Aernoud Fiolet

342 Chapter 14 query, except for the identification of trial participants and identification of drugs administration using local differences. All EHR entries were searched, with the exception of entries that were made for the purpose of the trial. Trial participants were identified by the software program based on their unique trial identification number as recorded in the EHR. Only EHR entries after trial enrolment were included. All potential endpoints yielded by the text-mining software were manually validated by two blinded investigators (PV and WD). The endpoints collected for the original trial functioned as reference standard. All potential (i.e., unadjudicated) endpoints during the trial were collected during patient-investigator interviews according to the original trial protocol. 15,16 The participant-investigator interviews took place during clinical evaluations and were scheduled at six-month intervals until the completion of the trial. All follow-up assessments were performed in person, if possible, or by telephone. The interviews were conducted following a standardized electronic case report form, with emphasis on any hospital admission in between visits. All investigator-reported endpoints from the trial were subjected to adjudication by a committee whose members were unaware of the trial-group assignments, following the definitions of an adjudication charter. To reflect actual trial practice, all analyses in the current study were performed using both the investigator-reported endpoints and adjudicated endpoints as reference. Outcome measures The primary endpoint was a composite of cardiovascular death, myocardial infarction, or ischemic stroke (major adverse cardiovascular events, or MACE). The secondary endpoint in the current analyses was a composite of cardiovascular death, myocardial infarction, ischemic stroke, or ischemia- driven coronary revascularization (the primary composite endpoint plus ischemia-driven coronary revascularization, orMACE+). Other secondary endpoints were all-cause mortality, the individual components of the composite primary and secondary composite outcome, venous-thrombotic embolism, atrial fibrillation, and diabetes mellitus. Definitions used for the outcomes were similar to those described in the endpoint adjudication charter of the original protocol. 15,16 Data analysis Accuracy of endpoint collection by EHR data retrieval was investigated by calculating sensitivity and specificity using investigator-reported endpoints and