Abstract
Type 2 diabetes (T2D) is known as one of the important risk factors for the severity and mortality of COVID-19. Here, we evaluate the impact of T2D and its genetic susceptibility on the severity and mortality of COVID-19, using 459,119 individuals in UK Biobank. Utilizing the polygenic risk scores (PRS) for T2D, we identified a significant association between T2D or T2D PRS, and COVID-19 severity. We further discovered the efficacy of vaccination and the pivotal role of T2D-related genetics in the pathogenesis of severe COVID-19. Moreover, we found that individuals with T2D or those in the high T2D PRS group had a significantly increased mortality rate. We also observed that the mortality rate for SARS-CoV-2-infected patients was approximately 2 to 7 times higher than for those not infected, depending on the time of infection. These findings emphasize the potential of T2D PRS in estimating the severity and mortality of COVID-19.
Similar content being viewed by others
Introduction
The coronavirus disease 2019 (COVID-19) is caused by an infection with the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and remains a global pandemic1. COVID-19 affects the respiratory system and other crucial bodily systems, presenting a spectrum of symptoms ranging from asymptomatic and mild cases to severe illnesses, including severe pneumonia, acute respiratory distress syndrome (ARDS), and even death2,3,4. Furthermore, a considerable portion of COVID-19 patients, at least one-third, develop what is termed post-acute COVID-19 syndrome, which encompasses respiratory complications, fatigue, nervous system disturbances, and musculoskeletal issues5. These persist for ~4 weeks or more following the initial onset of symptoms6. Despite rigorous vaccination efforts, the potential for severe and chronic outcomes of COVID-19 remains. This underscores the urgent need for enhanced characterization of individuals at a heightened risk for clinically serious disease outcomes.
A variety of studies have indicated that certain clinical risk factors are associated with increased severity of COVID-19, often leading to hospitalization or death. These factors include older age, male gender, smoking, and lower socioeconomic status7,8. Prevalent comorbidities, such as obesity9,10,11,12, type 2 diabetes (T2D)13,14, respiratory diseases, hypertension, cardiovascular disease (CVD)15,16, and dementia17, also correlate with poorer COVID-19 outcomes. Moreover, an individual’s genetic makeup significantly influences the degree of COVID-19 severity18. This is supported by extensive population-based genome-wide association studies (GWAS), which suggest that a number of common genetic variants can affect the severity of symptoms following a COVID-19 infection4,19. The genes most strongly associated with severe COVID-19 progression, as identified by GWAS, typically play roles in immune response mechanisms, lung disease pathology, CVD20, and T2D21.
Each individual variant associated with COVID-19 severity, identified in the original GWAS, has a minor effect when considered separately, but cumulatively, they contribute significantly to the genetic variation seen in the severity of COVID-19. By integrating these variants, such as single nucleotide polymorphisms (SNPs), into a polygenic risk score (PRS), we can more effectively identify patients at potential risk within a cohort. The COVID-19 PRS has been known to be associated with an increased risk of severity and morbidity from COVID-19 across multiple independent groups22,23. PRSs for other underlying diseases, such as obesity, asthma, and schizophrenia, have also been found to correlate with COVID-19 severity or disease progression23,24,25. However, most studies have primarily focused on the association between COVID-19 PRS or PRSs for other underlying diseases and the severity or clinical progression of the illness. Research on the effect of PRS on COVID-19 mortality is limited. Notably, only a few studies have assessed the impact of T2D or T2D PRS on the severity and mortality of COVID-19, despite T2D being recognized as one of the comorbidities strongly associated with severe COVID-19.
In this paper, we systematically investigate the association of T2D and T2D PRS with the severity and mortality of COVID-19. We also evaluate the effects of various SARS-CoV-2 variants, including the recent Omicron variants, and assess the impact of vaccinations. First, we estimate the genetic predisposition for T2D by computing the T2D PRS using GWAS summary statistics from the UK Biobank. We include body mass index (BMI) as a covariate in the GWAS when constructing the T2D PRS, to mitigate the confounding effect of BMI on T2D. This approach is informed by the Mendelian randomization (MR) studies on the causal effects of T2D on COVID-19 severity, which have yielded inconsistent results, largely attributable to the confounding impact of BMI on T2D26,27. Next, we assess the relationship between T2D, T2D PRS, and the severity of COVID-19 using proportional odds models. We then investigate the effect on survival times of T2D patients infected with SARS-CoV-2 using Cox proportional hazards models with time-dependent coefficients28 for SARS-CoV-2 infection to satisfy the proportional hazards assumption. In this model, we incorporate either T2D or T2D PRS to determine the impact of T2D or genetic risks for T2D on survival time. Additionally, we examine differences in survival times across SARS-CoV-2 variants and vaccination statuses. Lastly, we conduct a stratified survival analysis for COVID-19, T2D, and T2D PRS groups, and perform the pairwise log-rank test to identify significant differences in mortality between these groups.
Results
Ascertainment of T2D and SARS-CoV-2 cases
The overall study design is shown in Fig. 1 and the baseline characteristics for T2D and T2D-related traits are detailed in Table 1. Based on the recently updated datasets from the UK Biobank (up to December 2022), we ascertained the number of T2D cases and SARS-CoV-2 confirmed cases from 459,119 participants. We identified a total of 37,110 (8.1%) T2D patients from the self-reported data (data field 2443), hospital inpatient records, and death data (International Classification of Diseases, 10th Revision (ICD-10), code E11) (Fig. 2a). We obtained a total of 101,271 (22.1%) SARS-CoV-2 confirmed cases from the COVID-19 test data, hospital inpatient records, and death data (ICD-10, codes U07.1 and U07.4) (Fig. 2b).
We utilized 459,119 individuals in the UK Biobank to evaluate the impact of T2D and its genetic susceptibility on the severity and mortality of COVID-19. We calculated polygenic risk scores (PRS) based on summary statistics from the UK Biobank to capture the overall genetic susceptibility to T2D using 10-fold cross-validation. In our investigation, we examined the association of T2D and T2D PRS with COVID-19 severity using proportional odds models. We further explored the effect of T2D and T2D PRS on survival times in COVID-19 patients using Cox proportional hazard models.
a The total sample size for T2D patients is 37,110, identified from self-reported data, hospital inpatient data, and death data. b The total sample size for SARS-CoV-2 confirmed cases is 101,271, derived from laboratory-confirmed PCR test results, hospital inpatient data, and death data.
We prepared two COVID-19-related phenotypes: (1) all cases with reported SARS-CoV-2 infection regardless of symptoms (101,271 cases and 357,848 controls) and (2) individuals with severe COVID-19, defined as those who were hospitalized or died due to symptoms associated with the infection (7478 cases and 451,641 controls). Based on the SARS-CoV-2 infection dates and information on the proportions of all variants during the study period in the UK, we classified participants into seven variant groups: early variants, EU1 (20E), Alpha (20I), Delta (21J), Omicron1 (21K), Omicron2 (21L), and Omicron3 (22B) variants, defining early variants as all variants detected in the UK before the EU1 variant appeared. We randomly assigned all SARS-CoV-2-infected cases to one of the variant groups according to the proportions of variants on their SARS-CoV-2 infection date. The vaccination data contains self-reported vaccination status (data fields 27,983, 27,985) and dates (data fields 27,984, 27,986). The unvaccinated group (N = 279,223) was defined as participants who had no vaccination record, as well as those diagnosed with COVID-19 prior to vaccination or within 14 days of receiving the first dose29.
Heritability and genetic correlation
We performed GWAS analyses for T2D, SARS-CoV-2 infection, and severe COVID-19 (Supplementary Fig. 1), and then estimated heritability and genetic correlations between various phenotypes based on GWAS summary statistics. Supplementary Table 1 shows the estimated heritability in the diagonal elements and the estimated genetic correlation between pairs of traits in non-diagonal elements. The estimated heritability for T2D and severe COVID-19 were 0.154 (SE = 0.012) and 0.027 (SE = 0.009), respectively, implying that the phenotypic variation of T2D and severe COVID-19 can be explained approximately 15.4% and 2.7% by SNPs, respectively. The estimated genetic correlation between T2D and severe COVID-19 was 0.242 (SE = 0.086), indicating a potential shared genetic architecture between T2D and severe COVID-19. The estimated heritability and genetic correlation for other traits are described in Supplementary Note 1.
Genetic susceptibility
For the 459,119 participants, we computed individual PRS representing the genetic susceptibility for traits using a 10-fold cross-validation (CV), and then standardized the PRS. We further derived scores of adjusted PRS based on logistic or linear regression models with PRS, age, and gender as covariates. The estimated area under the curves (AUCs) of T2D PRS and severe COVID-19 were 0.641 and 0.518, and those of the adjusted T2D PRS and COVID-19 were 0.711 and 0.644, respectively (Supplementary Table 2). Supplementary Fig. 2 displays the ROC curves for PRS and adjusted T2D PRS and Supplementary Table 3 shows the estimated AUC for the candidate PRS across a range of tuning parameters. The PRS with the best discriminative capability is highlighted in bold type.
Association between the severity of COVID-19 and T2D
In order to identify the association between the severity of COVID-19 and T2D or its genetic susceptibility, we divided SARS-CoV-2-infected participants into three groups according to the severity of their COVID-19 symptoms: mild (1), moderate (2), and critical (3). We checked the suitability of the model by testing the proportional odds assumption based on the Brant test and found that the proportional odds model fit our data (Table 2). The associations of T2D and its genetic susceptibility with the severity of COVID-19 using the proportional odds model are described in Table 2. The estimated odds ratio (OR) of T2D was 2.824 (95% CI = 2.644–3.016), implying that the odds of COVID-19 severity among T2D patients was 2.824 times higher than in controls.
We included T2D PRS as a covariate and the estimated OR of T2D PRS was 1.080 (95% CI = 1.050–1.111), meaning that the odds of COVID-19 severity among individuals with a one-unit increase (=standard deviation [SD]) in T2D PRS were 1.080 times greater than those of the controls. When we included the PRS group as a covariate (with low PRS group as the reference), the estimated OR of the high PRS group compared to the low PRS group was 1.211 (95% CI = 1.102–1.330), meaning that the odds of COVID-19 severity in the high PRS group were 1.211 times greater than that in the low PRS group. We also included vaccination status as a covariate and found that the estimated OR of vaccination status was 0.206 (95% CI = 0.193–0.220), indicating that the odds of COVID-19 severity in the vaccinated group was 0.206 times less than that in the non-vaccinated group.
Survival analysis with all participants
To investigate the effect of T2D on survival times in patients with SARS-CoV-2 infection, we conducted survival analysis with all participants (Survival-All). We performed a multivariable Cox proportional hazards (PH) model including T2D and time-dependent coefficients for SARS-CoV-2 infection adjusted for other clinical variables, after confirming that the proportional hazards assumption was not violated for any covariates (Supplementary Table 4, Supplementary Figs. 3 and 4). The hazard ratio (HR) for T2D was 2.266 (95% CI = 2.164–2.372), implying that the risk of death for T2D patients was 2.266 times higher than for controls at any point in time, holding all other covariates constant (Table 3). The HR for SARS-CoV-2 infection within five intervals (i.e., before Delta, Delta, Omicron 1, 2, 3) were 7.333, 2.013, 1.864, 1.906, and 1.986, respectively. This means that the risks of death for SARS-CoV-2-infected individuals were 7.333 (95% CI = 6.902–7.791), 2.013 (95% CI = 1.811–2.238), 1.864 (95% CI = 1.636–2.123), 1.906 (95% CI = 1.692–2.146), and 1.986 (95% CI = 1.744–2.262) times higher than that for controls within five intervals, respectively (Table 3). The variable SARS-CoV-2 infection significantly affected all five intervals, with the effect in the first interval being much higher than that in the other intervals as expected. Figure 3a displays the estimated survival curves stratified with T2D and SARS-CoV-2 infection based on the Survival-All model. Four groups (i.e., controls, T2D patients, COVID-19 patients, T2D patients with SARS-CoV-2 infection) had significantly different (p < 0.05) survival probabilities based on pairwise log-rank tests. In general, SARS-CoV-2 infection led to decreased survival probability compared to controls, and SARS-CoV-2-infected T2D patients had dramatically decreased survival probabilities compared to other groups.
a The survival curves stratified by T2D and SARS-CoV-2 infection (N = 432K) based on the Survival-All model, adjusted for clinical covariates such as age, BMI, gender, genotyping array, and PC 1–4. b The survival curves stratified by PRS groups for T2D and SARS-CoV-2 infection (N = 432K) based on the Survival-All model, adjusted for clinical covariates. c The number of SARS-CoV-2 confirmed cases and total deaths due to COVID-19. d The proportion of SARS-CoV-2 variants in England during the study period.
We next included adjusted T2D PRS as a covariate in the Cox PH model, with adjustments for the clinical variables, to examine the effect of high genetic risk for T2D and SARS-CoV-2 infection on survival times. The HR of T2D PRS was 1.062 (95% CI = 1.042–1.081), indicating that the risk of death for individuals with a one-unit increase (=SD) in T2D PRS is 1.062 times higher than for controls (Table 3). This suggests that genetic susceptibility for T2D increases the risk of death. We further examined the effect of PRS group for T2D and SARS-CoV-2 infection on survival times. As in the proportional odds models, we included the PRS group as a covariate in the model, adjusting for potential confounders. The HR for the high PRS group was 1.181 (95% CI = 1.105–1.261), indicating that the risk of death for individuals in the high PRS group for T2D was 1.181 times greater than that for individuals in the low PRS group for T2D (Table 3). Figure 3b displays the estimated survival curves stratified with PRS group and SARS-CoV-2 infection, based on the Survival-All model. The survival probabilities for individuals without SARS-CoV-2 infection were similar among the three PRS groups, but the three PRS groups with SARS-CoV-2 infection had significantly distinct survival probabilities based on pairwise log-rank tests. SARS-CoV-2 infection generally decreased survival probabilities and, particularly, COVID-19 patients in high PRS group had the lowest survival probability, followed by the medium and low PRS groups. This indicates the potential genetic role of T2D in SARS-CoV-2 infection.
Figure 3c shows the number of confirmed COVID-19 cases and total deaths due to COVID-19, while Fig. 3d displays the proportion of COVID-19 variants in the UK during the study period. Survival probabilities for all groups considerably decreased between 230 and 470 days in Fig. 3a and b. This can be explained by the rapid increase in confirmed COVID-19 cases and deaths in the UK during the surge of the COVID-19 Alpha variant. The dominant COVID-19 variants transitioned to Alpha (red curve), Delta (green), and Omicron variants (purple, turquoise, and blue) as shown in Supplementary Fig. 5. The Alpha variants were more difficult to spread compared to the Delta variants, but the fatality rate of Alpha variants was generally higher than Delta variants, leading to a considerable increase in deaths during the Alpha variant surge in the UK. As in the proportional odds models, we included vaccination status as a covariate. The estimated HR of vaccination status was 0.198 (95% CI = 0.186–0.210), meaning that the risk of death for individuals in the vaccinated group was 0.198 times less than that in the non-vaccinated group (Table 3).
Survival analysis with SARS-CoV-2-infected Individuals
We restricted our study to only SARS-CoV-2-infected individuals (N = 101,086) to evaluate the effect of T2D on survival times in COVID-19 positive patients (Survival-COVID-19). We performed a multivariable Cox PH model similar to the Survival-All model after confirming that the proportional hazards assumption was not violated for all covariates (Supplementary Figs. 6 and 7). We found similar results for the HR of T2D, T2D PRS, PRS groups, and vaccination status to the Survival-All model (Table 4). Instead of including time-dependent coefficients for SARS-CoV-2 infection, we included an indicator for the variants (i.e., early, EU1, Alpha, Delta, Omicron1,2,3) as covariates in the model. Compared to the Omicron3 variant (reference), the HRs of Early, EU1, Alpha, Delta, Omicron1,2 variants were 7.993 (95% CI = 6.636–9.628), 4.454 (95% CI = 3.665–5.414), 5.291 (95% CI = 4.358–6.424), 1.359 (95% CI = 1.117–1.654), 0.847 (95% CI = 0.689–1.040), and 0.789 (95% CI = 0.644–0.968), respectively, indicating that the fatality rate was highest in the early variants, followed by Alpha, EU1, Delta, and Omicron variants.
Figure 4a and b show the estimated survival curves stratified with T2D and PRS groups based on the Survival-COVID-19 model, respectively. We found that T2D patients and controls had significantly different (P < 0.05) survival probabilities and the survival probabilities of the three PRS groups were also significantly different (P < 0.05), according to pairwise log-rank tests. To study the difference in survival probability between individuals aged below 60 years and above 60 years, we divided participants into two groups based on their ages. Survival curves stratified by age groups and T2D are displayed in Fig. 4c, and survival curves stratified by age groups and PRS groups for T2D are in Fig. 4d. Figure 4c and d clearly show that survival probabilities for T2D patients, or individuals with a high genetic predisposition for T2D in the aged population, decreased much more than those in the younger population when compared to controls.
a The survival curves stratified by T2D (N = 101K) based on the Survival-COVID-19 model, adjusted for clinical covariates such as age, BMI, gender, genotyping array, and PC 1–4. b The survival curves stratified by PRS groups for T2D (N = 101K) based on the Survival-COVID-19 model, adjusted for clinical covariates. c The survival curves stratified by age groups and T2D. d The survival curves stratified by age groups and PRS groups for T2D.
Assessment of survival model performances
To assess the performance of survival models, we computed the concordance index (C-index), which represents the proportion of observation pairs for which estimated risk scores based on Cox PH models correspond to their survival times. Based on 10-fold CV, we estimated Cox PH model-based risk scores for all individuals and compared their survival times to estimate the C-index (Supplementary Table 5). The C-index for the Survival-All model was 0.770 for T2D and 0.755 for T2D PRS groups, suggesting that, in rough terms, given two individuals from the data, the survival model can predict with 77% or 75.5% accuracy who will die first. The C-index for the Survival-COVID-19 model was 0.791 for T2D and 0.795 for T2D PRS groups.
Sensitivity analysis
In the sensitivity analysis, we computed individual T2D PRS based on the GWAS result from the Diabetes Genetics Replication And Meta-analysis (DIAGRAM) consortium for 459,119 participants30. To mitigate the confounding effect of BMI on T2D, we utilized T2D summary statistics adjusted for BMI from the DIAGRAM consortium, despite its smaller sample size of 26,676 T2D cases and 132,532 controls, compared to the BMI-unadjusted summary statistics. Supplementary Fig. 8 displays the scatter plot of T2D PRS between UK Biobank-derived PRS and DIAGRAM-derived PRS. We included three PRS groups (low, medium, and high) based on the DIAGRAM-based T2D PRS as covariates in both the Survival-All and the Survival-COVID-19 models. The estimated survival curves, stratified by T2D and PRS groups, as well as survival curves stratified by PRS groups, are represented in Supplementary Fig. 9a and b, respectively. The overall patterns of the survival curves are similar to the corresponding curves in Figs. 3b and 4b.
We presented the results for the association of T2D and T2D PRS using proportional odds models in Table 2, and the effects of T2D and T2D PRS on SARS-CoV-2 infection using logistic models in Supplementary Note 2, Supplementary Tables 6 and 7. The results from these two models differ significantly because the proportional odds models predict the three categories (i.e., moderate, severe, and critical) for COVID-19 severity in individuals infected with SARS-CoV-2, whereas the logistic models only consider the SARS-CoV-2 infection status among all participants. Given that over 22% of the current UK Biobank samples were infected with SARS-CoV-2, the infection status is no longer significantly associated with T2D and T2D PRS, although it was significantly associated during the initial phase of the COVID-19 pandemic due to its rarity. However, the severity of COVID-19 remains strongly associated with T2D and T2D PRS. Therefore, we primarily focused on the association between T2D and severe COVID-19, rather than on SARS-CoV-2 infection.
Partitioned T2D PRS
We constructed individual-level partitioned T2D PRS to capture the five physio-pathological components of T2D, namely beta-cell function, proinsulin, obesity, lipodystrophy, and liver/lipid metabolism31. The SNP lists and weights for these cluster-specific partitioned T2D PRS are detailed in Supplementary Table 8. Utilizing these SNP lists and weights, we calculated the partitioned T2D PRS for UK Biobank samples, presenting the AUC values in Supplementary Table 9. Subsequently, we conducted analyses using the newly estimated partitioned T2D PRS with proportional odds models and survival models. In Supplementary Table 10, the P-values for the OR of the beta-cell and proinsulin PRSs in the proportional odds models were not significant. In contrast, the P-values for the obesity, lipodystrophy, and liver/lipid PRSs were significant. This indicates that only PRSs for traits associated with insulin resistance (i.e. obesity, lipodystrophy, and liver/lipid), not insulin production and secretion (i.e. beta-cell and proinsulin), are significantly related to the severity of COVID-19. However, in Supplementary Tables 11 and 12, no P-values were significant, suggesting that none of the partitioned T2D PRSs are significantly associated with the mortality of COVID-19.
Discussion
In this paper, we investigated the association between T2D or genetic susceptibility to T2D, and COVID-19 severity using proportional odds regression and then systemically examine the impact of T2D in patients who tested positive for COVID-19 on survival time using multivariable Cox PH models with all individuals or only SARS-CoV-2-infected individuals. In the proportional odds models, we found a significant association between the risks of T2D or genetic susceptibility to T2D and the severity of COVID-19. This suggests that individuals with T2D or those genetically predisposed to T2D have a higher likelihood of experiencing severe COVID-19 symptoms, pointing to a role for T2D and T2D-related genetics in the pathogenesis of COVID-19 severity. In the Cox PH model with all individuals, we found that the risk of death for T2D patients and individuals in the high T2D PRS group were higher than for controls. Moreover, SARS-CoV-2-infected individuals with T2D or relatively high genetic risks for T2D showed a significant correlation with increased mortality, indicating a role for genetic interplay in the underlying pathogenesis of T2D and COVID-19. Lastly, when working only with SARS-CoV-2-infected individuals, we studied the association of T2D and T2D PRS with survival times and found that T2D patients who tested positive for COVID-19 had a significantly higher mortality rate than non-diabetic individuals who tested positive for COVID-19. We also discovered that the fatality rate was highest for the early variants, succeeded by the Alpha, EU1, Delta, and Omicron variants and the odds of severe COVID-19 were substantially lower in the vaccinated group compared to the non-vaccinated group.
The use of PRS in identifying the relationship between genetic susceptibility for T2D and severe COVID-19 has several clinical and research implications. First, our results underscored the importance of considering T2D status and genetic susceptibility to T2D in managing COVID-19. Early detection and management of COVID-19 in patients with T2D or high genetic risk for T2D might improve clinical outcomes. Second, in our analysis, we found participants who did not have T2D but were classified in the high PRS group for T2D were more likely to have severe COVID-19. It suggested the potential use of PRS to classify COVID-19 patients for better medical care since the T2D PRS proved useful in estimating the severity and mortality of COVID-19, even amongst individuals who did not currently have T2D. Third, the observed positive relationship between T2D PRS and COVID-19 provides an avenue for discovering the causal impact of novel genetic factors on the development of severe COVID-19. Lastly, the biological profiling of SARS-CoV-2-infected T2D patients may uncover the molecular mechanism underlying the interaction between host genetics and the pathogenesis of SARS-CoV-2 that links T2D to severe COVID-19. This could ultimately suggest new methods for prevention, prediction, and medical therapy for such patients.
This study has several limitations. First, we analyzed only individuals infected with SARS-CoV-2 prior to December 2022. The number of Omicron cases continued to increase after December 2022, and thus further analysis of these cases is necessary to evaluate the effect of T2D and susceptibility to T2D on the severity and mortality of COVID-19. Second, a variety of vaccines (e.g., Pfizer, Moderna, and Johnson & Johnson) have been continuously utilized worldwide, and thus we need to study the effect of these COVID-19 vaccines on the mortality rate. Thrid, we only included the Caucasian population, with an average age of 58 years and thus we should be cautious about generalizing these findings to other populations or age groups. Fourth, we found high genetic correlation between T2D and severity of COVID-19 and the usefulness of T2D PRS in estimating the severity and mortality of COVID-19 but we did not further investigate the shared genetic architecture between T2D and COVID-19. We are currently identifying their shared genetic loci based on cross-trait meta-analysis32 and detecting gene expression associations in specific tissues for T2D and COVID-19 based on transcriptome-wise association study (TWAS)33. Fifth, despite the high genetic correlation between T2D and severe COVID-19, the heritability of severe COVID-19 is considerably low (~3%) compared to that of T2D (~15%). Although the genetic correlation between T2D and severe COVID-19 is statistically significant, further careful investigation is needed to interpret the shared genetic architecture between them. Lastly, while our analysis primarily focused on the relationship between T2D and severe COVID-19 due to their well-established association, it is necessary to further examine the association between other underlying diseases (i.e. respiratory diseases, hypertension, cardiovascular disease, dementia) and severe COVID-19, given the considerable genetic correlation between these diseases and severe COVID-19.
In conclusion, our study was undertaken to investigate the association between T2D, or genetic susceptibility to T2D, and COVID-19 severity using UK Biobank data. We aimed to understand the impact of T2D in patients who tested positive for COVID-19 on survival time, and how genetic risks for T2D, calculated using individual PRS, influenced these outcomes. We found that T2D and genetic predisposition for T2D were significantly associated with the risk of COVID-19 severity, and an increased mortality rate and T2D PRS proved useful in estimating the severity and mortality of COVID-19. These observations should help clinicians categorize patients based on their underlying diseases (i.e. T2D), their PRS as well as the severity of COVID-19, to provide more effective treatment. Lastly, our findings should aid further research to delineate the complex interactions between SARS-CoV-2 infections, COVID-19 severity, airway inflammation, and other outcomes in patients with T2D.
Methods
Study population and design
The UK Biobank is a large-scale cohort study that includes over 500,000 participants aged between 40 and 69 at the time of recruitment from the UK during the period of 2006–201034. For our analysis, we utilized questionnaire-based self-reported data, hospital inpatient data, and death data, all updated as of December 19, 2022. We also incorporated COVID-19 test result data, which included records of polymerase chain reaction (PCR) tests for SARS-CoV-2. We restricted our study to only individuals of European ancestry based on the information of self-reported ethnicity, resulting in 459,119 individuals. The SNPs were filtered out if they had a minor allele frequency (MAF) of <0.01 and an imputation quality score of <0.8, leading to 9,572,556 SNPs.
Heritability and genetic correlations
The GWAS summary statistics of T2D and related traits were produced from a large-scale cohort data of 459,119 participants in the UK Biobank. These statistics were adjusted for potential confounders such as age, age squared, gender, genotype PCs, assessment array, and genotyping array using BOLT-LMM v2.3 software35. To estimate heritability and genetic correlations between T2D, its related traits, and COVID-19, we performed cross-trait LD score regression (LDSC) analysis based on GWAS summary statistics using the LDSC software36. We utilized the 1000 Genomes project data (phase 3 version 5) as a reference panel and restricted our study to those of European ancestry with common and well-imputed HapMap3 SNPs. We estimated heritability and genetic correlation between various traits, including T2D, SARS-CoV-2 infection, and severe COVID-19.
Polygenic risk score (PRS)
The PRS represents an individual’s genetic liability for a particular trait, calculated by multiplying the genotype dosage for each variant by its corresponding weight and summing across all variants37,38. To compute the individual PRS of 459,119 participants, we utilized a 10-fold cross-validation (CV) approach. We first generated 10 GWAS summary statistics with 10 training sets for each trait, then computed PRS for 10 validation sets using only HapMap3 SNPs using the LDpred method39, thus obtaining PRS scores for all individuals. We merged the PRS scores for the 10 folds after standardizing the PRS in each fold to avoid bias. The candidate PRSs were calculated with a range of tuning parameters (i.e., the fraction of causal variants), and the PRS with the b est discriminative capability was selected based on the maximal area under the curve (AUC).
Proportional odds regression
To investigate the effects of T2D and genetic susceptibility for T2D on the severity of COVID-19, we divided participants into three groups according to the severity of their COVID-19 symptoms: mild (1), moderate (2), and critical (3). If samples were obtained from the participants were hospitalized with COVID-19, they were classified as having moderate COVID-19. If the participants died due to COVID-19, they were classified as having critical COVID-19. Otherwise, they were classified as having mild COVID-19. With the proportional odds assumption, we performed a proportional odds regression for T2D or T2D PRS, and the analysis proceeded separately for T2D or T2D PRS: \({{\rm {logit}}}(P({Y}_{i}\le j))={\beta }_{0j}+{\beta }_{1}{X}_{i}+{{{{{{\boldsymbol{\beta }}}}}}}_{{{{{{\boldsymbol{2}}}}}}}{{{{{{\boldsymbol{C}}}}}}}_{{{{{{\boldsymbol{i}}}}}}}\) for \(j={{{{\mathrm{1,2,3}}}}}\) where \({Y}_{i}\) is an ordinal outcome for COVID-19 severity (1:mild, 2: moderate, 3:critical) of ith individual; \(j\) is the severity of COVID-19; \({X}_{i}\) is a T2D or T2D PRS; \({{{{{{\boldsymbol{C}}}}}}}_{{{{{{\boldsymbol{i}}}}}}}\) is a set of clinical covariates including age, gender, vaccination status, BMI, genotyping array, and the top four PCs. We also included an indicator for the three PRS groups for T2D (i.e., low, medium, and high) as a covariate (with low as the reference) in the model. The assessment of model performances and the assumption of constant effect sizes of risk factors were tested using Brant tests.
Survival analysis with all participants
To examine the effect of T2D and SARS-CoV-2 infection on survival time, we performed survival analysis using multivariable Cox proportional hazards (PH) models with all participants (Survival-All). We excluded individuals who died before and after the study period, resulting in a total of 431,975 participants included in the study. Out of these, 12,911 participants (2.99%) died during the study period, and the remaining ones (97.01%) were censored. Survival time was defined as the number of days from the start date of the study to the date of death or the last date of the study. The effect of SARS-CoV-2 infection is not constant over the study period, which is a violation of the proportional hazards assumption. Thus, we divided the analysis time into five time-intervals based on SARS-CoV-2 variants such as Delta (494 days until the proportion of the Delta variant had exceeded 95%), Omicron1 (672), Omicron2 (757), and Omicron3 (851) and Cox proportional hazards model was stratified for these time-intervals28. The effect of SARS-CoV-2 infection varied over time, which can be explored via stratification by time. We fit the following multivariable Cox proportional hazards models with time-dependent coefficients: \(h({t|}{{{{{{\boldsymbol{x}}}}}}}_{{{{{{\boldsymbol{i}}}}}}})={h}_{0}(t)\exp ({\beta }_{1}{X}_{1i}+{{{{{{\boldsymbol{\beta }}}}}}}_{{{{{{\boldsymbol{2}}}}}}}{{{{{{\boldsymbol{X}}}}}}}_{{{{{{\boldsymbol{2}}}}}}{{{{{\boldsymbol{i}}}}}}}+{{{{{{\boldsymbol{\beta }}}}}}}_{{{{{{\boldsymbol{3}}}}}}}{{{{{{\boldsymbol{C}}}}}}}_{{{{{{\boldsymbol{i}}}}}}})\) where \({X}_{1i}\) is a T2D or T2D PRS; \({{{{{{\boldsymbol{X}}}}}}}_{{{{{{\boldsymbol{2}}}}}}{{{{{\boldsymbol{i}}}}}}}\) is a set of indicators for one of the five time-intervals based on SARS-CoV-2 variants (i.e. \({{{{{{\boldsymbol{X}}}}}}}_{{{{{{\boldsymbol{2}}}}}}{{{{{\boldsymbol{i}}}}}}} = {\left(I\left(t \, \le \, 494\right),{I}\left(494 \, < \, t \, \le \, 672\right),{I}\left(672 \, < \, t\le 757\right),{I}\left(757 \, < \, t \, \le \, 851\right),{I}\left(851 < t\right)\right)}^{{{{{{\rm{T}}}}}}}\)); \({{{{{{\boldsymbol{C}}}}}}}_{{{{{{\boldsymbol{i}}}}}}}\) is a set of clinical covariates. We checked the proportional hazards assumption of this stratified Cox regression model. Again, we included an indicator for the three PRS groups for T2D (i.e., low, medium, and high) as a covariate in the model. We performed stratified Cox PH analysis and pairwise log-rank tests to assess the differences in survival probabilities for groups stratified by T2D and SARS-CoV-2 infection.
Survival analysis with only SARS-CoV-2-infected individuals
We further performed survival analysis with only SARS-CoV-2-infected individuals (Survival-COVID-19). For this analysis, we only included individuals with confirmed SARS-CoV-2 infection dates, leading to a total of 101,086 patients included in the study. Among these participants, the number of deaths was 2,836 (2.8%), while the number of censored individuals was 98,250 (97.2%). For the Survival-COVID-19 model, survival time was defined as the number of days from the date of SARS-CoV-2 infection to the date of death or the last date of the study. We performed multivariable Cox PH models, similar to the Survival-All model, but excluded the indicator for SARS-CoV-2 infection, as we only considered SARS-CoV-2-infected individuals for the Survival-COVID-19 model. Instead, we included an indicator for SARS-CoV-2 variants as a covariate in the model to investigate the effects of SARS-CoV-2 variants on the mortality rate. We also estimated survival probability stratified by T2D or T2D PRS groups, and the differences in survival probabilities across the groups were assessed using pairwise log-rank tests.
Statistics and reproducibility
Statistical analyses and figures were created using R version 4.2.3. We conducted the statistical analyses using multivariable logistic regression models, proportional odds models, and Cox proportional hazards models. All the codes used are publicly available on GitHub (https://github.com/wonilchung/CovidT2D) to guarantee the reproducibility of all analyses.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
All data are available for academic uses at https://github.com/wonilchung/CovidT2D. All other data are available from the corresponding author upon reasonable request.
References
Sohrabi, C. et al. World Health Organization declares global emergency: a review of the 2019 novel coronavirus (COVID-19). Int. J. Surg. 76, 71–76 (2020).
Organization, W.H. Coronavirus Disease 2019 (COVID-19): Situation Report, 73 (World Health Organization, 2020).
Rothan, H. A. & Byrareddy, S. N. The epidemiology and pathogenesis of coronavirus disease (COVID-19) outbreak. J. Autoimmun. 109, 102433 (2020).
Pathak, G. A. et al. A first update on mapping the human genetic architecture of COVID-19. Nature 608, E1–E10 (2022).
Lhuillier, E. et al. The impact of World Trade Center related medical conditions on the severity of COVID-19 disease and its long-term sequelae. Int. J. Environ. Res. Public Health 19, 6963 (2022).
Groff, D. et al. Short-term and long-term rates of postacute sequelae of SARS-CoV-2 infection: a systematic review. JAMA Netw. Open 4, e2128568–e2128568 (2021).
Williamson, E. J. et al. Factors associated with COVID-19-related death using OpenSAFELY. Nature 584, 430–436 (2020).
Biswas, M., Rahaman, S., Biswas, T. K., Haque, Z. & Ibrahim, B. Association of sex, age, and comorbidities with mortality in COVID-19 patients: a systematic review and meta-analysis. Intervirology 64, 36–47 (2021).
Ejaz, H. et al. COVID-19 and comorbidities: deleterious impact on infected patients. J. Infect. Public Health 13, 1833–1839 (2020).
Ryan, D. H., Ravussin, E. & Heymsfield, S. COVID 19 and the patient with obesity–the editors speak out. Obesity (Silver Spring, MD) 28, 847 (2020).
Simonnet, A. et al. High prevalence of obesity in severe acute respiratory syndrome coronavirus‐2 (SARS‐CoV‐2) requiring invasive mechanical ventilation. Obesity 28, 1195–1199 (2020).
Grasselli, G. et al. Baseline characteristics and outcomes of 1591 patients infected with SARS-CoV-2 admitted to ICUs of the Lombardy Region, Italy. JAMA 323, 1574–1581 (2020).
Gregory, J. M. et al. COVID-19 severity is tripled in the diabetes community: a prospective analysis of the pandemic’s impact in type 1 and type 2 diabetes. Diabetes Care 44, 526–532 (2021).
Norouzi, M. et al. Type-2 diabetes as a risk factor for severe COVID-19 infection. Microorganisms 9, 1211 (2021).
Zheng, Y.-Y., Ma, Y.-T., Zhang, J.-Y. & Xie, X. COVID-19 and the cardiovascular system. Nat. Rev. Cardiol. 17, 259–260 (2020).
Guzik, T. J. et al. COVID-19 and the cardiovascular system: implications for risk assessment, diagnosis, and treatment options. Cardiovasc. Res. 116, 1666–1687 (2020).
Hariyanto, T. I., Putri, C., Situmeang, R. F. V. & Kurniawan, A. Dementia is a predictor for mortality outcome from coronavirus disease 2019 (COVID-19) infection. Eur. Arch. Psychiatry Clin. Neurosci. 271, 393–395 (2021).
Niemi, M. E., Daly, M. J. & Ganna, A. The human genetic epidemiology of COVID-19. Nat. Rev. Genet. 23, 533–546 (2022).
COVID-19 Host Genetics Initiative. The COVID-19 host genetics initiative, a global initiative to elucidate the role of host genetic factors in susceptibility and severity of the SARS-CoV-2 virus pandemic. Eur. J. Hum. Genet. 28, 715–718 (2020).
Guo, H., Li, T. & Wen, H. Identifying shared genetic loci between coronavirus disease 2019 and cardiovascular diseases based on cross-trait meta-analysis. Front. Microbiol. 13, 993933 (2022).
Ni, J., Qiu, L.-J., Yin, K.-J., Chen, G.-M. & Pan, H.-F. Shared genetic architecture between type 2 diabetes and COVID-19 severity. J. Endocrinol. Investig. 46, 501–507 (2023).
Horowitz, J. E. et al. Genome-wide analysis provides genetic evidence that ACE2 influences COVID-19 risk and yields risk scores associated with severe disease. Nat. Genet. 54, 382–392 (2022).
Waszczuk, M. A. et al. Polygenic risk scores for asthma and allergic disease associate with COVID-19 severity in 9/11 responders. PLoS ONE 18, e0282271 (2023).
Zhu, Z. et al. Association of obesity and its genetic predisposition with the risk of severe COVID-19: analysis of population-based cohort data. Metabolism 112, 154345 (2020).
Alemany-Navarro, M. et al. Psychiatric polygenic risk as a predictor of COVID-19 risk and severity: insight into the genetic overlap between schizophrenia and COVID-19. Transl. Psychiatry 13, 189 (2023).
Qu, H.-Q., Qu, J., Glessner, J. & Hakonarson, H. Mendelian randomization study of obesity and type 2 diabetes in hospitalized COVID-19 patients. Metabolism 129, 155156 (2022).
Cao, H., Baranova, A., Wei, X., Wang, C. & Zhang, F. Bidirectional causal associations between type 2 diabetes and COVID‐19. J. Med. Virol. 95, e28100 (2023).
Zhang, Z., Reinikainen, J., Adeleke, K.A., Pieterse, M.E. & Groothuis-Oudshoorn, C.G.M. Time-varying covariates and coefficients in Cox regression models. Ann. Transl. Med. 6, 121 (2018).
Topless, R. K. et al. Gout and the risk of COVID-19 diagnosis and death in the UK Biobank: a population-based study. Lancet Rheumatol. 4, e274–e281 (2022).
Scott, R. A. et al. An expanded genome-wide association study of type 2 diabetes in Europeans. Diabetes 66, 2888–2902 (2017).
DiCorpo, D. et al. Type 2 diabetes partitioned polygenic scores associate with disease outcomes in 454,193 individuals across 13 cohorts. Diabetes Care 45, 674–683 (2022).
Lee, C. H., Shi, H., Pasaniuc, B., Eskin, E. & Han, B. PLEIO: a method to map and interpret pleiotropic loci with GWAS summary statistics. Am. J. Hum. Genet. 108, 36–48 (2021).
Gusev, A. et al. Integrative approaches for large-scale transcriptome-wide association studies. Nat. Genet. 48, 223–341 (2016).
Allen, N. E., Sudlow, C., Peakman, T. & Collins, R. UK biobank data: come and get it. Sci. Transl. Med. 6, 224ed4–224ed4 (2014).
Loh, P.-R. et al. Efficient Bayesian mixed-model analysis increases association power in large cohorts. Nat. Genet. 47, 284 (2015).
Bulik-Sullivan, B. K. et al. LD score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat. Genet. 47, 291 (2015).
Chung, W. et al. Efficient cross-trait penalized regression increases prediction accuracy in large cohorts using secondary phenotypes. Nat. Commun. 10, 569 (2019).
Chung, W. Statistical models and computational tools for predicting complex traits and diseases. Genom. Inform. 19, e36 (2021).
Vilhjálmsson, B. J. et al. Modeling linkage disequilibrium increases accuracy of polygenic risk scores. Am. J. Hum. Genet. 97, 576–592 (2015).
Acknowledgements
We appreciate the individuals who participated in the study. This research has been conducted using the UK Biobank Resource (application numbers 45052, 58105, 77890). This research was supported by the Bio and Medical Technology Development Program of the National Research Foundation (NRF) funded by the Korean government (2021M3E5E3081425). This work was also supported by the NRF grant funded by the Korea government (2020R1C1C1A01012657) and Basic Science Research Program of the NRF funded by the Korea government (2021R1A6A1A10044154). This work was also supported by the Soongsil University Research Fund.
Author information
Authors and Affiliations
Contributions
A.L., L.L., T.P., and W.C. conceived and designed the experiments. A.L., J.S., S.P., Y.C., G.K., J.L., and W.C. performed the experiments and analyzed the data. They were also responsible for generating the necessary data, materials, and analysis tools. A.L. and W.C. primarily wrote the manuscript. A.L., L.L., T.P., and W.C. reviewed and revised the manuscript.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Communications Biology thanks Marie Pigeyre and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Hélène Choquet and Luke R. Grinham. A peer review file is available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Lee, A., Seo, J., Park, S. et al. Type 2 diabetes and its genetic susceptibility are associated with increased severity and mortality of COVID-19 in UK Biobank. Commun Biol 7, 122 (2024). https://doi.org/10.1038/s42003-024-05799-1
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s42003-024-05799-1
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
