[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.

250]||ClickheretodownloadfreeAndroidapplicationfort

Saudi J Kidney Dis Transpl 2013;24(2):260-273

2013 Saudi Center for Organ Transplantation

Saudi Journal
of Kidney Diseases
and Transplantation

Original Article
Glycated Albumin versus Glycated Hemoglobin as Glycemic Indicator in
Hemodialysis Patients with Diabetes Mellitus: Variables that Influence
Dawlat Sany, Yasser Elshahawy, Walid Anwar
Nephrology Division, Faculty of Medicine, Ain Shams University, Cairo, Egypt
ABSTRACT. The significance of glycated albumin (GA) compared with casual plasma glucose
(PG) and glycated hemoglobin (HbA1c) was evaluated as an indicator of the glycemic control
state in hemodialysis (HD) patients with diabetes. In HD patients with diabetes (n = 25), the mean
PG, GA and HbA1c levels were 192.9 + 23 mg/dL, 278.8 + 43 mol/L and 5.9 + 0.5%, respectively, which were higher by 43.9%, 67.04% and 18%, respectively, compared with HD patients
without diabetes (n = 25). HbA1c levels were significantly lower than simultaneous PG and GA
values in those patients in comparison with the three parameters in patients who had diabetes
without renal dysfunction (n = 25). A significant negative correlation was found between GA and
serum albumin (r = 0.21, P <0.05) in HD patients with diabetes, whereas HbA1c correlated positively and negatively with hemoglobin (r = 0.11, P <0.01) and weekly dose of erythropoietin
injection (r = -0.19, P < 0.01), respectively. Although PG and GA did not differ significantly between HD patients with diabetes and with and without erythropoietin injection, HbA1c levels were
significantly higher in patients without erythropoietin. Categorization of glycemic control into
arbitrary quartiles by GA level led to better glycemic control in a significantly higher proportion
of HD patients with diabetes than those assessed by HA1c. Multiple regression analysis demonstrated that hemoglobin in addition to PG emerged as an independent factor associated with HbA1c
in HD patients with diabetes, while PG, body mass index and albumin were an independent factor
associated with GA. Conclusion: it is suggested that GA provides a significantly better measure to
estimate glycemic control in HD patients with diabetes and that the assessment of glycemic
control by HbA1c in these patients might lead to likely underestimation as a result of the increasing proportion of young erythrocyte by the use of erythropoietin.
Introduction
Strict glycemic control in patients with diabeCorrespondence to:
Dr. Dawlat Sany,
Nephrology Division, Faculty of Medicine,
University of Ain-Shams, Cairo, Egypt
E-mail: dollysany@gmail.com

tes decreases the incidence of diabetic complications,1 which can determine the quality of
life and prognosis of such patients.
A reduction of the risk for the development
of diabetic microangiopathy in patients with
type-2 diabetes by strict glycemic control was
demonstrated in the UK prospective diabetes
study.2 Recent clinical evidence has suggested
the favorable effects of strict glycemic control

[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.250]||ClickheretodownloadfreeAndroidapplicationfort

Glycated albumin versus glycated hemoglobin in HD patients

on cardiovascular disease, a main cause of

death in patients with diabetes.3 Several clinical tests are useful for measuring long-term
glycemic control in the general diabetic population. These same tests are routinely performed
in diabetic subjects with chronic kidney disease (CKD) and end-stage renal disease
(ESRD); however, their accuracy in these patients has not been rigorously tested.4 Hemoglobin A1c (HbA1c), the most widely used
assay, measures the percentage of circulating
hemoglobin that has chemically reacted with
glucose and reflects ambient blood glucose
control over the prior 120 days, with the most
profound effect in the preceding 30 days.5
HbA1c values are influenced significantly in
HD patients by either shortening of the life
span of erythrocytes6 or the changing proportion of young to old erythrocytes by erythropoietin use.7 If this significantly impacts
HbA1c, dialysis patients and clinicians would
be falsely comforted by relatively low HbA1c
values despite a high risk for subsequent cardiovascular disease and infectious complications. Previous studies attempting to address
this concern were underpowered.8 Recently,
serum glycated albumin (GA) was hypothesized to be an alternative marker for glycemic
control in patients with diabetes, which is not
affected by changes in the survival time of
erythrocytes in the case of type-2 diabetes with
hemoglobinopathy.9 Inaba et al10 determined
that HbA1c underestimated long-term glycemic control in dialysis patients with diabetes
after comparing the mean of random blood
glucose concentrations, HbA1c and percentage
of glycated albumin (% GA). They found that
the % GA assay provided a more accurate
assessment of glycemic control among Japanese
hemodialysis (HD) patients. The current study
attempted to validate this clinically important
result and extend its application to another
ethnic group (Egyptian).
Materials and Methods
This study was composed of 25 HD patients
with type-2 diabetes, 25 HD patients without
diabetes, 25 patients with type-2 diabetes and

261

chronic kidney disease and ten patients with

type-2 diabetes and normal renal function,
which was defined on the basis of serum creatinine levels of less than 1.2 mg/dL. The diagnosis of diabetes was based on a history of
diabetes or on the ADA criteria. Information
collected from participants included demographic data, height, weight (dry weight in HD
patients), duration of diabetes and duration of
HD. All patients provided written informed
consent before participation in this study.
Patients with diabetes were restricted to those
with stable blood glucose and whose diabetes
treatment had not been altered during the
preceding six months before the determination
of GA and HbA1c. Information on weekly doses of erythropoietin, which had not been
changed during the three months before determination of GA and HbA1c, was also obtained.
Exclusion criteria
Patients with hemoglobinopathy, anemia due
to causes other than CKD as hemolytic anemia, patients who had a history of overt blood
loss or who received a blood transfusion four
months prior to the study, patients with evidence of hyporvitaminosis C (Scurvy), evidence
of hepatic disorders, inflammatory disease or
thyroid disease, patients with heavy proteinuria or taking steroid therapy were all excluded from the study
Methods
GA assay: Diazyme enzymatic assay for GA
uses proteinase K to digest it into low molecular weight GA fragments and uses Diazymes
specific enzyme, a microorganism originated
amadoriase to catalyze the oxidative degradation of amadori products of GA fragments to
yield amino acids, glucosone and H2O2. The
H2O2 released is measured by a colorimetric
Trinder end-point reaction. The absorbance at
550 nm is proportional to the concentration of
GA. The following analytic concentrations were
not found to affect the assay: ascorbic acid
no interference up to 10 mg/dL; bilirubin no
interference up to 14.6 mg/dL; glucose no
interference up to 1000 mg/d; uric acid no
interference up to 15 mg/dL; and triglycerides

[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.250]||ClickheretodownloadfreeAndroidapplicationfort

262

Sany D, Elshahawy Y, Anwar W

Table 1. Comparison of demographic measures between the groups

Group I
Group II
Group III
Variables
(n = 10)
(n = 25)
(n = 25)
Age (years) (mean SD)
42.3 9
43.8 9
43.8 11
Gender
Male
3 (30%)
9 (36%)
14 (56%)
Female
7 (70%)
16 (64%)
11 (44%)

no interference up to 500 mg/dL.

HbA1c assay: Stanbio glycohemoglobin assay
is the quantitative colorimetric determination
of glycohemoglobin in whole blood.
Biochemical measurements
Blood was drawn without overnight fasting,
immediately before the morning Monday/Tuesday session of HD, to measure serum parameters in HD patients. In patients with diabetes with and without CKD, blood samples
were collected in the morning.
The mean values of the three-monthly measurements of casual plasma glucose (PG) that
were obtained during the two months before
determination of serum GA and HbA1c were
used in the analysis. Serum GA and HbA1c
were measured once, concomitant with the
determination of red blood cells, hemoglobin
(Hb), hematocrit, total protein, albumin, blood
urea nitrogen and creatinine. Blood was drawn
from the dialyzer circuit in subjects with
ESRD prior to the initiation of dialysis or administration of anticoagulants. Blood samples
were then divided with 5 mL sent for HbA1c
and 5 mL centrifuged with the serum frozen at
80C for measurement of GA, 8 mL was analyzed for total protein, albumin, blood urea
nitrogen and creatinine and 2 mL on EDTA for
CBC.
Statistical Analysis
Analysis of data was carried out using an
IBM computer and SPSS (Statistical Program
for Social Science, version 12). Data expressed

Group IV
(n = 25)
49.6 7

>0.05 NS

15 (60%)
10 (40%)

0.05
NS

as means and SD. One-way analysis of variance was used to compare more than two
groups as regards the quantitative variable.
Unpaired t-test was used to compare two groups
as regards the quantitative variable. The Mann
Whitney test was used instead of the unpaired
t-test in non-parametric data SD >50% mean.
Correlation co-efficients were calculated by
simple regression analysis and were used to
rank different variables against each other positively or inversely. Chi-square test, test,
was performed to compare the various distributions. Multiple logistic regression analysis
assessed the independent contribution of PG,
HbA1c and GA to the occurrence of diabetes.
Multiple regression analyses were performed
to explore the association of PG, hemoglobin,
albumin, erythropoietin dose and other variables with HbA1c and GA.
Results
Demographic and clinical characteristics of
the study population
Demographic characteristics of the study population is given in Table 1. Blood samples
were collected from 60 diabetic patients, ten
diabetic patients without CKD (group I), 25
diabetic patients with CKD (group II), 25 nondiabetic patients with ESRD receiving regular
HD treatments (group III) and 25 diabetic
patients with ESRD under regular HD (group
IV); 22 patients were receiving erythropoietin
with 4000 IU as the mean average weekly
dose. Group II had the highest body mass
index (BMI) and the most frequent obesity and

Table 2. Comparison of anthropometric measures between the groups.

Group I
Group II
Group III
Variables
(n = 10)
(n = 25)
(n = 25)
Weight (kg)
82.9 15
88.6 12
69.1 13
Height (m)
166.5 4.8
167 2.8
166 2.8
BMI (kg/m2)
29.8 4.9
31 4
24.8 4.1

Group IV
(n = 25)
79.4 14
166.9 4
28 4.9

P
<0.001HS
>0.05 NS
<0.001HS

[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.250]||ClickheretodownloadfreeAndroidapplicationfort

Glycated albumin versus glycated hemoglobin in HD patients

Table 3. Comparison of frequency of obesity between the groups.
Group I
Group II
Group III
Obesity
(n = 10)
(n = 25)
(n = 25)
Under weight
0
0
2 (8%)
Normal
1 (10%)
2 (8%)
13 (52%)
Over weight
4 (40%)
6 (24%)
6 (24%)
Obesity
5 (50%)
16 (64%)
4 (16%)
Morbid obesity
0
1 (4%)
0

263
Group IV
(n = 25)
0
6 (24%)
10 (40%)
9 (36%)
0

Table 4. Correlation between HbA1c and GA with BMI in the studied groups.
HbA1c (%)
Variables
Groups
r
P
Group I
-0.09
>0.05
BMI
Group II
-0.33
>0.05
(kg/m2)
Group III
-0.23
>0.05
Group IV
-0.20
>0.05

morbid obesity compared with the other subgroups; however, group III had the lowest
weight and BMI (P <0.001) (Tables 2 and 3).
GA and BMI showed negative correlations in
the studied groups (P <0.05) (Table 4). A significant difference between groups was noted
as regards different laboratory variables, except for total protein and serum albumin that
did not differ between groups (P <0.001) (Table
5). The prevalence of diabetic complications,
viz. diabetic peripheral neuropathy, diabetic
nephropathy, erectile dysfunction, diabetic retinopathy, cardiovascular disease and facial
palsy were 80%, 55%, 33.3%, 28.3%, 26.7%
and 1.7%, respectively.
Variation of casual PG levels during the study
period of 2 months
PG from patients with diabetes (n = 60) at
two months before, one month before and at the

X2

26

<0.001
HS

GA (mol/L)
r
P
-0.23
<0.05 S
-0.43
<0.05 S
-0.13
<0.05 S
-0.11
<0.05 S

time of measurement of GA and HbA1c were

198.8 44, 197.6 47.6 and 197.9 45.8
mg/dL, respectively. The correlation coefficients for PG between two and one months
before, between two and zero months before
and between one and zero months before were
r = 0.988 (P >0.05), r = 0.986 (P >0.05) and r
= 0.991 (P >0.05), respectively (Table 6).
These data suggested that glycemic control of
our patients with diabetes was stable during
the study period.
Glycemic control on the basis of PG, HbA1c
and GA values in diabetic patients
The mean PG, GA and HbA1c levels in the
HD patients with diabetes were 192.9 23
mg/dL, 278.8 43 mol/L and 5.9 0.5%,
respectively, all of which were significantly
higher than the corresponding values of 134
14 mg/dL, 166.9 25 mol/L and 5 0.6% in

Table 5. Comparison between groups as regards laboratory data.

Group I
Group II
Group III
Variables
(n = 10)
(n = 25)
(n = 25)
Hb (g/dL)
12.8 1
9.3 2.3
8.5 1.5
RBCs (million/mm3)
4.3 0.3
3.4 1
3 0.5
HCT (%)
41 3.5
30 8.6
29 5.6
Total protein (g/dL)
7.3 0.26
7 0.8
6.9 0.6
Albumin (g/dL)
3.9 0.2
3.6 0.3
3.7 0.4
Urea (mg/dL)
14 2.3
89 29
129 25.5
Creatinine (mg/dL)
0.85 0.21
3.8 1.4
7.6 1.6
HbA1c (%)
6.8 0.8
6.7 0.6
5 0.6
GA (mol/L)
190 67
304 81
166.9 25
GFR (mL/min/1.73 m2)
121 21
32+13
12.4 3.1
PG (mg/dL)
146.5 35
208 44
134 14

Group IV
(n = 25)
8.7 1.4
3 0.5
28 5
6.9 0.6
3.6 0.4
140 33
8.9 1.4
5.9 0.5
278.8 43
10.4 2
192.9 23

P
<0.001 HS
<0.001 HS
<0.001 HS
>0.05 NS
>0.05 NS
<0.001 HS
<0.001 HS
<0.001 HS
<0.001 HS
<0.001 HS
<0.001 HS

[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.250]||ClickheretodownloadfreeAndroidapplicationfort

264

Sany D, Elshahawy Y, Anwar W

Table 6. Variation of casual PG levels during the study period of 2 months in diabetic patients.
Std.
Paired
Correlation
Mean
N
t
deviation differences
coefficients P-value
(0 month)
198.88
60
45.960
RBS
1.267
1.550
0.991
0.13
(1 month)
197.62
60
46.829
(0 month)
198.88
60
45.960
RBS
0.917
0.920
0.986
0.36
(2 months) 197.97
60
44.965
(1 month)
197.62
60
46.829
RBS
-0.350
0.365
0.988
0.72
(2 months) 197.97
60
44.965

Sig
NS
NS
NS

Table 7. Comparison of the degrees of glycemic control on the basis of PG HbA1c and GA values in
diabetic patients.
Variables
Group I
Group II
Group IV
P
HbA1c (%)
6.8 0.8
6.7 0.6
5.9 0.5
<0.001 HS
GA (mol/L)
190 67
304 81
278.8 43
<0.001 HS
PG (mg/dL)
146.5 35
208 44
192.9 23
<0.001 HS
Albumin (g/dL)
3.9 0.2
3.6 0.3
3.6 0.4
>0.05 NS

the HD patients without diabetes. Meanwhile,

the mean PG and GA, levels in diabetic
patients without CKD were 146.5 35 mg/dL
and 190 67 umol/L, respectively, all of which
were lower than the corresponding values of
208 44 mg/ dL and 304 81 mol/L in
diabetic patients with CKD. HbA1c levels in
diabetic patients without CKD were 6.8 0.8%
and in diabetics with CKD patients were 6.7
0.6%, which were higher than that in diabetic
HD patients. A highly significant difference
between diabetic groups as regard PG, HbA1c
and GA values (P <0.001) was noted (Table 7).
Distribution of the degrees of glycemic control
on the basis of the HbA1c and GA values
HD patients with diabetes were divided into
four categories according to HbA1c values:
excellent (HbA1c 5.5%), good (5.5% < HbA1c
6.8%), fair (6.8% < HbA1c 7.6%), and poor
(HbA1c >7.6%). There were 13 (52%), seven
(28%), three (12%) and two (8%) in each
group, respectively. Glycemic control was also
assessed according to the GA values: excellent

(GA 122 mol/L), good (122 mol/L < GA

203 mol/L), fair (203 mol/L < GA 285
mol/L) and poor (GA >285 mol/L). There
were seven (28%), six (24%), four (16%) and
eight (32%) in each group, respectively (Table
8). The proportions of glycemic control that
were based on the HbA1c values were significantly different from those that were based
on the GA values.
Correlation between PG and GA or HbA1c in
HD patients with and without diabetes, and
diabetic patients with and without CKD
There was a significant and positive correlation between PG and serum GA (r = 0.54, P
<0.01) in HD patients with diabetes, (r = 0.48,
P <0.01) in diabetic patients without CKD, (r
= 0.58, P <0.01) in diabetic patients with CKD
and (r = 0.97, P <0.01) in non-diabetic HD patients. Also, with HbA1c (r = 0.51, P <0.01) in
HD patients with diabetes, (r = 0.61, P <0.01)
in diabetic patients without CKD and (r = 0.56,
P <0.01) in diabetic patients with CKD. As
shown, the relationship between PG and GA

Table 8. Distribution of the degrees of glycemic control on the basis of the HbA1c, GA and PG values in
group IV.
Glycemic
HbA1c
PG
GA
PG
control
(%)
No (%)
mg/d/L
(mol/L)
No (%)
mg/dL
Excellent
5.5
13 (52%)
170
122
7 (28%)
126
Good
>5.5 6.8
7 (28%)
>170 222
>122 203
6 (24%)
>126 140
Fair
>6.8 7.6
3 (12%)
>222 258
>203 285
4 (16%)
>140 200
Poor
>7.6
2 (8%)
>258
>285
8 (32%)
>200

[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.250]||ClickheretodownloadfreeAndroidapplicationfort

Glycated albumin versus glycated hemoglobin in HD patients

265

Positive correlation between HBA1C and PG in diabetic HD patients

Figure 1. Correlation between HBA1C and PG in group IV.

was identical between the HD patients with

diabetes and patients with diabetes without and
with CKD, although HbA1c values in comparison with those of PG seemed to be significantly lower in HD patients with diabetes than
in patients with diabetes and without CKD
(Table 5 and Figure 1).
Correlation between GA and serum albumin
and between HbA1c and hemoglobin levels in
HD patients with and without diabetes, and
diabetic patients with and without CKD.
The serum albumin and HbA1c in HD patients with diabetes ranged from 3 to 4.7 g/dL
and from 5.3 to 7.1%, respectively. A significant and negative correlation was found between GA and serum albumin levels (r = -0.21,

P <0.05) in diabetic HD patients, (r = -0.33, P

<0.05) in diabetic patients without CKD, (r = 0.30, P <0.05) in diabetic patients with CKD
and (r = -0.34, P <0.05) in non-diabetic HD
patients, although HbA1c did not correlate
with serum albumin levels (r = 0.13, P >0.05)
in HD patients with diabetes, (r = 0.09, P
>0.05) in diabetic patients without CKD, (r =
0.03, P >0.05) in diabetic with CKD and (r =
0.07, P >0.05) in non-diabetic HD patients. In
contrast, there was a significant and positive
correlation between HbA1c and hemoglobin
levels (r = 0.11, P <0.001) in diabetic HD, (r =
0.87, P <0.01) in diabetics without CKD, (r =
0.57, P <0.01) in diabetics with CKD and (r =
0.77, P <0.01) in non-diabetic HD patients,
although GA did not correlate with the serum

Table 9. Correlation between HbA1c, GA and other variables in group I and group II.
Group I
Group II
Variables
HbA1c%
GA
HbA1c%
GA
r
P
r
P
r
P
r
P
BMI (kg/m2)
-0.09 >0.05 -0.23
<0.05
-0.33
>0.05
-0.43
<0.05 S
Hb (g/dL)
0.87
<0.01
0.09
>0.05
0.57
<0.01HS
0.14
>0.05
HCT (%)
0.56
<0.05
0.16
>0.05
0.54
<0.01 HS
0.15
>0.05
Total protein (g/dL)
0.13
>0.05
0.25
>0.05
0.11
>0.05
0.21
>0.05
Albumin (g/dL)
0.09
>0.05
0.3
<0.05
0.03
>0.05
-0.30
<0.05 S
Urea (mg/dL)
0.12
>0.05
0.26
>0.05
0.16
>0.05
0.26
>0.05
Creatinine (mg/dL)
0.11
>0.05
0.19
>0.05
0.10
>0.05
0.01
>0.05
GA (mol/L)
0.67
<0.00
0.65
<0.01 HS
GFR (mL/min/1.73 m2)
0.15
>0.05
0.22
>0.05
-0.07
>0.05
0.12
>0.05
PG (mg/dL)
0.61
<0.01
0.48 <0.001
0.56
<0.01 HS
0.58
<0.01 HS

[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.250]||ClickheretodownloadfreeAndroidapplicationfort

266

Sany D, Elshahawy Y, Anwar W

Table 10. Correlation between HbA1c, GA and other variables in group III and group IV.
Group III
Group IV
Variables
HbA1c
GA
HbA1c
GA
r
P
r
P
r
P
r
P
BMI (kg/m2)
-0.23
>0.05
-0.13
<0.05 S
-0.20
>0.05
-0.11
<0.05 S
Hb (g/dL)
0.77
<0.01 HS
0.13
>0.05
0.11
<0.01 HS
0.07
>0.05
HCT (%)
0.74
<0.01 HS
0.11
>0.05
0.70
<0.01
0.10
>0.05
Total protein
0.18
>0.05
0.27
>0.05
0.17
>0.05
0.17
>0.05
(g/dL)
Albumin (g/dL)
0.07
>0.05
-0.34
<0.05 S
0.13
>0.05
-0.21
<0.05
Urea (mg/dL)
0.11
>0.05
0.16
>0.05
0.16
>0.05
0.15
>0.05
Creatinine (mg/dL)
0.15
>0.05
0.08
>0.05
0.12
>0.05
0.18
>0.05
GA (mol/L)
0.63
<0.01 HS
0.70
<0.01 HS
GFR (mL/min/1.73
0.27
>0.05
0.17
>0.05
-0.32
>0.05
0.16
>0.05
m2)
PG (mg/dL)
0.14
>0.05
0.97
<0.01 HS
0.51
<0.01 HS
0.54
<0.01 HS

hemoglobin levels (r = 0.07, P >0.05) in diabetic HD, (r = 0.09, P >0.05) in diabetics without CKD, (r = 0.14, P >0.05) in diabetics with
CKD and (r = 0.13, P >0.05) in non-diabetic
HD patients (Tables 9 and 10).
Correlation between serum GA and HbA1c
levels in HD patients with and without diabetes, patients with diabetes with and without
CKD
There was a significant and positive correlation between serum GA and HbA1c levels in
HD patients with diabetes (r = 0.70, P <0.01),
patients with diabetes and without CKD (r =

0.67, P <0.01), diabetic patients with CKD (r =

0.65, P <0.01) and non-diabetic HD patients (r
= 0.63, P <0.01) (Tables 9 and 10 and Figure
2).
Correlation of the weekly erythropoietin dose
with HbA1c, GA and PG in HD patients with
diabetes and non-diabetic HD patients
There was a significant and negative correlation between HbA1c and the weekly dose of
erythropoietin (r = -0.19, P <0.01) in both HD
patients with diabetes and non-diabetic HD
patients (r = -0.23, P <0.01), although GA did
not correlate well (r = 0.08, P >0.05) in HD pa-

Positive correlation between HBA1C versus GA in HD patients with diabetes

Figure 2. Correlation between HbA1C and GA in group IV.

[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.250]||ClickheretodownloadfreeAndroidapplicationfort

Glycated albumin versus glycated hemoglobin in HD patients

267

Table 11. Correlation of the weekly erythropoietin dose with HbA1c, GA and PG in group III.
Erythropoietin
Variables
r
P
Positive (n =10)
Negative (n = 15)
HbA1c (%)
4.61 0.8
5.01 1.13
-0.23
<0.01 HS
GA (mol/L)
169.8 27
165 21
0.11
>0.05 NS
PG (mg/dL)
137.2 25
133.3 27
0.15
>0.05 NS
Table 12. Correlation of the weekly erythropoietin dose with HbA1c, GA and PG group IV.
Erythropoietin
Variables
r
P
Positive (n =12)
Negative (n = 13)
HbA1c (%)
5.72 1.3
6.19 1.49
-0.19
<0.01 HS
GA (mol/L)
282.6 45
277.4 43
0.08
>0.05 NS
PG (mg/dL)
196.4 19
192.3 23
0.11
>0.05 NS

tients with diabetes, (r = 0.11, P >0.05) in nondiabetic HD patients. The average PG and GA
levels in the HD patients with diabetes and
without erythropoietin (n = 13) were 192.3
23 mg/dL and 277.4 43 mol/L and in nondiabetic HD patients without erythropoietin (n
= 15) these levels were (133.3 27 mg/dL and
165 21 umol/L), which were not significantly different from the respective values of
196.4 19 mg/dL and 282.6 45 mol/L in
those who received erythropoietin (n = 12) in
diabetic HD patients and (137.2 25 mg/dL,
169.8 27 umol/L) in those who received erythropoietin (n = 10) in non-diabetic HD patients. However, the HbA1c values were significantly higher in those who were not treated
with erythropoietin compared with those who
were treated with erythropoietin (6.19 1.49
versus 5.72 1.3) (P <0.05) in the HD patients

with diabetes and (5.01 1.13 versus 4.61

0.8) (P <0.05) in non-diabetic HD patients
(Tables 11 and 12).
PG, GA and HbA1c levels as monitoring tests
for diabetes in HD patients
At the cut-off value for PG, GA and HbA1c
(160 mg/dL, 231 umol/L, 6.5%), the sensitivity was (83%, 80%, 40%) and specificity was
(96%, 70%, 54%). Therefore, GA and PG were
better than HbA1c as the follow-up method for
diabetes in HD patients (Table 11 and Figure
3).
Multiple regression analysis of factors for
HbA1c and GA in HD patients with DM
In diabetic HD, multiple regression analysis
of various clinical variables was used to evaluate their independent association with HbA1c

Significant and positive correlation between HBA1C and Hb levels

Figure 3. Correlation between HBA1C and Hb levels in group IV.

[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.250]||ClickheretodownloadfreeAndroidapplicationfort

268

Sany D, Elshahawy Y, Walid A

Table 13. Validity of PG, GA and HbA1c as screening tests for diabetes.
HbA1c
GA
PG
6.5%
231 mol/L
160 mg/dL
40%
80%
83.3%
54%
70%
96%
56%
75%
98%
47%
85%
70%

Validity
Bets cut-off
Sensitivity
Specificity
PPV
NPV

Table 14. Multiple regression analysis of factors for HbA1c and GA in diabetic HD patients.
GA (mol/L)
HbA1c (%)
Clinical variables
Beta
Beta
P
95% CI
P
coefficient
coefficient
Hb (g/dL)
0.07
>0.05 NS
0.11.4
0.11
<0.01 HS
PG (mg/dL)
0.51
<0.01 HS
1.88
0.51
<0.01 HS
Albumin (g/dL)
-0.21
<0.05 S
18.1
0.13
>0.05 NS
BMI (kg/m2)
-0.11
<0.05 S
1.16
-0.20
>0.05 NS
GFR (mL/min)
0.16
>0.05 NS
0.16
-0.32
>0.05
ttt (Erythropoietin)
0.92
>0.05 NS
0.98
0.45
>0.05 NS

and GA values in HD patients with diabetes.

(Table 13 and Figure 4). Among various clinical variables that included average PG, serum
albumin, GFR, erythropoietin treatment and
hemoglobin, only average PG and hemoglobin
were independent factors associated with
HbA1c (P <0.01). Evaluation of the independent
factors that were associated with GA showed
that the average PG (P <0.01), serum albumin
and BMI (P <0.05) exhibited a significant and

95% CI
1.21.6
1.18
0.32
0.23
1.46
0.42

independent association with GA (Table 14).

The same was true for all diabetic patients
(Table 15).
Logistic regression analysis of PG, GA and
HbA1c with diabetes in HD patients
The independent contribution of PG, GA and
HbA1c to the probability of diabetes in HD
patients was assessed after adjustment for serum albumin and Hb by multiple logistic reg-

Figure 4. Receiver operator characteristics (ROC) curve between DM and non-DM as regards HBA1C
and GA.
GA level was found to have an optimum sensitivity of 80% and specificity of 70% for >231 mol/L cutoff. However, HBA1C level was found to have a sensitivity of 40% and specificity of 54% for a >6.5%
cut-off.

[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.250]||ClickheretodownloadfreeAndroidapplicationfort

Glycated albumin versus glycated hemoglobin in HD patients

Table 15. Multiple regression analysis of factors for HbA1c and GA in diabetic patients.
GA (mol/L)
HbA1c (%)
Clinical variables
Beta
Beta
P
95% CI
P
coefficient
coefficient
Hb (g/dL)
0.11
>0.05 NS
0.91.4
0.58
<0.01 HS
PG (mg/dL)
0.64
<0.01 HS
18
0.46
<0.01 HS
Albumin (g/dL)
-0.31
<0.05 S
1.28
0.08
>0.05 NS
BMI (kg/m2)
-0.26
<0.05 S
1.86
-0.21
>0.05 NS
GFR (mL/min/1.73 m2)
0.17
>0.05 NS
0.66
-0.20
>0.05 NS
ttt (Erythropoietin)
0.24
>0.05 NS
0.93
0.12
>0.05 NS

ression analysis. PG (per 10 mg/dL; odds ratio

[OR] 1.5; P <0.01), GA (per 10 umol/L; OR
1.6; P <0.01) and HbA1c (per 1.0%; OR 2.1; P
<0.01) were independent risk factors associated with diabetes in HD patients (Table 15).
Discussion
Accurate determination of glycemic control
is of paramount importance in the diabetic population, as improved glycemic control reduces micro- and macro-vascular complications
in patients with type-1 and type-2 diabetes
mellitus.11 HbA1c has been a cornerstone in
the evaluation of dialyzed and non-dialyzed
diabetic patients. This measurement relies on a
relatively stable RBC survival, a characteristic
typical of the general population, but not patients on HD. During HD, the uremic environment, blood loss during treatments and frequent
phlebotomy contribute to decreased RBC lifespan. Shortened RBC survival and red cell
transfusions are likely to lower the HbA1c,
potentially making it unreliable in assessing
glycemic control. In this study, the measurement
of GA was shown to provide a more relevant
method to assess glycemic control in HD
patients with diabetes, also demonstrating that
HbA1c, relative to GA, significantly underestimates glycemic control in diabetic dialysis
patients. In those with ESRD, lower HbA1c
values were also associated with lower hemoglobin concentration and higher doses of ery-

269

95% CI
11.8
13.8
0.36
0.86
1.18
0.12

thropoietin. Although PG was measured without overnight fasting, a previous report showed
that non-fasting, rather than fasting, PG was a
better marker of glycemic control in type-2
diabetes.12 Because the mean values of monthly
determined PG essentially were the same
throughout the study period, it was suggested
that glycemic control had been stable during
the two months before the determination of
GA and HbA1c and that a single determination
just before the Monday/Tuesday HD session
might be representative of glycemic control in
HD patients with diabetes. Although HbA1c and
GA reflect glycemic control during the preceding four to six weeks and one to two weeks,7
the stable glycemic control during the preceding months can negate the different impact of
acute changes of glycemic control between
HbA1c and GA in this study. Supportive of
this notion is that the correlation coefficient
between PG and HbA1c was similar with that
between PG and GA (Table 16). The correlation coefficients of PG at 2, 1 or 0 months
before with HbA1c were very similar to those
with GA. The degree with which serum GA
correlated with PG was identical between the
HD patients with and without diabetes and
patients with diabetes with and without CKD
(Table 10). The significantly lower value of
HbA1c relative to PG and GA in HD patients
with diabetes compared with the patients with
diabetes with and without CKD might suggest
that the measurement of HbA1c would result in

Table 16. Logistic regression analysis of PG, GA and HbA1c with diabetes.
Variables
Beta coefficient
P
Odd ratio (OR) (95% confidence interval)
GA (mol/L)
0.14
<0.01
1.6 (16.8)
HbA1c (%)
0.10
<0.01
2.1 (1.18)
PG (mg/dL)
0.8
<0.01
1.5 (18)

[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.250]||ClickheretodownloadfreeAndroidapplicationfort

270

Sany D, Elshahawy Y, Anwar W

Figure 5. ROC curve between DM and non-DM as regards PG

PG level was found to have an optimum sensitivity of 83.3% and specificity of 96% for >160 mg/dL
cut-off

the underestimation of glycemic control in HD

patients with diabetes (Figure 5).
As shown in Table 8, in HD patients with
diabetes, it was shown that a good category
of GA of <203 mol/L and HbA1c of <6.8%
results in a PG of <140 and 222 mg/dL, respectively. Therefore, the GA value of <203
mol/L was reasonably categorized into a good
category, as reflected by the PG value of <140
mg/dL. However, categorization of the HbA1c
value of <6.8% into a good category definitely
was an underestimation, as reflected by PG
values as high as 222 mg/dL. The mechanism
for the significantly lower HbA1c value in
those patients was explained by anemia and/or
erythropoietin injection, as reflected by a significant positive correlation of HbA1c with
hemoglobin and the negative correlation with
weekly dose of erythropoietin (Tables 9 and
12).
Multiple regression analysis demonstrated that
hemoglobin reduction was an independent factor that was associated significantly with the
HbA1c values (Table 14). In fact, the HbA1c
values were significantly lower in HD patients
who had diabetes and were treated with ery-

thropoietin compared with those without erythropoietin, although PG and GA did not differ
significantly between the two groups of patients (Table 12). The differences of the mean
HbA1c values between the HD patients with
diabetes and HD patients without diabetes
were smaller than those of PG and GA (Table
5), which is explained partly by a significantly
greater erythropoietin dose in the HD patients
with diabetes (Tables 11 and 12). Importantly,
although serum albumin correlated negatively
with GA (Tables 9, 10 and 14), it failed to be a
significant factor between different groups of
patients as the serum albumin levels were
above 3.5 gm/dL (Table 5). The factors that
were associated independently with the GA
value was the average PG, BMI and serum
albumin, while those associated with HbA1c
were the average PG and hemoglobin levels
(Table 14). In order to clarify the reasons for
the negative association of BMI with GA, one
could conclude that the obesity-related inflammation is connected to this negative association. Inflammation reduces the rate of albumin synthesis and increases its catabolic rate.13
Also, hyperinsulinemia in obese diabetic pa-

[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.250]||ClickheretodownloadfreeAndroidapplicationfort

Glycated albumin versus glycated hemoglobin in HD patients

tient leads to increased albumin turnover.

Tessari et al14 proved that albumin fractional
synthesis rates and absolute synthesis rates
increased by ~25% after hyperinsulinemia.
Multiple logistic regression analysis showed
that PG, GA and HbA1c were independent risk
factors associated with the prevalence of diabetes after adjustment for serum albumin and
increase of Hb A 1% is indicative of a 2.1-fold
increase to have diabetes, in contrast to a 1.6fold increase per 10 mol/L increase of GA
value and 1.5-fold increase per 10 mg/dL of
PG value (Table 15). It was suggested that an
increase of GA might be more highly indicative of diabetes than that of HbA1c.
The non-enzymatic glycation of various proteins is increased in patients with diabetes as a
result of sustained higher PG.15 The rate of
production also depends on the half-life of
each protein.16 HbA1c provides an integrated
measure of PG during the previous two to
three months as a result of the long life span of
erythrocytes (120 d),17 whereas GA has been
hypothesized to be a glycemic indicator during
the immediately previous two weeks.16 Although
a rapid change in glycemic control may reflect
a greater change of GA than HbA1c, this study
examined the significance of GA compared
with HbA1c under stationary state of diabetic
control, without any change of antidiabetic drugs
during the study period, and compared GA and
HbA1c values in patients with diabetes and
with and without CKD and non-diabetic HD
patients. A previous report7 showed that after
erythropoietin treatment, the HbA1c levels decreased with the increase of hematocrit in 15
HD patients without diabetes, although PG did
not change. Conversely, after stopping erythropoietin treatment, the HbA1c levels increased. Because erythropoietin accelerates the
production of new erythrocytes, the proportion
of young erythrocytes in peripheral blood must
increase after erythropoietin administration.
HbA1c is the product of the chemical condensation of hemoglobin and glucose, and the glycation rate of just-produced young erythrocytes is reported to be lower than that of old
cells.18 Therefore, it seems that the decrease of
HbA1c levels relative to PG or GA in HD pa-

271

tients who have diabetes and are treated with

erythropoietin might be due to the increasing
proportion of young erythrocytes over old erythrocytes in peripheral blood of those patients.7
Anemia that results from shorter life span of
erythrocytes theoretically suppresses HbA1c
values. Withdrawal of erythropoietin administration increases HbA1c values, although it
suppresses Hb levels.7 Therefore, a relationship between HbA1c and Hb could be controversial. These data may suggest that HbA1c is
not an ideal index for glycemic control in HD
patients who have diabetes and receive erythropoietin. Because approximately 70% of
dialysis patients undergo erythropoietin treatment, HbA1c might be an unsuitable marker to
reflect glycemic control in HD patients with
diabetes because of the false reduction of
HbA1c values as a result of the increasing proportion of young erythrocytes over old erythrocytes in peripheral blood of those who
receive erythropoietin; however, this was not
due to improvement of glycemic control, leading to the underestimation of integrated hyperglycemia when assessed by HbA1c value.
GA acquires biologic properties that are linked
to the pathogenesis of diabetic vascular complications,19 suggesting that GA not only is
significant as an indicator of hyperglycemia20
but also contributes directly to vascular injury. As such, GA is better than HbA1c in predicting the development of vascular complications in HD patients with diabetes.
However, a limitation of the GA assay also
exists. Albumin turnover should change in patients who are maintained on peritoneal dialysis and in patients who have CRF with massive proteinuria, in whom GA values theoretically should be reduced as a result of shorter
exposure to plasma albumin. Also, in the present study, we found that HbA1c of 6.5% or
greater provides sensitivity and specificity as a
screening test for diabetes, at 40% and 54%,
respectively, and positive predictive value test
of 56% and negative predictive value test of
47% (Table 13). Buell et al21 recently completed a similar analysis based on the 19992004
NHANES data. The diagnosis of diabetes was
considered established if FPG was 126 mg/dL

[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.250]||ClickheretodownloadfreeAndroidapplicationfort

272

Sany D, Elshahawy Y, Anwar W

or greater. Using an ROC analysis, they found

that HbA1c of 5.8% or greater is the point that
yielded the highest sum of sensitivity (86%)
and specificity (92%). They concluded that
HbA1c of 5.8% would be an appropriate cutpoint above which to proceed to further evaluation. Also, we found that GA of >231
mol/L or greater provides reasonable sensitivity and specificity as a screening test for
diabetes, at 80% and 70%, respectively, and
positive predictive value test of 75% and negative predictive value test of 85% (Table 13).
Cefalu et al22 found that GA with a cut-off of
250 mol/L1 for the second-generation assay,
has the sensitivity to detect diabetes was 81%,
specificity was 87% and positive predictive
value was 43%.
Potential weaknesses in this study include use
of random (not necessarily fasting) recent
blood glucose measures in participants on dialysis and those with and without nephropathy,
and the small number of patients. Fasting blood
glucose measurements cannot be readily obtained as patients on the afternoon hemodialysis shift are unable to report fasting and
those on the morning shift are typically away
from home for periods exceeding 6 h and are
not permitted to eat or drink during HD.
In conclusion, GA provides a significantly
better measure to estimate glycemic control in
HD patients with diabetes and the assessment
of glycemic control by HbA1c in those patients might lead to underestimation due to
EPO therapy and anemia.
References
1.

2.

3.

The Diabetes Control and Complications Trial

Research Group: The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulindependent diabetes mellitus. N Engl J Med
1993;329:97786.
UK Prospective Diabetes Study (UKPDS)
Group: Intensive blood-glucose control with
sulphonylureas or insulin compared with conventional treatment and risk of complications
in patients with type-2 diabetes (UKPDS 33).
Lancet 1998;352:83753.
Cao JJ, Hudson M, Jankowski M, Whitehouse

4.
5.
6.

7.

8.

9.

10.

11.

12.

13.

14.

F, Weaver WD. Relation of chronic and acute

glycemic control on mortality in acute myocardial infarction with diabetes mellitus Am J
Cardiol 2005;96:1836.
Calisti L, Tognetti S. Measure of glycosylated
hemoglobin. Acta Biomed 2005;76(Suppl 3):
5962.
Goldstein DE, Little RR, Lorenz RA, et al.
Tests of glycemia in diabetes. Diabetes Care
2004;27:176173.
Joy MS, Cefalu WT, Hogan SL, et al. Longterm glycemic control measurements in diabetic patients receiving hemodialysis. Am J
Kidney Dis 2002;39:297307.
Nakao T, Matsumoto H, Okada T, et al.
Influence of erythropoietin treatment on hemoglobin A1c levels in patients with chronic renal
failure on hemodialysis. Intern Med 1998;38:
82630.
Tzamaloukas AH, Hsi KC, Quintana BJ, et al.
Glycosylated hemoglobin measured by affinity
chromatography in diabetic and nondiabetic
patients on long-term dialysis therapy. West J
Med 1989;150:4159.
Kosecki SM, Rodgers PT, Adams MB. Glycemic monitoring in diabetics with sickle cell
plus beta-thalassemia hemoglobinopathy. Ann
Pharmacother 2005;39:155760.
Inaba M, Okuno S, Kumeda Y, et al. Glycated
albumin is a better glycemic indicator than
glycated hemoglobin values in hemodialysis
patients with diabetes: effect of anemia and
erythropoietin injection. J Am Soc Nephrol
2007;18:896903.
Adler AI, Stevens RJ, Manley SE, et al.
Development and progression of nephropathy
in type 2 diabetes: the United Kingdom Prospective Diabetes Study (UKPDS 64) Kidney
Int 2003;63:22532.
Avignon A, Radauceanu A, Monnier L. Nonfasting plasma glucose is a better marker of
diabetic control than fasting plasma glucose in
type 2 diabetes. Diabetes Care 1997;20:1822
6.
Koga M, Otsuki M, Matsumoto S, et al. Negative association of obesity and its related
chronic inflammation with serum glycated
albumin but not glycated hemoglobin levels.
Clinica Chimica Acta. 2007;378:4852.
Tessari P, Kiwanuka E, Millioni R, et al. Albumin and fibrinogen synthesis and insulin effect
in type 2 diabetic patients with normoalbuminuria. Diabetes Care 2006;29:3238.

[Downloadedfreefromhttp://www.sjkdt.orgonFriday,January09,2015,IP:202.162.199.250]||ClickheretodownloadfreeAndroidapplicationfort

Glycated albumin versus glycated hemoglobin in HD patients

15. Cohen MP. Nonenzymatic glycation: A central
mechanism in diabetic microvasculopathy. J
Diabet Comp1988;2:2147.
16. Schleicher ED, Olgemoller B, Wiedenmann E,
et al. Specific glycation of albumin depends on
its half-life. Clin Chem 1993;39:6258.
17. Bunn HF, Gabbay KH, Gallop PM. The glycosylation of hemoglobin: Relevance to diabetes
mellitus. Science 1978;20:217.
18. Fitzgibbons JF, Koler RD, Jones RT. Red cell
age-related changes of hemoglobins A1a_b
and A1c in normal and diabetic subjects. J Clin
Invest 1976;58:8204.
Cohen MP, Ziyadeh FN, Chen S. Amadorimodified glycated serum proteins and acelerated atherosclerosis in diabetes: Pathogenic

273

and therapeutic implications. J Lab Clin Med

2006;147:2119.
19. Amore A, Cirna P, Conti G, et al. Amadoriconfigurated albumin indices nitric oxidedependent apoptosis of endothelial cells: A
possible mechanism of diabetic vasculopathy.
Nephrol Dial Transplant 2004;19:5360.
20. Buell C, Kermah D, Davidson MB. Utility of
A1c for diabetes screening in the 1999-2004
NHANES population. Diabetes Care. 2007;30:
22335.
21. Cefalu WT, Ettinger WH, Rushing JT. Serum
glycated albumin as a screening test for
diabetes in the elderly: a pilot study. Journal of
the American Geriatrics Society. 1993;41:
10904.