Prealbumin is the best nutritional predictor of survival in hemodialysis and peritoneal dialysis

Prealbumin Is the Best Nutritional Predictor of Survival in Hemodialysis and Peritoneal Dialysis Rajanna Sreedhara, MD, Morrell Michael Avram, MD, Marta Blanco, MD, Rajesh Batish, MD, Mathew M. Avram, MD, and Neal Mittman, MD • Patients undergoing dialytic therapy for end-stage renal disease (ESRD) have greater morbidity and mortality than age-matched individuals with similar demographics in the general population. Risk factors for early death during treatment for ESRD include advanced age, diabetes, hypertension, and malnutrition. We questioned whether the level of serum prealbumin at the start of uremia therapy might serve as a marker of subsequent survival in patients treated with maintenance hemodialysis (HD) and peritoneal dialysis (PD). Study cohorts included 111 HD and 78 PD patients followed for up to 5 years. Selected demographic characteristics and biochemical variables were tested for correlation with survival in each cohort. Variables evaluated included age, race, gender, diabetic status, and serum concentrations of albumin, creatinine, cholesterol, and prealbumin. For comparison, expected survival was calculated with Cox proportional hazards analysis, which accounts for confounding variables. We found that a higher relative risk (RR) of death in HD patients correlated with older age, the diagnosis of diabetes, and a serum prealbumin < 30 mg/dL. In PD patients, older age and the presence of diabetes correlated with a higher RR of death than in the standard population. When nutritional variables were analyzed separately, prealbumin < 30 mg/dL was the strongest variable that predicted mortality in HD patients (RR = 2.64, P = 0.002) and also predicted increased risk of mortality in PD patients (RR = 1.8, P = 0.035). Observed and expected survival was significantly higher in patients with enrollment prealbumin greater than 30 mg/dL in both HD and PD. The serum prealbumin level correlated significantly with other measures of nutrition, including serum albumin, serum creatinine, and serum cholesterol, in both HD and PD patients. Among tested markers of nutritional status, prealbumin level appears to be the single best nutritional predictor of survival in ESRD patients. © 1996 by the National Kidney Foundation, Inc. INDEX WORDS: Hemodialysis; peritoneal dialysis; prealbumin; survival.

poor survival (comDISAPPOINTINGLY pared with registry reports from Europe and Japan) during treatment for end-stage renal disease (ESRD) in the United States has been attributed to various causes, including inadequate dialysis and malnutrition. According to the United States Renal Data System (USRDS), by the end of 1 year after the first 90 days of dialytic therapy, approximately 23.2% of ESRD patients die, a survival rate worse than that of patients with colon cancer, ~ Several lines of evidence indict malnutrition as a major contributor to this high mortality. 2-7 In an effort to discern those variables that might serve as markers for accelerated death, we explored a database containing demographic and biochemical information accumulated over the past 9 years of dialytic therapy at the Long Island College Hospital. Previously, we and others have reported that age, diabetes, and serum levels of visceral and somatic protein metabolism (albumin, cholesterol, and creatinine) correlate directly with survival in both hemodialysis (HD) and peritoneal dialysis (PD) patients. 5'8-1° Additionally, we previously reported a 7-year prospective study detailing the degree of association between multiple demographic and biochemical variables and survival of HD patients. ~ Serum prealbumin (recently termed transthy-

retin), a serum protein with rapid turnover and relatively short half-life, is a practical and sensitive marker of nutritional integrity] 2-~4 In a preliminary analysis of survival, we inferred that a serum prealbumin level of less than 30 mg/dL was associated with diminished survival at 36 to 48 months in both HD and PD patients. 9'1° Furthermore, the serum prealbumin level in HD patients had a greater correlation with other nutritional markers (serum cholesterol, serum apo-B, serum creatinine, and blood urea) than did serum From The Division of Nephrology, The Long Island College Hospital, Brooklyn, NE Received July 31, 1996; accepted in revised form August 17, 1996. Supported in part by the National Kidney Foundation of NY~J, the Nephrology Foundation of Brooklyn, and the National Kidney Foundation of NY/NJ Young Investigator Award (R.S.). Presented in part at the conference on Long Term Survival on Dialysis up to 30 Years, sponsored by The Avram Center for Kidney Diseases and The Long Island College Hospital, New York, NY, May 10, 1996. Address reprint requests to Morrell M. Avram, MD, FACP, Professor of Medicine, Chief Division of Nephrology, The Long Island College Hospital, Hicks Street at Atlantic Ave, Brooklym NY, 11201. © 1996 by the National Kidney Foundation, Inc. 0272-6386/96/2806-002253.00/0

American Journal of Kidney Diseases, Vol 28, No 6 (December), 1996: pp 937-942

937

938

S R E E D H A R A ET AL

albumin. 9 We now extend our analysis to compare the relative risks of mortality of prealbumin with that of other nutritional markers.

Table 1. Demographics of HD and PD Patients HD (n = 111) Variables

No.

PD (n = 78) %

No.

%

METHODS Patients Our study population consisted of 111 HD and 78 PD patients undergoing dialytic therapy between June 1991 and March 1992. Study subjects were monitored through May 1996. Demographic and biochemical data extracted from patient records included age, race, gender, diabetic status, cause of ESRD, and total months on dialysis at the time of enrollment. A fresh biochemical profile was obtained at enrollment as described below. Follow-up was censored at death, dialysis modality switch, renal transplantation, or transfer to another dialysis facility. Patients with known seropositivity to human immunodeficiency virus were excluded from the study.

Dialysis Prescription Hemodialysis patients were treated on Cobe Centry System 3 machines with bicarbonate-based dialysate and volumetric ultrafiltration control. Cellulose-based membranes were used for more than 98% of all treatments (predominantly cellulose acetate and cellulose triacetate). Dialyzers were not reused. The targeted urea reduction ratio (URR) was > 60% and validated by monthly blood sampling performed on the first treatment of the week (ie, 68 hours after the preceding dialysis). Most PD patients used four 2-liter exchanges per day, although the volume and frequency were adjusted when necessary based on patient size, peritoneal equilibration characteristics, and clinical condition of the patient. Weekly creatinine clearance was initially estimated by extrapolating the 4-hour peritoneal equilibration test (PET), and adjustments were made to maintain a creatinine clearance above 50 L/ wk/1.7 m 2. Equilibration characteristics were monitored semiannually with the PET and more often if clinically necessary. During the last 2.5 years of the study, actual urea and creatinine clearances were obtained by collecting urine and peritoneal fluid for 24 hours.

Biochemical Data On enrollment, nonfasting baseline blood samples were drawn before the initiation of dialysis. Multiphasic biochemistry screen including total cholesterol was obtained for each patient using the SMAC autoanalyzer (Technicon, Tarrytown, NY). Prealbumin was measured by rate nephelometry on a Beckman Array Protein System (Beckman Instruments Inc, Brea, CA). For each HD patient, serial monthly predialysis and 5-minute postdialysis blood urea nitrogen was measured, from which URR was calculated.

Statistical Analysis Results are reported as mean (_+SD) unless otherwise specified. Multivariate repeated measures ANOVA and chi-square analysis were used for comparisons between groups. Observed patient survival between groups of patients was analyzed by the Kaplan-Meier life-table method. Independent

Age (yr) Mean _+ SD Range Race White Black Hispanic Sex Male Femate Diabetes Yes No Prior months on dialysis Mean_+SD Range

59.9 _+ 15.5 18.8-89.1

54.0 _+ 15.9 27.2-84,9

30 61 20

27 55 18

18 41 19

23 53 24

54 57

49 51

34 44

44 56

42 69

38 62

28 50

36 64

56_+51 0-274

78_+ 11 0-81

predictors of mortality were determined using the Cox proportional hazards model, which allows for adjustment of confounding variables. In cross-sectional analyses, Spearman rank coefficients were used to determine the correlation between variables. Calculations were performed using SPSS Windows version 7.0 statistical software (SPSS Inc, Chicago, IL).

RESULTS

Listed in Table 1 are the demographic characteristics of HD and PD study patients. Among HD patients, the cause of ESRD was hypertension in 36%; diabetes, 34%; polycystic kidney disease, 8%; glomerulonephfitis, 7%; and other renal disorders, 15%. Mean patient survival after enrollment in the study was 2.91 years. The causes of ESRD in PD patients were similar (data not shown). In Table 2, cohorts of HD and PD patients are sorted by various nutritional indices, including serum albumin, prealbumin, creatinine, and cholesterol. When enrollment biochemical indices of nutrition were analyzed separately, a serum prealbumin level < 30 mg/dL showed the greatest correlation with increased mortality risk (RR = 2.64, P = 0.002) in HD patients (Table 3). Even among PD patients, enrollment prealbumin < 30 mg/dL imparted a mortality risk of 1.8 compared with patients with prealbumin > 30 mg/dL (P = 0.035). A creatinine enrollment level < 9 mg/dL also predicted significantly higher mortality risk in both HD and PD patients; whereas an albumin enrollment level < 3.5 g/dL

939

PREALBUMIN PREDICTS SURVIVAL IN DIALYSIS

Table 2. Biochemical Profile of HD and PD Patients HD (n = 111)

Table 4. Predictors of Mortality Up t o 5 Years in Hemodialysis (N = 111)

PD (n = 78) Variable

Variables

Albumin (g/dL) < 3.5 -> 3.5 Creatinine (mg/dL) < 9 -> 9 Cholesterol (mg/dL) < 200 -> 200 Prealbumin (mg/dL) < 30 >- 30

No

%

No

Relative Risk

P

%

15 96

14 86

30 48

38 62

22 89

20 80

20 58

26 74

84 27

76 24

30 48

38 62

74 37

67 33

25 53

32 68

predicted excess mortality in PD patients only (Table 3). However, patients with enrollment cholesterol < 200 mg/dL did not have any significant difference in relative risk of mortality compared with patients with cholesterol > 200 mg/dL. Employing the Cox proportional hazards analysis, in which adjustments are made for confounding variables, age (RR = 1.03, P = 0.004), diabetic status (RR -- 1.77, P = 0.05), and prealbumin < 30 mg/dL (RR = 2.32, P = 0.011) were predictive of higher mortality risk in HD patients. In this model, neither albumin < 3.5 g/ dL nor creatinine < 9 mg/dL were able to predict a significantly higher mortality risk in HD patients (Table 4). Observed and expected survival of HD patients by enrollment prealbumin level

Table 3. Predictive Value of Enrollment Levels of Albumin, Prealbumin, and Creatinine in HD and PD Patients

Prealbumin (mg/dL) < 30 v -> 30 Diabetes (yes v no) Creatinine (mg/dL) < 9 v > 9 Total cholesterol (mg/dL) < 200 v -> 200 Albumin (g/d L) < 3.5 v -> 3.5 Age (yr) Prior months on dialysis

2.32 1.77

0.011 0.050

1.58

0,225

1.20

0,585

1.11 1.03 1.005

0.764 0.004 0.103

NOTE. Data obtained by Cox proportional hazards analysis.

was significantly higher in patients with enrollment prealbumin greater than 30 mg/dL (Fig 1). When Cox analysis was performed in PD patients, only age (RR = 1.05, P = 0.001) and the presence of diabetes (RR = 2.59, P -- 0.006) emerged as significant variables that predicted excess mortality risk (Table 5). Prealbumin level did not predict mortality risk in PD patients in the Cox analysis. However, observed survival of PD patients with enrollment prealbumin level > 30 mg/dL was significantly higher, as shown in Fig 2. Prealbumin correlated positively and significantly with albumin, creafinine, and cholesterol in both HD and PD patients (Table 6). In addition, in HD patients prealbumin was also negatively correlated with age and positively with apo-B level. The level of prealbumin among diabetics was significantly lower than that in nondiabetics in both HD and PD patients (Table 7). DISCUSSION

Variable Albumin (g/dL) < 3.5 v -> 3.5 Creatinine (mg/dL) < 9 v -> 9 Total cholesterol (mg/dL) < 200 v -> 200 Prealbumin (mg/dL) < 30 v -> 30

HD(N = 1 1 1 )

PD(N-78)

RR

P

RR

1.47

0.26

2.1

0.019

2.0

0.035

2,7

0.003

1.73

0.073

0.98

0,96

2.64

0.002

1.8

0.035

Abbreviation: RR, relative risk of death.

P

Protein-energy malnutrition is present in a large proportion of dialysis patients. 6'7 In the multicenter Canadian Hemodialysis Morbidity Study, Churchill et al 8 noted a direct correlation between serum albumin concentration and morbidity and mortality. 8 Confirmatory evidence is present in USRDS reports, which consistently find that a low serum albumin level at initiation of dialytic therapy predicts excess mortality. 1 Teehan et a115 found that continuous ambulatory peritoneal dialysis (CAPD) patients who were

940

SREEDHARA ET AL

CumulativeSurvival(%) 10n_ 100.CumulativeSurvival(%) 90N ~ 3 . 1 ~ " ' l . . I

80,

~'

I

P<0.001

P
I

70,

6080~ . l b ~ l , > = 3 0 m g / d L

=30mgldL

4DI

<30mg/dL

30, ~

i

i i

20, A

20

B

Years of Followup

Years of

Followup

Fig 1. (A) Observed patient survival by enrollment level of prealbumin in 111 hemodialysis patients monitored for up to 5 years. (B) Expected patient survival in the same patient cohort by enrollment level of prealbumin after adjustments were made for confounding variables such as age, diabetes, albumin, creatinine, and prior months on dialysis.

not hospitalized had a higher normalized protein catabolic rate than those who were hospitalized (1.06 + 0.14 g/kg/d v 0.91 _+ 0.16 g/kg/d). 15 Based on the foregoing, the serum albumin concentration has been accepted as a valid marker of nutritional status and a predictor of subsequent survival of ESRD patients sustained by dialytic therapy. We and others have previously reported the predictive value for survival of several nutritional indices including serum albumin, serum creatinine, serum cholesterol, and other markers. 2-7 We have also previously demonstrated the predictive

value of serum prealbumin level in HD and PD patients for treatment duration of less than 3 years. 5'1° In the current report, we extend the predictive value of single measurements of serum creatinine and serum prealbumin at enrollment for patient survival for up to 5 years in both HD and PD patients (Table 3, Figs 1, 2). Noteworthy in our results is the finding that in HD patients, a serum prealbumin < 30 rag/ dL had the highest predictive power of mortality for up to 5 years with a relative risk of 2.64.

10(1.CumulativeSurvival(%) Table 5. Predictors of Mortality Up to 5 Years in Peritoneal Dialysis (N = 78)

P<0.03

Variable

Relative Risk

P

Diabetes (yes v no) Albumin (g/dL) < 3.5 v -> 3.5 Creatinine (mg/dL) < 9 v ~- 9 Age (yr) Cholesterol (mg/dL) < 200 v -> 200 Prior months on dialysis Prealbumin (mg/dL) < 30 v -> 30

2.59

0.006

1.82

0.130

1.39 1.05

0.395 0.001

1.00 0.98

0.997 0.404

0.71

0.447

60.

'°120

=30mg/dk (N=53)

, = ,I Years

NOTE. Data obtained by Cox proportional hazards analysis.

I ,

ofFollowup

Fig 2. Observed patient survival by enrollment level of prealbumin in 78 peritoneal dialysis patients monitored for up to 5 years.

PREALBUMIN PREDICTS SURVIVAL IN DIALYSIS

In this patient cohort, a low enrollment serum albumin level failed to correlate with 5-year survival. Likewise, in PD patients, low enrollment levels of serum creatinine, serum albumin, and serum prealbumin predicted significantly increased risk of mortality at 5 years. Applying the Cox proportional hazards analysis, which accounts for confounding variables, age, diabetes, and serum prealbumin < 30 mg/ dL significantly predicted survival up to 5 years (Table 4). Neither serum albumin nor serum creatinine were significant predictors of mortality up to 5 years. The Cox analysis of PD patients showed that only increasing age and diabetes were significant predictors of mortality up to 5 years (Table 5). Thus, in both HD and PD, the best nutritional marker for shortened survival is a low serum prealbumin level at the time of initiation of dialytic therapy. The inability of traditional nutritional markers such as serum albumin, serum creatinine, and serum cholesterol to predict long-term survival may be explained by their strong association with serum prealbumin (Table 6). Therefore, when serum prealbumin is added into the Cox model, the other variables lose their significance. We caution that serum prealbumin has a relatively short half-life, and therefore the ability of enrollment level of prealbumin to predict survival up to 5 years in ESRD patients cannot be directly attributed its ability to reflect momentary nutritional status. Conversely, the significant correlation of serum prealbumin with both somatic and visceral proteins, as well as with advancing age, could imply that patients with higher serum prealbumin levels are in a better nutritional state than those with low serum prealbumin levels. We believe that measurement of serum prealbumin concentration is a reliable method of identifying those ESRD patients at risk of reduced survival

Table 6. Correlations of Prealbumin With Age and Other Nutritional Markers Variables Age (yr) Albumin (g/d L) Creatinine (mg/dL) Cholesterol (mg/dL) * P < 0.01.

HD (n = 111)

PD (n = 78)

-0.21" 0,42* 0.53* 0.42*

-0.16 0.50* 0.32* 0.39*

941

Table 7. Prealbumin in Dialysis Patients by Diabetic Status

Prealbumin (mg/dL) HD

PD

Diabetic

Nondiabetic

P

25.3 _+ 6.4 (42) 31.4 _+ 7.5 (28)

27.9 +_ 5.8 (69) 35,3 _+ 9,6 (50)

0,028 0.052

NOTE. Data shown as mean + SD (number of patients

shown in parentheses).

on dialysis. One intriguing prospective study yet to be performed is to confirm the hypothesis that serial measurements of serum prealbumin may serve as a simple means to plot the course of response to nutritional intervention by malnourished ESRD patients. 16 An interesting finding in this study is that patients with diabetes have decreased levels of serum prealbumin. Whether this decrease reflects changes induced by metabolic disturbances of diabetes or results from the myriad of comorbid conditions found in diabetes such as gastroparesis and gastrointestinal neuropathy is unclear. Because both diabetes and serum prealbumin were significant predictors of excess mortality in the Cox model of HD patients, the poor survival among diabetics cannot be attributed solely to decreased levels of serum prealbumin (Table 4). Nevertheless, among PD patients, diabetes strongly predicted excess mortality, whereas serum prealbumin did not predict excess mortality in the Cox model (Table 5). In summary, we have shown that in ESRD patients subjected to dialytic therapy, the level of serum prealbumin at enrollment is the single strongest predictor of 5-year survival in HD patients. Serum prealbumin also significantly predicts survival at 5 years in PD patients. Serial measurement of serum prealbumin levels may therefore be included in the design of interventions to improve nutritional status in ESRD patients. In addition, the serum prealbumin level may be useful to identify patients at risk of nearterm death, signaling the need for appropriate intervention. ACKNOWLEDGMENT

The authors thank Joseph Feldman, PhD, fox"statistical data interpretationand Jyotiprakash Chattopadhyay,PhD, for scientific assistance.

942

SREEDHARA ET AL REFERENCES

1. U.S. Renal Data System: USRDS 1995 Annual Data Report. Bethesda, MD, The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, 1995 2. Acchiardo SR, Moore LW, Latour PA: Malnutrition as the main factor in morbidity and mortality of dialysis patients. Kidney Int 24:S119-$203, 1983 (suppl) 3. Lowrie EG, Lew NL: Death risk in hemodialysis: The predictive value of commonly measured variables and an evaluation of death rate differences between facilities. Am J Kidney Dis 15:458-482, 1990 4. Winchester JF: The albumin dilemma. Am J Kidney Dis 20:76-77, 1992 5. Goldwasser P, Mittman N, Antignani A, Burrell D, Michel MA, Collier J, Avram MM: Predictors of mortality in hemodialysis patients. J Am Soc Nephrol 3:1613-1622, 1993 6. Kopple JD: Effect of nutrition on morbidity and mortality in maintenance dialysis patients. Am J Kidney Dis 24:1002-1009, 1994 7. Bergstrom J: Nutrition and mortality in hemodialysis [editorial]. J Am Soc Nephrol 6:1329-1341, 1995 8. Churchill DN, Taylor DW, Cook RJ, LaPlante P, Barre P, Cartier P, Fay WP, Goldstein MB, Jindal K, Mandin H, McKenzie JK, Muirhead N, Parfrey PS, Posen GA, Slaughter D, Ulah RA, Werb R: Canadian hemodialysis morbidity study. Am J Kidney Dis 3:214-234, 1992

9. Goldwasser P, Michel M-A, Collier J, Mittman N, Fein P, Gusik S-A, Avram MM: Prealbumin and lipoprotein(a) in hemodialysis: Relationships with patient and vascular access survival. Am J Kidney Dis 22:215-225, 1993 10. Avram MM, Goldwasser P, Erroa M, Fein PA: Predictors of survival in continuous ambulatory peritoneal dialysis patients: The importance of prealbumin and other nutritional and metabolic markers. Am J Kidney Dis 23:91-98, 1994 11. Avram MM, Mittman N, Bonomini L, Chattopadhyay J, Fein PA: Markers for survival in dialysis: A seven year prospective study. Am J Kidney Dis 26:209-219, 1995 12. Ingenbleek Y, deVisscher M, deNayer P: Measurement of prealbumin as an index of protein-calorie malnutrition. Lancet 2:106-108, 1972 13. Tuten MB, Wogt S, Dasse F, Lieder Z: Utilization of prealbumin as a nutritional parameter. J Parenter Enteral Nutr 9:709-711, 1985 14. Ingenbleek Y, Young V: Transthyretin(prealbumin) in health and disease: Nutritional implications. Annu Rev Nutr 14:495-533, 1994 15. Teehan BP, Schleifer CR, Brown JM, Sigler MH, Raimondo J: Urea kinetic analysis and clinical outcome on CAPD: A five year longitudinal study. Adv Perit Dial 6:181185, 1990 16. Wolfson M, Strong C: Assessment of nutritional status in dialysis patients. Adv Renal Replacement Therapy 3:174179, 1996