Do published data in trials assessing cancer drugs reflect the real picture of efficacy and safety?

Abstract

Background: The reporting quality of publications is of vital importance to ensure accurate evidence dissemination. This study aimed to compare the consistency of results reporting between the ClinicalTrials.gov results database and the respective matching publications. Methods: We identified 323 phase III/IV cancer drug trials with a randomized controlled design and searched PubMed for publications in a 50% random sample (n=160). Data were extracted independently from ClinicalTrials.gov and publications. A scoring system was applied to determine characteristics associated with reporting quality. Results: Of 117 reviewed trials with publications, result reporting was significantly more complete in ClinicalTrials.gov for efficacy measurement (92.3% vs 90.6%), serious adverse events (SAEs; 100% vs 43.6%), and other adverse events (OAEs; 100% vs 62.4%). For trials with both posted and published results for design information (n=117), efficacy measurements (n=98), SAEs (n=51), and OAEs (n=73), discrepancies were found in 16 (13.7%), 38 (38.8%), 26 (51.0%), and 54 (74.0%) trials, respectively. Overreporting of treatment effects (7 trials) and alteration of primary end points favoring statistically significant outcomes (11 trials) were the major discrepancies in efficacy reporting; incomplete (66 trials) and underreporting (20 trials) of SAEs were the predominant issues in benefit/risk reporting. Median quality score was 21 (range, 14-28). Trials that had parallel assignment, were phase IV, had primary funding by industry, were completed after 2009, and had earlier results posted possessed better reporting quality. Conclusions: Although most trials showed reasonable completeness and consistency, some discrepancies are prevalent and persistent, jeopardizing evidence-based decision-making. Our findings highlight the need to consult results systematically from both ClinicalTrials.gov and publications.