Background: Achieving pathological complete response (pCR) to neoadjuvant chemotherapy (NAC) improves survival outcomes for breast cancer patients. Currently, conventional histopathological biomarkers predicting such responses are inconsistent. Studies investigating radiomic texture analysis from breast magnetic resonance imaging (MRI) to predict pCR have varied radiomic protocols introducing heterogeneity between results. Thus, the efficacy of radiomic profiles compared to conventional strategies to predict pCR are inconclusive.
Purpose: Comparing the predictive accuracy of different breast MRI radiomic protocols to identify the optimal strategy in predicting pCR to NAC.
Material and methods: A systematic review and network meta-analysis was performed according to PRISMA guidelines. Four databases were searched up to October 4th, 2021. Nine predictive strategies were compared, including conventional biomarker parameters, MRI radiomic analysis conducted before, during, or after NAC, combination strategies and nomographic methodology.
Results: 14 studies included radiomic data from 2,722 breast cancers, of which 994 were used in validation cohorts. All MRI derived radiomic features improved predictive accuracy when compared to biomarkers, except for pre-NAC MRI radiomics (odds ratio [OR]: 0.00; 95 % CI: -0.07-0.08). During-NAC and post-NAC MRI improved predictive accuracy compared to Pre-NAC MRI (OR: 0.14, 95 % CI: 0.02-0.26) and (OR: 0.26, 95 % CI: 0.07-0.45) respectively. Combining multiple MRIs did not improve predictive performance compared to Mid- or Post-NAC MRIs individually.
Conclusion: Radiomic analysis of breast MRIs improve identification of patients likely to achieve a pCR to NAC. Post-NAC MRI are the most accurate imaging method to extrapolate radiomic data to predict pCR.
Keywords: Breast; Magnetic resonance imaging; Neoadjuvant chemotherapy; Oncology; Pathological complete response; Prediction of response; Radiomics; Texture analysis.
Copyright © 2022 Elsevier B.V. All rights reserved.