The accuracy of incremental pre-operative breast MRI findings - Concordance with histopathology in the Swedish randomized multicenter POMB trial

Purpose: The Pre-Operative MRI of the Breast (POMB) trial was a randomized, prospective, multicenter trial evaluating the impact of pre-operative breast MRI on treatment regimens and short-term surgical outcomes in women up to 56 years of age with breast cancer. The purpose of this study was to evaluate the performance of pre-operative breast MRI in the POMB trial with respect to incremental MRI findings - over conventional breast imaging methods - and their concordance with histopathology. Patients and methods: Two-hundred and ten patients (n = 210) participating in the POMB trial underwent preoperative breast MRI at two Swedish breast units. Positive predictive values (PPV) for the incremental MRI findings were calculated for three subgroups of patients with: 1. alteration/alterations of treatment plan; 2. no alteration of treatment plan; and, 3. MRI-related conversion from BCS to mastectomy. Area under the receiver operating characteristic curve (AUC) was calculated using in-breast BI-RADS based ratings for the whole MRI group. Results: After exclusions a total number of 99 incremental findings in 78 patients were eligible for statistical analysis resulting in a PPV = 74%: (95% CI 60-84%) in 39 patients with MRI related alterations of initial treatment plans and 27%: (95% CI 14-44%) in 39 patients without. Positive predictive values of incremental findings decisive for specific treatment alteration/s were 83% (95% CI 68-92%) in patients with any alteration of initial treatment plans and 91% (95% CI 70-98%) for patients (n = 20/22) with conversion from breast conserving surgery to mastectomy. The empirical AUC for the incremental findings in the whole MRI group was 85% (95% CI 78-91%). Conclusion: Breast MRI, performed and evaluated together with conventional breast imaging methods can provide relevant information at a high degree of accuracy in the pre-operative setting.