Diabetes and prediabetes among women universally screened for gestational diabetes: a multi-ethnic, population-based, prospective study with eleven years follow-up

Abstract

Background: Gestational diabetes (GDM) is a strong risk factor for later development of diabetes. However, data are scarce on the long-term risk for diabetes or prediabetes diagnosed by HbA1c, in non-selected, multi-ethnic populations universally screened for GDM using the WHO₂₀₁₃ criteria. We aimed to investigate the development of diabetes or prediabetes eleven years after the index pregnancy and identify risk factors in pregnancy or shortly after.

Methods: A population-based cohort study of 360 women with complete eleven years follow-up data for diabetes (HbA1c ≥ 48 mmol/mol) or prediabetes_ADA (HbA1c 39-47 mmol/mol). Women were enrolled in gestational week 15 and universally screened with an oral glucose tolerance test in week 28. We performed least absolute shrinkage and selection operator (LASSO) regression to identify predictors of future diabetes or prediabetes_ADA and constructed a nomogram to predict individual risks.

Results: Diabetes or prediabetes_ADA combined, was found in 26.9%, and the prevalence was slightly higher in previous GDM compared with non-GDM women (35.6% versus 23.5%; p = 0.019). The relative risk (RR) for developing diabetes or prediabetes_ADA was moderately elevated in GDM compared with non-GDM women (1.4 [1.0, 1.9], p = 0.035). Seven women (1.9%) had diabetes and all of these except for one, had previous GDM. Hence, the crude prevalence was 5.8% among GDM women vs. 0.4% among non-GDM women. The RR for developing diabetes was substantially higher in GDM vs. non-GDM women (14.8 [2.6, 277.1], p = 0.012). Prediabetes_ADA was found in 25% and the RR for prediabetes_ADA was not significantly increased for GDM compared to non-GDM women (1.3 [0.9, 1.8], p = 0.143). Among Europeans, 17.0% had diabetes or prediabetes_ADA, compared to 43.0% among South Asians (p < 0.001) and 34.4% among other ethnicities (p = 0.002). The most significant predictors identified from the LASSO were HbA1c measured in early pregnancy, ethnicity, and a family history of diabetes.

Conclusions: The risk for developing diabetes was low, overall and among GDM women. Still GDM represented a strong risk for diabetes, but not for prediabetes_ADA. HbA1c early in pregnancy, non-European ethnicity, and a family history of diabetes were the strongest risk factors for developing diabetes or prediabetes_ADA.

Trial registration: STORK G2 Women and Risk of Diabetes. NCT03870724 (ClinicalTrials.gov). February 27th, 2019.

Keywords: Ethnicity; Gestational diabetes; Type 2 diabetes.

Fig. 1

Flow chart. HbA1c based on capillary dried blood spot (n = 360, 94%)

Fig. 2

Nomogram estimating the probability of having prediabetes or diabetes as a function of 11 predictors. The predictors were selected by the Lasso logistic regression model. For a specific individual, each predictor value is given a score and a total score obtained by adding the individual scores. The total score is projected on total points axis to obtain the probability of prediabetes or diabetes

Fig. 3

Prediabetes and diabetes by GDM (WHO₂₀₁₃) status

Fig. 4

Prediabetes and diabetes by GDM (WHO₂₀₁₃) status by ethnic group

Fig. 5

ROC curve from LASSO for the performance of the least absolute shrinkage and selection operator risk score in identifying women with increased risk for prediabetes and diabetes. The AUC under the ROC curve is 0.7983 (95% CI 0.7813, 0.8154)

Fig. 6

ROC curve from Nomogram for the performance of the least absolute shrinkage and selection operator risk score in identifying women with increased risk for prediabetes and diabetes. The AUC under the ROC curve is 0.8018 (95% CI 0.7831, 0.8186)