Residual proliferative cancer burden to predict long-term outcome following neoadjuvant chemotherapy

The purpose of this study was (i) to test the hypothesis that combining Ki67 with residual cancer burden (RCB) following neoadjuvant chemotherapy, as the residual proliferative cancer burden (RPCB), provides significantly more prognostic information than either alone; (ii) to determine whether also integrating information on ER and grade improves prognostic power. A total of 220 patients treated with neoadjuvant chemotherapy for primary breast cancer were included in the study. Analyses employed a Cox proportional hazard model. Prognostic indices (PIs) were created adding in Ki67, grade and ER to RCB. Leave-one-out cross-validation was used to reduce bias. The overall change in chi(2) of the best model for each index was used to compare the prognostic ability of the different indices. All PIs provided significant prognostic information for patients with residual disease following neoadjuvant chemotherapy. RPCB (chi(2) = 61.4) was significantly more prognostic than either RCB (chi(2) = 38.1) or Ki67 (chi(2) = 53.8) alone P < 0.001. A PI incorporating RCB, Ki67 grade and ER provided the most prognostic information overall and gave chi(2) = 73.8. This study provides proof of principle that the addition of post-treatment Ki67 to RCB improves the prediction of long-term outcome. Prediction may be further improved by addition of post-treatment grade and ER and warrants further investigation for estimating post-neoadjuvant risk of recurrence. These indices may have utility in stratifying patients for novel therapeutic interventions after neoadjuvant chemotherapy.