Pathological complete response and long-term outcomes after neoadjuvant therapy for breast cancer: a multicenter retrospective cohort study with exploratory subtype analyses

Background To examine differences in pathological complete response (pCR) rates across immunohistochemistry (IHC)-defined breast cancer surrogate subtypes and the association of pCR with recurrence-free survival (RFS) and breast cancer-specific survival (BCSS). Methods This retrospective multicenter study included 3,803 patients from the Shanghai Jiao Tong University Breast Cancer Database who received neoadjuvant therapy followed by surgery. Total pCR was defined as ypT0/is ypN0. Multivariable logistic and Cox regression assessed factors associated with pCR and its association with survival, respectively. Exploratory global subtype-by-pCR interaction tests were restricted to the remaining four subtypes with estimable within-subtype pCR contrasts. Sensitivity analyses included exploratory covariate-balancing analyses based on propensity score matching and overlap weighting, calendar-year adjustment, and subtype-stratified Cox models. Results Overall, 908 patients achieved pCR (23.88%), with rates differing across IHC-defined surrogate subtypes (P < 0.001). During a median follow-up of 46 months, pCR was associated with favorable RFS (adjusted HR 0.37, 95% CI 0.24–0.57) and BCSS (adjusted HR 0.30, 95% CI 0.18–0.50; both P < 0.001). Exploratory global interaction tests among the remaining four subtypes with estimable within-subtype pCR contrasts showed no statistical evidence of heterogeneity for RFS (P for interaction = 0.665) or BCSS (P for interaction = 0.718); Luminal A-like disease could not be included in these tests because no RFS or BCSS events occurred among the five patients who achieved pCR. Exploratory propensity-score matching yielded 907 pairs; matched-pair robust HRs were 0.43 (95% CI 0.27–0.69) for RFS and 0.38 (95% CI 0.22–0.64) for BCSS. Associations were consistent across the other sensitivity analyses. Conclusion pCR frequency varied across IHC-defined surrogate subtypes, and pCR was associated with favorable survival in the overall cohort. The exploratory interaction analyses did not establish effect modification among the four included subtypes; Luminal A-like disease could not be assessed for interaction because of sparse events. Subtype-specific survival estimates remain exploratory and require confirmation in adequately powered studies.