Nonalcoholic Fatty Liver Disease Is Associated With Decreased Lung Function

Chang-Hoon Lee; Seung Ho Choi; Goh Eun Chung; Boram Park; Min-Sun Kwak


Liver International. 2018;38(11):2091-2100. 

In This Article

Patients and Methods

Study Population

For this longitudinal study, 20 112 subjects who underwent routine health screening examinations including hepatic ultrasonography and pulmonary function tests (PFTs) with at least 3 years' follow-up from October 2003 to December 2016 were included. Flowchart of study population is included in Figure 1 and Appendix S1. The subjects in the NAFLD group were randomly selected from the database file with propensity score (PS) matching to the control group (1:1 ratio).

Figure 1.

Flow chart of study enrolment

This study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki and was approved by the Institutional Review Board of Seoul National University Hospital (H–1503–002–650). The requirement for informed consent from subjects was waived because the researchers accessed only de-identified database entries for analytical purposes.

Ultrasonographic Examinations and Definition of Nonalcoholic Fatty Liver Disease

Hepatic ultrasonography was performed by experienced radiologists who did not know the clinical or laboratory details of the participants at the time of the procedure. Fatty liver was diagnosed by ultrasonography (Aloka a–10 (Aloka, Tokyo, Japan), Siemens Acuson S2000™ ultrasound system (Siemens Medical Solutions, Erlangen, Germany), iU22 ultrasound system (Philips Medical Systems, Bothell, WA, USA), or LOGIQ E9 ultrasound system (GE Healthcare, Wauwatosa, WI, USA)) on the basis of liver brightness, evident contrast between the hepatic and renal parenchyma, vascular blurring, and deep attenuation.[19] NAFLD was defined as the presence of fatty liver, as determined by ultrasonography, with none of the following other possible causes of chronic liver disease: (i) significant alcohol consumption (defined as >30 g/d for men and >20 g/d for women), (ii) positivity for hepatitis B surface antigen or antibodies against the hepatitis C virus, and (iii) other known etiologies of chronic liver disease.

Noninvasive Measurement of Hepatic Fibrosis

Hepatic fibrosis was defined using a previously validated noninvasive prediction model including NAFLD fibrosis score (NFS)[20] and the Fibrosis–4 (FIB–4) index.[21] Calculations of NFS and FIB–4 score were performed only in subjects with NAFLD as determined by ultrasonography. The formulae for both scores are presented in Table S1. According to the NFS, subjects with NAFLD were divided into two groups, including those with low (NFS < −1.455) and intermediate to high probability (NFS ≥ −1.455) for advanced fibrosis.[20] Similarly, subjects with NAFLD were divided into two groups by FIB–4 as those with low (FIB–4 < 1.30) and intermediate to high probability for advanced fibrosis (FIB–4 ≥ 1.30).[21]


Spirometry was performed by trained technicians using a computerized spirometry system (Vmax Encore PFT system and MasterScreen Pneumo (SensorMedics, Yorba Linda, CA, USA)) according to the 2005 American Thoracic Society (ATS)/European Respiratory Society (ERS) recommendations.[22] Spirometry results were expressed as percentages of the predicted values, which were calculated by computer programs using Morris's predictive equations.[23] The results were reviewed by trained physicians. Post-bronchodilator testing was not performed. All measures were based on prebronchodilator values. Chronic obstructive pulmonary disease (COPD) was defined as follows: forced expiratory volume in 1–second (FEV1)/forced vital capacity (FVC) <0.7.

Clinical and Laboratory Assessments

Each subject completed a past medical history questionnaire including smoking; history of alcohol consumption; underlying diseases such as hypertension, diabetes, dyslipidemia, chronic liver disease, lung disease, and cancer; history of surgery; and medications taken. The smoking questionnaire gathered detailed information including current smoking status, amount smoked (packs/d), and duration of current and past smoking.

Subjects underwent anthropometric assessment and laboratory and radiological tests on the same day. The details of anthropometric assessment and laboratory tests are presented in Appendix S1.

Statistical Analysis

Groups were compared with adequate statistical methods: Continuous variables were expressed as means ± standard deviation, and categorical variables were expressed as numbers and percentages. Between-group comparisons were performed using Student's t test for continuous variables and Chi-square tests for categorical variables. Main outcome variables were FEV1 and FVC, and all analysis was separated by sex.

For cross-sectional analysis, we performed multiple regression and conducted leave-one-out cross validation to predict the FEV1 and FVC values and to compare predicted values of FEV1 and FVC at baseline between the NAFLD and non-NAFLD groups. In multivariate model 1, known major factors affecting lung function including age, anthropometric measures (body mass index (BMI)) and smoking that have been applied in large studies[24–26] were adjusted. In multivariate model 2 and 3, variables related to metabolic syndrome, which are associated with both lung function and NAFLD were added. To avoid collinearity, several variables that highly related with other variables were not included.

In longitudinal analysis, the effects of NAFLD on lung function decline rates were determined using linear mixed models with both a random slope and a random intercept. The models included terms for NAFLD, observation time in years, NAFLD-by-time interaction, and covariates.[24] Firstly, we analysed all included participants without (ie, unadjusted model) and with adjustment by covariates, including age, BMI and smoking (adjusted model 1). We also showed PS-adjusted model for all participants (adjusted model 2). PS was generated by logistic regression analysis with covariates including FVC, FEV1, systolic blood pressure, diastolic blood pressure, height, weight, BMI, waist circumference, fasting glucose, HbA1c, uric acid, cholesterol, total protein, albumin, bilirubin, alkaline phosphatase, alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma glutamyl transferase (GGT), blood urea nitrogen, creatinine, triglycerides, high-density lipoprotein (HDL), low-density lipoprotein (LDL), white blood cells, haemoglobin, platelets, total smoking (packs/y) and current smoking status. Secondly, we used PS-matched cohort. Patients with NAFLD were matched (1:1) to those without NAFLD on the basis of PS. The balancing in variables between groups was evaluated by both P–value and standardized mean difference (SMD). Although after PS matching, most variables were balanced between NAFLD and non-NAFLD group, several variables (triglyceride and BMI in male; fasting glucose and diabetes in female) were unbalanced (P < .05, Table S2). We analysed the PS-matched cohort without (ie, unadjusted model) and with those unbalanced covariates (adjusted model). All statistical analysis was performed by Stata 14.2 (StataCorp, College Station, TX, USA), and P <.05 were considered to indicate statistical significance.