Objective To explore the risk factors of upper limb lymphedema after breast cancer surgery and to construct and validate a risk prediction visualized model.

Methods A retrospective analysis was performed on the clinical data of 528 breast cancer patients who underwent surgery. The patients were randomly divided into training set(352 cases) and validation set (176 cases). The training set patients were divided into lymphedema group(67 cases) and non-lymphedema group(285 cases) based on whether they had upper limb lymphedema. Logistic regression analysis was used to screen risk factors for upper limb lymphedema after breast cancer surgery. A risk prediction visualized model was constructed and validated using R software.

Results The incidence of upper limb lymphedema after breast cancer surgery was 20.08% (106/528). Multivariate Logistic regression analysis showed that high body mass index (BMI), hypertension, TNM stage Ⅲ a, bilateral lesions, level Ⅲ of axillary lymph node dissection, and postoperative radiotherapy and chemotherapy were independent risk factors for upper limb lymphedema after breast cancer surgery (P < 0.05). A risk prediction nomogram model was constructed based on above six independent risk factors. The calibration curve of the model in the training set and validation set both fitted the ideal curve well; the receiver operating characteristic curve analysis showed that the area under the curve of the nomogram model in the training set and validation set for predicting upper limb lymphedema after surgery was 0.950 and 0.886, respectively; the decision curve analysis showed that the overall net benefitof the model in predicting upper limb lymphedema after surgery in the training set and validation set was higher than that of all patients receiving full intervention or no intervention.

Conclusion High BMI, hypertension, stage Ⅲ a of TNM, bilateral lesions, level Ⅲ of axillary lymph node dissection, and postoperative radiotherapy and chemotherapy are independent risk factors for upper limb lymphedema after breast cancer surgery. The risk prediction nomogram model constructed based on these factors has good predictive performance.