Participant data from Phase I.

Ovarian age assessment is an important indicator to evaluate the ovarian reserve function and reproductive potential of women. At present, the application of ovarian age prediction model in China needs further improvement and optimization to make it more suitable for the actual situation of women in China. In this study, we collected subjects and their data in three ways: firstly, we collected clinical data from a number of women go to local hospital, including healthy women and women with DOR or PCOS; secondly, we obtained data by recruited healthy women through CRO companies for a fee; thirdly, we collected data from a number of healthy women using WeChat applet. Using the data collected by CRO company and WeChat applet, we applied the generalized linear model to optimize the ovarian age prediction model. The optimized formula is: OvAge = exp (3.5254–0.0001*PRL-0.0231*AMH), where P = 0.8195 for PRL and P = 0.0003 for AMH. Applying the formula to the hospital population data set for testing, it showed that the predicted ovarian age in the healthy women was comparable to their actual age, with a root mean squared error (RMSE) = 5.6324. The prediction accuracy was high. These data suggest that our modification of the ovarian age prediction model is feasible and that the formula is currently a more appropriate model for ovarian age assessment in healthy Chinese women. This study explored a new way to collect clinical data, namely, an online ovarian age calculator developed based on a WeChat applet, which can collect data from a large number of subjects in a short period of time and is more economical, efficient, and convenient. In addition, this study introduced real data to optimize the model, which could provide insights for model localization and improvement.