Ovarian Cancer Lymphocele Risk: A Delphi-Based Retrospective Model Development and Validation

IntroductionLymphocele is a common complication following gynecologic cancer surgery. Given that lymphadenectomy for ovarian cancer differs from that for other gynecologic malignancies, it is essential to determine whether the risk factors for postoperative lymphocele are specific to this patient population.MethodA Delphi study was conducted to identify risk factors for postoperative lymphocele in ovarian cancer and to establish eligibility criteria for an institutional development cohort. Based on these findings, we retrospectively collected clinicopathological data from patients who underwent ovarian cancer staging surgery at our institution between 2019 and 2024. Postoperative lymphocele and baseline clinical variables were compared using chi-square tests and regression analyses. A nomogram for predicting lymphocele risk was developed using the least absolute shrinkage and selection operator (LASSO) method, and its performance was evaluated using calibration curves and decision curve analysis (DCA).ResultsFollowing three rounds of the Delphi survey, 13 risk factors met the criteria for expert consensus and boundary definition. Among 324 patients with ovarian cancer, multivariate analysis revealed that open surgery, a higher number of retrieved lymph nodes, and adjuvant therapy were independently associated with lymphocele. LASSO regression selected age, hypertension, diabetes, anemia, surgical approach, lymphadenectomy details (number and extent), lymph node metastasis, drainage, and adjuvant therapy as predictors for the nomogram. The model demonstrated good predictive performance (C-index: 0.784), which was confirmed by calibration and decision curve analyses.ConclusionBy incorporating the Delphi method into the development of our prediction model, we achieved a more comprehensive identification of risk factors for postoperative lymphocele in ovarian cancer than previously reported. We anticipate that this innovative model will facilitate broader adoption in future clinical practice and research.