Objective: To establish a predictive model for pathological complete response (pCR) in operable primary breast cancer after neoadjuvant chemotherapy (NAC).
Study design: Observational study. Place and Duration of the Study: Breast Centre, Shunyi District Health Care Hospital for Women and Children of Beijing, Beijing, China, from January 2010 to June 2023.
Methodology: Four hundred and fourteen operable invasive breast cancer patients who received NAC were included in this study. After a random assignment at a ratio of 7:3, 289 patients in the training set were analysed for model building, and the remaining 125 patients in the test set were used for validation. The definition of pCR was the absence of residual invasive disease in either the breasts or the axillary lymph nodes (ypT0 / is ypN0). After multivariate logistic regression analysis, a nomogram was drawn. In the validation phase, the receiver operating characteristic (ROC) curve and AUC were used for evaluation of discrimination, while the calibration plot and Hosmer-Lemeshow test for calibration. Additionally, a decision curve was drawn.
Results: A model containing 8 variables, including BMI, tumour size, histological grade, HR, HER2, axilla status, chemotherapy cycles, and regimens was built. After validation, the model had moderate discriminatory power [AUC, 0.831; 95% CI (0.733, 0.928)]. Calibration curve and Hosmer-Lemeshow goodness of fit (GOF) test (p = 0.1645) demonstrated that the model fitted well. Meanwhile, the decision curve analysis revealed that the model was beneficial to patients.
Conclusion: Model containing BMI, tumour size, histological grade, HR, HER2, axilla status, chemotherapy cycles, and regimens showed moderate discrimination and calibration abilities in predicting pCR.
Key words: Breast neoplasms, Neoadjuvant therapy, Surgery, Pathology, Nomogram.