Development and Validation of a Model Predicting Malignant Potential of Adnexal Masses in Areas with Scarcity of Ultrasound Resources

Oncology. 2024 Dec 23:1-22. doi: 10.1159/000542952. Online ahead of print.

Abstract

Introduction: Appropriately stratifying the risk of adnexal masses is of great importance. Many diagnostic algorithms have been devised, most of which rely on ultrasound features. However, some remote areas lack trained sonographers. This study aimed to develop an alternative model to distinguish between malignant and benign adnexal masses in resource-constrained settings using clinical information rather than ultrasound data.

Methods: The study included women diagnosed with an adnexal tumor and scheduled for surgery between 2020 and 2023. Participants were divided into two groups based on histopathology reports: those with malignant adnexal masses and those with benign ones. Univariate and multivariate logistic regression analyses were used to identify independent predictors of adnexal mass malignancy. The training set yielded a nomogram model, which was then validated in the validation set. The model's effectiveness was evaluated using receiver operating characteristic (ROC), calibration, and clinical decision analysis (DCA) curves.

Results: We randomly assigned 550 participants to the training and the validation sets in an 8:2 ratio. Logistic regression analyses identified age (OR = 1.044, P = 0.003), abdominal distension (OR = 0.139, P < 0.001), serum CA125 (OR = 1.007, P < 0.001), and serum carcinoembryonic antigen (CEA) (OR = 1.291, P = 0.004) as independent risk factors for predicting malignant adnexal tumors. A nomogram was constructed using these factors. The ROC curve showed an area under the curve (AUC) of 0.846 (95% confidence interval [CI]: 0.783, 0.908) in the training set and 0.817 (95% CI: 0.668, 0.966) in the validation set. The calibration curve showed good consistency between model predictions and actual outcomes. The DCA curve demonstrated a considerable clinical advantage afforded by the model.

Conclusions: The logistic regression model can aid gynecologists-particularly those in areas with limited access to skilled sonographers-in identifying patients at high risk and implementing appropriate management strategies.