Is controlling phosphorus by decreasing dietary protein intakebeneficial or harmful in persons with chronic kidney disease?1–4

Christian S Shinaberger, Sander Greenland, Joel D Kopple, David Van Wyck, Rajnish Mehrotra, Csaba P Kovesdy,and Kamyar Kalantar-Zadeh

ABSTRACTBackground: Dietary restrictions to control serum phosphorus,which are routinely recommended to persons with chronic kidneydisease, are usually associated with a reduction in protein intake.This may lead to protein-energy wasting and poor survival.Objective: We aimed to ascertain whether a decline in serum phos-phorus and a concomitant decline in protein intake are associatedwith an increase in the risk of death.Design: In a 3-y study (7/2001–6/2004) of 30 075 prevalent main-tenance hemodialysis (MHD) patients, we examined changes inserum phosphorus and in normalized protein nitrogen appearance(nPNA), a surrogate of dietary protein intake, during the first 6 moand the subsequent mortality. Four groups of MHD patients weredefined on the basis of the direction of the changes in serum phos-phorus and nPNA.Results: Baseline phosphorus had a J-shaped association with mor-tality, whereas higher baseline nPNA was linearly associated withgreater survival. Compared with MHD patients whose serum phos-phorus and nPNA both rose over 6 mo, those whose serum phos-phorus decreased but whose nPNA increased had greater survival,with a case mix–adjusted death risk ratio of 0.90 (95% confidencelimits: 0.86, 0.95; P � 0.001), whereas those whose phosphorusincreased but whose nPNA decreased or those whose phosphorusand nPNA both decreased had worse mortality with a risk ratio of1.11 (1.05,1.17; P � 0.001) and 1.06 (1.01,1.12; P � 0.02), respec-tively.Conclusions: The risk of controlling serum phosphorus by restrict-ing dietary protein intake may outweigh the benefit of controlledphosphorus and may lead to greater mortality. Additional studiesincluding randomized controlled trials should examine whether non-dietary control of phosphorus or restriction of nonprotein sources ofphosphorus is safer and more effective. Am J Clin Nutr 2008;88:1511–8.

INTRODUCTION

Hyperphosphatemia is a common disorder in persons withadvanced chronic kidney disease (CKD) (1, 2). Gradual declinein renal phosphorus clearance during the progression of CKDleads to increases in serum phosphorus concentrations (3). Thisdevelopment may result in additional mineral and bone disor-ders, such as the inhibition of 1-�-hydroxylation of 25-hydroxycalciferol via the hyperphosphatemia-induced fibroblastgrowth factor-23 pathway (4, 5). Both hyperphosphatemia andcalcitriol deficiency may result in hyperparathyroidism and renal

osteodystrophy (6). Hyperphosphatemia also may contribute toworsening vascular calcification and greater risk of cardiovas-cular morbidity (7, 8). Indeed, hyperphosphatemia is a knowndeath risk in both the general population (9) and CKD patients(10, 11), including maintenance dialysis patients (12–14), and itmay worsen the rate of CKD progression (15–17). Hence, cor-rection and prevention of hyperphosphatemia is a main compo-nent of the management of CKD. This goal is usually approachedboth by administering phosphorus binders and by restrictingdietary phosphorus intake (18–20).

Because foods high in protein are a main source of dietaryphosphorus, imposing dietary phosphorus restriction is oftenassociated with a reduction in dietary protein intake. The lattercan lead to malnutrition and protein-energy wasting (PEW),which are strong risk factors for greater risk of death in mainte-nance dialysis patients (21, 22). It is thus important to examinethe risks or benefits of restricting dietary protein intake to controlserum phosphorus concentrations in persons with renal insuffi-ciency who undergo maintenance hemodialysis (MHD) treat-ment. It is not clear whether a reduction in serum phosphorus byvirtue of a concurrent fall in protein intake is associated withbetter or worse survival. We hypothesized that a decline in serum

1 From the Harold Simmons Center for Kidney Disease Research andEpidemiology (CSS and KK-Z) and the Division of Nephrology and Hyper-tension (JDK, RM, and KK-Z), Los Angeles Biomedical Research Instituteat Harbor–UCLA Medical Center, Torrance, CA; and the David GeffenSchool of Medicine at UCLA, Los Angeles, CA (JDK, RM, and KK-Z); theDepartments of Epidemiology (CSS, KK-Z, and SG) and Family Health(JDK), UCLA School of Public Health, Los Angeles, CA; the Salem VAMedical Center, Salem, VA (CPK); DaVita, Inc, El Segundo, CA (DVW);and the Arizona Center on Aging, Arizona Health Sciences Center, Tucson,AZ (DVW).

2 The abstract for this article was presented orally at the InternationalSociety of Renal Nutrition and Metabolism conference, June 10–14, 2008, inMarseilles, France. The data in this report are part of the doctoral dissertationof the first author (CSS) in the Department of Epidemiology at the UCLASchool of Public Health, Los Angeles, CA.

3 Supported by grants no. R01DK078106 from the National Institutes ofHealth and 0655776Y from the American Heart Association (to KK-Z), byDaVita Clinical Research, and by a grant from Harold Simmons (to KK-Z).

4 Address reprint requests to K Kalantar-Zadeh, Harold Simmons Centerfor Kidney Disease Research & Epidemiology, Los Angeles BiomedicalResearch Institute at Harbor–UCLA Medical Center, 1124 West CarsonStreet, C1-Annex, Torrance, CA 90509-2910. E-mail: kamkal@ucla.edu.

Received July 8, 2008. Accepted for publication September 5, 2008.doi: 10.3945/ajcn.2008.26665.

1511Am J Clin Nutr 2008;88:1511–8. Printed in USA. © 2008 American Society for Nutrition

by guest on June 7, 2013ajcn.nutrition.org

Dow

nloaded from

phosphorus associated with a concomitant decline in proteinintake increases the risk of death, whereas controlling serumphosphorus without restricting dietary protein intake is associ-ated with improved survival in established MHD patients.

SUBJECTS AND METHODS

Patients

We examined data from all stage 5 CKD patients who under-went MHD treatment from July 1, 2001, to December 31, 2001,in any of the 560 outpatient dialysis facilities of a large dialysisorganization in the United States (DaVita, Inc; El Segundo, CA),and we followed those patients until death, censoring, or June 30,2004. Because of the large sample size, the anonymity of thepatients studied, and the nonintrusive nature of the research, therequirement for written informed consent was waived. The studywas approved by the Institutional Review Committees of bothLos Angeles Biomedical Research Institute at Harbor–UCLAand DaVita Clinical Research.

Clinical and demographic measures

The study cohort was described previously (14, 21, 23, 24). Tominimize measurement variability, all repeated measures foreach patient during the baseline calendar quarters (Q3 and Q42001) were averaged, and this average was used in all models.Dialysis vintage was defined as the duration of time between thefirst day of dialysis treatment and the first day that the patiententered the cohort. MHD patients for this study were �18 y oldand were required to have a dialysis vintage of �90 d during �1of the baseline calendar quarters.

The dose of the injectable medications administered in thedialysis clinics, including the 2 vitamin D receptor activatorsparicalcitol (Zemplar; Abbott, Abbott Park, IL) and calcitriol(Calcijex; Abbott) and recombinant human erythropoietin(rHuEPO, EPOGEN; Amgen, Inc, Thousand Oaks, CA), werecalculated for each baseline calendar quarter. The dates of deathor censoring events such as kidney transplantation or movingabroad were obtained for all patients who did not survive to June30, 2004 or who were lost before that date.

A history of diabetes mellitus was available in the database,whereas histories of tobacco smoking and preexisting comorbidconditions were obtained by linking the DaVita database to Med-ical Evidence Form 2728 (25). The preexisting comorbid condi-tions were categorized into 10 conditions: 1) ischemic heartdisease, 2) congestive heart failure, 3) the patient’s status aftercardiac arrest, 4) the patient’s status after myocardial infarction,5) pericarditis, 6) cardiac dysrhythmia, 7) cerebrovascularevents, 8) peripheral vascular disease, 9) chronic obstructivepulmonary disease, and 10) cancer.

Laboratory measures

Blood samples were drawn by using uniform techniques in allof the DaVita dialysis clinics, and they were transported to theDaVita Laboratory (DeLand, FL), typically within 24 h. Alllaboratory values were measured by automated and standardizedmethods in the DaVita Laboratory. Most laboratory values, in-cluding complete blood cell counts and serum concentrations ofurea nitrogen, creatinine, albumin, calcium, phosphorus, bicar-bonate, iron, and total iron–binding capacity (TIBC), were mea-sured monthly. Serum ferritin and intact parathyroid hormone

usually were measured quarterly. The conventional urea-kineticmeasure known as Kt/V (single pool) (a term in which K � rateof clearance; t � the amount of time of the session; and V � thearea distribution volume after hemolysis) was used to estimatethe dialysis dose. The normalized protein equivalent of totalnitrogen appearance (nPNA), also known as normalized proteincatabolic rate (21), was assessed monthly as a measure of proteinintake. Most blood samples were collected before dialysis; post-dialysis serum urea nitrogen was obtained to calculate urea ki-netics.

Statistical analysis

Analysis of variance was used to compare differences acrossgroups. The change in serum phosphorus and protein intake,represented by nPNA, was defined as the difference between the13-wk (3-mo) averaged values of the baseline calendar quarter(Q2 2001) and its subsequent quarter (Q3 2001). We used Coxproportional hazard models with 3 levels of regression adjust-ment. The first level comprised a minimally adjusted model thatincluded mortality as the outcome measure, a defined predictorsuch as baseline serum phosphorus or its change over time, andthe entry calendar quarter. The second level comprised casemix–adjusted models that included all of the above plus diabetesmellitus and the 10 preexisting comorbid conditions, history oftobacco smoking, categories of dialysis vintage (�6 mo, 6 mo–2y, 2–5 y, or �5 y), primary insurance (Medicare, Medicaid,private, or other), marital status (married, single, divorced, wid-owed and other, or unknown), the standardized mortality ratio ofthe dialysis clinic during entry quarter, dialysis dose as indicatedby Kt/V (single pool), the presence or absence of a dialysis cath-eter, and residual renal function during the entry quarter. Thethird level comprised malnutrition-inflammation complex syn-drome (MICS)–adjusted models that included all of the covari-ates in the case-mix model and �12 surrogates of nutritionalstatus and inflammation: the average doses of rHuEPO, calcit-riol, and paricalcitol and �9 laboratory variables with knownassociation with clinical outcomes in MHD patients—serum al-bumin, creatinine, TIBC, ferritin, calcium, and bicarbonate; pe-ripheral white blood cell count; lymphocyte percentage; andhemoglobin. In our view, results from the minimally adjustedmodels are likely to be underadjusted as a result of the omissionof potential confounders, whereas results from the MICS-adjusted models may be overadjusted as a result of the possibleinclusion of biological intermediates. We thus prefer to baseinferences on the case mix–adjusted models. Because we cannotbe certain which model is the best, however, we have performed3 levels of adjustments to provide the full spectrum of the results.

In models that examined the mortality predictability of base-line protein intake (nPNA) or its change over time, we did notadjust for other nutritional variables (eg, albumin, creatinine,TIBC, ot lymphocyte) that may be in the etiologic pathway. Wealso excluded Kt/V because of its mathematical correlation withnPNA (21, 26). To examine the combined association of thechanges in serum phosphorus and nPNA with death, we addedand subtracted the percentiles of the changes in these 2 measuresand analyzed the mortality predictability of the created “joint”scores. In an attempt to mitigate the effect of the regression to themean, all case mix– and MICS-adjusted models that examinedthe “change” as a mortality predictor were also controlled for thebaseline serum phosphorus and nPNA.

1512 SHINABERGER ET AL

by guest on June 7, 2013ajcn.nutrition.org

Dow

nloaded from

Nonlinear associations for continuous mortality predictorswere examined with the use of restricted cubic splines (27). Tolimit the instability of such models at extreme predictor levels,fitting was usually restricted to 95% of the predictor values byexcluding the upper and lower 1–2.5% of the values of the givenvariable.

Missing covariate data (�2% for most laboratory and demo-graphic variables and �3% for any of the 10 comorbid condi-tions) were imputed by the mean of the existing values, except forferritin and parathyroid hormone, for which the median was used.All analyses were carried out with STATA software (version10.0; Stata Corporation, College Station, TX).

RESULTS

The original 6-mo (July through December 2001) nationaldatabase of all DaVita MHD patients included 47 156 subjects.After deletion of those patients who did not continue hemodial-ysis treatment for �90 d by the middle of the each calendarquarter, 41 093 MHD patients remained for analysis, After ex-clusion of 306 patients without core data such as age or sex, 9644subjects without electronic records of phosphorus and nPNAmeasurements in both Q3 and Q4 of 2001, and 1068 patients(0.2%) whose data from the second and first calendar quarterswere the same (because the first quarter value may have been

TABLE 1Baseline data for 30 152 maintenance hemodialysis patients (July through December 2001), categorized into 4 groups according to the changes in dietaryprotein intake �represented by normalized protein equivalent of total nitrogen appearance (nPNA)� and serum phosphorus over 2 consecutive calendarquarters1

All patients

2 x 2 grouping based on the change in nPNA and phosphorus

Bothincreased

nPNA1 butphosphorus2

nPNA2 butphosphorus1

Bothdecreased P2

Patients �n (%)� 30 152 (100) 9788 (32) 6425 (21) 5322 (18) 8617 (29)Age (y) 60.7 � 15.23 60.4 � 15.2 60.1 � 15.4 60.8 � 15.0 61.5 � 14.9 �0.001Women (%) 47 47 45 47 48 �0.001Black (%) 36 36 35 37 35 0.02Hispanic (%) 16 15 17 15 17 �0.001Diabetes mellitus (%) 46 48 44 46 47 �0.001Dialysis vintage (mo) 61.2 � 41.4 59.6 � 40.3 60.7 � 42.3 63.4 � 42.6 62.1 � 41.3 �0.001Medicare as primary insurance (%) 69 69 69 70 69 0.7Married (%) 42 43 43 41 42 0.2History of comorbid states

AIDS (%) 1 1 1 1 1 0.05Cancer (%) 3 3 4 4 3 0.3Heart failure (%) 26 27 26 25 26 0.1PVD (%) 10 11 10 11 10 0.5Ischemic heart disease (%) 17 18 17 17 18 0.1Acute myocardial infarction (%) 5 5 5 5 5 0.4Chronic pulmonary disease (%) 5 5 5 5 5 0.6

Tobacco smoking (%) 5 5 5 5 5 0.8BMI (in kg/m2) 26.2 � 6.2 26.1 � 6.3 26.0 � 6.1 26.2 � 6.3 26.3 � 6.2 �0.001Kt/V (single dose) 1.56 � 0.30 1.53 � 0.30 1.52 � 0.30 1.60 � 0.31 1.59 � 0.30 �0.001nPNA (g � kg�1 � d�1) 1.01 � 0.24 0.94 � 0.22 0.96 � 0.22 1.06 � 0.24 1.08 � 0.25 �0.001

Change in nPNA (g � kg�1 � d�1) 0.02 � 0.20 0.17 � 0.15 0.14 � 0.12 �0.12 � 0.11 �0.16 � 0.13 �0.001Serum phosphorus (mg/dL) 5.8 � 1.5 5.4 � 1.4 6.1 � 1.5 5.4 � 1.4 6.1 � 1.5 �0.001

Change in phosphorus (mg/dL) 0.0 � 0.8 1.0 � 0.8 �0.7 � 0.7 0.7 � 0.6 �0.9 � 0.8 �0.001Serum albumin (g/dL) 3.83 � 0.34 3.80 � 0.36 3.84 � 0.34 3.84 � 0.34 3.85 � 0.33 �0.001

Change in albumin (mg/dL) 0.01 � 0.21 0.06 � 0.21 0.01 � 0.19 �0.01 � 0.20 �0.04 � 0.21 �0.001Serum creatinine (mg/dL) 9.7 � 3.2 9.3 � 3.1 9.8 � 3.3 9.8 � 3.2 9.9 � 3.2 �0.001

TIBC (mg/dL) 200 � 41 197 � 41 201 � 41 200 � 40 202 � 41 �0.001Ferritin (ng/mL) 691 � 498 691 � 495 663 � 478 714 � 517 699 � 505 �0.001Calcium (mg/dL) 9.3 � 0.7 9.3 � 0.7 9.3 � 0.7 9.3 � 0.7 9.3 � 0.7 �0.001Intact PTH (pg/mL) 346 � 374 317 � 341 359 � 404 351 � 369 366 � 386 �0.001Bicarbonate (mg/dL) 21.9 � 2.6 22.3 � 2.6 21.8 � 2.6 22.0 � 2.6 21.5 � 2.7 �0.001

Blood hemoglobin (g/dL) 12.0 � 1.1 11.9 � 1.2 12.1 � 1.1 11.9 � 1.2 12.1 � 1.1 �0.001WBC count (� 1000) 7.1 � 2.2 7.1 � 2.1 7.0 � 2.1 7.2 � 2.2 7.1 � 2.2 0.006Lymphocyte (%) 21 � 8 21 � 8 22 � 8 22 � 8 22 � 8 0.4

Receiving paricalcitol (%) 69 68 70 69 68 0.05Paricalcitol dose (�g/wk) 14.1 � 10.8 14.1 � 10.9 14.1 � 10.0 14.5 � 10.0 14.0 � 10.6 0.09

Receiving calcitriol (%) 5 4 5 6 5 0.03EPO dose (units/wk) 18.8 � 16.7 19.9 � 17.1 19.1 � 16.5 18.1 � 16.7 17.8 � 16.2 �0.001

1 TIBC, total iron–binding capacity; PTH, parathyroid hormone; WBC, white blood cell; EPO, erythropoietin. nPNA is also known as normalized proteincatabolized rate (nPCR).

2 P values among those who received paricalcitol (ANOVA).3 x� � SD (all such values).

PHOSPHORUS COMPARED WITH PROTEIN INTAKE 1513

by guest on June 7, 2013ajcn.nutrition.org

Dow

nloaded from

carried forward into the second quarter rather than measuredagain), the final cohort included 30 075 prevalent MHD patients.Analysis of the follow-up time started from the first day of thecalendar quarter in which the patient met inclusion criteria.

As shown in Table 1, patients were divided into 4 categoriesaccording to the concurrent fall or rise in their serum phosphorusand nPNA from the first (Q3) to the second (Q4) calendar (2001)quarter. Concordant (same-direction) change in both serum phos-phorus and nPNA occurred in 61% of patients, and discordant(opposite-direction) change occurred in the remaining 39%. Thebaseline serum phosphorus was 0.7 mg/dL lower in those whoseserum phosphorus rose, and the baseline nPNA was 0.12–0.14g � kg�1 � d�1 lower in those whose nPNA rose, whereas othervariables did not differ considerably across the 4 groups. Becausebaseline serum phosphorus and baseline nPNA were related to sub-sequent changes in serum phosphorus and nPNA, respectively, theywere included in all multivariate models. The association betweennPNA and serum phosphorus at baseline, which indicates that pa-tients with higher protein intake tended to have higher serum phos-phorus concentrations, is shown in Figure 1.

The mortality rate progressively decreased with increasingconcentrations of nPNA up to a threshold value of 1.4g � kg�1 � d�1, above which there was no further decline (Figure2A). In contrast, mortality increased both below serum phospho-rus concentrations of 3.5 mg/dL and above concentrations of 4.5mg/dL, which led to a J-shaped phenomenon (Figure 2B). In 54%of the MHD patients (n � 16 213), the 13-wk averaged nPNAconcentration rose over 6 mo, whereas in 45% it fell. As shownin Figure 2C, a fall in nPNA was incrementally associated withhigher mortality, with a calculated death risk ratio (RR) of 1.18(95% CL: 1.13, 1.22) for each 0.1 g � kg�1 � d�1 fall in nPNA.Even though a slight rise in nPNA up to 0.1 g � kg�1 � d�1 ap-peared to be associated with greater survival, further increase innPNA did not exhibit this trend. A rise in the averaged serumphosphorus over 6 mo was noted in 50% of MHD patients (n �15 112) and was associated with increased death risk, as depictedin Figure 2D; each 1 mg/dL increase in serum phosphorus wasassociated with a death RR of 1.13 (95% CL: 1.07, 1.20). Al-though a slight fall in serum phosphorus showed a trend toward

improved survival, an extreme change—whether rise or fall—inserum phosphorus of �1.0 mg/dL in 6 mo was associated with anincrease in the risk of death.

To evaluate the combined effect of the changes in serum phos-phorus and nPNA and their competing risks of death, we exam-ined the sum and the difference of percentile changes in these 2measures. For this analysis, each patient received a ranking scorebetween 0.01 and 0.99 corresponding to the percentile rank of thechange in phosphorus and a second ranking score correspondingto the change in nPNA. The sum of the 2 percentile scores yieldeda rank score for each patient between 0.02 and 1.98 (Figure 3A).The difference between the 2 percentile scores (nPNA scoreminus serum phosphorus score) yielded a rank number between�0.98 and 0.98 (Figure 3B). Mortality was highest at bothextremes of the sum of the percentiles (Figure 3A), but it declinedprogressively with increasing differences between percentiles.

We then calculated the death RR for the 4 combinations of thechanges in serum phosphorus and nPNA. As shown in Table 2and Figure 4, compared with the simultaneous rise in nPNA andserum phosphorus (reference), the rise in nPNA with a fall inserum phosphorus was associated with a lower risk, whereas theopposite difference (a rise in phosphorus with a fall in nPNA) anda concordant fall in both nPNA and phosphorus had a higher risk.Additional analyses using changes in serum albumin as a surro-gate of the nutritional status yielded similar death RRs, as shownin Table 2. Additional stratified analyses of the risk of a baselinenPNA of �1.0 g � kg�1 � d�1 (compared with �1.0 g � kg�1 �d�1) across the increments of serum phosphorus changes areshown in Figure 5; they indicate that, even though a rise in nPNAwas somewhat consistently associated with greater survival, thesurvival advantage was even stronger with simultaneous fall inserum phosphorus or with a drop in phosphorus of a larger mag-nitude.

DISCUSSION

We examined the ability of changes in serum phosphorus andnPNA, a surrogate of dietary protein intake, to predict death in30 075 MHD patients in a large dialysis organization. We foundthat a fall in the serum phosphorus concentration should promptan evaluation of dietary protein intake, and, if nPNA has alsofallen during the same period, the patient may be at greater riskof death. Another finding was that lowering serum phosphorus inan MHD patient, if it requires protein restriction, may be effec-tive in achieving the surrogate outcome but may present a risk ofdeath. These findings are most consistent with the hypothesis thatcontrolling serum phosphorus by restricting dietary protein in-take may indeed cause more harm than good in MHD patients.

In CKD patients, both disorders of minerals and bone metab-olism (MBD) (3, 28) and PEW (22, 29) are common and may berelated to exceptionally high mortality (13, 14, 21, 30, 31).Whereas both of these disorders are associated with poor clinicaloutcomes, these 2 seemingly separate disorders are usually as-sumed to have distinct and unrelated etiologies and to act throughdifferent clinical pathways.

Disorders of minerals and bone metabolism are believed todevelop with worsening hyperphosphatemia as a result of inad-equate renal phosphorus clearance, which leads to increasedactivation of fibroblast growth factor-23 and subsequent inhibi-tion of 1-�-hydroxylation of 25-hydroxyvitamin D, secondaryhyperparathyrodism, and renal osteodystrophy (1, 4, 5). PEW is

3.5

4

4.5

5

5.5

6

6.5

7

7.5

8

8.5

<0.6 0.6-0.7 0.7-0.8 0.8-0.9 0.9-1.0 1.0-1.1 1.1-1.2 1.2-1.3 1.3-1.4 ≥1.4

Dietary Protein Intake Estimated by nPNA

Ser

um P

hosp

horu

s (m

g/dL

) .

(g kg d )–1–1. .

FIGURE 1. The association between the baseline dietary protein intake,represented by 13-wk averaged normalized protein equivalent of total nitro-gen appearance (nPNA), and 13-wk averaged serum phosphorus in 30 075maintenance hemodialysis (MHD) patients in the baseline calendar quarter ofthe cohort. Error bars represent SD. P for trend � 0.001.

1514 SHINABERGER ET AL

by guest on June 7, 2013ajcn.nutrition.org

Dow

nloaded from

believed to result from inadequate protein intake (32) due toanorexia from the uremic state (33) and other conditions thatrestrict oral food ingestion in MHD patients, and it is usuallyassociated with chronic inflammation, sarcopenia, hypoalbu-minemia, and weight loss (22, 34). Hence, restricting dietaryphosphorus intake and increasing dietary protein intake are rec-ommended to most persons with advanced CKD, especiallythose undergoing MHD.

Nonetheless, the resulting prevention of disorders of mineralsand bone metabolism may be at the expense of worsening PEW,and vice versa, because higher protein intake may lead to higherserum phosphorus concentrations (Figure 1). This therapeuticconundrum is encountered frequently during the medical care ofMHD patients (35). Many physicians and dietitians are not surewhether they should reinforce dietary restrictions in their MHDpatients (restrictions that often include significant protein restric-tion, which is intended to achieve a serum phosphorus concen-tration within the recommended target zone) (35), or whetherthey should liberalize or encourage protein intake to improve

nutritional status and prevent hypoalbuminemia (which is asso-ciated with elevated risk of death). Indeed, the lower mortality inAfrican American MHD patients may be related to higher proteinintake at the expense of worsening hyperphosphatemia (36). Ourfindings further support the idea that the risk of controlling serumphosphorus by imposing dietary protein restriction may out-weigh the benefit. However, reduced protein intake may be theresult of the poor appetite that is common in these patients inde-pendent of restricting or liberalizing dietary intake (33, 37).

Our finding of a J-shaped association of baseline serum phospho-rus concentrations with survival (Figure 2B) is consistent with thefindings of several previous studies (12–14). Whereas the greaterrisk of death observed with higher phosphorus concentrations hasbiologic support (such as worsening secondary hyperparathyroid-ism and vascular calcification), the association of a low phosphorusconcentration and death risk is more poorly understood. The latterassociation may be due to the decline in protein intake and worsen-ing PEW in persons with exceptionally low phosphorus. In a recentstudy (14), low serum phosphorus concentrations showed a much

-0.2

0.0

0.2

0.4

0.6

0.8

Log

arit

hm o

f D

eat

h R

ate

Ra

tio

0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6

Estimated Protein Intake (nPNA )

-0.5

0.0

0.5

1.0

1.5

2.0

Log

arit

hm o

f D

eat

h R

ate

Ra

tio

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Serum Phosphorus (mg/dL)

� �

�

-0.2

0.0

0.2

0.4

0.6

Loa

grit

hm o

f D

eat

h R

ate

Ra

tio

-0.2 0.0 0.20.1-0.1-0.3 0.30.3

nPNA change over 2 consecutive calendar quarters

Fall in Protein Intake (nPNA)

Rise in Protein Intake (nPNA)

-0.2

0.0

0.2

0.4

0.6

Log

arit

hm o

f D

eat

h R

ate

Ra

tio

-2.0 -1.0 0.0 1.0 2.0

Serum phosphorus change over 2 consecutive calendar quarters (mg/dL)

Fall in Serum Phosphorus

Rise in Serum Phosphorus

�

(g kg d )–1–1. .

(g kg d )–1–1. .

FIGURE 2. Comparisons of the 3-y mortality predictabilities of the baseline dietary protein intake [represented by the normalized protein equivalent of totalnitrogen appearance (nPNA)] and the serum phosphorus concentration and of their changes over time by using Cox regression models in 30 075 maintenancehemodialysis (MHD) patients. The y-axis shows the logarithm of the risk ratio of all-cause mortality over 3 y of observation, ie, July 2001 through June 2004.The multivariable regression spline models are adjusted for case-mix and MICS. Dashed lines are 95% pointwise confidence levels. A: baseline, 13-wk averagednPNA. B: baseline, 13-wk averaged serum phosphorus concentration. C: changes in the 13-wk averaged nPNA over 2 consecutive calendar quarters. D: changesin the 13-wk averaged serum phosphorus concentration over 2 consecutive calendar quarters.

PHOSPHORUS COMPARED WITH PROTEIN INTAKE 1515

by guest on June 7, 2013ajcn.nutrition.org

Dow

nloaded from

stronger unadjusted association with death risk than did hyperphos-phatemia, but, after extensive multivariate adjustment, the associa-tion weakened. The remaining association may be due to the excep-tionally strong effect of PEW, which would act as a residualconfounder. Similarly, the U-shaped association in Figure 2D indi-cates that a rise in serum phosphorus over time is associated withincreasing mortality, whereas a major fall in serum phosphorus isalso associated with death risk, perhaps because of its link to PEWand low protein intake.

We also found that higher baseline nPNA was associated withlower mortality (Figure 2A), which is consistent with our previ-ous study (21). The nPNA, also known as normal protein cata-bolic rate, is a urea kinetic-based estimate of dietary protein

intake in MHD patients when minimal or no residual renal func-tion is assumed (38, 39). However, nPNA is collinear with Kt/V,because both nPNA and Kt/V are calculated by using the sameurea nitrogen concentrations (38, 39). Higher protein intakesmay indeed lead to lower mortality by virtue of improving nu-tritional status; that is why we did not adjust for additional nu-tritional markers in the models presented in Figure 2. Whereas amoderate rise in nPNA was associated with lower mortality, amore drastic rise in nPNA was associated with a paradoxicaltrend toward higher mortality, as shown in Figure 2C. The latterassociation may result from worsening hyperphosphatemia fromhigher protein intake, which once again may illustrate the coun-tervailing risks and benefits of high protein intake.

-0.1

0.0

0.1

0.2

Log

arit

hm o

f D

eat

h R

ate

Ra

tio

0.0 0.5 1.0 1.5 2.0Sum of changes in percentile nPNA and serum phosphorus

Highest Fall in nPNAHighest Fall in PO4

Highest Rise in nPNAHighest Rise in PO4

Fall in nPNAand

Fall in PO4

Rise in nPNAand

Rise in PO4

-0.2

0.0

0.2

0.4

0.6

Log

arit

hm o

f D

eat

h R

ate

Ra

tio

-1.0 -0.5 0.0 0.5 1.0

Difference between changes in percentile nPNA and serum phosphorus

Highest Fall in nPNAHighest Rise in PO4

Highest Rise in nPNAHighest Fall in PO4

Fall in nPNAbut

Rise in PO4

Rise in nPNAbut

Fall in PO4

A B

FIGURE 3. Mortality predictability of the combinations of the percentiles of the changes in dietary protein intake, represented by normalized protein equivalentof total nitrogen appearance (nPNA), and serum phosphorus (PO4) concentration in 30 075 maintenance hemodialysis (MHD) patients. The y-axis shows the logarithmof the risk ratio of all-cause mortality over 3 y based on a multivariable Cox regression spline model, adjusted for case-mix and malnutrition-inflammation complexsyndrome (MICS). Dashed lines are 95% pointwise confidence levels. Each patient received a percentile score between 0.01 and 0.99 according to the percentile rankof the change in nPNA or serum phosphorus. The sum of the 2 percentile scores for each patient resulted in a number between 0.02 and 1.98 (A), whereas the differencebetween nPNA and serum phosphorus concentration in each patient resulted in a number between �0.98 and 0.98 (B).

TABLE 2Death rate ratios (RRs) and 95% confidence levels (CL) based on Cox regression models across the 4 different combinations of changes in nutritionalstatus and serum phosphorus over 2 consecutive calendar quarters in 30 152 maintenance hemodialysis patients1

Combinations of changes

Case mix–adjusted2 Case mix– and MICS–adjusted2,3

RR (95% CL) P RR (95% CL) P

Changes in nPNA and phosphorusBoth increased 1.0 1.0nPNA1 but phosphorus2 0.90 (0.86–0.95) �0.001 0.93 (0.88–0.99) 0.01nPNA2 but phosphorus1 1.11 (1.05–1.17) �0.001 1.10 (1.04–1.16) �0.001Both decreased 1.06 (1.01–1.12) 0.02 1.09 (1.04–1.16) �0.001

Changes in albumin and phosphorusBoth increased 1.0 1.0Albumin1 but phosphorus2 0.91 (0.86–0.96) 0.002 0.92 (0.87–0.98) 0.008Albumin2 but phosphorus1 1.28 (1.22–1.37) �0.001 1.28 (1.21–1.35) �0.001Both decreased 1.24 (1.18–1.31) �0.001 1.26 (1.19–1.33) �0.001

1 MICS, malnutrition-inflammation complex syndrome; nPNA, normalized protein equivalent of total nitrogen appearance. In the upper section, thenutritional status is represented by the dietary protein intake, measured indirectly via nPNA �also known as normalized protein catabolic rate (nPCR)�; in thelower section, the nutritional status is represented by the serum albumin concentration.

2 Case mix–adjusted models included adjustment for age, sex, diabetes mellitus, standardized mortality ratio, race, dialysis vintage, primary insurance,marital status, dialysis dose, dialysis catheter, and baseline comorbid states.

3 The MICS model covariates include all case-mix covariates plus surrogates of malnutrition and inflammation, as described in Statistical analysis.

1516 SHINABERGER ET AL

by guest on June 7, 2013ajcn.nutrition.org

Dow

nloaded from

The present study was limited insofar as it was observationaland record-based, which limited direct inferences to associationsamong available measurements, rather than to effects of primaryvariables. In particular, we lacked direct measurements of dietaryprotein intake and had no data on oral medication, especiallyphosphorus binders, although data on injectable medicationswere available and adjusted for. Our analysis was further limitedto a 3-y period, although this period is crucial because almost halfof MHD patients die within 3 y of dialysis initiation. Analysesover longer periods would have to control for changes in practice

patterns. The strengths of the present study included uniformlaboratory measurements and the fact that all laboratory datawere obtained from a single facility, the large sample size, and theavailability of 3-mo averaged laboratory data.

In conclusion, it is plausible that the risk of controllingserum phosphorus by imposing dietary protein restriction mayoutweigh the benefit of phosphorus control in MHD patients.The persistent association between low protein intake andworse survival may indicate that methods other than restrict-ing protein intake should be sought to restrict dietary phos-phorus intake. More attention to nonprotein sources of phos-phorus such as food additives or highly processedconvenience foods is warranted (18). Because higher proteinintake and a concurrent decline in serum phosphorus appear tobe associated with the lowest mortality, diligent use of potentphosphorus binders may be helpful, especially if these bindersdo not lead to an excessive calcium load or pill burden, al-though binder choice remains a topic of debate (40 – 43). Inany event, our results underscore the need for clinical trials todetermine the treatment protocols that offer the greatest sur-vival advantage for MHD patients and to ascertain whethernondietary control of phosphorus or restriction of nonproteinsources of phosphorus is safer and more effective.

The authors’ responsibilities were as follows—CSS: proposed the hypoth-esis and contributed to the design of the study, collation and analyses of data,and writing and revision of the manuscript; KK-Z (principal investigator onthe supporting research grants): contributed to obtaining funding for thestudy, collation and partial analysis of data, and writing and revision of themanuscript; CPK: contributed to constructing the STATA codes for the cubicsplines analyses and to manuscript preparation; JDK and SG: contributed tothe analysis of the data and reviewed and approved the final manuscript; andDVW: contributed to the provision of data and final review and approval ofthe manuscript. KK-Z, CPK, RM, and JDK have received grants, honoraria,

D eath R ate R atio

0.8 1.0

stra

taN.a.N .

npcr22

npcr22-cm

npcr33

npcr33-cm

hhhhh

npcr44

npcr44-cm

M inim ally adjustedC ase m ix– adjusted

Both protein intake and serum PO4 increased (reference)

Protein intake increased,but serum PO4 decreased

Protein intake decreased, but serum PO4 increased

Both protein intake and serum PO4 decreased

0.9 1.1

FIGURE 4. The 3-y death risk ratios of the 4 combinations of the changes in serum phosphorus (PO4) concentration and dietary protein intake, measuredvia normalized protein equivalent of total nitrogen appearance (nPNA), over 2 consecutive calendar quarters in 30 075 maintenance hemodialysis (MHD)patients by using the Cox regression model.

Death Rate Ratio1

chan

ge

in p

ho

sph

oru

s

STRATAALL

all-cmall-micsN.a.N.<=1.5

<=1.5-CM<=1.5-MICS

-1.5to-1.0-1.5to-1.0-CM

-1.5to-1.0-MICS-1.0to-0.5

-1.0to-0.5-CM-1.0to-0.5-MICS

-0.5to0.5-0.5to0.5-CM

-0.5to0.5-MICS0.5to1.0

0.5to1.0-CM0.5to1.0-MICS

1.0to1.51.0to1.5-CM

1.0to1.5-MICS>=1.5

>=1.5-CM>=1.5-MICS

All patients

≤1.5 g/dL (fall)

-1.5 to -1.0

-1.0 to -0.5

-0.5 to +0.5 (unchanged)

+0.5 to +1.0

+1.0 to +1.5

>+1.5 g/dL (rise)

Risk of death for nPNA ≥1.0 (vs. <1.0)

UnadjustedCase mix–adjusted

0.7 0.8 0.9 1.2

(g kg d )–1–1. .

FIGURE 5. The 3-y death risk ratio of the baseline dietary protein intake[normalized protein equivalent of total nitrogen appearance (nPNA)] �1.0(compared with �1.0 g � kg�1 � d�1) across the increments of changes inserum phosphorus concentration over 2 consecutive calendar quarters in30 075 maintenance hemodialysis (MHD) patients by using the Cox regres-sion model.

PHOSPHORUS COMPARED WITH PROTEIN INTAKE 1517

by guest on June 7, 2013ajcn.nutrition.org

Dow

nloaded from

or both from any or all of Amgen (manufacturer of Sensipar), Abbott (man-ufacturer of Calcijex and Zemplar), Genzyme (manufacturer of Hectoral,Renagel, and RenVela), and Shire (manufacturer of Fosrenol). None of theother authors had a personal or financial conflict of interest.

REFERENCES1. Hruska KA, Teitelbaum SL. Renal osteodystrophy. N Engl J Med 1995;

333:166–74.2. Moe S, Drueke T, Cunningham J, et al; Kidney Disease: Improving

Global Outcomes (KDIGO). Definition, evaluation, and classification ofrenal osteodystrophy: a position statement from Kidney Disease: Im-proving Global Outcomes (KDIGO). Kidney Int 2006;69:1945–53.

3. Levin A, Bakris GL, Molitch M, et al. Prevalence of abnormal serumvitamin D, PTH, calcium, and phosphorus in patients with chronic kid-ney disease: results of the study to evaluate early kidney disease. KidneyInt 2007;71:31–8.

4. Gutierrez O, Isakova T, Rhee E, et al. Fibroblast growth factor-23 mit-igates hyperphosphatemia but accentuates calcitriol deficiency inchronic kidney disease. J Am Soc Nephrol 2005;16:2205–15.

5. Gupta A, Winer K, Econs MJ, Marx SJ, Collins MT. FGF-23 is elevatedby chronic hyperphosphatemia. J Clin Endocrinol Metab 2004;89:4489–92.

6. Hernandez JD, Wesseling K, Salusky IB. Role of parathyroid hormoneand therapy with active vitamin D sterols in renal osteodystrophy. SeminDial 2005;18:290–5.

7. Hruska KA, Saab G, Mathew S, Lund R. Renal osteodystrophy, phos-phate homeostasis, and vascular calcification. Semin Dial 2007;20:309–15.

8. Achinger SG, Ayus JC. Left ventricular hypertrophy: is hyperphos-phatemia among dialysis patients a risk factor? J Am Soc Nephrol 2006;17:S255-61.

9. Tonelli M, Sacks F, Pfeffer M, Gao Z, Curhan G. Relation betweenserum phosphate level and cardiovascular event rate in people withcoronary disease. Circulation 2005;112:2627–33.

10. Voormolen N, Noordzij M, Grootendorst DC, et al. High plasma phos-phate as a risk factor for decline in renal function and mortality inpre-dialysis patients. Nephrol Dial Transplant 2007;22:2909–16.

11. Kestenbaum B, Sampson JN, Rudser KD, et al. Serum phosphate levelsand mortality risk among people with chronic kidney disease. J Am SocNephrol 2005;16:520–8.

12. Block GA, Hulbert-Shearon TE, Levin NW, Port FK. Association ofserum phosphorus and calcium x phosphate product with mortality riskin chronic hemodialysis patients: a national study. Am J Kidney Dis1998;31:607–17.

13. Block GA, Klassen PS, Lazarus JM, Ofsthun N, Lowrie EG, ChertowGM. Mineral metabolism, mortality, and morbidity in maintenance he-modialysis. J Am Soc Nephrol 2004;15:2208–18.

14. Kalantar-Zadeh K, Kuwae N, Regidor DL, et al. Survival predictabilityof time-varying indicators of bone disease in maintenance hemodialysispatients. Kidney Int 2006;70:771–80.

15. Schwarz S, Trivedi BK, Kalantar-Zadeh K, Kovesdy CP. Association ofdisorders in mineral metabolism with progression of chronic kidneydisease Clin J Am Soc Nephrol 2006;1:825–31.

16. Kovesdy CP, Kalantar-Zadeh K. Serum phosphorus and the risk ofprogression of chronic kidney disease. Nephrol Dial Transplant 2007;(22):3679–80.

17. Kovesdy UP, Kalantar-Zadeh K. Bone and mineral disorders in pre-dialysis CKD. Int Urol Nephrol 2008;40(2):427–40.

18. Uribarri J. Phosphorus homeostasis in normal health and in chronickidney disease patients with special emphasis on dietary phosphorusintake. Semin Dial 2007;20:295–301.

19. Block G, Uribarri J, Coladonato JA, et al. How should hyperphos-phatemia be managed in dialysis patients? Semin Dial 2002;15:315–28.

20. Nolan CR, Qunibi WY. Treatment of hyperphosphatemia in patientswith chronic kidney disease on maintenance hemodialysis. Kidney IntSuppl 2005;(95):S13–20.

21. Shinaberger CS, Kilpatrick RD, Regidor DL, et al. Longitudinal asso-ciations between dietary protein intake and survival in hemodialysispatients. Am J Kidney Dis 2006;48:37–49.

22. Kalantar-Zadeh K, Ikizler TA, Block G, Avram MM, Kopple JD.Malnutrition-inflammation complex syndrome in dialysis patients:causes and consequences. Am J Kidney Dis 2003;42:864–81.

23. Wu DY, Shinaberger CS, Regidor DL, McAllister CJ, Kopple JD,Kalantar-Zadeh K. Association between serum bicarbonate and death inhemodialysis patients: is it better to be acidotic or alkalotic? Clin J AmSoc Nephrol 2006;1:70–8.

24. Kovesdy CP, Regidor DL, Mehrotra R, et al. Serum and dialysate po-tassium concentrations and survival in hemodialysis patients. Clin J AmSoc Nephrol 2007;2:999–1007.

25. Longenecker JC, Coresh J, Klag MJ, et al. Validation of comorbidconditions on the end-stage renal disease medical evidence report: theCHOICE study. Choices for Healthy Outcomes in Caring for ESRD.J Am Soc Nephrol 2000;11:520–9.

26. Kalantar-Zadeh K, Supasyndh O, Lehn RS, McAllister CJ, Kopple JD.Normalized protein nitrogen appearance is correlated with hospitaliza-tion and mortality in hemodialysis patients with Kt/V greater than 1.20.J Ren Nutr 2003;13:15–25.

27. Durrleman S, Simon R. Flexible regression models with cubic splines.Stat Med 1989;8:551–61.

28. Moe SM, Drueke T, Lameire N, Eknoyan G. Chronic kidney disease-mineral-bone disorder: a new paradigm. Adv Chronic Kidney Dis 2007;14:3–12.

29. Fouque D, Kalantar-Zadeh K, Kopple J, et al. A proposed nomenclatureand diagnostic criteria for protein-energy wasting in acute and chronickidney disease. Kidney Int 2007;73:391–9. (Published erratum appearsin Kidney Int 2008;74:393.)

30. Melamed ML, Eustace JA, Plantinga L, et al. Changes in serum calcium,phosphate, and PTH and the risk of death in incident dialysis patients: alongitudinal study. Kidney Int 2006;70:351–7.

31. Kalantar-Zadeh K, Kilpatrick RD, Kuwae N, et al. Revisiting mortalitypredictability of serum albumin in the dialysis population: time depen-dency, longitudinal changes and population-attributable fraction. Neph-rol Dial Transplant 2005;20:1880–8.

32. Ikizler TA. Protein and energy: recommended intake and nutrient sup-plementation in chronic dialysis patients. Semin Dial 2004;17:471–8.

33. Kalantar-Zadeh K, Block G, McAllister CJ, Humphreys MH, KoppleJD. Appetite and inflammation, nutrition, anemia and clinical outcomein hemodialysis patients. Am J Clin Nutr 2004;80:299–307.

34. Ikizler TA. Effects of hemodialysis on protein metabolism. J Ren Nutr2005;15:39–43.

35. Martin CJ, Reams SM. The renal dietitian’s role in managing hyper-phosphatemia and secondary hyperparathyroidism in dialysis patients: anational survey. J Ren Nutr 2003;13:133–6.

36. Kalantar-Zadeh K, Kovesdy CP, Derose SF, Horwich TB, Fonarow GC.Racial and survival paradoxes in chronic kidney disease. Nat Clin PractNephrol 2007;3:493–506.

37. Carrero JJ, Qureshi AR, Axelsson J, et al. Comparison of nutritional andinflammatory markers in dialysis patients with reduced appetite. Am JClin Nutr 2007;85:695–701.

38. Kloppenburg WD, Stegeman CA, de Jong PE, Huisman RM. Relatingprotein intake to nutritional status in haemodialysis patients: how tonormalize the protein equivalent of total nitrogen appearance (PNA)?Nephrol Dial Transplant 1999;14:2165–72.

39. Kopple JD, Gao XL, Qing DP. Dietary protein, urea nitrogen appearanceand total nitrogen appearance in chronic renal failure and CAPD pa-tients. Kidney Int 1997;52:486–94.

40. Kovesdy CP, Mehrotra R, Kalantar-Zadeh K. Battleground: chronickidney disorders mineral and bone disease—calcium obsession, vitaminD, and binder confusion. Clin J Am Soc Nephrol 2008;3:168–73.

41. Shantouf R, Budoff MJ, Ahmadi N, Tiano J, Flores F, Kalantar-Zadeh K.Effects of sevelamer and calcium-based phosphate binders on lipid andinflammatory markers in hemodialysis patients. Am J Nephrol 2007;28:275–9.

42. Cozzolino M, Brancaccio D. Hyperphosphatemia in dialysis patients:the therapeutic role of lanthanum carbonate. Int J Artif Organs 2007;30:293–300.

43. Qunibi WY, Nolan CR. Treatment of hyperphosphatemia in patientswith chronic kidney disease on maintenance hemodialysis: results of theCARE study. Kidney Int Suppl 2004;(90):S33–8.

1518 SHINABERGER ET AL

by guest on June 7, 2013ajcn.nutrition.org

Dow

nloaded from