Use of cumulative meta-analysis in the design, monitoring, and final analysis of a clinical trial: a case study

From 1983 to 1987, the Department of Veterans Affairs (DVA) Cooperative Studies Program (CSP) conducted a multicenter clinical trial (CSP #207) to determine whether four different antiplatelet regimens compared to placebo could prevent the occlusion of grafts following coronary artery bypass surgery. The study showed that all of the active regimens tended to be better than placebo and that the three regimens containing aspirin were statistically significantly better. A cumulative meta-analysis of 12 trials performed shortly before the end of CSP #207 raised the issue as to whether the meta-analysis, if done earlier, would have changed the conduct of the trial. At the start of the planning period, one trail of size n = 37 had been published with a nonsignificant odds ratio (OR) of 0.74 (95% CI: 0.18, 3.12). At the time that CSP # 207 was approved by the DVA Cooperative Studies Evaluation Committee, two trials had been published (cumulative n = 150, OR = 0.44, 95% CI 0.19, 0.99). At the time patient intake started, five trials showed cumulative n = 769, OR = 0.42, 95% CI = 0.26, 0.68. Although the first 6-month CSP #207 progress report showed no treatment effect, by the time of the 12-month review by the Data Monitoring Board (DMB) a trend was developing in favor of active treatment. If the results of the meta-analysis had been available to the DMB at that time, conceivably the Board would have recommended stopping the placebo arm because of a convincing treatment effect based on the totality of the evidence. Cumulative meta-analysis could be useful as an adjunct in the planning, conduct, and final analysis of a clinical trial. It could also be used as one piece of evidence in the monitoring of the ongoing phase of a trial.