P-hacking in clinical trials: a meta-analytical approach / Norman Belas/Paul Bengart/Bodo Vogt

Abstract

Clinical trials play a decisive role in the drug approval processes. By completing a p-curve analysis of a newly compiled data set that consists of thousands of clinical trials, we substantiate that the occurrence of p-hacking in clinical trials is not merely hypothetical. Medical and pharmaceutical research consists of both primary and secondary study endpoints. The primary finding covers the main effect, which directly influences the approval process, while the secondary outcome delivers further additional information. For primary p-curves, we observed an abnormal increase in the p-value frequency at common significance thresholds, while the secondary p-curves exhibited no such anomaly.