Objective: The objective of the study was to investigate the association between thyroid stimulating hormone (TSH) and incident non-alcoholic fatty liver disease (NAFLD) in middle-aged and elderly population in a 3-year follow up study. Methods: 222 subclinically hypothyroid (TSH > 4.50 mIU/L) and 727 euthyroid people aged 40~75 years who were free of fatty liver disease at baseline in a community were included in the current study. The study population was screened initially in 2011 and re-evaluated in 2014. On both occasions assessed by structured interview via a questionnaire on general information, anthropometric measurements, thyroid hormones, biochemical and serological tests, liver ultrasound. NAFLD was diagnosed on ultrasound criteria, safe alcohol consumption. Results: Among 222 baseline subclinical hypothyroidism patients, 62 (27.93%) reverted to euthyroidism at a 3-year follow-up. The 3-year incidence of subclinical hypothyroidism was 16.64% (121/727). Higher TSH levels at baseline were associated with higher TC, triglyceride and AST levels (P <0.05). The incidence rate of NAFLD increased with increasing quartiles of baseline TSH level in female(P <0.05).Logistic regression analyses showed, in female, after adjusting for age in model 1, higher baseline TSH quartiles were associated with the diagnosis of NAFLD; after adjustment for age, baseline BMI, baseline triglyceride, baseline HbA1c in model 2, this association remained significant; when data were adjusted for the change in BMI and change in HbA1c in model 3, the association was no longer significant. Conclusions: TSH at baseline was an influence factor for 3-year risk of incident NAFLD in female in middle-aged and elderly population.