OBJECTIVE: To identify the issues of using overall survival (OS) as a primary endpoint in the presence of crossover and the statistical analyses available to adjust for confounded OS due to crossover in oncology clinical trials.
METHODS: An indirect comparison was conducted between pazopanib and sunitinib in advanced renal cell carcinoma. Statistical adjustment methods were used to estimate the true comparative effectiveness of these treatments. Recently, a head-to-head trial comparing pazopanib and sunitinib was completed. This provided the opportunity to compare the OS treatment effect estimated for pazopanib versus sunitinib using indirect comparison and statistical adjustment techniques with that observed in the head-to-head trial.
RESULTS: Using a rank-preserving structural failure time model to adjust for crossover in the pazopanib registration trial, the indirect comparison of pazopanib versus sunitinib resulted in an OS hazard ratio (HR) of 0.97, while an unadjusted analysis resulted in an OS HR of 1.96. The head-to-head trial reported a final OS HR of 0.92 for pazopanib versus sunitinib.
CONCLUSION: This case study supports the need to adjust for confounded OS due to crossover, which enables trials to meet ethical standards and provides decision makers with a more accurate estimate of treatment benefit.