Background: Different ultrasound based models, rules and descriptors are used to characterise ovarian masses. LR2 is a logistic regression model with six variables to assess the preoperative risk of ovarian malignancy. Simple rules (SR) are ten features of an adnexal mass that may be identified using ultrasonography. Simple descriptors (SD) are five ultrasound characteristics and one based on age and serum CA 125 that enable an intuitive classification of adnexal masses.
Objectives: The main aim was to assess the performance of the conventional (Risk of Malignancy Index RMI) and the IOTA (International Ovarian Tumour Analysis Group) models (LR2, SR and SD) and subjective assessment (SA) by examiners of varied ultrasound training and experience (level II).
Methods: We carried out a prospective single-blinded randomised controlled trial (IOTA 4a) to investigate the referral pattern following the use of RMI and LR2. In IOTA 4b, diagnostic performance measures for LR2, RMI, SR, SD and SA were compared in a three-year multicentre prospective observational study. A prospective comparison was also carried out between LR2 and ROMA (Risk of Ovarian Malignancy Algorithm).
Results: In the IOTA 4a RCT, LR2 correctly classified 100% of benign masses as low risk compared to 70% for RMI (difference of 30%. p<0.0001). In IOTA 4b study, a higher AUC for LR2 compared to RMI was observed in premenopausal women (0.93 and 0.83 respectively. p=0.05). Overall the combination of SD and SR were able to characterise 89% of masses (DOR of 348). Subjective assessment by level II ultrasound examiners had a DOR of 85.56. In relation to ROMA, the AUC of LR2 for discriminating between benign and malignant adnexal masses (0.952) was higher than that of ROMA (0.893) (p=0.0004).
Conclusion: The overall test performance of IOTA LR2, SR and SD was maintained in the hands of examiners with varying levels of training and experience. This performance was generally better than the RMI. LR2 showed a better diagnostic performance than ROMA.