Circulating microRNAs (miRNAs) are promising non-invasive biomarkers for cancer detection; however, their diagnostic performance across breast cancer molecular subtypes and their incremental value beyond demographic and epidemiological variables remain incompletely characterized. We conducted a multicenter case–control study including 317 breast cancer cases and 127 population-based controls. Serum levels of 44 literature-derived miRNAs were quantified by RT-qPCR. Feature selection was performed using LASSO penalization, followed by multivariable logistic regression to estimate odds ratios (ORs) with 95% confidence intervals (CIs). Models were adjusted for demographic and epidemiological covariates. Predictive performance was assessed using repeated fivefold cross-validation and reported as area under the curve (AUC) with bootstrap bias-corrected 95% CIs. Incorporating demographic and epidemiological covariates enhanced discrimination overall (AUC = 0.908 vs. 0.802 unadjusted) and across subtypes. The most notable improvements were observed in Luminal A (0.896 vs. 0.751) and Luminal B (0.894 vs. 0.768), while HER2-positive and Basal-like tumors already showed high performance (AUC = 0.965 and 0.989, respectively). Among the 12 miRNAs selected by LASSO, miR-21-5p and miR-423-3p were consistently elevated in cases, particularly in HER2-positive and Basal-like tumors, whereas miR-101-3p, miR-146a-5p, and miR-29a-3p showed reproducibly lower levels across multiple subtypes, consistent with oncogenic and tumor-suppressive roles, respectively. Circulating miRNA signatures, especially when integrated with demographic and epidemiological information, demonstrate high discriminatory power for breast cancer detection across molecular subtypes. These results support subtype-aware, minimally invasive strategies for screening and risk stratification using miRNA-based models. Prospective validation in independent cohorts is warranted to confirm clinical utility.
Gómez-Acebo et al. (Sat,) studied this question.