Comparison between publicly accessible publications, registries and protocols of phase III trials indicated persistence of selective outcome reporting - Journal of Clinical Epidemiology

Abstract:  Objectives

The decision to make protocols of phase III randomized clinical trials (RCTs) publicly accessible by leading journals was a landmark event in clinical trial reporting. Here we compared primary outcomes defined in protocols with those in publications describing the trials, and in trial registration.

Study design and setting

We identified phase III RCTs published between January 1, 2012 and June 30, 2015 in The New England Journal of Medicine, The Lancet, The Journal of the American Medical Association and The BMJ with available protocols. Consistency in primary outcomes between protocols and registries (articles) were evaluated.

Results

We identified 299 phase III RCTs with available protocols in this analysis. Out of them, 25(8.4%) trials had some discrepancy for primary outcomes between publications and protocols. Types of discrepancies included protocol-defined primary outcome reported as non-primary outcome in publication (11 trials, 3.7%), protocol-defined primary outcome omitted in publication (10 trials, 3.3%), new primary outcome introduced in publication (8 trials, 2.7%), protocol-defined non-primary outcome reported as primary outcome in publication (4 trials, 1.3%) and different timing of assessment of primary outcome (4 trials, 1.3%). Out of trials with discrepancies in primary outcome, 15 trials (60.0%) had discrepancies that favored statistically significant results. Registration could be seen as a valid surrogate of protocol in 237 of 299 trials (79.3%) with regard to primary outcome.

Conclusions

Despite unrestricted public access to protocols, selective outcome reporting persists in a small fraction of phase III RCTs. Only studies from four leading journals were included, which may cause selection bias and limit the generalizability of this finding.