A risk prediction model for medical treatment failure in tubal pregnancy

Objective: Methotrexate is an alternative treatment for tubal pregnancy. However, despite initial treatment, similar to 15% of women eventually require surgery. This study aimed to identify the risk factors for medical treatment failure in tubal pregnancy and apply them to a risk prediction model. Study design: This single-center retrospective cohort study included 123 participants initially treated medically for tubal pregnancy between January 2006 and December 2015. Logistic regression analysis was used to construct a risk prediction model (visually presented as a nomogram) for medical treatment failure. Model performance was assessed using discrimination and calibration. The medical treatment failure rate was 36.6%. The prediction model integrated the presence of a gestational sac, ectopic mass size, and follow-up beta-human chorionic gonadotropin levels above cut-off values on days 4 and 7. The model used the following cut-off values: increased beta-human chorionic gonadotropin levels by 1028.6 mIU/mL, 1.0457-fold higher than baseline level on day 4; and increased beta-human chorionic gonadotropin levels by 1233 m1U/mL, 1.3025-fold higher than baseline level on day 7. Results: The corresponding areas under the receiver-operating characteristic curves were 0.8135 (95% confidence interval, 0.733-0.893) for the day 4 model and 0.8600 for the day 7 model (95% confidence interval, 0.788-0.932). Comparison of the day 4 and 7 models revealed no significant difference in their predictive abilities (P=0.4318). Conclusions: This model identified a substantial proportion of the participants who experienced medical treatment failure for tubal pregnancy. It was visualized as a nomogram, facilitating clinical application. (C) 2018 Elsevier B.V. All rights reserved.