Longitudinal changes in food patterns predict changes in weight and body mass index and the effects are greatest in obese women

The prevalence of obesity is increasing in most parts of the world. The objective of this study was to examine whether changes in food patterns were associated with changes in BMI among women over 9 y. Data were from 33,840 women participating in the Swedish Mammography Cohort in 1987 and 1997. Diet was assessed with a FFQ at both time points and 4 food patterns were derived using confirmatory factor analysis (Healthy, Western/Swedish, Alcohol, and Sweets). Our exposure variables were defined as change in factor score for each food pattern from 1987 to 1997. Our outcome variable was defined as change in BMI during the same period, and we also examined change in weight. Normal weight and overweight women had positive changes in weight and BMI during follow-up, whereas obese women had negative changes in BMI and weight; we observed a significant interaction between change in food patterns and baseline BMI. Obese subjects who increased their factor score for the Healthy pattern had larger decreases in BMI (beta = -0.18 kg/m(2) for a 1 unit increase in SD score, Cl: -0.26 to -0.10 ; P < 0.0001), whereas normal weight and overweight women who increased their Healthy pattern score had smaller increases in BMI (-0.05 kg/m(2) and -0.11 kg/m(2), respectively; P < 0.05 for both). Obese women who increased their Healthy pattern score > 3 SD had almost a 4 kg decrease in weight and a 1 unit decrease in BMI at follow-up. In conclusion, changes in eating patterns were significantly related to changes in BMI over 9 y and the effect was modified by baseline BMI. Pattern analysis is helpful in generating hypotheses regarding the role of diet in obesity, and additional research is needed to understand which elements of our patterns are protective or risk factors for weight gain.