Serum Albumin and Risk of Myocardial Infarction and All-Cause Mortality in the Framingham Offspring Study
Background— Coronary disease remains the leading cause of death in the United States. The association between serum albumin and cardiovascular disease remains controversial. We used data collected prospectively from participants of the Framingham Offspring Study to assess whether a lower concentration of serum albumin was associated with an increased risk of myocardial infarction (MI) and all-cause mortality.
Methods and Results— During 21.9 years of mean follow-up, 280 cases of MI occurred. From the highest to the lowest tertile of serum albumin, crude incidence rates of MI were 26.7, 46.7, and 67.8 cases per 10 000 person-years, respectively, for men and 5.9, 15.0, and 16.8 cases per 10 000 person-years, respectively, for women. In a Mantel-Haenszel method adjusting for age, total cholesterol, and hypertension, lower serum albumin was associated with an increased risk of MI in both sexes. From the highest to the lowest tertile of albumin, the adjusted hazard ratios (95% CI) of MI were 1.0 (reference), 1.25 (0.84 to 1.84), and 1.49 (1.01 to 2.21), respectively, for men and 1.0, 1.79 (0.88 to 3.65), and 2.12 (1.06 to 4.27), respectively, for women. The albumin-MI association was stronger among hypertensive subjects in both sexes. In addition, low albumin was associated with an increased rate of all-cause mortality in women.
Conclusions— Lower serum albumin concentrations appear to be associated with an increased risk of coronary disease in both sexes and with all-cause mortality in women and could help along with traditional risk factors in identifying people at risk of MI.
Received August 6, 2002; revision received September 24, 2002; accepted September 26, 2002.
Coronary heart disease (CHD) is the leading cause of death in industrialized nations.1,2⇓ Traditional risk factors for CHD such as hypertension, smoking, diabetes, and hyperlipidemia may not be able to predict cardiovascular events accurately in older men and women.3–5⇓⇓ Early detection and prevention of CHD, especially among the elderly, remains a major public health issue.
Synthesized in the liver, serum albumin concentration falls (≈20%) during an inflammatory process.6 Albumin is inversely correlated with age, smoking, obesity, and blood pressure.6,7⇓ It is not clear whether low serum albumin level is a nonspecific, prognostic variable, a marker for subclinical disease, or whether it is part of the causal mechanism leading to CHD.6 Prior epidemiologic studies have reported an inverse association between serum albumin and CHD6,8⇓ and stroke.9 In a meta-analysis, low albumin was associated with a 50% increased risk of CHD.6 In contrast, several observational studies did not find an association between serum albumin and CHD.10–14⇓⇓⇓⇓
Previous studies have suggested that a lower concentration of serum albumin is associated with a 2-fold increased risk of total cardiovascular mortality,6,10,15–17⇓⇓⇓⇓ all-cause mortality,10,15,16,18,19⇓⇓⇓⇓ and cancer mortality.16,20⇓ Contrary to these studies, Law et al11 did not find evidence for an association between lower levels of serum albumin and mortality from cardiovascular disease or cancer.
The aim of this project was to evaluate whether lower serum albumin was associated with an increased risk of myocardial infarction (MI) and all-cause mortality in the Framingham Offspring Study.
The Framingham Offspring Study is a prospective study started in 1971 consisting of 5124 offspring (and their spouses) of the original Framingham cohort. Subjects in this cohort have been reexamined 8 years after the baseline examination and every 4 years thereafter. Detailed descriptions of the Framingham Offspring Study have been published previously.21 Informed consent is obtained repeatedly from study participants, and the study protocol was approved by the Institutional Review Board of Boston Medical Center.
Serum Albumin Measurement
During the first and second clinical examinations, blood samples were collected and serum albumin was measured by a thin film adaptation of a bromcresol green colorimetric procedure using the Vitros analyzer.22 Only the baseline albumin levels were used for the initial analyses, although both measurements were used in the time-dependent Cox regression models.
Cases of MI or death that occurred between the first (1971 to 1975) and fifth (1991 to 1995) examinations were included in these analyses. A detailed description of cardiovascular events has been published previously.23 Briefly, at each examination, an extensive cardiovascular disease history was obtained and physical examination, 12-lead ECG, and various blood chemistries were completed. Occurrence of MI was determined on the basis of the review of medical history of chest pain, ECG changes (Q wave), cardiac enzymes, and medical records. A review panel of 3 physicians (including cardiologists) made the final diagnostic determination of the presence of MI (recognized [Framingham event code 01 to 03] and unrecognized MI [code 04 to 05]).
For subjects who died, the cause of death was determined through medical history, interview of next of kin, review of medical records, death certificate, and review of the National Death Index. All-cause mortality referred to death from any cause.
Bilirubin was measured at the first and second examinations by use of a colorimetric method. Blood pressure was measured at each visit on seated subjects at rest. Hypertension was defined as a systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg or treatment for hypertension. Diabetes mellitus was defined as a history of diabetes mellitus, current treatment with a hypoglycemic medication, or fasting glucose ≥7.0 mmol/L (126 mg/dL). Total cholesterol was measured with the use of a manual Abell-Kendall procedure.24 Data on alcohol consumption, smoking, physical activity, and education were collected using a standardized questionnaire. Existing comorbid conditions were ascertained using the medical history and a review of medical records.
We created sex-specific tertiles of serum albumin and created 2 indicator variables. For each outcome, person-time of follow-up was calculated from the first examination until the occurrence of the outcome of interest, or loss to follow-up, or censoring date. We used the Mantel-Haenszel method to calculate incidence rate ratios and incidence rate differences within each sex, by using the highest tertile of albumin as the reference category.
For assessment of confounding, we initially created categories of age (5-year categories), body mass index (BMI) (quartiles), total cholesterol (quartiles), HDL-cholesterol (quartiles), smoking (never, former, and current smokers of 1 to 20, 20 to 30, 31+ cigarettes/d), alcohol intake 0, 1 to 6, 7 to 12, 13 to 24, and >24 g/d of total ethanol, bilirubin (quartiles), diabetes mellitus (yes/no), and blood pressure status: hypertensive subjects (systolic blood pressure of at least 140 mm Hg, diastolic blood pressure of at least 90 mm Hg, or treatment for hypertension) and normotensive subjects with systolic blood pressure (a) <120, (b) between 120 and 129, or (c) between 130 and 139 mm Hg. Each variable was assessed individually and in combination with other relevant confounders. When a linear relation between a covariate and the outcome was detected, the covariate was used as a continuous variable in the regression model. The final analyses retained age, total cholesterol, and blood pressure as major confounders. This final model yielded comparable results to the Cox regression model adjusting for age, BMI, blood pressure categories (4 categories as above), total cholesterol, HDL-cholesterol, pack-years of smoking, alcohol consumption, and diabetes mellitus. We repeated the primary analysis after exclusion of 61 cases of unrecognized MI (28 silent, Framingham event code 04). We conducted analyses stratified by cholesterol groups (high [at least 5.17 mmol/L] and low [below 5.17 mmol/L]) and hypertension status (yes/no). In addition, we examined the albumin-MI association stratified by diabetes mellitus (yes/no), smoking status (never, former, and current smokers), alcohol status (yes/no), BMI (using sex-specific medians as cut points), and age (≤40/>40 years). For all-cause mortality, we assessed the sensitivity of the findings to preexisting diseases by excluding deaths that occurred during the first 5 years of follow-up.
Because serum albumin concentrations may vary over time, we used a time-dependent Cox regression model for the primary outcome (MI). For these analyses, albumin and all time-dependent covariates were updated at the second examination.
Among 4506 subjects (48% men) who had serum albumin measurements and who were free of MI at baseline, the mean age was 37.6 years. The mean albumin was 47.8 g/L for men and 45.6 g/L for women. Tables 1 and 2⇓ show the baseline characteristics of the study participants by tertiles of serum albumin. During a mean follow-up of 21.9 years, MI developed in 217 men and 63 women. In a Cox regression model adjusting for age, BMI, total cholesterol, HDL-cholesterol, alcohol intake, smoking, serum bilirubin, and blood pressure categories, the hazard ratios (HR) (95% CI) of MI in men were 1.0, 1.25 (0.84 to 1.84), and 1.49 (1.01 to 2.21), respectively, from the highest to the lowest tertile of albumin (Table 3). Corresponding values for women were 1.0, 1.79 (0.88 to 3.65), and 2.12 (1.06 to 4.27), respectively. Excluding 61 cases of unrecognized cases of MI made the results stronger (multivariate HR [95% CI] from the highest to the lowest tertile of albumin were 1.0, 1.29 [0.84 to 2.00], and 1.55 [1.01 to 2.40] for men and 1.0, 2.04 [0.86 to 4.83], and 2.78 [1.20 to 6.47] for women).
Using the time-dependent Cox regression, from the highest to the lowest tertile of serum albumin, HRs of MI (95% CI) were 1.0, 1.16 (0.77 to 1.75), and 1.52 (1.01 to 2.29) for men and 1.0, 1.55 (0.72 to 3.33), and 2.10 (1.02 to 4.33) for women, adjusting for age, BMI, total cholesterol, HDL-cholesterol, alcohol intake, smoking, serum bilirubin, and blood pressure categories.
We stratified by high versus low levels of serum cholesterol (<5.17 mmol/L versus ≥5.17 mmol/L) and found similar relations between albumin and MI in both groups: From the highest to the lowest tertile of albumin, multivariate adjusted relative risks of MI were 1.0, 1.3 (0.7 to 2.3), and 1.8 (1.0 to 3.2), respectively, among subjects with low cholesterol and 1.0, 1.5 (1.0 to 2.3), and 1.9 (1.2 to 2.8), respectively, for the high cholesterol group.
In the multivariate Cox regression, there was evidence for effect modification by hypertension (Table 4). From the highest to the lowest tertile of albumin, multivariate HRs (95% CI) of MI were 1.0, 0.86 (0.52 to 1.41), and 1.05 (0.65 to 1.70), respectively, for normotensive men and 1.0, 2.56 (1.29 to 5.05), and 2.64 (1.31 to 5.32), respectively, for hypertensive men. In women, corresponding values were 1.0, 1.01 (0.40 to 2.54), and 1.54 (0.66 to 3.56), respectively, for normotensive and 1.0, 3.92 (1.09 to 14.18), and 3.42 (0.92 to 12.69), respectively, for hypertensive subjects. In additional analyses, low albumin was associated with an increased risk of MI when stratified by age, BMI, and diabetes mellitus (data not shown). Although low serum albumin was related to an increased rate of MI among current drinkers and current smokers, we did not have enough cases of MI among current nondrinkers, former smokers, and never-smokers for stable estimates (data not shown).
From the highest to the lowest tertile of albumin, multivariate adjusted HRs (95% CI) for total mortality rates were 1.0, 0.85 (0.61 to 1.16), and 1.24 (0.91 to 1.68), respectively, for men and 1.0, 1.34 (0.91 to 1.98), and 1.59 (1.09 to 2.33), respectively, for women (Table 5). Exclusion of deaths occurring within the first 5 years of follow-up had little effect on the results (HR: 1.0, 0.84 (0.60 to 1.18), 1.22 (0.88 to 1.68), respectively, for men and 1.0, 1.38 (0.92 to 2.06), 1.58 (1.06 to 2.35), respectively, for women.
In this prospective study, we found that lower concentrations of serum albumin were associated with an increased risk of MI among men and women. The association persisted after adjusting for traditional risk factors. The albumin-MI association was modified by hypertension, in that hypertensive subjects showed a steeper relation of albumin to MI risk. There was evidence for an increased risk of all-cause mortality with lower albumin levels among women.
Albumin and MI
Several observational studies did not find an association between serum albumin and CHD.10–14⇓⇓⇓⇓ It has been suggested that the relation between serum albumin and CHD may vary across sex, age, and level of serum cholesterol. Corti and colleagues12 found an association between serum albumin and CHD among older women but not among older men (≥71 years of age). In the Zutphen Elderly study,10 serum albumin was not associated with a 5-year incidence of CHD among men ≥64 years of age; however, an association between serum albumin and CHD incidence was observed among men with elevated serum cholesterol (>6.4 mmol/L). Gillum and Makuc15 did not find an association between albumin and CHD among men ≥65 years of age. In contrast, other studies have reported a positive association between low albumin and increased risk of CHD. In the NHANES I study, Gillum and Makuc15 reported an increased risk of CHD incidence with low albumin in men <65 years of age and women 45 to 74 years of age. Other researchers have demonstrated a detrimental effect of low albumin on the risk of CHD6,8⇓ and stroke.9
There are limitations to the albumin-CHD relation from those previous studies. First, most of the previous studies had a relatively short follow-up time (from 1 to 5 years). Second, all of these studies used a single measurement of albumin at baseline and are thus subject to regression dilution bias.7 Third, residual confounding might be an issue because not all studies controlled for alcohol use, diabetes mellitus, physical activity, and dietary factors known to influence both serum albumin and CHD. In the current project, we have data accumulated over 21.9 years of follow-up. In addition, we have measurements on serum albumin at baseline and at the second examination (an average of 8 years later). A further strength of this study is the ability to update important covariates at the time of second albumin measurement. With the use of time-dependent Cox regression, we had little evidence for regression dilution bias.
Although our findings suggest that low serum albumin may be a risk factor for MI, it is possible that low albumin might not have a direct causal role for CHD but could be an indicator of an underlying chronic condition.
Albumin and Mortality
Although Law et al11 did not find evidence for an association between a lower level of serum albumin and death from cardiovascular disease and cancer, the majority of published data are in favor of detrimental effects of low serum albumin on death. Most studies report that a lower concentration of serum albumin is associated with a 2-fold increased risk of total cardiovascular mortality10,25⇓ and all-cause mortality.10,15,16⇓⇓ Results from a meta-analysis showed that the combined relative risk for all-cause mortality associated with low albumin was 1.9 (95% CI: 1.6 to 2.3).6
Our findings of an albumin-mortality relation among women were consistent with the majority of published data, and there was little confounding by other cardiovascular risk factors. However, among men, adjustment for confounding weakened the association between albumin and mortality. We do not have a good explanation for the sex differences. It is possible that women may be more susceptible to low levels of albumin than men as the result of biology or lifestyle factors. Women may have had different dietary habits than men; however, the lack of data on dietary intake precludes the testing of this hypothesis. If low serum albumin were merely an indicator of some chronic condition without a causal role on CHD or mortality, then the observed sex difference could be due to differential burdens of disease across sexes.
Potential Mechanisms of Effect
It is unclear whether the prognostic value of low albumin simply reflects inflammation or if there is an independent effect of albumin. Reuben and colleagues25 have shown that low albumin is associated with an increased mortality rate among healthy subjects without evidence of inflammation (IL-6 level <3.2 pg/mL). There is substantial evidence to indicate that low albumin might be causally related to cardiovascular disease and death. Albumin has antioxidant properties. Albumin at concentrations less than physiological can inhibit copper-stimulated peroxidation and hemolysis.26 Albumin also inhibits the production of free hydroxyl radicals from systems containing copper ions and H2O226 and is able to scavenge peroxy radicals.27 Albumin also inhibits copper-dependent lipid peroxidation systems.7 LDL oxidation is one of the early steps in the atherosclerotic process.9,28⇓ Serum albumin may inhibit endothelial apoptosis.29
First, subjects in this study were white, thus limiting the generalizability of our findings. Second, there is a possibility of a residual dilution bias because we only had two measurements of serum albumin over the follow-up time. Third, we did not have nutrition data to evaluate the role of nutrition status on the studied associations. Fourth, information bias and unmeasured confounding could have influenced our results. Fifth, some of our results (ie, Table 4) had wide confidence intervals and could have been observed by chance. Last, the observational nature of our study limits causal inference.
In conclusion, low serum albumin appears to be a risk factor for MI. In addition, there is evidence suggesting that low albumin may also be a predictor of all-cause mortality in women. Because it is easy to measure at a relatively low cost, albumin in conjunction with other traditional risk factors may aid in identifying those at risk for MI and premature death.
This study was supported by the National Heart, Lung, and Blood Institute, the National Institutes of Health (NIH/NHLBI Contract N01-HC-38038), and the Institute on Lifestyle and Health, Boston University School of Medicine.
- ↵American Heart Association. Heart and Stroke Statistical Update. Dallas, Tex: American Heart Association; 2002.
- ↵Kuller LH, Eichner JE, Orchard TJ, et al. The relation between serum albumin levels and risk of coronary heart disease in the Multiple Risk Factor Intervention Trial. Am J Epidemiol. 1991; 134: 1266–1277.
- ↵Gillum RF, Ingram DD, Makuc DM. Relation between serum albumin concentration and stroke incidence and death: the NHANES I Epidemiologic Follow-up Study. Am J Epidemiol. 1994; 140: 876–888.
- ↵Law MR, Morris JK, Wald NJ, et al. Serum albumin and mortality in the BUPA study: British United Provident Association. Int J Epidemiol. 1994; 23: 38–41.
- ↵Kuller LH, Tracy RP, Shaten J, et al. Relation of C-reactive protein and coronary heart disease in the MRFIT nested case-control study: Multiple Risk Factor Intervention Trial. Am J Epidemiol. 1996; 144: 537–547.
- ↵Tracy RP, Lemaitre RN, Psaty BM, et al. Relationship of C-reactive protein to risk of cardiovascular disease in the elderly: results from the Cardiovascular Health Study and the Rural Health Promotion Project. Arterioscler Thromb Vasc Biol. 1997; 17: 1121–1127.
- ↵Kannel WB, Feinleib M, McNamara PM, et al. An investigation of coronary heart disease in families: the Framingham offspring study. Am J Epidemiol. 1979; 110: 281–290.
- ↵Abell LL, Levy BB, Brodie BB, et al. A simplified method or estimation of total cholesterol in serum and demonstration of its specificity. J Biol Chem. 1952; 195: 357–366.
- ↵Zoellner H, Hofler M, Beckmann R, et al. Serum albumin is a specific inhibitor of apoptosis in human endothelial cells. J Cell Sci. 1996; 109: 2571–2580.