Randomised controlled trials in oncology closed early for benefit: Trends in methodology, results, and interpretation

PURPOSE: To assess methodology, results and interpretation of oncology randomised controlled trials closed early for benefit (RCTCEB). METHODS: Structured literature search (1950-2008) to identify all published oncology RCTCEB. We then searched for related follow-up articles and conference abstracts to evaluate whether study results and conclusions changed with longer follow-up. A standardised data abstraction process captured information related to statistical methodology, details of interim analyses, results and conclusions. Original articles and follow-up reports were compared for results of primary end-point and author conclusions. RESULTS: We identified 71 RCTCEB. In 16 articles (23%) the study primary end-point was not explicitly stated. Most trials were open to accrual (47/71, 66%) at the time of closure. Formal interim analysis was performed in 65 (92%) trials of which 72% (47/65) was reported as planned; 82% (53/65) reported stopping rules. Trials on average accrued 75% of the planned sample size. Amongst the 23 (32%) RCTCEB with follow-up reports, in only one case did the study results or conclusions change substantially. CONCLUSIONS: While the majority of oncology RCTCEB follows rigourous methodological principles, an important percentage includes limitations in design and/or analysis. Amongst the 23 studies with subsequent follow-up reports, initial results were confirmed in 22 (96%).