Abstract
1031 Background: The FDA and EMA have published guidance for industry on the use of pCR as a surrogate endpoint to support accelerated approval in the neoadjuvant treatment of high-risk, EBC. The CTNeoBC consortium (Cortazar 2014), Berruti (2014) and Korn (2015) each used meta analyses to explore the trial level association between pCR odds ratio and the event hazards ratio but did not reach strong association. There are uncertainties raised in both guidances in determining the magnitude of increase in effect size in pCR that is reasonably likely to predict long term clinical benefit. Methods: We introduce a Cox regression model of the relationship among treatment, event hazard, and time dependent effect of pCR as joint co-variates. The trial level association between treatment and event hazard is derived and then shown to be linked with pCR rates. The methodaccounts for the magnitude of the patient level association, which is likely important in predicting clinical benefit. We apply the method to re-analyze the trial level data from CTNeoBC and derive an empirical relationship as an indirect validation between pCR and long term outcome. Results: For the 8 clinical trials included in the CTNeoBC analysis for trial level association, the predictor based on pCR and patient level association served well under the assumption of perfect surrogate endpoint. The difference away from the perfect surrogate prediction was to follow normal distribution based on Kolmogorov-Smirnov test. For event free survival (EFS), the Student's t test p-value of 0.02 suggested non-zero central location with the mean value of -0.161 (s.e. 0.058). For overall survival (OS), the Student's t test p-value of 0.0032 suggested non-zero central location with the mean value of -0.157 (s.e. 0.0393). The observed hazard ratios fit well with the predicted hazard ratios for both EFS and OS and suggest plausible trial level association with the new predictor. Conclusions: Our model predicts well the correlation between pCR and EFS as well as OS. This should serve as a supporting method to help assessment of positive results of clinical studies in the neoadjuvant setting.