SNP indicates single-nucleotide polymorphism. A, Unadjusted odds ratios of CAD according to HbA1c value at study entry (top tertile vs lower 2 tertiles) and genotypes at rs2383206. Individuals with no risk alleles and HbA1c in the lower 2 tertiles serve as a reference. B, Unadjusted odds ratios of CAD according to the time-weighted average HbA1c during the 7 years before study entry (top tertile vs lower 2 tertiles) and genotypes at rs2383206. The top tertile boundaries were 7.6 for HbA1c at examination and 7.9 for the average HbA1c in the years before examination.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Doria A, Wojcik J, Xu R, et al. Interaction Between Poor Glycemic Control and 9p21 Locus on Risk of Coronary Artery Disease in Type 2 Diabetes. JAMA. 2008;300(20):2389–2397. doi:10.1001/jama.2008.649
Author Affiliations: Research Division, Joslin Diabetes Center (Drs Doria, Wojcik, Xu, and Warram, and Mr Nolan), Department of Medicine, Harvard Medical School (Drs Doria, Wojcik, Xu, Gervino, Hauser, and Hu); Cardiovascular Division, Beth Israel Deaconess Medical Center (Drs Gervino and Hauser); Cardiovascular Medicine Division, Caritas St Elizabeth's Medical Center (Dr Johnstone); Departments of Nutrition (Dr Hu), and Epidemiology, Harvard School of Public Health (Drs Doria and Hu), Boston, Massachusetts.
Context A common allele on chromosome 9p21 has been repeatedly associated with increased risk of coronary artery disease (CAD) in the general population. However, the magnitude of this effect in the population with diabetes has not been well characterized.
Objective To examine the association of the 9p21 variant with CAD in individuals with type 2 diabetes and evaluate its interaction with poor glycemic control.
Design, Setting, and Participants (1) Case-control study of 734 type 2 diabetes patients (322 with angiographically diagnosed CAD and 412 with no evidence of CAD) who were recruited between 2001 and 2006 at the Joslin Clinic, Beth Israel Deaconess Medical Center; and (2) independent cohort study of 475 type 2 diabetes patients from the Joslin Clinic whose survival status was monitored from their recruitment between 1993 and 1996 until December 31, 2004. Participants for both studies were genotyped for a representative single-nucleotide polymorphism at 9p21 (rs2383206) and characterized for their long-term glycemic control by averaging multiple hemoglobin A1c (HbA1c) measurements taken in the years before study entry.
Main Outcome Measures For the case-control study, association between single-nucleotide polymorphism rs2383206 and CAD defined as angiographically documented stenosis greater than 50% in a major coronary artery or a main branch thereof was assessed and for the cohort study, cumulative 10-year mortality was documented.
Results Individuals who were homozygous for the risk allele were significantly more frequent among case than control participants (42.3% vs 28.9P = .0002). This association was unaffected by adjustment for cardiovascular risk factors, but the effect of the risk genotype was significantly magnified (adjusted P for interaction = .048) in the presence of poor glycemic control (worst tertile of the distribution of HbA1c at examination). Relative to the CAD risk for patients with neither a 9p21 risk allele nor poor glycemic control, the CAD odds for participants having 2 risk alleles but not poor glycemic control were increased 2-fold (odds ratio [OR], 1.99; 95% confidence interval [CI], 1.17-3.41), whereas the odds for study participants with the same genotype and with poor glycemic control were increased 4-fold (OR, 4.27; 95% CI, 2.26-8.01). The interaction was stronger (adjusted P = .005) when a measure of long-term glycemic control (7-year average rather than most recent HbA1c) was used with ORs of 7.83 (95% CI, 3.49-17.6) for participants having 2 risk alleles and a history of poor glycemia and 1.54 (95% CI, 0.72-3.30) for participants with the same genotype but without this exposure. A similar interaction between 9p21 variant and poor glycemic control was observed with respect to cumulative 10-year mortality in the cohort study (43.6% in patients with 2 risk alleles and poor glycemic control, 23.1% in individuals with only the 2 risk alleles, 30.0% in individuals with only poor glycemic control, and 31.6% in individuals with neither factor, P for interaction, = .036).
Conclusion In this study population, the CAD risk associated with the 9p21 variant was increased in the presence of poor glycemic control in type 2 diabetes.
Among the known risk factors for cardiovascular disease, diabetes mellitus ranks as one of the most potent. It increases the lifetime risk of a major cardiac event by a factor of 2 to 4 relative to individuals without diabetes1 and compounds the effect of such events with increased risks of repeat myocardial infarction, congestive heart failure, and death.2 It is also a major determinant of peripheral artery disease and stroke.3,4 These profound effects result from an acceleration of atherosclerosis induced by hyperglycemia and other aspects of the diabetic milieu, such as hyperlipidemia (particularly small low-density lipoprotein cholesterol particles) and hypertension.5
Based on family studies, a substantial proportion of cardiovascular risk is under the control of genetic factors.6,7 While this evidence has been gathered mainly in the general population, studies of indicators of preclinical atherosclerosis suggest that genetic factors also play a role in the development of atherosclerosis in the presence of diabetes.8,9 One may suppose that the genes that play a role in the absence of diabetes are likely to play a role also in its presence. If so, it invites the question of whether the increased cardiovascular disease risk observed in diabetes is mediated by interaction of those same genes with the diabetic milieu.
Four recent genome-wide association studies found association of a common allele on chromosome 9p21 with coronary artery disease (CAD) in the general population.10-13 In this study, we examined the association of this allele with coronary artery disease in individuals with type 2 diabetes and whether the association is modified by the severity of hyperglycemia, the defining characteristic of diabetes.
Case-Control Study. We studied a group of 734 individuals with type 2 diabetes living in greater Boston, Massachusetts and receiving treatment at the Joslin Clinic, the Beth Israel Deaconess Medical Center, or both facilities. All participants were self-reported as non-Hispanic white and all provided written informed consent. Study protocol and informed consent procedures were approved by the Joslin Committee on Human Studies and the Beth Israel Deaconess Medical Center Committee on Clinical Investigations.
Type 2 diabetes was defined as diabetes that was diagnosed at age 30 years or older according to American Diabetes Association criteria14 and did not require insulin treatment for at least 2 years after its diagnosis. Study participants included 322 cases with diagnosed CAD and 412 control participants who did not have clinical evidence of CAD. Case participants with CAD were a random sample of patients with type 2 diabetes who had a stenosis greater than 50% in a major coronary artery or a main branch thereof that was documented by cardiac catheterization at the Beth Israel Deaconess Medical Center between 2001 and 2006. All eligible participants were enrolled in the study at the time of catheterization and examined within 1 month following the procedure. Sixty-two percent of the case patients received diabetes management care at the Joslin Clinic.
Control participants without CAD were randomly selected from among 903 Joslin patients who were identified between 2001 and 2006 as fulfilling the following criteria: (1) current age between 55 and 74 years; (2) type 2 diabetes for 5 years or more; (3) negative cardiovascular history (ie, normal resting electrocardiogram, absence of cardiac symptoms, and no hospitalization for cardiovascular events); and (4) normal response to an exercise treadmill test15 performed for screening purposes. All control participants were recruited within 6 months following the exercise treadmill test. History of myocardial infarction, smoking, hypertension, and hypercholesterolemia and treatment with glucose-lowering drugs were determined by a questionnaire administered at the time of examination. Data regarding medications were confirmed by review of medical records.
Prospective Study. The interaction between 9p21 variant and poor glycemic control with respect to all-cause and cardiovascular mortality was evaluated in a group of 516 Joslin patients who had been recruited between 1993 and 1996 for genetic studies of type 2 diabetes and its complications. The present study was limited to the 475 members of the cohort for whom DNA samples were still available in 2007. Eleven of these participants overlapped with individuals in the case-control study. The study protocol and informed consent procedures were approved by the Joslin Committee on Human Studies. All participants were self-reported non-Hispanic white and provided written informed consent.
Study participants were a random sample of type 2 diabetes patients from the Joslin Clinic enriched with individuals with proteinuria.16 All participants had diabetes diagnosed after age 25 years according to World Health Organization criteria and were treated with diet or oral agents for at least 2 years following diagnosis. Their survival status was updated as of December 31, 2004, by matching with the National Death Index, and causes of death were extracted for deceased cohort members. A death was ascribed to cardiovascular causes if the primary cause of death was International Classification of Diseases, Ninth Revision code 401 to 448.9 or International Classification of Diseases, Tenth Revision code I10 to I74.9, or if diabetes or renal failure were listed as the primary cause of death and cardiovascular disease was the secondary cause.
For the case-control study, HbA1c was measured at examination on a random blood sample by high-performance (pressure) liquid chromatography (Tosoh Bioscience, South San Francisco, California) by the Joslin Clinical Laboratory. The intra- and interassay coefficients of variation for this measurement were 0.25% and 2.1%, respectively. Additional HbA1c values measured between 2 months to 7 years before study entry were abstracted from the Joslin electronic medical records. For the prospective study, all HbA1 and HbA1c values measured in these participants between 1990 and study entry were abstracted from the Joslin electronic medical records. HbA1 values were converted to HbA1c values as previously described.17
All study participants were typed for SNP rs2383206 and rs10757278 by the Joslin DERC Genetics Core by means of TaqMan assays implemented on an ABI PRISM 7700 HT Sequence Detection System (Applied Biosystems, Foster City, California). Genotyping quality was tested by including 6 blinded duplicate samples in each 96-well assay. The average agreement rate of duplicate samples was greater than 99%.
Case-Control Study. All statistical analyses were conducted using SAS statistical software version 9.1 (SAS Institute Inc, Cary, North Carolina). Genotype distributions were tested at both polymorphic loci for departure from Hardy-Weinberg equilibrium and compared between study groups by Fisher exact tests. Allele frequencies were derived from genotype counts and compared between groups also by Fisher exact tests. Odds ratios (ORs) of CAD for SNP rs2383206 and other relevant predictors were estimated by logistic regression analysis using first a univariable model for each predictor and then a multivariable model including all variables showing a significant effect (P < .05). For examining the interaction of genotype with hyperglycemia, rs2383206 was represented as an additive model (number of risk alleles), HbA1c as an indicator variable for the highest tertile, and the interaction as the product of the HBA1c variable by an indicator variable for patients homozygous for the risk allele.
Power was estimated by means of the software QUANTO (http://hydra.usc.edu/GxE), assuming a risk allele frequency of 0.55. For both SNPs, there was 80% power (α = .05) to detect associations with CAD with ORs as low as 1.35 per risk allele and to detect a 2.5-fold difference in the ORs for risk allele homozygotes between the top and lower 2 tertiles of HbA1c. The latter calculation assumed a CAD prevalence of 0.25 and marginal ORs of 1.80 for G/G vs A/G+A/A and 1.45 for the top tertile vs the lower 2 tertiles of HbA1c.
Prospective Study. Life-table methods were used to estimate the cumulative 10-year mortality and its standard error within each stratum defined by degree of glycemic control and rs2383206 genotype. The significance of the interaction between these 2 variables was determined by comparing the effect of glycemic control on cumulative mortality in G/G homozygotes with that in carriers of other genotypes. This linear contrast, divided by its standard error, was compared with the standard normal distribution. The presence of interaction was also tested by adding a cross-product term to a Cox proportional hazard model including glycemic control and rs2383206 genotype together with age at baseline and sex as main effects. The assumption of proportionality of the hazards was tested by adding time interaction terms to the model.
Clinical characteristics of the study participants with type 2 diabetes are summarized in Table 1 according to their coronary artery disease (CAD) status. Case participants had significant CAD (angiographically-confirmed) and control participants had a negative cardiovascular history and normal exercise treadmill test. Age at examination, age at diagnosis of diabetes, and body weight were similar in the 2 groups. HbA1c (a measure of poor glycemic control) averaged slightly higher in patients with CAD rather than without. This difference was entirely due to an excess of CAD cases in the worst tertile of HbA1c, consistent with a nonlinear relationship between poor glycemic control and CAD risk. Treatment with insulin was more frequent in cases with CAD than control participants (50.9% as compared with 40.0%), as was a history of hypertension (80.4% as compared with 70.4%). A history of smoking was almost twice as common in case participants as in control participants (65.5% as compared with 37.6%). Almost half of the cases with CAD (45.7%) had a previous myocardial infarction.
Two SNPs on chromosome 9 (rs2383206 and rs10757278) that were reported to be associated with CAD in the general population10,11 were genotyped in both study groups. For both polymorphisms, the genotypes were in Hardy-Weinberg equilibrium in case and control participants. The genotype distributions for both SNPs were significantly different between case and control participants (P = .0002 for rs2383206 and P = .0049 for rs10757278; Table 2). For both, homozygotes for the G allele were more frequent in study participants with CAD than those without, the same pattern reported in the general population.10,11 No significant differences in genotype distributions were observed between case participants who received treatment at the Joslin Clinic and those who did not, or between participants who had a previous myocardial infarction and those who had not.
Haplotype analysis indicated that the effect of rs10757278 was secondary to its strong linkage disequilibrium with rs2383206 (D’ = 1, r2 = 0.78). The G allele of rs2383206 was associated with CAD regardless of the rs10757278 allele that was present on the same haplotype, whereas the A allele of rs10757278 was protective only when it occurred together with the protective allele of rs2383206. The primary role of rs2383206 in our study population was confirmed by the fact that the association with CAD disappeared for rs10757278 (P = .51), whereas it remained significant with rs2383206 (P = .013) when the 2 SNPs were analyzed together in a multivariable model. Thus, only rs2383206 was considered in further analyses.
The significant associations with CAD in Table 1 and those concerning the genotypes for rs2383206 are expressed again in Table 3 as ORs, both from univariable analysis and after adjustment in a multivariable model that included all the variables in the table. The univariable ORs were 1.45 (95% CI, 0.94-2.22) for rs2383206 heterozygotes and 2.37 (95% CI, 1.52-3.70) for allele G homozygotes, consistent with an additive mode of inheritance.
Among the other variables, association with smoking was strongest (OR, 3.24; 95% CI, 2.39-4.40), followed by sex (OR, 1.75; 95% CI, 1.28-2.38), antihypertensive therapy (OR, 1.73; 95% CI, 1.22-2.45), insulin therapy (OR, 1.53; 95% CI, 1.13-2.08), poor glycemic control (OR, 1.44; 95% CI, 0.98-2.1, for the highest tertile of HbA1c at examination vs the lowest), and history of hypercholesterolemia (OR, 1.29; 95% CI, 0.87-1.92). Except for insulin therapy and history of hypercholesterolemia, all variables remained significant in a multivariable model including all other predictors. The ORs for rs2383206, both for G/G homozygotes and for heterozygotes, were similar to those obtained in univariable analysis, indicating that the effect of this SNP was not mediated by an effect on the other cardiovascular risk factors.
The magnitude of the association between 9p21 locus and CAD appeared to be larger than that described in the general population. In a meta-analysis of the studies by Helgadottir et al10 and McPherson et al,11 including a total of 9583 CAD case participants and 32 292 control participants from the general population,10,11 the ORs were 1.24 for heterozygotes (95% CI, 1.17-1.32) and 1.46 for risk allele homozygotes (95% CI, 1.37-1.57), as compared with 1.45 (95% CI, 0.94-2.22) and 2.37 (95% CI, 1.52-3.70) in our study.
The contrast between the ORs in our study and those from the meta-analysis was significant in the case of risk allele homozygotes (P = .037) but not for heterozygotes (P = .49). The contrast with studies12,13 that used SNPs in incomplete linkage disequilibrium with rs2383206 gave similar results but did not reach significance for either genotype. While the higher OR for risk allele homozygotes in our study could plausibly be attributed to selection of control participants from the extreme low end of a liability distribution, it was also plausibly due to an enhanced effect of the G/G genotype in the intensely atherogenic milieu present in diabetes. In particular, we considered the possibility of an interaction with hyperglycemia, since it is the distinguishing characteristic of diabetes and excess glucose has potent proatherogenic effects in vitro.18
To explore this hypothesis, we divided study participants according to the 3 rs2383206 genotypes and whether they were in the worst tertile of HbA1c values at examination (HbA1c >7.6), the measure of glycemic control most associated with CAD in Table 1. Relative to the CAD risk for patients with neither an rs238206 risk allele nor the worst glycemic control, the risks for those with only poor glycemic control, or only 1 risk allele, or only 1 risk allele and poor glycemic control were similarly increased, although not significantly (Figure, A; OR, 1.25; 95% CI, 0.56-2.82; OR, 1.47; 95% CI, 0.87-2.46, and OR, 1.70; 95% CI, 0.86-3.01, respectively).
By contrast, the CAD odds for participants having 2 risk alleles but not poor glycemic control were increased 2-fold (OR, 1.99; 95% CI, 1.17-3.41), whereas the odds for study participants with the same genotype but poor glycemic control were increased 4-fold (OR, 4.27; 95% CI, 2.26-8.01). This effect magnification had a P value for interaction (ie, deviation from additivity in the log scale) between G/G genotype and glycemic control of 0.071 in a univariable analysis and 0.048 in a multivariable model including other cardiovascular risk factors.
Excluding the 60 days prior to examination, 68% of the study group (347 control and 151 case participants) had at least 2 HbA1c measurements in their medical records during the preceding 7 years (median, 11 measurements; interquartile range [IQR], 6-16). As compared with the rest of study group, these participants were younger at diabetes diagnosis (mean [SD], 51  years as compared with 55  years) and included fewer women (41% vs 56.9% in control participants and 24% vs 35% in case participants). In this subset, the magnitude of the interaction between the rs2383206 G/G genotype and poor glycemic control, when defined by the HbA1c at examination, was similar to that in the full case-control study (unadjusted regression coefficient for interaction, [SE] 0.49 [0.40], and 0.58 [0.32], respectively). HbA1c measurements spanned 1 year before study entry in 6.4% of participants, 2 years in 10.3%, 3 years in 10.9%, 4 years in 13.6%, and 5 years or greater in 58.8%.
Although the time-weighted average HbA1c value over these years was significantly correlated with the HbA1c value at examination (Spearman rank correlation, ρ = 0.64; P < .0001), use of the top tertile of the average values as the criterion for poor glycemic control resulted in a stronger interaction with the G/G genotype at rs2383206 (unadjusted regression coefficient for interaction, [SE] 1.11 [0.40]; Figure, B). The OR was 7.83 (95% CI, 3.49-17.6) for participants having both the G/G genotype and long-term poor glycemic control as compared with an OR of 1.54 (95% CI, 0.72-3.30) for participants with the G/G genotype but not long-term poor glycemia (Figure, B). This interaction was significant (unadjusted P = .0052; adjusted P = .0049) despite the reduced sample size and was not affected by further adjustment for mean arterial pressure and low-density lipoprotein and high-density lipoprotein cholesterol levels, suggesting that it was not due to a confounding effect of worse blood pressure or lipid control among individuals with high HbA1c values.
A similar interaction between 9p21 high-risk genotype and poor glycemic control was observed with respect to mortality in a study of 475 Joslin patients who had been recruited between 1993 and 1996 for genetic studies of type 2 diabetes and its complications (Table 4). As in the case-control study, multiple HbA1c measurements spanning several years before study entry were available for these participants in the Joslin medical records (median number of measures, 10; IQR, 6-15). On average, HbA1c levels were about 1 point higher than in the case-control study (8.3% vs 7.4%) due to a secular trend toward lower HbA1c values in the Joslin population during the last decade or so.
Table 5 shows the cumulative 10-year mortality in this cohort according to whether or not individuals carried the G/G genotype and whether or not they had a history of poor glycemic control (defined as the top tertile of the average HbA1c level before study entry [HbA1c >8.9]). Among individuals who did not carry the G/G genotype, no significant differences were observed in either all-cause or cardiovascular mortality between glycemic control groups. By contrast, among G/G carriers, mortality was about twice as high in individuals with poor glycemic control as compared with those in relatively good control (P = .021 for all-cause and (P = .020 for cardiovascular mortality; Table 5).
The P value for interaction between poor glycemic control and G/G genotype was .036 for all-cause and 0.049 for cardiovascular mortality. Similar evidence of interaction was obtained by means of a proportional hazard regression analysis adjusted for age and sex (P = .028 and P = .060 for all-cause and CVD mortality, respectively). Exclusion of 11 individuals who were also part of the case-control study did not change these results.
One or more genetic variants located on chromosome 9p21 and tagged by SNP rs2383206 are major risk factors for coronary artery disease among individuals with type 2 diabetes. In our population of diabetic participants, this effect is stronger than that reported in the general population due to a positive interaction between the genetic variant(s) and hyperglycemia. As was found in the general population, this association is not mediated by an effect of these genetic variants on other cardiovascular risk factors, since it is not attenuated by adjustment for these variables.10,11 This synergism between 9p21 locus and hyperglycemia on the risk of coronary artery disease translates into a similar interaction with respect to cardiovascular mortality among individuals with type 2 diabetes.
Other genes, including ADIPOQ, ADIPOR1, ENPP1, and TNFAIP3, have been reported to host polymorphisms influencing cardiovascular risk in type 2 diabetes.19-24 However, the present finding stands out from the previous ones in 2 respects. First, it concerns a genetic effect that was identified through a genome-wide approach and has been extensively replicated in the general population.10-13 Second, it is the first to demonstrate synergism with poor glycemic control.
The interaction between the 9p21 allele and glycemic control may help explain the discrepancy between the potent proatherogenic effects of glucose observed in vitro and the evidence from large clinical trials of limited benefit of good glycemic control on cardiovascular outcomes in diabetic participants.2,18,25-28 Poor glycemic control has an especially strong impact on cardiovascular risk in individuals who are homozygous for allele G at rs2383206, about 30% of individuals with type 2 diabetes. The other 70% are not as sensitive to the atherogenic effects of hyperglycemia. This heterogeneity could explain the past difficulties in demonstrating an association between glycemic control and cardiovascular outcomes.
Our findings are at variance with those by Broadbent et al,29 who found that the strength of the association between the 9p21 variant and CAD was similar among participants with diabetes and those without. That study, however, included individuals with both type 1 and type 2 diabetes. Furthermore, the estimate of the association between 9p21 variants and CAD in the diabetic stratum was based on a small number of control participants (n = 156) in whom asymptomatic CAD had not been excluded as had been in ours. Most importantly, that study did not present data on history of glycemic control. Thus, a fair comparison with our study is not possible.
SNP rs2383206 is placed in a 190 Kb region of high linkage disequilibrium containing 2 known genes (CDKN2A and CDKN2B), which code for 3 proteins (p16INK4a, ARF, and p15INK4b) that are expressed at high levels in a wide range of cell types, including endothelial and inflammatory cells. All 3 proteins are inhibitors of cyclin-dependent kinases controlling cell proliferation, cell aging, and apoptosis—functions that are all potentially relevant to the atherosclerotic process.30-32 Excess glucose is believed to foster atherosclerosis through multiple pathways involving the accrual of advance glycation end products, activation of protein kinase C, increased production of polyols and hexosamine, and increased oxidative stress.18,25
At what level the cellular pathways controlled by the 9p21 polymorphism(s) and those induced by high-glucose intersect remain to be determined. However, one possibility is that metabolic alterations associated with poor glycemic control, rather than hyperglycemia per se, are the actual factors responsible for the synergism with the 9p21 locus. The finding that the interaction was unaffected by adjustment for some of these metabolic traits such as blood pressure and cholesterol levels at examination is against this hypothesis, but further analyses using more precise indicators based on repeated measures are certainly warranted.
Another aspect that remains to be fully clarified is whether the association with type 2 diabetes reported in a region immediately centromeric to that associated with CAD33-35 plays any role in the observed interaction. However, a recent multicentric study has shown that the SNP associated with type 2 diabetes (rs10811661) is not associated with increased risk of CAD or other arterial disorders.36
Whether knowledge of modest genetic effects can improve disease prediction and treatment of common disorders remains uncertain.37 However, the magnitude of the joint effect of poor glycemic control and 9p21 locus favors some clinical benefit. If the probability of clinically significant CAD is approximately 30% for unselected type 2 diabetes participants, one can estimate from our data that this probability increases to 60% for individuals with diabetes who have Hba1c values in the upper tertile of the distribution and carry the high-risk genotype. Conversely, the availability of a test improving prediction does not necessarily imply that such test should be adopted in clinical practice. This decision should be based on an investigation of the cost-effectiveness of prevention strategies targeted at high-risk individuals rather than to the entire population of diabetic participants. This analysis must weigh the costs of the genetic test and available prevention strategies and the effectiveness of these strategies specifically in these high-risk patients.
Our study has 2 unique strengths, namely the contrast achieved by comparing angiographically-confirmed CAD case participants with control participants for whom CAD was ruled out by an exercise stress test, and the accurate assessment of long-term glycemic control through multiple HbA1c measurements spanning many years before study entry, in contrast to the crosssectional measurements available to most studies of CAD in type 2 diabetes.
However, some limitations of this study should be acknowledged. One is that the analysis of the interaction between genotype and glycemic control was based on small effective sample sizes. This translated into relatively high P values for interaction (.004-.05), raising the concern of a false-positive result. The fact that we found a similar interaction between glycemic control and the 9p21 locus for a related outcome (cardiovascular mortality) in an independent study based on a different design makes this possibility less likely, but additional replication studies are necessary to establish the interaction with statistical confidence.
Another limitation concerns generalizability. Case and control participants were recruited at the Joslin Diabetes Center and Beth Israel Deaconess Medical Center. It is possible that the strength of the association in our study may have been overestimated due to selective referral of especially severe CAD cases to these specialized centers. However, this does not seem to be the case since only half of the case participants had 3 stenotic vessels and fewer than half had a previous myocardial infarction. Generalizability is also affected by the fact that the additional risk to patients with the risk genotype was assessed against a select group of control participants rather than the general population of patients with type 2 diabetes.
A third potential limitation concerns the effectiveness of the exercise treadmill test to exclude CAD in control participants. Large meta-analyses evaluating the accuracy of the exercise treadmill test for the detection of CAD in the setting of normal ECG have determined the sensitivity of this test to be 72% with a specificity of 77%.15,38,39 Although the prevalence of asymptomatic CAD in diabetic patients has not been thoroughly investigated, a single, large prospective study found myocardial perfusion defects in 16% of asymptomatic diabetic patients.40 Assuming this prevalence of CAD in the population eligible for entry into the control group, the negative predictive value of our combined selection criteria is 93.5%. Therefore, only a small proportion of the individuals in the control group might have had asymptomatic obstructive CAD. Furthermore, such misclassification, if present, would have biased the results toward the null hypothesis, making our findings of association even more notable.
In conclusion, 9p21 locus and poor glycemic control interact in determining the odds of CAD in type 2 diabetes. This finding may have implications for the understanding of atherogenesis in diabetes and for the design of more effective prevention strategies. More broadly, it illustrates the complex etiology of multifactorial disorders and highlights the importance of accounting for gene-environment and gene-gene interactions in the quest for genetic factors contributing to these conditions.
Corresponding Author: Alessandro Doria, MD, PhD, MPH, Section on Genetics and Epidemiology, Joslin Diabetes Center, One Joslin Place, Boston, MA 02215 (email@example.com).
Author Contributions: Dr Doria had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Doria, Gervino, Hauser, Johnstone, Warram.
Acquisition of data: Doria, Wojcik, Xu, Gervino, Hauser, Johnstone, Nolan, Warram.
Analysis and interpretation of data: Doria, Wojcik, Xu, Gervino, Hauser, Hu, Warram.
Drafting of the manuscript: Doria, Wojcik, Xu, Hauser, Hu, Warram.
Critical revision of the manuscript for important intellectual content: Doria, Wojcik, Hauser, Johnstone, Hu, Gervino, Nolan, Warram.
Statistical analysis: Doria, Hu, Warram.
Obtained funding: Doria, Hu.
Administrative, technical, or material support: Doria, Wojcik, Xu, Nolan, Johnstone, Warram.
Study supervision: Doria, Gervino, Hauser, Warram.
Financial Disclosures: None reported.
Funding/Support:This study was supported by National Institutes of Health grants HL73168, HL71981, and DK36836 (Genetics Core of the Diabetes and Endocrinology Research Center at the Joslin Diabetes Center), and a grant from the Donald W. Reynolds Foundation.
Role of the Sponsors: None of the funding agencies had any role in the design and conduct of the study, in the collection, management, analysis, and interpretation of the data, or in the preparation, review, or approval of the manuscript.
Additional Contributions: We are grateful to Vincenzo Trischitta, MD, University of Rome, Rome, Italy; and Andrzej S. Krolewski, MD, PhD, Joslin Diabetes Center, Boston, for their valuable comments, and to Richard W. Nesto, MD, Lahey Clinic, Burlington, Massachusetts, for his initial help with the recruitment of study participants. The aforementioned individuals did not receive compensation for their contributions to this article. We are also grateful to the following employees of the Joslin Diabetes Center for their technical help: Christine Powers, BS, Ryan Thompson, BS, Maya Becker, BA, Aviva Bashan, BA, Jill Duffy, BS, Helen Kim, BS, Rachel Sagor, BA, and Celeste Amundsen, BA. We acknowledge the invaluable contribution by the individuals who participated in this study.
Create a personal account or sign in to: