Evolution in the eligibility criteria of randomized controlled trials for systemic cancer therapies

• Randomized controlled trial (RCT) eligibility criteria may lower generalizability.
• We assess changes in eligibility criteria of RCTs over time.
• An increased use of eligibility criteria in practice-changing RCTs is identified.
• An increase in organ-specific and comorbidity-based exclusion criteria is noted.

BackgroundEligibility criteria in randomized controlled trials (RCTs) reduce inter-patient heterogeneity, but may reduce generalizability of results. Here, we explore temporal changes in eligibility criteria of practice-changing RCTs for systemic cancer therapies and in the proportion of patients excluded from these trials after application of eligibility criteria.MethodsAn electronic search identified practice-changing RCTs published in six major journals between July 2010 and December 2012. Trial protocols were identified through journal websites and communication with authors or study sponsors. Eligibility criteria were extracted from protocols. The number of patients excluded after application of eligibility criteria was extracted from the CONSORT diagrams and text of publications. Changes in eligibility criteria over time were assessed by logistic regression and meta-regression was carried out to evaluate the impact of year of protocol on the proportion of patients who were excluded after screening.ResultsEighty-six protocols written between 1987 and 2012 were included. Over time, there has been an increasing frequency of exclusion of patients with prior cerebrovascular events (OR 1.34, p = 0.003), coagulation/bleeding disorders (OR 1.34, p = 0.006), prior gastrointestinal bleeding (OR 1.33, p = 0.01), cardiac co-morbidities (OR 1.24, p = 0.004) and exclusion based on concurrent medication (OR 1.19, p = 0.01). There has been a decrease in upper age limit usage (OR 0.83, p = 0.01) and leukopenia (OR 0.83, p = 0.009). The proportion of patients excluded from trials has increased from 9% prior to 2000 to 18% after 2010 (p-value for trend <0.001).ConclusionsRCTs have become less representative of cancer patients treated in routine practice with increased use of organ-specific and co-morbidity-based exclusion criteria.