Failed Randomized Clinical Trials in Radiation Oncology: What Can We Learn?

<bold>Purpose: </bold>Randomized clinical trials (RCTs) are essential to evidence-based medicine, yet a significant proportion fail to be completed. In radiation oncology, factors contributing to trial failure are not well understood. We sought to compare completed and incomplete clinical trials involving radiation therapy (RT) to identify predictors of trial failure.<bold>Methods and Materials: </bold>We undertook a review of ClinicalTrials.gov to identify RCTs involving RT. Eligible trials mandated external beam RT in ≥1 arm of the study and were registered between September 27, 2007, and December 31, 2010. Univariate and multivariate logistic regression analyses were performed to determine factors predictive of trial failure.<bold>Results: </bold>We included 134 eligible studies, of which 94 (70.1%) were successful and 40 (29.9%) failed. The reasons for trial failure were categorized as follows: lack of accrual (57.5%), inadequate funding (15.0%), drug unavailability (7.5%), interim data-monitoring report recommendations (7.5%), and other (12.5%). Over time, significantly more trials were failing to be completed (P = .010), with rates increasing from 11.8% (before 2007) to 34.0% (2007-2008) to 39.5% (2009-2012). On univariate analysis, predictors of failure were trials with a surgical comparator (odds ratio [OR], 8.12; P = .013), government sponsorship (vs non-government; OR, 3.67; P = .025), inclusion of a safety endpoint (OR, 2.85; P = .022), and studies starting after 2006 (P = .033). On multivariate analysis, surgical trials were strongly predictive of failure (OR, 12.30; P = .025) while behavioral trials were associated with success (OR, 0.11; P = .045).<bold>Conclusions: </bold>RT RCTs involving ≥1 surgical arms are at a very high risk of failure, with 75% failing to be completed. In contrast, behavioral studies are associated with study completion, with 94% of studies being successful. Future RT trials involving surgical interventions should consider novel methods to reduce the risk of trial failure. [ABSTRACT FROM AUTHOR]