2.1 General characteristics of the study participants
A total of 101,510 individuals participated in the 2006 annual health check-up, and data for 49,518 participants were included in the statistical analyses after applying the inclusion and exclusion criteria (Fig. 1). The mean age of the participants was 50.10 ± 12.86 years; 36,117 males and 13,401 females were included. The participants were divided into 4 groups according to the TG/HDL-C quartile: 12,384 participants in the first quartile group (T1 group; TG/HDL-C < 0.52), 12,526 participants in the second quartile group (T2 group; 0.52 ≤ TG/HDL-C < 0.75), 12289 participants in the third quartile group (T3 group; 0.75 ≤ TG/HDL-C < 1.13), And 12319 participants in the fourth quartile group (T4 group; TG/HDL-C ≥ 1.13). Statistically significant (P < 0.05) between-group differences were observed for participants’ age, proportion of males, BMI, WC, SBP, DBP, FBG, TG, TC, HDL-C, LDL-C, ALT, Hs-CRP, UA, smoking status, physical activity status, hypertension status, diabetes status, high school education status and above, lipid-lowering drug use, and TG/HDL-C (Table 1). The data showed that participants in the subgroup with a higher TG/HDL-C ratio also had a greater proportion of males; greater BMI, WC, SBP, DBP, FBG, TG, TC, LDL-C, ALT, Hs-CRP, and UA; and greater proportions of smoker status, hypertension status, and diabetes status.
2.2 Follow-up and incidence of NAFLD
The average follow-up time for the 49518 participants was 7.62 ± 3.99 years, and a total of 24,838 participants were diagnosed with NAFLD; 17,928 of those participants were male, and 6,910 were female. The person-year incidence rates of the individuals in the T1 ~ T4 groups were 48.34/1,000 person-years, 57.30/1,000 person-years, 70.69/1,000 person-years, and 93.08/1,000 person-years, respectively. The cumulative incidences of NAFLD in the T1 to T4 groups were 59.16%, 65.04%, 71.27% and 79.28%, respectively. The difference in cumulative incidence between the four groups was statistically significant according to the log-rank test (χ2 = 1785.71 p < 0.001) (Fig. 2).
2.3 Analysis of risk factors for NAFLD
According to the multivariate Cox proportional hazards model, age, sex, WC, TC, LDL-C, UA, Hs-CRP, hypertension status, smoking status, physical activity status and ALT were found to be influential factors in the development of NAFLD (Pp < 0.05). Model 1 was a one-way analysis with HRs (95% CIs) of 1.20 (1.16–1.25), 1.50 (1.45–1.56), and 2.02 (1.95–2.10) in groups T2, T3 and T4, respectively, compared with those in group T1 (p for trend < 0.05). Model 2 was adjusted for age and sex, and the HRs (95% CIs) were 1.21 (1.16–1.25), 1.51 (1.45–1.57), and 2.03 (1.96–2.11) for groups T2, T3 and T4, respectively, compared with those in group T1 (P for trend < 0.05). Model 3 was further adjusted for WC, TC, LDL-C, ALT, UA, Hs-CRP, diabetes status, hypertension status, smoking status, physical activity status, high school education status and above and lipid-lowering drug use on the basis of Model 2; the HRs (95% CIs) for groups T2, T3 and T4 compared with those of group T1 were 1.13 (1.08 ~ 1.17), 1.32 (1.28 ~ 1.38), and 1.60 (1.54 ~ 1.66) (P for trend < 0.05) (Table 2). The HR (95% CI) corresponding to each standard-deviation increase in TG/HDL-C in Model 3 was 1.10 (1.09 ~ 1.11) (P < 0.05). The RCS was plotted using knots at the 5th, 50th, and 95th percentiles for the continuous TG/HDL-C variable. After adjusting for sex, age, WC, TC, LDL-C, ALT, UA, Hs-CRP, diabetes status, hypertension status, smoking status, physical activity status, high school education status and above, and lipid-lowering medication use, the risk of NAFLD was linearly associated with the TG/HDL-C ratio. The result of the linear hypothesis test of the association between the TG/HDL-C ratio and the risk of NAFLD was χ2 = 756.77 (P < 0.001), and the result of the nonlinear test was χ2 = 209.06 (P < 0.001), which indicated that there was a nonlinear relationship between the TG/HDL-C ratio and the risk of NAFLD (Fig. 3).
2.4 Stratified analysis
Stratified analyses were also conducted to further investigate the association between the TG/HDL-C ratio and the development of NAFLD in populations with different characteristics. A Cox proportional hazards model was constructed with the occurrence of NAFLD as the dependent variable, TG/HDL-C quartile grouping as the independent variable and T1 as the control group. Cox proportional hazards models were constructed after stratifying the participants by sex, and the adjusted confounders were the same as those in Model 3 except for sex. In the age stratification, Cox proportional hazards analyses were repeated after the participants were divided into two groups based on whether they were older than 40 years of age, and the adjusted confounders were the same as those in Model 3 except for age. In the stratification based on WC, the overall group was divided into abdominal obesity and normal WC groups based on whether the male participants had a WC greater than or equal to 90 cm and whether the female participants had a WC greater than or equal to 85 cm. Cox proportional models were constructed again afterwards, with adjusted confounders identical to those of Model 3 except for WC. This study also included stratified analyses for hypertension status and diabetes status. In both stratified analyses, the adjusted confounders were the same as those in Model 3 except for the respective stratification factor. Each stratified analysis showed the same result as that of the main analysis: as the TG/HDL-C ratio increased, so did the risk of developing NAFLD. In addition, there are several other specific results in our stratified analyses: female participants had a greater risk of NAFLD than male participants did in the same TG/HDL-C groups according to sex. In terms of age, the risk of NAFLD in the young population was slightly greater in the T2 and T3 groups than in the older group at the same level and was slightly lower in the T4 group than in the older group. In terms of WC, we found that the NAFLD risk for participants in the T3 and T4 groups who had a normal WC was greater than that in participants with the same TG/HDL-C ratio who had abdominal obesity. Moreover, in terms of both hypertension and diabetes stratification, the risk of NAFLD was greater in the T4 group without metabolic disease than in the T4 group with metabolic disease (Fig. 4).
2.5 Competing risk model
Considering that the death of the participants would affect the overall incidence of NAFLD, we further analysed the risk of mortality via a competing risk model. After we adjusted for sex, age, WC, TC, LDL-C, ALT, UA, Hs-CRP, diabetes status, hypertension status, smoking status, physical activity status, high school education status and above, and lipid-lowering drug use, the results showed that, compared with those of the T1 group, the HRs (95% Cis) of the T2, T3, and T4 groups were 1.13 (1.08 ~ 1.17), 1.32 (1.27 ~ 1.37), and 1.61 (1.55 ~ 1.67), respectively (P for trend < 0.001) (Table 3).
2.6 Sensitivity analyses
Sensitivity analysis 1 was conducted as follows. A Cox proportional hazards model was constructed after excluding 3678 participants whose follow-up time was less than 2 years, and the adjusted confounders were the same as those in Model 3. Compared with those of the T1 group, the HRs (95% CIs) of the T2, T3, and T4 groups were 1.10 (1.06–1.15), 1.30 (1.25–1.35), and 1.56 (1.50–1.63), respectively (P for trend < 0.001). Sensitivity analysis 2 was conducted as follows. After excluding 242 participants taking lipid-lowering drugs at baseline, a Cox proportional hazards model was constructed, and the adjusted confounders were the same as those in Model 3 except for the use of lipid-lowering drugs; moreover, the HRs (95% CIs) of the T2, T3, and T4 groups compared with the T1 group were 1.13 (1.09–1.17), 1.33 (1.28–1.38), and 1.60 (1.55–1.67), respectively (P for trend < 0.001) (Table 4).