Survival in elderly men in relation to midlife and current BMI

Background: the relationship between BMI and mortality is U-shaped in the elderly but may be modified by midlife BMI and change in weight. Objective: to elucidate mortality prediction by BMI trajectory in older men. Subjects: the Oslo cohort of men born in 1923-32 were screened for BMI and cardiovascular risk in 1972-73. Survivors were rescreened at ages of 68-77 years, and all-cause mortality was followed from 2000 to 2011. Methods: we calculated Cox regression proportional hazards for 11-year mortality rates in relation to BMI change among 5,240 men with no reported disease in 1972-73 and complete data. Models were adjusted for demographics, medications and disease. Men with BMI < 25 kg/m(2) in midlife (1972-73) and in 2000 were the reference group. Results: men whose BMI changed from < 25 kg/m(2) in midlife to 25-29.9 kg/m(2) in 2000 carried the lowest mortality risk (hazard ratio [HR] 0.84; 95% confidence interval [CI] 0.75-0.94). Men with BMI >= 30 kg/m(2) in midlife were at highest risk (HR 1.79; 95% CI 1.12-2.88 if reclassified to 25-29.9 kg/m(2) in 2000 and HR 1.39; 95% CI 1.05-1.85 if BMI remained >= 30 kg/m(2) in 2000). Men with BMI 25-29.9 kg/m(2) in midlife that reclassified to < 25 kg/m(2) in 2000 had increased risk. Findings were similar when percentage change in BMI was the outcome. Conclusion: survival in older men with normal weight at midlife was associated with BMI gain after midlife while midlife obesity increased risk regardless of subsequent change.