Aims: Gestational diabetes (GDM) affects a growing number of women and identification of individuals at risk, e.g., with risk prediction models, would be important. However, the performance of GDM risk scores has not been optimal. Here, we assess the impact of GDM heterogeneity on the performance of two top-rated GDM risk scores.
Methods: This is a substudy of the RADIEL trial-a lifestyle intervention study including women at high GDM risk. We assessed the GDM risk score by Teede and that developed by Van Leeuwen in our high-risk cohort of 510 women. To investigate the heterogeneity of GDM, we further divided the women according to GDM history, BMI, and parity. With the goal of identifying novel predictors of GDM, we further analyzed 319 women with normal glucose tolerance in the first trimester.
Results: Both risk scores underestimated GDM incidence in our high-risk cohort. Among women with a BMI ≥ 30 kg/m2 and/or previous GDM, 49.4% developed GDM and 37.4% received the diagnosis already in the first trimester. Van Leeuwen score estimated a 19% probability of GDM and Teede succeeded in risk identification in 61%. The lowest performance of the risk scores was seen among the non-obese women. Fasting plasma glucose, HbA1c, and family history of diabetes were predictors of GDM in the total study population. Analysis of subgroups did not provide any further information.
Conclusions: Our findings suggest that the marked heterogeneity of GDM challenges the development of risk scores for detection of GDM.
Keywords: Gestational diabetes; Heterogeneity; Obesity; Prediction of diabetes; Pregnancy; Screening.