Performance of Selected Models for Predicting Malignancy in Ovarian Tumors in Relation to the Degree of Diagnostic Uncertainty by Subjective Assessment With Ultrasound.

OBJECTIVES

The study's main aim was to evaluate the relationship between the performance of predictive models for differential diagnoses of ovarian tumors and levels of diagnostic confidence in subjective assessment (SA) with ultrasound. The second aim was to identify the parameters that differentiate between malignant and benign tumors among tumors initially diagnosed as uncertain by SA.

METHODS

The study included 250 (55%) benign ovarian masses and 201 (45%) malignant tumors. According to ultrasound findings, the tumors were divided into 6 groups: certainly benign, probably benign, uncertain but benign, uncertain but malignant, probably malignant, and certainly malignant. The performance of the risk of malignancy index, International Ovarian Tumor Analysis assessment of different neoplasias in the adnexa model, and International Ovarian Tumor Analysis logistic regression model 2 was analyzed in subgroups as follows: SA-certain tumors (including certainly benign and certainly malignant) versus SA-probable tumors (probably benign and probably malignant) versus SA-uncertain tumors (uncertain but benign and uncertain but malignant).

RESULTS

We found a progressive decrease in the performance of all models in association with the increased uncertainty in SA. The areas under the receiver operating characteristic curve for the risk of malignancy index, logistic regression model 2, and assessment of different neoplasias in the adnexa model decreased between the SA-certain and SA-uncertain groups by 20%, 28%, and 20%, respectively. The presence of solid parts and a high color score were the discriminatory features between uncertain but benign and uncertain but malignant tumors.

CONCLUSIONS

Studies are needed that focus on the subgroup of ovarian tumors that are difficult to classify by SA. In cases of uncertain tumors by SA, the presence of solid components or a high color score should prompt a gynecologic oncology clinic referral.