92 ABSTRACT Introduction: Non-invasive diagnosis of adenomyosis remains challenging as there is still no consensus on diagnostic criteria. This indicates a current need for a non-invasive diagnostic tool. This study aims to externally validate a previously developed prediction model by Rees et al. to predict likelihood of histopathological adenomyosis diagnosis based on MRI Materials and methods: This single-centre, observational, retrospective cohort study took place in a non-academic teaching hospital in the Netherlands. Patients were included if they had undergone a hysterectomy on suspicion of benign pathology between 2014 and 2022 with a pre-operative pelvic MRI. The MRIs were retrospectively assessed for adenomyosis markers. The developed model was applied to the patients in this external dataset. The prediction model utilized several clinical factors and MRI factors such as mean junctional zone (JZ) thickness, JZ Dif->5mm, JZ/myometrium ratio >0.40, and presence of high signal intensity (HSI) foci. The predictive performance of the model was assessed using the receiver operating characteristic (ROC) curve analysis and its calibration and discrimination were evaluated. Results: Out of 195 patients, 78 patients (40%) received a diagnosis of adenomyosis based on histopathology. The previously developed model showed good external validity in this population with an Area Under the Curve (AUC) of 0.831 (95%CI 0.761 – 0.901). As for calibration, the Hosmer-Lemeshow test did not show significant difference between the predicted and observed outcome (chi-square 4.398, p = 0.820). Conclusions: The developed model showed good to excellent discriminative performance in this external cohort in predicting the adenomyosis diagnosis based on MRI in individual patients. Given the model’s accurate performance after external validation, its implementation in daily clinical practice could be considered.
RkJQdWJsaXNoZXIy MTk4NDMw