Abstract P6-06-37: Predicting improvements in survival based on improvements in pathologic response rate to neoadjuvant chemotherapy in different breast cancer subtypes

Abstract

Abstract Background: Individuals with excellent pathologic response (complete response or minimal residual cancer burden, pCR+RCB-I) to neoadjuvant chemotherapy have prolonged survival, and several chemotherapy regimens have resulted in improved pathologic response rates. However, how to estimate expected improvements in disease free survival (DFS) in a clinical trial based on improvement in pathologic response remains uncertain. The purpose of this study was to develop a statistical tool to estimate improvements in DFS based in improvements in pCR/RCB-I in triple negative (TNBC) and HR+/HER2- breast cancer subtypes. Methods: 387 clinical stage II (73%) and III (23%) breast cancers who received neoadjuvant T/FAC chemotherapy at MD Anderson Cancer Center were included in this analysis (N = 127 TNBC, N = 260 HR+/HER2-). Patients received adjuvant endocrine therapy if HR+. To evaluate the association between pCR rate and survival we used within-subtype stratified bootstrap analysis with biased resampling from the two response groups (pCR and RCB-I) to generate 500 bootstrapped populations with a range of different response rates. Survival was based on the Kaplan-Meier estimator with its variance obtained from the bootstrap standard error. Power was estimated as the probability of no overlap in the 95% log confidence intervals of survival in 500 bootstrap replicates of populations of different sizes and response rates. Results: Excellent pathologic response (pCR or RCB-I), was observed in 50% and 23% in the TNBC and ER+/HER2- cohorts respectively. The median follow-up was 7.6 years (range 0.1 to 13.4 years); 48 and 59 relapses occurred in the TNBC and HR+ subtypes, respectively. Bootstrap analysis showed a linear dependence of the 5-year DFS on the pCR/RCB-I rate, with a slope of 0.472 for TNBC suggesting a 4.7% improvement in DFS for every 10% increase in response rate. A more modest slope of 0.129 was observed for ER+/HER2- cancers. The sample size of a randomized 2-arm study required to show with 80% power a statistically significant improvement in 5-year DFS that corresponds to 75% response rate compared to baseline response rate of 50% and 23% in TNBC and ER+/HER2- would be 1144 and 1688 cases respectively. We are providing an open-access web-based calculator to estimate improvements in DFS and for sample size calculations for randomized clinical trials with combined endpoints of pCR/RCB-I and survival. Conclusions: We observed a linear increase in DFS with increase in pathologic response rate in the evaluated cohort. The slope depends on breast cancer subtype - it is greater in TNBC than in ER+ cancers - and expected to be influenced by the stage distribution in the study population. These results could help provide a basis for powering studies with combined response and survival endpoints. Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P6-06-37.