Obstructive sleep apnea mediates genetic risk of Diabetes Mellitus in Hispanic and Latino communities

Abstract

Background: Obstructive sleep apnea (OSA), and other sleep disorders, are associated with increased risk of developing diabetes mellitus (DM). We examined whether sleep disorders influence the genetic risk of developing diabetes in Hispanic/Latino individuals.

Methods: We developed Type 2 Diabetes (T2D) polygenic risk score (T2D-PRS) useful in admixed Hispanic/Latino individuals. We estimated the association of the T2D-PRS with cross-sectional (n = 12,342) and incident (n = 6965) DM in the Hispanic Community Health Study/Study of Latinos (ages 18-76, 50.9% female). We conducted a mediation analysis with T2D-PRS as an exposure, incident DM as an outcome, and OSA as a mediator. Additionally, we performed Mendelian randomization (MR) analysis to assess the causal relationship between T2D and OSA.

Results: Here, we show that a 1 standard deviation increase in T2D-PRS has DM adjusted odds ratio (OR) = 2.67, 95% CI [2.40; 2.97] and a higher incident DM rate (incident rate ratio (IRR) = 2.02, 95% CI [1.75; 2.33]). In a stratified analysis based on OSA severity categories the associations are stronger in individuals with mild OSA compared to those with moderate to severe OSA. Mediation analysis suggests that OSA mediates the T2D-PRS association with DM. In two-sample MR analysis, T2D has a causal effect on OSA, OR = 1.03, 95% CI [1.01; 1.05], and OSA has a causal effect on T2D, with OR = 2.34, 95% CI [1.59; 3.44].

Conclusions: These results support a causal association between OSA and DM, with OSA mediating up to 4.7% of the genetic risk for DM. OSA treatment may reduce DM prevalence.

Fig. 1. T2D-PRSs associations with DM.

a Proportion of individuals by DM status over time (between clinic’s visit 1 and 2) stratified by T2D-PRS quartiles. The number of individuals represented by each tile is written on the tile. Note: “Persistent” refers to having the same DM status in both visit 1 and visit 2; “Worsen” refers to change in DM status from normoglycemic at visit 1 to hyperglycemic at visit 2 or from hyperglycemic in visit 1 to diabetic at visit 2; “Improve” refers to change in DM status from hyperglycemic at visit 1 to normoglycemic at visit 2 or from diabetic in visit 1 to hyperglycemic or normoglycemic at visit 2. b Estimated OR per 1 SD increase in T2D-PRSs in association with DM at baseline in HCHS/SOL individuals. c Estimated IRR per 1 SD increase in T2D-PRSs in association with incident DM in individuals free of DM at baseline. In b and c colors correspond to different PRSs, as denoted in the legend. d Association of 1 SD increased in mgbPRSsum with DM and incident DM in HCHS/SOL individuals stratified by OSA severity levels and overall dataset. Colors correspond to OSA strata, as denoted in the legend. All p-values were obtained from the 1 degree-of-freedom Wald test. OSA severity levels were defined based on the respiratory even index (REI): mild OSA was defined as 15 ≥ REI ≥ 5, moderate-to-severe OSA was defined as REI ≥ 15, and REI < 5 was considered no OSA. Throughout, error bars represent 95% confidence intervals. All models were adjusted for age, sex, BMI, study center and 5 genetic PCs. OR odds ratios, IRR incidence rate ratios, AUC Area Under the ROC (receiver operating characteristic) Curve, T2D type 2 diabetes, PRSs polygenic risk scores, DM diabetes mellitus, HCHS/SOL Hispanic Community Health Study/Study of Latinos, EDS excessive daytime sleepiness, OSA obstructive sleep apnea, SD standard deviation.

Fig. 2. Associations of T2D-PRSs with poor sleep phenotypes, estimated mediation effect by OSA, causal effects of T2D on OSA and OSA on T2D.

a Estimated OR per 1 SD increase of mgbPRSsum in association with poor sleep health at baseline in HCHS/SOL individuals. The comparison categories are individuals without the stated poor sleep phenotype (e.g., short sleep versus individuals who do not have short sleep, etc.). b Distribution of mgbPRSsum computed over all individuals with genetic data and DM in visit 2 (N = 2483), horizontal dashed lines denote quantiles of the PRS values Q₀–Q₄. c Estimated percents of risk mediation by mild-to-severe OSA in the association between T2D-PRS and incident DM in individuals who participated at the second visit to a clinic (N = 6291). Darker shades correspond to higher estimates. Estimates are provided for set values of the T2D-PRS, selected according to the distribution quantiles. Significance codes: 0 >= ‘***’ <0.001 >= ‘**’ <0.01 >= ‘*’ <0.05 ‘’ <0.1. d Estimated causal effect of T2D on OSA based on SNPs selected using p-value threshold <5 × 10⁻⁸ in BMI-adjusted and BMI-unadjusted T2D GWASs. e Estimated causal effect of OSA on T2D based on SNPs selected using p-value threshold <5 × 10⁻⁸ in BMI-adjusted and BMI-unadjusted OSA GWASs. Estimates are provided from a few MR methods (primary method: IVW, secondary: MR-PRESSO, and MR-RAPS), denoted by different colors. Results are not presented for MR-PRESSO in panel (d) for estimated effect of OSA on T2D in BMI adjusted analysis because they were the same as the results for the IVW method. Throughout, error bars represent 95% confidence intervals. All models were adjusted for age, sex, BMI, study center and 5 genetic PCs. T2D type 2 diabetes, OSA obstructive sleep apnea, IVW inverse variance weighted, BMI body mass index, AUC Area Under the ROC (receiver operating characteristic) Curve, SD standard deviation, SNPs Single-nucleotide polymorphism, GWAS genome wide association study