AUTHOR=Mu Fangxiang , Huo Huyan , Wang Chen , Hu Ning , Wang Fang TITLE=A new prognostic model for recurrent pregnancy loss: assessment of thyroid and thromboelastograph parameters JOURNAL=Frontiers in Endocrinology VOLUME=15 YEAR=2024 URL=https://www.frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2024.1415786 DOI=10.3389/fendo.2024.1415786 ISSN=1664-2392 ABSTRACT=Objective

This study aimed to identify predictors associated with thyroid function and thromboelastograph (TEG) examination parameters and establish a nomogram for predicting the risk of subsequent pregnancy loss in recurrent pregnancy loss (RPL).

Methods

In this retrospective study, we analyzed the medical records of 575 RPL patients treated at Lanzhou University Second Hospital, China, between September 2020 and December 2022, as a training cohort. We also included 272 RPL patients from Ruian People’s Hospital between January 2020 and July 2022 as external validation cohort. Predictors included pre-pregnancy thyroid function and TEG examination parameters. The study outcome was pregnancy loss before 24 weeks of gestation. Variable selection was performed using least absolute shrinkage and selection operator regression and stepwise regression analyses, and the prediction model was developed using multivariable logistic regression. The study evaluated the model’s performance using the area under the curve (AUC), calibration curve, and decision curve analysis. Additionally, dynamic and static nomograms were constructed to provide a visual representation of the models.

Results

The predictors used to develop the model were body mass index, previous pregnancy losses, triiodothyronine, free thyroxine, thyroid stimulating hormone, lysis at 30 minutes, and estimated percent lysis which were determined by the multivariable logistic regression with the minimum Akaike information criterion of 605.1. The model demonstrated good discrimination with an AUC of 0.767 (95%CI 0.725-0.808), and the Hosmer-Lemeshow test indicated good fitness of the predicting variables with a P value of 0.491. Identically, external validation confirmed that the model exhibited good performance with an AUC of 0.738. Moreover, the clinical decision curve showed a positive net benefit in the prediction model. Meanwhile, the web version we created was easy to use. The risk stratification indicated that high-risk patients with a risk score >147.9 had a higher chance of pregnancy loss (OR=6.05, 95%CI 4.09-8.97).

Conclusions

This nomogram well-predicted the risk of future pregnancy loss in RPL and can be used by clinicians to identify high-risk patients and provide a reference for pregnancy management of RPL.