Rate of successful replication as measured by different criteria.

Results are shown as the percentage of studies with eligible replications in which the original result was replicated by each of 5 different criteria. For the statistical significance criterion, we excluded 7 phase 1 trials that did not report p-values and 2 studies with non-significant results. For these studies, the aggregate criterion (CI overlap + statistical significance) only considers CI overlap. The 2 negative studies are included in the “statistical significance (including negative studies)” criterion as a sensitivity analysis.