JAGS 39:891-894, 3991
Detectable Serum Levels of Tumor Necrosis Factor Alpha May Predict Early Mortality in Elderlv Institutionalized Patients . I
Arshag D. Mooradian, MD* Richard L. Reed, MD,* Dan Osterweil, MD,t and Philip Scuderi, P h w
Objective: To determine if detectable serum tumor necrosis factor alpha (TNR levels are associated with higher mortality in nursing home residents. Subjects and methods: The basal serum concentrations of TNF and interleukin-2 alpha (IL-2)were measured in 129 elderly nursing home patients (mean age of 89 years), and survival in the cohort was monitored over a 13month period. Results: At 4 months follow-up, seven out of 33 patients with detectable serum TNF levels had died (21.2%), and only three out of 96 patients with undetectable serum TNF levels had died (3.2%) (PCO.001).The
difference in mortality remained significant up to 13 months of follow-up (PCO.05). Those with detectable serum TNF levels and those with undetectable levels were comparable in age, body mass index, hematocrit, lymphocyte counts, and serum lever of albumin, prealbwmin, and retinol-binding protein. When patients with detectable serum IL-1 levels were compared to those with undetectable levels, there were no significant differences in mortality over a 23-month period. Conclusion: Detectable serum TNF levels in elderly nursing home patients may be a predictor of early mortality. J Am Geriatr SOC395391494, 1991
dministration of tumor necrosis factor alpha nursing home patients who had had their serum TNF (TNF) to experimental animals and humans and interleukin-1 alpha (IL-1) measured. The results results in a myriad of metabolic and hemo- suggest that detectable serum levels of TNF, but not of dynamic changes.'-' However, the biolog- IL-1,may be a predictor of early mortality in nursing ical significance of serum TNF levels in human subjects home residents. is still not clear. Previous studies have found that aging is not associated with significant alterations in either MATERIALS AND METHODS basal serum levels of TNF9," or in TNF secretory Human Serum Samples A totaI of 129 nursing capacity of mononuclear cells in culture." We had also reported that elderly patients with detectable serum home residents with a mean age of 89.1 years f 5.0 TNF levels were more likely to have low serum con- were recruited from the Los Angeles Jewish Homes for centrations of 3,5,3' triiodothyronine.'2 These obser- the Aging. All the participants were residents of a vations suggest that detectable serum levels of TNF in skilled nursing home facility, and the average length elderly patients may have biological implications even of stay for the group before the study was 18 months. though they do not correlate with body weight loss?- Since glucocorticoids may suppress TNF production,L3 patients on systemic or topical glucocorticoid therapy l2 To test this hypothesis, we examined the death rate and cause of mortality in a cohort of 129 elderly were excluded. Blood specimens were collected in nonheparinized tubes, and sera were separated within an hour of blood collection to avoid artifactual changes in serum TNF measurement^.'^ For each patient, the folFrom the *ArizonaCenter on Aging, University of Arizona College lowing information was recorded: major diagnoses, of Medicine, Tucson, Arizona; and the tJewish Homes for the Aging, Los Angeles, California, and $Miles, Inc., Berkeley, California. number of prescription drugs, and body height and This work was supported in part by the Flinn Foundation. weight changes over a 6-month observation period. Address correspondence and reprint requests to Arshag D. MoorThe data pertaining to c l i c a l diagnoses and drugs adian, MD, Division of Endocrinology, St. Louis University Medical School, 1402 So. Grand Blvd., St. Louis, MO 63104. used were extracted from the patients' records. The
A
0 1991 by the American Geriatrics Society
892
MOORADIAN ET AL
laboratory evaluation included a complete blood count, serum albumin, cholesterol, prealbumin, and retinolbinding protein. A detailed account of the clinical characteristics and serum TNF levels of most of these patients (127 out of 129) has been previously reported? The results of thyroid function tests in some of these patients have also been previously reported." Over a 13-month follow-up period, survival in this group was monitored, and the circumstances of each death were observed. None of the patients involved in this study was discharged or lost to follow-up.
JAGS-SEPTEMBER 1991-VOL. 39,NO.9
between the two survival curves at 4 and 13 months after determination of TNF levels.
RESULTS Of the 129 patients tested, 33 had detectable serum levels of TNF while in 96 they were below 40 pg/mL (2 PM) (the lower limit of the assay detection). The mean serum TNF concentration in the group with detectable levels was 264.0 f 138 pg/mL (13.2 f 6.9 PM)with a range of 150 to 825 pg/mL (7.5-41.25 PM). The two groups were similar in mean age, body weight, TNF and IL-1 ELISA Systems An ELISA system height, hematocrit, concentration of hemoglobin, the specific for human TNF was used to detect serum white blood cell count and serum concentrations of levels. The details of this assay system have been cholesterol, albumin, pre-albumin, and retinol-binding previously p~b1ished.l~ In brief, 96-Well Immulon I1 protein (Table 1). Although the percentage of white plastic plates were coated with mAb 6E, a murine blood cells that were identified as lymphocytes in monoclonal antibody specific for human TNF. This patients with undetectable serum TNF levels (26.16% antibody was a gift from Genetech, Inc. (South San f: 9.51%) was somewhat lower than the percentage Francisco, CA). Sera were added to the wells and found in those with detectable serum TNF levels incubated for 1 hour at room temperature. The plates (31.27% k 14.02%),the total lymphocyte count in the were washed three times with PBS containing 0.05% two groups was not significantly different. In addition, Tween-20, and a rabbit anti-TNF reagent was subse- the two groups were similar in the prevalence rates of quently added. After a similar incubation and wash various diseases such as diabetes, dementia, thyroid procedure, a peroxidase-conjugated goat anti-rabbit disease, congestive heart failure, depression, cancer, renal failure, and urinary tract infections or bed sores. IgG was added to the wells. The peroxidase substrate The prevalence of these diseases in the two groups used was 2,2'-azino-bis-3-ethylbenzthiazoline-6-~~1fonic acid (Sigma, Inc.). Optical density readings were studied is indicated in a previous report.' The mean made on an ELISA reader set at 405 nm within 10 number of medical diagnoses in the group with detectminutes of substrate addition. The concentration of able serum TNF levels (4.13 f 0.8) and the number of TNF in sera was determined by comparison with a set medications used (9.5 f 4.8) were not different from of standards made with recombinant human TNF. the number of diagnoses (4.21 f 0.9) and the number (Genetech Inc., South San Francisco, CA). The lower of medications used (9.9 f 4.6) by those who did not limit of detection with this assay system is 40 pg/mL have detectable serum TNF levels. The types of drugs (2 PM).The interassay coefficient of variation was 4.8% used by these patients were also similar. at 300 pg/mL. The ELISA for human IL-1 alpha has been described TABLE 1. ELDERLY NURSING HOME PATIENTS previously" and was carried out using essentially the WITH DETECTABLE SERUM TNF LEVELS AND same protocol used for the TNF assay. In this assay THOSE WITH UNDETECTABLE LEVELS system, 10 p g of ascitic protein containing the IL-1 Patients Patients specific mAb C12 (Olympus, Inc., Lake Success, NY) Detectable TNF Undetectable TNF (n = 33) (n = 96) was used to coat each well of the assay plates. A rabbit anti-IL-1 alpha specific anti-serum was used as the 89.17 f 5.06 88.85 f 3.87 Age second reagent. The serum concentrationsof IL-1 alpha Sex M:F 6:27 15:81 54.24 f 10.78 53.63 f 10.30 were determined by comparison with a set of standards Body weight (Kg) 153.60f 13.55 151.17& 12.86 made by adding known amounts of recombinant hu- Height (Cm) 38.64 f 3.69 Hematomit (%) 38.06 f 4.19 man IL-1 alpha to serum. The lower limit of detection 12.71 f 1.15 12.50 f 1.35 Hemoglobin (g/dL) with this assay system is 50 pg/mL (2.8 PM). The White blood cell 7137.3 & 2290.1 7709.4 f 2760.2 interassay coefficient of variation was 5.1%at 300 pg/ count/mm3 26.16 & 9.59* mL. All the samples were measured in the same day Percent lymphocytes 31.27 f 14.02 (%I and in duplicates. 3.59 f 0.39 3.51 f 0.43 All the results are reported as mean f SD. Statistical Albumin (gm/dL) 210.3 f 41.7 Cholesterol (mg/dL) 201.4 f 56.7 analysis was done with the two-tailed Student's t test Prealbumin (mg/dL) 29.8 f 14.5 27.1 f 12.9 for unpaired variables with Bonferoni's correction. Pro- Retinol binding pro4.56 2 1.45 4.28 1.36 tein (mg/dL) portions were compared using the chi square test. A Kaplan-Meier analysis of survival was carried 0 ~ t . I ~ Mean & SD. The log-rank test'* was used to evaluate the differences * P < 0.05.
*
JAGS-SEPTEMBER 1991-VOL. 39, NO.9
-l
g, 85 m c c
n 75
O 1 ;
-
TUMOR NECROSIS FACTOR IN THE ELDERLY
LI
893
I
TNF Positive (11-33) TNF Negative (n-96)
; ; ; : ; ; ; ; ; ;0;1;*:3 Months After Initial Measurement
FIGURE 1. Percentagesof 33 elderly nursing home patients with detectable serum levels of TNF (TNF positive) and the percentages of 9 6 patients with undetectable serum levels (TNF negative) who survived during the 13-month followup. The log-rank statistic showed a P value of CO.001 at 4 months and (0.05 at 13 months of follow-up.
The survival analysis (Kaplan-Meier Estimates) of patients with detectable serum TNF levels (TNF-positive) and those with undetectable levels (TNF-negative) is shown in Figure 1. By 4 months of follow-up, seven out of 33 TNF-positive patients died (21.2%),and only three out of 96 TNF-negative patients had died (3.1%) (P
Detectable Serum Levels of Tumor Necrosis Factor Alpha May Predict Early Mortality in Elderlv Institutionalized Patients . I
Arshag D. Mooradian, MD* Richard L. Reed, MD,* Dan Osterweil, MD,t and Philip Scuderi, P h w
Objective: To determine if detectable serum tumor necrosis factor alpha (TNR levels are associated with higher mortality in nursing home residents. Subjects and methods: The basal serum concentrations of TNF and interleukin-2 alpha (IL-2)were measured in 129 elderly nursing home patients (mean age of 89 years), and survival in the cohort was monitored over a 13month period. Results: At 4 months follow-up, seven out of 33 patients with detectable serum TNF levels had died (21.2%), and only three out of 96 patients with undetectable serum TNF levels had died (3.2%) (PCO.001).The
difference in mortality remained significant up to 13 months of follow-up (PCO.05). Those with detectable serum TNF levels and those with undetectable levels were comparable in age, body mass index, hematocrit, lymphocyte counts, and serum lever of albumin, prealbwmin, and retinol-binding protein. When patients with detectable serum IL-1 levels were compared to those with undetectable levels, there were no significant differences in mortality over a 23-month period. Conclusion: Detectable serum TNF levels in elderly nursing home patients may be a predictor of early mortality. J Am Geriatr SOC395391494, 1991
dministration of tumor necrosis factor alpha nursing home patients who had had their serum TNF (TNF) to experimental animals and humans and interleukin-1 alpha (IL-1) measured. The results results in a myriad of metabolic and hemo- suggest that detectable serum levels of TNF, but not of dynamic changes.'-' However, the biolog- IL-1,may be a predictor of early mortality in nursing ical significance of serum TNF levels in human subjects home residents. is still not clear. Previous studies have found that aging is not associated with significant alterations in either MATERIALS AND METHODS basal serum levels of TNF9," or in TNF secretory Human Serum Samples A totaI of 129 nursing capacity of mononuclear cells in culture." We had also reported that elderly patients with detectable serum home residents with a mean age of 89.1 years f 5.0 TNF levels were more likely to have low serum con- were recruited from the Los Angeles Jewish Homes for centrations of 3,5,3' triiodothyronine.'2 These obser- the Aging. All the participants were residents of a vations suggest that detectable serum levels of TNF in skilled nursing home facility, and the average length elderly patients may have biological implications even of stay for the group before the study was 18 months. though they do not correlate with body weight loss?- Since glucocorticoids may suppress TNF production,L3 patients on systemic or topical glucocorticoid therapy l2 To test this hypothesis, we examined the death rate and cause of mortality in a cohort of 129 elderly were excluded. Blood specimens were collected in nonheparinized tubes, and sera were separated within an hour of blood collection to avoid artifactual changes in serum TNF measurement^.'^ For each patient, the folFrom the *ArizonaCenter on Aging, University of Arizona College lowing information was recorded: major diagnoses, of Medicine, Tucson, Arizona; and the tJewish Homes for the Aging, Los Angeles, California, and $Miles, Inc., Berkeley, California. number of prescription drugs, and body height and This work was supported in part by the Flinn Foundation. weight changes over a 6-month observation period. Address correspondence and reprint requests to Arshag D. MoorThe data pertaining to c l i c a l diagnoses and drugs adian, MD, Division of Endocrinology, St. Louis University Medical School, 1402 So. Grand Blvd., St. Louis, MO 63104. used were extracted from the patients' records. The
A
0 1991 by the American Geriatrics Society
892
MOORADIAN ET AL
laboratory evaluation included a complete blood count, serum albumin, cholesterol, prealbumin, and retinolbinding protein. A detailed account of the clinical characteristics and serum TNF levels of most of these patients (127 out of 129) has been previously reported? The results of thyroid function tests in some of these patients have also been previously reported." Over a 13-month follow-up period, survival in this group was monitored, and the circumstances of each death were observed. None of the patients involved in this study was discharged or lost to follow-up.
JAGS-SEPTEMBER 1991-VOL. 39,NO.9
between the two survival curves at 4 and 13 months after determination of TNF levels.
RESULTS Of the 129 patients tested, 33 had detectable serum levels of TNF while in 96 they were below 40 pg/mL (2 PM) (the lower limit of the assay detection). The mean serum TNF concentration in the group with detectable levels was 264.0 f 138 pg/mL (13.2 f 6.9 PM)with a range of 150 to 825 pg/mL (7.5-41.25 PM). The two groups were similar in mean age, body weight, TNF and IL-1 ELISA Systems An ELISA system height, hematocrit, concentration of hemoglobin, the specific for human TNF was used to detect serum white blood cell count and serum concentrations of levels. The details of this assay system have been cholesterol, albumin, pre-albumin, and retinol-binding previously p~b1ished.l~ In brief, 96-Well Immulon I1 protein (Table 1). Although the percentage of white plastic plates were coated with mAb 6E, a murine blood cells that were identified as lymphocytes in monoclonal antibody specific for human TNF. This patients with undetectable serum TNF levels (26.16% antibody was a gift from Genetech, Inc. (South San f: 9.51%) was somewhat lower than the percentage Francisco, CA). Sera were added to the wells and found in those with detectable serum TNF levels incubated for 1 hour at room temperature. The plates (31.27% k 14.02%),the total lymphocyte count in the were washed three times with PBS containing 0.05% two groups was not significantly different. In addition, Tween-20, and a rabbit anti-TNF reagent was subse- the two groups were similar in the prevalence rates of quently added. After a similar incubation and wash various diseases such as diabetes, dementia, thyroid procedure, a peroxidase-conjugated goat anti-rabbit disease, congestive heart failure, depression, cancer, renal failure, and urinary tract infections or bed sores. IgG was added to the wells. The peroxidase substrate The prevalence of these diseases in the two groups used was 2,2'-azino-bis-3-ethylbenzthiazoline-6-~~1fonic acid (Sigma, Inc.). Optical density readings were studied is indicated in a previous report.' The mean made on an ELISA reader set at 405 nm within 10 number of medical diagnoses in the group with detectminutes of substrate addition. The concentration of able serum TNF levels (4.13 f 0.8) and the number of TNF in sera was determined by comparison with a set medications used (9.5 f 4.8) were not different from of standards made with recombinant human TNF. the number of diagnoses (4.21 f 0.9) and the number (Genetech Inc., South San Francisco, CA). The lower of medications used (9.9 f 4.6) by those who did not limit of detection with this assay system is 40 pg/mL have detectable serum TNF levels. The types of drugs (2 PM).The interassay coefficient of variation was 4.8% used by these patients were also similar. at 300 pg/mL. The ELISA for human IL-1 alpha has been described TABLE 1. ELDERLY NURSING HOME PATIENTS previously" and was carried out using essentially the WITH DETECTABLE SERUM TNF LEVELS AND same protocol used for the TNF assay. In this assay THOSE WITH UNDETECTABLE LEVELS system, 10 p g of ascitic protein containing the IL-1 Patients Patients specific mAb C12 (Olympus, Inc., Lake Success, NY) Detectable TNF Undetectable TNF (n = 33) (n = 96) was used to coat each well of the assay plates. A rabbit anti-IL-1 alpha specific anti-serum was used as the 89.17 f 5.06 88.85 f 3.87 Age second reagent. The serum concentrationsof IL-1 alpha Sex M:F 6:27 15:81 54.24 f 10.78 53.63 f 10.30 were determined by comparison with a set of standards Body weight (Kg) 153.60f 13.55 151.17& 12.86 made by adding known amounts of recombinant hu- Height (Cm) 38.64 f 3.69 Hematomit (%) 38.06 f 4.19 man IL-1 alpha to serum. The lower limit of detection 12.71 f 1.15 12.50 f 1.35 Hemoglobin (g/dL) with this assay system is 50 pg/mL (2.8 PM). The White blood cell 7137.3 & 2290.1 7709.4 f 2760.2 interassay coefficient of variation was 5.1%at 300 pg/ count/mm3 26.16 & 9.59* mL. All the samples were measured in the same day Percent lymphocytes 31.27 f 14.02 (%I and in duplicates. 3.59 f 0.39 3.51 f 0.43 All the results are reported as mean f SD. Statistical Albumin (gm/dL) 210.3 f 41.7 Cholesterol (mg/dL) 201.4 f 56.7 analysis was done with the two-tailed Student's t test Prealbumin (mg/dL) 29.8 f 14.5 27.1 f 12.9 for unpaired variables with Bonferoni's correction. Pro- Retinol binding pro4.56 2 1.45 4.28 1.36 tein (mg/dL) portions were compared using the chi square test. A Kaplan-Meier analysis of survival was carried 0 ~ t . I ~ Mean & SD. The log-rank test'* was used to evaluate the differences * P < 0.05.
*
JAGS-SEPTEMBER 1991-VOL. 39, NO.9
-l
g, 85 m c c
n 75
O 1 ;
-
TUMOR NECROSIS FACTOR IN THE ELDERLY
LI
893
I
TNF Positive (11-33) TNF Negative (n-96)
; ; ; : ; ; ; ; ; ;0;1;*:3 Months After Initial Measurement
FIGURE 1. Percentagesof 33 elderly nursing home patients with detectable serum levels of TNF (TNF positive) and the percentages of 9 6 patients with undetectable serum levels (TNF negative) who survived during the 13-month followup. The log-rank statistic showed a P value of CO.001 at 4 months and (0.05 at 13 months of follow-up.
The survival analysis (Kaplan-Meier Estimates) of patients with detectable serum TNF levels (TNF-positive) and those with undetectable levels (TNF-negative) is shown in Figure 1. By 4 months of follow-up, seven out of 33 TNF-positive patients died (21.2%),and only three out of 96 TNF-negative patients had died (3.1%) (P
