Development of a multivariable prognostic model for pathological response using anatomical and pathological variables in early triple-negative breast cancer treated with neoadjuvant chemo-immunotherapy: Initial results.

612 Background: The combination of chemotherapy and immune checkpoint inhibitors is the standard of care for most patients with early triple negative breast cancer (eTNBC). Higher residual cancer burden (RCB) after receiving neoadjuvant chemo-immunotherapy (NCI) is associated with worse outcomes; however, reliable tools to predict pathological response remain limited. Methods: We conducted a multicenter prospective cohort study including patients with eTNBC treated with NCI. Anatomical and pathological variables were collected. Multivariable logistic regression was used to develop a prognostic model for pathological response. RCB 0–1 was the primary outcome. p < 0.1 was used for the selection of candidate parameters, and p < 0.05 for the final assessment of the multivariate model. ROC curves and the Hosmer-Lemmershow tests were applied to assess the model performance Results: A total of 109 patients with eTNBC with NCI from six centers in Argentina were included, with tumor samples centrally collected and reviewed for centralized pathological assessment. Regarding the immune microenvironment, qualitative tumor-infiltrating lymphocytes (TILs) were categorized according to the International Immuno-Oncology Biomarker Working Group classification as low (score 1) in 44.4% and intermediate-to-high (scores 2–3) in 55.6% of patients. Pathologic complete response (pCR) was achieved in 64.8% of patients. Histological grade (Grade 3 vs. 2: OR 3.2; 95% CI 0.95–11), TIL categories (TILs score 2: OR 3.4; 95% CI 0.84–14 and score 3: OR 6.3; 95% CI 1.1–37) and axillary node involvement (present vs. absent: OR 0.28; 95% CI 0.083–0.95) were selected for the final multivariate model. The final model showed good discrimination (AUC = 0.82) with adequate calibration, confirmed by a non-significant Hosmer–Lemeshow test (χ² = 3.28, p = 0.35). Conclusions: The model enabled stratification of patients into distinct risk groups with significantly different probabilities of achieving RCB scores of 0 or 1. This approach may provide additional information for risk stratification and prognosis, improving our understanding of which patients are more likely to achieve pathological response. Further studies will assess the external validation of the model.