The success rates of medicinal treatments for EPL were highly variable, both between and within therapy regimens. Although certain predictors were established up to this point, no clinical prediction model has been developed. Researchers analyzed a multicenter randomized controlled experiment conducted at 17 institutions between June 28, 2018, and January 8, 2020.  Women with a nonviable pregnancy between 6 and 14 weeks of gestation chose medical treatment after at least 1 week of expectant care. Literature and expert opinion were used to select potential predictors for the successful medical therapy of EPL. Investigators conducted an internal validation of the prediction model using bootstrapping methods. About 237 out of 344 women were treated successfully with medical EPL (68.9%). The model contained the following variables: mifepristone use, body mass index, the number of previous uterine aspirations, and the existence of modest clinical signs (little vaginal bleeding or a few abdominal cramps) at treatment initiation. The model can moderately differentiate between treatment success and failure, with an AUC of 67.6% (95% CI=64.9–70.3%). Comparing projected and observed probabilities of success, the model fits well, but it might have underestimated treatment success in women with a predicted likelihood of success of approximately ∼70%. Most women (90.4%) were Caucasian, which might have led to suboptimal model performance in non-Caucasian populations. The study group could not externally validate the model’s performance and clinical impact, and the prediction model’s accuracy was only moderate at 0.67.

Source:academic.oup.com/humrep/article/37/5/936/6554030