Risk factors and prediction model for persistent breast-cancer-related lymphedema: a 5-year cohort study.

PURPOSE: Breast-cancer-related lymphedema (BCRL) can be a transient or persistent condition. The aims of this study were to (1) identify and weigh the risk factors for persistent lymphedema (PLE) among all patients with BCRL and (2) establish a prediction model for the occurrence of PLE.

METHODS: A cohort of 342 patients with BCRL with a median follow-up of 5 years after the onset of swelling was analyzed. PLE was defined as a hardening of the subcutaneous tissue, the persistence of the circumferential difference (CD) between arms, or a flare-up of swelling during follow-up. Multiple logistic regression was used to identify risk factors for PLE, including tumors, treatments, and patient-related factors. The prediction accuracy of the model was assessed using the area under the receiver operating characteristic curve (AUC).

RESULTS: Of the 342 patients with BCRL, 229 (67%) had PLE. Multiple logistic regression analysis revealed that the number of lymph node metastases (p = 0.012), the maximal CD between arms at the first occurrence of swelling (p < 0.001), and the largest difference during follow-up (p < 0.001) were significant predictors for PLE. The corresponding AUC was 0.908. Although inclusion of body weight gains (p = 0.008) and maximal CD at the latest follow-up (p = 0.002) increased the analytical accuracy (AUC = 0.920), the resulting AUC values (p = 0.113) were not significantly different.

CONCLUSIONS: BCRL is persistent in two thirds of patients. Patients with more lymph node metastases, weight gain, and larger CD since the onset of swelling and during follow-up have an increased likelihood of developing PLE.