Early Termination of the Stroke Prevention in Atrial Fibrillation I Trial: Protecting Participant Interests in the Face of Scientific Uncertainties and the Cruel Play of Chance

Abstract

The Stroke Prevention in Atrial Fibrillation (SPAF) I trial evaluated aspirin and warfarin for prevention of stroke and nonCNS emboli in elderly patients with nonvalvular atrial fibrillation. Participants were categorized as either warfarin-eligible or warfarin-ineligible based on contraindications to or refusal of anticoagulation, and interim efficacy monitoring examined treatment effects separately by warfarin eligibility. The planned primary analyses compared aspirin to placebo among all participants and warfarin to placebo among warfarin-eligible patients. The study was terminated early following the second interim analysis due to a large reduction in thromboembolic events by aspirin versus placebo among the subgroup of warfarin-eligible participants (1 vs. 18, respectively, relative risk reduction = 94%, p < 0.001). This reduction was not evident among warfarin-ineligible patients (25 vs. 28, respectively, relative risk reduction = 8%, p = 0.8). The reduction by aspirin vs. placebo for all aspirin-assigned patients (the planned primary analysis) was significant (26 vs. 46, respectively, relative risk reduction = 42%, p = 0.02), but this resulted from pooling of subgroups with dissimilar responses. While the extreme effect of aspirin in anticoagulation-eligible participants was suspected to be due to the play of chance, termination of the SPAF I trial was justified to protect the interests of warfarin-eligible participants assigned placebo. The potential implications of interim efficacy monitoring of multiple subgroups should be carefully considered when planning interim monitoring.