Prospective observational cohort study: Computational models for early prediction of ongoing pregnancy in fresh IVF/ICSI-ET protocols

Purpose This study sought to identify the significant factors related to ongoing pregnancy (OP) and to discover the most reliable model to distinguish OP from non-OP in early gestational age. Methods A total of 1650 cycles were enrolled in this study. Univariate Logistic Regression was used to identify the predictors included in multivariable analysis. The dataset was then randomly split into training set and test set with proportion of 70% and 30%. Forward stepwise multivariable logistic regression with 5-fold cross validation was used to build the final mathematic model. The performance of the model was determined by the arguments of test set. The area under receiver operating characteristic curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and misclassification rate (MR) were then calculated for model evaluation. Results Seven predictors were related to OP by univariate analysis. The serum hCG level on 14th day post-embryo-transfer (hCG14) and 21th day post-embryo-transfer (hCG21) were linear correlated. Therefore, different multivariate regression models were built using hCG14 or hCG21, respectively. After multivariate regression with 5-fold validation, the final indicators in model-1 were age_group, hCG21 and hCG21/hCG14, while age_group, hCG14, and calculated 48-hour-rising-ratio of hCG were the significant predictors in model-2. Model-2 showed better sensitivity and NPV, lower MR, and similar specificity and PPV. Conclusion This study provided an effective mathematic model for early prediction of OP. The model could be of better clinical significance, especially for clinical counseling to manage patients' stress and anxiety, and for early warning of threatened miscarriage.