The current analysis was designed to examine further the association between total serum carotenoids and a series of preselected variables (dietary intakes, anthropometric parameters, serum cholesterol, vitamin A, transport proteins, and others) using 85 elderly patients. Simple correlation between serum carotene and carotene intake, as estimated by a three-day weighed method, was 0.19 (P = 0.08), which was substantially increased (r = 0.29, P = 0.01) after adjustments for a selected set of variables. Standardized coefficients of multivariate regression indicated that the most important predictors of serum carotene were serum cholesterol (beta = 0.38), total serum proteins (beta = -0.35), and sex (beta = 0.34), followed by carotene intake (beta = 0.28) and midarm circumference (beta = 0.20). These variables accounted for 46% of the variance. Results suggest that serum carotene may be related to protein-energy status in hospitalized elderly, but further investigation should be directed to serum carotene in undernourished elderly. Nevertheless, total serum proteins could be an important factor in any attempt to correlate carotene intake and blood concentration.