Predicting outcomes of in vitro fertilization in women with infertility

Abstract

Background. The development of predictors of the infertility treatment effectiveness by in vitro fertilization is one of the most important tasks of modern medicine. Despite numerous studies, there is still no single point of view regarding the significance of the influence of certain factors on the in vitro fertilization results.
 Aim. Creation of a predictive model for the outcomes of in vitro fertilization for women suffering from infertility.
 Material and methods. A retrospective analysis of 518 case histories of patients suffering from infertility and undergoing an in vitro fertilization program from 2015 to 2020 at the Central Clinic in Baku was carried out. Of these, 234 women (main group) became pregnant after in vitro fertilization, while 284 (control group) did not. Due to this, an individual card and an algorithm for examining patients in order to predict the results of in vitro fertilization were developed. At the prospective stage of the work, the results were predicted using logistic regression analysis. The data obtained from clinical and laboratory studies were processed by the methods of variational statistics in the statistical analysis system Statistica 10. The KolmogorovSmirnov, ShapiroWilkie, and Leven tests were used. For comparative analysis, Student's t-test and the MannWhitney method were applied.
 Results. Based on the model we proposed using logistic regression, the main predictors of in vitro fertilization outcomes were identified: the absence of a previous pregnancy, the outcome of previous pregnancies with secondary infertility, age, spontaneous loss of uterine pregnancy, previous in vitro fertilization with a live birth, the number of mature oocytes prior to assisted reproductive technology and the number good quality embryos on the day of transfer. The sensitivity of this forecast was 84.7% (61 and 9; p=0.000), the specificity was 88.8% (11 and 72; p=0.000). It was found that forecasting using the model is 44.3 times more correct than if the prognosis of the outcome of in vitro fertilization was performed randomly.
 Conclusion. The developed logistic regression forecasting model allows in the vast majority of cases to correctly predict the outcome of in vitro fertilization.