107 Discussion Non-invasive diagnosis of adenomyosis still remains challenging. Despite great efforts to create a non-invasive classification system for adenomyosis, there is as yet no clinically implemented prediction tool for MRI diagnosis and no external validation studies were performed as far as we know (16,43). Therefore, in this retrospective cohort study, we performed the first external validation of a promising developed prediction model by Rees et al. to make this model applicable for diagnosis and patient counselling in daily practice. The model was developed with a patient population from the Catharina Hospital Eindhoven in the Netherlands. An explanation for the chosen variables in this model is described in previous study (204). When applying the model to a comparable population from a different centre, the model showed good performance with an AUC of 0.831 (95%CI 0.761 – 0.901). The model was well-calibrated for our population. As in the initial study, presence of HSI foci was the most significant predictive variable. To assess the generalisability of the developed prediction model, baseline characteristics of the external validation dataset were compared with the original dataset used in the development of the model. The age and BMI at MRI of both datasets do not show great difference, only one year or one point, which is arguably of little clinical relevance. In contrast to the initial study, this validation population did not show a statistically significant difference in a history of curettage between the groups of patients with and without a histopathological diagnosis of adenomyosis (12.8% vs. 10.3% respectively, p = 0.579). Only 22 patients in the entire external validation cohort had a history of curettage compared to 43 patients in the original dataset (22.1% in adenomyosis patients, p=0.002). A potential explanation for this difference in the prevalence of curettage could be that the surgery was completed elsewhere and thus not registered as clearly in the hospital database. The mean number of pregnancies was slightly higher in the adenomyosis group in the original dataset compared to the adenomyosis group in the external validation dataset (3.0 vs. 2.34 respectively). Despite the fact that this cohort did not show significance between the groups for history of curettage, adenomyosis has been associated with infertility and increased risk of miscarriage (81,85,208). Due to this increased miscarriage risk, women with adenomyosis may need to undergo curettage more often, which is why this was included in the model.
RkJQdWJsaXNoZXIy MTk4NDMw