Validation of Caprini risk model and development of a biomarker-driven predictive model for postoperative venous thromboembolism in gynecologic cancer surgery

Background Gynecologic cancer patients face a notably high postoperative venous thromboembolism (VTE) incidence, while the current standard Caprini Scale demonstrates limited discriminative power in this population. This study therefore aimed to evaluate the Caprini Scale's predictive performance and develop an enhanced predictive model incorporating dynamic clinical variables and cancer-specific biomarkers for optimized risk stratification and prophylaxis guidance. Methods A single-center retrospective cohort study analyzed 140 gynecologic cancer patients (28 VTE cases, 112 controls) undergoing surgery between April 2021 and April 2025. Preoperative and postoperative VTE risk was dynamically evaluated using the 2013 Caprini Scale. Clinical variables, including D-dimer levels, perioperative transfusion, and Caprini score dynamics, were analyzed via Lasso regression and multivariable logistic regression to construct a predictive nomogram. Model performance was assessed through discrimination, calibration and clinical utility. Results The study revealed no significant differences in preoperative or postoperative Caprini scores between groups, but the postoperative-preoperative score difference was significantly higher in VTE patients (5.09 ± 1.23 vs. 4.27 ± 1.28, P  = 0.0067), with 69.6% of VTE cases classified as postoperative "very high risk" versus 58.4% of controls ( P  = 0.0207). Multivariate analysis identified increased age, elevated D-dimer levels, perioperative transfusion, and Caprini score change as independent predictors. The integrated model demonstrated robust discrimination (training AUC = 0.841; validation AUC = 0.812), excellent calibration, and clinical utility. Conclusion The novel model integrating age, D-dimer elevation, perioperative transfusion, and Caprini score dynamics offered improved risk stratification. Future multicenter prospective studies with larger cohorts are required to validate and refine this model, particularly incorporating cancer stage-specific adjustments and long-term outcome monitoring.