Monash Centre for Health Research and Implementation, Faculty of Medicine, Nursing and Health Sciences, Monash University, Melbourne, Australia.
Deakin Rural Health, School of Medicine, Deakin University, Warrnambool, Australia.
Curr Diab Rep. 2023 Sep;23(9):231-243. doi: 10.1007/s11892-023-01516-0. Epub 2023 Jun 9.
Despite the crucial role that prediction models play in guiding early risk stratification and timely intervention to prevent type 2 diabetes after gestational diabetes mellitus (GDM), their use is not widespread in clinical practice. The purpose of this review is to examine the methodological characteristics and quality of existing prognostic models predicting postpartum glucose intolerance following GDM.
A systematic review was conducted on relevant risk prediction models, resulting in 15 eligible publications from research groups in various countries. Our review found that traditional statistical models were more common than machine learning models, and only two were assessed to have a low risk of bias. Seven were internally validated, but none were externally validated. Model discrimination and calibration were done in 13 and four studies, respectively. Various predictors were identified, including body mass index, fasting glucose concentration during pregnancy, maternal age, family history of diabetes, biochemical variables, oral glucose tolerance test, use of insulin in pregnancy, postnatal fasting glucose level, genetic risk factors, hemoglobin A1c, and weight. The existing prognostic models for glucose intolerance following GDM have various methodological shortcomings, with only a few models being assessed to have low risk of bias and validated internally. Future research should prioritize the development of robust, high-quality risk prediction models that follow appropriate guidelines, in order to advance this area and improve early risk stratification and intervention for glucose intolerance and type 2 diabetes among women who have had GDM.
尽管预测模型在指导妊娠期糖尿病（GDM）后 2 型糖尿病的早期风险分层和及时干预方面发挥着关键作用，但它们在临床实践中的应用并不广泛。本综述旨在检查现有的预测 GDM 后产后葡萄糖耐量不良的预后模型的方法学特征和质量。
对相关风险预测模型进行了系统评价，结果来自不同国家研究小组的 15 篇合格文献。我们的综述发现，传统的统计模型比机器学习模型更为常见，只有两个模型被评估为低偏倚风险。其中 7 个进行了内部验证，但没有一个进行了外部验证。分别有 13 项和 4 项研究对模型的区分度和校准度进行了评估。确定了各种预测因子，包括体重指数、妊娠期间空腹血糖浓度、产妇年龄、糖尿病家族史、生化变量、口服葡萄糖耐量试验、妊娠期间胰岛素使用、产后空腹血糖水平、遗传风险因素、糖化血红蛋白和体重。现有的 GDM 后葡萄糖耐量不良的预测模型存在各种方法学缺陷，只有少数模型被评估为低偏倚风险且进行了内部验证。未来的研究应优先开发稳健、高质量的风险预测模型，并遵循适当的指南，以推进该领域的发展，并改善 GDM 女性的葡萄糖耐量不良和 2 型糖尿病的早期风险分层和干预。
