Objective: To evaluate the association between weight variability and risk of mortality in women.
Design: Prospective cohort study, 1986-1991.
Subjects: Participants were a population-based sample of 33,760 Iowa women aged 55-69 y, free of cancer and heart disease, who completed a mailed questionnaire including self-reported weight at ages 18, 30, 40, 50 y, and currently. Weight variability was defined by (1) a measure of deviation from the linear regression of each woman's five weights on age (the root mean squared error, RMSE) and (2) categories of weight gain or loss.
Measurements: All-cause (n = 1068) and cause-specific mortality.
Results: After adjustment for age and the regression slope of weight on age, women who displayed higher weight variability (RMSE) over their adult years had an increased subsequent 6-y mortality. The relative risks of death for increasing quartiles of RMSE were 1.00 (referent), 1.17, 1.45 and 1.82 (ptrend < 0.001). Positive linear trends were seen across quartiles of RMSE for cardiovascular disease, cancer, and for non-cancer/non-cardiovascular deaths. These associations were attenuated increasingly with (1) statistical adjustment for body mass index, waist-to-hip ratio, smoking, education level, physical activity, alcohol intake, marital status and hormone replacement therapy; (2) further adjustment for diabetes and hypertension; and (3) exclusion of women in poor or fair health in 1986. Even so, with all adjustments, relative risks of coronary heart disease mortality for increasing quartiles of RMSE were 1.00 (referent), 1.22, 1.63 and 1.67 (ptrend = 0.049). Using the categorical approach, age-adjusted risks of death were highest in women who had a large weight loss (> 10%) or a large cycle of weight change (> or = 10% loss-gain or gain-loss), compared with women who had a stable weight (within 5%). Adjustment for covariates attenuated these relative risks.
Conclusions: Women who displayed greater weight variability in adulthood had an increased risk of dying in later life, especially from coronary heart disease. However, to a considerable degree this association seems to be due to other unhealthy characteristics and pre-existing disease among those displaying increased weight variability.