Benefits of preoperative MRI in breast cancer surgery studied in a large population-based cancer registry

Background: Although evidence for the benefits of preoperative MRI in breast cancer is lacking, use of MRI is increasing and characterized by large interhospital variation. The aim of the study was to evaluate MRI use and surgical outcomes retrospectively. Methods: Women with invasive breast cancer (pT1-3) or ductal carcinoma in situ (DCIS), diagnosed in 2011-2013, were selected from the Netherlands Cancer Registry and subdivided into the following groups: invasive cancer, high-grade DCIS, non-palpable cancer, age 40 years or less, and invasive lobular cancer. Associations between preoperative MRI use and initial mastectomy, resection margin after breast-conserving surgery (BCS), re-excision after BCS, and final mastectomy were analysed. Results: In total, 5514 women were included in the study; 1637 (34.1 per cent) of 4801 women with invasive cancer and 150 (21.0 per cent) of 713 with DCIS had preoperative MRI. Positive resection margins were found in 18.1 per cent women who had MRI and in 15.1 per cent of those who did not (adjusted odds ratio (OR) 1.20, 95 per cent c.i. 1.00 to 1.45), with no differences in subgroups. Re-excision rates were 9.8 per cent in the MRI group and 7.2 per cent in the no-MRI group (adjusted OR 1.33, 1.04 to 1.70), with no differences in subgroups. In the MRI group, 38.8 per cent of patients ultimately underwent mastectomy, compared with 24.2 per cent in the no-MRI group (adjusted OR 2.13, 1.87 to 2.41). This difference was not found for patients aged 40 years or less, or for those diagnosed with lobular cancer. Conclusion: No subgroup was identified in which preoperative MRI influenced the risk of margin involvement or re-excision rate after BCS. MRI was significantly associated with more extensive surgery, except in patients aged 40 years or less and those with invasive lobular cancer. These results suggest that use of preoperative MRI should be more targeted, and that general, widespread use be discouraged.