Primary weight maintenance: an observational study exploring candidate variables for intervention

BACKGROUND: Previous studies have focused on weight maintenance following weight loss, i.e. secondary weight maintenance (SWM). The long-term results of SWM have been rather modest and it has been suggested that preventing initial weight gain, i.e. primary weight maintenance (PWM), may be more successful. Therefore, developing a prevention strategy focused on PWM, enabling normal weight or overweight individuals to maintain their weight, would be of great interest. The aim of this study was to identify attitudes, strategies, and behaviors that are predictive of PWM in different age, sex and BMI groups in Northern Sweden. METHODS: A questionnaire was mailed to 3497 individuals in a Swedish population that had two measured weights taken ten years apart, as participants in the Västerbotten Intervention Programme. Subjects were between 41–63 years of age at the time of the survey, had a baseline BMI of 20–30, and a ten year percent change in BMI greater than -3%. The respondents were divided into twelve subgroups based on baseline age (30, 40 and 50), sex and BMI (normal weight and overweight). Analysis of variance (ANOVA), correlation, and linear regression were performed to identify independent predictors of PWM. RESULTS: Of the 166 predictors tested, 152 (91.6%) were predictive of PWM in at least one subgroup. However, only 7 of these 152 variables (4.6%) were significant in 6 subgroups or more. The number of significant predictors of PWM was higher for male (35.8) than female (27.5) subgroups (p=0.044). There was a tendency (non significant) for normal weight subgroups to have a higher number of predictors (35.3) than overweight subgroups (28.0). Adjusted R-squared values ranged from 0.1 to 0.420. CONCLUSIONS: The large number of PWM predictors identified, and accompanying high R-squared values, provide a promising first step towards the development of PWM interventions. The large disparity in the pattern of significant variables between subgroups suggests that these interventions should be tailored to the person’s demographic (age, sex and BMI). The next steps should be directed towards evaluation of these predictors for causal potential.