OP06.01: Head‐to‐head comparisons of RMI and ADNEX model to diagnose ovarian malignancy: systematic review and meta‐analysis

Conduct a systematic review and meta-analysis to compare the performance of the ADNEX model and the RMI regarding discrimination between benign and malignant adnexal masses (AUC) and clinical utility (PROSPERO, CRD42023449454). We searched the literature from inception to 10/11/2023 for studies that externally validated ADNEX (with or without CA125) and RMI in the same patients and reported their performance. All articles were screened, and data was extracted by two independent reviewers. Disagreements were resolved through discussion. We performed a meta-analysis of the AUC and net benefit for studies presenting those metrics. Net benefit was summarised through p(useful), which is the probability that a model is clinically useful in a new centre. We conducted independent meta-analyses for populations including and excluding non-surgically managed patients. 10 studies (7700 tumours) across 14 countries and 27 centres were included. Median study sample size was 306 (range 100-4905). The pooled AUC for ADNEX varied between 0.91 and 0.94 depending on the version (with or without CA125) and population (see table 1). For RMI, the pooled AUC varied between 0.84 and 0.89. In operated patients, p(useful) was between 92 and 96% for ADNEX and between 12 and 15% for RMI. Including also conservatively managed patients, p(useful) was 99% for ADNEX and 58% for RMI. ADNEX had higher discrimination between benign and malignant tumours than RMI. Importantly, ADNEX was almost always clinically useful to support decision-making. Clinical utility of RMI was dramatically lower. ADNEX without CA125 versus RMI ADNEX with CA125 versus RMI