Abstract
Background Adjuvant chemotherapy decisions in early-stage, hormone receptor-positive, HER2-negative breast cancer traditionally rely on clinicopathological features such as tumor size, grade, and lymph node status. However, multigene expression assays like MammaPrint offer additional prognostic information that may alter treatment recommendations. This study aimed to assess the level of agreement between MammaPrint-based genomic risk classification and chemotherapy recommendations derived from National Comprehensive Cancer Network (NCCN)-based clinical criteria in a cohort of Bosnia and Herzegovina breast cancer patients. Methods We retrospectively analyzed 66 patients with HR+/HER2-, node-negative early breast cancer treated between 2023 and 2024. All patients underwent MammaPrint testing, which classified tumors as either low risk or high risk for distant recurrence. Clinical risk was assessed using a simplified NCCN-guided algorithm, in which chemotherapy was recommended for tumors >2 cm and/or grade three histology. The primary outcome was the rate of concordance between genomic and clinical risk classifications. Statistical analysis included descriptive summaries, cross-tabulation, and Cohen's kappa to evaluate agreement. Results Of the 66 patients analyzed, MammaPrint classified 27 (40.9%) as high risk and 39 (59.1%) as low risk. Based on NCCN criteria, 36 patients (54.5%) were considered clinically high-risk and recommended for chemotherapy, while 30 (45.5%) were not. Concordance between genomic and clinical classifications was observed in 37 patients (56.1%), while 29 patients (43.9%) showed discordant results. The most common discordance pattern was a clinically high-risk but genomically low-risk classification, observed in 19 cases (65.5% of discordant pairs). Cohen's kappa for agreement between methods was 0.136, indicating slight agreement beyond chance. McNemar's test yielded a χ² value of 10.0 (p = 0.036), suggesting significant asymmetry in discordance patterns. Conclusion This study highlights a substantial rate of discordance between MammaPrint genomic risk and NCCN-based clinical risk assessment. In our cohort, reliance on clinicopathological features alone would have led to different chemotherapy recommendations in over half of the cases. These findings support the clinical utility of multigene assays in refining adjuvant treatment decisions and reducing potential overtreatment in early breast cancer.