Objective: Patient characteristics, CA125 level and two-dimensional (2D) ultrasonography can be used to predict the probability of malignancy of an ovarian mass. Three-dimensional (3D) ultrasonography might also contribute to the prediction of malignancy. We evaluated whether addition of 3D features to a diagnostic model could improve the discriminative capacity of the model.
Methods: This multicenter prospective study was approved by the institutional review board. Women with an adnexal mass scheduled for surgery underwent 2D and 3D ultrasonographic examination in the week prior to surgery. Stepwise logistic regression was used to construct two models for the prediction of malignancy: a model based on patient characteristics, level of CA125 and 2D ultrasonography and a second model based on patient characteristics, level of CA125, 2D and 3D ultrasonography. Receiver operator characteristic (ROC) curve analysis was used to compare the capacity of the two models to discriminate between benign and malignant adnexal masses.
Results: We included 181 women with an adnexal mass, of which144 were benign and 37 showed malignancy on histopathology. The 3D model discriminated better between benign and malignant adnexal masses than the 2D model (areas under the ROC curve of 0.92 and 0.82, respectively, p=0.02). The calibration of both models was good.
Conclusion: In the assessment of the ovarian mass, the use of 3D ultrasonography significantly improves the prediction of malignancy as compared to patient characteristics and 2D ultrasonography.