Keywords
HbA1c - glycated hemoglobin - eAG - estimated average blood glucose
Introduction
The prevalence of diabetes mellitus (DM) in India has risen from 7.1% in 2009 to 8.9%
in 2019. Currently, 25.2 million adults are estimated to have impaired glucose tolerance,
which is estimated to increase to 35.7 million in the year 2045. India ranks second
after China in the global diabetes epidemic with 77 million people with diabetes.[1] In Delhi, capital of India, 25.2% of the population was estimated to have diabetes.[2] Among the various biochemical markers associated with DM diagnosis and management,
glycated hemoglobin A1c (HbA1c) is of utmost importance owing to its utility as a
reliable marker to assess timely control over the preceding 2 to 3 months.[3] It is recommended that diabetic patients have their HbA1c levels checked at least
two times per year because quantitative and direct relationships have been identified
between HbA1c concentration and the risk of diabetic microvascular complications.[4] Therefore, clinicians use HbA1c test results to guide treatment decisions, and the
test has become the cornerstone for assessing diabetes care.[5] The conventional approach for the expression of HbA1c values is percent (%) of total
hemoglobin, which is not easily comprehensible for a DM patient with nonmedical background.[6]
In 2008, Nathan et al conducted the International HbA1c-Derived Average Glucose (ADAG)
trial, which established a linear dependence between HbA1c and averaged plasma glucose
levels, and a simple mathematical equation for the calculation of estimated average
glucose (eAG) level using the HbA1c level was introduced.[7] The relationship between HbA1C and eAG is described by the equation 28.7 × A1C –
46.7 = eAG. This equation has been extensively evaluated since then, and citing eAG
values with HbA1c laboratory reports has become a common practice. Still most clinical
laboratories have not yet started reporting eAG values and a widespread understanding
of its utility in the medical fraternity is missing.
Objectives
-
To determine the statistical correlation between HBA1C with random blood sugar (RBS)
values both in diabetic and prediabetic subjects.
-
To determine the statistical correlation between eAG derived from HbA1C using the
Nathan's regression equation with RBS both in diabetic and prediabetic subjects.
-
To analyze the significance of eAG as opposed to HbA1C as a marker of long-term glycemic
control in DM.
Materials and Methods
Study Design
This hospital-based retrospective analytical cohort study was conducted at the Clinical
Laboratory, Department of Biochemistry, Hindu Rao Hospital (900-bedded tertiary care
hospital), New Delhi, India after approval from the institutional ethical review committee
(IEC/NDMC/2022/131).
Sample Selection and Sample Size
The study group was selected from patient reporting to the laboratory for HbA1c estimation.
The simple random sampling technique was used to obtain laboratory records of both
sexes in the age range of 12 to 90 years presenting as outpatients. Pregnant females
and patients diagnosed with renal disorder were excluded from the study. For an estimated
prevalence of 25.2% of diabetes in Delhi population[2] and with 5% absolute precision, 95% confidence interval, and 10% confounding variable,
we needed a sample size of 320.
Data Collection
The random blood glucose and HbA1c levels of 461 patient samples (178 male and 283
female) were included in the study. Blood samples were taken on the same day for the
determination of both RBS and HbA1c. The eAG levels (mg/dL) were calculated using
the following formula: 28.7 × HbA1c – 46.7. The samples were divided into four groups
on the basis of HbA1c levels as group 1: HbA1c greater than 9% (poorly controlled
diabetic), group 2: HbA1c 6.5 to 9% (fairly controlled diabetic), group 3: HbA1c 5.7
to 6.4% (prediabetic), and group 4: HbA1c less than 5.7% (nondiabetic).[8]
Glucose levels were determined using the glucose oxidase method in Transasia Erba
analyzer with commercially available Agappe kits. HbA1c levels were determined using
high-performance liquid chromatographic method on Biorad D10 analyzer with Biorad
kits.
Statistical Analysis
Presentation of categorical variables was done in the form of number and percentage
(%). Quantitative data were presented as the means ± standard deviation (SD) and as
median with 25th and 75th percentiles (interquartile range). Data normality was checked
by using the Kolmogorov–Smirnov test. The cases in which the data was not normally
distributed, we used nonparametric tests. The following statistical tests were applied
for the results:
-
The association of the variables which were quantitative and not normally distributed
in nature was analyzed using Mann–Whitney U test (for two groups) and Kruskal–Wallis test (for more than two groups). Wilcoxon
signed ranks test was used for comparison of RBS (mg/dL) and eAG (mg/dL).
-
Spearman's rank correlation coefficient was used for correlation of RBS (mg/dL) with
HbA1c (%) and RBS (mg/dL) with eAG (mg/dL).
Data entry was done in Microsoft Excel spreadsheet and the final analysis was done
with the use of Statistical Package for Social Sciences (SPSS) software (ver 21.0,
IBM, Chicago, Illinois, United States). A p-value of less than 0.05 was considered statistically significant.
Results
A total of 461 patients' data (178 males, 283 females) were recorded for the study.
Age range was between 12 and 90 years and mean age ± SD was 46.89 ± 13.2 years. The
mean ± SD values of HbA1c, eAG, and RBS of the total population were 7.8 ± 2.54%,
177.2 ± 72.89 mg/dL, and 169.53 ± 86.1 mg/dL, respectively ([Table 1], [Fig. 1A] and [B]). There was a statistically significant difference found in all the dependent variables
(HbA1c/RBS/eAG) in two independent groups, that is, males and females ([Table 2]). It showed males have significant higher values of HbA1c, RBS, and eAG as compared
with females ([Fig. 2]). The study sample was divided into four groups on the basis of HbA1c values (group
1 HbA1c > 9%, group 2 HbA1c 6.5–9%, group 3 HbA1c 5.7–6.4%, and group 4 < 5.7%). A
nonparametric test applied to the four groups showed a statistically significant difference
in their RBS and eAG values ([Table 3], [Fig. 3]). As shown in [Table 4] ([Fig. 4A] and [B]) there was a statistically significant positive correlation in both cases: RBS versus
HbA1c and RBS versus eAG (r = 0.782, p ≤ 0.0001, R
2 = 0.612) for the whole study sample. We also observed similar positive correlation
in group 1 sample (r = 0.447, p ≤ 0.0001, R
2 = 0.1998) and group 2 sample (r = 0.322, p ≤ 0.0001, R
2 = 0.1037). No statistically significant relationship was observed between RBS and
eAG in group 3 (r = 0.111, p > 0.05) and group 4 (r = 0.082, p > 0.05).
Table 1
Distribution of baseline characteristics of study subjects
|
Baseline characteristics
|
Frequency
|
Percentage
|
|
Gender
|
|
Female
|
283
|
61.39
|
|
Male
|
178
|
38.61
|
|
HbA1c (%)
|
|
> 9% (poorly controlled diabetic)
|
131
|
28.42
|
|
6.5–9% (fairly controlled diabetic)
|
152
|
32.97
|
|
5.7–6.4% (prediabetic)
|
79
|
17.14
|
|
< 5.7% (nondiabetic)
|
99
|
21.48
|
|
Mean ± SD
|
7.8 ± 2.54
|
|
Median (25th–75th percentile)
|
7 (5.8–9.5)
|
|
Range
|
4–16.6
|
|
Age (y)
|
|
Mean ± SD
|
46.89 ± 13.2
|
|
Median (25th–75th percentile)
|
47 (38–57)
|
|
Range
|
12–90
|
|
Random blood sugar (mg/dL)
|
|
Mean ± SD
|
169.53 ± 86.1
|
|
Median (25th–75th percentile)
|
139 (108–204)
|
|
Range
|
61–627
|
|
Estimated average glucose (mg/dL)
|
|
Mean ± SD
|
177.2 ± 72.89
|
|
Median (25th–75th percentile)
|
154.2 (119.76–225.95)
|
|
Range
|
68.1–429.72
|
Abbreviations: HbA1c, hemoglobin A1c; SD, standard deviation.
Fig. 1 (A) Distribution of baseline characteristics of study subjects. (B) Descriptive statistics of age (years), hemoglobin A1c (HbA1c) (%), random blood
sugar (mg/dL), and estimated average glucose (mg/dL) of study subjects.
Table 2
Association of parameters with gender
|
Parameters
|
Female (n = 283)
|
Male (n = 178)
|
Total
|
p-Value
|
|
Age (y)
|
|
Mean ± SD
|
45.41 ± 13.31
|
49.25 ± 12.68
|
46.89 ± 13.19
|
0.004[a]
|
|
Median (25th–75th percentile)
|
45 (35–56)
|
50 (41–57.75)
|
47 (38–57)
|
|
Range
|
18–80
|
12–90
|
12–90
|
|
HbA1c (%)
|
|
Mean ± SD
|
7.53 ± 2.38
|
8.23 ± 2.73
|
7.8 ± 2.54
|
0.008[a]
|
|
Median (25th–75th percentile)
|
6.8 (5.6–8.95)
|
7.4 (6.1–10.075)
|
7 (5.8–9.5)
|
|
Range
|
4.1–14.4
|
4–16.6
|
4–16.6
|
|
Random blood sugar (mg/dL)
|
|
Mean ± SD
|
162.14 ± 81.64
|
181.27 ± 91.75
|
169.53 ± 86.1
|
0.022[a]
|
|
Median (25th–75th percentile)
|
134 (106.5–190)
|
150 (114.25–228.75)
|
139 (108–204)
|
|
Range
|
61–627
|
71–500
|
61–627
|
|
Estimated average glucose (mg/dL)
|
|
Mean ± SD
|
169.42 ± 68.22
|
189.56 ± 78.38
|
177.2 ± 72.89
|
0.008[a]
|
|
Median (25th–75th percentile)
|
148.46 (114.02–210.165)
|
165.68 (128.37–242.453)
|
154.2 (119.76–225.95)
|
|
Range
|
70.97–366.58
|
68.1–429.72
|
68.1–429.72
|
Abbreviations: HbA1c, hemoglobin A1c; SD, standard deviation.
a Mann–Whitney U test.
Fig. 2 Association of parameters with gender (nonparametric variables).
Table 3
Association of parameters with HbA1c (%)
|
Parameters
|
> 9% (poorly controlled diabetic)
(n = 131)
|
6.5–9% (fairly controlled diabetic)
(n = 152)
|
5.7–6.4% (prediabetic)
(n = 79)
|
< 5.7% (nondiabetic)
(n = 99)
|
Total
|
p-Value
|
|
Age (y)
|
|
Mean ± SD
|
49.19 ± 11.52
|
50.89 ± 11.11
|
48.09 ± 11.39
|
36.76 ± 14.52
|
46.89 ± 13.19
|
< 0.0001[a]
|
|
Median (25th–75th percentile)
|
50 (40–56.5)
|
51 (44–60)
|
50 (40.5–56)
|
35 (26–42)
|
47 (38–57)
|
|
Range
|
18–80
|
21–85
|
24–78
|
12–90
|
12–90
|
|
HbA1c (%)
|
|
Mean ± SD
|
11.25 ± 1.6
|
7.49 ± 0.73
|
6.06 ± 0.22
|
5.1 ± 0.38
|
7.8 ± 2.54
|
< 0.0001[a]
|
|
Median (25th–75th percentile)
|
10.9 (10–12.1)
|
7.4 (6.875–8.1)
|
6.1 (5.9–6.2)
|
5.2 (4.85–5.4)
|
7 (5.8–9.5)
|
|
Range
|
9.1–16.6
|
6.5–9
|
5.7–6.4
|
4–5.6
|
4–16.6
|
|
Random blood sugar (mg/dL)
|
|
Mean ± SD
|
258.16 ± 96.56
|
162.29 ± 51.68
|
121.54 ± 32.5
|
101.65 ± 19.37
|
169.53 ± 86.1
|
< 0.0001[a]
|
|
Median (25th–75th percentile)
|
243 (197–322)
|
151.5 (128.75–176.25)
|
114 (104–127.5)
|
97 (89–110.5)
|
139 (108–204)
|
|
Range
|
82–627
|
85–385
|
61–295
|
69–164
|
61–627
|
|
Estimated average glucose (mg/dL)
|
|
Mean ± SD
|
276.23 ± 45.87
|
168.29 ± 21.04
|
127.28 ± 6.32
|
99.67 ± 10.88
|
177.2 ± 72.89
|
< 0.0001[a]
|
|
Median (25th–75th percentile)
|
266.13 (240.3–300.57)
|
165.68 (150.612–185.77)
|
128.37 (122.63–131.24)
|
102.54 (92.495–108.28)
|
154.2 (119.76–225.95)
|
|
Range
|
214.47–429.72
|
139.85–211.6
|
116.89–136.98
|
68.1–114.02
|
68.1–429.72
|
Abbreviations: HbA1c, hemoglobin A1c; SD, standard deviation.
a Kruskal–Wallis test.
Fig. 3 Association of parameters with hemoglobin A1c (HbA1c) (%) (nonparametric variables).
Table 4
Correlation of random blood sugar (mg/dL) with HbA1c (%) and random blood sugar (mg/dL)
with estimated average glucose (mg/dL)
|
Variables
|
Random blood sugar (mg/dL) and HbA1c (%)
|
Random blood sugar (mg/dL) and estimated average glucose (mg/dL)
|
|
Total study subjects
|
r
|
0.782
|
0.782
|
|
R
2
|
0.612
|
0.612
|
|
p-value
|
< 0.0001
|
< 0.0001
|
|
Female
|
r
|
0.788
|
0.788
|
|
R
2
|
0.621
|
0.621
|
|
p-value
|
< 0.0001
|
< 0.0001
|
|
Male
|
r
|
0.763
|
0.763
|
|
R
2
|
0.582
|
0.582
|
|
p-value
|
< 0.0001
|
< 0.0001
|
|
> 9% (poorly controlled diabetic)
|
r
|
0.447
|
0.447
|
|
R
2
|
0.1998
|
0.1998
|
|
p-value
|
< 0.0001
|
< 0.0001
|
|
6.5–9% (fairly controlled diabetic)
|
r
|
0.322
|
0.322
|
|
R
2
|
0.1037
|
0.1037
|
|
p-value
|
0.0001
|
0.0001
|
|
5.7–6.4% (prediabetic)
|
r
|
0.111
|
0.111
|
|
R
2
|
0.0123
|
0.0123
|
|
p-value
|
0.328
|
0.328
|
|
< 5.7% (nondiabetic)
|
r
|
0.082
|
0.082
|
|
R
2
|
0.0067
|
0.0067
|
|
p-value
|
0.422
|
0.422
|
Abbreviation: HbA1c, hemoglobin A1c.
Note: Spearman rank correlation coefficient.
Fig. 4 (A) Correlation of hemoglobin A1c (HbA1c) (%) with random blood sugar (mg/dL) in total
study subjects. (B) Correlation of estimated average glucose with random blood sugar (mg/dL) in total
study subjects.
[Table 5] ([Fig. 5]) depicts these median values of RBS and eAG show statistically significant difference
(p < 0.001).
Table 5
Descriptive statistics of glycemic parameter of study subjects
|
Glycemic parameter
|
Mean ± SD
|
Median (25th–75th percentile)
|
Range
|
p-Value
|
|
Random blood sugar (mg/dL)
|
169.53 ± 86.1
|
139 (108–204)
|
61–627
|
< 0.0001[a]
|
|
Estimated average glucose (mg/dL)
|
177.2 ± 72.89
|
154.2 (119.76–225.95)
|
68.1–429.72
|
Abbreviation: SD, standard deviation.
a Wilcoxon signed ranks test.
Fig. 5 Descriptive statistics of random blood sugar (mg/dL) and estimated average glucose
(mg/dL) of study subjects.
Discussion
One limitation often associated with HbA1c is the reporting units of mmol/mol and
%, which differs from the usual units of blood glucose monitoring, that is, mg/dL,
often creating a confusing situation for the patients as well as clinicians for comprehension.[10]
To overcome these limitations, international bodies including the American Diabetes
Association and the International Diabetes Federation proposed a mathematical expression
termed eAG, which facilitates comprehension of HbA1c values in units parallel to self-monitoring.[11] Various guidelines recommend reporting eAG with every HbA1c report; however, it
is not widely practiced by the majority of laboratories, and advocacy is required
regarding its use based on evaluation in the local population.[12] With this perspective in mind, we planned to study the association between RBS and
eAG in a cohort of subjects.
This is the first study in Indian population correlating eAG values with RBS values
in both diabetics and nondiabetic subjects. In this study, we found statistically
significant correlation of RBS with eAG in total study subjects and diabetics (poorly
controlled and fairly controlled groups) but no significant correlation was found
between eAG and RBS in nondiabetic and prediabetic groups, which is similar to Kim
et al findings. We also found in our study that RBS values cannot be used interchangeably
with eAG values. Most of the below mentioned studies highlighted an association between
eAG/HbA1c and RBS/fasting blood sugar (FBS)/postprandial blood sugar (PPBS)/self-monitored
mean blood glucose (MBG) in diabetics, this association had not been checked in diabetic,
prediabetic, and nondiabetic subgroups separately, possibly due to the study design
which only included diabetics.
These studies represent association in patients labeled with DM divided into three
groups on the basis of FBS/PPBS values as good control, moderate control, and poorly
controlled glycemic state subjects, stating the statistically significant correlation
of eAG with FBS and PPBS in total diabetic population under study and poorly controlled
diabetic subjects. A study by Kariyawasan found a significant statistical correlation
in both FBS and PPBS with eAG in the groups of patients with moderately poor control.
In those with markedly poor control the FBS did not show a statistical correlation
with eAG, as opposed to the PPBS.[9] Bozkaya et al have found that a strong positive correlation exists between fasting
plasma glucose (FPG) levels and estimated average blood glucose levels (r = 0.757, p = 0.05).[10] Rosediani et al revealed that both PPBS and FBS correlated significantly with HbA1c
but PPBS showed better correlation with HbA1c than FBS (r = 0.604 vs. 0.575).[13] Mahato et al found statistically significant correlation of eAG with FBS (r = 0.61, p < 0.001) and post prandial (PP) blood sugar levels (r = 0.65, p < 0.001).[14] Kim et al found that FPG showed a moderate correlation with eAG (r = 0.672, p < 0.001) in all subjects but when diabetic and nondiabetic subjects were divided
into subgroups according to the FPG level, the correlation between eAG and FPG decreased
in both subgroups as the FPG level decreased.[15] Guan et al found the relationship between HbA1c and FPG changed according to the
different FPG ranges. [16] Azim et al found direct correlation between HbA1c and RBS in diabetics.[17] Ram et al found significant difference in eAG and FBS values.[18] Nkoana and Khine found a positive correlation between self-monitored MBG and HbA1c
in all participants (R
2 = 0.69, p < 0.0001) but clinically significant differences between MBG and eAG values.[19]
There were few limitations of this study. Notably, we did not categorized patients
into type 1 and type 2 diabetics. Moreover, as the complete blood picture was not
available for all subjects, anemic and other endocrine disorder cases were not excluded.
Conclusion
In conclusion, in poorly controlled diabetic care-sensitive group, eAG can serve as
easily comprehensible way to determine average glucose levels with the same reporting
units for self-blood glucose monitoring. This will supplement clinicians to facilitate
care and counsel patients in a more convincing way. Moreover, it can serve as a useful
measure for clinical laboratories of government hospitals in developing countries
to enhance the quality of reporting at no added substantial cost.
