Artificial intelligence for predicting pathological complete response to neoadjuvant therapy in triple-negative breast cancer: A systematic review and meta-analysis

Background: Artificial intelligence (AI) has emerged as a promising approach to predict pathological complete response (pCR) to neoadjuvant systemic therapy (NAST) in early triple-negative breast cancer (eTNBC), but the available evidence remains methodologically heterogeneous and has not been quantitatively synthesized. Methods: We conducted a systematic review and meta-analysis of studies evaluating AI-based models for pCR prediction after NAST in eTNBC. PubMed/MEDLINE, Embase, and CENTRAL were searched up to January 1, 2026. Eligible studies included original investigations applying machine learning or deep learning to radiological, pathological, clinical, or omics data, with pCR discrimination reported at least as area under the receiver operating characteristic curve (AUC). Random-effects meta-analysis of logit-transformed AUCs was performed on studies with sufficient data to reconstruct AUC variance. Results: Fifty-eight studies were included in the systematic review, and 20 cohorts from 19 studies were eligible for quantitative synthesis. Overall, AI-based models showed high discriminative performance for predicting pCR, with a pooled AUC of 0.79 (95% CI, 0.75-0.83). However, substantial heterogeneity was observed (I2 = 73.1%). Models incorporating radiomics and longitudinal data showed numerically higher AUCs than non-radiomics and baseline-only approaches, respectively, although subgroup differences were not statistically significant. Multimodal models did not significantly outperform unimodal models. Funnel plot asymmetry suggested possible small-study effects or publication bias. Conclusions: AI-based models demonstrate promising accuracy for predicting pCR after NAST in eTNBC, but their clinical applicability is limited by substantial heterogeneity, inconsistent validation strategies, and potential publication bias. Prospective multicenter studies with standardized reporting and robust external validation are mandatory before routine clinical implementation.