Comparison of primary endpoints between publications, registries, and protocols of phase III cancer clinical trials

BACKGROUND: Decisions by leading journals to require trial registration and to make protocols of phase III randomized clinical trials (RCTs) publicly accessible were landmark events in clinical trial reporting. MATERIALS AND METHODS: We identified phase III cancer RCTs published between 2013 and 2015 in New England Journal of Medicine (NEJM), The Lancet, The lancet Oncology, JAMA and Journal of Clinical Oncology (JCO). RESULTS: We identified 345 reports of phase III RCTs of which 217 (62.9%) had available protocols. The availability rates for NEJM, The Lancet, The Lancet Oncology, JAMA and JCO were 98.0%, 33.3%, 22.7%, 55.6% and 88.3%, respectively. Journal and publication year were significantly associated with protocol availability. Eight of 215 trials (3.7%) with English language protocols had a discrepancy in primary endpoints between publication and protocol. Discrepancies of primary endpoints between protocol and registration existed in 16 (7.7%) of 209 trials. CONCLUSIONS: The policy of providing protocols with articles reporting RCTs has not been enforced rigorously. Selective reporting of primary endpoints only remains in a small fraction of phase III trials. Further improvement in consistency between primary endpoints registered and that in protocol is necessary.