Current issues of ACP Journal Club are published in Annals of Internal Medicine


Decreased serum albumin levels predicted a small increase in mortality in healthy elderly persons

ACP J Club.1992 Nov-Dec;117:91. doi:10.7326/ACPJC-1992-117-3-091

Source Citation

Klonoff-Cohen H, Barrett-Connor EL, Edelstein SL. Albumin levels as a predictor of mortality in the healthy elderly. J Clin Epidemiol. 1992 Mar;45:207-12.



To determine whether low serum levels of albumin predict death independently of age and disease and whether higher serum globulin levels are associated with age and increased mortality.


A cohort was followed for 3 years.


University-affiliated research clinic attended by members of the Rancho Bernardo Heart and Chronic Disease study.


Upper middle-class persons (1045 men and 1297 women), aged 50 to 89 years, who were not acutely ill or institutionalized, were included in the study and followed for a mean of 1.4 years. Participants with medical disorders known to affect albumin and globulin levels were classified as "diseased" and analyzed separately. Vital status was known for each person. 99% of eligible clinic attenders participated.

Assessment of risk factors

Information about the participants was obtained at a clinic visit and included medical history; use of medications, alcohol, and tobacco; body mass index; and serum albumin (normal range 35 to 50 g/L) and globulin (normal range 24 to 35 g/L) measurements. 85% to 95% of a subsample of the diseases reported by participants (including metabolic, infectious, vascular, and neoplastic diseases) were confirmed by hospital records.

Main outcome measures

Relation of occurrence of death to lower serum albumin and to higher serum globulin. Estimated relative risks (relative odds [RO]) were calculated for 1 standard deviation (SD) changes in serum albumin.

Main results

Men and women in the older 10-year age groups had progressively lower mean serum albumin and higher serum globulin than younger persons, regardless of baseline health status (P for linear trends < 0.001). 21 of the healthy participants (2%) and 62 of the diseased participants (5%) died. In logistic regression adjusted for age and sex, a 1 SD reduction in albumin was associated with increased mortality overall (RO 1.24, P = 0.04); in healthy persons (RO 1.30, nonsignificant [NS]); and in the diseased group (RO 1.19, NS). Similar results were obtained after adjustment for lifestyle factors (RO overall 1.22, P = 0.05). Serum globulin increased with age (P < 0.001), but, after statistical adjustment, globulin was not positively associated with mortality.


Serum albumin levels decreased and serum globulin levels increased as middle-aged-to-elderly persons grew older. Reduced serum albumin was associated with increased 3-year mortality.

Source of funding: National Institute of Aging.

For article reprint: Dr. E.L. Barrett-Connor, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0607, USA.


Decreases in serum albumin concentration are characteristic of hepatic and renal diseases, malnutrition, and stresses, including surgery and trauma. The study by Klonoff-Cohen and colleagues took advantage of information that is readily available from routine blood tests. The presence or absence of disease was based on self-report. Validation methods were not described in detail, and disease severity was not accounted for. This cohort study resembles a cross-sectional study because serum albumin and disease state were only assessed at entry. Mortality was determined at a minimum of 1 year of follow-up, with average follow-up being 1.5 years, and only 83 participants died. On the basis of studies of other physiologic variables that decline with age (e.g., maximal exercise capacity), cross-sectional studies underestimate so-called "age-related" change (1).

The association of lower albumin levels with age and mortality is not surprising. The decrease may be caused by the reduction in fractional rate of albumin synthesis and a control level that is set at a lower set point. As the authors state, "it is unlikely that albumin levels can be used as a screening test in individuals, because the differences are small." The mortality effect is modest at best, and the statistical significance (P = 0.04) was based on 1-sided significance testing.

What is the take-home message here? Perhaps the clinical importance of the slightly lower, but still normal, albumin levels associated with chronologically older age in "healthy persons" is that it should serve to remind clinicians that albumin-bound drugs (e.g., phenytoin and warfarin) may give more intense "free drug effects" in older persons and that elderly patients are particularly susceptible to the acute effects of multiple drug therapy (2).

Eric B. Larson, MD, MPH
University of Washington Medical CenterSeattle, Washington, USA


1. Larson EB, Bruce RA. Arch Intern Med. 1987;147:353-6.

2. Vestal RE. In: Hazard WR, Andreas R, Bierman EL, Blass JP; eds. Principles of Geriatric Medicine and Gerontology. 2d ed. New York: McGraw-Hill; 1990:201-11.