jurnal imey

8/9/2019 jurnal imey

1/9

Glycated Albumin and Risk of Death and

Hospitalizations in Diabetic Dialysis PatientsBarry I. Freedman,* Lilian Andries,* Zak K. Shihabi,† Michael V. Rocco,* Joyce R. Byers,* Cesar Y. Cardona,* Michael A. Pickard,‡ David L. Henderson,‡ Margie V. Sadler,‡ Leah M. Courchene,‡ Jean R. Jordan,§

Somer S. Balderston,§ Angie D. Graham,§ Vicki L. Mauck,§ Gregory B. Russell, and Anthony J. Bleyer*

SummaryBackground and objectives Relative to hemoglobin (Hb) A1c, glycated albumin (GA) more accurately reflectsglycemic control in patients with diabetes mellitus and ESRD. We determined the association between GA,HbA1c, and glucose levels with survival and hospitalizations in diabetic dialysis patients.

Design, setting, participants, & measurements Quarterly GA levels were measured for up to 2.33 years in444 prevalent patients with diabetes and ESRD. Proportional hazard time-dependent covariate models were

computed with adjustment for demographic characteristics, comorbidities, and laboratory variables. Similaranalyses were performed for available HbA

1c and monthly random serum glucose determinations.

Results The participants were 53% male, 54% African American, 43% Caucasian, 90% on hemodialysis, witha mean (SD) age of 62 (12) years and median follow-up duration of 2.25 years. Mean SD GA and HbA1cwere 21.5% 6.0% (median 20.4%), and 6.9% 1.6% (median 6.6%), respectively. There were 156 deathsduring the observation period. In best-fit models, predictors of death included increasing GA, increasingage, presence of peripheral vascular disease, decreasing serum albumin, and decreasing hemoglobin con-centrations. HbA1c and random serum glucose concentrations were not predictive of survival. IncreasingGA levels were associated with hospitalization in the 17 days after measurement, whereas HbA1c was not.

Conclusions In contrast to the HbA1c and random serum glucose values, GA accurately predicts the risk of death and hospitalizations in patients with diabetes mellitus and ESRD. The GA assay should be consid-

ered by clinicians who care for patients with diabetes on dialysis.Clin J Am Soc Nephrol 6: 1635–1643, 2011. doi: 10.2215/CJN.11491210

IntroductionThe results of the hemoglobin (Hb) A1c assay arefalsely low in diabetic patients with ESRD on hemo-

dialysis and peritoneal dialysis (1–4). In addition,HbA1c appears to be less precise in patients with

advanced stages of chronic kidney disease (CKD),particularly in anemic subjects receiving erythropoie-

tin (5). This effect likely reflects shortened red bloodcell (RBC) survival in advanced kidney disease, with

reduced time available for glucose and hemoglobin tochemically interact (6). In contrast, the glycated albu-

min (GA) level more accurately reflects recent glyce-mic control on the basis of fasting blood sugar con-centrations in diabetic patients with ESRD on dialysis

and advanced CKD (1–5).

Whether HbA1c correlates with the risk of death or

hospitalization in patients with diabetes and ESRDremains controversial (7,8). The appropriate measure

to use to establish the presence of hyperglycemiaremains an important consideration, because diabetes

is the leading cause of kidney failure worldwide.

Improved glucose control reduces CKD progression

rates and has beneficial effects on diabetic retinopathyand possibly cardiovascular disease (9–11).

This study assessed relationships between GA withpatient survival and hospitalizations in subjects withdiabetes on dialysis. Participants were followed lon-gitudinally for up to 2.33 years at dialysis facilitiesoperated by Wake Forest University Health Sciences(WFUHS) with identical dialysis management proto-cols.

Materials and MethodsPatients

All of the patients with diabetes mellitus and ESRDtreated at WFUHS-operated outpatient dialysis facil-ities in North Carolina were invited to participate in alongitudinal, observational study assessing the asso-ciation between GA and the dialysis outcomes of patient survival and hospitalizations. Patients withadvanced HIV infection or malignancies with esti-mated survival of less than 2 years were excluded.Enrollment was performed between January 1, 2007and June 30, 2007. Thereafter, refrigerated serum sam-

ples were sent from Meridian Laboratory Corporation

Departments of *Internal Medicine-Nephrology,†Pathology, andBiostatistical Sciences,Wake Forest School of Medicine, Winston-Salem, North Carolina;‡Meridian LaboratoryCorporation, Charlotte,North Carolina; and§Wake Forest Out-Patient DialysisProgram, Winston-Salem, North Carolina

Correspondence:Dr. Barry I. Freedman,Section on Nephrology,Wake Forest School of Medicine, MedicalCenter Boulevard,Winston-Salem, NC27157-1053. Phone:336-716-6192; Fax:336-716-4318; E-mail:[email protected]

www.cjasn.org Vol 6 July, 2011 Copyright © 2011 by the American Society of Nephrology 1635

Article


2/9

(Charlotte, NC) to Wake Forest quarterly through Septem- ber 15, 2009. The serum samples were centrifuged andstored at 80°C for determination of GA within 180 daysof receipt. Asahi Kasei Pharma Corporation (Tokyo, Japan)was the study sponsor.

All of the subjects were actively prescribed oral hypo-

glycemic agents and/or insulin and had clinical diagnosesof type 1 or type 2 diabetes mellitus. No medical interven-tions were made on the basis of GA results, because thetreating physicians were not provided the results of thisresearch test. Demographic and medical information wasprovided by participants; comorbidities were extractedfrom the Centers for Medicare and Medicaid Services(CMS) 2728 form as reported at dialysis initiation. Studysubjects provided written informed consent, and the studywas approved by the institutional review board at theWake Forest School of Medicine.

Participants had routine monthly random serum glucosedeterminations. HbA1c was evaluated when ordered bytreating physicians. The results of quarterly hemoglobin,serum phosphorus, and serum albumin (bromocresol-green[BCG] method) concentrations were recorded throughout thestudy for use as covariates in the analyses. Patients receivinga kidney transplant or transferring from WFUHS dialysisfacilities were censored at final follow-up. All hospitalizationsand deaths were recorded.

The study sponsor approved the investigator-initiatedstudy design and provided GA assay kits and reagents.The data sets were maintained, and statistical analyseswere performed at Wake Forest by the investigative team;interim results were not shared with the sponsor. Thesponsor did not participate in data analysis or manuscriptpreparation. Asahi Kasei Pharma Corporation received fi-

nal results on August 14 2010, 10 months after the lastfollow-up.

GA and HbA1c

AssaysSerum albumin concentrations for determination of

GA were measured at Wake Forest WFUHS using the bromocresol-purple (BCP) assay (modified BCP methodusing the LUCICA® GA-L kit; Asahi Kasei Pharma Cor-poration) calibrated to the standards of the College of American Pathologists. GA was measured using the LU-CICA® GA-L kit (Asahi Kasei Pharma Corporation) onserum samples. This kit uses an enzymatic method forconverting GA to glycated amino acids. The glycatedamino acids are oxidized with the formation of hydrogenperoxide, which is coupled to a dye yielding a purple-bluecolor. The GA analysis was performed on the automatedADVIA 1650 instrument (Siemens Medical Solutions Diag-nostics, Tarrytown, NY). As per the manufacturer’s in-structions, the GA (%) is computed as (GA modifiedBCP serum albumin) 100 ( 1.14) (2.9).

HbA1c

was analyzed at Meridian Laboratory using theTosoh Automated Glycohemoglobin Analyzer HLC723-G8(Tokyo, Japan). This method uses nonporous ion-exchangeHPLC for precise separation of the stable form of HbA1cfrom other hemoglobin fractions. Meridian Laboratorymeasured routine monthly serum albumin concentrations

with the BCG method on the Beckman-Olympus AU2700chemistry analyzer (Center Valley, PA). BCG serum albu-

mins were used as a covariate in the analyses, whereasmodified BCP serum albumins, more precise in nephrop-athy, were used to compute GA.

Statistical Analyses

The primary study outcome was the all-cause mortalityrate. Secondary outcomes included all-cause hospitaliza-tion (cardiovascular disease, infectious complications, pe-ripheral vascular disease (PVD), access-related, and other)and all-cause hospitalization rate. Ascertainment of deathand hospitalizations were based upon monthly logs atWFUHS dialysis units. Whenever patients failed to re-port for outpatient dialysis treatments, facility staff de-termined the reason by contacting patients, relatives,hospitals, and physicians. This provided a full account-ing of deaths and hospitalizations. Reported deaths wereconfirmed by matching with completed CMS ESRDDeath Notification 2746 forms. Hospitalizations were

confirmed by matching dates of absence with dischargesummaries. These techniques allowed for close trackingof deaths and missed treatments.

Descriptive statistics were computed, including themeans and SD for continuous measures and frequenciesand proportions for categorical data. To compare propor-tions of hospitalization between quintiles within the threeprimary outcomes, a chi-squared test was used. In this testand in modeling of survival data, each analysis was con-ceptualized as a separate outcome. As such, no adjustmentwas made in our significance level for multiple compari-sons. Cox regression models using PROC PHREG (version9.2; SAS Software, Cary, NC) were utilized to create pro-

portional hazards models with time-dependent covariatesfor the outcomes of overall survival time and hospitaliza-tion events. Participant outcomes were censored at kidneytransplantation or transfer of care from WFUHS facilities.Because the time-dependent covariates could change atirregularly spaced time points (all of the laboratory mea-sures were not gathered at the same intervals), these mod-els were built with that structure in mind. For covariates, if a measure was not gathered concurrently with the out-come measure, the closest preceding measurement was brought forward and used in the analysis. For the outcomevariables (GA and HbA1c), values were never broughtforward. SAS PROC MIXED was utilized; when the out-

come measure was missing, that observation was not usedin the analysis.

A priori, analysis was performed using univariate andspecific multivariate models. The multivariate models forGA, HbA

1c, and glucose levels were constructed sequen-

tially using only demographic characteristics (age at en-rollment, gender, race, and body mass index [BMI]), thenadding comorbidities (malignant disease [excluding non-melanoma skin cancer], cardiovascular disease, and PVDon the basis of the CMS 2728 form completed at dialysisinitiation), and finally by adding laboratory values (quar-terly BCG serum albumin, serum phosphorus, and hemo-globin from Meridian Laboratory). After full adjustment, a

best-fit model was computed for the outcomes of patientsurvival and hospitalization.

1636 Clinical Journal of the American Society of Nephrology


3/9

ResultsFour hundred sixty diabetic individuals performing he-

modialysis or peritoneal dialysis were approached, and444 (96.52%) agreed to participate. Demographic and lab-oratory characteristics at enrollment are provided in Table 1,with comorbidities at dialysis initiation.

Study follow-up encompassed 9574 patient months,with mean (SD) participant follow-up 25.75 (4.77) months(median follow-up, 27.25 months [range, 0.56 to 27.78]).During the study, 3050 GA assays (median 8 per subject,mean (SD) 6.87 (2.85) per subject), 2652 quarterly randomserum glucose concentrations (median 7 per subject, mean(SD) 6.09 (2.45) per subject), and 1747 HbA1c measure-ments (median 3 per subject, mean (SD) 3.93 (2.85) persubject) were recorded. One hundred fifty-six deaths(35.14%) were recorded. Forty-one participants were cen-sored: 23 were censored after kidney transplantation (aftera mean duration of 13.27 months), 16 moved from the areaand were lost to follow-up (after a mean duration of 16.57months), and 2 regained kidney function (after a mean of 6.10 months).

The mean (SD) median GA level during follow-up was21.52% (6.0%) 20.36% (range, 10.6% to 50.88%; interquar-tile range, 7.34%; 25th and 75th percentiles, 17.3% and24.62%). The mean intrapatient coefficient of variationfor patients with at least three GA measures (397 pa-tients) was 13.35% (range, 0.6% to 49.93%). The mean

(SD) median HbA1c level during follow-up was 6.91%(1.6%) 6.6% (range, 4.3% to 22.48%; interquartile range,1.74%, 25th and 75th percentiles, 5.9% and 7.61%). Themean intrapatient coefficient of variation for patientswith at least three HbA1c measures (276 patients) was11.07% (range, 1.5% to 142.17%). The mean (SD) randomglucose concentration during follow-up was 178.19(91.24) mg/dl. The mean intrapatient coefficient of vari-ation for patients with at least three glucose measures(387 patients) was 30.51% (range, 0.9% to 105.05%). Themost intrapatient variability in longitudinal measure-ments was observed with the random blood glucoseconcentration, with lower variability seen with HbA1cand GA. Figure 1 reveals the relationship between GAand HbA1c results in patients with diabetes and ESRD.

Table 1. Demographic and laboratory characteristics of study participants at enrollment

Variable n Percentage Mean (SD) Median

Age (years) 444 62.3 (12.4) 63Gender

female 210 47.3male 234 52.7

RaceAfrican American 238 53.6Caucasian 191 43.0

other 15 3.4Dialysishemodialysis 401 90.3peritoneal 43 9.7

Diabetes type1 38 8.62 406 91.4

ACE/ARBno 229 51.6yes 215 48.4

EPO useno 28 6.3yes 416 93.7

BMI (kg/m2) 444 29.4 (7.3) 28.6Diabetes duration (years) 444 18.5 (10.8) 18.0

ESRD duration (years) 444 2.9 (2.6) 2.2HbA1c (%) 443 6.8 (1.5) 6.5GA (%) 443 19.1 (6.1) 17.7Serum albumin (g/dl) 444 3.9 (0.4) 4.0Hemoglobin (g/dl) 444 11.8 (1.2) 11.8URR (%) 401 72.5 (9.6) 73.0Ferritin (ng/ml) 444 579 (407) 496Iron saturation (%) 444 28.0 (13.0) 25.0BUN (mg/dl) 444 47.7 (16.4) 47.0Serum creatinine (mg/dl) 444 7.9 (2.8) 7.7Serum aspartate aminotransferase (units/L) 444 19.3 (8.1) 17.0Serum glucose (mg/dl) 444 169 (63) 158Serum phosphorus (mg/dl) 408 5.9 (2.6) 5.7

ACE/ARB, angiotensin-converting enzyme inhibitor/angiotensin receptor blocker; BMI, body mass index; BUN, blood ureanitrogen; EPO, erythropoietin; GA, glycated albumin; Hb, hemoglobin; URR, urea-reduction ratio.

Clin J Am Soc Nephrol 6: 1635–1643, July, 2011 Glycated Albumin in ESRD, Freedman et al. 1637


4/9

The association of GA levels and patient survival withpredefined univariate and multivariate models is shown inTable 2. In the univariate model, an increasing GA levelwas associated with a nonsignificant increased risk of death (P 0.12). When age, gender, race, and BMI areentered into the model, there was a statistically significantassociation between higher GA levels and risk of mortality(P 0.02). The addition of comorbid conditions did notchange the level of significance. In the full model, includ-ing comorbid conditions, demographic characteristics, andlaboratory values, there was a nonsignificant trend towardan association between higher GA levels and higher mor-tality rates (P 0.07). Finally, in the best-fit model for

mortality, only GA, age, PVD, hemoglobin, and serumalbumin were significant predictors of patient survival. Inthis model, the risk of mortality increased by 14% for each5% rise in GA. We then evaluated the effect of GA onsurvival solely in the sample of diabetic patients perform-ing hemodialysis. In this analysis, P values for the associ-ation of GA with survival were P 0.10 (unadjusted), P 0.015 (adjusted for demographic characteristics), P 0.013(adjusted for demographics, comorbidities, and dialysisvintage), P 0.044 (fully adjusted), and P 0.02 (best-fitmodel).

The relationship between random blood glucose concen-trations and mortality is shown in Table 3. Glucose con-centration was not significantly associated with survival inany model. The relationship between HbA1c levels andmortality is shown in Table 4. In the univariate model,higher HbA

1c values were associated with lower mortality

(P 0.017). The association between HbA1c and mortalitywas no longer statistically significant when other patientparameters were entered into the model.

During the observation period, 86.71% of participantshad at least one hospitalization. A total of 1428 individualhospitalizations encompassing 10,773 inpatient days wererecorded. Among patients with at least one hospitalization,there was an average of 2.76 2.73 hospitalizations perpatient year. The median number of days of hospitalizationper year was 10.55 days (25th and 75th percentiles, 4.15 and

29.46 days, respectively). Comparison of the characteristicsof hospitalized versus nonhospitalized patients is shown in

G A

10

20

30

40

50

60

HbA1c

4 6 8 10 12 14 16 18

Figure 1. | The relationship between glycated albumin (GA) (%)and hemoglobin (Hb) A

1c (%) in patients with diabetes on dialysis.

T a b

l e 2 .

G l y c a t e d a l b u m i n G A a n d p a t i e n t s

u r v i v a l o n d i a l y s i s

V a r i a b l e

P , H R ( 9 5 % C I )

S i n g l e V a r i a b l e

( n

4 4 4 , 3 0 5 0 o b s )

A d d D e m

o g r a p h i c s

( n

4 4 3 , 3 0 3 7 o b s )

A d d C o m o r b i d i t i e s

( n

4 4 3 , 3 0 3 7 o b s )

A d d H e m o g l o b i n ,

P h o s p h o r u s , A l b u m i n

( n

4 2 5 , 3 0 1 3 o b s )

B e

s t - f i t M o d e l

( n

4 2 7 , 3 0 1 9 o b s )

G A

( H R

5 % i n c r e a s e )

0 . 1 2 , 1 . 0 9 ( 0 . 9 8 t o 1 . 2 1 )

0 . 0 2 , 1 . 1 4 ( 1 . 0 2 t o 1 . 2 8 )

0 . 0 2 , 1 . 1 5 ( 1 . 0 2 t o 1 . 2 9 )

0 . 0 7 , 1 . 1 2 ( 0 . 9 9 t o 1 . 2 7 )

0 . 0 3 , 1

. 1 4 ( 1 . 0 1 t o 1 . 2 8 )

A g e ( H R

1 - y e a r i n c r e a s e )

0 . 0 0 1 , 1 . 0 3

( 1 . 0 2 t o 1 . 0 5 )

0 . 0 0 1 , 1 . 0 3 ( 1 . 0 2 t o 1 . 0 5 )

0 . 0 0 1 , 1 . 0 3 ( 1 . 0 2 t o 1 . 0 5 )

0 . 0 0 1 ,

1 . 0 3 ( 1 . 0 2 t o 1 . 0 5 )

G e n d e r ( f e m a l e

r e f e r e n c e )

0 . 4 0 , 1 . 1 5 ( 0 . 8 3 t o 1 . 5 8 )

0 . 2 3 , 1 . 2 2 ( 0 . 8 8 t o 1 . 6 9 )

0 . 5 4 , 1 . 1 1 ( 0 . 8 0 t o 1 . 5 5 )

R a c

e ( A f r i c a n A m e r i c a n v e r s u s C a u c a s i a n )

0 . 0 8 , 0 . 7 0 ( 0 . 5 1 t o 0 . 9 6 )

0 . 0 6 , 0 . 6 8 ( 0 . 4 9 t o 0 . 9 4 )

0 . 1 2 , 0 . 7 1 ( 0 . 5 1 t o 0 . 9 8 )

B M

I ( H R

1 k g / m

2

i n c r e a s e )

0 . 0 0 8 , 0 . 9 6

( 0 . 9 4 t o 0 . 9 9 )

0 . 0 2 , 0 . 9 7 ( 0 . 9 4 t o 0 . 9 9 )

0 . 1 1 , 0 . 9 8 ( 0 . 9 5 t o 1 . 0 1 )

C V D ( n o v e r s u s y e s )

0 . 9 8 , 1 . 0 0 ( 0 . 7 2 t o 1 . 4 0 )

0 . 6 8 , 1 . 0 8 ( 0 . 7 7 t o 1 . 5 1 )

C a n c e r ( n o v e r s u s y e s )

0 . 5 0 , 1 . 2 3 ( 0 . 6 7 t o 2 . 2 6 )

0 . 3 9 , 1 . 3 1 ( 0 . 7 1 t o 2 . 4 1 )

P V D ( n o v e r s u s y e s )

0 . 0 0 1 , 0 . 4 9 ( 0 . 3 4 t o 0 . 7 0 )

0 . 0 0 1 , 0 . 5 1 ( 0 . 3 5 t o 0 . 7 4 )

0 . 0 0 1 ,

0 . 5 1 ( 0 . 3 6 t o 0 . 7 3 )

V i n

t a g e ( H R

1 - y e a r i n c r e a s e )

0 . 7 0 , 1 . 0 1 ( 0 . 9 5 t o 1 . 0 8 )

0 . 4 9 , 1 . 0 2 ( 0 . 9 6 t o 1 . 0 9 )

H e m o g l o b i n ( H R

1 g / d l i n c r e a s e )

0 . 0 0 1 , 0 . 8 2 ( 0 . 7 3 t o 0 . 9 1 )

0 . 0 0 1 ,

0 . 8 2 ( 0 . 7 4 t o 0 . 9 3 )

P h o

s p h o r u s ( H R

1 m g / d l i n c r e a s e )

0 . 9 1 , 0 . 9 9 ( 0 . 9 1 t o 1 . 0 9 )

A l b

u m i n ( H R

1 g / d l i n c r e a s e )

0 . 0 0 1 , 0 . 3 7 ( 0 . 2 7 t o 0 . 5 2 )

0 . 0 0 1 ,

0 . 3 5 ( 0 . 2 6 t o 0 . 4 8 )

B M

I , b o d y m a s s i n d e x ; C I , c o n f i d e n c e i n t e r v

a l ; C V D , c a r d i o v a s c u l a r d i s e a s e ; G A , g l y c a t e d a l b u m i n ; H R ,

h a z a r d r a t i o ; o b s , n u m b e r o f u n i q u e o b s e r v a t i o n s ; P V D , p e r i p h e r a l v a s c u l a r

d i s e a s e .



5/9


6/9

Table 5. Patients who were not hospitalized had higher baseline serum albumin concentrations. The mean GA,HbA1c, and random serum-glucose concentrations aver-aged over the course of the study did not differ signifi-cantly in patients with versus without hospitalizations.

The relationship between measures of glycemic controland hospitalizations is shown in Table 6. In the univariatemodel, increasing GA was associated with risk of hospi-talizations within 17 days of measurement (P 0.003).When age, gender, race, and BMI are entered into themodel, there was a statistically significant association be-

tween higher GA and risk of hospitalization (P

0.005).The addition of comorbid conditions did not change thelevel of significance. In the full model, including comorbidconditions, demographic characteristics, and laboratoryvalues, there was a strong significant association betweenhigher GA levels and risk of hospitalization (P 0.02). Inthe best-fit model, only GA, gender, serum phosphorus,and PVD were associated with hospitalizations (P 0.005).Random glucose levels were predictive of hospitalizationin the univariate (P 0.03) and all multivariate models(e.g. P 0.01 full model; P 0.01 best-fit model). HbA1cwas not predictive of hospitalization in any model.

Figure 2 reveals the graded relationship between higherGA levels and higher hospitalization rates within 17 daysof measurement (similar results seen within 30 days of measurement; data not shown). The rate of hospitaliza-tions per quintile of GA differed significantly between thelowest and highest quintiles (5.90% versus 9.67%, respec-tively; P 0.02) and was also significantly different com-paring the second lowest with the highest quintiles (6.23%versus 9.67%, respectively; P 0.03). No graded relation-ships were seen for hospitalization rates with either HbA1cor blood glucose levels (Figure 2). Comparison of hospi-talization rates by quintiles of HbA

1c level or blood glucose

level did not show any statistically significant differences.The association of the outer quintiles of HbA1c, represent-ing the lowest and highest hospitalization rates (6.7% and

8.82%, respectively), with hospitalization rates was nonsig-nificant (P 0.32). Similar results were seen for bloodglucose levels. The greatest difference in hospitalizationrates was seen with the second (7.29%) and highest quintile(9.42%); the association between these quintiles of bloodglucose levels and hospitalization rates was nonsignificant(P 0.22). These data provide strong evidence of theimportance of glycemic control assessed using GA withhospitalization rates.

DiscussionOur report is the first to prospectively evaluate the

relationship between longitudinal measurements of GAon survival and hospitalizations in a relatively largesample of African American and Caucasian patientswith ESRD and diabetes mellitus. Higher GA levels weresignificantly associated with poorer survival and in-creased hospitalization rates in dialysis patients withdiabetes mellitus using multivariate analyses. In theseanalyses, risk of death was also predicted by older age,PVD, and lower hemoglobin and serum albumin con-centrations. We did not find an association betweenpatient survival and glycemic control measured by ei-ther HbA1c or random blood glucose levels. Our find-ings are complimentary to those of Fukuoka et al. (12),who reported that high GA values (29%) at dialysis

initiation were associated with poorer patient survival in98 Japanese patients with diabetes mellitus followed for

Table 5. Demographic and laboratory characteristics of hospitalized and nonhospitalized study participants

Variable n % Mean (SD) P

Age (years)hospitalized 385 62.5 (12.4) 0.30nonhospitalized 59 60.8 (12.6)

Gender 0.33female

hospitalized 186 48.3nonhospitalized 24 40.7

malehospitalized 199 51.7nonhospitalized 35 59.3

Race 0.81African American


Caucasianhospitalized 208 54.2nonhospitalized 30 50.9

otherhospitalized 12 3.1nonhospitalized 2 3.4

Dialysis 0.99hemodialysis


peritonealhospitalized 52 13.5nonhospitalized 8 13.6

BMI (kg/m2) 0.53hospitalized 385 29.4 (7.5)nonhospitalized 59 30.0 (6.1)

ESRD duration (years) 0.97hospitalized 385 2.9 (2.6)nonhospitalized 59 2.9 (2.6)

Serum albumin (g/dl) 0.01hospitalized 385 3.9 (0.4)nonhospitalized 59 4.1 (0.3)

Hemoglobin (g/dl) 0.37hospitalized 385 11.8 (1.2)nonhospitalized 59 11.9 (1.1)

Serum phosphorus(mg/dl)

0.88

hospitalized 378 5.7 (1.7)nonhospitalized 58 5.8 (1.6)

GA (%)* 0.39hospitalized 385 21.6 (6.2)nonhospitalized 59 21.0 (4.9)

HbA1c (%)* 0.90hospitalized 385 6.9 (1.7)nonhospitalized 59 6.9 (1.3)

glucose (mg/dl)* 0.51hospitalized 385 173 (70)nonhospitalized 59 167 (65)

BMI, body mass index; GA, glycated albumin; Hb,hemoglobin, *averaged over the course of the study.



7/9

a mean duration of 47.7 months, whereas HbA1c did notpredict survival.

Other investigators have analyzed the role of glycemiccontrol measured by HbA

1c levels on patient outcomes.

Kalantar-Zadeh et al. (7) assessed patient survival in23,618 dialysis patients with diabetes; subjects had atleast one measurement of HbA1c between July 2001 and June 2004. Unadjusted analyses paradoxically revealedlower death rates with higher HbA

1c. After adjusting for

potential confounders, increasing HbA1c levels were sig-nificantly associated with poorer survival. HbA1c wasrobustly associated with survival among nonanemic di-alysis patients, suggesting that shortened RBC survivalunderlies the inaccuracy of this measure in ESRD. Wil-

liams et al. (8) assessed the effects of HbA1c on survivalin 24,875 dialysis patients and reached the opposite con-

clusion. In contrast to Kalantar-Zadeh et al. (7), survivalrates observed in a 12-month period after measurementof HbA1c ranged from 80% to 85% across differentHbA1c strata. Kaplan-Meier survival curves grouped bylevel of HbA1c revealed no correlation between glycemiaand 12-month survival. A subsequent report by theseinvestigators revealed that only extremely high and ex-tremely low HbA

1c values were associated with hospi-

talization risk in diabetic hemodialysis patients (13).These authors concluded that prospective studies wereneeded to determine whether meeting the currently rec-ommended HbA1c targets improve outcomes withoutposing additional risks in the diabetic dialysis popula-tion. Our longitudinal study was prospective and obser-vational yet failed to detect a relationship betweenHbA

1c and risk of death or hospitalization.

The Alberta Kidney Disease Registry recently reportedthe effects of HbA1c in 1484 incident dialysis patientsinitiating renal replacement therapy between 2001 and2007 (14). Not all participants in this retrospective cohortstudy had diabetes; some were at high risk for diabetes.Casual blood glucose and HbA1c results were recorded.As in the report of Williams et al. (8), these measureswere not found to predict survival on dialysis.

In this report, HbA1c was negatively associated withdialysis survival in the unadjusted model (higher HbA1cwas associated with a reduced risk of death). A similarfinding was reported by Kalantar-Zadeh et al. (7). In ourstudy, the negative effect disappears after adjustment fordemographic characteristics. As proposed by Kalantar-Za-deh et al. (7), we hypothesize that the unadjusted HbA1cfindings relate to the effects of improved nutrition. Mitt-man et al. (15) also evaluated the relationship between baseline and 6-month HbA

1c and albumin-corrected serum

fructosamine levels on patient morbidity in 100 hemodial-ysis patients. Higher serum fructosamine levels, but notHbA1c levels, were associated with higher hospitalization

rates and infection rates; mortality was not assessed. BothGA and serum fructosamine have shorter half-lives than

Table 6. Regression models of glycemic markers and risk of hospitalization at 17 days

Model

P, HR (95% CI)

Glucose(n 435)

HbA1c(n 444)

Glycated Albumin(n 444)

Univariate analysis 0.03, 1.01 (1.001 to 1.02) 0.54, 1.013 (0.97 to 1.06) 0.003, 1.024 (1.009 to 1.041)Multivariate model, including

demographic variables0.02, 1.01 (1.002 to 1.02) 0.54, 1.014 (0.97 to 1.06) 0.005, 1.026 (1.008 to 1.044)

Including demographic variables,comorbid conditions anddialysis vintage

0.025, 1.01 (1.001 to 1.02) 0.69, 1.011 (0.96 to 1.07) 0.01, 1.025 (1.006 to 1.044)

Including demographic variables,dialysis vintage, comorbidconditions and lab values

0.014, 1.01 (1.002 to 1.02) 0.66, 1.013 (0.96 to 1.07) 0.02, 1.022 (1.004 to 1.040)

Best-fit (includes gender,phosphorus, and PVD)

0.007, 1.01 (1.003 to 1.02) 0.30, 1.020 (0.97 to 1.08) 0.005, 1.023 (1.006 to 1.039)

The hazard ratio (HR) indicates the ratio for 10 mg/dl change in glucose, 5 units of change in glycated albumin (GA), 1 unitchange in HbA

1c. CI, confidence interval; Hb, hemoglobin.

Figure 2. | Proportion of individuals hospitalized within 17 days of determination of glycemic measurements, by quintile. Rates of hos-pitalizations per quintile of glycated albumin (GA) differed signifi-cantly between the lowest and highest (5.9% versus 9.7%, respec-tively; P 0.02) and between the second lowest and second highest(6.2% versus 9.7%, respectively; P 0.03). Significant graded rela-tionships were not seen for hospitalization rates with either hemo-globin (Hb) A1c or serum glucose quintiles.



8/9

HbA1c and reflect more recent glycemic control. In addi-tion, neither of these assays is affected by hemoglobinlevels, and they both appear to be minimally affected byshortened RBC survival. We note that diabetes medica-tions may have to be discontinued in some dialysis pa-tients because of higher frequencies of hypoglycemic epi-

sodes (16). This may further confound the association of glycemic markers with outcomes in this unique popula-tion.

Our study identified a strong, graded relationship be-tween higher GA and higher risk of hospitalization in theensuing 17 days. We chose the 17-day period as the timeframe to assess hospitalizations because GA predomi-nantly reflects blood sugar control over the preceding 17days (17). Hospitalization analyses were repeated using a30-day period after measurements because HbA1c predom-inantly reflects blood sugar control over the preceding 30days (17), and similar results were seen (data not shown).Prior studies have shown that higher HbA1c is a risk factor

for hospitalization because of heart failure in nondialysispatients (18). Because cardiac disease is the leading causeof morbidity and mortality in dialysis patients, it is possi- ble that the GA may be a marker for cardiac events in thispopulation. HbA1c may be a better marker for hospitaliza-tion in the general population than in ESRD because of differences in RBC survival.

Our study has limitations. We elected to adjust for de-mographic and clinical variables but not dialysis adequacyor modality (hemodialysis versus peritoneal dialysis). Weelected not to adjust for dialysis adequacy because themajority of patients met Kidney Disease Outcomes QualityInitiative guidelines for dialysis adequacy, and little vari-ation was seen in dialysis dose. Recent data suggest nomajor differences in patient survival for peritoneal dialysisand hemodialysis (19), and our results were relatively con-sistent when analyses were limited to the hemodialysissample. Finally, serum phosphorus concentrations did notsignificantly affect dialysis survival, although they affectedhospitalizations. The lack of effect on survival was unex-pected and may be explained by the modest sample size orrelatively low serum phosphorus levels in the cohort. Mostparticipants were below the cut points at which highermortality rates are seen (20). Activated forms of vitamin Dand vitamin D analogues were widely prescribed (95%),so adjustment was not made for these medications.

We conclude that the GA assay, reflecting glycemic con-trol for an approximate 17-day period preceding measure-ment, predicts mortality and hospitalization in patientswith diabetes mellitus on dialysis. In contrast, HbA1c wasnot predictive of survival or hospitalizations in these pa-tients. We conclude that clinicians should consider mea-suring GA instead of HbA

1c in patients with diabetes on

dialysis.

Acknowledgments

Asahi Kasei Pharma Corporation provided partial support for

this study. Dr. Freedman independently developed the study

protocol. Data collection and analyses were performed at the

Wake Forest School of Medicine, and manuscript preparation was

solely at the discretion of Dr. Freedman.

Disclosures

None.

References1. Inaba M, Okuno S, Kumeda Y, Yamada S, Imanishi Y, Tabata

T, Okamura M, Okada S, Yamakawa T, Ishimura E,Nishizawa Y: Glycated albumin is a better glycemic indicatorthan glycated hemoglobin values in hemodialysis patientswith diabetes: Effect of anemia and erythropoietin injection.

J Am Soc Nephrol 18: 896–903, 20072. Peacock TP, Shihabi ZK, Bleyer AJ, Dolbare EL, Byers JR,

Knovich MA, Calles-Escandon J, Russell GB, Freedman BI:Comparison of glycated albumin and hemoglobin A(1c) lev-els in diabetic subjects on hemodialysis. Kidney Int 73:1062–1068, 2008

3. Nagayama H, Inaba M, Okabe R, Emoto M, Ishimura E, Oka-zaki S, Nishizawa Y: Glycated albumin as an improved indi-cator of glycemic control in hemodialysis patients with type 2diabetes based on fasting plasma glucose and oral glucosetolerance test. Biomed Pharmacother 63: 236–240, 2009

4. Freedman BI, Shenoy RN, Planer JA, Clay KD, Shihabi ZK, Burkart JM, Cardona CY, Andries L, Peacock TP, Sabio H, Byers JR, Russell

GB, Bleyer AJ: Comparison of glycated albumin and hemoglobinA1C concentrations in diabetic subjects on peritoneal and hemodi-alysis. Perit Dial Int 30: 72–79, 2010

5. Freedman BI, Shihabi ZK, Andries L, Cardona CY, PeacockTP, Byers JR, Russell GB, Stratta RJ, Bleyer AJ: Relationshipbetween assays of glycemia in diabetic subjects with ad-vanced chronic kidney disease. Am J Nephrol 31: 375–379,2010

6. Viljoen M, de Oliveira AA, Milne FJ: Physical properties of the red blood cells in chronic renal failure. Nephron 59:271–278, 1991

7. Kalantar-Zadeh K, Kopple JD, Regidor DL, Jing J, ShinabergerCS, Aronovitz J, McAllister CJ, Whellan D, Sharma K: A1Cand survival in maintenance hemodialysis patients. Diabetes Care 30: 1049–1055, 2007

8. Williams ME, Lacson E Jr, Teng M, Ofsthun N, Lazarus JM:

Hemodialyzed type I and type II diabetic patients in the US:Characteristics, glycemic control, and survival. Kidney Int 70:1503–1509, 2006

9. Stratton IM, Adler AI, Neil HA, Matthews DR, Manley SE,Cull CA, Hadden D, Turner RC, Holman RR: Association of glycaemia with macrovascular and microvascular complica-tions of type 2 diabetes (UKPDS 35): Prospective observa-tional study. BMJ 321: 405–412, 2000

10. Adler AI, Stevens RJ, Manley SE, Bilous RW, Cull CA, Hol-man RR: Development and progression of nephropathy intype 2 diabetes: The United Kingdom Prospective DiabetesStudy (UKPDS 64). Kidney Int 63: 225–232, 2003

11. The American Diabetes Association (ADA) has been activelyinvolved in the development and dissemination of diabetescare standards, guidelines, and related documents for manyyears. Diabetes Care 32[Suppl 1]: S1–S2, 2009

12. Fukuoka K, Nakao K, Morimoto H, Nakao A, Takatori Y,Arimoto K, Taki M, Wada J, Makino H: Glycated albuminlevels predict long-term survival in diabetic patients undergo-ing haemodialysis. Nephrology 13: 278–283, 2008

13. Williams ME, Lacson E Jr, Teng M, Hakim RM, Lazarus JM:Extremes of glycemic control (HbA1c) increase hospitaliza-tion risk in diabetic hemodialysis patients in the USA. Am J Nephrol 29: 54–61, 2009

14. Shurraw S, Majumdar SR, Thadhani R, Wiebe N, Tonelli M:Glycemic control and the risk of death in 1,484 patients re-ceiving maintenance hemodialysis. Am J Kidney Dis 55: 875–884, 2010

15. Mittman N, Desiraju B, Meyer KB, Chattopadhyay J, AvramMM: Treatment of secondary hyperparathyroidism in ESRD: A2-year, single-center crossover study. Kidney Int 78: S33–S36,2010

16. Kovesdy CP, Park JC, Kalantar-Zadeh K: Glycemic controland burnt-out diabetes in ESRD. Semin Dial 23: 148–156,2010



9/9

17. Tahara Y, Shima K: Kinetics of HbA1c, glycated albumin, andfructosamine and analysis of their weight functions against pre-ceding plasma glucose level. Diabetes Care 18: 440–447, 1995

18. Gerstein HC, Swedberg K, Carlsson J, McMuarry JJ, Michel-son EL, Olofssom B: The hemoglobin A1c level as a progres-sive risk factor for cardiovascular death, hospitalization forheart failure, or death in patients with chronic heart failure.Arch Intern Med 168: 1699–1704, 2008

19. Traynor JP, Thomson PC, Simpson K, Ayansina DT,Prescott GJ, Mactier RA: Comparison of patient survival innon-diabetic transplant-listed patients initially treated withhaemodialysis or peritoneal dialysis. Nephrol Dial Trans- plant 26: 245–252, 2010

20. Isakova T, Gutierrez OM, Chang Y, Shah A, Tamez H,Smith K, Thadhani R, Wolf M: Phosphorus binders and

survival on hemodialysis. J Am Soc Nephrol 20: 388–396,2009

Received: December 18, 2010 Accepted: March 17, 2011

Published online ahead of print. Publication date available at

www.cjasn.org.See related editorial, “Assesment of Glycemic Control in

Dialysis Pateints with Diabetes: Glycosylated Hemoglobin or

Glycated Albumin?, ” on pages 1520–1522.

Access to UpToDate on-line is available for additional clinical

information at www.cjasn.org.


Documents

jurnal imey