Thomson

Stroke and the statistics of the aspirin/clopidogrel secondaryprevention trialsGeorge Howard, Leslie A. McClure, John W. Krakauer and Christopher S. Coffey Four randomized trials have investigated the combination of Clopidogrel versus Aspirin in Patients at Risk of Ischemic Events clopidogrel plus aspirin for secondary prevention of Clopidogrel for High Atherothrombotic Risk and Ischemic vascular outcomes in 54 949 patients. Here we argue that Clopidogrel in Unstable Angina to Prevent Recurrent Events attempts to translate the results of these trials into clinical Management of Atherothombosis with Clopidogrel in High-risk practice have proven frustrating because of the following statistical considerations: differences in study populations and study design make comparisons difficult (comparisonsof ‘apples and oranges’), incomplete factorial designsprevent proper contrasts (examining ‘bits and pieces’ of a ß 2007 Lippincott Williams & Wilkins1350-7540 larger picture), results concern widely different vasculardiseases (‘puzzling subgroups’), and negative results areeasily misinterpreted.
Recent findings Between 1996 and 2004 three major randomized trials The relative efficacy of different combinations of aspirin assessed combinations of aspirin and clopidogrel, finding: and clopidogrel to prevent cardiovascular events have Clopidogrel versus Aspirin in Patients at Risk of Ischemic been contrasted in several large randomized clinical Events (CAPRIE) in favor of clopidogrel alone versus aspirin trials including Clopidogrel versus Aspirin in Patients alone, Clopidogrel in Unstable Angina to Prevent Recurrent at Risk of Ischemic Events (CAPRIE) , Clopidogrel in Events (CURE) in favor of clopidogrel plus aspirin versus Unstable Angina to Prevent Recurrent Events (CURE) aspirin alone, and Management of Atherothombosis with and Management of Atherothombosis with Clopido- Clopidogrel in High-risk Patients (MATCH) in favor of grel in High-risk Patients (MATCH) Together these clopidogrel plus aspirin versus clopidogrel alone. A recently trials have randomized 39 346 patients, but as yet the completed fourth trial (Clopidogrel for High appropriate treatment for primary and secondary pre- Atherothrombotic Risk and Ischemic Stabilization, vention of stroke remains a matter of debate While Management and Avoidance; CHARISMA) was a ‘negative individually each of these studies was conducted with study’ comparing aspirin alone to aspirin plus clopidogrel.
the highest level of quality, they have failed to provide conclusive guidance for secondary prevention because Even after four large randomized trials we still do not know they provide only comparisons between ‘apples and the optimal treatment for secondary prevention of stroke.
oranges’, include only ‘bits and pieces’ of the infor- We suggest that subsequent trials should focus on a mation needed, include puzzling subgroups, and are particular vascular disease and test hypotheses that relate subject to misinterpretation in the case of a negative study. The surrounding confusion is a product ofthese shortcomings in information. We will first review the three studies CAPRIE, CURE and MATCH. We then see if a more recent fourth study clarified concernsand uncertainties about the first three.
Curr Opin Neurol 20:71–77. ß 2007 Lippincott Williams & Wilkins.
A brief overview of the first three studies is necessary to Department of Biostatistics, University of Alabama at Birmingham, Birmingham, appreciate the confusion resulting from their interpre- Correspondence to George Howard, Dr PH, Professor and Chair, Department ofBiostatistics, UAB School of Public Health, Birmingham, AL 35294-0022, USAE-mail: CAPRIE randomized 19 185 symptomatic patients toeither clopidogrel (75 mg) or aspirin (325 mg). By design, Current Opinion in Neurology 2007, 20:71–77 one-third of the patients had a previous stroke, one thirdhad a previous myocardial infarction (MI) and one-thirdhad peripheral vascular disease. Patients were followed forbetween one and 3 years (median follow-up of 1.91 years).
The study showed an 8.7% (95% CI 0.3%–16.5%; Copyright Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
P ¼ 0.043) reduction in the primary endpoint of ischemic (clopidogrel plus aspirin) as compared to the aspirin stroke, MI, or vascular death in favor of clopidogrel.
CURE randomized 12 652 patients with unstable angina Both of these studies, however, used a composite end- in the past 24 h to either aspirin alone (75 mg to 325 mg) point including both stroke and myocardial infarction.
versus aspirin plus clopidogrel (loading dose of 300 mg Also, as seen in the magnitude of the effect in orally followed by 75 mg). The study followed patients both CAPRIE and CURE was substantially larger for the for between 3 and 12 months and showed a 20% (95% CI prevention of MI (effect 0.82 and 0.77 respectively) than 10–28%) reduction in the composite outcome of death for stroke (0.95 and 0.86 respectively). As such, much of from cardiovascular causes, nonfatal myocardial infarc- the overall significant effect in CAPRIE and CURE, both tion, or stroke in favor of clopidogrel plus aspirin.
containing populations at high risk for coronary disease, isattributable to treatment effects on coronary disease, with MATCH randomized 7599 stroke or tranisent ischemic only a minor effect on stroke endpoints.
attack (TIA) patients who had at least one additionalcardiovascular risk factor to receive clopidogrel alone In contrast, all patients enrolled in MATCH had a history (75 mg) versus aspirin (75 mg) plus clopidogrel (75 mg).
of stroke or TIA but only 5% of their population had a The study followed patients for 18 months and showed a history of prior MI. Hence, patients in MATCH were 6.4% (À4.6% to 16.3%) reduction on a composite end- likely to have a higher propensity towards stroke and a point of ischemic stroke, myocardial infarction, vascular lower propensity for coronary events. As can be seen in death, or rehospitalization for acute ischemia (including combination treatment (aspirin plus clopidogrel) rehospitalization for transient ischemic attack, angina relative to colpidogrel alone, as tested in MATCH, pectoris, or worsening of peripheral arterial disease) in provided approximately as much benefit in preventing favor of the combined clopidogrel plus aspirin treatment.
strokes as the addition of aspirin to clopidogrel in CUREor the difference between aspirin and clopidogrel in CAPRIE (0.93 compared to 0.86 and 0.95 respectively).
CAPRIE and CURE both showed significant treatment The benefit of combination therapy in the prevention of differences on the composite endpoint, in favor of clo- MIs in MATCH, however, was substantially less (0.95) pidogrel and clopidogrel plus aspirin, respectively, while than the effects observed in CAPRIE and CURE (odds MATCH failed to reach the traditional level of signifi- ratios of 0.82 and 0.77 respectively). As such, in the cance. The significance observed in CAPRIE and CURE MATCH population at higher risk for stroke but lower could be attributed to a true difference between the risk for coronary events, a similar treatment effect on treatments considered, while the lack of a significant stroke was observed, but the substantial effect on coro- difference in MATCH could be due to fundamental nary events that contributed a major effect in CAPRIE For example, both CAPRIE and CURE were conducted The significant effects observed in CURE and CAPRIE in populations that contained a substantial portion of but not in MATCH could be attributable to true diffe- patients with likely coronary heart disease: in CAPRIE, rences in treatment effects between the assigned treat- one-third of the patient population had previous MI, and ments. It is also possible, however, that differences in the CURE randomized patients with recent unstable angina.
design of the study, including the eligibility criteria, As can be seen in in CAPRIE the crude odds of could contribute to the differences in outcome. Specifi- the primary endpoint was only 0.91 (95% CI 0.84–0.997) cally, could the observed effects in CAPRIE and CURE times as great in the group assigned to clopidogrel as and the lack of a significant effect in MATCH be attribu- compared to those assigned to aspirin. Likewise in table to differences in both the study populations and CURE, the hazard was only 0.80 (95% CI: 0.72–0.90) the endpoints? Likewise, could the inconsistencies be times as great in the group with combination treatment attributable to other differences, such as the difference in Table 1 Crude results from CAPRIE, CURE and MATCH, both for the overall and by components of the composite endpoint CAPRIE, Clopidogrel versus Aspirin in Patients at Risk of Ischemic Events; CURE, Clopidogrel in Unstable Angina to Prevent Recurrent Events;MATCH, Management of Atherothombosis with Clopidogrel in High-risk Patients.
Copyright Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
The clopidogrel studies: statistics Howard et al.
Figure 1 Design of the European Stroke Prevention Study 2 compared with existing studies (a) The two-by-two factorial designof the European Stroke Prevention Study 2 (ESPS2) study allowingfor assessment of ‘main effects’ for clopidogrel and aspirin fromexisting studies.
the dose of aspirin or clopidogrel used? Frankly, who introduces similar problems, may make this point clear.
knows: making comparisons between apples and oranges On November 19th, 2004, the University of North Carolina Tar Heel basketball team was ranked numbertwo in the country, when they lost to Santa Clara University (Santa Clara 77, UNC 66). This established The European Stroke Prevention Study 2 (ESPS2) used a Santa Clara as a better team than UNC. Subsequently, on two-by-two factorial design to assess the efficacy of December 29th, Yale defeated Santa Clara 90 to 84, aspirin and dipyridamole individually or in combination establishing Yale as superior to Santa Clara. Since: This powerful design allows for the assess- UNC is the number two team in the nation, and Yale ment of the following: the effect of aspirin, a comparison is superior to Santa Clara, who is superior to UNC, then of the pooled data from those receiving neither aspirin Yale is better than the second best team in the nation, nor dipyridamole plus those on dipyridamole only com- thus establishing Yale as the number one team in the pared to the pooled data of those receiving aspirin only nation. In addition, New Mexico, Pacific, Cal Poly, UC and those receiving both aspirin and dipyridamole; the Irvine, Arizona State, Central Connecticut State, Brig- effect of dipyridamole, a comparison of the pooled data ham Young, Gonzaga, and San Francisco had also all from those receiving neither aspirin nor dipyridamole and defeated Santa Clara during the year, hence they should those receiving aspirin only versus the pooled data of also have been the number one team in the nation. This those receiving dipyridamole only and those receiving ludicrous example of logic is faulty because of the both aspirin and dipyridamole; and the potential of a uncertainty involved in sporting events: a victory on synergistic (or nonadditive) effect of aspirin and dipyri- one night does not establish a team as ‘globally’ superior damole, i.e. if the efficacy gains among those receiving to another. In much the same way, however, there is both aspirin and dipyridamole represents the sum of the uncertainty in research studies, and differences between gains from taking aspirin and the gains from taking treatments are also estimated with error. Few people are comfortable drawing transitive relationships in sports (goYale!); perhaps we should be equally cautious drawing In contrast, shows the comparisons that can be such transitive conclusions from bits and pieces of incom- made on the basis of studies conducted on aspirin and plete study designs (as we have in Unfortu- clopidogrel. Interestingly, each of these major studies nately, the illogic of using the transitive property to shares one treatment with each of the other two major compare studies means that physicians are left in a studies. Thus, instead of having the complete factorial situation where large studies have been performed but design we have only bits and pieces from which to draw their results cannot be combined in a coherent way.
The transitive property in mathematics states that if A is As the largest study and the study providing a direct greater than B, and if B is greater than C, then A is greater comparison of monotherapy with aspirin versus clopido- than C. It is tempting to use this type of logic to make grel, an understanding of the results of CAPRIE is central decisions as to which of the three alternative active to the decision of prescribing treatments to prevent treatments (aspirin, clopidogrel or aspirin plus clopido- stroke. Overall, clopidogrel was superior to aspirin in grel) is superior. The uncertainty associated with the CAPRIE, showing an 8.7% (95% CI 0.3–16.5%) benefit comparisons within each of the studies makes this logic problematic. A sports analogy, where uncertainty traditional level of significance (P ¼ 0.043). As discussed Copyright Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Figure 2 Estimated composite treatment effect in Clopidogrel versus Aspirin in Patients at Risk of Ischemic Events (CAPRIE) Estimated composite treatmenteffect (estimate and 95% CI) in (CAPRIE), with treatment effectsabove zero showing benefit for clopidogrel and effects below zerobenefit for aspirin. Effects are shown for the overall study findingsand by study subpopulations. MI, myocardial infarction; PVD,peripheral vascular disease.
above, patients in CAPRIE were drawn from three A final point should be made here about MATCH. In this patient sub-populations in approximately equal repre- study, approximately 70% of the patients had diabetes sentation: those having had a stroke, MI or peripheral and 54% of the recurrent strokes were small vessel in vascular disease (PVD). The interpretation of CAPRIE, nature. Thus, it is likely that a very different mechanism however, is substantially complicated by the presence of is in play when compared to coronary artery disease. Thus a significant interaction within the study population we do not know if the outcome would have been different (P ¼ 0.042) indicating that it is likely that the relative if only patients with stroke from large artery disease had benefit of clopidogrel differs depending on the study sub-population As is evident in the overalltreatment effect is largely affected by a 23.8% benefit in Interpretation of ‘negative’ studies? the clopidogrel group for the PVD subpopulation. The A study is considered ‘negative’ when it fails to reject the large effect among PVD patients was substantially null hypothesis of no difference between treatment dampened by a 3.7% benefit for aspirin in the MI groups. Frequently, results from negative studies (such subpopulation (i.e. nonsignificantly and in the opposite as MATCH) are described thus: ‘there are no differences direction). The 7.3% effect in favor of clopidogrel in between treatment groups’. This interpretation is simply the stroke subpopulation was intermediate and similar incorrect. There are two major reasons why a trial may fail to the overall effect; however, this difference did not to show a significant difference: the difference is in fact reach a level of statistical significance (P ¼ 0.26).
small or does not exist, or the study was not adequatelypowered to document differences that truly do exist. As When faced with the decision of whether to prescribe such, rather than ‘there are no differences between clopidogrel to a stroke patient, the information from treatment groups’, the correct interpretation is that there CAPRIE is confusing at best. It is not surprising that is no evidence in these data that indicates which treat- statistical significance was not achieved in the stroke ment is superior: a substantially different statement. As subpopulation since the study was not powered to an example, consider the statement ‘all swans are white’.
achieve significance in the subpopulations. Had the To examine this statement, a sample of swans is drawn.
effects across the subpopulations been consistent, then There are two possible outcomes: all the swans in the it would be reasonable to proceed with treatment of the sample are white; at least one swan in the sample is not stroke patient, based on the results from the CAPRIE white. The latter outcome establishes the falsehood of study. The evidence of a substantial difference between the statement (i.e., rejects the null hypothesis that all the patient subpopulations implies that a general ‘aver- swans are white). The first outcome, however, does not age’ effect across the patient population as a whole is not prove the statement since, had we taken a different a reasonable interpretation of the results. The lack of a sample, we may have found a non-white swan. Obviously, significant association in stroke patients raises the ques- although the statement can never be proven, a larger tion of whether they are more similar to MI patients (for sample of swans provides more definitive evidence that whom no effect was observed), more similar to PVD the probability that the statement is false is likely to be patients (for whom a substantial effect was observed), or intermediate between the two. When treating a strokepatient, this presents the clinician with the dilemma of Hypothesis testing in clinical trials follows very similar interpreting the results based on confusing subgroups.
principles. In hypothesis testing, there are two types of Copyright Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
The clopidogrel studies: statistics Howard et al.
errors that can be made. A type I error is made if we So, does MATCH confirm or refute an effect of combi- incorrectly reject the null hypothesis. In the swan analogy nation therapy? The observed effect was a 6.4% (95% CI this would be equivalent to saying that all swans are À4.6% to 16.3%) event reduction for combination white, when in fact they are not (perhaps we had an therapy as compared to monotherapy with clopidogrel.
unusual sample with a ‘dirty’ swan). Similarly, a type II The old way of thinking is that since this does not differ error is made if we fail to reject the null hypothesis when significantly from a 0% difference, this is a negative study it is false. In the swan analogy, this would be equivalent to and thus does not support combination therapy. A new having a sample of all white swans and failing to reject the way of thinking is (hopefully) emerging, however, by assumption that all are white. Statistical power, which is which we recognize that the information available from defined as one (1.0) minus the probability of a type II the study resides in the estimated difference and its confi- error, is therefore defined as the probability that the study dence limits. Taking this approach combination therapy will reject the null hypothesis of no treatment difference showed a trend in favor of being superior; it is true that it is if there truly is a difference. Thus, failing to reject not significantly different from a 0% effect, and so we hypotheses for sufficiently powered tests provides more should interpret this with caution; but it is also not sig- definitive evidence that the likelihood that the null nificantly different than the a priori difference of 14%, hypothesis is false is very small. Hence, in interpreting which was selected as being a clinically meaningful differ- the evidence from a negative study, one should examine ence. That is, while we cannot say that there was clearly a whether the study was adequately powered to detect treatment difference, we also cannot say that the observed difference was less than the a priori ‘clinically meaningful’difference that the study was powered to detect.
So, what difference was the MATCH study designed todetect, and to what degree was the study powered to How should we interpret a negative study such as detect this difference? The MATCH study was designed MATCH? It is clear that we should not say that based to detect a 14% treatment difference, with 80% power. In on this study, there no is evidence of difference between MATCH, this means that even if a 14% treatment treatments. It is also clear that we cannot say that this difference did exist there was a 20% chance that the establishes combination therapy as the better treatment.
study would fail to reject the null hypothesis of no If one had to comment in this uncertain situation, it is difference. Just as 5% is the standard for rejecting the reasonable to say that our best guess is that combination null hypothesis (i.e., P < 0.05), studies are frequently therapy results in a 6.4% reduction in risk; however, this designed with either 80% or 90% power to detect a is an uncertain guess and it is not unreasonable to expect treatment difference. Interestingly, 5%, 80%, and 90% to see as much as a 4.6% harm, or as much as a 16.3% are simply all arbitrary numbers: why do we require benefit. In summary, we should interpret these findings P < 0.05 to reject a hypothesis, yet we are comfortable We should all reflect on the meaning of a negative study with 80% power. First, the MATCH trial did nothing One could only hope that adding new information from a incorrect: 80% power is an ‘industry standard’ number more recent study would clarify our understanding of the (much like P < 0.05). It is surprising (at least to this clinical usefulness of clopidogrel. The recently reported author), however, that as a scientific community, in partnership with industry, we are comfortable commit- Ischemic Stabilization, Management and Avoidance ting the financial resources required to mount a trial of the (CHARISMA) trial however, has, if anything, done scope of MATCH and exposing patients to randomly the opposite. CHARISMA recruited 15 603 patients with assigned treatments under a situation where there is a one clinically evident cardiovascular disease (‘symptomatic’ in five chance of not detecting a treatment effect even if patients). Like the CURE study, CHARISMA rando-mized patients to either aspirin/clopidogrel or aspirin/ MATCH was designed to detect a 14% difference. Is placebo. While all patients recruited into CURE had this a reasonable effect? Perhaps the most reasonable unstable angina, the CHARISMA population included comparison is the observed 13% treatment effect for com- approximately 11% with ‘angina with documented multi- bined treatment (dipyridamole plus aspirin) versus single vessel coronary disease’, 26% with other evidence of treatment in ESPS2 This implies a difference of coronary disease (interventions or myocardial infarc- 14% is achievable, but it is still slightly larger than tions), 28% with stroke, 18% with peripheral vascular achieved in ESPS2. It is relatively small, however, com- disease, and 21% asymptomatic patients with multiple pared to the 20% difference between combination treat- risk factors (percents sum to more than 100% because ment and treatment with aspirin observed in CURE.
Copyright Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
It is an informative exercise to see if we could have the number of patients were to soar into six digits (not to predicted the results of CHARISMA. We would surely mention active comparison trials with alternative promi- have been heavily influenced by the CURE results, sing treatments such as dipyridomole) Thus, what do showing a striking 20% (95% CI 10–28%) benefit for we tell our patients: aspirin, clopidogrel, or both? What treatment with clopidogrel. We would also have also been does the information from randomizing 54 949 patients influenced by the CAPRIE results where there was a tell us? All it tells us is that we can be sure that we are nonsignificant harm for treatment with clopidogrel unsure of the appropriate approach. CAPRIE tells us that among patients with coronary disease, but a significant monotherapy with clopidogrel is slightly superior to benefit for patients with peripheral vascular disease and a aspirin, but not really for treating stroke patients, and nonsignificant benefit among patients with stroke.
the benefit of clopidogrel apparently is dependent on the Unlike CURE (which included only coronary patients), medical history of the patient. CURE tells us combi- CHARISMA was ‘enriched’ by patients with stroke and nation therapy is clearly superior to monotherapy with peripheral vascular disease, and as such one could specu- aspirin, but this was conducted in a population of almost late that the benefit of clopidogrel in CHARISMA would exclusively coronary patients. MATCH, a study actually actually be larger than CURE. The exact opposite performed in stroke patients, suggests a benefit for combination therapy but fails to reach a level wheredefinitive statements can be made. CHARISMA tells CHARISMA failed to find a benefit for the addition of us to avoid dual antiplatelet therapy altogether.
clopidogrel to aspirin with a relative risk of 0.93 (95% CI 0.83 – 1.05; P ¼ 0.22), basically a finding that leads tothe response ‘now what?’ This result is close to a This sorry state of affairs suggests that we need to do perfect teaching example of the dangers of generalizing something differently. Although we may not be at the findings of a study to new populations (apples and the point where we can substitute ‘one size fits all’ oranges), the problems of focusing on findings in sub- megatrials for personalized patient management based groups (confusing subgroups), attempting to make on genotype surely a more homogenous disease interpretations between different drug contrasts (bits mechanism approach should be employed. After all, in and pieces), and how to interpret a negative study. As the field of secondary stroke prevention, it has been such, rather than clarifying our understanding, it is those trials that have focused on a particular stroke hard not to conclude that CHARISMA has only mechanism, carotid stenosis and atrial fibrillation, further clouded the situation. To make matters worse, that have yielded significant and large effect sizes.
CHARISMA reported an a priori hypothesized sub- Subgroup analysis is not a good alternative. It is an group analysis, which showed a borderline significant approach, as we have seen, which is fraught with (P ¼ 0.045) interaction between treatment and sympto- difficulties and suggests differences that then go on matic status with significant protection in the sympto- to be proven wrong when tested directly Without a matic population (relative risk 0.88; 95% CI 0.77 – 0.998; change, it can be envisaged that the next trial will be P ¼ 0.046) but nonsignificant harm in the asymptomatic population (relative risk 1.2; 95% CI 0.91 – 1.59). Asacknowledged by the authors, this finding was the only one of only several subgroups where effect modification Papers of particular interest, published within the annual period of review, have was found; however, the paper suggests (by showing results for 12 subgroup analyses) that at least 12 such analyses were conducted, and in this case under the Additional references related to this topic can also be found in the Current null hypothesis that clopidogrel has no effect there is a World Literature section in this issue (p. 85).
46% (1 – (1 – 0.05)12) chance that at least one subgroup CAPRIE Steering Committee. A randomised, blinded, trial of clopidogrel versus aspirin in patients at risk of ischaemic events (CAPRIE). Lancet1996; 348:1329–1339.
would be reported. Depressingly, after a remarkable The Clopidogrel in Unstable Angina to Prevent Recurrent Event Trial Inves- 54 949 patients randomized (the 39 246 in CAPRIE, tigators. Effects of clopidogrel in addition to aspirin in patients with acute CURE and MATCH plus the 15 603 in CHARISMA), coronary syndromes without ST-segment elevation. N Engl J Med 2001;345:494–502.
the authors of CHARISMA concluded that the Diener HC, Bogousslavsky J, Brass LM, et al. Aspirin and clopidogrel finding in symptomatic patients ‘requires further compared with clopidogrel alone after recent ischaemic stroke or transient ischaemic attack in high-risk patients (MATCH): randomised, double-blind,placebo-controlled trial. Lancet 2004; 34:331–337.
Hankey GJ. Is clopidogrel the antiplatelet drug of choice for high-risk patients with stroke/TIA?: No. J Thromb Haemost 2005; 3:1137 – It is hard not to be pessimistic about whether the piece- meal approach employed to date will ever answer the Diener HC. Is clopidogrel the antiplatelet drug of choice for high-riskpatients with stroke/TIA?: Yes. J Thromb Haemost 2005; 1133– dual antiplatelet therapy question definitively, even if Copyright Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
The clopidogrel studies: statistics Howard et al.
Diener HC, Forbes CC, Sivenius J, et al. European Stroke Prevention Study 2 Halkes PH, van Gijn J, Kappelle LJ, et al. ESPRIT Study Group. Aspirin Dipyridamole and acetylsalicylic acid in the secondary prevention of stroke.
plus dipyridamole versus aspirin alone after cerebral ischaemia of arterial origin (ESPRIT): randomised controlled trial. Lancet 2006; 367:1665 – Bhatt DL, Fox KAA, Hacke W, et al. Clopidogrel and aspirin versus aspirin alone for the prevention of atherothrombotic events. N Engl J Med 2006; Maseri A. Atillio Maseri, MD, FACC, discusses the role of biological clinical research in cardiovascular disease and the move from megatrials to perso- CHARISMA recruited 15 603 patients with clinically evident cardiovascular nalized patient management. Circulation 2006; 113:29–30.
disease (‘symptomatic’ patients) or multiple risk factors (‘asymptomatic’ pa-tients), and failed to show a difference by treatment assignment with a relative 10 Rothwell PM. Treating individuals 2. Subgroup analysis in randomized con- risk of 0.93 (95% CI 0.83– 1.05; P ¼ 0.22) for aspirin plus clopidogrel versus trolled trials: importance, indications, and interpretation. Lancet 2005; Copyright Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Source: http://columbiampl.org/pdf/Papers/Howard07AspirinTrials.pdf