Integrating genetic, clinical, and behavioural factors may improve the prediction of type 2 diabetes
F Escorcio Silva, M I Mendonca, G Abreu, F Sousa, M Ferreira, S Freitas, M Rodrigues, E Henriques, S Borges, G Guerra, A Cardoso, I Ornelas, A Drumond, A C Sousa, R Palma Dos ReisAbstract
Introduction
Type 2 diabetes mellitus (T2DM) is one of the fastest-growing noncommunicable multifactorial and polygenic diseases, leading to many health complications and significant morbidity and mortality.
Aim
Investigate the predictive capacity of a polygenic risk score with genetic variants associated with T2DM (PRS-D), and evaluate the increase in predictive value when other behavioural and clinical risk factors are added.
Methods
A case–control study was conducted with T2DM patients from our centre’s dataset and controls without diabetes or pre-diabetes. DNA samples were genotyped for 15 genetic polymorphisms associated with T2DM and oxidation (IGF2BP2 rs4402960, PPARG rs1801282, TCF7L2 rs7903146, SLC30A8 rs1326634, MC4R rs17782313, ADIPOQ rs266729, FTO rs8050136, TAS2R50 rs1376251, HNF4A rs1884613, MTHFD1L rs6922269, MTHFR677 rs1801133, MTHFR rs1298, PON108 rs705379, PON192 rs662, PON55 rs854560). A weighted PRS-D was calculated by summing all risk alleles multiplied by their respective odds ratios (ORs). T2DM relative risk was assessed using multivariate logistic regression, and the area under the receiver operating characteristic curve (AUC) evaluated behavioural and clinical factors (hypertension, obesity, and sedentary lifestyle) alone and combined with PRS-D.
Results
Our population included 2,097 participants: 846 T2DM patients (mean age 55.3±6.9; 78.7% male) and 1,251 controls (mean age 51.4±8.5; 71.9% male). PRS-D was significantly higher in patients than controls (12.5±2.6 vs. 12.0±2.5; p<0.0001). Bivariate analysis of risk factors showed that patients had 79.9% hypertension, 44.3% obesity, and 63.6% sedentary lifestyle. Adding PRS-D to risk factors improved basal model discrimination, increasing the AUC for sedentary lifestyle (0.614→0.691), hypertension (0.665→0.710), and obesity (0.598→0.689). The DeLong test confirmed significance (p<0.0001). After multivariate analysis adjusting for age and sex, physical inactivity, obesity, and hypertension, the PRS-D remained significantly associated with T2DM. PRS-D increased risk by 9.1% per unit (OR=1.091; 95% CI: 1.050–1.133), while sedentary lifestyle (OR=2.121), obesity (OR=1.588), and hypertension (OR=3.042) showed robust independent associations.
Conclusions
PRS-D is associated with T2DM susceptibility and improves disease prediction even with a limited number of loci. PRS may support early risk stratification and personalised prevention strategies. Further studies may optimise its use for clinical and public health applications, though classical risk factors remain the strongest predictors.Integrating factors for diabetes prediction