Methods
Data Sources and Search Strategy
Established methods were used in compliance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement in healthcare interventions. We screened Medline, Embase, the Cumulative Index to Nursing and Allied Health Literature, Scopus, Web of Science, the Cochrane Register of Controlled Clinical Trials, as well as congress proceedings from major cardiac societies, for randomised data comparing different durations of dual antiplatelet therapy. Dual antiplatelet therapy was defined as aspirin plus a P2Y12 receptor inhibitor, after percutaneous coronary intervention with implantation of a drug eluting stent. The search period took place from 1 January 2002 to 16 February 2015.
Search terms according to medical subjects headings were: "DAPT", "dual antiplatelet therapy", "clopidogrel", "Plavix", "prasugrel", "Efient", "ticagrelor", "Brilinta", "thienopyridine", "P2Y12", "shortened DAPT", "prolonged DAPT", "extended DAPT", "premature cessation", "early discontinuation", "randomised trial", and "trial". No language or publication status restriction was imposed. The most updated or inclusive data for each study were used for abstraction. In addition, landmark analysis data at 12 months were available from the original PROlonging Dual antIplatelet treatment after Grading stent-induced intimal hyperplasia studY (PRODIGY) and were therefore incorporated into the present article.
Study Design and Selection Criteria
The design of the current meta-analysis compared two strategies of dual antiplatelet therapy involving three durations after percutaneous coronary intervention with drug eluting stent implantation. The first comparison was between a short term (<12 months) and 12 month therapy, and the second between an extended duration (>12 months) and 12 month therapy. The original PRODIGY randomised controlled trial assigned patients to either six or 24 month durations. Because the randomisation process in PRODIGY began one month after the index percutaneous coronary intervention, the availability of landmark data at 12 months allowed inclusion of the study in the short term versus 12 month comparison, after censoring events that occurred after 12 months and keeping the original randomisation design. We did additional sensitivity analyses by including PRODIGY trial data in the extended duration versus 12 month comparison. The analyses included only events that occurred beyond 12 months in both study arms (postrandomisation subgroups).
The main exclusion criteria for this meta-analysis were: observational design, patients without documented coronary artery disease or patients with peripheral or cerebrovascular disease, percutaneous coronary intervention without stents or with bare metal stents only, and duration timeframes of dual antiplatelet therapy selected by the meta-analysis not reported. Two independent reviewers (VS and MK) selected the studies for inclusion and extracted the study characteristics and relevant outcomes; divergences were solved by consensus after discussion with a third reviewer (EPN). Three authors (EPN, MK, and VS) independently reassessed the trials' eligibility and ranked their risk of bias. Risk of bias was graded using the components recommended by the Cochrane Collaboration—that is, random sequence generation; allocation concealment; blinding of participants, personnel, and outcome assessors; incomplete outcome data; selective outcome reporting; and other sources of bias.
Outcome Measures
Primary clinical endpoints were cardiovascular mortality, myocardial infarction, stent thrombosis, major bleeding, and overall mortality; secondary endpoints were repeat revascularisation, and cerebrovascular accident, and the combination of cardiac and cerebrovascular accidents. We classified stent thrombosis as definite/probable, definite, late (between 30 days and one year after percutaneous coronary intervention), and very late (>one year after percutaneous coronary intervention) according to criteria from the Academic Research Consortium. For major bleeding, trial definitions were applied. Major bleeding according to TIMI criteria, and a composite endpoint of major adverse cardiac and cerebrovascular accidents were also assessed.
Statistical Analyses
Data were analysed according to the intention to treat principle. Odds ratios and 95% confidence intervals were used as summary statistics. Heterogeneity was assessed by Cochran's Q test. We also used the statistical inconsistency test (I=(Q−df)/Q×100%, where Q=χ statistic and df=its degrees of freedom) to overcome the low statistical power of Cochran's Q test. Pooled odds ratios were calculated using a fixed effect model with the Mantel-Haenszel method, because of the absence of moderate or significant inconsistency (>50%) across studies. We also did prespecified sensitivity analyses using a random effects model. Potential publication bias was examined by constructing funnel plots for the clinical outcomes in which the standard error of the log of the odds ratio was plotted against the odds ratio. The asymmetry of the plot was estimated both visually and by Harbord's regression test. Prespecified analyses assessed the effect of different durations of dual antiplatelet therapy in the following subgroups: age older than 65 years or younger than 65 years, patients with or without acute coronary syndrome, and those treated with either clopidogrel or new P2Y12 inhibitors (prasugrel and ticagrelor). P<0.05 was considered significant and reported as two sided.