Angina and NonST-Segment Elevation Myocardial Infarction Matthew D. Solomon, MD, PHD,*y Alan S. Go, MD,*yz David Shilane, PHD,y Derek B. Boothroyd, PHD,y Thomas K. Leong, MPH,* Dhruv S. Kazi, MD, MSC, MS,zx Tara I. Chang, MD, MS,y Mark A. Hlatky, MDy Oakland, Stanford, and San Francisco, California Objectives This study sought to examine the effectiveness of clopidogrel in real-world, medically managed patients with unstable angina (UA) or nonST-segment elevation myocardial infarction (NSTEMI). Background Although clinical trials have demonstrated the efcacy of clopidogrel to reduce cardiovascular (CV) morbidity and mortality in medically managed patients with UA or NSTEMI, the effectiveness of clopidogrel in actual clinical practice is less certain. Methods A retrospective cohort study was conducted of Kaiser Permanente Northern California members without known coronary artery disease or prior clopidogrel use who presented with UA or NSTEMI between 2003 and 2008 and were medically managed (i.e., no percutaneous coronary intervention or coronary artery bypass grafting during the index hospitalization or within 7 days post-discharge). Over 2 years of follow-up, we measured the association between clopidogrel use and all-cause mortality, hospital stay for MI, and a composite endpoint of death or MI using propensity-matched multivariable Cox analyses. Results We identied 16,365 patients with incident UA (35%) or NSTEMI (65%); 36% of these patients were prescribed clopidogrel within 7 days of discharge. In 8,562 propensity scorematched patients, clopidogrel users had lower rates of all-cause mortality (8.3% vs. 13.0%; p < 0.01; adjusted hazard ratio [HR]: 0.63; 95% condence interval [CI]: 0.54 to 0.72) and the composite of death or MI (13.5% vs. 17.4%; p < 0.01; HR: 0.74, CI: 0.66 to 0.84), but not MI alone (6.7% vs. 7.2%; p 0.30; HR: 0.93, CI: 0.78 to 1.11), compared with nonusers of clopidogrel. The association between clopidogrel use and the composite of death or MI was signicant only among patients presenting with NSTEMI (HR: 0.67; CI: 0.59 to 0.76; p int < 0.01), not among those presenting with UA (HR: 1.25; CI: 0.94 to 1.67). Conclusions In a large, community-based cohort of patients who were medically managed after UA/NSTEMI, clopidogrel use was associated with a lower risk of death and MI, particularly among patients with NSTEMI. (J Am Coll Cardiol 2014;63:224957) 2014 by the American College of Cardiology Foundation In 2001, the CURE (Clopidogrel in Unstable Angina to Prevent Recurrent Events) trial showed that dual- antiplatelet therapy with aspirin and clopidogrel improved outcomes among patients with unstable angina (UA) or nonST-segment myocardial infarction (NSTEMI) (1). Patients randomly assigned to long-term treatment with aspirin in combination with clopidogrel had a lower rate of the combined endpoint of CV death, nonfatal MI, or stroke within 1 year compared with patients assigned to aspirin alone. On the basis of the results of CURE, clopidogrel was given a Class I recommendation in the American College of Cardiology/American Heart Association (ACC/AHA) guideline statements for patients treated medically after UA or NSTEMI for 1 month and ideally up to 9 months (2). See page 2258 From the *Division of Research, Kaiser Permanente Northern California, Oakland, California; yStanford University School of Medicine, Stanford, California; zUniversity of California, San Francisco, San Francisco, California; and the xSan Francisco General Hospital, San Francisco California. This work is supported by grant 0875162N from the American Heart Association, Dallas, Texas. Dr. Go is the recipient of a research grant from Genentech. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Deepak L. Bhatt, MD, MPH, served as Guest Editor for this paper. Manuscript received September 26, 2013; revised manuscript received January 24, 2014, accepted February 5, 2014. Journal of the American College of Cardiology Vol. 63, No. 21, 2014 2014 by the American College of Cardiology Foundation ISSN 0735-1097/$36.00 Published by Elsevier Inc. http://dx.doi.org/10.1016/j.jacc.2014.02.586 Most patients in CURE were medically managed (only one- fth initially underwent revascu- larization procedures), and the rate of subsequent revasculariza- tion during the rst 12 months was low (15% in the clopidogrel arm of the study and 14% in the aspirin arm). Although clo- pidogrel therapy led to similar risk reductions in patients who were medically managed or who underwent revascularization, the role of clopidogrel in percutane- ous coronary intervention (PCI) has received more attention, both periprocedurally (311) and post- PCI (12). Perhaps as a result, rates of clopidogrel use after PCI are extremely high (13). In contrast, 50% to 60% of patients with UA or NSTEMI in clinical practice are medically managed (i.e., do not undergo revascularization during the index hospital stay) (1416). In addition, clopidogrel use in medically managed patients with UA or NSTEMI remains low; less than one-half of medically managed patients with UA or NSTEMI will receive clopi- dogrel on discharge (13,1719). Further, the impact of clo- pidogrel among medically managed patients with UA or NSTEMI who are treated outside a clinical trial remains unclear. We therefore examined the effectiveness of clopi- dogrel in a real-world, community-based cohort of patients who were medically managed after hospital discharge for UA or NSTEMI. Methods This study was approved by the Institutional Review Boards of the Kaiser Foundation Research Institute (Oakland, California) and of Stanford University (Stanford, Califor- nia). Waiver of informed consent was obtained because of the nature of the study. Source population. The source population was from Kaiser Permanente Northern California, a large, inte- grated healthcare delivery system that provides care to >3.2 million individuals in San Francisco and the greater Bay Area. The health plan membership is broadly repre- sentative of the local and statewide population, apart from slightly lower representation at the extremes of age and income. Study population. We assembled an incident disease cohort by rst identifying all adult (age >30 years) members hospitalized for an initial episode of UA or NSTEMI be- tween January 1, 2003 and December 31, 2007. UA was dened by a primary hospital discharge diagnosis (Interna- tional Classication of Diseases-Ninth Edition) of 411.x or a combination of a primary discharge diagnosis code of 414.0 and a secondary discharge diagnosis of 411.x for the same hospital stay. NSTEMI was dened by a primary diagnosis code of 410.7, 410.8, or 410.9. To ensure an incident disease cohort, we excluded any patients with earlier diagnoses of UA, NSTEMI, ST-segment elevation myo- cardial infarction (STEMI), coronary artery bypass grafting (CABG), or PCI for at least the previous 12 months, although most patients had a much longer (>12 months) lookback period for these diagnoses and procedures. We dened the index date for this analysis as 7 days after hospital discharge, to allow subjects the opportunity to ll a clopidogrel prescription after hospital stay. To identify medi- cally managed patients, we excluded patients who received any coronary revascularization procedures (CABG or PCI) during their initial hospital stay or within 7 days of hospital discharge. Patients with any clinical event, including UA, MI, or death, within the rst 7 days post-discharge were also excluded. We adopted a new-user design (20) by excluding patients who had used clopidogrel in the 120 days before developing UA or NSTEMI. To match the eligibility criteria of the CURE study, we also excluded patients with an earlier history of stroke or transient ischemic attack (TIA), as well as patients taking warfarin or those with uncontrolled hypertension (systolic blood pressure >180 mm Hg) at baseline. We also excluded patients with a history of maintenance dialysis or organ transplantation. Covariates and exposure of interest. We dened clopi- dogrel users as patients who lled a prescription within 7 days post-discharge. To mimic an intention-to-treat design, we dened nonusers as patients who never started clopidogrel or who did so after the rst 7 days of follow-up, and we did not allow patients to cross over into the clopi- dogrel arm of the study. To capture complete data on de- mographic characteristics, comorbidities, and medication use, we restricted the analysis to patients with complete demographic data and at least 12 months of continuous membership and continuous pharmacy benet before the index date. We obtained data on age, sex, and self-reported race or ethnicity from health plan databases. We identied co- morbid conditions up to 4 years before the index date by using previously validated approaches using diagnosis and procedure codes and current procedural terminology codes recorded in health plan hospital stay, ambulatory, laboratory, and pharmacy databases for the following comorbidities: a history of heart failure, peripheral arterial disease, valvular heart disease, diabetes mellitus, hypertension, dyslipidemia, or bleeding during hospital stay (2127). We used pharmacy data to identify post-discharge use of angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, beta- blockers, and statins. We also included indicator variables for the year of the index date to control for secular trends, an indicator for the type of presenting event (UA or NSTEMI), and an indicator for patients in the group that had not received clopidogrel who underwent PCI between days 7 and 14 and who subsequently started taking clopidogrel. Abbreviations and Acronyms ACS = acute coronary syndrome (s) CABG = coronary artery bypass grafting CAD = coronary artery disease CV = cardiovascular MI = myocardial infarction NSTEMI = nonST-segment elevation myocardial infarction PCI = percutaneous coronary intervention PPI = proton pump inhibitor TIA = transient ischemic attack UA = unstable angina Solomon et al. JACC Vol. 63, No. 21, 2014 Clopidogrel in Unstable Angina or NSTEMI June 3, 2014:224957 2250 Outcomes. We examined a composite outcome of all-cause mortality and acute myocardial infarction (MI) requiring hospital stay during the 2 years after discharge, as well as the individual endpoints of death and MI. All-cause mortality was identied from health plan databases, supplemented by linkage with Social Security Administration vital status les and California state death certicate records. MI was dened as a hospital admission with a primary diagnosis code of 410.x1. Propensity score matching. We used propensity score matching to control selection bias and to create demo- graphically and clinically comparable cohorts (28). The propensity score for use of clopidogrel within 7 days was developed on the basis of the following variables: age; race or ethnicity; sex; year of index date; type of presenting event (UA or NSTEMI); smoking status; estimated glomerular ltration rate; systolic blood pressure; a history of bleeding, heart failure, hypertension, hyperlipidemia, or diabetes; and the use of angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, beta-blockers, or statins. Pa- tients in the clopidogrel-treated cohort were matched with the group that had not received clopidogrel in a 1:1 ratio by using a modied greedy matching algorithm without replacement, on the basis of propensity scores within 1%. Statistical analysis. We conducted the analysis in an intention-to-treat framework (i.e., not an on-treatment analysis because we did not allow crossover between study arms after 7 days post-discharge). We compared means of baseline characteristics by using 2-sided Student t tests. We used Kaplan-Meier estimators to describe survival functions for the selected outcomes and multivariable Cox regression analysis to test the association of clopidogrel use with outcomes, after adjustment for the covariates described earlier. We conducted analyses on unadjusted and propensity- matched models and performed sensitivity analyses, and we examined interaction effects (p int ) between outcomes and the type of initial presentation (UA or NSTEMI) and several patient subgroups (history of diabetes, history of smoking, and older age). Sensitivity analyses were performed as follows: by varying the time window after discharge that dened clopi- dogrel use; by including patients who were excluded by our eligibility criteria (patients using warfarin at baseline, dialysis recipients, patients with very high blood pressure, and patients withanearlier stroke or TIA); and, to examine whether proton pump inhibitor (PPI) use modies the effect of clopidogrel in this population, by restricting analyses to those patients who lled baseline prescriptions for PPIs in both the clopidogrel user group and the nonuser group. All analyses were per- formed with the R statistical package, version 2.15.3 (R Development Team, Vienna, Austria). Results We identied 39,455 patients 30 years of age or older with incident disease between 2003 and 2008. After applying the exclusion criteria, the analysis sample included 16,365 eligible patients, 5,961 (36%) of whom lled a prescription for clopidogrel with the rst 7 days after discharge (Fig. 1). In the overall study population, patients who used clopi- dogrel were younger and more likely to be male and to smoke, but they were otherwise healthier with fewer comorbidities (Table 1). Use of CV medications before the incident event was lower among clopidogrel users (see Table 1), but after hospital discharge, CV drug use was Figure 1 Cohort Assembly Final cohort derivation for clopidogrel users and nonusers after the application of study eligibility criteria. CABG coronary artery bypass grafting; CHD coronary heart disease; D/C discharge; HTN hypertension; MI myocardial infarction; NSTEMI nonST-segment elevation myocardial infarction; PCI percutaneous coronary intervention; SBP systolic blood pressure; STEMI ST-segment elevation myocardial infarction, TIA transient ischemic attack; UA unstable angina. JACC Vol. 63, No. 21, 2014 Solomon et al. June 3, 2014:224957 Clopidogrel in Unstable Angina or NSTEMI 2251 higher among clopidogrel users (Table 1). Mean length of follow-up was 976 days for clopidogrel users and 876 days for clopidogrel nonusers. Median and mean duration of continuous clopidogrel use among users was 188 days (interquartile range: 82 to 603 days) and 206 days, respec- tively. Most (97%) clopidogrel users lled their prescriptions on the day of discharge. Patients who did not ll a prescription for clopidogrel within 7 days of discharge from the index hospital stay had a low rate of subsequent initiation of clopidogrel: of the 10,404 initial nonusers, 3.4% initiated clopidogrel between 7 and 30 days after discharge, and another 7.4% initiated clopidogrel after 30 days post-discharge. More than one- third (38%) of the patients who initiated clopidogrel more than 7 days after discharge (332 of 859) did so after subsequent PCIs, most of which occurred well after discharge (24% between days 7 and 30, 58% between 30 and 365 days, and 18% between 1 and 2 years post- discharge). The c-statistic for the propensity score model for pre- dicting clopidogrel use within 7 days after discharge was 0.72. We matched 72% of clopidogrel users (4,281 of 5,961) with nonusers (1:1 match) within 1% (2 decimal places) of propensity score; 97% of pairs were matched within 0.1% (3 decimal places). Baseline covariates were well balanced in the matched cohort (see Table 1), and there were no statistically signicant differences among measured cova- riates for clopidogrel users and nonusers. In unadjusted analyses, the composite endpoint of all- cause mortality and MI requiring hospital stay over 2-years of follow-up was lower in patients receiving clopidogrel in both the overall population and in the propensity score Table 1 Characteristics of Study Population Unmatched Propensity-Matched Clopidogrel (n 5,961) No Clopidogrel (n 10,404) p Value Clopidogrel (n 4,281) No Clopidogrel (n 4,281) p Value Presentation UA, % 31.7 36.8 d 36.0 36.2 d NSTEMI, % 68.3 63.2 <0.01 64.0 63.8 0.84 Demographics Age (mean), yrs 65.8 70.5 <0.01 68.1 68.0 0.64 Male, % 66.3 53.1 <0.01 60.1 61.0 0.36 White race, % 73.0 73.4 0.56 73.1 73.3 0.83 Comorbidities Smoker, % 33.1 30.7 <0.01 31.6 31.9 0.75 Heart failure, % 4.9 11.6 <0.01 6.7 7.6 0.12 Bleeding, % 1.7 3.6 <0.01 2.2 2.1 0.83 HTN, % 58.9 65.2 <0.01 64.1 64.4 0.80 Diabetes, % 28.4 32.1 <0.01 32.1 31.7 0.69 Hyperlipidemia,% 84.3 80.0 <0.01 84.6 84.7 0.93 Patient characteristics eGFR >90, % 7.2 6.1 <0.01 6.9 6.0 0.29 eGFR 6090, % 36.7 31.8 35.9 34.7 eGFR 4560, % 17.7 20.4 20.0 19.9 eGFR 3045, % 9.3 14.6 11.9 12.0 eGFR 1530, % 3.0 6.0 3.9 4.1 eGFR <15, % 0.4 1.0 0.5 0.7 eGFR, unknown, % 25.7 20.1 20.9 22.6 SBP <120, % 20.1 23.9 <0.01 22.0 22.8 0.75 SBP 120130, % 16.5 16.3 16.7 16.6 SBP 130140, % 24.5 22.8 24.1 23.5 SBP 140160, % 20.1 19.2 19.6 19.5 SBP 160180, % 6.1 6.6 6.7 6.1 SBP 180, % 0.0 0.0 0.0 0.0 SBP, unknown, % 12.7 11.2 11.0 11.5 Rx use after event* Statin, % 90.8 70.4 <0.01 87.2 86.9 0.68 ACEI/ARB, % 74.7 58.0 <0.01 68.9 69.6 0.47 Beta-blocker, % 91.6 74.8 <0.01 88.7 88.3 0.57 PPI, % 20.3 23.1 <0.01 21.7 22.9 0.19 *Prior Rx or started within 7 days of discharge. ACEI angiotensin converting enzyme inhibitor; ARB angiotensin receptor blocker; eGFR estimated glomerular ltration rate; HTN hypertension; NSTEMI nonST-segment elevation myocardial infarction; PPI proton pump inhibitor; Rx treatment; SBP systolic blood pressure; UA unstable angina. Solomon et al. JACC Vol. 63, No. 21, 2014 Clopidogrel in Unstable Angina or NSTEMI June 3, 2014:224957 2252 matched cohort (Fig. 2). In the propensity-matched cohort, the composite of death or MI was lower among clopidogrel users (13.5% vs. 17.4%; p < 0.01; adjusted hazard ratio [HR]: 0.74, 95% condence interval [CI]: 0.66 to 0.84) (Fig. 3). All-cause mortality over 2 years was also lower for clopidogrel users (8.3% vs. 13.0%; p<0.01; HR: 0.63, CI: 0.54 to 0.72), but the rate of acute MI was not signif- icantly different (6.7% vs. 7.2%; p 0.30; HR: 0.93; CI: 0.78 to 1.11). There was evidence of treatment effect modication by clinical presentation (NSTEMI vs. UA) and by a history of smoking, but not by a history of diabetes. The impact of older age yielded mixed results, with a signicant interaction for the composite of death and MI, but not for the indi- vidual endpoint of all-cause mortality (Figs. 3 and 4). Compared with patients who presented with UA, patients who presented with NSTEMI had a stronger association of clopidogrel use with the composite of death or MI (HR: 0.67 vs. 1.25; p int < 0.01), as well as with the individual endpoints of mortality (HR: 0.56 vs. 1.20; p int < 0.01), and acute MI (0.84 vs. 1.38; p int 0.03). The association of clopidogrel with the composite of death or MI was similar, however, among patients with and without diabetes (HR: 0.83 vs. 0.69; p int 0.14), but it was stronger among patients more than 70 years of age (HR: 0.70 vs. 0.88; p int 0.04), as well as among nonsmokers (HR: 0.69 vs. 0.86; p int 0.02). Results were similar for the individual endpoint of all-cause mortality, although the interaction between older age and clopidogrel use was not signicant (Fig. 4). Results were not materially changed in sensitivity analyses that dened clopidogrel use as any prescription lled within 30 days, that included subjects excluded by our eligibility criteria (i.e., patients taking warfarin, or undergoing dialysis, or with very high blood pressure, or with an earlier stroke or TIA), and that restricted the analysis to patients taking PPIs at baseline (data not shown). Discussion In our analysis of a large, community-based cohort of medi- cally managed patients with UA or NSTEMI, initiation of Figure 2 Kaplan-Meier Curves for All-Cause Mortality and Acute MI Unadjusted Kaplan-Meier estimates for all-cause mortality and acute myocardial infarction (MI) for clopidogrel users and nonusers. JACC Vol. 63, No. 21, 2014 Solomon et al. June 3, 2014:224957 Clopidogrel in Unstable Angina or NSTEMI 2253 clopidogrel within the rst week after hospital discharge was associated with a lower risk of the composite outcome of death or MI over 2 years of follow-up. The association be- tween clopidogrel use and outcomes was signicantly greater among patients who presented with NSTEMI, among nonsmokers, and among older patients. The reduction in the composite endpoint of death and MI replicates, in a real-world cohort, the ndings of the land- mark CURE trial. Indeed, the HR for the composite pri- mary outcome in CURE (HR: 0.80) was similar to the HR for the composite outcome of death or MI in our study (HR: 0.74). In CURE, as well as in CRUSADE, early adminis- tration of clopidogrel was associated with a signicant improvement of in-hospital and 1-year outcomes (1,18). Our study demonstrates that long-term outcomes are also improved for survivors of coronary events requiring hospital stay and further supports the use of clopidogrel in medically managed patients with acute coronary syndrome (ACS). Although signicant advances in antiplatelet therapy have occurred, the use of these agents would not have affected patient selection in our study because these agents were approved for use after our study period. Interestingly, clopidogrel use did not reduce the risk of MI requiring hospital stay in this study cohort. These results differ from those of the CURE trial (1), in which clopidogrel use was associated with a reduction in the composite endpoint of CV death, nonfatal MI, and stroke, but among individual endpoints, clopidogrel use reduced the risk of MI (notably, only Q-wave MIs), but not CV mortality. These differences in individual endpoints may be chance ndings, may represent secular trends in outcomes for patients with UA or NSTEMI unrelated to antiplatelet therapy, or may be the result of important differences in study design between the CURE trial and our study. First, ours was a retrospective study of administrative and clinical data from a real-world cohort, not a prospective randomized controlled trial. Real-world observational data can be analyzed reliably by applying design and analytic techniques such as new-user design, restriction of the pop- ulation to trial-eligible patients, propensity score matching, and multivariable adjustment; nevertheless, these techniques may not entirely eliminate selection bias. However, a key strength of observational research is its ability to examine more diverse populations and to generalize trial ndings to real-world patients and practice settings. Another important distinction between our study and CURE is the denition of study endpoints. We do not have data on the cause of death (e.g., CV mortality), and the MI endpoint in our study was a mixture of fatal and nonfatal events. Patients who died out of hospital of an MI were coded as a death, not an MI, in our data. Thus, a strict comparison of the MI endpoint between CURE and our study is not feasible. Patients taking clopidogrel in our study may have been less likely to suffer fatal, out-of-hospital MIs, thus leading to fewer deaths among users while increasing the number of patients who survived to be hospitalized for Figure 3 Multivariable Cox Regression Results Among Propensity ScoreMatched Cohort for All-Cause Mortality, Acute MI, and the Composite Endpoint of Death or MI Multivariable Cox regression results among propensity scorematched cohort for all-cause mortality, acute myocardial infarction (MI), and the composite endpoint of death or MI among all patients and among patients presenting with either un- stable angina (UA) or nonST-segment elevation myocardial infarction (NSTEMI). CI condence interval; HR hazard ratio. Figure 4 Multivariable Cox Regression Results Among Propensity ScoreMatched Cohort for All-Cause Mortality and the Composite Endpoint of Death or MI Among Patient Subgroups Multivariable Cox regression results among propensity scorematched cohort for all-cause mortality and the composite endpoint of death or myocardial infarction (MI) among patient subgroups including diabetic patients, older adults (age >70 years), and smokers. CI condence interval; HR hazard ratio. Solomon et al. JACC Vol. 63, No. 21, 2014 Clopidogrel in Unstable Angina or NSTEMI June 3, 2014:224957 2254 an MI. The CURE data support this potential explanation. In CURE, clopidogrel decreased the number of Q-wave MIs, which may be more likely to cause sudden cardiac death than nonQ-wave MIs. It is possible that clopidogrel reduces the severity of an MI, such that otherwise fatal, out-of-hospital MIs become less severe, nonfatal MIs that allow more clopidogrel users to make it to the hospital for treatment. Given the difculty of identifying the cause of death in observational data, the composite outcome of death and MI requiring hospital stay is the most comparable endpoint from our study. Interestingly, in a substudy of the PLATO (Platelet Inhibition and Patient Outcomes) trial of more than 5,000 medically managed patients, ticagrelor reduced mortality compared with clopidogrel, but it did not reduce the rate of MI (29). Future research should examine the mechanism driving mortality reductions that are disproportionate to reductions in MI for patients receiving these agents. Although we did not examine stroke outcomes, in the CURE trial and other trials of antiplatelet and anticoagulant use in ACS and high-risk groups (3033), strokes were uncommon, and stroke rates were not different among treatment groups. The CAPRIE (Clopidogrel Versus Aspirin in Patients at Risk of Ischaemic Events) trial found no difference in CV events among more than 12,000 pa- tients with a previous history of stroke. Another difference between this study and CURE was that we analyzed a longer follow-up period (2 years vs. 1 year). We excluded patients who received revascularization either in the hospital or within the rst week post-discharge, whereas 15% of patients in the CURE trial received in- hospital revascularization. Finally, we excluded patients with a previous history of coronary artery disease (CAD), whereas one-third of patients in the CURE trial had a pre- vious MI, and nearly 20% had previously undergone CABG or PCI. The combination of these differences may possibly account for the disparities seen in the individual endpoints. An important nding from this study was that a patients clinical presentation affected the effectiveness of clopidogrel. Clopidogrel was associated with a signicant improvement in outcomes only among patients with NSTEMI, but not among patients presenting with UA. Patients with NSTEMI had elevated cardiac enzymes, an objective marker of myocardial damage, whereas patients with UA were identied by using more subjective signs and symptoms. Therefore, some patients with a discharge diagnosis of UA in our study may not have had underlying CAD and would have been unlikely to benet from clopidogrel. In addition, although both NSTEMI and UA are usually caused by atherosclerotic plaque rupture and superimposed throm- bosis, long-term antiplatelet therapy may be more effective in patients with more severe ischemia that ultimately leads to infarction, as occurs in NSTEMI but not UA. Patients with NSTEMI and UA may also differ in other factors that in- uence the risk of MI or death, such as thrombogenicity, which could affect the response to clopidogrel. Our results suggesting that patients with NSTEMI may receive a greater benet from clopidogrel than patients with UA are consistent with a subanalysis of CURE that demonstrated a greater absolute risk reduction among high- risk subjects (Thrombolysis In Myocardial Infarction risk score of 5 to 7) (34). Similarly, in the secondary prevention subgroup analysis of the CHARISMA (Clopidogrel for High Atherothrombotic Risk and Ischemic Stabilization, Management, and Avoidance) trial (35), patients with prior MI had a larger protective effect from clopidogrel than did patients with ischemic stroke or peripheral vascular disease, with effect sizes similar to those in our study (HR: 0.74 [95% CI: 0.66 to 0.84] for death and MI in our study; HR: 0.83 [95% CI: 0.72 to 0.96] for CV death, MI, or stroke in CHARISMA). Further, in CHARISMA, there was no benet from clopidogrel among patients with CAD but without MI (HR: 1.10 [95% CI: 0.77 to 1.58]), similar to our nding for patients with UA (HR: 1.25 [95% CI: 0.94 to 1.67]). In addition, the ability to discern MI in clinical trials is likely better than in regular clinical practice because rigorous protocols and additional scrutiny are in place to monitor study patients, and the coronary events identied through our claims-based algorithms may represent clinically larger MIs. Unlike results from a Danish registry (36), we found similar effectiveness for clopidogrel among patients with and without diabetes, but we did nd a similar direction of the effect (i.e., trend toward less effectiveness among patients with diabetes) (Fig. 4). We also found a stronger association for clopidogrel with outcomes among nonsmokers than among smokers. These results differ from those of CAPRIE (37) and other trials, although this discrepancy may signify differences in characterization of smoking status in routine care from those in a clinical trial. Our data suggesting that clopidogrel may be more protective among older patients (age 70 years), at least for our composite outcome, is consistent with ndings of the CURE trial and may reect greater benet of clopidogrel among higher-risk patients. This nding is important because concern over the risk of bleeding from dual-antiplatelet therapy can often tilt the perceived benet/risk ratio against aggressive treatment in older populations. In TRILOGY ACS (Targeted Platelet Inhibition to Clarify the Optimal Strategy to Medically Manage Acute Coronary Syndromes), a dedicated trial of prasugrel versus clopidogrel in medically managed patients, older patients did not receive a benet from more potent platelet inhibition with prasugrel (33). Less than one-half (36%) of medically managed patients with UA or NSTEMI were prescribed clopidogrel during the study period (2003 to 2008), even though it was given a Class I indication in the 2002 AHA/ACC guidelines (2). In similar patient populations, the CRUSADE (Can Rapid Risk Stratication of Unstable Angina Patients Suppress Adverse Outcomes With Early Implementation of the ACC/AHA Guidelines) registry reported clopidogrel adherence rates of 28% in 2002 and 53% in 2005 (13), the JACC Vol. 63, No. 21, 2014 Solomon et al. June 3, 2014:224957 Clopidogrel in Unstable Angina or NSTEMI 2255 GRACE registry (Global Registry of Acute Coronary Events) reported almost 50% usage of clopidogrel in Ca- nadian patients between 2003 and 2007 (17), and in the ACTION Registry-GWTG (Get With the Guidelines), 55% of medically managed patients with NSTEMI be- tween 2009 and 2011 were discharged on clopidogrel (15). Both GRACE and CRUSADE demonstrated increasing use of clopidogrel for UA or NSTEMI over time (18,19). Although the reported use of clopidogrel was slightly lower in this study, given the wide availability of PCI for Kaiser Permanente Northern California members, the population of medically managed patients with ACS in our cohort may be sicker than the U.S. average, with more clinical con- traindications to clopidogrel use. Nevertheless, this nding lends additional support to more rigorous use of clopidogrel use among medically managed patients with ACS. Indeed, the results of the medically managed arm of PLATO, in which ticagrelor was superior to clopidogrel on the com- posite outcome of death, MI, and stroke, suggest that potent platelet inhibition is indeed benecial in this patient population (29). Despite recent advances in antiplatelet therapy, clopidogrel remains a viable option; TRILOGY ACS found no benet for prasugrel over clopidogrel in a similar patient population. Study limitations. Our analysis had limitations. It was not possible to conrm the use of aspirin therapy because most aspirin use is over the counter and is not captured within our pharmacy dispensing database. We could not ascertain the cause of death, details of the coronary event, the extent and severity of CAD (i.e., single-vessel or multivessel disease), or heart failure severity. Even though we used advanced sta- tistical methods, including propensity score matching and covariate adjustment, we cannot rule out residual or un- measured confounding. In addition, although our study provides insights into clopidogrel use over the past decade, changes in guideline recommendations over this period may have altered selection of patients for invasive versus con- servative strategies. Conclusions Examining the strength and durability of clinical trial ndings is a core mission of comparative effectiveness research. Our results show that clopidogrel use initiated within 7 days post-discharge was associated with improved outcomes among medically managed patients after ACS. These ndings suggest that the results of antiplatelet trials for this patient population translate well to real world practice. Reprint requests and correspondence: Dr. Matthew D. Solomon, Kaiser Permanente Oakland Medical Center, 3600 Broadway, Oakland, California 94611. E-mail: Matthew.D.Solomon@kp. org. REFERENCES 1. Yusuf S, Zhao F, Mehta SR, Chrolavicius S, Tognoni G, Fox KK. Effects of clopidogrel in addition to aspirin in patients with acute coronary syndromes without ST-segment elevation. N Engl J Med 2001;345:494502. 2. Braunwald E, Antman EM, Beasley JW, et al. ACC/AHA 2002 guideline update for the management of patients with unstable angina and nonST-segment elevation myocardial infarctiondsummary article: 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 2002;40:136674. 3. Mehta SR, Yusuf S, Peters RJ, et al. Effects of pretreatment with clopidogrel and aspirin followed by long-term therapy in patients un- dergoing percutaneous coronary intervention: the PCI-CURE study. Lancet 2001;358:52733. 4. Steinhubl SR, Berger PB, Mann JT 3rd, et al. Early and sustained dual oral antiplatelet therapy following percutaneous coronary intervention: a randomized controlled trial. JAMA 2002;288:241120. 5. Patti G, Colonna G, Pasceri V, Pepe LL, Montinaro A, Di Sciascio G. Randomized trial of high loading dose of clopidogrel for reduction of periprocedural myocardial infarction in patients undergoing coronary intervention: results from the ARMYDA-2 (Antiplatelet therapy for Reduction of MYocardial Damage during Angioplasty) study. Circu- lation 2005;111:2099106. 6. von Beckerath N, Taubert D, Pogatsa-Murray G, Schomig E, Kastrati A, Schomig A. Absorption, metabolization, and antiplatelet effects of 300-, 600-, and 900-mg loading doses of clopidogrel: results of the ISAR-CHOICE (Intracoronary Stenting and Antithrombotic Regimen: Choose Between 3 High Oral Doses for Immediate Clopi- dogrel Effect). Trial. Circulation 2005;112:294650. 7. Montalescot G, Sideris G, Meuleman C, et al. A randomized com- parison of high clopidogrel loading doses in patients with non ST-segment elevation acute coronary syndromes: the ALBION (Assessment of the Best Loading Dose of Clopidogrel to Blunt Platelet Activation, Inammation and Ongoing Necrosis) trial. J Am Coll Cardiol 2006;48:9318. 8. LAllier PL, Ducrocq G, Pranno N, et al. Clopidogrel 600-mg double loading dose achieves stronger platelet inhibition than conventional regimens: results from the PREPAIR randomized study. J Am Coll Cardiol 2008;51:106672. 9. Schiele F, Meneveau N, Bassand JP. Routine pre-treatment with clo- pidogrel before diagnostic coronary angiography: the question is right, but what about the answer? Eur Heart J 2008;29:14757. 10. Di Sciascio G, Patti G, Pasceri V, et al. Effectiveness of in-laboratory high-dose clopidogrel loading versus routine pre-load in patients undergoing percutaneous coronary intervention: results of the ARMYDA-5 PRELOAD (Antiplatelet therapy for Reduction of MYocardial Damage during Angioplasty) randomized trial. J Am Coll Cardiol 2010;56:5507. 11. Ho PM, Tsai TT, Maddox TM, et al. Delays in lling clopidogrel prescription after hospital discharge and adverse outcomes after drug- eluting stent implantation: implications for transitions of care. Circ Cardiovasc Qual Outcomes 2010;3:2616. 12. Bertrand ME, Rupprecht HJ, Urban P, Gershlick AH, for the CLASSICS Investigators. Double-blind study of the safety of clopi- dogrel with and without a loading dose in combination with aspirin compared with ticlopidine in combination with aspirin after coronary stenting: the Clopidogrel Aspirin Stent International Cooperative Study (CLASSICS). Circulation 2000;102:6249. 13. Amsterdam EA, Peterson ED, Ou FS, et al. Comparative trends in guidelines adherence among patients with nonST-segment elevation acute coronary syndromes treated with invasive versus conservative management strategies: results from the CRUSADE quality improvement initiative. Am Heart J 2009;158:74854. 14. Van de Werf F, Gore JM, Avezum A, et al. Access to catheterisation facilities in patients admitted with acute coronary syndrome: multina- tional registry study. BMJ 2005;330:441. 15. Maddox TM, Ho PM, Tsai TT, et al. Clopidogrel use and hospital quality in medically managed patients with nonST-segment-elevation myocardial infarction. Circ Cardiovasc Qual Outcomes 2012;5: 52331. Solomon et al. JACC Vol. 63, No. 21, 2014 Clopidogrel in Unstable Angina or NSTEMI June 3, 2014:224957 2256 16. de Winter RJ, Windhausen F, Cornel JH, et al. Early invasive versus selectively invasive management for acute coronary syndromes. N Engl J Med 2005;353:1095104. 17. Banihashemi B, Goodman SG, Yan RT, et al. Underutilization of clopidogrel and glycoprotein IIb/IIIa inhibitors in nonST-elevation acute coronary syndrome patients: the Canadian global registry of acute coronary events (GRACE) experience. Am Heart J 2009;158:91724. 18. Mehta RH, Roe MT, Chen AY, et al. Recent trends in the care of pa- tients with nonST-segment elevation acute coronary syndromes: insights from the CRUSADE initiative. Arch Intern Med 2006;166:202734. 19. Rao RV, Goodman SG, Yan RT, et al. Temporal trends and patterns of early clopidogrel use across the spectrum of acute coronary syn- dromes. Am Heart J 2009;157:64250. 20. Ray WA. Evaluating medication effects outside of clinical trials: new- user designs. Am J Epidemiol 2003;158:91520. 21. Go AS, Chertow GM, Fan D, McCulloch CE, Hsu CY. Chronic kidney disease and the risks of death, cardiovascular events, and hospital stay. N Engl J Med 2004;351:1296305. 22. Go AS, Lee WY, Yang J, Lo JC, Gurwitz JH. Statin therapy and risks for death and hospital stay in chronic heart failure. JAMA 2006;296:210511. 23. Selby JV, Ray GT, Zhang D, Colby CJ. Excess costs of medical care for patients with diabetes in a managed care population. Diabetes Care 1997;20:1396402. 24. Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol In Adults. Executive summary of the third report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol In Adults (Adult Treatment Panel III). JAMA 2001;285:248697. 25. Levey AS, Coresh J, Greene T, et al. Using standardized serum creatinine values in the modication of diet in renal disease study equation for estimating glomerular ltration rate. Ann Intern Med 2006;145:24754. 26. National Kidney Foundation. K/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classication, and stratication. Am J Kidney Dis 2002;39 Suppl 1:S1266. 27. Fireman BH, Fehrenbacher L, Gruskin EP, Ray GT. Cost of care for patients in cancer clinical trials. J Natl Cancer Inst 2000;92:13642. 28. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika 1983;70:4155. 29. James SK, Roe MT, Cannon CP, et al. Ticagrelor versus clopidogrel in patients with acute coronary syndromes intended for non-invasive management: substudy from prospective randomised PLATelet inhi- bition and patient Outcomes (PLATO) trial. BMJ 2011;342:d3527. 30. Wiviott SD, Braunwald E, McCabe CH, et al. Prasugrel versus clo- pidogrel in patients with acute coronary syndromes. N Engl J Med 2007;357:200115. 31. CAPRIE Steering Committee. A randomised, blinded, trial of clopi- dogrel versus aspirin in patients at risk of ischaemic events (CAPRIE). CAPRIE Steering Committee. Lancet 1996;348:132939. 32. Mega JL, Braunwald E, Wiviott SD, et al. Rivaroxaban in patients with a recent acute coronary syndrome. N Engl J Med 2012;366:919. 33. Roe MT, Armstrong PW, Fox KA, et al. Prasugrel versus clopidogrel for acute coronary syndromes without revascularization. N Engl J Med 2012;367:1297309. 34. Budaj A, Yusuf S, Mehta SR, et al. Benet of clopidogrel in patients with acute coronary syndromes without ST-segment elevation in various risk groups. Circulation 2002;106:16226. 35. Bhatt DL, Flather MD, Hacke W, et al. Patients with prior myocardial infarction, stroke, or symptomatic peripheral arterial disease in the CHARISMA trial. J Am Coll Cardiol 2007;49:19828. 36. Andersson C, Lyngbaek S, Nguyen CD, et al. Association of clopi- dogrel treatment with risk of mortality and cardiovascular events following myocardial infarction in patients with and without diabetes. JAMA 2012;308:8829. 37. Gurbel PA, Nolin TD, Tantry US. Clopidogrel efcacy and cigarette smoking status. JAMA 2012;307:24956. Key Words: acute coronary syndrome(s) - clopidogrel - outcomes. JACC Vol. 63, No. 21, 2014 Solomon et al. June 3, 2014:224957 Clopidogrel in Unstable Angina or NSTEMI 2257 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.