Comparative effectiveness of preoperative localization techniques for non-palpable breast lesions: multicentre real-world study

Background: The increasing detection of non-palpable breast lesions has made accurate preoperative localization essential tooptimize breast-conserving surgery. Although multiple localization methods exist, there is still a lack of robust, large-scale,multicentre evaluations comparing different techniques.Methods: The LOCALIZATION01 study compares real-world data from 13 breast units in Italy and Switzerland on the impact oflocalization techniques on breast-conserving surgery for non-palpable lesions between 2016 and 2024. Four localization techniques were compared: wire-guided (WGL), radio-guided (ROLL), magnetic seed (MSL), and carbon (CL). The main outcomes were margin status, calculated resection ratio, postoperative complications, and surgical time. Subgroup analyses were performed for body mass index, lesion morphology and histopathology.Results: In total, 3241 patients were enrolled (ROLL 985, MSL 592, WGL 1079, and CL 585). ROLL achieved the highest rate of negative surgical margins, significantly outperforming MSL, WGL, and CL (97.5% versus 94.7% versus 94.5% versus 90.6%, respectively; P < 0.05). CL was associated with the highest postoperative complications rate (16.7%) versus ROLL (4.1%), MSL (4.5%), and WGL (2.1%) (P < 0.0001). The surgical time for MSL was significantly shorter when compared with WGL (46 versus 70 minutes (min); P < 0.0001) and CL (55 min; P < 0.0001). WGL had the most favourable calculated resection ratio (2.4), followed by MSL (2.6), ROLL (2.7), and CL (3.0). Multivariable analysis identified CL as an independent predictor of positive margins (odds ratio 1.82; P = 0.004), whereas ROLL was protective (odds ratio 0.45; P = 0.009).Conclusion: ROLL and MSL outperformed WGL and CL across multiple endpoints. CL data revealed objective limitations that suggest caution in its use. A personalized approach considering lesion morphology, body mass index, and logistics is recommended.