Weight variability in a population-based sample of older women: reliability and intercorrelation of measures.

Six measures of weight variability were examined in a cohort of 29,015 postmenopausal women. Recalled weight at ages 18, 30, 40 and 50 years, current weight at baseline and at each of three biennial follow ups (approximate ages 62, 64, 66, 68 years), and recalled episodes of intentional and unintentional weight loss were used to construct (1) the coefficient of variation (CV) in body weight, (2) weight change categories (cycling, weight gain, weight loss and stable weight), (3) the root mean square error of variation (RMSE) around the slope of weight versus age, (4) the number of intentional weight loss episodes of 5 or more pounds, (5) the number of unintentional weight loss episodes of 20 or more pounds and (6) a categorical measure of intentional and unintentional weight loss episodes of > = 20 lb. The nine-month test-retest reliability correlations for the measures of lifetime history of intentional and unintentional weight loss were 0.80 and 0.62, respectively. Correlations between the different weight variability measures were positive but weak, suggesting that they reflect different aspects of weight variability. The RMSE discriminated categorically defined cyclers from weight gainers, but the CV did not. The weight change categories were more sensitive to age-related weight changes than the CV or RMSE. Studies examining the relationship between weight variability and health outcomes need to include measures that distinguish intentionality, short-term versus long term variability, and the magnitude, direction, and frequency of weight change.