Association of morbidity with markers of nutrition and inflammation in chronic hemodialysis patients: A prospective study

Transcription

1 Kidney International, Vol. 55 (1999), pp Association of morbidity with markers of nutrition and inflammation in chronic hemodialysis patients: A prospective study T. ALP IKIZLER, REBECCA L. WINGARD, JANICE HARVELL, YU SHYR, and RAYMOND M. HAKIM Department of Medicine, Division of Nephrology and Department of Biostatistics, Vanderbilt University Medical Center, Nashville, Tennessee, USA Association of morbidity with markers of nutrition and inof was 7% higher (relative risk 1.07) for a CRP concentration flammation in chronic hemodialysis patients: A prospective 0.92 mg/dl and was 30% (relative risk 1.30) higher for a study. CRP concentration of 3.4 mg/dl. When a reactance value of Background. Numerous studies suggest a strong association 70 ohms was considered as a reference range with a relative between nutrition and clinical outcome in chronic hemodialysis risk of 1.0, the relative risk of hospitalization increased to 1.09 (CHD) patients. Nevertheless, the pathophysiological link befor for a reactance value of 43 ohms and further increased to 1.14 tween malnutrition and morbidity remains to be clarified. In a reactance value of 31 ohms. addition, recent evidence suggests that nutritional indices may Conclusions. The results of this study strongly indicate that reflect an inflammatory response, as well as protein calorie both nutritional status and inflammatory response are indepen- malnutrition. In this study, we prospectively assessed the relareactance dent predictors of hospitalization in CHD patients. CRP and tive importance of markers of nutritional status and inflammation. values by BIA are reliable indicators of hospitaliza- tory response as determinants of hospitalization in CHD patransferrin Visceral proteins such as serum albumin, prealbumin, and tients. are influenced by inflammation when predicting Methods. The study consisted of serial measurements of conpitalizations hospitalization. When short-term clinical outcomes such as hoscentrations of serum albumin, creatinine, transferrin, prealbunutrition are considered, markers of both inflammation and min, C-reactive protein (CRP), and reactance values by bioelectrical should be evaluated. impedance analysis (BIA) as an indirect measure of lean body mass every 3 months over a period of 15 months in 73 CHD patients. Outcome was determined by hospitalizations over the subsequent three months following each collection of The mortality and morbidity rate of treated end-stage data. renal disease (ESRD) patients remain high [1, 2]. In the Results. Patients who required hospitalization in the three United States, the life expectancy of ESRD patients is months following each of the measurement sets had signifi- 20 to 25 years less than the normal age-sex-race matched cantly different values for all parameters than patients who U.S. population over the age of 45. Despite recent adwere not hospitalized. Thus, serum albumin ( vs. vances in our understanding of the uremic state and g/dl), serum creatinine ( vs mg/dl), serum transferrin ( vs mg/dl), serum improvements in the science and technology of renal prealbumin ( vs mg/dl), and reactance replacement therapy, the prognosis of this patient popu- ( vs ohms) were higher for patients lation remains poor. not hospitalized, whereas CRP ( vs Among many factors, including treatment characterismg/dl) was lower in patients who were not hospitalized. All tics and comorbid conditions, protein and calorie malnudifferences were statistically significant (P 0.05 for all paramtrition has been shown to be a major risk factor for eters). When multivariate analysis was performed, serum CRP and reactance values were the only statistically significant predictors increased mortality in the chronic hemodialysis (CHD) of hospitalization (P 0.05 for both). When a serum patient population [3 6]. Protein and calorie malnutriincreased CRP concentration of 0.12 mg/dl was considered as a reference tion can be characterized by an insidious loss of somatic range (relative risk 1.0), the relative risk for hospitalization proteins, reflected by a decrease in lean body mass and serum creatinine concentration, as well as by visceral Key words: malnutrition, C-reative protein, inflammatory response, proteins, as assessed by serum albumin, prealbumin, and end-stage renal disease, uremia, bioelectrical impedance analysis. transferrin concentrations. Using these indices, several prospective and retrospective studies have suggested Received for publication January 26, 1998 and in revised form November 3, 1998 that protein and calorie malnutrition is a significant de- Accepted for publication December 17, 1998 terminant of clinical outcomes in chronic dialysis patients 1999 by the International Society of Nephrology [7 14]. 1945

2 1946 Ikizler et al: Morbidity and markers of nutrition Table 1. Demographic characteristics of the sion criteria included CHD for at least three months and study patients (N 73) the ability to sign an informed consent form. The study Gender (M/F) 53% M/47% F period was between February 1, 1996, through May 31, Race 27% White/73% African American Age years , and included all consenting patients surveyed by Cause of ESRD 30% Diabetes May 31, The overall participation rate into the 46% Hypertension study was approximately 60%. Patients were censored 4% Glomerulonephritis from the study if death or transplantation occurred or 4% PKD 16% Unknown if they were either transferred to another facility or Abbreviations are: M, male; F, female; PKD, polycystic kidney disease. switched to peritoneal dialysis. The study protocol was approved by the institutional review board, and a written informed consent form was obtained from patients. Despite the numerous studies suggesting the strong association between malnutrition and mortality in CHD patients, the pathophysiological link between these two conditions remains to be clarified. Complicating this linkage is the fact that serum albumin, the most common and usually the singular parameter used for assessment of nutritional status, and more recent nutritional indices such as serum prealbumin and transferrin are also acutephase reactants and their serum concentrations are profoundly affected by the presence of an inflammatory response. Therefore, it is not clear whether clinical outcome in CHD patients associated with these markers is a reflection of nutritional status or inflammatory response. Indeed, recent studies suggest that a combined state of poor nutritional status and inflammation predisposes ESRD patients to poor clinical outcome [15]. Another limitation of the available studies on the association of nutrition with patient outcome is that these studies are, in general, a cross-sectional design. Longitudinal studies evaluating the clinical significance of nutritional parameters over time in CHD are lacking. Finally, the majority of the studies on clinical outcomes are limited to mortality, and studies on morbidity (that is, hospitalization) are few. In this prospective study, we evaluated the association of several well-defined markers of nutritional status, as well as a marker of inflammation with hospitalization over a period of 15 months in a stable CHD patient population. Our specific aim was to evaluate the relative importance of nutritional status and inflammatory response as determinants of hospitalization in CHD patients. METHODS Patient characteristics A total of 73 patients on CHD were included in the study. The demographic characteristics of the study population are depicted in Table 1. In general, the study population differed little from the average characteristics of the CHD population reported by the United States Renal Data System. All patients dialyzed at the Vanderbilt University Medical Center Outpatient Hemodialysis Center were asked to participate in the study. The inclu- Study design The study was a prospective cohort design. There were no interventions specific to the study, and dialysis pre- scriptions were determined by the attending nephrolo- gist. The study consisted of serial measurements of study parameters every 3 months over a period of 15 months. The study parameters were serum albumin, creatinine, transferrin, prealbumin, and C-reactive protein (CRP) as biochemical parameters, as well as reactance values by bioelectrical impedance analysis (BIA) as an indirect measure of lean body mass. Sample collection was completed within the first week of each month of data analysis. Blood samples were collected predialysis in a nonfasting state from each study patient by the study coordinator. On the day of the sample collection, patients underwent BIA at approximately 30-minutes postdialysis. During the study period, the study coordinator moni- tored all morbid events for all study patients. Specific causes of hospitalization (grouped as infectious causes, cardiac causes, vascular access-related causes, and other causes classified according to ICD-9 codes) were also monitored. All hospitalizations were recorded for each three-month interval following the data collection. In an attempt to find the predictive value of these parameters in terms of near-term morbidity, analysis with regard to hospitalization was limited to the subsequent three months following each data collection. Of note, the final analysis does not include vascular access-related hospi- talization. Mortality data were also collected during the study period. The overall mortality of the study patients was 9.6% (seven deaths in 73 patients) during the 15-month study period, an annual mortality rate of 7.7%. Because this rate was insignificant for any meaningful statistical analysis, mortality was not analyzed as an outcome mea- sure in this study. All patients underwent formal urea kinetic modeling once a month during the study period. The minimum dialysis dose practiced at the dialysis unit was Kt/V greater than or equal to 1.4 (single pool). All patients were dialyzed with biocompatible membranes. Study patients calcium, phosphorus, total bicarbonate, and he-

3 Ikizler et al: Morbidity and markers of nutrition 1947 Statistical analysis Demographic information and descriptive statistics are presented in table form. Data were transformed using appropriate transformations such as the log or square root. Tests of hypotheses concerning comparisons were completed using the paired t-test or the Wilcoxon rank sum test. For both count and binary multiple time points data, the generalized estimating equation (GEE) method sta- tistical procedure was used [17, 18]. This method adjusts for an intracorrelation effect for the patients with multi- ple hospitalizations and is a repeated-measures analysis for correlated dichotomous outcomes (hospitalization yes or no) and a set of predictors. Specifically, multiple measures over time for each study subject were taken into account and adjusted by this method. The possible two-way interactions such as the interaction between serum albumin and CRP, the interaction between serum albumin and serum prealbumin, and the interaction between serum creatinine and reactance were included in the full statistical model for testing the statistical signifi- cance of interactions. None of the two-way interaction terms reached the 5% statistically significant level in the full statistical model; therefore, the final statistical model in this article does not include such interaction terms. Data are presented as mean sd. Differences were considered statistically significant when P was less than All statistical analyses were completed using SAS version RESULTS matocrit levels were included in regular patient monthly laboratory reports. Every effort was made to keep laboratory parameters within acceptable ranges through established protocols. Recombinant human erythropoietin was prescribed to keep hematocrit within the range of 33 to 36%. All measurements were done at the clinical and special chemistry laboratories at Vanderbilt University Medical Center. Serum albumin was analyzed using bromcresol green technique. Serum prealbumin was analyzed by an antigen antibody complex assay, and serum transferrin was analyzed by turbidimetric reading (Hitachi 717). CRP was measured using nephelometric analysis. Bioelectrical impedance analysis was done using a hand- held device (Quantum, BIA 101Q; RJL Systems, Clinton Twp., MI, USA). All tests were done 30 minutes follow- ing the termination of hemodialysis. In brief, subjects were placed in a supine position with their arms at but not touching their sides and with their legs separated. Electrodes were attached to their right hand and foot, and a high-frequency, alternating low voltage (approximately 1 khz) was passed across the limbs; the current passing through the body was measured. From this measurement, reactance and resistance values were measured. The validity of BIA in CHD patients has been proven using deuterium oxide and sodium bromide isotope dilution studies [16]. These values allowed the calculation of body composition, including lean body mass according to standard equations from the manufacturer. In this study, reactance values were used as a marker of metabolically active body compartment, that is, body cell mass or lean body mass, to eliminate any potential flaws associated with equations that use normal population adjustments. The mean values for the biochemical parameters that were followed during the study period are depicted in Table 2. The overall 15-month mean concentration for serum albumin was g/dl (range 1.40 to 4.80 g/dl), for serum creatinine mg/dl (range 1.70 to 24.3 mg/dl), for serum prealbumin mg/dl (range 8.8 to 59.0 mg/dl), for serum transferrin mg/dl (range 107 to 347 mg/dl), and for serum CRP mg/dl (range 0.09 to 11.8 mg/dl). The reactance values obtained by BIA are also depicted in Table 2. The overall mean reactance value was ohms (range 14 to 95 ohms). During the study period (3 months following each data collection for a total of 15 months), a total of 84 nonvascular access-related hospitalization episodes were recorded. Twenty-four out of 73 patients were not hospi- talized during the study. Of those 84 hospitalizations within each three-month period, 25 patients were hospi- talized only once. Fourteen were hospitalized twice, and 10 were hospitalized three times or more. For any three- month period, only the initial hospitalizations following data collection were used for analysis. The median length of stay was five days for the hospitalizations (range 1 to 59 days). Because the primary outcome of the study was hospitalization, the study parameters were analyzed according to subsequent hospitalizations over the three months following the collection of data. Specifically, the mean ( sd) of the study parameters in patients without subse- quent hospitalization was compared with the mean ( sd) of the study parameters in patients with subsequent hospitalization within the following three months. Table 2 shows the mean values for each study parameter grouped according to whether patients were hospitalized or not hospitalized in the subsequent three months. As can be seen, all study parameters were statistically significantly different between the two groups. Of the biochemical nutritional markers, mean serum albumin concentration was g/dl for patients who did not have a subsequent hospitalization as compared with g/dl for patients who had a hospitalization episode (P 0.05). The mean serum creatinine concentration was mg/dl for patients who did not have a subse- quent hospitalization as compared with mg/dl

4 1948 Ikizler et al: Morbidity and markers of nutrition Table 2. The mean values for each study parameter grouped according to patients with subsequent hospitalization or patients with no hospitalization CRP SAlb S Cr S Prealb S Transferrin Reactance mg/dl g/dl mg/dl ohms All (12-month average) Hospitalization c a b a b c No hospitalization Abbreviations are: CRP, C-reative protein, S alb, serum albumin; S Cr, serum creatinine; S Prealb, serum prealbumin; S Tranferrin, serum transferrin. a P 0.05 vs. No hospitalization b P 0.01 vs. No hospitalization c P vs. No hospitalization Fig. 1. The percentage of hospitalized patients for each quartile of serum albumin concentrations (P 0.05 by regression analysis). Fig. 2. The percentage of hospitalized patients for each quartile of reactance values (P by regression analysis). for patients who had a hospitalization episode (P 0.005). The mean serum prealbumin concentration was mg/dl for nonhospitalized patients versus mg/dl for hospitalized patients (P 0.05). The mean serum transferrin concentration was mg/dl for patients who did not have a subsequent hospitalization as compared with mg/dl for patients who had a hospitalization episode (P 0.009). The reactance values as marker for lean body mass were ohms for patients who did not have a subsequent hospitalization compared with ohms for patients who had a hospitalization episode (P ). Serum CRP concentration as a marker of inflamma- tion also showed significant difference between the two groups. Specifically, the mean CRP concentration was mg/dl for patients without hospitalization, whereas it was mg/dl for patients with a hospitalization episode (P ). Age, race, and gender were not statistically associated with hospitalizations. We further analyzed the data in quartiles with regard to percentage of patients hospitalized in each quartile. Figures 1, 2, and 3 depict this analysis for different ranges of serum albumin, reactance, and CRP, respectively. Specifically, in the group with serum albumin concentrations in the lowest quartile (serum albumin less than 3.7 Fig. 3. The percentage of hospitalized patients for each quartile of serum C-reactive protein concentrations (P by regression analysis). g/dl), 43% of patients were hospitalized, whereas only 17% of the patients were hospitalized in the group with serum albumin concentrations in the highest quartile (serum albumin more than 4.1 g/dl; Fig. 1). Reactance values also showed a similar trend with patients in the lowest quartile (less than 38 ohms), having a 33% rate of hospitalization, whereas only 16% of the patients were hospitalized in the highest quartile (higher than 57 ohms).

5 Ikizler et al: Morbidity and markers of nutrition 1949 Fig. 4. The estimation of relative risk of hospitalization within the subsequent three months of data collection with regard to serum C-reactive protein concentrations. A serum concentration of 0.12 mg/ dl is considered as the reference range (relative risk 1.0). Data are controlled for all study parameters. Fig. 5. The estimation of relative risk of hospitalization within the subsequent three months of data collection with regard to reactance values. A measurement of 70 ohms is considered as the reference range (relative risk 1.0). Data are controlled for all study parameters. risk of 1.0 (average of reactance values for 75th to 100th On the other hand, in the lowest CRP quartile (less than percentiles), the adjusted relative risk of hospitalization 0.2 mg/dl), only 15% of the patients were hospitalized, increased to 1.06 for a reactance value of 53 ohms (averwhereas in the highest quartile (higher than 1.4 mg/dl), age of reactance values for 50th to 75th percentile) to 48% of the patients were hospitalized for a reactance value of 43 ohms (average of re- Because all of the study parameters showed significant actance values for 25th to 50th percentile) and further differences with regard to predicting hospitalization by increased to 1.14 for a reactance value of 31 ohms (averunivariate analysis, a multivariate analysis was performed age of reactance values for 0 to 25th percentile). to differentiate each parameter s predictive power. When We further repeated multivariate analysis removing all of the studied parameters were included within the CRP and reactance from the analyzed parameters. Semodel, serum CRP and reactance values were the only rum transferrin was the only parameter that reached statistically significant predictors of hospitalization (P statistical significance in the analysis (P 0.045). None 0.05 for both). None of the other variables that were of the other parameters, including serum albumin, serum significant by univariate analysis was found to be signifi- creatinine, and serum prealbumin, were statistically sigcant by multivariate analysis. nificantly associated with hospitalization. In order to evaluate the adjusted relative risk of hospitalization with regard to different concentrations of CRP, we grouped CRP concentrations according to quartiles. DISCUSSION Figure 4 depicts a diagram of the adjusted ratio of rela- This prospective study was performed in order to evaluate tive risk of hospitalization at several concentrations of several clinically relevant issues related to nutrition serum CRP. When the serum CRP concentration of 0.12 that were not studied systematically in previous studies. mg/dl (average of CRP values for 0 to 25th percentile) First, we attempted to define the association between was considered as a reference range (relative risk of 1.0), nutritional status and hospitalization as an important there was an increase in the risk of hospitalization for measure of morbidity. We also evaluated the relationship increasing serum CRP concentrations. The relative risk between inflammatory response and nutritional status for hospitalization was 2% higher (relative risk 1.02) and its predictive power for hospitalization. The results for a CRP concentration of 0.38 mg/dl (average of CRP of this study strongly indicate that both inflammation values for 25th to 50th percentile), 7% higher (relative assessed by CRP concentrations and nutritional status risk 1.07) for CRP concentration of 1.4 mg/dl (average indicated by reactance values as a marker of lean body of CRP values for 50th to 75th percentile), and 30% mass are independent predictors of hospitalization in (relative risk 1.30) higher for a CRP concentration CHD patients. of 3.4 mg/dl (average of CRP values for 75th to 100th A series of studies has suggested that nutritional status percentile). correlates with clinical outcome in CHD patients [5, 6, Figure 5 shows a similar analysis for reactance values ]; however, many of these studies have drawbacks. Specifically, when a reactance value of 70 ohms was An important consideration is the completeness of the considered as a reference range and assigned a relative nutritional markers used for assessment of nutritional

6 1950 Ikizler et al: Morbidity and markers of nutrition multivariate analysis, which included CRP in the model. This is an important finding when one considers that serum albumin, serum prealbumin, and serum trans- ferrin are all proposed to be acute-phase reactants and acute illnesses can trigger an abrupt decrease in their concentrations independent of changes in nutrient intake [15, 33 35]. This implies that these markers are indeed influenced by inflammation and that in this particular group of CHD patients, they may be considered as mark- ers of inflammatory response and not just nutritional status. Although this suggests that there may be a biolog- ical interaction between these parameters and CRP, we were not able find any statistically significant interactions between these parameters in our multivariate analysis. This may be secondary to the characteristics of our pa- tient population, as well as the relatively small sample size studied. Morbidity was used as the primary outcome measure in this study, which also distinguishes it from the other available studies in the literature. Although mortality is clearly the ultimate end point for any clinical outcomes study, the morbidity in CHD patients represents a major health expenditure in the United States as well as in other countries. In fact, the health care cost of treating the U.S. ESRD program exceeds $10 billion annually, with a mean hospital admission rate of 1.72 admissions per patient year at risk and length of stay of 15 hospital days per patient year at risk, for the average of 1994 through 1996 [1]. It is therefore important to identify the factors that influence morbidity in CHD patients. This may allow timely and appropriate interventions to prevent subsequent hospitalizations. Other clinically relevant information that can be in- ferred from this study is that estimation of somatic pro- tein stores is also a reliable indicator of clinical outcome. Reactance values were significantly associated with sub- sequent hospitalizations in this study. Reactance values were used in the study as the primary measure rather than percentage lean body mass because it is this value that predicts the viability of the metabolically active cells, that is, lean body mass, and therefore it alleviates any potential flaws with unconfirmed regression equations [16]. Once more reliable ESRD-specific equations are developed, BIA may be readily used as an inexpensive and practical method for the estimation of body composi- tion, and may represent an advantage when compared with other methods (that is, dual-energy x-ray absorpti- ometry, and prompt neutron activation analysis), which are limited because of their cost and availability. In spite of the intriguing results of this study, several potential shortcomings should be considered. Specifi- cally, the patient population is small and follow-up is limited to 15 months, which prohibits evaluation of mortality as a clinical outcome. It should also be noted that the information presented in this study reflects statistical status in these studies. Specifically, serum albumin has been the most common index of nutritional status used in many, if not all, of the available studies. It is becoming increasingly clear that use of serum albumin as a singular index of nutritional status is complicated by several factors. Even though serum albumin is a reliable index of visceral protein stores, it is a rather late marker for changes in nutrient intake and hence nutritional status because of its long half-life [25 27]. In this study, a variety of nutritional markers was used to assess both visceral as well as somatic protein stores. The results showed a consistent predictive ability for all nutritional markers with regard to hospitalization using the univariate analysis. In addition to serum albumin, serum creatinine, serum transferrin, and serum albumin were all capable of predicting hospitalization in CHD patients. Furthermore, reactance values, which provide an accurate reflection of functional capacity of lean body mass, also significantly predicted hospitalization. Another parameter that was evaluated in this study was serum CRP. CRP is a well-known nonspecific marker of inflammatory response in multiple patient popula- tions. It is known to increase during states of inflamma- tion such as infections and recent surgery. A recent study also related CRP values to risk of myocardial infarction and stroke [28]. CRP has also been evaluated in ESRD patients and is reported to be elevated compared with healthy populations [29, 30]. This increase has been linked to multiple factors, including effects of hemodialy- sis procedure, biocompatibility of the dialysis mem- branes, as well as multiple hospitalizations because of infections and/or other causes [31, 32]. Our study also found CRP to be a powerful predictor of hospitalization in CHD patients. Indeed, it was one of two significant predictive parameters in the multivariate analysis. This is an expected finding given the fact that CRP is a reliable marker for inflammation in all populations, including healthy cohorts. What is as important is that along with CRP, reactance values also independently predicted hos- pitalizations in this group of CHD patients. Reactance is a well-established marker of lean body mass. Indeed, it is suggested to be a more reliable marker of overall nutritional status than lean body mass because it directly reflects the functional capacity of somatic protein stores. This finding suggests that poor nutritional status, independent of the presence of an inflammatory response, is also associated with more hospitalizations. To our knowledge, this is the first study that has systematically shown this association between nutritional status, inflammation, and hospitalization. It is also important to note that the effects of poor nutrition and inflammation are independent, and their concomitant presence would be additive in increasing the risk of hospitalizations in CHD patients. Interestingly, serum albumin as well as serum trans- ferrin and serum prealbumin lost their predictive power for hospitalization once the analysis was adjusted for

7 Ikizler et al: Morbidity and markers of nutrition 1951 correlations, and the clinical applicability of similar associsis. J Am Soc Nephrol 3: , 1993 M, Collier T, Avram MM: Predictors of mortality on hemodialy- ations to larger patient populations needs to be confirmed 13. Oksa H, Ahonen K, Pasternack A, Marnela KM: Malnutrition with further studies. Nonetheless, the results of this study in hemodialysis patients. Scand J Urol Nephrol 25: , 1991 are very consistent and highly significant with each statisof measured variables to death risk among hemodialysis patients, 14. Lowrie EG, Huang WH, Lew NL, Liu Y: The relative contribution tical analysis applied, suggesting that these data could in Death on Hemodialysis, edited by Friedman EA, Amsterdam, be generalized for clinical practice. Kluwer Academic, 1994, p 121 In summary, the results of this study strongly indicate 15. Lowrie EG: Conceptual model for a core pathobiology of uremia with special references to anemia, malnourishment and mortality that both nutritional status and inflammatory response among dialysis patients. Semin Dial 10: , 1997 are independent predictors of hospitalization in this pa- 16. Chertow GM, Lowrie EG, Wilmore DW, Gonzales J, Lew NL, tient population. Specifically, CRP and reactance values Ling J, Leboff MS, Gottlieb MN, Huang W, Zebrowski B, Col- lege J, Lazarus JM: Nutritional assessment with bioelectrical imby BIA are reliable and independent indicators of hospi- pedance analysis in maintenance hemodialysis patients. JAmSoc talization. Visceral proteins such as serum albumin, se- Nephrol 6:75 81, 1995 rum prealbumin, and serum transferrin are influenced 17. Liang KY, Zeger SL: Longitudinal data analysis using generalized linear models. Biometrika 73:13 22, 1986 by inflammation. Therefore, when short-term clinical out- 18. Diggle P, Liang KY, Zeger SL: Analysis of Longitudinal Data. comes such as hospitalizations are considered, markers Oxford, Claredon Press, 1994 of both inflammation and nutrition should be evaluated. 19. Acchiardo SR, Moore LW, Latour PA: Malnutrition as the main factor in morbidity and mortality of hemodialysis patients. Kidney Int 24(Suppl 16):S199 S203, 1983 ACKNOWLEDGMENTS 20. Avram MM, Mittman N, Bonomini L, Chattopadhyay J, Fein P: Markers for survival in dialysis: A seven-year prospective study. This study is supported in part by National Institutes of Health Am J Kidney Dis 26: , 1995 grant #RO1 DK and FDA Grant #FDR We wish 21. Bergström J:Nutrition and mortality in hemodialysis. J Am Soc to gratefully acknowledge the patients and staff of the Vanderbilt Nephrol 6: , 1995 Outpatient Dialysis Unit for their valuable participation and coopera- 22. Cano N, Feinandez JP, Lacombe P, Lankester M, Pascal S, tion in the study. Defayolle M, Labastie J, Saingra S: Statistical selection of nutritional parameters in hemodialysis patients. Kidney Int 32(Suppl Reprint requests to T. Alp Ikizler, M.D., Vanderbilt University Medi- 22):S178 S180, 1987 cal Center, st Avenue South and Garland, Division of Nephrol- 23. Pollock CA, Ibels LS, Ayass W, Caterson RJ, Waugh DA, ogy, S-3223 MCN, Nashville, Tennessee , USA. Macadam C, Pennock Y, Mahony JF: Total body nitrogen as a alp.ikizler@mcmail.vanderbilt.edu prognostic marker in maintenance dialysis. J Am Soc Nephrol 6:82 88, 1995 REFERENCES 24. Owen WF, Lowrie EG: C-reactive protein as an outcome predictor for maintenance hemodialysis patients. Kidney Int 54: , United States Renal Data System: Excerpts from United States 25. Klein S: The myth of serum albumin as a measure of nutritional Renal Data System 1997 Annual Data Report. Am J Kidney Dis status. Gastroenterology 99: , (Suppl 1):S1 S195, Winkler MF, Pomp A, Caldwell MD, Albina JE: Transitional 2. Ikizler TA, Hakim RM: Nutrition in end-stage renal disease. Kid- feeding: The relationship between nutritional intake and plasma ney Int 50: , 1996 protein concentrations. J Am Diet Assoc 89: , Parker TFI, Laird NM, Lowrie EG: Comparison of the study 27. Jacobs DO, Scheltinga MRM: Metabolic assessment, in Clinical groups in the national cooperative dialysis study and a description Nutrition: Parenteral Nutrition, edited by Rombeau JL, Caldwell of morbidity, mortality, and patient withdrawal. Kidney Int MD, Philadelphia, W.B. Saunders, 1993, p (Suppl 13):S42 S49, Ridker PM, Cushman M, Stampfer MJ, Tracy RP, Hennekens 4. Owen JRWF, Lew NL, Liu Y, Lowrie EG, Lazarus JM: The urea CH: Inflammation, aspirin, and the risk of cardiovascular disease reduction ratio and serum albumin concentrations as predictors in apparently healthy men. N Engl J Med 336: , 1997 of mortality in patients undergoing hemodialysis. N Engl J Med 29. Haubitz M, Schulze M, Koch KM: Increase of C-reactive protein 329: , 1993 serum values following haemodialysis. Nephrol Dial Transplant 5. Lowrie EG, Lew NL: Death risk in hemodialysis patients: The 5: , 1990 predictive value of commonly measured variables and an evalua- 30. Docci D, Bilancioni R, Buscaroli A, Baldrati L, Capponcini tion of death rate differences between facilities. Am J Kidney Dis C, Mengozzi S, Turci F, Feletti C: Elevated serum levels of 15: , 1990 C-reactive protein in hemodialysis patients. Nephron 56: , 6. Kopple JD: Effect of nutrition on morbidity and mortality in main tenance dialysis patients. Am J Kidney Dis 24: , Yeun JY, Kaysen GA: Acute phase proteins and peritoneal dialy- 7. Churchill DN, Taylor DW, Cook RJ, Al E: Canadian hemodial- sate albumin loss are the main determinants of serum albumin in ysis morbidity study. Am J Kidney Dis 19: , 1992 peritoneal dialysis patients. Am J Kidney Dis 30: , Collins AJ, Ma JZ, Umen A, Keshaviah P: Urea index and 32. McIntyre C, Harper I, Macdougall IC, Raine AE, Williams other predictors of hemodialysis patient survival. Am J Kidney A, Baker LR: Serum C-reactive protein as a marker for infection Dis 23: , 1993 and inflammation in regular dialysis patients. Clin Nephrol 48: Iseki K, Kawazoe N, Fukiyama K: Serum albumin is a strong 374, 1997 predictor of death in chronic dialysis patients. Kidney Int 44: Kaysen GA, Rathore V, Shearer GC, Depner TA: Mechanisms 119, 1993 of hypoalbuminemia in hemodialysis patients. Kidney Int 48: Verdery RB, Goldberg AP: Hypocholesterolemia as a predictor 516, 1995 of death: A prospective study of 224 nursing home residents. 34. Kaysen GA, Stevenson FT, Depner TA: Determinants of albumin J Gerontol 46:M84 M90, 1994 concentration in hemodialysis patients. Am J Kidney Dis 29: Goldwasser P, Michel MA, Collier J, Mittman N, Fein P, Gusik 668, 1997 SA, Avram MM: Prealbumin and lipoprotein(a) in hemodialysis: 35. Qureshi AR, Alvestrand A, Danielsson A, Divino-Filho JC, Relationships with patient and vascular access survival. Am J Kid- Gutierrez A, Lindholm B, Bergström J:Factors predicting malney Dis 22: , 1993 nutrition in hemodialysis patients: A cross-sectional study. Kidney 12. Goldwasser P, Mittman M, Antignani A, Burrel D, Michel Int 53: , 1998