LAUSR

Using administrative claims data to ascertain clinical events may be a pragmatic way to make randomized, controlled trials cheaper and more efficient. However, little is known about the accuracy of claims data for estimating treatment effects of randomized therapies. To evaluate this further, we compared outcomes ascertained with Medicare claims linked with adjudicated outcomes in the DAPT Study (Dual Antiplatelet Therapy) (https://www.clinicaltrials.gov; registration number NCT00977938), a randomized trial comparing 30 and 12 months of DAPT with aspirin and a P2Y12 inhibitor after percutaneous coronary intervention (PCI).1 This investigation was part of the EXTEND Study (Extending Trial-Based Evaluations of Medical Therapies Using Novel Sources of Data)2,3 and includes DAPT Study data linked to the American College of Cardiology’s NCDR (National Cardiovascular Data Registry) CathPCI Registry and Medicare fee-for-service claims. Analyses were approved by the Institutional Review Board at Beth Israel Deaconess Medical Center with waiver of informed consent. The DAPT Study enrolled participants between August 13, 2009, and July 1, 2011. At 12 months after PCI, patients without ischemic or bleeding events in the previous year were randomized to placebo (12 months of DAPT) or continued P2Y12 inhibitor for 18 months (30 months of DAPT). This analysis included all US patients in the DAPT Study who were ≥65 years of age and could be linked via the NCDR CathPCI Registry to Medicare claims using deterministic algorithms based on patient identifiers. We compared treatment effects using claims-derived versus adjudicated events for extended-duration DAPT versus standard-duration DAPT on major adverse cardiovascular and cerebrovascular events (MACCE; a composite of death, myocardial infarction [MI], or stroke), MI, and major bleeding. Events in trial data were based on adjudication by the DAPT Study clinical events committee, and claims events were based on International Classification of Diseases, Ninth Revision diagnosis codes from inpatient Medicare data. Follow-up was from randomization to 18 months after randomization (12–30 months after PCI). Absolute risk differences using Kaplan-Meier methods and hazard ratios (HRs) using Cox regression were compared with the use of claims-derived and adjudicated events, accounting for competing risk of death based on Fine-Gray methods and intraclass correlation using robust covariance estimation.4,5 Given the post hoc nature of this analysis, all findings were considered exploratory. Among 11 648 randomized patients, 1336 (11.5%) were included in this analysis. Patients outside the United States (n=1756), <65 years of age (n=5984), lacking sufficient identifying information (n=1880), or with Medicare Advantage (n=692) were excluded. In the linked cohort, 658 were randomized to 30 months of DAPT Kamil F. Faridi, MD Hector Tamez, MD, MPH Jordan B. Strom, MD, MSc Yang Song, MS Neel M. Butala, MD, MBA Changyu Shen, PhD Eric A. Secemsky, MD, MSc Laura Mauri, MD, MSc Jeptha P. Curtis, MD C. Michael Gibson, MD, MS Robert W. Yeh , MD, MSc