The Association Between Breast Tumor Immune Subtypes and Breast Cancer Outcomes

Background: The tumor immune microenvironment (TIME) is important for breast cancer outcomes, with strong prognostic relevance in estrogen receptor–negative (ER−) disease. However, the prognostic significance of immune infiltration in estrogen receptor–positive (ER+) breast cancer remains unclear, due in part to methodological limitations and inadequate consideration of immune heterogeneity. This dissertation evaluates whether immune subtypes defined using transcriptomic and spatial protein-based approaches provide robust and long-term prognostic information across breast cancer subtypes.Methods: Data were drawn from the population-based Carolina Breast Cancer Study (CBCS). RNA-based immune subtypes were identified using latent class analysis (LCA) applied to expression of 48 immune-related genes among 2,783 tumors, with disease-free interval (DFI) evaluated in early (≤5 years) and late (5–10 years conditional) time windows. External validation was conducted in the METABRIC and SCAN-B cohorts. Complementary spatial immune profiling was performed using multiplex immunofluorescence (IF) on 1,665 tumors to define adaptive immune subtypes based on six T-cell markers, assess intratumoral heterogeneity, and compare multi-marker IF, RNA-based, and CD8-only classifiers. Associations with DFI were estimated using Cox proportional hazards models stratified by ER status.ivResults: LCA identified immune classes broadly characterized as Adaptive, Innate, and Quiet. Adaptive immune class membership was associated with improved early DFI in both ER+ and ER− breast cancers and with substantially reduced late recurrence risk, particularly among ER+ tumors. Innate and Quiet immune classes were consistently associated with elevated late recurrence risk. External validation showed partial reproducibility, reflecting differences in cohort composition and gene expression platforms. Multiplex IF analyses demonstrated that adaptive-low tumors had worse long-term DFI, though associations were weaker than those observed with RNA-based classification. Importantly, intratumoral immune heterogeneity, defined by discordant immune status across tumor cores, was associated with poorer outcomes. Multi-marker IF outperformed CD8-only classification.Conclusions: Immune microenvironments measured at diagnosis provide durable prognostic information across breast cancer subtypes, including ER+ disease, when immune classification captures both immune composition and heterogeneity. Integrated transcriptomic and spatial immune profiling offers a refined framework for long-term risk stratification and immune biomarker development in breast cancer.