Using UK/ASRM models and local model to predict the prognosis of IVF in China: a retrospective study

Objective: To validate four pretreatment prediction models which were developed based on UK/US population (McLernon 2016 model, Luke model, Dhillon model, and McLernon 2022 model) and compare them with a model developed based on local data. Design: Retrospective development and validation of prediction models. Setting: University-affiliated hospital Patient(s): A total of 26035 Chinese patients underwent OPU cycles between January 2013 and December 2020. Methods: UK/US model performance was externally validated. Local models were established with XGboost, Lasso, and generalized linear model algorithms. Discriminatory power and calibration of the models were compared. Main Outcome Measure(s): The areas under the curves (AUC) of the Receiver Operator Characteristic (ROC) and calibration curves. Result(s): The AUCs for McLernon 2016 model, Luke model, Dhillon model, and McLernon 2022 model were 0.68 (95% CI: 0.675-0.689), 0.67 (95% CI: 0.666-0.68), 0.685((95% CI: 0.678-0.692), and 0.673((95% CI: 0.666-0.68), respectively. The local model yielded an AUC of 0.709(95% CI: 0.702-0.716) with key predictors including age, duration of infertility, and endocrine parameters. All external models suggested underestimation. Among the external models, the rescaled McLernon 2022 model demonstrated the best calibration (slope 1.12, intercept 0.06). Conclusion(s): A patient in China may find the published pretreatment prediction models based on UK/US population provide similar discriminatory power with reasonable AUCs and underestimated predictions.