Tumor-associated macrophage (TAM) score as an independent predictor of recurrence in residual disease after neoadjuvant chemotherapy and for identification of high-risk patients for adjuvant escalation.

Abstract

522 Background: Tumor-associated macrophages (TAMs) drive immune evasion and chemotherapy resistance in breast cancer through tumor microenvironment reprogramming. We developed a composite TAM score integrating transcriptional programs with spatial IHC polarization markers to stratify recurrence risk and enrich macrophage-modulating trials, hypothesizing it would improve prognostic accuracy in treatment-naïve disease and predict neoadjuvant chemotherapy (NACT) response. Methods: TAM score was derived from scRNA-seq atlases (n=4), filtering survival-associated genes in METABRIC/TCGA (n=2,509), and weighting IHC markers CD163/CD206 normalized to CD68/CSF1R via Cox coefficients. Validation: treatment-naïve invasive carcinomas (n=83; follow-up 57 months; 24 events). Primary endpoint: recurrence-free survival (RFS); multivariable Cox adjusted for age, tumor size, nodal status, grade, subtype, treatment. Predictive utility assessed in matched primary/post-NACT residual tumors (n=45) correlating TAM dynamics with RCB. Results: High TAM score (≥median) independently predicted worse RFS (HR 3.8, 95% CI 1.6-9.1, p<0.05) and outperformed CD163/CD68 alone (HR 2.1, p<0.05); 5-year RFS was 68% vs 91% (TAM-high vs TAM-low). Prognostic value persisted in node-negative (HR 4.2, p=0.02) and luminal B (HR 3.5, p=0.01) subsets. In the NACT cohort, baseline TAM score inversely correlated with pCR (12% vs 38%, p=0.04). Among residual disease (RCB-II/III, n=34), persistent TAM-high post-NACT identified 3-year RFS 52% vs 85% in TAM-low/normalized cases (HR 4.1, p=0.01), independent of RCB; CD163+/CD206+ expansion occurred in 76% of RCB-III vs 29% of RCB-I (p=0.008). Conclusions: This TAM score predicts recurrence from pre-treatment tissue and identifies macrophage-driven resistance persisting post-NACT, nominating TAM-high residual disease for macrophage-modulating trials. Translational evidence supports this: CSF1R inhibition reduced immunosuppressive TAMs, increased M1-like programs, decreased PD-L1/PD-L2, and enhanced anti-PD-1 efficacy preclinically with ex vivo confirmation. Early-phase data (rebastinib plus chemotherapy in HER2-negative metastatic disease) demonstrate feasibility, supporting prioritization of TAM-high patients with residual disease for trials combining TAM modulation with checkpoint therapy. TAM score performance in treatment-naïve and NACT cohorts. Endpoint TAM-High TAM-Low HR scores (95% CI) p-value Treatment Naïve cohort 5 year RFS (Overall) 68% 91% 3.8 (1.6-9.1) <0.05 RFS (node-negative) 4.0 0.02 RFS (luminal B)-- 3.5 0.01 CD163/CD68 alone 2.1 <0.05 NACT Cohort (n=45) Baseline pCR rate 12% 38%- 0.04 3-year RFS(RCB-II/III n=34) 52% 85% 4.1 0.01 CD163+/CD206+ expansion 76% (RCB-III) 29%(RCB-I)- 0.008