| Background:Various longitudinal cohort studies have demonstrated that non-alcoholic fatty liver disease (NAFLD) was associated with an increased risk of metabolic syndrome (MetS), while other studies have concluded that MetS was a predictor of NAFLD. It remained unclear whether NAFLD was a cause or effect of MetS. We aimed to test a hypothesis that there is reciprocal causality between NAFLD and MetS.Methods:Based on the routine health check-up system, we set up bi-directional longitudinal cohorts (sub-cohort A and B) in urban Han Chinese population from2005to2011in Health Management of Shandong Provincial Qianfoshan Hospital and Shandong Provincial Hospital. Participants who had at least2times health check-up were recruited in our study, with the first check-up data as baseline and the last as end-point. Sub-cohort A was selected from those who were free of neither MetS nor its components (obesity, dyslipidemia, hyperglycemia, and hypertension) at baseline and to follow up MetS. Sub-cohort B was selected from those who were free of NAFLD at baseline and to follow up NAFLD. A total of8,426participants were included in sub-cohort A and16,110in sub-cohort B. Generalized estimating equation (GEE) combined with causal graphical model was used for the reciprocal causal inference. Results:a) The incidence density (ID) of MetS was2.47per100person-years in sub-cohort A, while17.39per100person-years of NAFLD in sub-cohort B. b) GEE analyses showed that NAFLD was a potential cause for MetS (RR=5.23,[CI:3.50-7.81]; P<0.0001) after adjusting baseage, gender, time, GGT, TP, A/G, BUN, S-CREA, CHOL, Hb, MCV, WBC, diet, smoking and exercise, and NAFLD was also a predictor for individual components of MetS-obesity, diabetes, hypertension and dyslipidemia (RR=3.69,2.45,2.60,3.71). Vice verse, MetS was a cause for NAFLD (RR=2.55,[CI:2.23-2.92]; P<0.0001) and so were individual components (RR=3.29,1.75,1.92,2.36). c) Causal inference from Bayesian network illustrated that the total effect of NAFLD on MetS was2.49%, while19.92%of MetS on NAFLD. There were36causal pathways existing from NAFLD to MetS within the causal network. Generally, the most important causal pathways were that NAFLD led to elevated gamma-glutamyl transpeptidase (GGT), then to dysliplidemia, followed by hypertension, and finally resulted in MetS; the second important causal pathways were that persistent NAFLD led to obesity, and then to diabetes, or dysliplidemia, hypertension, and finally resulted in MetS; while, if the elevated CHOL appeared in the pathway, the probability resulting in MetS would become low. There were98causal pathways from MetS/its components to NAFLD within the causal network. The dominant causal pathway was that dysliplidemia led to other components (diabetes, hypertension, etc.) and finally resulted in NAFLD. In addition, SD, HCT, or Hb might play an important role in the above causal pathways. The second important causal pathway was beginning with hypertension, followed by beginning with diabetes, and obesity.Conclusion:It should be emphasized that the incidence density of NAFLD and MetS were very high in urban Han Chinese population. There was reciprocal causality between NAFLD and MetS, which could be mainly explained by insulin resistance. |
