Determining lifestyle correlates of body mass index using multilevel analyses: The Tromso study, 1979-2001

Increases in overweight and obesity have been observed globally in both developed and developing countries. The authors assessed the relation between lifestyle factors and body mass index (BMI) (weight (kg)/height (m)(2)) in a population-based longitudinal study, using BMI and its subsequent change as responses in a multilevel model. The authors included 11,115 men and women aged 20-61 years at baseline who were living in the municipality of Tromso, Norway, and who participated in three or four consecutive health surveys between 1979-1980 and 2001. Baseline age, physical activity at work, coffee consumption, and desired BMI (i.e., the BMI that the subjects reported they would like to have) were positively associated with baseline BMI, whereas height, alcohol consumption, leisure-time physical activity, and level of education were inversely associated. Most relations were found to be stronger in women than in men. Clinically relevant effect sizes were observed for most of the significant associations, especially in women. For instance, on an ordinal scale, a one-category increase in educational level would decrease the mean baseline BMI among women by 0.30 kg/m(2). Significant associations between several lifestyle factors and subsequent BMI change revealed that observed baseline associations were strengthened over time, especially in women.