RESUMO
BACKGROUND: Independent adjudication of clinical trial events is traditionally performed by physicians on a clinical event classification (CEC) committee. OBJECTIVES: The experience of the centralized CEC group of the APEX-AMI trial is described. This group adjudicated key secondary pre-specified outcome measures of congestive heart failure (CHF) and cardiogenic shock through 90 days using an algorithmic approach for some events. METHODS: Data were collected via an electronic data capture (EDC) tool on all subjects, and additional information was provided via EDC for patients identified by site investigators with CHF or shock. Two strategies were used to adjudicate potential events: 1) a computer algorithm (followed by physician confirmation) analyzed data to determine whether events met trial end point definitions; or 2) physician review was used if EDC data were inadequate to allow classification by algorithm. RESULTS: Of 5745 patients, 282 suspected cardiogenic shock and 465 suspected CHF events were identified. The computer algorithm or physicians confirmed 196/282 cardiogenic shock and 277/465 CHF end points. Overall, 242/742 (32.6%) of suspected events were classified by algorithm. Of the 500 events not resolved by computer algorithm, the CEC physicians agreed with site investigator assessments in 126/277 (45%) of CHF and 151/196 (77%) of cardiogenic shock events. The CEC committee completed adjudication of all suspected 30- and 90-day CHF and cardiogenic shock events within 7 days of the last patient 30-day follow-up visit and within 1 day of the last patient 90-day follow-up visit. Only 27% of patients required source document collection in addition to EDC-collected information. CONCLUSIONS: A complementary approach of a computerized assessment and physician review was used in the CEC effort of the APEX-AMI trial. The algorithm categorized approximately one third of suspected CHF/cardiogenic shock events. The APEX-AMI CEC experience shows that an algorithmic approach may be a useful strategy for end point evaluation but requires validation.
