Abstract Background Trastuzumab deruxtecan (T-DXd) has transformed the treatment paradigm for HER2-expressing breast cancer, including HER2-low disease. However, biomarkers associated with T-DXd activity remain poorly defined. We conducted a comprehensive analysis of clinical, genomic, and immune correlates of T-DXd outcomes to identify molecular determinants of therapeutic benefit and resistance. Methods We retrospectively analyzed 2 independent cohorts of patients with advanced breast cancer treated with T-DXd at Dana-Farber Cancer Institute and Yale Cancer Center between 2018 and 2024. We included patients with ≥2 tumor blocks with HER2 immunohistochemistry (IHC) assessments prior to T-DXd. Clinical data on 524 patients were manually reviewed, and genomic profiling and immune microenvironment assessments were performed on a subset of patients. Multivariable Cox proportional hazards models evaluated associations between molecular features and overall survival (OS) and time to next treatment (TTNT). Results Among 524 patients, HER2 IHC discordance between sequential tumor biopsies was observed in 20% of patients and was independently associated with significantly worse OS (hazard ratio [HR] = 0.67; P = .012) and TTNT (HR = 0.65; P = .002), resembling outcomes seen in HER2 0 tumors. Genomic analysis revealed that PTEN mutations correlated with inferior TTNT (HR = 2.2; q = 0.068), whereas ERBB2 amplifications predicted improved OS and TTNT. An inflamed tumor microenvironment determined by digital pathology was associated with significantly poorer TTNT outcomes (median TTNT = 5.5 months) compared with immune-desert phenotypes (median OS = 9.6 months, P = .03). Conclusions This study identifies HER2 IHC discordance and specific genomic and immune features as prognostic biomarkers of T-DXd efficacy. These findings warrant prospective validation and may inform biomarker-driven strategies to optimize T-DXd therapy in HER2-expressing malignancies.