Predicting the efficacy of vaginal misoprostol in the management of early pregnancy failures: A signal center study

Abstract

Introduction: Early pregnancy loss is a situation commonly experienced by pregnant women. Management includes expectant, medical and surgical evacuation. Medical management is effective and relatively safe option and involves use of vaginal misoprostol which was recently approved to use in Sri Lanka. Proper patient selection for the medical management will improve the success rate.

Objectives: Aim is to develop a predictive model using different factors to determine the efficacy of vaginal misoprostol in medical management of early pregnancy losses.

Method: This was a single center retrospective descriptive study and was carried out in Teaching Hospital, Mahamodara, Galle. Eligibility criteria was medically managed early pregnancies (<12 weeks of gestation) adherent to Health Ministry guidelines. Data was retrieved using bed head tickets and was entered to data retrieval sheet. Orange Statistics 3.26, a leading free and open-source software was used for the development, testing and validation of predictive model. One hundred thirty four cases, approximately 80% of total (5% precision level), were randomly selected using randomization to train, build, score and evaluate predictive models. The remaining 20% of the cases were used once to validate each model.

Results: 167 women were recruited for the analysis. Overall success rate of vaginal misoprostol was 62.3%. Majority of women (76.6%) didn’t have significant adverse effects for misoprostol. Out of the twenty clinical and radiological parameters, only the presence of vaginal bleeding weakly associated with the success of medical management. Other factors were not statistically significant enough to develop a predictive model.

Conclusion: Management of early pregnancy losses with misoprostol was relatively safe and successful in majority of patients. However, it was unable to develop a predictive model with selected factors in this study, except presence of per vaginal bleeding prior to treatment yield high info gain.