Type 2 diabetes and its genetic susceptibility are associated with increased severity and mortality of COVID-19 in UK Biobank

The study addresses how type 2 diabetes (T2D) and genetic susceptibility to T2D influence COVID-19 severity and mortality. COVID-19 presents a wide clinical spectrum from asymptomatic to fatal disease and remains a major public health concern despite vaccination. Prior evidence implicates demographic factors (age, sex), behaviors (smoking), socioeconomic status, comorbidities (obesity, T2D, respiratory and cardiovascular diseases, dementia), and host genetics in severe outcomes. Genome-wide association studies have identified variants associated with COVID-19 severity, and polygenic risk scores (PRS) derived from common variants can stratify risk. However, effects of PRS—particularly T2D PRS—on COVID-19 mortality have been understudied. This work aims to systematically evaluate associations of T2D and T2D PRS with COVID-19 severity and mortality, account for SARS-CoV-2 variants (including Omicron) and vaccination, and clarify potential confounding by BMI.

Multiple studies report that older age, male sex, smoking, and lower socioeconomic status correlate with severe COVID-19. Comorbidities such as obesity, T2D, respiratory disease, hypertension, CVD, and dementia are linked to worse outcomes. Large-scale GWAS indicate host genetic variation contributes to COVID-19 severity; PRS for COVID-19 has been associated with severe disease in independent cohorts. PRS for other diseases (e.g., obesity, asthma, schizophrenia) also show relationships with COVID-19 outcomes. Nonetheless, prior work has focused mainly on severity or progression rather than mortality, and few studies have specifically assessed the impact of T2D and T2D PRS on both severity and mortality of COVID-19. Mendelian randomization studies evaluating T2D’s causal effect on COVID-19 severity have been inconsistent, potentially due to confounding by BMI.

Data: UK Biobank cohort with 459,119 participants; updated self-reported data, hospital inpatient records, death records, and PCR test data through December 2022. T2D cases identified from self-report, hospital, and death data (ICD-10 E11). SARS-CoV-2 infection identified from test results, inpatient, and death data (ICD-10 U07.1/U07.4). Vaccination status from self-reported fields; vaccinated defined considering timing relative to infection and 14 days post-first dose. Genetics: GWAS conducted within UK Biobank for T2D, SARS-CoV-2 infection, and severe COVID-19, adjusting for age, sex, genotype PCs, assessment array, and genotyping quality. Heritability and genetic correlation estimated via LD score regression (LDSC) using 1000 Genomes Phase 3 v5 reference, primarily among European ancestry. PRS: Individual T2D PRS computed using LDpred with 10-fold cross-validation to avoid overfitting. Candidate PRS across fraction-causal tuning parameters evaluated by AUC; best PRS selected. PRS standardized; adjusted PRS scores derived via models including age and gender. Partitioned T2D PRS constructed for five pathophysiologic components (beta-cell function, proinsulin, obesity, lipodystrophy, liver/lipid metabolism) using published SNP sets and weights. Sensitivity PRS built using BMI-adjusted T2D GWAS from DIAGRAM consortium. Severity analysis: Among SARS-CoV-2-positive individuals, severity categorized as mild, moderate (hospitalized), or critical (COVID-19 death). Proportional odds regression modeled severity with predictors: T2D, T2D PRS (continuous per SD), or PRS group (low/medium/high), plus clinical covariates (age, sex, BMI, genotyping array, PCs 1–4), with/without vaccination status. Brant tests evaluated proportional odds assumption. Survival analysis (all participants): Multivariable Cox proportional hazards models assessed mortality risk with T2D and SARS-CoV-2 infection using time-dependent coefficients (five time intervals aligned with variant waves: pre-Delta, Delta, Omicron sub-periods). Covariates included age, sex, BMI, genotyping array, PCs 1–4; vaccination status was added in extended models. Survival time defined from study start to death or censoring; proportional hazards assumptions examined; pairwise log-rank tests compared stratified survival curves. Survival analysis (infected only): Cox PH models among infected individuals (N≈10,106) included indicators for variant periods (Early, EU1, Alpha, Delta, Omicron1/2/3; Omicron3 as reference), T2D or T2D PRS/PRS group, clinical covariates, and vaccination. Model performance assessed via 10-fold cross-validated C-index. Analyses performed in R 4.2.3; code available at the cited GitHub repository.

- Sample ascertainment: Within the UK Biobank, T2D cases totaled 37,110; SARS-CoV-2 confirmed cases totaled 101,271 across data sources. - Heritability/genetic correlation: SNP-heritability estimated at 0.12 (SE 0.012) for T2D and 0.27 (SE 0.009) for severe COVID-19. Genetic correlation between T2D and severe COVID-19 was 0.324 (SE 0.086), indicating shared genetic architecture. - PRS discrimination: AUCs for T2D PRS and adjusted T2D PRS were 0.641 and 0.711, respectively. For severe COVID-19, unadjusted and adjusted models had AUCs 0.518 and 0.644, respectively. - Severity associations (proportional odds models): • T2D associated with higher severity: OR 2.824 (95% CI 2.644–3.016). • T2D PRS (per SD) associated with higher severity: OR 1.080 (95% CI 1.050–1.111). • High vs low PRS group: OR 1.211 (95% CI 1.120–1.323). • Vaccination associated with lower severity: OR 0.206 (95% CI 0.193–0.220). • Proportional odds assumption supported by Brant tests. - Survival (all participants): • T2D increased mortality risk: HR ≈ 2.266 (95% CI 2.164–2.372). • SARS-CoV-2 infection increased mortality risk with time-interval-specific HRs: ~7.733 in earliest interval; ~2.013, 1.864, 1.906, 1.986 in subsequent intervals. • T2D PRS increased mortality risk: HR 1.062 (95% CI 1.042–1.081) per SD. • Vaccination associated with lower mortality: HR 0.198 (95% CI 0.186–0.210). • C-index for Survival-All: 0.770 (T2D model) and 0.755 (PRS group model). - Survival (infected only): • T2D increased mortality: HR 2.360 (95% CI 2.154–2.584). • T2D PRS per SD increased mortality: HR 1.062 (95% CI 1.022–1.104). • High PRS group vs low: HR 1.190 (95% CI 1.034–1.370). • Variant effects vs Omicron3 (reference) showed substantially higher mortality risk for Early (HR ~7.993), EU1 (~4.454), Alpha (~5.291), and elevated risk for Delta (~1.359). Omicron periods had lower or comparable risks. • Vaccination reduced mortality risk in infected cohorts. • C-index for Survival-COVID-19: 0.791 (T2D) and 0.795 (PRS group). - Partitioned T2D PRS: Obesity, lipodystrophy, and liver/lipid metabolism PRS (insulin resistance-related) were significantly associated with COVID-19 severity; beta-cell and proinsulin PRS were not. None of the partitioned PRS were significantly associated with COVID-19 mortality.

Findings demonstrate that both clinical T2D and genetic susceptibility to T2D (PRS) are independently associated with greater COVID-19 severity and higher mortality. The positive genetic correlation between T2D and severe COVID-19 suggests shared genetic underpinnings. Vaccination markedly reduces both severity and mortality risks. Mortality risk among infected individuals varied across variant waves, being highest during early and Alpha periods, attenuated during Delta, and lowest during Omicron periods. Notably, individuals without diagnosed T2D but with high T2D PRS were at elevated risk for severe outcomes, highlighting the utility of PRS for risk stratification beyond clinical diagnosis. Partitioned PRS results implicate insulin resistance-related pathways (obesity, lipodystrophy, liver/lipid) rather than insulin secretion pathways (beta-cell, proinsulin) in severe COVID-19 pathogenesis. These results support integrating T2D status and T2D PRS into clinical risk assessments for COVID-19 to prioritize monitoring and interventions, and motivate mechanistic studies on the metabolic-genetic links to severe COVID-19.

This study systematically shows that T2D and higher genetic susceptibility to T2D are associated with increased COVID-19 severity and mortality in the UK Biobank cohort. Vaccination substantially mitigates these risks, and mortality risk differs by SARS-CoV-2 variant period. T2D PRS can aid in identifying high-risk individuals, including those without diagnosed T2D, informing clinical triage and management. Future research should investigate underlying biological mechanisms connecting insulin resistance to severe COVID-19, validate PRS-based risk stratification across ancestries and age groups, and integrate lifestyle and comorbidity data to refine predictive models.

The cohort’s average age (~58 years) may limit generalizability to younger populations. Confounding by lifestyle factors and comorbidities remains a concern and warrants ongoing evaluation. Some analyses (e.g., LDSC) focused on European ancestry, which may limit applicability to other ancestries. PRS partition components were not associated with mortality, suggesting limited utility for predicting death outcomes. Differences between proportional odds and logistic infection-status models underscore sensitivity to outcome definitions and pandemic phase prevalence.