Quality of reports related to survival events in randomized clinical trials: A review of recent cancer trials published in eight major journals

Abstract

6611 Background: About ten years ago, Altman pointed the low quality of survival analyses published in cancer journals. Despite the development of Consort statement, the lack of quality standards remained in the reports of randomized clinical trials (RCT).The aim of this study was to evaluate quality of definitions and reports of survival endpoints in recent cancer RCTs. Methods: A computerized search in Medline databases identified 274 cancer RCTs published from January 2004 to December 2004 in 4 general medical journals (BMJ, JAMA, Lancet, NEJM) and 4 clinical oncology journals (BJC, Cancer, JCO, JNCI). Eligible papers were these that reported primary analyses of RCT with survival endpoints. Three methodologists have reviewed and scored 1/3 of papers according to 7 key points: prevalence of adequate definition of survival end-points (time to origin, survival events, censures) and relevant information about their analyses (Kaplan-meier method, confidence interval, number of events, patients at risk). Reliability of scoring was checked among a random subsample of 30 papers with an independent biostatistician (kappa coefficient). Results: Following eligibility screening, 149 papers were not included. Among the 125 selected papers, 104 trials were of phase III (83%) and 98 publications (78%) were obtained from the 4 oncology journals. Among these RCTs, a total of 268 survivals endpoints were recorded and overall survival (OS) was the most frequent outcome (118 terms, 44%). Survival terms were adequately defined in 116 endpoints (43%) and OS was most frequent adequately defined (49%). Accurate information about analysis was found in 51 endpoints (19%). The less adequate information was patient's number at risk (55%). Finally, according to the 7 key methodological points, optimal quality of report was found in 19 of the endpoints (7 %) i.e.6 articles (kappa value: 0.65). Conclusions: Majority of published papers failed to provide an adequate definition and analysis reports of survival endpoints. Finally survival definitions add another source of variability actually not controlled. Readers should critically pay attention to the reported definitions since the lack of these elements may lead to misinterpretation of the RCT results. No significant financial relationships to disclose.