Background: The present study aims at defining a body mass index (BMI) threshold for risk of being underweight in elderly persons on the basis of the BMI distribution in a large Italian population-based sample and on its ability to predict short-term mortality.
Methods: At baseline (1992), BMI was obtained for 3110 (1663 males and 1447 females) persons aged 65-84 participating in the Italian Longitudinal Study on Aging (ILSA). BMI and risk factors (age, sex, education, smoking status, disability, and disease status) have been considered for their potential association with 4-year all-cause mortality. Information on vital status at 1995 was obtained for 2551 participants.
Results: The fifth centile of BMI was well approximated by a value of 20 for both sexes. Also in both sexes, at a BMI value of 24 the a posteriori probability of death started to increase, doubling at a value of 22 for men and 20 for women. Crude mortality was 14.6% for men and 9.8% for women. The hazard ratios and confidence intervals (CIs) comparing mortality for each BMI two-unit class to the 26-28 class, after adjusting for confounding variables, showed significantly higher rates only for BMI values below 20 (2.9; 95% CI, 1.2-7.0), although a consistent increase in hazard ratio (1.6; 95% CI, 0.9-3.0) already appeared for the 20-22 BMI group.
Conclusions: Our study confirms that low BMI is an independent predictive factor of short-term mortality in elderly persons. A BMI value of 20 kg/m2 seems to be a reliable threshold for defining underweight elderly persons at high risk. Nevertheless, more careful clinical and nutritional management should also be applied to elderly persons with higher BMI values.