Nomogram for prediction of gestational diabetes mellitus in urban, Chinese, pregnant women

Abstract

Background: This study sought to develop and validate a nomogram for prediction of gestational diabetes mellitus (GDM) in an urban, Chinese, antenatal population. Methods: Age, pre-pregnancy body mass index (BMI), fasting plasma glucose (FPG) in the first trimester and diabetes in first degree relatives were incorporated as validated risk factors. A prediction model (nomogram) for GDM was developed using multiple logistic regression analysis, from a retrospective study conducted on 3956 women who underwent their first antenatal visit during 2015 in Shanghai. Performance of the nomogram was assessed through discrimination and calibration. We refined the predicting model with t-distributed stochastic neighbor embedding (t-SNE) to distinguish GDM from non-GDM. The results were validated using bootstrap resampling and a prospective cohort of 6572 women during 2016 at the same institution. Results: Advanced age, pre-pregnancy BMI, high first-trimester, fasting, plasma glucose, and, a family history of diabetes were positively correlated with the development of GDM. This model had an area under the receiver operating characteristic (ROC) curve of 0.69 [95% CI:0.67-0.72, p < 0.0001]. The calibration curve for probability of GDM showed good consistency between nomogram prediction and actual observation. In the validation cohort, the ROC curve was 0.70 [95% CI: 0.68-0.72, p < 0.0001] and the calibration plot was well calibrated. In exploratory and validation cohorts, the distinct regions of GDM and non-GDM were distinctly separated in the t-SNE, generating transitional boundaries in the image by color difference. Decision curve analysis showed that the model had a positive net benefit at threshold between 0.05 and 0.78. Conclusions: This study demonstrates the ability of our model to predict the development of GDM in women, during early stage of pregnancy.