A model for malignancy probability prediction of adnexal masses.

ABSTRACT

OBJECTIVE: To develop a model for pre-operative malignancy probability determination in a patient with an adnexal tumor or tumors by the application of multivariate logistic regression analysis to variables at the time of pelvic sonography. METHOD: Pre-operative ultrasound examination including Doppler analysis was performed on 117 consecutive women scheduled for surgery because of an adnexal mass or masses. Each tumor was classified as probably benign or malignant using a subjective evaluation system on the gray-scale morphological images. Then, Doppler sonography was carried out. The resistance index (RI) and pulsatility index (PI) of the vessel with the highest velocity were recorded. Multivariate logistic regression analysis was performed with the histological outcome as the dependent variable. Independent variables included patient's age, menopausal status, gray-scale morphological data, RI and PI. The probability of malignancy was formulated from statistical analysis. RESULTS: There were 117 women included in the study, 83 (71%) with histologically benign and 34 (29%) with histologically malignant ovarian tumors. Regression analysis on the five variables resulted in the retention of only patient's age, morphological data and RI as significant contributing factors for malignancy prediction. The probability of malignancy was 1/(1+e(-z)) where e was the base value for natural logarithms and z was the regression equation -3.6355 + 1.8028 (age) + 2.1047 (morphological data) + 2.9816 (RI). CONCLUSION: A model for estimation of probability of malignancy for an adnexal tumor was derived using multivariate logistic regression analysis. The prediction should be more accurate than that from either gray-scale ultrasound imaging or Doppler velocimetry alone. The test of the model is now on-going.