D18. Risk of Breast Cancer-Related Lymphedema: Development and Validation of a Clinically-Relevant, Patient-facing Prediction Model

Abstract

PURPOSE: Breast cancer related lymphedema (BCRL) may be prevented with immediate lymphatic reconstruction (ILR). However, prioritizing patients for ILR based on risk is not possible since current predictive models include obscure datapoints and omit key risk factors. The purpose of this study was to create a simple predictive model for BCRL. METHODS: A prospective cohort of patients treated for breast cancer with either ALND or SLNB and followed with arm volume measurements was used to develop two predictive models via logistic regression: a preoperative model (model 1) and a postoperative model (model 2). Model 1 was externally validated using a cohort of 34,438 patients with an ICD diagnosis of breast cancer. RESULTS: The prospective cohort consisted of 1,882 patients with an 11.6% rate of BCRL at an average follow-up of 3.9 (SD 1.8) years. Model 1 included: age, weight, height, race, ALND or SLNB, any radiation therapy, and any chemotherapy. Model 2 included: age, weight, ALND or SLNB, any chemotherapy, and current arm swelling. Accuracy was 75.0% for model 1 (sensitivity 71.6%, specificity 75.4%, AUC 78.9%) and 87.7% for model 2 (sensitivity 66.1%, specificity 90.5%, AUC 85.7%) at a cutoff of 19.0%. Both models demonstrated high AUCs on external (model 1, 71.4%) or internal (model 2, 82.5%) validation. CONCLUSION: Preoperative and postoperative prediction models for BCRL can be highly accurate and clinically-relevant tools comprised of inputs that are easily accessible to patients and providers. These models should be used to identify at-risk patients for perioperative risk reduction measures and surveillance.