Prediction Models for Lower Limb Lymphedema after Surgery in Patients with Gynecological Malignancies: A Systematic Review

The incidence of gynecological malignancies continues to increase worldwide. Lower limb lymphedema is perhaps the most dreaded long-term complication related to gynecological malignancy surgery. Identifying the early risk factors for lower limb lymphedema can facilitate targeted prevention and improve the prognosis of patients. To systematically review and evaluate existing studies on prediction models for lower limb lymphedema after surgery in patients with gynecological malignancies. PubMed, Embase, Web of Science, the Cochrane Library, CINAHL, China National Knowledge Infrastructure, Wanfang Database, China Science and Technology Journal Database, China Biomedical Literature Database, ProQuest Dissertations & Theses, medRxiv, and Google Scholar were systematically searched from inception to December 1, 2024. Studies were appraised critically, and data were extracted by two authors independently based on the Prediction Model Risk of Bias Assessment Tool and Data Extraction for Systematic Reviews of Prediction Modeling Studies. A total of 9 studies were included, involving 10 relevant prediction models. Five studies reported calibration; validation involved internal ( n = 7) or both ( n = 2). Discrimination varied across models (the area under the curve [AUC] range: 0.63–1.00), with 9 of 10 models reporting AUC values >0.70. However, calibration reporting was incomplete, and all studies were rated as high risk of bias. Although all studies demonstrated good applicability, all were rated as high risk of bias. Across all included studies, age, body mass index, hypertension, diabetes, tumor stage, postoperative drainage time, lymph node dissection, and radiotherapy were the most frequently reported predictors. Among them, lymph node dissection–related variables were included in eight of the ten models, while radiotherapy was included in 7 of 10 models. Although several models reported moderate to good discrimination, overall methodological quality was limited. The predominance of high risk of bias, scarce external validation, and restricted geographic diversity constrains the strength of the current evidence. More rigorously designed and externally validated models are needed before routine clinical implementation.