OBJECTIVETo assess the predictive value of hemoglobin (HGB) levels for non-alcoholic fatty liver disease (NAFLD) by performing a prospective cohort study of NAFLD incidence in a healthy population.

METHODSA total of 2840 individuals in the Xinjiang province were enrolled in the study from 2008 to 2011, based on liver ultrasound showing no evidence of fatty liver disease and the discovery of no major risk factors upon interview. All participants completed an epidemiological questionnaire survey, a physical examination, an abdominal ultrasonography, and gave blood for biochemistry testing. The hazard ratios of NAFLD were compared when the participants were grouped according to HGB level (g/L in quintiles): Q1, less than or equal to 145 for males and less than or equal to 123 for females; Q2, > 145 to less than or equal to 151 for males and > 123 to less than or equal to 129 for females; Q3, > 151 to less than or equal to 155 for males and >129 to less than or equal to 134 for females; Q4, > 155 to less than or equal to 161 for males and > 134 to less than or equal to 139 for females; Q5, > 161 for males and > 139 for females. Between-group comparison of measurement data was carried out by t-test and of percentage or count data by chi-square test. Between group comparison of the HGB level was carried out by one-way ANOVA. The prospective association between HGB levels and NAFLD was assessed by conditioned logistic regression analysis.

RESULTSThe values of body mass index, blood pressure, and triglyceride level were significantly higher in the participants with elevated serum uric acid quartiles. Within the 3-year study period, NAFLD was newly diagnosed in 19.6% of the male participants and 10.1% of the female participants; the difference between males and females reached the threshold of statistical significance (X2 = 51.043, P less than 0.01). The incidence of NAFLD in the quintile groups was 6.10% in Q1, 10.50% in Q2, 13.13% in Q3, 16.95% in Q4, and 22.03% in Q5 (X2 = 70.495, P less than 0.01), and the increasing trend with elevated HGB quartiles was significant (P less than 0.01). In multivariate logistic regression analysis, with adjustment for sex, age, race, metabolic syndrome and its components, the hazard ratios for incidence of fatty liver comparing Q2 to Q5 of HGB concentration to Q1 were 1.125, 1.325, 1.516 and 1.982.

CONCLUSIONElevated HGB concentration is predictive of NAFLD in otherwise healthy subjects and may be used for screening during a routine health check-up.