BACKGROUND: Separate analysis of cause-specific treatment effects is important for interpreting results of randomized trials. We sought to determine the extent to which cause-specific effects on primary events are reported in contemporary randomized controlled trials in oncology. METHODS: We screened 833 randomized trials published in eight leading medical journals between January 2006 and December 2009. We excluded prevention studies (n=52), secondary reports (n=100), and one retracted study. Analysis was further restricted to 116 trials in non-metastatic/recurrent cancer that used an event-free survival primary endpoint. For each study included in the analysis, we evaluated whether treatment effects on both cancer and non-cancer events comprising the primary endpoint were reported separately and whether statistical analysis was provided. RESULTS: Of the 116 randomized trials, 47 (40%; 95% confidence interval (CI), 32-50%) reported effects on both cancer and non-cancer events comprising the primary endpoint, with statistical analysis provided in 13 (11%; 95% CI, 7-19%). Twenty-six trials (22%; 95% CI, 15-31%) reported effects on cancer but not non-cancer events, with statistical analysis provided in 11 (9%; 95% CI, 5-17%). In 43 studies (37%; 95% CI, 28-47%), no effects on cancer-specific components of the primary endpoint were given. Of these, 33 studies (28%; 95% CI, 21-38%) did report effects on some cancer-specific event, while ten (9%; 95% CI, 4-16%) did not report effects of treatment on any cancer event. DISCUSSION: Many randomized trials in oncology do not report cause-specific effects on primary events. Increased specificity is needed in the design and reporting of cancer clinical trials.
BACKGROUND: Separate analysis of cause-specific treatment effects is important for interpreting results of randomized trials. We sought to determine the extent to which cause-specific effects on primary events are reported in contemporary randomized controlled trials in oncology. METHODS: We screened 833 randomized trials published in eight leading medical journals between January 2006 and December 2009. We excluded prevention studies (n=52), secondary reports (n=100), and one retracted study. Analysis was further restricted to 116 trials in non-metastatic/recurrent cancer that used an event-free survival primary endpoint. For each study included in the analysis, we evaluated whether treatment effects on both cancer and non-cancer events comprising the primary endpoint were reported separately and whether statistical analysis was provided. RESULTS: Of the 116 randomized trials, 47 (40%; 95% confidence interval (CI), 32-50%) reported effects on both cancer and non-cancer events comprising the primary endpoint, with statistical analysis provided in 13 (11%; 95% CI, 7-19%). Twenty-six trials (22%; 95% CI, 15-31%) reported effects on cancer but not non-cancer events, with statistical analysis provided in 11 (9%; 95% CI, 5-17%). In 43 studies (37%; 95% CI, 28-47%), no effects on cancer-specific components of the primary endpoint were given. Of these, 33 studies (28%; 95% CI, 21-38%) did report effects on some cancer-specific event, while ten (9%; 95% CI, 4-16%) did not report effects of treatment on any cancer event. DISCUSSION: Many randomized trials in oncology do not report cause-specific effects on primary events. Increased specificity is needed in the design and reporting of cancer clinical trials.