Association between Inflammation-Associated Cytokines, Serum Albumins, and Mortality in the Elderly

Association between InflammationAssociated Cytokines, Serum Albumins, and Mortality in the Elderly Dennis H. Sullivan, MD, Paula K. Roberson, PhD, Larry E. Johnson, MD, PhD, Priya Mendiratta, MD, Melinda M. Bopp, BS, and Osama Bishara, MD Objective: To investigate the association between serum albumin, prealbumin, various serum inflammation associated-cytokines, and mortality in older geriatric recuperative care patients. Design: A prospective cohort study. Setting: A geriatric rehabilitation unit of a universityaffiliated Department of Veterans Affairs hospital. Participants: Participants were 53 geriatric patients (mean age 78 ⫾ 7.3, 96% male) admitted to a Geriatric Evaluation and Management (GEM) unit. Patients with documented near-terminal medical disorder, overt infections, and any systemic or localized inflammatory disorders were excluded. Measurements: Inflammation-associated cytokines (IL-8, IL-1␤, IL-6, IL-10, TNF-␣), albumin, prealbumin, and C-reactive protein were measured at hospital discharge and each subject was then tracked for 1 year.

Numerous studies of older adult populations have found strong associations between serum albumin and the risk of Geriatric Research Education and Clinical Center, Central Arkansas Veterans Healthcare System, Little Rock, AR (D.H.S.); Donald W. Reynolds Department of Geriatrics, University of Arkansas for Medical Sciences Little Rock, AR (D.H.S., L.E.J., P.M., M.M.B.); Department of Biostatistics, University of Arkansas for Medical Sciences Little Rock, AR (P.K.R.); Geriatrics and Extended Care Service, Central Arkansas Veterans Healthcare System Little Rock, AR (L.E.J.); Department of Internal Medicine, Ohio State University Columbus, OH (O.B.). This study was supported by a Medical Research Endowment Award from the University of Arkansas for Medical Sciences. This was an investigator-initiated study that was funded by the University of Arkansas for Medical Sciences. The sponsor did not contribute to any part of the study or the preparation of the manuscript. D.H.S. conceptualized and designed the study, interpreted the results, and analysis, and was primary author of the manuscript. P.K.R. was involved in study concept and design, data analysis, and critical revision of the manuscript. L.E.J. and P.M. were involved in the study concept and design and interpretation of data. M.M.B. was involved in study concept and design, acquisition of data, and preparation of the manuscript. O.B. conceptualized and designed the study, interpreted the results, and analysis, and critical revision of the manuscript. Address correspondence to Dennis H. Sullivan, MD, Geriatric Research Education Clinical Center (3J/NLR), Central Arkansas Veterans Healthcare System, 4300 W. 7th Street, Little Rock, AR 72205. E-mail: [email protected]

Copyright ©2007 American Medical Directors Association DOI: 10.1016/j.jamda.2007.04.004 458 Sullivan et al

Main results: By Cox Proportional-Hazards Regression analysis, the strongest predictor of mortality within 6 months of study entry was the serum IL-6. For each log increase in IL-6, there was nearly a 9-fold greater 6-month mortality risk (RR 8.99, 95% CI 1.65 to 49.03). The association between albumin and mortality was no longer significant after controlling for IL-6. There was a strong inverse correlation between IL-6 and both albumin (R2 0.39, P ⬍ .001) and prealbumin (R2 0.41, P ⬍ .001). Conclusion: Subclinical inflammation appears to be an important factor contributing to low serum albumins in older recuperative care patients and may confound the association between albumin and mortality in this population. More in-depth studies of these associations are warranted. (J Am Med Dir Assoc 2007; 8: 458–463) Keywords: Elderly; protein energy undernutrition; inflammatory cytokines; albumin; mortality

subsequent mortality.1,2 For this reason, albumin can be considered to be an important indicator of health status in these populations. Whether it is more specifically a marker of protein-energy nutritional status is less certain. Although albumin production by the liver is dependent on an adequate nutrient intake,3,4 other disease processes also affect the serum albumin concentration.3,5 Of particular relevance, albumin is known to be a negative acute phase reactant.3,5 In response to major surgery, sepsis, or other systemic inflammatory events, the serum albumin concentration can drop by 10 to 30 g/L within a few days.6,7 This drop in albumin appears to be mediated by the inflammation-associated cytokines which include interleukin (IL)-6, IL-1␤, tumor necrosis factor (TNF)-␣, and possibly IL- 8 and IL-10.8,9 It is therefore clear that albumin has very low specificity as a nutritional marker following an acute inflammatory event. In this situation, the albumin-mortality association is probably reflective of the seriousness of the underlying inflammatory event. The significance of low serum albumin is more controversial when there is no overt inflammation and in the recuperative period after the acute inflammatory events have apparently resolved. In these situations, it is not clear whether persistent hypoalbuminemia represents poor nutrient intake, JAMDA – September 2007

ongoing subclinical inflammation, or another disease process. Because the association between albumin and mortality remains very strong in recuperative care and outpatient settings, this is an important issue.2,10,11 A better understanding of this controversy may lead to improved therapeutic interventions aimed at reducing the high mortality rate of hypoalbuminemic patients. In this study, we examined the interrelationships among serum albumin, various serum inflammation associated– cytokines, nutritional reserve (body mass index [BMI]), and mortality in a select population of older adults admitted to a geriatric recuperative care and rehabilitation unit. All of the subjects were free of preexisting conditions known to cause hypoalbuminemia. The purpose of the study was to look for evidence that subclinical inflammation is an important determinant of persistent hypoalbuminemia and increased mortality in this population. METHODS Enrollment This was a prospective study of 53 patients recruited from the inpatient Geriatric Evaluation and Management Unit (GEM) or a transitional care bed within the Nursing Home Care Unit (NHCU) of a university-affiliated Veterans Administration (VA) hospital. Subjects were eligible for entry into the study if they were ready for discharge home, 65 years of age or older, and free of cancer, ongoing infections, and other known inflammatory disorders for at least 14 days. Patients were considered to be free of infection if they maintained a normal core temperature (⬍100°C) and a white blood cell count of less than 10.8 cells/mm3 off antibiotics. All patients admitted to these wards routinely undergo comprehensive ongoing medical, functional, and nutritional assessments that include biweekly serum albumin determinations. Eligibility was also based on the latest serum albumin obtained more than 14 days after admission and closest to the planned discharge. Based on prior evaluations of this population, it was anticipated that the proportion of otherwise eligible patients with predischarge albumins higher than 35 g/L would be greater than the proportion with albumins lower than 30 g/L. For this reason, we oversampled for patients with low albumins. To do this, patients were stratified into 3 groups based on their serum albumin values obtained after day 14 and after they met the study entry criteria. These groups were as follows: (1) the low albumin group (serum albumin level ⱕ 30 g/L), (2) intermediate albumin group (serum albumin level ⬎30 –35 g/L ), and (3) high albumin group (serum albumin level ⬎35 g/L). To have approximately equal numbers of subjects from each of the 3 groups, patients to be recruited into the study were chosen randomly from groups 2 and 3 using a block recruitment design (50% from the intermediate group, and 33% in the high albumin group). This was done using computer-generated random numbers. All patients in group 1 were considered eligible. Between April and December 2004, 106 patients were identified to be potentially eligible for study entry. Per the above recruitment protocol, 55 were selected. All 55 received ORIGINAL STUDIES

oral and written explanations of the study, including possible risks involved, in accordance with the ethical standards of the Department of Veterans Affairs and the Human Research Advisory Committee of the University of Arkansas for Medical Sciences. Two of these patients declined our offer of study admission. The remaining 53 entered the study after signing Health Insurance Portability and Accountability Act (HIPAA) and informed-consent documents. At study entry, clinical data were collected by the study investigators including clinical diagnoses, medications, height and weight, demographics, and functional status as assessed using the Katz Index of Activities of Daily Living (ADL) Scale.12 Subsequently, each patient was tracked (clinic visits and/or by phone) for 1 year postenrollment to identify those who died. Measurement of Cytokines and C-Reactive Protein At study entry, serum was obtained from each subject. One sample was sent for measurement of C-reactive protein (CRP), albumin, and prealbumin. A second tube was placed immediately on ice, rapidly transported to the laboratory, and then spun down. The serum was then flash frozen and stored at –70°C until analyzed for select inflammation-associated cytokines including IL-1␤, IL-6, IL-8, IL-10, and TNF-␣. A Cytometric Bead Array (CBA) analysis was used to simultaneously quantify each of these cytokines in a single sample. The CBA beads were analyzed with a FACSCalibur flow cytometer (BD Biosciences, San Jose, CA) and the quantity of each of the respective cytokines was calculated with the proprietary software. This assay system has been extensively tested for sensitivity, specificity, and reliability.13–15 The sensitivity of the assays is in the 1.5 to 7 pg/mL range, and intraand interassay reproducibility was determined to be in the 2% to 8% range across a wide concentration range. Because the assays are run against standards, linearity of the standard curves is also important. It has been determined that the slope of matched assays for each cytokine is in the range of 1.00 ⫾ 0.04. The assays are also highly specific with no known cross-reactivity or background detection of other proteins. For this panel of cytokines, the CBA was also validated against an enzyme-linked immunosorbent assay (ELISA). The ELISA and CBA values showed good correlation with coefficients of variation of less than 10% for each cytokine. For this study, duplicate CBA analyses were performed on all samples and the average of the 2 measures was used in the analyses. Albumin and prealbumin were measured using standard procedures. A highly sensitive latex-based immunoassay was used to determine the levels of CRP. Statistics The primary outcome was 6-month mortality. One-year mortality was also examined. In the first set of analyses, the relationship between each variable of interest and mortality was analyzed using univariable Cox Proportional-Hazards regression. Because of their skewed distributions, many of the continuous variables were either log transformed or expressed as categorical groupings. BMI, weight loss, prealbumin, CRP, and each of the inflammation-associated cytokines were log Sullivan et al 459

Table 1. Characteristics of Study Subjects All Subjects (n ⴝ 53)

Variable Age, y, mean ⫾ SD Body mass index, kg/m2, mean ⫾ SD Education, y, mean ⫾ SD Albumin, g/L, mean ⫾ SD Prealbumin, mg/L, mean ⫾ SD C-reactive protein, mg/L, median (interquartile range) IL-6, pg/mL, median (interquartile range) TNF-␣, pg/mL, median (interquartile range) IL-6 to IL-10 ratio, median (interquartile range) Total no. of medications, mean ⫾ SD No. of prescription medications, mean ⫾ SD No. of active problems, mean ⫾ SD Percentage of weight lost in previous year,* median (interquartile range) Weight as a percentage of usual, pre-illness weight,* mean ⫾ SD Charlson’s Co-Morbidity Index,† mean ⫾ SD Married, n (%) White race, n (%) Male, n (%) Independent in all activities of daily living,‡ n (%) Most prevalent active medical problems,§ n (%) Hypertension Deconditioned Congestive heart failure Arthritis Benign prostatic hypertrophy Coronary artery disease Diabetes mellitus, type 2 Depression Chronic Obstructive Pulmonary Disease Chronic renal insufficiency储

78.9 ⫾ 7.3 25.3 ⫾ 4.1 11.4 ⫾ 3.3 33.9 ⫾ 3.8 21.8 ⫾ 6.2 0.87 (0.42–1.55) 6.3 (3.8–2.5) 0.00 (0.00–0.00) 3.2 (1.00–6.97) 13.3 ⫾ 4.6 8.4 ⫾ 3.3 9.9 ⫾ 3.0 5.1 (0.0–8.8) 92.4 ⫾ 9.0 3.0 ⫾ 1.9 29 (54.7) 48 (90.6) 51 (96.2) 29 (54.7) 38 (71.7) 30 (56.6) 23 (43.4) 22 (41.5) 22 (41.5) 22 (41.5) 20 (37.7) 20 (37.7) 15 (28.3) 14 (26.4)

IL, interleukin; TNF-␣, tumor necrosis factor alpha. * Based on review of prior medical records. † Charlson’s Co-Morbidity Index.18 ‡ Independent in all of the basic activities of daily living (ADLs; bathing, dressing, toileting, transferring, continence, and feeding) as measured using the Katz Index of ADLs. § Includes both primary and secondary diagnoses. 储 Defined as a blood urea nitrogen greater than 30 mg/dL.

transformed. The log transformed IL-6 (inflammatory) to IL-10 (anti-inflammatory) cytokine ratio was also included as an inflammatory marker.16 Functional status was dichotomized (completely independent in all basic ADLs versus other). Albumin was categorized into 4 groups (25 to ⬍30 g/L, 30 to ⬍35 g/L, 35 to ⬍40 g/L, and 40 to ⬍45 g/L) based on previous studies and commonly used reference ranges.2,17 Only those variables found to be significantly associated with the outcome by univariable analysis were included in the subsequent multivariable analyses. In the second step, a series of multivariable Cox Proportional-Hazards Regression analyses were run using a stepwise procedure to determine which of the variables within a given set were the strongest predictors of mortality. For these analyses, the database variables were categorized into 1 of 3 sets: an albumin, a marker of inflammation (ie, CRP and cytokines), or a health status indicator (ie, BMI, Charlson’s CoMorbidity Index,18 functional status, and diagnostic and demographic variables). For each analysis, albumin was included in the analysis along with all of the preselected database variables from either or both of the other 2 sets. A similar 460 Sullivan et al

process was used to examine prealbumin. Select results were also examined graphically using Kaplan-Meier survival curves and analyzed using the log rank test. In the second part of the study, the relationship between albumin and each inflammation-associated cytokine was determined using least-squared regression analysis. For these analyses, the continuous variables (ie, raw data) were used. A multivariable regression analysis using a stepwise procedure was then employed to determine if the correlation between albumin and any cytokine remained significant after controlling for health status (as defined in the preceding paragraph). All data analyses were conducted using SAS software (version 9.1, SAS, Cary, NC). A 2-sided value of P ⬍ .05 was considered significant. RESULTS The characteristics of the 53 study subjects are shown in Table 1. Their ages ranged from 65 to 93 years and the majority consisted of white (91%) males (96%). The most prevalent active medical problems are also listed. JAMDA – September 2007

Unadjusted RR (95% CI) Six-month Mortality† Albumin, per 5 g/L increase C-reactive protein, per log increase IL-6, per log increase IL-6 to IL-10 ratio, per log increase Chronic renal insufficiency One-year Mortality‡ Albumin, per 5 g/L increase IL-6, per log increase IL-6 to IL-10 ratio, per log increase Independent in all activities of daily living§ Charlson’s Co-Morbidity Index Congestive heart failure Chronic renal insufficiency

0.35 (0.14–0.86) 5.15 (1.04–25.41) 8.99 (1.65–49.03) 8.55 (1.60–45.76) 5.16 (1.23–21.62) 0.42 (0.20–0.90) 2.07 (1.15–3.72) 6.28 (1.54–25.55) 0.16 (0.04–0.76) 1.46 (1.08–1.97) 4.10 (1.09–15.48) 5.94 (1.73–20.37)

IL, interleukin. * Based on univariable Cox Proportional-Hazards Regression analysis. † During the first 6 months of observation, 8 (15.1%) subjects died. ‡ During the 1-year period of observation, 11 (20.8%) subjects died. § Independent in all of the basic activities of daily living (ADLs; bathing, dressing, toileting, transferring, continence, and feeding) as measured using the Katz Index of ADLs.

Predictors of Mortality The primary outcome was 6-month mortality. During the first 6 months of observation, 8 (15.1%) subjects died. As shown in Table 2, albumin was a powerful predictor of 6-month mortality as were IL-6, IL-6 to IL-10 ratio, CRP, and renal insufficiency (defined as blood urea nitrogen greater than 30 mg/dL). When all 4 of these variables were included in the multivariable Cox Proportional-Hazards Regression analysis, the strongest predictor of mortality within 6 months (and the only variable to enter the model) was the serum IL-6. For each log increase in IL-6, there was nearly a 9-fold greater 6-month mortality risk (relative risk [RR] 8.99, 95% confidence interval [CI] 1.65 to 49.03). The relationship between IL-6 and mortality is shown graphically in Figure 1. The association between albumin and mortality was no longer significant once IL-6 entered the model (P ⫽ .584). The secondary outcome, 1-year mortality, was analyzed in the same manner. During the 1-year period of observation, 11 (20.8%) subjects died. As shown in Table 2, albumin, IL-6, and the IL-6 to IL-10 ratio were all significantly associated with 1-year mortality risk. Several health status indicators were also strong predictors of 1-year mortality including functional status, Charlson’s Co-Morbidity Index, and the diagnoses of renal insufficiency and congestive heart failure (CHF). When albumin and the 2 inflammation variables (IL-6 and IL-6 to IL-10 ratio) were included in the multivariable Cox Proportional-Hazards Regression analysis, the strongest predictor of mortality (and the only variable to enter the model) was the serum IL-6 to IL-10 ratio. When all 8 variables were included in the multivariable analysis, only ORIGINAL STUDIES

renal insufficiency entered the model. Subjects who had renal insufficiency had a nearly 7-fold increased risk of death compared with those without this diagnosis (RR 6.9, 95% CI 2.0 to 24.4). Prealbumin was not significantly associated with either 6-month or 1-year mortality risk (P ⬎ .05 for both analyses). Relationship between Albumin and the Inflammation-associated Cytokines By univariable analysis, albumin was inversely correlated with IL-6 (R2 0.39, P ⬍ .001), CRP (R2 0.14, P ⫽ .006), and the IL-6 to IL-10 ratio (R2 0.13, P ⫽ .007). When the inflammatory and health status variables were included in a multivariable analysis, IL-6 was the first variable to enter the model followed by functional status and education level. The strength of the correlation between albumin and IL-6 was unchanged and the model R2 was 0.53 (P ⬍ .001). After controlling for health status, IL-6 was also strongly correlated with prealbumin (partial R2 0.41, P ⬍ .001). DISCUSSION There are several important findings from this study. First, it confirms the importance of albumin as an indicator of health status. Even though the study was limited to a select population of geriatric recuperative care and rehabilitation patients who were doing well and nearing discharge home, a low albumin was associated with a poor prognosis for survival. Prior studies of older patients conducted in acute care hospitals, nursing homes, rehabilitation units, and outpatient settings have found a similar strong association between serum albumin and mortality.1,2,10,11,19 The consistency of these findings substantiates the importance of albumin as a prognostic marker. However, it remains to be determined why a low albumin has so much prognostic significance even in non–acutely ill older adults.

100

Probability of Survival (%)

Table 2. Variables Associated with Mortality Risk (n ⫽ 53)*

Log IL-6 < 1 (n=37)

90 80 70

Log IL-6 > 1 (n=16)

60 50 40

0

50

100

150

200

Days Fig. 1. Six-month survival based on Kaplan-Meier survival curves for the study population stratified into two groups based on serum IL-6. The groups are as defined within the figure and the text. The difference in survival between the two groups was highly significant by the log rank test (p⫽0.003). Sullivan et al 461

Relating to this last issue, this study also provides evidence to suggest that subclinical inflammation may be an important factor contributing both to low serum albumins and the associated increased mortality in non–acutely ill hypoalbuminemic patients. IL-6 in particular was strongly correlated with the serum albumin concentration. The ability of IL-6 and other inflammation-associated cytokines to lower serum albumin is well established.8,9 Like albumin, several markers of inflammation were also found to be significantly associated with mortality risk in this study. After controlling for these markers, the association between albumin and mortality no longer reached statistical significance. The fact that controlling for diagnoses and other health status indicators diminished the strength of the association between the inflammation-associated cytokines and mortality suggests that inflammation itself may be only a marker of ongoing chronic illness. Whether there is a causal relationship between high serum cytokine concentrations and mortality remains to be determined. Although this study provides further evidence that inflammation is an important determinant of the serum albumin concentration, it does not directly address the issue of whether albumin is ever a valid indicator of nutritional status or the adequacy of a patient’s nutrient intake. Serial measures of albumin, inflammatory markers, and nutrient intake may provide further clarification of this issue. Available evidence from nutrition intervention studies suggests that albumin is not responsive to improved nutrient intake.20 However, these studies did not control for the possible effects of inflammation. Given the known complex interrelationship between albumin, inflammation, and nutritional status, this issue needs further exploration. This study had a number of limitations. The approach to subject selection and follow-up monitoring was one. Patients were considered to be free of overt inflammation if they were free of cancer, ongoing infections, and other known inflammatory disorders for at least 14 days. This may not have been an adequate amount of time for the effects of recent inflammatory events to completely resolve. Because it was a crosssectional study, it was not possible to assess for any change in serum cytokine concentrations with time or to determine whether the trajectory of change was influenced by nutritional status or other disease processes. Prior studies have shown that older patients who remain hypoalbuminemic 90 days after hospital discharge have higher 1-year mortality than do those whose albumin concentrations rise during this time period.2 Whether changes in serum cytokine concentrations precede or follow changes in serum albumin remain to be determined. What effect such changes have on survival also needs to be investigated. There were other study limitations as well. The study population was relatively small and consisted primarily of white males. For this reason, the influence of race and gender on the interaction between albumin, cytokines, and mortality could not be discerned and we did not have an adequate sample size to examine complex multivariable analytic models. The cytokines and the method used to measure their concentrations in the serum also may not have been optimal. A CBA analysis was 462 Sullivan et al

used to measure the serum cytokine concentrations. Although the CBA was validated against ELISA for this panel of cytokines, it did not appear to have adequate sensitivity for some of the cytokines. For example, the median value for TNF-␣ was 0 (as shown in Table 1). Stronger associations between mortality and some of the cytokines may have been found had the serum concentrations of these cytokines been measured using ELISA. Because TNF production itself is episodic, the plasma half-life short, and random sampling too imprecise, it may also be necessary to assay for the soluble TNF-␣ receptors, TNFR-p75 and TNFR-p55.21 CONCLUSION Subclinical inflammation appears to be an important factor contributing to the low serum albumins in older recuperative care patients and may confound the association between albumin and mortality in this population. More in-depth studies of these associations are warranted. REFERENCES 1. Goldwasser P, Feldman J. Association of serum albumin and mortality risk. J Clin Epidemiol 1997;50(6):693–703. 2. Sullivan DH, Roberson PK, Bopp MM. Hypoalbuminemia 3 Months after hospital discharge: Significance for long-term survival. J Am Geriatr Soc 2005;53(7):1222–1226. 3. Rothschild MA, Oratz M, Schreiber SS. Serum albumin. Hepatology 1988;8:385– 401. 4. Doweiko JP, Nompleggi DJ. Role of albumin in human physiology and pathophysiology. JPEN J Parenter Enteral Nutr 1991;15:207–211. 5. Doweiko JP, Nompleggi DJ. The role of albumin in human physiology and pathophysiology, Part III: Albumin and disease states. JPEN J Parenter Enteral Nutr 1991;15:476 – 483. 6. Puskarich-May CL, Sullivan DH, Nelson CL, et al. The change in serum protein concentration in response to the stress of total joint surgery: a comparison of older versus younger patients. J Am Geriatr Soc 1996;44:555–558. 7. Hoye RC, Bennett SH, Geelhoed GW, et al. Fluid volume and albumin kinetics occurring with major surgery. JAMA 1972;222(10):1255–1261. 8. Johnson AM. Low levels of plasma proteins: Malnutrition or inflammation? Clin Chem Lab Med 1999;37(2):91–96. 9. Gabay C, Kushner I. Acute-phase proteins and other systemic responses to inflammation. N Engl J Med 1999;340(6):448 – 454. 10. Sullivan DH, Walls RC. The risk of life-threatening complications in a select population of geriatric patients: The impact of nutritional status. J Am Coll Nutr 1995;1:29 –36. 11. Reuben DB, Ferrucci L, Wallace R, et al. The prognostic value of serum albumin in healthy older persons with low and high serum interleukin-6 (IL-6) levels. J Am Geriatr Soc 2000;48(11):1404 –1407. 12. Katz S, Ford AB, Moskowitz RW, et al. Studies of illness in the aged. The index of ADL: A standardized measure of biological and psychosocial function. JAMA 1963;185(12):914 –919. 13. Morgan E, Varro R, Sepulveda H, et al. Cytometric bead array: A multiplexed assay platform with applications in various areas of biology. Clin Immunol 2004;110[3]:252–266. 14. Camilla C, Mely L, Magnan A, et al. Flow cytometric microsphere-based immunoassay: Analysis of secreted cytokines in whole-blood samples from asthmatics. Clin Diagn Lab Immunol 2001;8(4):776 –784. 15. Carson R, Vignali D. Simultaneous quantitation of fifteen cytokines using a multiplexed flow cytometric assay. J Immunol Methods 1999;227:41–52. 16. Simovic MO, Bonham MJ, bu-Zidan FM, et al. Anti-inflammatory cytokine response and clinical outcome in acute pancreatitis. Crit Care Med 1999;27(12):2662–2665. 17. Sullivan DH, Walls RC. Protein-energy undernutrition and the risk of mortality within six years of hospital discharge. J Am Coll Nutr 1998; 17(6):571–578. JAMDA – September 2007

18. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Dis 1987;40:373–383. 19. Rudman D, Feller AG, Nagraj HS, et al. Relation of serum albumin concentration to death rate in nursing home men. JPEN J Parenter Enteral Nutr 1987;11:360 –363.

ORIGINAL STUDIES

20. Bernstein L. Measurement of visceral protein status in assessing protein and energy malnutrition: Standard of care. Prealbumin in Nutritional Care Consensus Group. Nutrition 1995;11(2):169 –171. 21. Hasegawa Y, Sawada M, Ozaki N, et al. Increased soluble tumor necrosis factor receptor levels in the serum of elderly people. Gerontology 2000; 46(4):185–188.

Sullivan et al 463