Optimizing gestational diabetes diagnostic criteria to predict adverse perinatal outcomes in the United Arab Emirates: The Mutaba'ah Study

Abstract

Background: Gestational diabetes mellitus (GDM) affects 25% of pregnancies in the United Arab Emirates (UAE), and there is a need for evidence-based diagnostic criteria. This study aimed to determine the optimal diagnostic criteria for GDM in the Emirati population based on predicting adverse perinatal outcomes.

Methods: A total of 2,449 eligible pregnancies from "The Mutaba'ah Study" birth cohort were screened using OGTT between 24 and 32 weeks from May 2017 to March 2021. We compared the prediction of adverse perinatal outcomes [Large for Gestational Age (LGA) and Composite Outcome] risk by four GDM diagnostic criteria (IADPSG, NICE2015, WHO1999, and ADIPS1998) using adjusted regression models. We then developed a new GDM diagnostic threshold for this population (using an aOR of 1.75 recommended by the IADPSG consensus panel). The new criteria was validated and compared with other criteria using risk analyses, c-statistic (AUC), integrated discrimination improvement (IDI), and net reclassification improvement (NRI).

Results: Of the four criteria assessed, IADPSG was the best predictor for large for gestational age (LGA) (aOR 1·77, 95%CI 1·36-2·29) and composite outcome (aOR 1·49, 95%CI 1·19-1·86). The newly developed criteria showed even stronger associations than the IADPSG [LGA (aOR 1·93, 95% CI 1·48-2·53); Composite Outcome (aOR 1·62, 95% CI 1·28-2·05)]. The new criteria model had good discrimination properties for LGA prediction (AUC 0·78; 95%CI 0·68-0·88) and composite outcome prediction (AUC 0.73; 95%CI 0.57-0.83). The new criteria model also correctly reclassifies 49.4% of patients based on LGA risk [NRI; 0·494 (p=0·043)], whereas the IADPSG did not significantly reclassify these patients [NRI; 0·202 (p=0·409)]. For composite outcome prediction, the NRIs for both models were not statistically significant. The new criteria model also improved the discrimination slope (IDI) for LGA prediction by 42.2%, whereas IADPSG improved it by only 9.0%. For the composite outcome prediction, the new criteria model improved by 5.0% vs. IADPSG by 1.3%.

Conclusions: Following the development of a new threshold, the GDM diagnostic criteria defined in this study predicted adverse perinatal outcomes better and demonstrated more optimal clinical utility compared with the existing criteria in this population; hence, adopting it could minimize the burden of GDM adverse perinatal outcomes.

Keywords: IADPSG; United Arab Emirates; diabetes; diagnostic criteria; gestational diabetes mellitus (GDM); oral glucose tolerance test (OGTT); perinatal outcomes; risk stratification.

Figure 1

Marginsplot graphing statistics from the fitted, adjusted model (FPG vs. composite outcome) (N = 2,449). The composite outcome is defined as one or more of LGA, NICU admission, abnormal APGAR score, caesarean delivery, premature delivery, and preeclampsia. Model adjusted for age, gravidity, booking BMI, education, employment, family history of type 2 DM, previous GDM, study center, maternal smoking, passive smoking, physical activity, antepartum hemorrhage, previous macrosomia, and family history of hypertension.