TL;DR
This study develops and validates a polygenic risk score (PRS) for type 2 diabetes mellitus (T2DM) specifically for the Korean population. The T2DM PRS not only predicts the progression from normal glucose tolerance to prediabetes and then to T2DM but also identifies individuals at higher risk for insulin treatment and those belonging to a severe diabetes subgroup characterized by insulin resistance and beta-cell dysfunction.
What Problem Does This Paper Address?
Type 2 diabetes mellitus (T2DM) is a growing global health concern, with increasing prevalence in Korea. Early identification of individuals at high risk is crucial for timely intervention and prevention of complications. While genome-wide association studies (GWAS) have identified numerous genetic variants associated with T2DM risk, the clinical utility of polygenic risk scores (PRS) for T2DM, particularly in East Asian populations, remains underexplored. Existing PRS research has often been limited to cross-sectional data or included only a small number of risk variants, and has predominantly focused on European populations. This paper aims to bridge this gap by establishing and evaluating a T2DM PRS for the Korean population, assessing its prognostic value and potential as a clinical instrument.
Methods
The researchers employed a multi-step approach:
- PRS Construction: They first constructed a T2DM PRS using a large East Asian biobank dataset (269,487 samples) through a meta-analysis of GWAS summary statistics from the Korean Genome and Epidemiology Study (KoGES_HEXA) and Biobank Japan. Two PRS construction methods, Lassosum and PRS-CS, were evaluated, with Lassosum ultimately chosen for its slightly better performance in the PRS-only model.
- Validation and Prediction: The predictive capability of the established T2DM PRS was then assessed in a prospective longitudinal study using data from the KoGES Ansan and Ansung cohorts, which included baseline data and seven follow-up phases over 14 years.
- Progression Prediction: Cox regression models were used to evaluate the PRS's ability to predict:
- Progression from non-diabetes to T2DM.
- Progression from normal glucose tolerance (NGT) to prediabetes.
- Progression from prediabetes to T2DM.
- Progression to insulin prescription among T2DM patients.
- T2DM complications.
- Subgroup Analysis: To explore T2DM heterogeneity, patients were classified into four subgroups using k-means clustering based on BMI, age at diagnosis, HOMA-B, HOMA-IR, and HbA1c. The PRS was then analyzed within these subgroups.
- Model Performance Evaluation: The predictive accuracy of models incorporating the PRS was compared to models without it using Harrell's C-indices.
Datasets
- Training Data:
- Korean Genome and Epidemiology Study (KoGES_HEXA) cohort (58,622 participants).
- Biobank Japan (BBJ) data (downloaded summary statistics).
- A combined meta-analysis of KoGES_HEXA and BBJ data (269,487 samples) was used for PRS training.
- Validation Data:
- KoGES_CAVAS cohort (n=8105) was used for hyperparameter selection.
- KoGES_Ansan and Ansung cohorts (n=5490) were used for prospective longitudinal study and evaluation of T2DM PRS.
- UK Biobank (UKBB) (1503 East Asian samples) was used for external validation of the T2DM PRS.
Key Insights and Ideas
- East Asian Specificity: The study successfully developed and validated a T2DM PRS tailored for the Korean population, addressing a critical gap in PRS research for East Asian ancestry.
- Beyond Simple Prediction: The T2DM PRS demonstrates utility not just for predicting T2DM onset but also for forecasting the transition through intermediate glycemic stages (NGT to prediabetes, prediabetes to T2DM).
- Predicting Treatment Needs: A higher PRS is linked to a greater likelihood of requiring insulin treatment among T2DM patients, suggesting it could help anticipate future treatment demands.
- Identifying Severe Subtypes: The PRS is significantly associated with a "severe diabetes subgroup" characterized by insulin resistance and impaired beta-cell function, hinting at its potential role in personalized diabetes management and understanding disease heterogeneity.
- Enhanced Predictive Power: Incorporating the T2DM PRS into prediction models significantly improves their accuracy (higher Harrell's C-indices) compared to models relying solely on traditional clinical risk factors.
Key Results
- The T2DM PRS predicted the transition from normal glucose tolerance to prediabetes and from prediabetes to T2DM.
- T2DM patients in the top-decile PRS group were 1.69 times more likely to be treated with insulin (p=2.31E-02).
- T2DM PRS values were significantly higher in the severe diabetes subgroup (p=0.0012).
- Prediction models with the T2DM PRS showed significantly greater Harrell's C-indices compared to models without it. For instance, adding standardized PRS to a model with sex, age, and family history improved the C-index from 0.586 to 0.631.
- The top-decile PRS group showed significantly higher progression rates from non-diabetes to prediabetes and from prediabetes to T2DM.
Limitations and Future Work
- T1DM Distinction: The study acknowledges that distinguishing between type 1 and type 2 diabetes can be unclear, and while they excluded T1DM patients based on age and applied a robust criterion for diagnosis, this remains a potential limitation.
- Medication and Complication Data: Information on oral diabetes medication, insulin dosage, and detailed T2DM complication history was self-reported and may have limitations.
- C-peptide and Autoantibody Data: The study did not include C-peptide tests or diabetes autoantibody results, which could further refine T2DM subtyping.
- Future Directions: The authors suggest that implementing this PRS as a clinical assessment tool could aid in T2DM screening and prognosis, potentially preventing complications through early intervention and strict glycemic control. Further research is needed to understand the biological pathways linking PRS to specific T2DM complications.
Summary updated