A systematic review of first trimester prediction models for gestational diabetes mellitus (GDM)
MetadataShow full item record
OBJECTIVE: The aim of this study is to perform a systematic review of available obstetric first trimester prediction models based on maternal characteristics for gestational diabetes mellitus (GDM). METHODS: The review included a comprehensive search of following electronic bibliographic databases: PubMed/MEDLINE from July 1st 2017 until September 30th 2021. Prognostic models published before April 1st 2017 were obtained from a previously conducted reviews. The studies were considered eligible if they met the pre-established criteria as follows: (1) the article must describe either the development or external validation of a prediction model, or an update to a previously developed model; (2) the model in question must contain multiple predictors; (3) the predictors used must be routinely collected in Dutch Obstetric Care; (4) the predictors must be available and/or measured prior to 16 weeks and 0 days of gestation; (5) The study population must comprise pregnant women; (6) the model must be based on weighted risk predictors that have been identified through multivariate analysis; and (7) the model must be used to predict the GDM outcomes. Data was collected through a pre-determined data extraction form, which included specific items related to study type, domain, outcome, development and validation, model performance (measured by AUROC), risk of bias, and applicability. To evaluate the risk of bias and applicability, the prediction model risk of bias assessment tool (PROBAST) was utilized. The systematic review adhered to the PRISMA guidelines for reporting. RESULTS: In this study, 24 studies on GDM prediction models were selected for analysis, with 12 studies retrieved from the latest databases after rigorous selection. The final analysis included 20 models developed for GDM and 57 models externally validated for GDM. The developing models demonstrated AUROC values ranging from 0.64 to 0.88 (mean), but their performance in external validation studies was slightly lower, with AUROC values ranging from 0.60 to 0.87 (mean). Compared to all other models evaluated, Nanda's model has demonstrated a relatively stable performance in terms of AUROC values across both self-validation and external validation (0.73-0.79). Which suggests that Nanda's model may be more reliable and consistent in predicting outcomes compared to other models. CONCLUSIONS: By analyzing and summarizing existing literature, this study presents information on current GDM prediction models with maternal characteristics and offers suggestions on the selection of predictors for future models, serving as a useful reference for future model development and enhancement. The study's findings suggest that although the model developed by Nanda et al. shows promise for predicting GDM, the other models exhibit lower performance. As technology and research improve, we expect better GDM prediction models in the future.