Comparison of reporting phase I trial results in ClinicalTrials.gov and matched publications

Background Data on completeness of reporting of phase I cancer clinical trials in publications are lacking. Methods The ClinicalTrials.​gov database was searched for completed adult phase I cancer trials with reported results. PubMed was searched for matching primary publications published prior to November 1, 2016. Reporting in primary publications was compared with the ClinicalTrials.​gov database using a 28-point score (2=complete; 1=partial; 0=no reporting) for 14 items related to study design, outcome measures and safety profile. Inconsistencies between primary publications and ClinicalTrials.​gov were recorded. Linear regression was used to identify factors associated with incomplete reporting. Results After a review of 583 trials in ClinicalTrials.​gov, 163 matching primary publications were identified. Publications reported outcomes that did not appear in ClinicalTrials.​gov in 25% of trials. Outcomes were upgraded, downgraded or omitted in publications in 47% of trials. The overall median reporting score was 23/28 (interquartile range 21–25). Incompletely reported items in >25% publications were: inclusion criteria (29%), primary outcome definition (26%), secondary outcome definitions (53%), adverse events (71%), serious adverse events (80%) and dates of study start and database lock (91%). Higher reporting scores were associated with phase I (vs phase I/II) trials (p<0.001), multicenter trials (p<0.001) and publication in journals with lower impact factor (p=0.004). Conclusions Reported results in primary publications for early phase cancer trials are frequently inconsistent or incomplete compared with ClinicalTrials.​gov entries. ClinicalTrials.​gov may provide more comprehensive data from new cancer drug trials.