Clinical Utility of Risk Models to Refer Patients with Adnexal Masses to Specialized Oncology Care: Multicenter External Validation Using Decision Curve Analysis
Abstract
Purpose: To evaluate the utility of preoperative diagnostic models for ovarian cancer based on ultrasound and/or biomarkers for referring patients to specialized oncology care. The investigated models were RMI, ROMA, and 3 models from the International Ovarian Tumor Analysis (IOTA) group [LR2, ADNEX, and the Simple Rules risk score (SRRisk)].
Experimental Design: A secondary analysis of prospectively collected data from 2 cross-sectional cohort studies was performed to externally validate diagnostic models. A total of 2,763 patients (2,403 in dataset 1 and 360 in dataset 2) from 18 centers (11 oncology centers and 7 nononcology hospitals) in 6 countries participated.
Excised tissue was histologically classified as benign or malignant. The clinical utility of the preoperative diagnostic models was assessed with net benefit (NB) at a range of risk thresholds (5%-50% risk of malignancy) to refer patients to specialized oncology care.
We visualized results with decision curves and generated bootstrap confidence intervals. Results: The prevalence of malignancy was 41% in dataset 1 and 40% in dataset 2.
For thresholds up to 10% to 15%, RMI and ROMA had a lower NB than referring all patients. SRRisks and ADNEX demonstrated the highest NB.
At a threshold of 20%, the NBs of ADNEX, SRrisks, and RMI were 0.348, 0.350, and 0.270, respectively. Results by menopausal status and type of center (oncology vs. nononcology) were similar.
Conclusions: All tested IOTA methods, especially ADNEX and SRRisks, are clinically more useful than RMI and ROMA to select patients with adnexal masses for specialized oncology care.