Baseline characteristics
In total, 4,007 eligible subjects (596 non-OSA, and 3,411 moderate-to-severe OSA) were enrolled in this study (see flow chart in Figure 1). Of the 4007 participants enrolled, 596 were Non-OSA, 831 were moderate OSA and 2580 were severe OSA. Of the non-OSA subjects, the median age was 34 (range 29-43), the median HOMA-IR was 1.172 (range 1.11-2.55), median serum APOA-I was 1.092(range 0.092-1.15) g/L, medium serum APOB was 0.77(range 0.65-0.89) g/L, the percentage of insulin resistance was 26.8%, the percentage of metabolic syndrome was 27.2%, the median AHI value was 2 (range 0.8-3.4) events/h. Compared with non-OSA, patients with OSA were older and had higher serum concentrations of glucose, insulin, sleep parameters and ratio of smoking, drinking, prevalence of insulin resistance and percentage of MetS. OSA patients had higher levels of anthropometric parameter, such as SBP, DBP, BMI, NC, WC, HC, WHR. With the exception of serum APOA-I, all biochemistry parameters, demographic parameters and sleep parameters were also significantly different among the groups (Table 1). There were more subjects with insulin resistance and MetS in OSA group compared to non-OSA (P<0.001). The percentages of insulin resistance in non-OSA, moderate, and severe OSA were 26.8%, 46.8%, and 63.4%, respectively. The percentages of MetS in non-OSA, moderate, and severe OSA were 27.2%, 50.2%, and 64.3%, respectively.
The basic characteristics of the SNPs and SNP’s β coefficients in GRS construction are listed in Table 2. APOA-I SNP rs964184, rs10047462, rs888246 had negative correlations (ß=-0.013, -0.008, -0.014, respectively) while rs9804646 had positive correlation with serum APOA-I levels (ß=0.015). APOB SNP 1042031, rs693, rs1367117 associated with APOB levels positively (ß=0.017, 0.009, 0.012, respectively), while rs2854725, rs12713956 associated with APOB levels negatively (both ß=-0.03).
We also assessed the clinical characteristics of the participants in the top quintiles of the APOA GRS and APOB GRS compared with those in the bottom fifth quintiles (Tables S1 and S2, respectively). The basic characteristics of insulin, insulin resistance, and TG were lower in the highest quintile of APOA GRS than in the lowest quintile of APOA GRS, whereas HDL-C, LDL-C, APOA, and APOA/APOB were higher (all P < 0.05; Table S1). As expected, TC, APOB, and LDL-C levels were higher in the highest quintile of APOB GRS than in the lowest quintile of APOB GRS (all P < 0.05, Table S2).
Associations between common APOA-I and APOB genotypes and serum APOA-I and APOB levels
All SNPs of APOA-I and APOB were in Hardy-Weinberg equilibrium (P > 0.05). The associations of SNPs with serum APOA-I and APOB levels are summarized in Table 2. APOA-I rs964184 and rs9804646 were associated with APOA-I (β = −0.013, P = 0.021; β = 0.015, P = 0.006). APOB SNP rs2854725 was significantly associated with APOB (β = −0.03, P < 0.001). APOB SNPs rs1367117 and rs12713956 were moderately associated with increased APOB levels (β = 0.012, P = 0.008; β = −0.03, P = 0.004, respectively) (Table 2).
Associations of serum APOA-I and APOB and their GRS with insulin resistance and MetS risks
The associations between each SNP of APOA-I and APOB with insulin resistance and MetS are listed in Table S3. APOA-I SNPs rs9804646 and rs888246 were associated with insulin resistance (OR = 0.856, 95% confidence interval [CI]: 0.756–0.968, P = 0.013; OR = 1.340, 95% CI: 1.069–1.680, P = 0.011) after adjustment. APOA-I SNPs rs964184, rs9804646, and rs888246 were significantly associated with MetS (OR = 1.353, 95% CI: 1.201–1.523, P < 0.01; OR = 0.777, 95% CI: 0.69–0.874, P < 0.01; OR = 1.274, 95% CI: 1.024–1.586, P = 0.03, respectively) after adjustment. For APOB, only rs2854725 was associated with MetS (OR = 0.829, 95% CI: 0.718–0.956, P = 0.01) after adjusting for age, gender, and BMI.
Serum APOA-I levels decreased the risk of insulin resistance and MetS (Table 3) (OR=0.573. P < 0.001; OR=0.131, P < 0.001), and had no relationship with insulin resistance after adjustment (OR=0.8233, P = 0.308). Serum APOA-I levels remained to decrease the risk of MetS after adjustment (OR=0.09, P < 0.001). Serum APOB levels increased the risk of insulin resistance and MetS (OR=0.573, P < 0.001; OR=0.131, P < 0.001), which remained after adjusting for age, gender, and BMI (OR=3.168, P < 0.001; OR=6.098, P < 0.001). APOA-I GRS decreased insulin resistance (OR = 0.923, 95% CI: 0.880–0.968, P < 0.001) and MetS (OR = 0.886, 95% CI: 0.845–0.929, P < 0.001) significantly, which remained after adjusting for age, gender, and BMI (all P < 0.001). APOB GRS was not associated with insulin resistance and MetS (P > 0.05). Subjects with OSA had higher APOB/APOA-I and APOB levels (OR = 7.610, 95% CI: 4.886–11.852, P < 0.001; OR = 31.683, 95% CI: 18.231–55.059, P < 0.001, respectively), even after adjusting for age, gender, and BMI.
We also stratified APOA-I and APOB GRS into quintiles. When compared with the bottom quintile, subjects in the top quintile of the APOA-I GRS group had a lower risk of insulin resistance and MetS (Table 4) [OR = 0.753 (0.63–0.90), P = 0.002; OR = 0.651 (0.546–0.777), P < 0.001], even after adjusting for age, gender, and BMI [OR = 0.761 (0.623–0.929), P = 0.007; OR = 0.637 (0.526–0.773), P < 0.001]. APOB GRS was not associated with insulin resistance (OR=1.364, P=0.610), and had weak associations with MetS (OR=1.072, P=0.042). Linear regression analysis revealed that APOA-I GRS was associated with decreased insulin, TG, HOMA-IR, APOB/APOA-I, and elevated HDL-C, LDL-C, and APOA-I levels (all P < 0.05, Table S4), even after adjustment. APOB GRS was associated with elevated TC, LDL-C, APOB, and APOB/APOA-I levels (all P < 0.001, Table S5).
Percentages of independent contributors of HOMA-IR
To reveal the percentages of independent contributors of HOMA-IR, stepwise multivariate linear regression analysis was performed. APOA-I GRS, age, gender, and BMI were included in model 1. APOA GRS, gender, and BMI explained 0.099%, 0.14%, and 18% of HOMA-IR, respectively (Table S6). As AHI was identified as an important marker of OSA, we included AHI in model 2, APOA-I GRS, gender, BMI, and AHI explained 0.1%, 0.14%, 18%, and 0.94% of HOMA-IR, respectively (all P < 0.05, Table S6).