Fig. 3: Effect of equally incomplete ascertainment of events in both control and experimental patients. Panel A displays the true effect of the experimental treatment: a relative risk reduction of 1/5, generating an absolute risk reduction signal of 0.10 whose confidence interval excludes zero. If the experimental and control patients are equally affected by this incomplete ascertainment (missing, say, 25% of events in each group) the misclassification of events depicted in panel B would occur. As a consequence (panel C), although the relative risk reduction is preserved, the absolute risk reduction signal declines from 0.10 to 0.075, its confidence interval now crosses zero, and the trial result becomes indeterminate.