Comparing Digital Mammography and Contrast-Enhanced Magnetic Resonance Imaging for Distinguishing Benign and Malignant Breast Lesions: A Propensity Score - Matched Retrospective Study

Abstract

Background: Retrospective comparisons between imaging modalities commonly are confounded by systematic differences between groups between the characteristics of the patients undergoing different tests. Objective: To determine whether contrast-enhanced magnetic resonance imaging (CEMRI) discriminates malignant breast lesions better than digital mammography (DM) after the offsetting of baseline imbalances with the use of propensity score matching (PSM). Materials and Methods: Records of 480 women evaluated for suspicious breast lesions between 2020 and 2024 were reviewed retrospectively. Of these, 300 underwent DM and 180 underwent CEMRI. Propensity scores were estimated based on age, body mass index (BMI), lesion size, breast density, parity, and family history. One-to-one nearest-neighbor matching with a caliper of 0.2 standard deviations of the logit propensity score yielded 160 matched pairs. Diagnostic accuracy, confirmed by histopathology, was compared using McNemar's test. Results: Before matching, patients undergoing CEMRI were younger (mean age, 44.2 years vs 52.1 years; P < 0.001), had lower BMI (24.8 vs 26.5; P = 0.008), denser breasts (64.1% vs 38.7%; P = 0.002), lower parity (1.8 vs 2.1; P = 0.03), and higher family history prevalence (25.6% vs 14.3%; P = 0.01). After matching, covariates were balanced (all standardized mean differences < 0.1; Rubin's B = 18.3; Rubin's R = 1.1). CEMRI demonstrated higher sensitivity (94.4% vs 83.1%; P = 0.01) with comparable specificity (86.9% vs 87.5%; P = 0.73). Conclusion: After PSM correction for the patients' characteristics, CEMRI exhibited increased sensitivity in malignancy detection compared with DM in favor of deploying CEMRI in dense-breast or at-risk populations.