Prediction of thrombophilia in patients with unexplained recurrent pregnancy loss using a statistical model

Objective To establish a statistical model to predict thrombophilia in patients with unexplained recurrent pregnancy loss ( URPL ). Methods A retrospective case–control study was conducted at Ren Ji Hospital, Shanghai, China, from March 2014 to October 2016. The levels of D‐dimer ( DD ), fibrinogen degradation products ( FDP ), activated partial thromboplastin time ( APTT ), prothrombin time ( PT ), thrombin time ( TT ), fibrinogen (Fg), and platelet aggregation in response to arachidonic acid ( AA ) and adenosine diphosphate ( ADP ) were collected. Receiver operating characteristic curve analysis was used to analyze data from 158 UPRL patients (≥3 previous first trimester pregnancy losses with unexplained etiology) and 131 non‐ RPL patients (no history of recurrent pregnancy loss). A logistic regression model ( LRM ) was built and the model was externally validated in another group of patients. Results The LRM included AA , DD , FDP , TT , APTT , and PT . The overall accuracy of the LRM was 80.9%, with sensitivity and specificity of 78.5% and 78.3%, respectively. The diagnostic threshold of the possibility of the LRM was 0.6492, with a sensitivity of 78.5% and a specificity of 78.3%. Subsequently, the LRM was validated with an overall accuracy of 83.6%. Conclusion The LRM is a valuable model for prediction of thrombophilia in URPL patients.