RESUMO
Objective: To investigate the correlation between clinicopathological features and Miller/Payne (MP) grading system of breast carcinoma after neoadjuvant treatment and to establish novel prediction models. Methods: A total of 1 053 cases of invasive breast carcinoma NOS that undertaken neoadjuvant treatment according to Guidelines of CSCO for Breast Cancer were selected at the Key Laboratory of Carcinogenesis and Translational Research, Peking University Cancer Hospital & Institute from September 2016 to September 2019, and the clinical, pathologic data, MP grading and immunohistochemical staining were evaluated. Statistical analysis was conducted using R software. Several novel computer models on prediction of MP grading were established and validated. Results: Among 1 053 patients who accepted neoadjuvant treatment, 316 patients (316/1 053, 30%) were evaluated as MP5 postoperatively, and 737 patients (737/1 053, 70%) did not meet MP5 level. MP5 had significant association with histological grade, ER and PR expression, HER2 status, Ki-67 index and molecular classification (P<0.05). Univariate/multivariate logistic regression analyses further showed that the above clinicopathological features were also independent influencing factors of MP5 grade; five-fold cross-validation was used to evaluate the performance of the models, and the sensitivity and specificity of different models were obtained. Conclusions: MP grading of invasive breast carcinoma NOS after neoadjuvant treatment is associated with high histological grade, negative ER and PR expression, HER2 positivity, high Ki-67 index and molecular classification, which are independent influence factors. GBM model recommended through comparison can provide some help for clinical diagnosis and treatment.
