• Breast Cancer
• Concordance Kappa Cohen Coefficient
• Magnetic Resonance Imaging
• Neoadjuvant Therapy
• Pathological Complete Response
• Radiological Complete Response

Abstract
Background: The complete pathological response after NAC is an important prognostic factor for breast cancer. The capability of imaging to assess the complete response to NAC before surgery remains an open challenge. The reliability of MRI in the evaluation of the complete pathological response after NAC has been assessed in view of limiting oncoplastic surgery to non-responders.

Methods: A retrospective analysis has been conducted in a single institution from January to December 2021 on 104 consecutive women (mean age, 52 years; range, 29-80 years) who underwent NAC, breast MRI before and after NAC, and surgery for biopsy-confirmed invasive breast cancer. The radiological complete response (rCR) has been evaluated by breast imaging as the absence of both early and late enhancement on MRI after NAC. The diagnostic performance of MRI in predicting the pathologic complete response (pCR) has been evaluated for all patients and for all molecular subtypes of breast cancer (luminal A, luminal B, HER2 positive, and triple-negative). The actual pCR has been assessed by pathologic examination.

Results: Radiological complete response has been observed in 39/104 patients enrolled in the analysis (37.5%), and pathological complete response in 52/104 patients (50%). An agreement analysis has been carried out using the Cohen's Kappa test for the evaluation of the concordance between the radiological and pathological complete responses, with "moderate" agreement (Cohen's Kappa 0.361) in the general population. The same analysis has been carried out by molecular subtypes and the concordance is "slight-fair" for luminal A and HER2 positive (Cohen's Kappa respectively 0.167 and 0.128) and "moderate" for luminal B and triple-negative (Cohen's Kappa respectively 0.475 and 0.407).
Overall, MRI has shown a sensitivity of 81% and a specificity of 57% for pCR detection. The highest positive predictive value, equal to 87.5%, has been obtained for Luminal B; the highest negative predictive value, also equal to 87.5%, for triple-negative breast cancer.

Conclusions: The agreement analysis between the results of MRI and pCR indicates that the diagnostic accuracy of MRI depends on the molecular subtypes, with the best results obtained for luminal B and triple-negative breast cancer. The findings obtained in the present study are encouraging. However, it is not yet possible to conclude on their basis that MRI is a fully reliable tool for pCR assessment, which therefore must still be carried out by histopathological evaluation. More conclusive results can be obtained by expanding the population of the dataset and exploring the role of radiogenomics.