Kaplan-Meier curves for patients with and without use of clopidogrel: propensity score−matched sub-group analyses. C statistics for propensity-score models were 0.80 for patients with diabetes and 0.85 for patients without diabetes. Diabetes-subgroup comprised 2005 clopidogrel-treated patients and 2005 non–clopidogrel-treated patients; the nondiabetes subgroup comprised 11 410 clopidogrel-treated patients and 11 410 non–clopidogrel-treated patients. P for interaction was obtained from adjusted Cox regression models between diabetes and nondiabetes subgroups: 0.02 for all-cause mortality, 0.01 for cardiovascular-related mortality, 0.09 for the combination end points.
Anderson C, Lyngbæk S, Nguyen CD, et al. Association of Clopidogrel Treatment With
Risk of Mortality and Cardiovascular Events Following Myocardial Infarction in Patients
With and Without Diabetes. JAMA. 2012;308(9):882-889.
eFigure 1. Persistence Rates
eFigure 2. Hazard Ratios
eTable. Baseline Characteristics of Propensity-Matched Subgroup
Charlotte Andersson, Stig Lyngbæk, Cu Dinh Nguyen, Mia Nielsen, Gunnar H. Gislason, Lars Køber, Christian Torp-Pedersen. Association of Clopidogrel Treatment With Risk of Mortality and Cardiovascular Events Following Myocardial Infarction in Patients With and Without Diabetes. JAMA. 2012;308(9):882–889. doi:10.1001/2012.jama.10779
Author Affiliations: Department of Cardiology, Gentofte Hospital (Drs Andersson, Lyngbæk, Nguyen, Gislason, and Torp-Pedersen and Ms Nielsen) and The Heart Center, Department of Cardiology (Dr Køber), Rigshospitalet, University of Copenhagen, Denmark.
Context Pharmacodynamic studies have shown that persistently high platelet reactivity is common in patients with diabetes in spite of clopidogrel treatment. Clinical trials have not convincingly demonstrated that clopidogrel benefits patients with diabetes as much patients without diabetes.
Objectives To estimate the clinical effectiveness associated with clopidogrel treatment after myocardial infarction (MI) in patients with diabetes.
Design, Setting, and Patients By individual-level linkage of the Danish nationwide administrative registries between 2002-2009, patients who were hospitalized with incident MI and who had survived and not undergone coronary artery bypass surgery 30 days after discharge were followed up for as long as 1 year (maximally until December 31, 2009). Adjusted for age, sex, comorbidity, calendar year, concomitant pharmacotherapy, and invasive interventions, hazard ratios that were associated with clopidogrel in patients with and without diabetes were analyzed by Cox proportional-hazard models and propensity score−matched models.
Main Outcome Measures All-cause mortality, cardiovascular mortality, and a composite end point of recurrent MI and all-cause mortality.
Results Of the 58 851 patients included in the study, 7247 (12%) had diabetes and 35 380 (60%) received clopidogrel. In total, 1790 patients (25%) with diabetes and 7931 patients (15%) without diabetes met the composite end point. Of these, 1225 (17%) with and 5377 (10%) without diabetes died. In total, 978 patients (80%) with and 4100 patients (76%) without diabetes died of events of cardiovascular origin. For patients with diabetes who were treated with clopidogrel, the unadjusted mortality rates (events/100 person-years) were 13.4 (95% CI, 12.8-14.0) vs 29.3 (95% CI, 28.3-30.4) for those not treated. For patients without diabetes who were treated with clopidogrel, the unadjusted mortality rates were 6.4 (95% CI, 6.3-6.6) vs 21.3 (95% CI, 21.0-21.7) for those not treated. However, among patients with diabetes vs those without diabetes, clopidogrel was associated with less effectiveness for all-cause mortality (HR, 0.89 [95% CI, 0.79-1.00] vs 0.75 [95% CI, 0.70-0.80]; P for interaction, .001) and for cardiovascular mortality (HR, 0.93 [95% CI, 0.81-1.06] vs 0.77 [95% CI, 0.72-0.83]; P for interaction, .01) but not for the composite end point (HR, 1.00 [95% CI, 0.91-1.10] vs 0.91 [95% CI, 0.87-0.96]; P for interaction, .08). Propensity score−matched models gave similar results.
Conclusion Among patients with diabetes compared with patients without diabetes, the use of conventional clopidogrel treatment after MI was associated with lower reduction in the risk of all-cause death and cardiovascular death.
Patients with diabetes have an increased risk of ischemic adverse events and death compared with patients without diabetes.1- 4 Some of this increase in risk may relate to platelet hyperreactivity and altered platelet receptor- and intracellular-signaling pathways; including an upregulation of the P2Y12 pathway seen in patients with diabetes.5 The P2Y12 pathway is the therapeutic target used for inhibition of platelet activation in clopidogrel treatment. Consequently, platelets in patients with diabetes often display persistently high platelet reactivity in spite of clopidogrel treatment, as evaluated by platelet function assays.6,7
High platelet reactivity during steady-state treatment with clopidogrel has been shown to increase the risk of major adverse cardiovascular events in patients with diabetes,8 but the overall clinical efficacy of clopidogrel treatment in patients with diabetes is not well investigated. Data from a few randomized clinical trials comparing dual antiplatelet therapy with clopidogrel plus aspirin, with aspirin as monotherapy in patients at high risk of ischemic events, have not convincingly demonstrated any improvement in prognosis for patients with diabetes, as opposed to a clear reduction in event rates for patients without diabetes, although no interactions were presented.9- 11
Currently, there are thienopyridine analogues available on the market for secondary prevention of myocardial infarction (MI) other than clopidogrel. Both prasugrel and ticagrelor have shown to cause a greater inhibition of the P2Y12 pathway than clopidogrel but have also shown to cause more bleedings.12,13 Because future selection of antithrombotic agents may depend on the presence of diabetes, we analyzed the outcomes associated with clopidogrel treatment after MI in patients with and without diabetes using the nationwide administrative health care related registries available in Denmark.
Danish health care is based on a tax-financed system that provides all citizens with equal access. Because individuals are given a personal and permanent civil registration number at the time of birth or immigration, administrative nationwide registries allow data retrieval at the individual level.
All hospitalizations and invasive procedures have been registered since 1978 in the Danish National Patient Register. Information includes dates and discharge diagnoses (1 main diagnosis and any bidiagnoses) registered according to the International Classification of Diseases (ICD) system. The MI diagnosis has been validated with good sensitivity, specificity, and positive predictive values.14 Dispensed prescriptions have consecutively been registered according to the Anatomical Therapeutic Chemical (ATC) classification system in the Danish Register of Medicinal Product Statistics since 1995. The registry includes data on amount and strength of dispensed tablets as well as dispensing dates, which has been shown to be accurate.15
All death certificates are registered in the Danish National Causes of Deaths register on immediate and contributing causes of deaths. The Danish National Population registry updates information on vital status on all Danish residents, and deaths are registered no more than 2 weeks after occurrence.
Through individual-level linkage of the administrative registries, all patients who were hospitalized with incident MI between 2002 and 2009 (main diagnoses; ICD-10 codes I21 and I22) were identified. Study start was defined as 30 days after discharge from index MI. This quarantine period was necessary to allow patients time for claiming a prescription for clopidogrel and concomitant pharmacotherapy, thereby avoiding immortal-time bias. Patients who had received a coronary artery bypass graft (CABG) surgery during this period were excluded because patients undergoing the procedure are prone to stop clopidogrel treatment after surgery.16
The following end points were defined: all-cause mortality, cardiovascular mortality (immediate or contributing cause of death; ICD-10 code I00-I99), and a combination of all-cause mortality and recurrent MI (ICD-10 code I21-I22). For analyses of the respective end points, patients were followed up for no longer than 365 days from study start and were censored after 365 days, at the time of an event, or at the end of 2009.
Comorbidity, defined as in the Ontario acute myocardial infarction mortality prediction rules,17 was identified by designated diagnoses at discharge from the index hospitalization and from other hospitalizations up to a year before MI. However, heart failure was not based on hospitalization diagnoses because of a very low sensitivity in the registries (29%).18 Instead, loop diuretic dosages (ATC code C03C) used at study baseline (ie, 30 days after discharge) were used as a proxy for heart failure and heart failure severity. This method has been used previously and has shown to correlate well with New York Heart Association functional class and the risk of mortality in patients with MI and heart failure.19,20 Mean dosages were calculated by dividing the amount of dispensed tablets with the dispensing time intervals, as described previously.19 Baseline use and persistence rates of clopidogrel (ATC code B01AC04) treatment were also calculated by this method.21
Concomitant pharmacotherapy was defined as at least 1 claimed prescription of a specific agent between 90 days prior to index hospitalization and 30 days after discharge (ATC codes: β-blockers, C07; statins, C10A; renin–angiotensin system [RAS] inhibitors, C09; thiazides, C03A; spironolactone or eplerenone, C03D; calcium-channel blockers, C08; digoxin, C01AA05; vitamin K antagonists, B01AA0; aspirin, B01AC06; metformin, A10BA02; sulfonylureas, A10BB; and insulin, A10A). In accordance with previous work,4 diabetes was defined as at least 1 claimed prescription of glucose-lowering medications (ATC A10); for the present study at any time between 90 days prior to hospitalization and 30 days after discharge.
Treatment with percutaneous coronary intervention (PCI; procedure codes, KFNG [Nordic classification system of operations]) was identified as early (day 0-1) and late (days 2-30) after index MI.
The study was approved by The Danish Data Protection Agency. Registries were available at Statistics Denmark in an anonymous set-up, disabling identification of individuals but enabling individual-level linkage between registries. In Denmark, retrospective registry-based studies do not need ethical approval or participant informed consent. The Danish Data Protection Agency has approved our use of registries.
Tests for differences in baseline characteristics between clopidogrel and non–clopidogrel-treated individuals were performed with χ2 and t tests for discrete and continuous variables, respectively. Multivariable Cox-proportional regression analyses were used to calculate the hazard ratios associated with clopidogrel treatment in patients with and without diabetes. Two different dummy variables according to clopidogrel treatment were created; 1 for patients with diabetes and 1 for patients without diabetes; enabling the whole population to be analyzed in the same model. All models were adjusted for age, sex, concomitant pharmacotherapy, heart failure severity group, calendar year, average income in a 5-year period prior to hospitalization, comorbidity, PCI treatment day 0 through 1, and PCI treatment days 2 through 30. Tests for differences in outcomes associated with clopidogrel treatment in patients with and without diabetes were performed by inclusion of an interaction term in the overall model. All tests were 2-sided and P <.05 was considered statistically significant.
To ensure robustness of our findings, additional propensity-based subgroup analyses were performed. In these, patients using clopidogrel were matched with patients not using clopidogrel (1:1) on all baseline characteristics from Table 1 using the Greedy matching macro (http://mayoresearch.mayo.edu/mayo/research/biostat/upload/gmatch.sas; accessed on August 15, 2011). Finally, to ensure that numbers were not affected by the use of a 30-day quarantine period, all main analyses were repeated with the use of a 7-day quarantine period. Cumulative incidence curves and graphs over persistence rates were generated using the Kaplan-Meier method and test for differences between strata were done by log-rank test. Persistence curves were divided into 2 periods because recommended treatment length with clopidogrel was changed from 6 months in 2002-2003 to 12 months in 2004.22
All analyses were performed using SAS version 9.2 (SAS Institute Inc).
Of the 58 851 patients—7247 of whom (12%) had diabetes—who were included in the analyses, 35 380 (60%) received clopidogrel at study baseline. Characteristics of patients with and without diabetes stratified by clopidogrel treatment are shown in Table 1. For both groups, clopidogrel users were younger, more often men, had lower prevalence of comorbidity, and less often had heart failure than the nonclopidogrel users. Persistence rates of clopidogrel treatment were comparable in patients with and without diabetes, eFigure 1.
Patients were followed up for a median of 365 days (range, 0-365 days). In total, 1790 patients (25%) with and 7931 patients (15%) without diabetes met the composite end point. Of these, 1225 (17%) with and 5377 (10%) without diabetes died. In all, 978 patients (80%) with and 4100 patients (76%) without diabetes died because of cardiovascular-related events.
Crude incidence rates for the 3 defined end points in patients with and without diabetes stratified by clopidogrel treatment are shown in Table 2. Of patients with diabetes, those who took clopidogrel had an all-cause mortality per 100 person-years of 13.4 (95% CI, 12.8-14.0) and those who were not taking clopidogrel had an all-cause rate of 29.3 (95% CI, 28.3-30.4). Of patients who did not have diabetes, those who took clopidogrel had an all-cause mortality of 6.4 (95% CI, 6.3-6.6) and those who did not take clopidogrel had an all-cause mortality of 21.3 (95% CI, 21.0-21.7). Adjusted for other variables (hazard ratios [HRs] associated with other predictive variables are shown in the eFigure 2, patients with diabetes were found to have a smaller relative risk reduction than patients without diabetes (Figure 1).
When adjusted, this finding applied for the all-cause mortality end point (HR, 0.89 [95% CI, 0.79-1.00] for patients with diabetes vs HR, 0.75 [95% CI, 0.70-0.80] for patients without diabetes, P for interaction, .001) and the cardiovascular mortality end point (HR, 0.93 [95% CI, 0.81-1.06] for patients with vs HR, 0.77 [95% CI, 0.72-0.83] for patients without diabetes, P for interaction, .01), but not for the composite end point (HR, 1.00 [95% CI, 0.91-1.10] vs 0.91 [0.87-0.96] P for interaction, .08). Stratified according to PCI treatment, similar (nonsignificant) trends were seen for patients treated with PCI and for those treated with colopidogrel (Figure 2). There was no evidence of a similar differential effectiveness associated with aspirin treatment among patients with and without diabetes (P for interactions: .77 for all-cause mortality, .67 for cardiovascular, and .66 for the combined end point).
In subanalyses, there was no evidence toward a differential effectiveness of clopidogrel in patients with diabetes treated with and without metformin (P for interactions: .63 for all-cause mortality, .37 for cardiovascular mortality, and .63 for the combined end point), or with and without insulin (P for interactions: .43 for all-cause mortality, .34 for cardiovascular mortality, and .51 for the combined end point).
The propensity score−matching identified 4010 patients with diabetes, 2005 of whom were treated with clopidogrel and 2005 were not and found 22 820 patients who did not have diabetes, 11 410 of whom were treated with clopidogrel and 11 410 who were not. C statistics for propensity-score models were 0.80 for those with and 0.85 for those without diabetes. Baseline characteristics were comparable between patients treated with clopidogrel and those who were
not (eTable). Similar to the main analysis, these analyses revealed a diminished effectiveness associated with clopidogrel treatment in patients with diabetes; Figure 3.
Starting the analyses only 7 days after discharge instead of 30 days provided similar results. Hazard ratios for all-cause mortality for clopidogrel treatment were 0.80 (95% CI, 0.76-0.84) for patients with vs 0.62 (95% CI, 0.59-0.66) for patients without diabetes (P for interaction, <.001); for cardiovascular mortality, 0.82 (95% CI, 0.72-0.92) for patients with vs 0.62 (95% CI, 0.58-0.67) for patients without diabetes (P for interaction, .0001); and for the combined end point, 0.80 (95% CI, 0.74-0.88) for patients with vs 0.72 (95% CI, 0.69-0.75) for patients without diabetes (P for interaction, .02).
Using a cohort of nearly 60 000 patients with first-time MI between 2002 and 2009, the present analysis demonstrated a reduced 1-year clinical effectiveness associated with clopidogrel treatment in patients with diabetes. This supports previous knowledge of platelets in diabetes showing high platelet reactivity and diminished responsiveness to standard clopidogrel treatment.
In the present analysis, clopidogrel treatment was associated with a relative risk reduction of 25% for all-cause mortality, 23% for cardiovascular mortality, and 9% for the combination of recurrent MI and all-cause mortality in patients without diabetes, and a relative risk reduction of 11% for all-cause mortality but no significant reduction in cardiovascular mortality or the combined end point in patients with diabetes. These numbers correspond well with the findings from randomized clinical trials, supporting the validity of our registry-based analyses.10,11 Because previous randomized clinical trials were not designed to investigate the relative efficacy of standard clopidogrel treatment in patients with diabetes, it has been impossible to conclude definitively whether a statistically significant diminished effectiveness of clopidogrel treatment is present for patients with diabetes or whether previous findings may be due to chance and small numbers of patients in the diabetes subgroups. This study supports the conclusion that there may be a difference of effect of clopidogrel among those with diabetes compared with those without it.
Two large randomized clinical trials investigating the long-term outcomes of clopidogrel for patients at high risk of cardiovascular events have presented subgroup analyses of patients with diabetes, and neither of them have demonstrated any clear beneficial effect of clopidogrel treatment. The Clopidogrel for the Reduction of Events During Observation (CREDO) trial evaluated the effect of 75 mg of clopidogrel over placebo on 1-year incidence of the composite of death, MI, or stroke in 2116 patients who were planned for elective PCI or who had a high likelihood of undergoing PCI.10 Subgroup analyses revealed no significant effect of clopidogrel in the diabetes subgroup (n = 560; relative risk reduction, 11.2% [95% CI, −46.8% to 46.2%]) as opposed to a clear risk-reducing effect of clopidogrel in the nondiabetes subgroup (relative risk reduction, 32.8% [6.8%-51.6%]).10
The Clopidogrel for High Atherothrombotic Risk and Ischemic Stabilization, Management, and Avoidance (CHARISMA) trial enrolled 15 603 high-risk patients without acute coronary syndrome, but with clinically evident cardiovascular disease or multiple risk factors.9 The study compared the efficacy of 75 mg of clopidogrel to placebo in addition to low-dose aspirin (75-162 mg) on the risk of a composite end point of stroke, MI, or cardiovascular death.9 The trial showed no benefit of 75 mg of clopidogrel during 28 months of follow-up. Interestingly, however, subgroup analyses showed a clear trend toward a risk-reducing effect of clopidogrel in the nondiabetic subgroup as opposed to a neutral outcome in the diabetes subgroup. Because diabetes was prevalent in 43% of the total study population, it may be speculated that exclusion of patients with diabetes prior to enrollment would have resulted in a significantly greater efficacy of clopidogrel than aspirin.
Current US guidelines recommend 12 months of dual platelet inhibition with a thienopyridine analogue (clopidogrel or prasugrel) and aspirin after acute coronary syndromes to prevent recurrent ischemic events but do not endorse one thienopyridine analogue over another.23,24 The guidelines acknowledge the superiority of prasugrel over clopidogrel for reduction of adverse events overall but also emphasize the increased risk of bleedings associated with treatment with prasugrel.23,24
Data from the Trial to Assess Improvement in Therapeutic Outcomes by Optimizing Platelet Inhibition With Prasugrel-Thrombolysis in Myocardial Infarction 38 (TRITON-TIMI 38), a randomized clinical trial comparing clopidogrel with prasugrel after MI, indicated that treatment with prasugrel may have a greater clinical benefit in patients with diabetes compared with patients without diabetes, which—although speculative—could relate to a limited effect of clopidogrel rather than enhanced effects of prasugrel in this particular subgroup.25 In this context, prasugrel treatment has shown to cause greater inhibition of platelet aggregation and a lower rate of nonresponders than clopidogrel.26
Although more study is needed, prasugrel may constitute an attractive alternative to clopidogrel in patients with diabetes with acute coronary syndromes, especially if recurrent ischemic events have occurred during clopidogrel treatment. This is congruent with recent European guidelines that recommend ticagrelor or prasugrel use before clopidogrel in acute coronary syndromes.27 It is however still unknown whether prasugrel and high doses of clopidogrel have comparable clinical efficacy in patients with diabetes because randomized studies with hard end points comparing these 2 combinations do not exist. Platelet inhibition with high-dose clopidogrel (150 mg) has been compared with standard dose prasugrel (10 mg) in a cohort of patients with planned PCI (30% of whom had diabetes). In this study, prasugrel was associated with more consistent levels of platelet inhibition than clopidogrel treatment, which may support the choice of prasugrel use before clopidogrel for patients with diabetes, but more studies are needed to establish the optimal antiplatelet treatment strategy.28
We acknowledge that no definite conclusions on causal mechanisms can be drawn from observational studies. Although several differences between patients receiving and not receiving clopidogrel could be identified and adjusted for in the present analyses, unmeasured confounders may have influenced the results. Contrary to randomized clinical trials, the study population was unselected, which may explain the rather high mortality rates found in the present analyses compared with those seen in randomized trials, but it cannot be excluded that some of the findings may have related to confounding by indication. Specifically, data on stents and stent types, body mass index, hematology and blood biochemistry, and left ventricular ejection fraction were not available, which—despite propensity score−matched subgroup analyses revealing similar results—may have influenced the findings. Furthermore, no information was available on smoking status, which may influence the efficacy of clopidogrel.29 Finally, the use of loop diuretics as a proxy of heart failure and heart failure severity also comprises a limitation of the present study.
In summary, data from previous clinical trials and data from the present analyses strongly suggest that patients with diabetes have a significantly diminished relative effectiveness of conventional platelet-inhibition with clopidogrel after MI compared with patients without diabetes. It should however be emphasized that considering the relatively higher absolute risks found for patients with diabetes, use of clopidogrel may still translate into a significant reduction in event rates for patients with diabetes, which data from the subgroup analyses supported. Available data nevertheless raise a possibility that patients with diabetes may benefit from a more potent platelet inhibitor strategy to achieve a relative risk reduction similar to patients without diabetes.25
Corresponding Author: Charlotte Andersson, MD, PhD, Department of Cardiology, Gentofte Hospital, Niels Andersens vej 65, 2900 Hellerup, Denmark (firstname.lastname@example.org).
Author Contributions: Dr Anderson 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: Lyngbaek, Gislason, Torp-Pedersen, Andersson.
Analysis and interpretation of data: Lyngbaek, Nguyen, Nielsen, Gislason, Køber, Torp-Pedersen, Andersson.
Drafting of the manuscript: Lyngbaek, Andersson.
Critical revision of the manuscript for important intellectual content: Lyngbaek, Nguyen, Nielsen, Gislason, Køber, Torp-Pedersen, Andersson.
Statistical analysis: Køber, Torp-Pedersen, Andersson.
Obtained funding: Torp-Pedersen.
Administrative, technical, or material support: Torp-Pedersen.
Study supervision: Gislason, Køber, Torp-Pedersen.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Køber reported that he has received honorarium as speaker for Servier. Dr Torp-Pedersen reported that he has served on advisory boards and served on study steering committees for Cardiome, Sanofi, and Merck; and has obtained research grants from Bristol-Myers Squibb, and Sanofi. Dr Andersson reported receiving a month's salary for work on the study.
Funding/Support: The study was funded by an internal research foundation grant from Department of Cardiology, Gentofte Hospital.
Role of the Sponsor: The foundation had no influence on design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.
This article was corrected for errors on September 11, 2012.