Abstract:
|
A recent surge of cancer immunotherapy trials has revealed both a great potential of such treatments over standard chemotherapy or placebo and the significant challenges of appropriately designing and analyzing such trials. Delayed treatment effect and crossover after progression are among the reasons leading to a violation of proportional hazards (PH) assumption, which is the bedrock for survival analysis methods, such as log rank test and Cox PH regression, for correct and efficient inference on treatment effect. We systematically review the results of immunotherapy trials of advanced non-small cell lung cancers published in the past decade and provide assessment on the extent of PH violation and its impact of clinical trial findings. We also discuss potential solutions to appropriately design, monitor and analyze clinical trials where non-proportional hazards data are expected. One focus is to answer whether alternative methods, including weighted log rank test and restricted means survival (RMST) ratio, perform better relative to the standard methods.
|