Analysis of Risk Factors for Tumor Regression Patterns After Neoadjuvant Chemotherapy in Breast Cancer and the Construction of Its Predictive Model
10.11714/jsysu.med.YX20260083
- VernacularTitle:乳腺癌新辅助化疗后肿瘤退缩模式的影响因素分析及其预测模型的构建
- Author:
Liming XIA
1
;
Wei GUO
2
;
Lin WANG
3
;
Kunjian XIA
2
Author Information
1. Department of General Surgery， Leping People's Hospital， Jingdezhen 333000， China
2. Department of General Surgery， The Second Affiliated Hospital of Jiujiang University， Jiujiang 332005， China
3. Department of Breast Surgery， The Second Affiliated Hospital of Nanchang University， Nanchang 330006， China
- Publication Type:Journal Article
- Keywords:
breast cancer;
neoadjuvant chemotherapy;
tumor regression pattern;
predictive model
- From:
Journal of Sun Yat-sen University(Medical Sciences)
2026;47(4):756-766
- CountryChina
- Language:Chinese
-
Abstract:
ObjectiveTo explore the influencing factors of tumor regression patterns after neoadjuvant chemotherapy （NACT） in breast cancer， construct and validate a predictive model for tumor regression patterns. MethodsA total of 1 314 breast cancer patients who received NACT between March 1， 2019， and March 1， 2024， were selected as study subjects and randomly divided into a validation cohort （n=301） and a study cohort （n=603）. Within the study cohort， patients were further stratified based on post-NACT tumor regression patterns into a concentric regression group（case group）（n=387） and a non-concentric regression group（control group）（n=216）. General clinicopathological data were collected and compared between the two groups. Unconditional logistic regression was used to analyze the influencing factors of tumor regression patterns post-NACT and establish a predictive model. The Hosmer-Lemeshow test and receiver operating characteristic （ROC） curve were employed to evaluate the model's goodness-of-fit and predictive performance， followed by external validation. ResultsAmong all subjects in the study cohort， 174 patients （28.9%） achieved pathological complete response （pCR）， 213 （35.3%） showed unifocal regression， 92 （15.3%） had multifocal regression， 78 （12.9%） presented main residual lesions with satellite lesions， 43 （7.1%） had stable disease， and 3 （0.5%） experienced disease progression. Significant differences were observed between the two groups in age， estrogen receptor（ER）， progesterone receptor（PR）， Ki-67， human epidermal growth factor receptor 2（HER2）， histological grade， prognostic nutritional index（PNI）， and chemotherapy regimen （all P ＜ 0.05）. Logistic regression analysis revealed that age （OR=0.563， 95% CI： 0.358–0.885）， ER （OR=0.521， 95% CI： 0.350–0.773）， PR（OR=0.552， 95% CI： 0.379–0.806）， HER2（OR=3.729， 95% CI： 2.488–5.590）， Ki-67 （OR=1.804， 95% CI： 1.071–3.039）， PNI （OR=2.285， 95% CI： 1.307–3.997） and histological grade （Ⅲ） （OR=2.194 95% CI： 1.283–3.751） were independent influencing factors for concentric regression post-NACT （all P ＜ 0.05）. The Hosmer-Lemeshow test for the logistic regression model yielded a P-value of 0.927， and the area under the ROC curve was 0.731 （95%CI： 0.690-0.772， P ＜ 0.001）. External validation demonstrated that the area under the ROC curve of this prediction model was 0.764 （95% CI： 0.709–0.820， P＜0.001）， with a sensitivity of 75.0% and a specificity of 65.7%. The Youden's index was 0.407. ConclusionThis logistic regression model exhibits high predictive value and provides a reference for clinicians to predict tumor regression patterns in breast cancer patients after NACT.
