Predictors of replication success.

Table shows comparisons of potential predictors of replicability between replicated and contradicted studies using different definitions of replication success. P-values are shown for a Mann-Whitney test (for continuous variables) or Fisher’s exact test (for categorical variables) comparing replicated and contradicted studies. For the significance criterion, only 15 studies (i.e., initially positive RCTs) were considered. Study design is not included as a predictor of statistical significance, as phase 1 trials did not use significance testing. Sample size for each group varies according to the specific criteria: statistical significance: 13 replicated, 2 contradicted; CI overlap: 21 replicated, 3 contradicted; both criteria: 20 replicated, 4 contradicted. IQR = interquartile range.