A quantitative study of tumor response and progression-free survival as surrogate endpoints for overall survival in first-line treatment of metastatic breast cancer

Abstract

Background: The traditional endpoint to assess the efficacy of chemotherapies for metastatic breast cancer is overall survival (OS), but this endpoint can only be observed after prolonged follow-up. Moreover, with the number of active compounds available in this disease, any potential effect of first-line therapies on OS may be confounded by the effects of second-line therapies. It is therefore of interest to investigate whether tumor response or progression-free survival (PFS) could replace OS as the primary endpoint in randomized trials for patients with metastatic breast cancer. Methods: Individual patient data were collected on 4,256 patients in 12 randomized trials comparing an anthracycline (alone or in combination) with a taxane (alone or in combination with an anthracycline). Tumor response was assessed using WHO criteria. Progression-free survival was calculated from date of randomization to progression or death from any cause. Overall survival was calculated from date of randomization to death from any cause. Surrogacy was assessed through the correlation between the endpoints as well as the correlation between the treatment effects on the endpoints. Correlation coefficients were estimated along with their 95% confidence interval (CI). Results: Median follow-up of alive patients was 41 months, median OS 20 months, median PFS 7 months. Tumor response was strongly associated with PFS (survival odds ratio=5.8, 95% CI=5.0-6.5) and with OS (survival odds ratio=3.0, 95% CI=2.6-3.3). The correlation coefficient between response log odds ratios and PFS log hazard ratios was 0.97 (95% CI=0.74-1.21). The correlation coefficient between response log odds ratios and OS log hazard ratios was 0.56 (95% CI=-0.33-1.44). The rank correlation between PFS and OS was 0.695 (95% CI=0.693-0.696). The correlation between the log hazard ratios for PFS and OS was 0.44 (95% CI=- 0.41 -1.30). Conclusions: Tumor response is predictive of longer PFS and longer OS. Treatment effects on response are strongly correlated with treatment effects on PFS but not on OS. PFS is poorly correlated with OS, and treatment effects on PFS are poorly correlated with treatment effects on OS. These results indicate that tumor response is an acceptable surrogate for PFS but no endpoint is a good surrogate for OS in these trials. Associations between treatment effects are not precisely estimated despite the relatively large sample sizes.