Upper Extremity Lymphedema​​​: Performance of a Risk Prediction Model in Women Undergoing Surgery for Breast Cancer.

Abstract

Introduction: Upper extremity lymphedema is a common complication following breast cancer treatment, and supportive tools are valuable for screening, preventive strategies, and early therapeutic intervention. In 2020, Kwan and colleagues proposed a formula to predict the risk of lymphedema.

Objective: To evaluate the performance of a predictive model for lymphedema development (classified as low, moderate, or high risk) in a population of Brazilian women.

Methods: This retrospective study included 190 women who underwent surgery for breast cancer, with at least 6 months of postoperative follow-up. Clinical data were collected, and arm volumetry was performed. The Kwan model was applied using the following variables: age, body mass index (BMI), mammographic density, number of positive lymph nodes, and type of axillary surgery (sentinel lymph node biopsy or axillary lymph node dissection). Predicted values were compared with observed outcomes and categorized into low, moderate, or high risk.

Results: The model demonstrated low predictive accuracy (RMSE=223.54; R2=-0.18). The overall accuracy rate was 76%, but sensitivity was low (48%), with a positive predictive value of 22%. Specificity was acceptable (79%), and the negative predictive value was high (92%). However, the low incidence of lymphedema in the study population should be considered.

Conclusion: The model showed limited performance in predicting positive cases of lymphedema, but may have utility as a rule-out (screening) tool. New models incorporating broader clinical variables and technologies, such as machine learning, may provide improved positive predictive accuracy.