Understanding the potential contribution of polygenic risk scores to the prediction of gestational and type 2 diabetes in women from British Pakistani and Bangladeshi groups: a cohort study in Genes and Health

Abstract

Background: British Pakistani and Bangladeshi (BPB) women have disproportionately high rates of gestational diabetes mellitus (GDM), with prevalence estimates up to three times higher than in the general population. They are also at increased risk of progressing to type 2 diabetes, leading to significant health complications. Despite this, predictive models tailored to this high-risk, yet understudied group are lacking.

Objective: To investigate whether combining genetic and traditional clinical data improves risk prediction of GDM and progression to type 2 diabetes among BPB women. We hypothesized that incorporating polygenic risk scores (PRS) would enhance the predictive accuracy of existing models.

Study design: An observational cohort study utilizing the Genes & Health dataset, which includes comprehensive electronic health records. Women who gave birth between 2000 and 2023, both with and without a history of GDM, were included. Controls were defined as women without a GDM diagnosis during this period but who had a birth record. A total of 117 type 2 diabetes or GDM PRS were tested to determine the optimal PRS based on predictive performance metrics. The best-performing PRS was integrated with clinical variables for statistical analyses, including descriptive statistics, chi-square tests, logistic regression, and receiver operating characteristic curve analysis.

Results: Of 13,489 women with birth records, 10,931 were included in the analysis, with 29.3% developing GDM. Women with GDM were older (mean age 31.7 years, P<.001) and had a higher BMI (mean 28.4 kg/m², P<.001) compared to controls. The optimal PRS demonstrated a strong association with GDM risk; women in the highest PRS decile had significantly increased odds of developing GDM (OR 5.66, 95% CI [4.59, 7.01], P=3.62×10^-58). Furthermore, the risk of converting from GDM to type 2 diabetes was 30% in the highest PRS decile, compared to 19% among all GDM cases and 11% in the lowest decile. Incorporating genetic risk factors with clinical data improved the C-statistic for predicting type 2 diabetes following GDM from 0.62 to 0.67 (P=4.58×10^-6), indicating better model discrimination.

Conclusion: The integration of genetic assessment with traditional clinical factors significantly enhances risk prediction for BPB women at high risk of developing type 2 diabetes after GDM. These findings support the implementation of targeted interventions and personalized monitoring strategies in this high-risk population. Future research should focus on validating these predictive models in external cohorts and exploring their integration into clinical practice to improve health outcomes.

Keywords: South Asian; gestational diabetes; polygenic risk; prediction model; pregnancy complications; prognosis; risk stratification.

Figure 1

Strobe chart

Figure 2

Polygenic risk score distribution amongst women with and without GDM GDM status: 0=women who did not have GDM; 1=women who had GDM. GDM, gestational diabetes

Figure 3

Risk of developing gestational diabetes by polygenic risk score decile PRS, polygenic risk score.

Figure 4

ROC curves demonstrating improvement in prediction of GDM with utilization of genetic information Comparing three models: the Null model, which includes only traditional population risk factors (age, BMI, parity, and ethnicity); the PRS model, which includes only the polygenic risk score; and the Full model, which combines both the population risk factors and the polygenic risk score. AUC, area under the curve; PRS, polygenic risk score; ROC, receiver operating characteristic.

Figure 5

Polygenic risk score distribution amongst women with GDM who progressed to type 2 diabetes versus those who did not T2DM status: 0=women who did progress to type 2 diabetes; 1=women who progressed to type 2 diabetes. T2DM, type 2 diabetes

Figure 6

ROC curves demonstrating improvement in prediction of type 2 diabetes with utilization of genetic information Comparing three models: the Null model, which includes only traditional population risk factors (age, BMI, parity, and ethnicity); the PRS model, which includes only the polygenic risk score; and the Full model, which combines both the population risk factors and the polygenic risk score. AUC, area under the curve; PRS, polygenic risk score; ROC, receiver operating characteristic.