Table 2

Association between baseline body mass index and the development of fatty liver

BMI (kg/m2) categoryPerson-yearsIncident caseAge-adjusted HR (95% CI)Multivariate HR* (95% CI)HR (95% CI)‡ in the model using time-dependent variables
Ultrasonographically detected fatty liver (USFL)
    <18.5579.740.26 (0.10 to 0.71)0.32 (0.12 to 0.87)0.25 (0.06 to 1.02)
    18.5–22.98746.12411.00 (reference)1.00 (reference)1.00 (reference)
    23.0–24.94807.72221.65 (1.38 to 1.99)1.27 (1.05 to 1.53)1.71 (1.39 to 2.10)
    ⩾25.02696.31552.08 (1.70 to 2.54)1.37 (1.10 to 1.71)1.94 (1.54 to 2.44)
    p For trend<0.001<0.001<0.001
USFL + elevated ALT
    <22.9§9551.0641.00 (reference)1.00 (reference)1.00 (reference)
    23.0–24.95048.3531.53 (1.06 to 2.20)1.03 (0.71 to 1.51)2.12 (1.38 to 3.25)
    ⩾25.02860.2422.15 (1.45 to 3.17)1.15 (0.75 to 1.75)2.55 (1.60 to 4.04)
    p For trend<0.0010.542<0.001
  • *Estimated from Cox proportional hazard models adjusted for age, smoking status, HDL-C, triglycerides, ALT, uric acid and HOMA-IR at baseline.

  • ‡Estimated from extended Cox proportional hazard models with BMI as a time-dependent variable adjusted for other covariates (age, smoking status, HDL-C, triglycerides, ALT, uric acid, and HOMA-IR) over time as time-dependent variables.

  • §The BMI categories of <18.5 and 18.5–22.9 were combined due to the small number of incident cases.

  • ALT, alanine aminotransferase; BMI, body mass index; CI, confidence intervals; HDL-C, high-density lipoprotein-cholesterol; HOMA-IR, homeostasis model assessment of insulin resistance; HR, hazard ratios.