Study design and participants
Details of the Tianjin Chronic Low-Grade Systemic Inflammation and Health (TCLSIH) Cohort Study have been described elsewhere(17). Briefly, participants were randomly recruited between January 2010 and December 2016 from the general population in Tianjin, China. The inclusion criteria for the TCLSIH cohort study were men and women who were 18 years and older living in Tianjin, China for at least 5 years. Subjects in the present study were sampled by a random process, using a random number generator. Nearly all occupations are covered in this study, and we also included retired individuals living in residential communities. Therefore, the sample population used here is representative of the general adult population in Tianjin, a typical city in north China. All participants received at least 2 health examinations in our study (including liver ultrasound examination, anthropometric measurements, and blood tests) and completed a structured questionnaire survey(18). The questionnaire consisted of the following contents: age, gender, smoking and drinking habits, history of diseases (cardiovascular disease, hypertension, hyperlipidemia, and diabetes), and family history of diseases (cardiovascular disease, hypertension, hyperlipidemia, and diabetes). The reliability and validity of the questionnaire have been assessed, with the Spearman’s rank correlation coefficient of 0.67 and 0.58, separately. Written informed consent was obtained from all participants. The ethical protocol of this study was approved by the Medical Ethics Committee of the Tianjin Medical University with the reference number of TMUhMEC 201430, in accordance with the 1975 Declaration of Helsinki (as revised in 1983).
From 2010 to 2016, a total of 90,536 participants received health examinations. We excluded 170 participants who had missing data on alanine aminotransferase, 4,013 participants had excessive alcohol intake (>140 g/week in males and >70 g/week in females), and 28,935 participants who had NAFLD at baseline. Moreover, we excluded 767 participants with other liver diseases (including autoimmune liver diseases, chronic hepatitis B or C, cirrhotic or operation on liver), and those with a history of cardiovascular disease (n = 5,475) or cancer (n = 1,039), and those aged < 25 years (n = 3,996). Furthermore, participants were also excluded if they were recruited in 2016 (n = 4,894) or were lost in following up (n = 5,253). Furthermore, based on the prevalence of NAFLD in Chinese population and on the principle of 10 outcome events per variable (19), the sample size was calculated. Finally, a total of 35,994 participants were available for analysis (follow-up rate: 87%; followed up for 2-5.5 y; mean duration of follow-up (standard deviation): 2.6 (1.6)).
Assessment of height
Standing height without shoes was measured to the nearest 0.1 cm using an automatic BMI measuring stadiometer with a precision of 0.1 cm and a range of 0.9–2.50 m (BSM370, Chungcheongnam-do, Korea). In order to investigate how height level is associated with NAFLD, we divided males and female participants into 5 categories (quintiles) according to height, in cm (range), as follows: (1) Level 1 (148.5-167.7), Level 2 (167.8-171.2), Level 3 (171.3-174.3), Level 4 (174.4-178.1), and Level 5 (178.2-204.1) in males; (2) Level 1 (138.0-156.2), Level 2 (156.3-159.5), Level 3 (159.6-162.1), Level 4 (162.2-165.4), and Level 5 (165.5-184.6) in females.
Diagnosis of NAFLD
Real-time ultrasonography performed by trained and certified technicians was used to diagnose NAFLD. Participants were considered to have NAFLD if (1) they had a self-reported alcohol intake of < 140 g/week and < 70 g/week for males and females, respectively; (2) and at least two of the following abnormal findings of abdominal ultrasound images: diffusely increased liver near field ultrasound echo; increased liver echotexture, compared to the kidneys; vascular blurring and the gradual attenuation of far field ultrasound echo(20). Inter-observer variations for NAFLD status (yes or no) were evaluated in a subsample of 200 participants from the TCLSIH study. The Kappa coefficient was 0.90, and the total agreement was 96.4%.
Assessment of other variables
Waist circumference (WC) was measured using a nonelastic plastic anthropometric tape at the level of umbilicus with subjects standing and breathing normally. Waist-to-height ratio was calculated by dividing WC (cm) by the subjects’ height (cm). Participants rest for at least 5 minutes in a seated position prior to blood pressure measurements. Blood pressure was measured twice from participants’ upper right arms using the TM-2655 device (A&D Company Ltd, Tokyo, Japan), and the blood pressure value was recorded in average. If the first two results are quite different, additional measurements was carried out until stabilization. The mean of the two closest readings (including the last reading) was calculated to determine the reported BP for each participant. Hypertension was finally assessed and diagnosed by physicians according to the criteria of the JNC 8: hypertension was defined as SBP ≥140 mmHg and/or DBP ≥90 mmHg or having history of hypertension or using antihypertensive drugs(21). Fasting blood samples for the analysis of biochemical values were collected in siliconized vacuum plastic tubes. Fasting blood glucose was measured by the glucose oxidase method, triglycerides were measured by enzymatic methods, low-density lipoprotein cholesterol was measured by the polyvinyl sulfuric acid precipitation method, high-density lipoprotein cholesterol was measured by the chemical precipitation method, and alanine aminotransferase was measured by International Federation of Clinical Chemists (IFCC) method using reagents from Roche Diagnostics on an automatic biochemistry analyzer (Roche Cobas 8000 modular analyzer, Mannheim, Germany). Diabetes was defined as FBG levels ≥7.0 mmol/L or having history of diabetes. Hyperlipidemia was defined as TC ≥5.17 mmol/L or TG ≥1.7 mmol/L or LDL ≥3.37 mmol/L or history of hyperlipidemia. We defined metabolic syndrome (MetS) according to the American Heart Association scientific statements of 2009(22).
Body weight was measured by an automatic body mass index (BMI) measuring stadiometer (BSM370, Chungcheongnam-do, Korea), accurate to 0.1 kg, with participants wearing only light clothing and no shoes. BMI was calculated as weight (kg) divided by squared height (m2). Based on the World Health Organization recommendations for Chinese people, underweight was defined as BMI < 18.5kg/m2, normal weight was defined as 18.5 kg/m2 ≤ BMI < 23.0 kg/m2, overweight was defined as 23 kg/m2 ≤ BMI < 27.5 kg/m2, and obesity was defined as BMI ≥ 27.5kg/m2(23). Information on family history of cardiovascular disease, hypertension, hyperlipidemia, and diabetes and lifestyle and health-related habits was assessed at baseline using a structured questionnaire. Smoking status was grouped in three: smoker, ex-smoker or nonsmoker and drinking status was classified as everyday, sometime, ex-drinker or nondrinker by self-reporting.
Baseline characteristics of participants were compared using analysis of variance for continuous variables and logistic regression analysis for categorical variables. Continuous variables were shown as geometric mean (95% confidence interval (CI)), and categorical variables were presented as percentage. Cumulative event rates for incident NAFLD were estimated by Kaplan-Meier survival curves, and equalities were compared with the log-rank test.
We tested the interaction between height and the confounding factors, including age, sex, waist circumference, BMI, smoking status, alcohol drinking status, Mets, and family history of disease (cardiovascular disease, hypertension, hyperlipidemia, and diabetes), separately. The interaction between sex and height was statistically significant (P < 0.0001), while the P values for interaction between height and other confounding factors were > 0.1. Therefore, we analyzed the association between height and NAFLD stratified by sex. We fitted four Cox proportional hazards regression models to evaluate the association between baseline height and incident NAFLD. The initial model was unadjusted model (crude model). Model 2 was adjusted for age and WC. In model 3, we additionally adjusted for smoking status, alcohol drinking status, Mets, family history of cardiovascular disease, hypertension, hyperlipidemia, and diabetes. In model 4, we further adjusted for baseline BMI. In model 5, we further adjusted for baseline waist-to-height ratio. Moreover, we adjusted the history of disease (hypertension, diabetes mellitus, and hyperlipidemia) or the subject's blood pressure, fasting blood glucose, triglycerides, low-density lipoprotein cholesterol, and high-density lipoprotein cholesterol to replace the metabolic syndrome in the final multiple-adjusted model, separately. All P values for linear trends were calculated using the median value for each quintile. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Inc.). Two-tailed P <0.05 was considered as statistically significant.