Depression Is Associated With Non-alcoholic Fatty Liver Disease Among Adults in the United States

Donghee Kim; Eric R. Yoo; Andrew A. Li; Sean P. Tighe; George Cholankeril; Stephen A. Harrison; Aijaz Ahmed


Aliment Pharmacol Ther. 2019;50(5):590-598. 

In This Article


We analysed a total of 10 484 individuals (mean age was 47.0 years and 48.8% men). Table 1 describes the clinical and demographic characteristics of the population according to the presence of depression. There were noticeable differences in the clinical features in the presence of depression. Compared to individuals without depression, those with depression were more likely to be older, women, diabetic, hypertensive and smoker. Individuals with depression also had a higher BMI and waist circumference, higher levels of total cholesterol, triglyceride and fasting glucose and gamma-glutamyltransferase compared to those without depression.

As shown in Table 2, when we defined NAFLD using USFLI as the NAFLD prediction model, the weighted prevalence of depression_med was 8.5% in individuals without NAFLD and 13.7% in individuals with NAFLD (P < 0.001). The weighted prevalence of depression_PHQ9 was 6.2% in individuals without NAFLD vs 9.8% in those with NAFLD (P < 0.001). Individuals with NAFLD also had a higher prevalence of functional impairment due to depression than those without NAFLD. When we defined NAFLD using HSI and FLI, similar and significant differences were observed. Weighted proportions of depression and functional impairment due to depression were not different between individuals with NAFLD-related advanced fibrosis and those without.

Results of logistic regression analyses are tabulated in Table 3. Depression_med was associated with the higher prevalence of NAFLD using USFLI as the tool for NAFLD prediction in the age, sex-adjusted models (odds ratio [OR] 1.87; 95% CI 1.54-2.28). After adjusting for age, sex, Race/ethnicity, educational level, marital status, economic status, smoking status, hypertension, physical activity, and total cholesterol, depression was associated with 78% higher odds of NAFLD (OR 1.78; 95% CI 1.47-2.15). The addition of diabetes, waist circumference, and BMI to the model reduced the ORs for depression, but the significance persisted (OR 1.48; 95% CI 1.17-1.87). As shown in Table S2, further adjustment for high-density lipoprotein-cholesterol and fasting triglycerides resulted in attenuation, but the association remained significant (OR 1.39; 95% CI 1.06-1.82). When we considered insulin resistance by HOMA-IR, this association between depression_med and NAFLD attenuated and remained statistically significant (OR 1.35; 95% CI 1.00-1.83, P = 0.049). These results suggest that the association between depression and NAFLD might be mediated, in part, by diabetes, general/abdominal obesity and insulin resistance. When we performed sensitivity analysis using a different definition of depression, depression_PHQ9 was associated with NAFLD (the ORs for NAFLD comparing the individuals with depression to those without depression were 1.44 [95% CI 1.09-1.91], P = 0.012) in the multivariate model 2. When we performed sensitivity analyses using HSI and FLI (Table 3 and Table S3), similar and significant associations between depression and NAFLD were observed in the univariate and multivariate models regardless of differences in the definition of depression.

As shown in Table 4, in the age, sex-adjusted logistic regression analysis, depression_med was not associated with advanced fibrosis. This insignificant association persisted in the multivariate models. Sensitivity analyses were conducted utilising the different definitions of depression (depression_PHQ9) and functional impairment and different definitions of NAFLD (Table 4 and Table S3). Overall, the results were identical to those obtained using the original definition.