Kidney International, Vol. 62 (2002), pp. 2238–2245

DIALYSIS – TRANSPLANTATION

Mortality risk in hemodialysis patients and changes innutritional indicators: DOPPS

TRINH B. PIFER, KEITH P. MCCULLOUGH, FRIEDRICH K. PORT, DAVID A. GOODKIN,BRADLEY J. MARONI, PHILIP J. HELD, and ERIC W. YOUNG

University Renal Research and Education Association, Department of Medicine, Veterans Affairs Medical Center, and Divisionof Nephrology, University of Michigan, Ann Arbor, Michigan; and Amgen, Inc., Thousand Oaks, California, USA

Mortality risk in hemodialysis patients and changes in nutritional Nutritional problems are common among patientsindicators: DOPPS. with end-stage renal disease (ESRD) and are associated

Background. Nutritional status is strongly associated with with higher rates of morbidity and mortality, as describedoutcomes among hemodialysis patients. We analyzed the inde-in a recent Dialysis Outcomes Quality Initiative (DOQI)pendent predictive value of several readily measured nutritional[1]. Several anthropomorphic and laboratory measure-indicators, including a modified subjective global assessment

(mSGA), body mass index (BMI), serum albumin, serum creat- ments correlate with nutritional status, although no sin-inine, normalized protein catabolic rate (nPCR), serum bicar- gle measurement provides a complete and unambiguousbonate, lymphocyte count, and neutrophil count, using baseline

assessment. Several potential indicators are strongly in-and six-month follow-up measurements.fluenced by non-nutritional factors such as inflammationMethods. The study sample consisted of 7719 U.S. adult

hemodialysis patients enrolled in the international Dialysis and volume overload, which are common in dialysis pa-Outcomes and Practice Patterns Study (DOPPS), a prospective tients. The use of a broad panel of putative indicatorsobservational study that includes a random sample of hemodi- may best facilitate the epidemiologic and clinical assess-alysis patients from 145 dialysis facilities in the United States.

ment of nutritional status [2, 3].Cox regression was used to estimate the relative risk of mortal-The mortality risks associated with several readily mea-ity associated with differences in measurements at baseline and

six months later. Each analysis was adjusted for age, race, sex, sured nutritional indictors were evaluated, including aand 15 summary comorbid conditions. modified subjective global assessment (mSGA), body mass

Results. Lower baseline measurements of mSGA, BMI, se-index (BMI), serum albumin, serum creatinine, normal-rum albumin, serum creatinine, and lymphocyte count wereized protein catabolic rate (nPCR), and lymphocyte count.independently associated with significantly higher risk of mor-

tality. During six-month follow-up, decreases in BMI, serum Also, the associations between mortality risk and serumalbumin, and serum creatinine were also associated with sig- bicarbonate and neutrophil count were examined, asnificantly higher mortality risk. The risk of mortality increased these indicators are used often to identify acidosis orwith higher baseline and six-month increases in neutrophil count.

inflammation, respectively. The most important evalua-Conclusions. This study confirms that several readily-mea-tion looked at the mortality risks associated with longitu-sured nutritional indicators predict mortality among hemodial-

ysis patients and that changes in indicator values over six months dinal changes in selected indicators which have not beenprovide additional important prognostic information. Interven- well reported. This study was part of the larger Dialysistions that modify these indicators of nutritional status may have

Outcomes and Practice Pattern Study (DOPPS), a longi-an important impact on the survival of hemodialysis patients.tudinal, observational study of chronic hemodialysis pa-tients involving nationally representative samples of he-modialysis facilities and patients from seven countries:France, Germany, Italy, Japan, Spain, the United King-dom, and the United States. The details of the studyhave been described previously [4].

Key words: albumin, body mass index, creatinine, Dialysis Outcomesand Practice Patterns Study, end-stage renal disease, hemodialysis, lym-phocyte, mortality, neutrophil, normalized protein catabolic rate, nutri-

METHODStional status, subjective global assessment.

The current study was restricted to patients from dial-Received for publication November 8, 2001ysis facilities in the United States (N � 145 facilities).and in revised form May 3, 2002

Accepted for publication July 8, 2002 Within each facility, a random sample of 20 to 40 hemodi-alysis patients was selected for participation in DOPPS. 2002 by the International Society of Nephrology

2238

Pifer et al: Nutrition and mortality in HD patients 2239

Table 1. Modified subjective global assessment (mSGA)

Nutrition status

mSGA variable Normal Moderate Severe

Recent weight loss No recent weight loss 5–10% weight loss in 6 months �10% weight loss in �6 monthsVisual somatic protein wasting

(provider judgment) None None Cachectic appearanceLoss of appetite (patient reported) None Moderate ExtremeNausea/vomiting (patient reported) None Moderate ExtremeEnergy level (patient reported) Rarely tired Sometimes tired Always tiredDisease burden (patient reported) Good/best possible health Moderate health Bad/worse possible health

Classification rules:1. “Loss of appetite” and “nausea/vomiting” only count as a single factor.2. “Energy level” and “disease burden” only count as a single factor.3. Severely malnourished � at least three factors at the severe malnutrition level.4. Moderately malnourished � if not severely malnourished, any combination of three factors at the moderate or the severe malnutrition level.

Patients who were lost to follow-up (died or departed Some patients entered the study at onset of ESRD,while others entered after being on dialysis for severalfrom the facility) were replaced by newly enrolled pa-

tients. Clinical data were collected for each subject at years. Statistical adjustment for patient years of ESRDprior to their entry into the study was made using leftstudy enrollment and approximately every four months

thereafter. In addition, each patient was asked to com- truncation [6, 7]. Left-truncation allows for descriptionof the survival experience beginning at the start ofplete a questionnaire that included quality of life and

nutritional history items. ESRD, and under this method patients only contributeto the study during times when they are actually at risk.Data were collected for several putative nutritional

indicators including BMI, serum albumin, serum creati- All analyses were performed using SAS software (SASnine, nPCR, lymphocyte count, and mSGA. In addition, Institute, Cary, NC, USA) [8].data were collected on neutrophil count and serum bicar- Baseline characteristics were tested using a full samplebonate as indicators of inflammation and acidosis. BMI of 7719 patients. The six-month longitudinal nutritionalwas based on the patients’ post-dialysis weight, and labo- data were available for 3739 patients. This decrease re-ratory values were based on pre-dialysis measurements. sulted primarily from questionnaires not being returnedThe nPCR was determined using the two-point model within 60 days of the desired six-month anniversary ofof urea kinetics published by Depner and Daugirdas [5]: the baseline measurement (29% of all patients with firstnPCR (g/kg/day) � C0/[a � bKt/V � c/(Kt/V)] � 0.0168, measurement). Attrition (death, transplantation, and de-where C0 is the pre-dialysis BUN in mg/dL and Kt/V is parture from a study facility) also reduced the samplethe single-pool estimate of dialysis dose. For patients size (18% of all patients with first measurement). Allwho received dialysis three times weekly, appropriate models contained both the six-month longitudinal changescoefficients (a, b, c) were used, depending on whether and the baseline values for each of the individual nutri-laboratory measurements were obtained at the beginning, tional parameters except mSGA. Where possible, missingmiddle, or end of the week. The mSGA was determined nutritional indicator data were handled by creating a bi-at baseline for each patient, based on the caregivers’ nary missing data variable. This approach yielded an unbi-responses to questions about patient weight loss and ased estimate of the effects of nutritional indicators onphysical appearance and on patient responses to ques- mortality. Each putative nutritional indicator was groupedtions about appetite, nausea, energy level, and disease into quartiles and evaluated in a separate model. In addi-burden. Table 1 describes how this information was used tion, all indicators were entered in an overall multivariateto classify patients into one of three mSGA categories: model as continuous variables (except for mSGA, whichnormal, moderately, or severely malnourished. was entered as a categorical variable). A P value less

Patient characteristics were summarized using stan- than 0.05 was considered statistically significant.dard descriptive statistics. Cox regression was used toestimate the risk of mortality associated with both the

RESULTSbaseline and the six-month change data for the nutri-tional indicators. All models were adjusted for age, race, Table 2 shows demographic and comorbid characteris-

tics for the patient sample (N � 7719). Patient character-sex, time since onset of ESRD (through left truncation),and 15 summary comorbid conditions that were found to istics were similar to other representative samples of

chronic dialysis patients reported in contemporaneousindependently predict mortality among DOPPS patients(Table 2). studies [9]. Table 3 summarizes the patients’ nutritional

Pifer et al: Nutrition and mortality in HD patients2240

Table 2. Summary of patient demographics and comorbidities

Factor Mean (SD)

Age years 60.8 (15.6)Race % black vs. non-black 34.3Gender % male 55.3Time since onset ESRD years 2.1 (3.6)Cancer (other than skin) % 10.9Cerebrovascular disease % 18.6Congestive heart failure % 47.1Coronary heart disease % 50.2Diabetes mellitus % 49.1Dyspnea % 34.4 Fig. 1. Relative risk of mortality and categories of modified subjectiveHIV/AIDS % 2.3 global assessment (mSGA). All categories of SGA, demographics, andHypertension % 84.8 15 comorbid conditions are shown together in a Cox model. *P � 0.05Lung disease % 14.0 vs. normal category.Neurological disease % 11.0Other cardiovascular diseasea % 33.3Peptic ulcer % 9.0Peripheral vascular disease % 26.4Psychiatric disorder % 25.8 the highest quartile. Furthermore, mortality risk signifi-Recurrent cellulitis/gangrene % 10.3 cantly increased among patients with a decrease in BMI

a Includes cardiac arrest, atrial fibrillation, other arrhythmia, permanent pace- of greater than 3.5% (Fig. 2B). A 3.5% decrease in BMImaker implanted, pericarditis, prosthetic heart valve (s/p valve replacement)

reflects a 3.5% loss in body weight.A strong inverse association was observed between

mortality and the serum albumin concentration (Fig.Table 3. Summary of putative nutritional indicators 3A). The relationship did not appear to reach a plateau

even in the range of relatively normal albumin concen-Nutritional indicator Mean (SD)trations, as shown by the significant differences in riskModerate malnutrition % 7.6 (26.6)between the third and fourth quartiles. Mortality riskSevere malnutrition % 11.0 (31.3)

Baseline BMI kg/m2 25.4 (5.9) also was strongly associated with a decline in the serumHeight m 1.7 (0.11) albumin concentration during the six-month follow-upPost-dialysis weight kg 73.1 (19.7)

(Fig. 3B). A decline in the serum albumin concentrationBaseline albumin g/dL 3.6 (0.6)Baseline creatinine mg/dL 8.8 (3.7) of greater than 5.3% was associated with a near doublingBaseline nPCR 1.0 (0.3) of mortality risk (P � 0.001).Baseline lymphocyte cells/mm3 1494.4 (1314.5)

Mortality risk also was inversely associated with theBaseline bicarbonate mEq/L 20.9 (4.2)Baseline neutrophil cells/mm3 5287.4 (2752.2) baseline serum creatinine concentration (Fig. 4A) and a�BMI % 2.2 (7.9) decrease in the serum creatinine concentration (Fig. 4B).�Albumin % 8.8 (16.7)

In both cases, the mortality risk was 60 to 70% higher�Creatinine % 17.4 (33.3)�nPCR % 7.6 (33.8) in the lowest quartile group compared with the highest�Lymphocyte % 21.5 (33.6) quartile group.�Bicarbonate % 3.3 (24.6)

The lowest quartile of baseline lymphocyte count was�Neutrophil % 14.3 (26.9)associated with increased risk of mortality, but a six-month change in lymphocyte count was not significantlyrelated to death risk (Fig. 5). In contrast, mortality in-creased in association with higher baseline neutrophilcharacteristics and changes (�) during the six-month fol-

low-up. The restricted sample (N � 3739) used to analyze count and longitudinal increases in the neutrophil count(Fig. 6). A strong inverse correlation was observed be-the impact of changes in the putative nutritional indica-

tors was similar to the larger sample in all respects (data tween neutrophil count and serum albumin (R � �0.206,P � 0.0001) and between neutrophil count and serumnot shown).

Patients with severely malnourished mSGA scores had creatinine (R2 � �0.172, P � 0.0001).There was no association between mortality and base-a 33% higher mortality risk, and patients with moderate

malnutrition had a 5% higher mortality risk than patients line measurements of nPCR. A large six-month declinein nPCR (�15.6%) was associated with higher mortalitywith normal mSGA scores (Fig. 1). Unlike other putative

indicators, mSGA was not measured longitudinally, as (RR � 1.22), but this finding was not statistically signifi-cant (not shown). For serum bicarbonate concentration,these data were not available at the six-month follow-up.

Mortality risk was higher for patients with lower BMI, there was no association between mortality and eitherthe baseline level or six-month change (data not shown).and there was no indication of increased risk associated

with obesity (Fig. 2A). Patients in the lowest BMI quartile Table 4 displays the results of the individual modelsfor each indicator as a continuous variable, except forhad a mortality risk 60% higher than that of patients in

Pifer et al: Nutrition and mortality in HD patients 2241

Fig. 2. Relative risk of mortality and quartilesof body mass index (BMI), adjusted for base-line BMI (A), �BMI (B), demographics, and15 comorbid conditions. *P � 0.05 and †P �0.001 vs. 4th quartile.

Fig. 3. Relative risk of mortality and quartilesof serum albumin, adjusted for baseline albu-min (A), �albumin (B), demographics, and 15comorbid conditions. *P � 0.05 and †P �0.001 vs. 4th quartile.

mSGA, which is a categorical variable. Each model was nutrition. The results illustrate the potential value of usinga variety of measures to assess nutritional status in epide-adjusted for the factors listed in Table 2. The resultsmiological studies and, potentially, for clinical evalua-of an overall Cox regression model that simultaneouslytion. A major strength of this study is the longitudinalincluded all nutritional indicators entered as continuousmeasurements of nutritional data for a large, representa-variables (except for mSGA) is shown in the right col-tive sample of hemodialysis patients in the United States.umn. In the overall model, significant associations per-Although nutritional status was not independently cor-sisted between mortality and BMI, serum albumin, six-roborated, the validity of the results is partly affirmedmonth change in albumin, serum creatinine, neutrophilby the strength of the associations between mortalitycount, and mSGA.and putative nutritional indicators. Furthermore, severalassociations observed in this study have been described

DISCUSSION previously in ESRD patients.This study investigated the impact of several putative Mortality risk was significantly associated with mSGA.

nutritional indicators on mortality in chronic hemodialy- The mSGA is based on subjective and objective aspectssis patients. Mortality risk was significantly associated of the patients’ medical history and physical examina-

tion, including any recent weight loss, dietary intake,with clinical measurements that are possibly linked to

Pifer et al: Nutrition and mortality in HD patients2242

Fig. 4. Relative risk of mortality and quartilesof serum creatinine, adjusted for baseline cre-atinine (A), �creatinine (B), demographics,and 15 comorbid conditions. *P � 0.05 and†P � 0.001 vs. 4th quartile.

Fig. 5. Relative risk of mortality and quartilesof lymphocyte count, adjusted for baselinelymphocyte count (A), �lymphocyte count (B),demographics, and 15 comorbid conditions.*P � 0.05 vs. 4th quartile.

Fig. 6. Relative risk of mortality and quartilesof neutrophil count, adjusted for neutrophilcount (A), �neutrophil count (B), demograph-ics, and 15 comorbid conditions. *P � 0.05and †P � 0.001 vs. 4th quartile.

Pifer et al: Nutrition and mortality in HD patients 2243

Table 4. Adjusted relative risk (RR) for mortality by nutritional eral population. It has been speculated that this protec-indicators as continuous variables

tive effect may be due to the benefits of greater energyOverall reserves if hemodialysis adds to catabolism [13].

Nutritional indicator Isolated multivariate Serum albumin is a marker of visceral protein stores.(vs. reference group or per one SD increase) modelsa modela

However, serum albumin is affected also by non-nutri-Severe SGA vs. normal 1.33b 1.25b

tional factors such as infection or inflammation, hydra-Moderate SGA vs. normal 1.05 0.97BMI per 5.9 kg/m2 0.81c 0.88c tion status, and anabolic or catabolic processes [17, 18].

�BMI per 10% 0.89 1.02 Nonetheless, we found a significant inverse associationAlbumin per 0.56 g/dL 0.74c 0.78c

between mortality and both baseline measures and longi-�Albumin per 20% 0.75c 0.76c

Creatinine per 3.7 mg/dL 0.78c 0.85c tudinal change in serum albumin concentration, consis-�Creatinine per 39% 0.88 0.92 tent with findings in other studies [19–23]. The strong

nPCR per 0.28 units 1.02 NAassociation between mortality and albumin persisted in�nPCR per 34% 0.98 NA

Lymphocyte count per 1300 cells/mm3 0.98 0.95 the model that included all the nutritional variables that�Lymphocyte count per 34% 0.99 0.93 we studied (Table 4). Despite its limitations, albumin is

Bicarbonate per 4.2 mEq/L 0.97 NAa potent and robust indicator of both nutritional status�Bicarbonate per 25% 0.97 NA

Neutrophil count per 2800 cells/mm3 1.26c 1.09b and mortality risk.�Neutrophil count per 27% 1.00 1.0 The serum creatinine concentration is related to nutri-NA, did not significantly predict mortality in Overall Model. tional status in that it reflects somatic protein stores,a All models adjusted for age, race, sex, duration of ESRD, and 15 comorbid

muscle mass, and dietary protein intake. However, manyfactorsb P � 0.05, cP � 0.001 other factors influence the serum creatinine concentra-

tion, including age, sex, race, dialysis dose, and residualrenal function [24]. In any case, low serum creatininelevels have been shown to be highly predictive of future

gastrointestinal symptoms, and a visual assessment ofmortality in dialysis patients [19, 23, 25], as confirmed

subcutaneous fat [10–12]. For this study, we developedin this study. We also found that mortality risk was associ-

a modified mSGA scoring system (Table 1) based on ated with a fall in the serum creatinine concentrationinformation provided on standardized questionnaires by over six months. Although not yet used for this purpose,the nurse study coordinators and the patients. Given its baseline and longitudinal measures of serum creatininestrong association with mortality, it appears that mSGA concentration could be helpful as clinical indicators ofprovides a meaningful assessment of nutritional status. high-risk patients.This simple and inexpensive indicator could be useful Nutritional and inflammatory status can influence thefor identifying high-risk patients who might benefit from concentrations of various white blood cells. It also hasnutritional interventions. been shown that lymphocyte proliferation and function

Anthropometric measurements are frequently used to are impaired by protein and calorie malnutrition [26].indicate protein-energy nutritional status. BMI provides In this study, an increased mortality risk was associateda simple and inexpensive proxy for body mass adjusted with a low baseline lymphocyte count. Mortality risk wasfor height. Hence, it provides a more accurate measure not associated with changes in lymphocyte count overof overweight and obesity than relying on weight alone. six months, however. Owen and Lowrie also reported aThe following BMI guidelines were set by the World strong, inverse relationship between total lymphocyteHealth Organization for the general population: normal count and the risk of death [23]. The association between(BMI 18 to 25), overweight (BMI �25), and obese (BMI lymphocyte count and mortality could be explained by�30). However, there are limitations associated with nutritional or immunological mechanisms.BMI, especially its inability to differentiate between Infectious and inflammatory processes are stimuli ofbone, muscle, and fat compartments. In addition, BMI neutrophil proliferation. In our study, risk of mortalitymay be misleading in the presence of edema, which is was directly associated with both the neutrophil count atcommonly present in the hemodialysis population. De- baseline and with six-month increases in the neutrophilspite these limitations, many studies have shown BMI count. These findings suggest that the presence of anto be highly correlated with both morbidity and mortality infection or inflammatory response, as indicated by a[13–16]. In our analyses using post-dialysis weight of high or rising neutrophil count, is associated with higherpatients for BMI calculations, we observed a significantly mortality. Furthermore, we observed a strong inversehigher risk of mortality in patients with a lower BMI at correlation between neutrophil count and serum albuminbaseline. Furthermore, a large decrease in BMI over and creatinine, consistent with previous studies [23].six months also was associated with increased risk of These findings suggest that infection or inflammationmortality. A high BMI seems to have a protective effect has a potentially negative effect on serum albumin and

creatinine, causing decreased levels of these nutritionalfor chronic hemodialysis patients, in contrast to the gen-

Pifer et al: Nutrition and mortality in HD patients2244

indicators. It follows that it is important to search for the REFERENCESpresence of inflammation when using these laboratory 1. National Kidney Foundation: K/DOQI clinical practice guide-values as nutritional indicators. lines for nutrition in chronic renal failure. Am J Kidney Dis

35(Suppl 2):S1–S140, 2000The nPCR is widely interpreted as a measure of di-2. Simko M, Cowell C, Gilbride J (editors): Nutrition Assessment:etary protein intake (DPI), assuming a state of protein

A Comprehensive Guide for Planning Intervention (2nd ed). Gaith-balance. A value for nPCR is much simpler to determine ersburg, MD, Aspen Publishers, 1995

3. Blumenkrantz MJ, Kopple JD, Gutman RA, et al: Methods forthan DPI from diet diaries or interviews [5, 27]. However,assessing nutritional status of patients with renal failure. Am J Clinwe did not observe any relationship between risk of mor-Nutr 33:1567–1585, 1980tality and either baseline or six-month follow-up measure- 4. Young EW, Goodkin DA, Mapes DL, et al: The Dialysis Out-

ments of nPCR. This finding raises concerns about the comes and Practice Patterns Study (DOPPS): An internationalhemodialysis study. Kidney Int 57(Suppl 74):S71–S81, 2000assumptions behind the use of nPCR as a measure of

5. Depner TA, Daugirdas JT: Equations for normalized proteinDPI. In particular, the assumption that patients maintaincatabolic rate based on two-point modeling of hemodialysis urea

a steady protein balance state is challenged by the sub- kinetics. J Am Soc Nephrol 7:780–785, 1996stantial changes observed in other putative nutritional 6. Klein JP, Moeschberger ML (editors): Survival Analysis: Tech-

niques for Censored and Truncated Data. New York, Springer-indicators over the six-month period (see quartile rangesVerlag, 1997in Figs. 2 to 6). It is possible that the precision of the 7. Thernaeu TM, Grambsch PM (editors): Modeling Survival Data:

nPCR measurement is affected by lack of uniformity of Extending the Cox Model. New York, Springer-Verlag, 20008. SAS Institute Inc: SAS/STAT� Version 8. Cary, NC, SAS Institutethe post-dialysis blood flow for BUN. Variability in the

Inc, 1999, 3884 ppnPCR measurements would bias the results, showing a9. U.S. Renal Data System: USRDS 1997 Annual Data Report. Na-

lack of correlation with other measures. tional Institutes of Health, National Institutes of Diabetes andSystemic acidosis, as indicated by a low serum bicar- Digestive and Kidney Diseases, Bethesda, 1997

10. Kalantar-Zadeh K, Kleiner M, Dunne E, et al: A modifiedbonate concentration, is associated with increased pro-quantitative subjective global assessment of nutrition for dialysistein degradation and decreased albumin synthesis [28].patients. Nephrol Dial Transplant 14:1732–1738, 1999

The correction of acidosis has been associated with in- 11. Enia G, Sicuso C, Alati G, Zoccali C: Subjective global assess-ment of nutrition in dialysis patients. Nephrol Dial Transplantcreased serum albumin concentration, decreased protein8:1094–1098, 1993degradation rates, and increased body weight [29, 30].

12. McCann L: Subjective global assessment as it pertains to the nutri-However, other studies failed to show any adverse effect tional status of dialysis patients. Dial Transplant 25:199–202, 1996of acidosis [31]. We did not observe any significant associ- 13. Leavey SF, Strawderman RL, Jones CA, et al: Simple nutritional

indicators as independent predictors of mortality in hemodialysisations between mortality and either baseline bicarbonatepatients. Am J Kidney Dis 31:997–1006, 1998levels or six-month changes in bicarbonate concentra-

14. Kopple JD, Zhu X, Lew NL, Lowrie EG: Body weight-for-heighttion. Protein intake contributes to acid production, rai- relationships predict mortality in maintenance hemodialysis pa-sing the possibility that a higher dietary intake may offset tients. Kidney Int 56:1136–1148, 1999

15. Fleischman E, Teal N, Dudley J, et al: Influence of excess weightthe presumed detrimental effects of acidosis, as pre-on mortality and hospital stay in 1346 hemodialysis patients. Kidneyviously suggested by Leavey et al [21].Int 55:1560–1567, 1999

16. Calle EE, Thun MJ, Petrell JM, et al: Body-mass index andConclusion mortality in a prospective cohort of US adults. N Engl J Med

341:1097–1105, 1999Assessing the nutritional status of chronic hemodialy-17. Kaysen GA, Stevenson FT, Depner TA: Determinants of albuminsis patients is a challenging and complicated process be- concentration in hemodialysis patients. Am J Kidney Dis 29:658–

cause there is no one practical measurement that func- 668, 199718. Dutton J, Campbell H, Tanner J, Richards N: Pre-dialysis serumtions well for both research and clinical purposes. The

albumin is a poor indicator of nutritional status in stable chronicuse of a broad panel of indicators may well provide thehaemodialysis patients. EDNTNA ERCA J 25:36–37, 1999best approach. Indicators such as mSGA, BMI, serum 19. Lowrie EG, Lew NL: Death risk in hemodialysis patients: The

albumin, serum creatinine, lymphocyte count, and neu- predictive value of commonly measured variables and an evalua-tion of death rate differences between facilities. Am J Kidney Distrophil count are independently associated with mortal-15:458–482, 1990ity and could prove valuable as such a panel. Interven-

20. Goldwasser P, Mittman N, Antignani A, et al: Predictors oftions that modify these indicators of nutritional status mortality in hemodialysis patients. J Am Soc Nephrol 3:1613–1622,may have an important impact on the mortality of hemo- 1993

21. Leavey SF, Strawderman RL, Young EW, et al: Cross-sectionaldialysis patients.and longitudinal predictors of serum albumin in hemodialysis pa-tients. Kidney Int 58:2119–2128, 2000

ACKNOWLEDGMENT 22. Owen WF, Lew NL, Liu Y, et al: The urea reduction ratio andserum albumin concentration as predictors of mortality in patientsThis study was supported by a grant from Kirin-Amgen, Inc.undergoing hemodialysis. N Engl J Med 329:1001–1006, 1993

23. Owen WF, Lowrie EG: C-reactive protein as an outcome predictorReprint requests to Trinh B. Pifer, M.P.H., University Renal Researchfor maintenance hemodialysis patients. Kidney Int 54:627–636, 1998and Education Association, 315 W. Huron, Suite 360, Ann Arbor, Michi-

24. Mitch EM, Walser M: Creatinine metabolism in chronic renalgan 48103 USA.E-mail: tpifer@urrea.org failure. Clin Sci 58:327–335, 1980

Pifer et al: Nutrition and mortality in HD patients 2245

25. U.S. Renal Data System: USRDS 1992 Annual Data Report. Am bolic acidosis decreases albumin synthesis and induces negativenitrogen balance in humans. J Clin Invest 95:39–45, 1995J Kidney Dis 20(Suppl 2):32–38, 1992

29. Graham KA, Reaich D, Channon SM, et al: Correction of acidosis26. McMurray DN, Loomis SA, Casazza LJ: Development of im-in hemodialysis decreases whole-body protein degradation. J Ampaired cell-mediated immunity in mild and moderate malnutrition.Soc Nephrol 8:632–637, 1997Am J Clin Nutr 34:68–77, 1981

30. Brady JP, Hasbargen JA: Correction of metabolic acidosis and27. Panzetta G, Tessitore N, Faccini G, Maschio G: The protein its effect on albumin in chronic hemodialysis patients. Am J Kidneycatabolic rate as a measure of protein intake in dialysis patients: Dis 31:35–40, 1998Usefulness and limits. Nephrol Dial Transplant 5(Suppl 1):125–127, 31. Movilli E, Bossini N, Viola BF, et al: Evidence for an independent1990 role of metabolic acidosis on nutritional status in haemodialysis

patients. Nephrol Dial Transplant 13:674–678, 199828. Ballmer PE, McNurlan MA, Hulter HN, et al: Chronic meta-