Diagnostic rentability of IOTA models for differentiating between benign and malignant complex adnexal masses

Abstract

Purpose

To evaluate the diagnostic accuracy of the International Ovarian Tumor Analysis (IOTA) Logistic Regression Model 1, 2 (LR1, LR2) ADNEX model and IOTA Simple Rules, in the pre-surgical evaluation of ovarian tumors classified as complex adnexal masses.

Methods

This is a cross-sectional observational study of diagnostic accuracy. We will select patients, who undergo surgical intervention due to adnexal mass with indeterminate, intermediate or high suspicion of malignancy (GI-RADS 4–5), as assessed by an expert ultrasound operator. We analyzed and compared the diagnostic performance and predictive capacity of the different models in the studied population, and also we analyzed each model by creating subgroups based on menopausal status.

Results

One hundred thirty five malignant masses (45%), one hundred forty benign (46.7%) and twenty five border line (8.3%) were included.

LR1 and LR2 models, and ADNEX were applicable to all lesions; however, in 72 lesions (24%), the simple rules were inconclusive.

We observed better performance for LR1 and simple rules, based on the following results: Sensitivity: LR1 91%. Specificity: simple rules 86%. PPV: simple rules 79%. NPV: LR1 88%.

Conclusions

Our study suggests that the subgroup of patients with complex adnexal masses, the IOTA risk stratification through LR1 shows higher sensitivity for risk stratification of malignancy, while simple rules has the highest specificity and diagnostic accuracy. However, it is inconclusive in one out of every four adnexal masses. Additionally, LR1–LR2 and ADNEX do not show significant differences in diagnostic accuracy.