criterion,17 Net Reclassification Improvement [NRI],18 and Integrated Discrimination Improvement18) to assess the value of adding rs10116277 to the clinical model of all-cause mortality.
A bootstrapping procedure was applied to estimate the covariate adjusted HR of homozygous minor allele carriers vs the other patients (ie, heterozygous or homozygous carriers of major allele) and to mitigate the potential for overfitting.19,20 To assess uncertainty about the estimated HR, the studentized bootstrap CI was calculated. We further hypothesized that these associations would provide additional predictive value, over and above traditional clinical scores commonly used for predicting postoperative mortality after CABG surgery. To establish the value of adding rs10116277 to an established risk prediction model, an individual risk score was calculated for each patient using the EuroSCORE logistic model.21 To estimate the independent prognostic utility of rs10116277 on EuroSCORE-estimated all-cause mortality, a Cox proportional hazard model was constructed with and without the rs10116277 variant.
Statistical and genetic association analyses were performed using SAS (version 9.1.3; SAS Institute) and JMP 8.0 (SAS Institute). Data are presented as medians with interquartile ranges (interquartile range, 25%–75%) unless otherwise stated. Continuous variables were compared using the analysis of variance and the Wilcoxon-Mann-Whitney rank-sum test. Categorical variables were compared with the χ2 test or Fisher exact test, as appropriate. P<0.05 was considered significant.
Patient characteristics stratified by mortality after primary CABG surgery are shown in Table 1. Ninety-seven patients (13%) died within 5 years (median follow-up, 4.96 years; range, 3.9–5.0 years), including 6 patients who died within 30 days of surgery. Thirty-eight patients died of cardiovascular causes. Patients who were older, had lower preoperative left ventricular ejection fraction, a history of pulmonary disease or hypercholesterolemia, higher serum creatinine, lower hematocrit, or who were not using aspirin or HMG-CoA reductase inhibitors were more likely to die during the follow-up period. Similarly, elevated preoperative B-type natriuretic peptide and creatine kinase MB were associated with a higher risk of death. Neither preoperative cardiac troponin I nor cardiac troponin I on postoperative day 1 adjusted the association between rs10116277 and all-cause mortality.
Neither the degree of coronary artery stenosis by angiography nor the number of vessels grafted intraoperatively was significantly different between patients who died and those who survived. Patients who subsequently died within 5 years were hospitalized after CABG surgery for a mean of 3 additional days and spent a mean of 14 more hours in the intensive care unit.
In the Cox proportional hazard model of mortality, age, left ventricular ejection fraction, history of pulmonary disease, and rs10116277 were independently associated with increased mortality (Table 2). Age predicted mortality in a nonlinear yet progressive fashion, with the upper quintile demonstrating increasing significance with the risk of death (age ≥75 years; HR, 20.5; 95% CI, 4.8–88.0; P<0.0001). Patients with 2 copies of the minor allele of rs10116277, compared to patients with either the heterozygote or the major allele, had a significantly greater risk of dying (HR, 2.0; 95% CI, 1.2–3.5; P=0.003; Figure Kaplan–Meier curve).
Addition of rs10116277 using a recessive genetic model to the Cox proportional hazard model for mortality significantly improved overall model prediction (P<0.0001). Possessing 2 copies of the minor allele of rs10116277 significantly increased mortality when assessed using a Cox proportional hazard model (HR, 1.7; 95% CI, 1.1–2.7; P=0.026) using the 2 global measures of model fit: −2LL and Akaike information criterion (Table 3). Adding rs10116277 to the Cox proportional hazard model has a marginally significant yet fairly substantial reclassification improvement (≈20% net gain as assessed by National Death Index), whereas Integrated Discrimination Improvement suggests a modest and marginally significant improvement for average sensitivity and specificity (<3% gain; Table 4). Bootstrapping showed a similar significant result with a very tight 95% CI (HR, 1.65; 95% CI, 1.60–1.84; P=0.031).
The Cox proportional hazard model for cardiovascular mortality was underpowered to show significance for predicting the outcome. Nevertheless, the HR showed similar directionality to all-cause mortality (HR, 1.41; 95% CI, 0.65–3.07; P=0.39; EuroSCORE HR, 1.55; 95% CI, 0.74–3.3; P=0.25).
The preoperative logistic EuroSCORE was significantly higher in those patients who died during follow-up (median, 4.8%; interquartile range, 2.4–8.3) compared to those who survived (median, 2.1%; interquartile range, 1.2–3.8; P<0.001; Table 1). The addition of rs10116277 using a recessive genetic model to the logistic EuroSCORE significantly improved model prediction for mortality (HR, 1.82; 95% CI, 1.15–2.88; P=0.01) as assessed by the improvement in the 4 global measures of model fit: −2LL, Akaike information criterion, NRI, and Integrated Discrimination Improvement (Tables 3, 4⇑).
In this investigation, we demonstrate that white patients carrying 2 copies of the minor allele of a common 9p21 variant rs10116277 have a significantly greater risk of dying within 5 years of CABG surgery. In addition, this variant contributes additional predictive value to the well-validated EuroSCORE prediction of postoperative mortality. In our population, 18% of patients are homozygous for rs10116277, which is similar to the 22% in the white HapMap cohort, making this a common disease-linked genotype with a novel association to a long-term outcome.
The rs10116277 genetic variant and SNP in strong linkage disequilibrium have been associated with coronary artery calcification, angiographic CAD, and myocardial infarction.3,22,23 We recently have shown an association between the 9p21 variant and PMI, independent of CAD severity.9 In addition, we have shown that a direct link between the extent of PMI and mortality exists.16 We now show an association between the 9p21 variant and all-cause mortality with a trend for cardiovascular mortality, independent of extent of myocardial injury, suggesting that this variant may have a direct role in postoperative morbidity.
Mechanisms of Association
The mechanism of the association between 9p21 variants and mortality is unexplained. A recent study found that probands with the 9p21 risk allele for cardiovascular disease had decreased expression of the CDKN2A/2B locus in peripheral blood leukocytes.24 The CDKN2A and CDKN2B play a critical role in regulating cell aging, cell proliferation, senescence, and apoptosis. CDKN2A overexpression in mice leads to increased tumor suppression, whereas inactivation of CDKN2B or CDKN2A allows cells to escape cell cycle arrest or become senescent.8 CDKN2A expression correlates with histopathologic changes in a hypertensive rat myocardium model and therefore could be the result of widespread irreversible cell cycle arrest with accumulation of senescent cells.25
Cellular senescence is classified as an irreversible arrest of cell proliferation with maintenance of cell function, as opposed to apoptosis, which is defined as controlled autodigestion of the cell leading to the formation of apoptotic bodies within the intact cell plasma membrane.26 Senescence and apoptosis are 2 distinct pathways. Senescent cells are resistant to programmed cell death which leads to accumulation of these cells in injured tissue, and may have substantial effects on healthy neighboring tissue, with long-term consequences for the aging process and disease states.26
Similarly, ANRIL, a noncoding RNA gene, spans the entire 9p21.3 region and is highly expressed in human myocardium.6 RNA plays a role at all levels of gene regulation, including the control of chromosome dynamics, splicing, translational inhibition, and mRNA destruction. The ANRIL gene does not share any nucleotide sequences with CDKN2B, but it overlaps the 2 CDKN2B exons and possibly shares a bidirectional promoter.6 Therefore, it is plausible that ANRIL coordinates CDKN expression6,24 or exerts effects on translocated genes27 also contributing to cellular senescence with subsequent increased mortality.
In our population, ischemia-reperfusion injury is an obligate component of surgery from aortic cross-clamping and cardioplegia administration. Given ANRIL’s apparent relation to CDKN2B, one could hypothesize that the explanation for our findings of significantly earlier mortality in patients carrying the 9p21 risk allele is that the timed insult of surgery with CPB provides a stimulus that unmasks and accelerates the effects of the 9p21 variant. Patients with the risk allele are more likely to accumulate senescent cells or even show evidence of necrotic cell death with rapid cellular edema and lysis. This could also explain the higher levels of cardiac troponin I in the risk allele group. Therefore, patients with the 9p21 risk allele may be more likely to die earlier.
Although genetic variation in the 9p21 locus is associated with incident cardiovascular disease across many different populations and has a high frequency of the risk allele, the utility of screening for this polymorphism to improve risk prediction remains unclear. We observed significant incremental improvement by adding 9p21 variants into a clinical predictive model of perioperative mortality. Similarly, in the white population of the Atherosclerosis Risk in Communities study (n=10 004), adding 9p21 variation to traditional risk factors improved CAD risk prediction and reclassification, particularly in higher-risk categories.28 Conversely, in the Women’s Genome Health Study (n=22 129), genetic variation in 9p21 did not improve discrimination or classification of predicted risk.29 Therefore, further investigation is warranted before the genetic variation in 9p21 can be routinely incorporated into perioperative risk prediction models.
Consideration for certain limitations of this investigation is warranted. We have not yet replicated our findings in a separate cardiac surgical cohort. However, the 9p21 region is the most replicated locus for CAD and myocardial infarction to date across multiple populations. In addition, we emphasize that this study only examined whites and should not be extrapolated to other racial groups, for whom the frequency of the minor allele of rs1011627 is lower.
We observed significance of the association between rs10116277 only for a recessive genetic model. Other studies have demonstrated significance for the additive genetic model, which we did not observe. We are likely underpowered to observe a difference in effect between 0 and 1 copy.
Use of HMG-CoA reductase inhibitors, aspirin, and beta blockers has been associated with reduced mortality after cardiac surgery.30 We do not have follow-up data on medication use and medication compliance of patients after discharge from the hospital.
We also observed a significant improvement in model prediction using the conservative methods likelihood ratio test and Akaike information criterion, which penalize for the addition of multiple variables; however, the improvement in receiver-operator characteristic curve was not significant. This highlights the deficiencies of receiver-operator characteristic curves, which may be commonly used but are not very sensitive or powered to see a difference in effect. Consequently, in our data even a large change in probability value (P change from 10−32 to 10−35) was not sensitive enough to show a significant difference.
The 9p21 variant rs10116277, which is within or adjacent to genes known to be involved in cell-cycle regulation, cell senescence, and apoptosis, is significantly and independently associated with all-cause mortality after primary CABG surgery and increases the predictive value of the well-validated EuroSCORE.
In our population, ischemia-reperfusion injury is an obligate component of surgery from aortic cross-clamping, cardioplegia administration, and CPB. Ischemia-reperfusion injury may unmask and accelerate the effects of the 9p21 variant. Thus, hypothetically, patients with the risk allele may be more susceptible to senescent cell accumulation or even necrotic cell death with rapid cellular edema and lysis. Further investigation is warranted to elucidate this mechanism.
The authors acknowledge the outstanding contributory efforts of the CABG Genomics research staff: James Gosnell, RN; Kujtim Bodinaku, MD; Jai Madan, MD, MPH; Svetlana Gorbatov, MPH; James Chen, RN; and Isabella Candelaria. The authors thank all study subjects who participated in the CABG Genomics Program and surgeons who identified their patients.
Sources of Funding
Supported by a Bayer Fellowship in Blood Conservation grant, Wyckoff, NJ (J.D.M.); by Biosite Inc, San Diego, CA; by a Society of Cardiovascular Anesthesiologists research starter grant, Richmond, VA (J.D.M.); by a grant from NIH NHLBI HL-068774, Bethesda, MD (S.C.B.); and by a Mercedes Concepcion Faculty Development Fellowship and Scholar in Translational Anesthesia research grant from the Department of Anesthesiology, Perioperative, and Pain Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA (J.D.M.).
S.K.S. and S.C.B. contributed equally to this work.
Presented at the 2009 American Heart Association meeting in Orlando, Fla, November 14–18, 2009.
Arnett DK, Baird AE, Barkley RA, Basson CT, Boerwinkle E, Ganesh SK, Herrington DM, Hong Y, Jaquish C, McDermott DA, O'Donnell CJ. Relevance of genetics and genomics for prevention and treatment of cardiovascular disease: a scientific statement from the American Heart Association Council on Epidemiology and Prevention, the Stroke Council, and the Functional Genomics and Translational Biology Interdisciplinary Working Group. Circulation. 2007; 115: 2878–2901.
McPherson R, Pertsemlidis A, Kavaslar N, Stewart A, Roberts R, Cox DR, Hinds DA, Pennacchio LA, Tybjaerg-Hansen A, Folsom AR, Boerwinkle E, Hobbs HH, Cohen JC. A common allele on chromosome 9 associated with coronary heart disease. Science. 2007; 316: 1488–1491.
Pasmant E, Laurendeau I, Heron D, Vidaud M, Vidaud D, Bieche I. Characterization of a germ-line deletion, including the entire INK4/ARF locus, in a melanoma-neural system tumor family: identification of ANRIL, an antisense noncoding RNA whose expression coclusters with ARF. Cancer Res. 2007; 67: 3963–3969.
Broadbent HM, Peden JF, Lorkowski S, Goel A, Ongen H, Green F, Clarke R, Collins R, Franzosi MG, Tognoni G, Seedorf U, Rust S, Eriksson P, Hamsten A, Farrall M. Susceptibility to coronary artery disease and diabetes is encoded by distinct, tightly linked SNPs in the ANRIL locus on chromosome 9p. Hum Mol Genet. 2008; 17: 806–814.
Priebe HJ. Perioperative myocardial infarction–aetiology and prevention. Br J Anaesth. 2005; 95: 3–19.
Braunwald E, Antman EM, Beasley JW, Califf RM, Cheitlin MD, Hochman JS, Jones RH, Kereiakes D, Kupersmith J, Levin TN, Pepine CJ, Schaeffer JW, Smith EE III, Steward DE, Theroux P. ACC/AHA guidelines for the management of patients with unstable angina and non-ST-segment elevation myocardial infarction. A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee on the Management of Patients With Unstable Angina). J Am Coll Cardiol. 2000; 36: 970–1062.
Croal BL, Hillis GS, Gibson PH, Fazal MT, El-Shafei H, Gibson G, Jeffrey RR, Buchan KG, West D, Cuthbertson BH. Relationship between postoperative cardiac troponin I levels and outcome of cardiac surgery. Circulation. 2006; 114: 1468–1475.
Chen JC, Kaul P, Levy JH, Haverich A, Menasche P, Smith PK, Carrier M, Verrier ED, Van de Werf F, Burge R, Finnegan P, Mark DB, Shernan SK. Myocardial infarction following coronary artery bypass graft surgery increases healthcare resource utilization. Crit Care Med. 2007; 35: 1296–1301.
Muehlschlegel JD, Perry TE, Liu K-Y, Nascimben L, Fox AA, Collard CD, Avery EG, Aranki SF, D'Ambra MN, Shernan SK, Body SC, for the CGI. Troponin is superior to electrocardiogram and creatinine kinase MB for predicting clinically significant myocardial injury after coronary artery bypass grafting. Eur Heart J. 2009; 30: 1574–1583.
Canty A, Ripley B. Boot: Bootstrap R (S-Plus) Functions. Version 1.2–36; 2009. Available at: http://CRAN.R-project.org/package=boot.
Davison AC, Hinkley DV. Bootstrap Methods and their Application. New York, NY: Cambridge University Press; 1997.
Roques F, Michel P, Goldstone AR, Nashef SA. The logistic EuroSCORE. Eur Heart J. 2003; 24: 881–882.
Kathiresan S, Voight BF, Purcell S, Musunuru K, Ardissino D, Mannucci PM, Anand S, Engert JC, Samani NJ, Schunkert H, Erdmann J, Reilly MP, Rader DJ, Morgan T, Spertus JA. Genome-wide association of early-onset myocardial infarction with single nucleotide polymorphisms and copy number variants. Nat Genet. 2009; 41: 334–341.
Liu Y, Sanoff HK, Cho H, Burd CE, Torrice C, Mohlke KL, Ibrahim JG, Thomas NE, Sharpless NE. INK4/ARF transcript expression is associated with chromosome 9p21 variants linked to atherosclerosis. PLoS ONE. 2009; 4: e5027.
Westhoff JH, Hilgers KF, Steinbach MP, Hartner A, Klanke B, Amann K, Melk A. Hypertension induces somatic cellular senescence in rats and humans by induction of cell cycle inhibitor p16INK4a. Hypertension. 2008; 52: 123–129.
Brautbar A, Ballantyne CM, Lawson K, Nambi V, Chambless L, Folsom AR, Willerson JT, Boerwinkle E. Impact of adding a single allele in the 9p21 locus to traditional risk factors on reclassification of coronary heart disease risk and implications for lipid-modifying therapy in the Atherosclerosis Risk in Communities study. Circ Cardiovasc Genet. 2009; 2: 279–285.