J. Clin. Med. 2020, 9, x doi: FOR PEER REVIEW www.mdpi.comjournal jcm [606825]

J. Clin. Med. 2020, 9, x; doi: FOR PEER REVIEW www.mdpi.com/journal/ jcm
Article
Limited agreement between classifications of
diabetes and prediabetes resulting from the OGTT,
HbA1c, and fasting glu cose tests: Evidence based on
7412 U.S. adults
Larry A. Tucker
Affiliation: College of Life Sciences, Brigham Young University, Provo, Utah, USA 84602
Email: [anonimizat]
Received: date; Accepted: date; Published: date

Abstract: Accurate diagnosis of diabetes is cri tical for prompt treatment. This investigation was
designed to determine the degree of concordance resulting from diabetes tests of fasting plasma
glucose (FPG) and hemoglobin A1c (A1c) compared to the oral glucose tolerance test (OGTT) in
adults undiagnos ed with diabetes. A secondary aim was to assess agreement among the 3 tests for
prediabetes. A third objective was to measure concordance within subsamples of U.S. women and
men, and within young, middle -age, and older adults. A total of 7,412 randomly sel ected adults
from the National Health and Nutrition Examination Survey (NHANES) were included. Sample
weights were used in the analyses so the results can be generalized to all U.S. adults. With outcomes
classified as normal, prediabetes, or diabetes, acco rding to standard guidelines, overall agreement
was low. With an OGTT diagnosis of diabetes, the A1c test agreed only 34% of the time and the FPG
assay resulted in only 44% concordance. Delimited to older adults, agreement between the OGTT
and A1c was only 25%, and between the OGTT and FPG concordance was only 33.5%. Clearly, when
the outcome is a diagnosis of diabetes, the A1c and FPG tests disagree with the OGTT far more than
they agree.
Keywords: hyperglycemia ; glycated hemoglobin ; NHANES ; blood sugar ; sensitivity ; specificity

1. Introduction
Nearly a half -billion individuals have diabetes, almost 10% of the world’s adults [1]. By 2045,
the prevalence is expected to reach 700 million. In 2019, more tha n four million deaths resulted from
diabetes and its complications [1]. Clearly, diabetes is one of the most devastating global diseases.
Many individuals with diabetes have not been diagnosed. According to the International
Diabetes Federation, one -half of those with diabetes do not know they have the disease [1]. As a
result, treatment is delayed, and health risks are increased. Millions of other adults have impaired
glucose tolerance or prediabetes. Collectively, diabetes, undiagnosed diabetes, and prediabetes have
created one of the most significant worldwide p ublic health challenges.
The human and economic costs associated with diabetes are staggering. Consequently,
significant efforts are expended each year with the goal of improving the management of diabetes.
However, before diabetes can be successfully tre ated, it must be accurately diagnosed.
The 2019 “Standards of Medical Care in Diabetes” by the American Diabetes Association (ADA)
[2] lists three methods for diagnosing diabetes: 1) fasting plasma glucose (FPG), 2) the oral glucose

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 2 of 16
tolerance test (OGTT), or 3) glycated hemoglobin, HbA1c (A1c). The ADA document also indicates
that these same three tests may be employed to detect predi abetes .
Well -accepted criteria have been established for differentiating among those with diabetes,
prediabetes, or normal glucose metabolism. The 2019 ADA report states that the FPG, OGTT, and
A1c are each appropriate for diagnosing diabetes, generally. However, research shows that
agreement between the FPG and the OGTT is less than impressive, and classifications based on the
glucose -based protocols and A1c are often not in alignment . Research shows that the OGTT tends to
diagnose more individuals with diabetes or prediabetes than the FPG or A1c tests [3].
Because accurate diagnosis is critical for prompt and appropriate treatment of diabetes and
prediabetes, a variety of investigations have been conducted to determine the relationship betwee n
outcomes produced by the OGTT, FPG, and the A1c assays [4-8]. Results have been mixed. Moreover,
few of the studies have had good exte rnal validity, and those with generalizable findings have
become outdated. Consequently, the present investigation was conducted.
The primary aim of the current study was to evaluate the extent of agreement resulting from
diabetes tests of fasting plasma g lucose (FPG) and hemoglobin A1c (A1c) compared to the oral
glucose tolerance test (OGTT) in previously undiagnosed adults. A secondary purpose was to
identify the degree of concordance among the three diagnostic tests for prediabetes. A third objective
was to determine the extent of concordance within subsamples of U.S. women and men, and within
young, middle -age, and older U.S. adults.
2. Materials & Methods
2.1. Sample
Data for the present investigation were obtained from the ongoing National Health and
Nutrition Examination Survey (NHANES) [9]. NHANES is administered by the U.S. Centers for
Disease Control and Prevention (CDC) under the direction of the National Center for Health
Statistics. The surve y is conducted by the CDC to evaluate the health and lifestyle of individuals
living within the country. A complex, multi -level sampling strategy is employed so findings can be
generalized nationwide.
NHANES gathers data using two -year cycles. The present study employed data from the 2009 –
2010, 2011 -2012, 2013 -2014, and 2015 -2016 cycles. All the NHANES data sets are available online for
free [9]. Data derived from the OGTT, A1c, and FPG assays for the N HANES 2017 -2018 cycle are not
yet available and therefore could not be incorporated into the present study.
Participants were 20 -80 years of age. To be included in the study, subjects had to have complete
data for age, sex, the oral glucose tolerance test (OGTT), hemoglobin A1c (A1c), and fasting plasma
glucose (FPG). Information about race, body weight, BMI, and smoking were also collected to help
describe the sample.
Adults who reported they were on medication to control their blood sugar were excluded from
the sample (n=1,170). Women who were pregnant were also excluded (n=14). A total of 164 adults
refused to drink or did not consume all the OGTT glucose solution, and another 18 were excluded
because they reported they did not fast for at least 9 hours , the minimum NHANES used for the FPG
test. There were 66 excluded because they came late or left early, 42 who had unsuccess
venipunctures, 16 who were faint during the blood draw, 2 who refused venipuncture, and 551 who
were excluded by NHANES for “other ” reasons.
Prior to the fasting blood glucose test, subjects were given a fasting questionnaire to assess
potential issues associated with fasting [10]. The fasting questionnaire asked about the consumption
of coffee or tea, alcohol, gum, mints, lozenges, cough drops, antacids, laxatives, antidiarrheals, and
dietary supplements during the fasting period. As part of the current study, participants who
reported consuming a prohibited item while fasting were compared to those who reported no intake
of any food or beverage, except water, to determine if fasting b lood glucose levels were affected by
intake of the prohibited item. There were no differences in fasting blood glucose levels associated

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 3 of 16
with any of the items, except alcohol. Therefore, subjects who reported consuming alcohol while
“fasting” were excluded from the study (n=44).
There were 7,412 participants, 3,809 women and 3,603 men, who fasted, completed the 3
required blood tests, and fulfilled the other study requirements. Written consent was obtained from
each subject. Data collection was carried out following the rules of the Declaration of Helsinki. The
Ethics Review Board of the National Center for Health Statistics approved data collection and posting
the data online for public use (ethical approval number) [11].
2.2. Methods
A total of 8 variables were used in this study. The key variables were outcomes of the oral
glucose tolerance test, the glycated hemoglobin test (A1c), and the fasting plasma glucose test (FPG).
Data on age and sex were used to subdivide the sample. Information about race, height and weight,
body mass index (BMI) and cigarette smoking were collected to help describe the sample.
2.2.1. Race
The following categories were used by NHANES to define race: Mexican American, Non –
Hispanic black, Non -Hispanic white, Other Hispanic, and Other race or Multi -racial.
2.2.2. Height and Weight
A digital scale was employed to measure body weight. During the assessment, subjects wore
only their underwear, a disposable paper gown, and foam slip pers [12]. A stadiometer was used to
measure standing height. Height was assessed with both feet flat on the floor and toes angled
outwards. Heels, buttocks, shoulder blades, and back of the head were required to be against the wall
[12].
2.2.3. Body Mass Index
Both height and weight were used to calculate the body mass index (BMI). BMI is a common
measure of body weight independent of height. It is calculated as weight in kilograms divided by
height in meters squar ed.
2.2.4. Smoking
Cigarette smoking was assessed by questionnaire. Individuals reported the number of cigarettes
they smoked per day over the past 30 days.
2.2.5. Oral Glucose Tolerance Test and the Fasting Glucose Test
The NHANES quality control and qu ality assurance protocols met the 1988 Clinical Laboratory
Improvement Act mandates. Plasma specimens were processed, stored and shipped to Fairview
Medical Center Laboratory at the University of Minnesota, Minneapolis Minnesota, U.S.A., for
analysis. Gluc ose levels were measured by a hexokinase method [13]. It is an endpoint enzymatic
method with a sample blank correction. According to NHANES, “using this enzymatic method,
glucose is converted to glucose -6-phosphate (G -6-P) by hexokinase in the presence of ATP, a
phosphate donor. Glucose -6-phosphate dehydrogenase then converts the G -6-P to gluconate -6-P in
the presence of NADP+. As the N ADP+ is reduced to NADPH during this reaction, the resulting
increase in absorbance at 340 nm (secondary wavelength = 700 nm) is measured” [13]. This is an
endpoint reaction that is specific for glucose. Detailed specimen collection and processing
instructions are discussed in the NHANES Laborator y/Medical Technologists Procedures
Manual , available online [14].
A fasting glucose blood test was performed on all participants who were examined in the
morning session, after fasting at least 9 hours. After the initial blood draw, which was used to assess

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 4 of 16
fasting glucose levels, participants were asked to drink a 75 -gram dose of glucose, TrutolTM. A
second venipuncture was performed 2 hours (± 15 minutes) after drinking the TrutolTM solution.
There were no changes to equipment, lab methods, or the lab site associated with the OGTT and
the FPG test until the 2015 -2016 N HANES cycle [13]. During prior years, the Cobas C501 analyzer
(Roche Diagnostics, Basel, Switzerland) was employed to measure glucose levels. NHANES then
changed to the Cobas C311 analyzer. To equalize the results, NHANES compared a random sample
of 165 individuals using both machines. They determined that the C311 device produced glucose
levels that were 2% higher than the values from t he Cobas C501 analyzer, with the same variation. A
weighted Deming regression was employed to adjust the 2015 -2016 plasma glucose values so they
matched the results derived from the C501 analyzer. The correlation between the bridging
measurements was 0.999 (P<0.0001) [13].
2.2.6. Glycohemoglobin (A1c)
Similar to the OGTT and FPG test, blood specimens were processed, stored and shipped to
Fairview Medical Center Laboratory at the University of Minnesota, Minneapolis Minnesota, USA
for analysis of glycohemoglobin (A1c). There were no changes to equipment, lab methods, or the lab
site associated with the A1c test during the eight years inclu ded in the present investigation. The
Tosoh Automated Glycohemoglobin Analyzer HLC -723G8 (South San Francisco, California, USA)
was used to measure A1c [15]. According to NHANES, “in this assay, the stable (SA1c) and labile
(LA1c) A1c forms can be individually resolved on the chromatogram without manual pretreatment,
allowing accurate measurement of the stable form of HbA1c. The analyzer dilutes the whole blood
specimen with a hemolysis solution, and then injects a small volume of the treated specimen onto the
HPLC analytical column. Separation is achieved by utilizing differences in ionic interactions between
the cation exchange group on the column resin surface an d the hemoglobin components. The
hemoglobin fractions (A1c, A1b, F, LA1c, SA1c, A0 and H -Var) are subsequently removed from the
column material by step -wise elution using elution buffers each with a different salt concentration.
The separated hemoglobin co mponents pass through the photometer flow cell where the analyzer
measures changes in absorbance at 415 nm. The analyzer integrates and reduces the raw data, and
then calculates the relative percentages of each hemoglobin fraction” [15]. Comprehensive specimen
collection and processing instructions are discussed in the NHANES Laboratory Procedures Manual,
which is available online [14].
2.3. Diabetes classifications
In their 2019 “Standards of Medical Care” recommendations, the American Diabetes Association
(ADA) states that diabetes may be diagnosed using the fasting plasma glucose (FPG) test, the OGTT,
or the A 1c assay [2]. The ADA document also indicates that the same assessments can be used to
diagnose prediabetes [2].
Well -accepted criteria have been established for the diagnosis of diabetes: 1) FPG ≥ 126 mg/dL
(7.0 mmol/L). 2) For the OGTT, two -hours after consuming a 75 -gram anhydrous gluc ose load
dissolved in water, a plasma glucose level ≥ 200 mg/dL (11.1 mmol/L). 3) A1C ≥ 6.5% (48 mmol/mol).
Using the same tests, prediabetes is defined as 1) FPG 1 00 mg/dL (5.6 mmol/L) to 125 mg/dL (6.9
mmol/L), 2) OGTT at 2 hours: 140 mg/dL (7.8 mmol/L) to 199 mg/dL (11.0 mmol/L), 3) A1C: 5.7% (39
mmol/mol) to 6.4% (47 mmol/mol) [2].
2.4. Statistical Analysis
NHANES assigns each participant a person -level sample weight, which allows findings to be
generalized to all civilian, non -institutionalized adults living in the United States. The sample weights
represent the unequal probability of selection, nonresponse corrections, and adjustments for
independent population controls [16].
For the current st udy, sample weights were based on 8 years of the Oral Glucose Tolerance Test
(OGTT) records, as recommended by NHANES. SAS SurveyMeans was used to generate weighted

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 5 of 16
means and SAS SurveyFreq was utilized to provide weighted frequencies, including agreement and
disagreement between diabetes classifications for the OGTT, A1c, and FPG tests. SAS SurveyFreq,
with the Jackknife variance method, was used to calculate weighted Kappa values. Test sensitivity
and specificity were used to describe the diagnostic accur acy of the FPG and A1c tests compared to
the OGTT. Sensitivity and specificity were calculated using weighted prevalence rather than the
sample size (i.e., n) within each cell of the 2×2 matrices. The statistical analyses were calculated using
SAS Version 9.4 (SAS Institute, Inc., Cary, NC).
3. Results
Individual sample weights provided by NHANES were used for each analysis, so all the results
are generalizable to the U.S. adult population. Average ( ± SE) age of the sample was 46.2 ±0.3 years
and the range was 20 -80 years. Means (±SE) for OGTT, A1c, and FPG were 115.0±0.6 mg/dL, 5.5±0.01
%, and 100.1±0.3 mg/dL, respectively. The sample was comprised of 3809 women and 3603 men, a
total of 7412 adults. The racial composition of the sample, as defined by NHANES , was 66.8% non –
Hispanic white; 11.0% non -Hispanic black; 8.4% Mexican American; 7.7% Other or multi -race; and
6.0% Other Hispanic. Approximately 25% of the participants were from each two -year sampling
cycle: 2009 -2010, 2011 -2012, 2013 -2014, and 2015 -2016 . Table 1 displays a range of percentiles,
including the median (50th percentile), for the key continuous variables for women and men
considered separately and for all subjects combined.
Table 1. Percentiles with standard errors for continuous variables i n U.S. women (n=3,809) and men
(n=3,603), and combined data (n=7,412) .
Variable Percentile (± SE)
5th 25th 50th 75th 95th
2 hr OGTT (mg/dL)
Women 68.0±1.3 87.8±0.6 105.1 ±0.8 131.6 ±1.2 203.0 ±2.9
Men 58.6±1.3 84.7±0.7 103.1 ±0.9 129.0 ±1.1 198.0 ±4.8
Combined 62.4±1.0 86.1±0.6 104.3 ±0.6 130.4 ±0.8 201.0 ±2.7
A1c (%)
Women 4.8±0.02 5.1±0.01 5.4±0.01 5.6±0.01 6.1±0.02
Men 4.7±0.02 5.1±0.01 5.4±0.01 5.6±0.01 6.1±0.04
Combined 4.8±0.01 5.1±0.01 5.4±0.01 5.6±0.01 6.1±0.02
Fasting Plasma Glucose (mg/dL)
Women 81.7±0.3 89.6±0.2 95.4±0.3 102.7 ±0.4 118.8 ±1.0
Men 85.6±0.6 93.6±0.3 99.6±0.3 106.8 ±0.4 124.0±1.0
Combined 82.8±0.3 91.4±0.2 97.6±0.2 104.9 ±0.3 121.1 ±0.8
Age (years)
Women 21.5±0.3 31.6±0.5 46.1±0.6 59.3±0.4 77.3±0.7
Men 21.3±0.2 30.9±0.4 43.2±0.7 57.3±0.6 74.2±0.6
Combined 21.4±0.8 31.2±0.4 44.6±0.6 58.4±0.2 76.0±0.5
Body weight (kg)
Women 50.3±0.4 61.7±0.4 71.9±0.5 86.3±0.8 115.6 ±1.7
Men 61.6±0.5 74.6±0.5 85.8±0.7 99.5±0.8 125.1 ±1.6
Combined 53.4±0.4 66.7±0.3 79.1±0.4 93.8±0.5 121.7 ±1.1
Smoking (cigarettes per day)
Women 0.0±0.4 0.0±0.4 0.0±0.4 0.0±0.4 14.5±1.3
Men 0.0±0.3 0.0±0.3 0.0±0.3 0.0±0.3 19.4±1.2
Combined 0.0±0.3 0.0±0.3 0.0±0.3 0.0±0.3 19.1±1.2
BMI
Women 19.7±0.2 23.5±0.2 27.5±0.2 32.9±0.3 42.8±0.6
Men 20.6±0.2 24.6±0.2 27.8±0.2 31.6±0.2 39.4±0.6
Combined 20.0±0.1 24.0±0.1 27.7±0.1 32.1±0.2 41.5±0.5
SE is standard error of the percentage.
3.1. Comparing the OGTT and A1c

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 6 of 16
For each analysis comparing the OGTT and the A1c, the OGTT was used as the referent. As
shown in Table 2, with all subjects combined together (n=7412) and with the OGTT and the A1c
outcomes classified as normal, prediabetes, or diabetic according to standa rd guidelines, the
weighted Kappa (±SE) was 0.355±0.015. A total of 34.1% of participants classified as diabetic
according to the OGTT were also classified as diabetic according to their A1c findings. Of U.S. adults
determined to be prediabetic by the OGTT , 46.1% were also defined as prediabetic using the A1c test.
Adults defined by their OGTT results as normal were mostly found to be classified as normal by their
A1c results, with 80.7% concordance. With the OGTT and A1c outcomes summarized using a simple
2×2 matrix (non -diabetic or diabetic for the OGTT and the A1c), rather than the 3×3 matrix displayed
in Table 2, sensitivity was 34.1% and specificity was 99.6%.
As shown in Table 2, agreement between the OGTT and the A1c classifications of normal and
prediabetes were almost identical for women and men. However, concordance was lower for women
defined as diabetic than men classified as diabetic, with agreement levels of 29.5% for U.S. women
and 39.8% for U.S. men. Weighted Kappa (±SE) for the 3×3 matrix sh own in Table 2 comparing the
OGTT and the A1c results for women was 0.346±0.016. For U.S. men, weighted Kappa (±SE) was
0.364±0.024. Using a 2×2 matrix, with classifications of non -diabetic or diabetic, sensitivity was 29.5%
for women only and specificity was 99.5%. With the sample limited to men, sensitivity was 39.8%
and specificity was 99.6%, after accounting for individual sample weights.
Table 2. Agreement between test results for the OGTT and the A1c in U.S. women and men analyzed
separately and comb ined, 2009 -2016.
OGTT Classification A1c Classification
All Subjects A1c: Normal A1c: Prediabetes A1c: Diabetes
OGTT: Normal
n=5678 n=4295
row %: 80.7 n=1370
row %: 19.2 n=13
row %: 0.1
OGTT: Prediabetes
n=1221 n=581
row %: 51.8 n=599
row %: 46.1 n=41
row %: 2.1
OGTT: Diabetes
n=513 n=83
row %: 18.6 n=244
row %: 47.3 n=186
row %: 34.1
Column: n=7412 Column: n=4959 Column: n=2213 Column: n=240
Women Only A1c: Normal A1c: Prediabetes A1c: Diabetes
OGTT: Normal
n=2922 n=2231
row %: 80.4 n=686
row %: 19.5 n=5
row %: 0.1
OGTT: Prediabetes
n=625 n=298
row %: 51.1 n=306
row %: 46.4 n=21
row %: 2.5
OGTT: Diabetes
n=262 n=47
row %: 20.9 n=129
row %: 49.5 n=86
row %: 29.5
Column: n=3809 Column: n=2576 Column: n=1121 Column: n=112
Men Only A1c: Normal A1c: Prediabetes A1c: Diabetes
OGTT: Normal
n=2756 n=2064
row %: 81.1 n=684
row %: 18.8 n=8
row %: 0.1
OGTT: Prediabetes
n=596 n=283
row %: 52.6 n=293
row %: 45.7 n=20
row %: 1.7
OGTT: Diabetes
n=251 n=36
row %: 15.6 n=115
row %: 44.6 n=100
row %: 39.8
Column: n=3603 Column: n=2383 Column: n=1092 Column: n=128
Note: SE: standard error of the weighted Kappa. OGTT: Oral glucose tolerance test. For the 2 -hr
OGTT, “Normal” was defined as a plasma glucose level of < 140 mg/dL (< 7.8 mmol/L), “Prediabetes”
was 140 mg/dL (7.8 mmol/L) to 199 mg/dL (11.0 mmol/L), and “Diabetes” was defined as a 2 -hr
glucose level ≥ 200 mg/dL (11.1 mmol/L). For the A1c test , “Normal” was < 5.7% (< 39 mmol/mol),
“Prediabetes” was 5.7% (39 mmol/mol) to 6.4% (47 mmol/mol), and a d iagnosis of “Diabetes” was
defined as an A1C ≥ 6.5% (≥ 48 mmol/mol). Because NHANES sample weights were applied to each
participant, the sample size of each category should be interpreted using row percentages, which have
been adjusted based on individual sample weights, not “n”. Results using “n” do not reflect the effect
of the NHANES sample weights.

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 7 of 16
Table 3 shows the level of agreement between the OGTT classifications and the A1c results by
age group, 20 -39, 40 -59, and 60 -80 years. Among young adults (n=2669), those found to be diabetic
using the OGTT were classified as diabetic using the A1c test less than half the time, with 46.4%
agreement. Agreement between classifications of prediabetes for both tests was much lower,
however, with 26.1% concordanc e. Individuals classified as normal by the OGTT were also likely to
be labeled normal by the A1c assessment, with 90.3% concordance. Weighted Kappa (± SE) for U.S.
young adults for the OGTT and the A1c was 0.277±0.030. Based on a 2×2 matrix, with subjects
classified as non -diabetic or diabetic, sensitivity was 46.4% and specificity was 99.9%, after taking
into account individual sample weights.
Among middle -age adults (n=2617), as shown in Table 3, individuals diagnosed as diabetic using
the OGTT were also identified as diabetic by the A1c assessment 42.1% of the time. Those labeled
prediabetic by the OGTT were also classified as prediabetic using the A1c test 45.9% of the time. There
was 77.8% overlap for those identified as normal using the OGTT and normal according to the A1c
assessment. The weighted Kappa (± SE) results for U.S. middle -age adults comparing agreement
between the OGTT and the A1c was 0.330±0.025. Displayed as a basic 2×2 matrix, with categories of
non-diabetic or diabetic, sensitivity was 4 2.1% and specificity was 99.5%, based on calculations
performed using NHANES sample weights.
As shown in Table 3, older U.S. adults (n=2126) who were labeled diabetic by the OGTT were
also labeled diabetic by the A1c assessment 25.2% of the time. Agreement for classifications of
prediabetes was 55.8% and overlap for classifications of normal was 64.0%. For older U.S. adults, the
weighted Kappa (± SE) results comparing the OGTT and the A1c in the 3×3 matrix shown in Table 3
was 0.301±0.024. Based on a 2×2 ma trix, with participants identified as either non -diabetic or diabetic,
sensitivity was 25.2% in older U.S. adults and specificity was 99.1%, after applying individual sample
weights.
Table 3. Agreement between diabetes classifications based on the OGTT an d the A1c in U.S. adults
analyzed by age group, 2009 -2016.
OGTT Classification A1c Classification
Young Adults A1c: Normal A1c: Prediabetes A1c: Diabetes
OGTT: Normal
n=2400 n=2117
row %: 90.3 n=282
row %: 9.7 n=1
row %: 0.0
OGTT: Prediabetes
n=569 n=150
row %: 72.9 n=60
row %: 26.1 n=3
row %: 1.0
OGTT: Diabetes
n=292 n=9
row %: 19.5 n=16
row %: 34.1 n=31
row %: 46.4
Column: n=2669 Column: n=2276 Column: n=358 Column: n=35
Middle -age Adults A1c: Normal A1c: Prediabetes A1c: Diabetes
OGTT: Normal
n=2013 n=1441
row %: 77.8 n=566
row %: 22.1 n=6
row %: 0.2
OGTT: Prediabetes
n=439 n=204
row %: 52.1 n=220
row %: 45.9 n=15
row %: 2.0
OGTT: Diabetes
n=165 n=30
row %: 22.5 n=57
row %: 35.4 n=78
row %: 42.1
Column: n=2617 Column: n=1675 Column: n=843 Column: n=99
Older Adults A1c: Normal A1c: Prediabetes A1c: Diabetes
OGTT: Normal
n=1265 n=737
row %: 64.0 n=522
row %: 35.9 n=6
row %: 0.2
OGTT: Prediabetes
n=569 n=227
row %: 41.3 n=319
row %: 55.8 n=23
row %: 2.8
OGTT: Diabetes
n=292 n=44
row %: 15.5 n=171
row %: 59.2 n=77
row %: 25.2
Column: n=2126 Column: n=1008 Column: n=1012 Column: n=106
Note: SE: standard error of the weighted Kappa. OGTT: Oral glucose tolerance test. For the 2 -hr
OGTT, “Normal” was defined as a plasma glucose level of < 140 mg/dL (< 7.8 mmol/L), “Prediabetes”
was 140 mg/dL (7.8 mmol/L) to 199 mg/dL (11.0 mmol/L), and “Di abetes” was defined as a 2 -hr

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 8 of 16
glucose level ≥ 200 mg/dL (11.1 mmol/L). For the A1c test , “Normal” was < 5.7% (< 39 mmol/mol),
“Prediabetes” was 5.7% (39 mmol/mol) to 6.4% (47 mmol/mol), and a diagnosis of “Diabetes” was
defined as an A1C ≥ 6.5% (≥ 48 mmol /mol). Because NHANES sample weights were applied to each
participant, the sample size of each category should be interpreted using row percentages, which have
been adjusted based on individual sample weights, not “n”. Results using “n” do not reflect the effect
of the NHANES sample weights.
3.2. Comparing the OGTT and the Fasting Plasma Glucose Test (FPG)
Focusing on the total sample (n=7412), as displayed in Table 4, with the OGTT and the fasting
plasma glucose (FPG) outcomes categorized as normal, predia betes, or diabetes, the weighted Kappa
(±SE) was 0.310±0.011. A total of 44.3% of the adults identified as diabetic according to their OGTT
were also diagnosed as diabetic based on their FPG results. Adults labeled prediabetic by the OGTT
were typically de fined as prediabetic using the FPG, with 63.4% agreement. Adults defined by their
OGTT results as normal were usually classified as normal also by their FPG results, with 64.8%
concordance. With the OGTT and the FPG classifications summarized using a simpl e 2×2 matrix (i.e.,
non-diabetic or diabetic), sensitivity was 44.3% and specificity was 98.7%, based on results derived
using individual sample weights and all participants.
As shown in Table 4, concordance between the OGTT and the FPG classifications of normal,
prediabetes, and diabetes were different for men and women. Agreement based on diagnoses of
diabetes using the OGTT, followed by FPG results also concluding diabetes, was higher in men than
women, with 57.9% concordance compared to 33.7% agreement. For classifications of prediabetes,
agreement was higher in men than women, with 59.2% concordance in women and 68.0% in men.
Concordance was higher for women defined as normal compared to men classified as normal, with
OGTT and FPG agreement levels of 73 .4% for U.S. women and 55.3% for men.
Weighted Kappa (±SE) for the 3×3 matrix shown in Table 4 comparing the OGTT classifications
and the FPG results for women was 0.351±0.016. For U.S. men, weighted Kappa (±SE) was
0.274±0.019. Using a straightforward 2x 2 matrix, with classifications of non -diabetic or diabetic,
sensitivity was 33.7% for women and specificity was 99.2%%, calculated using NHANES sample
weights. With the sample delimited to men, sensitivity was 57.9% and specificity was 98.1%, after
account ing for individual sample weights.
Table 4. Agreement between diabetes classifications for the OGTT and the FPG test in U.S. women
and men analyzed separately and combined, 2009 -2016. .
OGTT Classification Fasting Plasma Glucose Classification
All Subjects FPG: Normal FPG: Prediabetes FPG: Diabetes
OGTT: Normal
n=5678 n=3595
row %: 64.8 n=2043
row %: 34.7 n=40
row %: 0.5
OGTT: Prediabetes
n=1221 n=381
row %: 30.9 n=758
row %: 63.4 n=82
row %: 5.8
OGTT: Diabetes
n=513 n=37
row %: 8.9 n=248
row %: 46.8 n=228
row %: 44.3
Column: n=7412 Column: n=4013 Column: n=3049 Column: n=350
Women Only FPG: Normal FPG: Prediabetes FPG: Diabetes
OGTT: Normal
n=2922 n=2130
row: 73.4% n=781
row: 26.3% n=11
row: 0.3%
OGTT: Prediabetes
n=625 n=232
row: 36.8% n=360
row: 59.2% n=33
row: 4.0%
OGTT: Diabetes
n=262 n=30
row: 13.5% n=143
row: 52.8% n=89
row: 33.7%
Column: n=3809 Column: n=2392 Column: n=1284 Column: n=133
Men Only FPG: Normal FPG: Prediabetes FPG: Diabetes
OGTT: Normal
n=2756 n=1465
row: 55.3% n=1262
row: 43.8% n=29
row: 0.8%
OGTT: Prediabetes n=149 n=398 n=49

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 9 of 16
n=596 row: 24.3% row: 68.0% row: 7.7%
OGTT: Diabetes
n=251 n=7
row: 3.0% n=105
row: 39.0% n=139
row: 57.9%
Column: n=3603 Column: n=1621 Column: n=1765 Column: n=217
Note: SE: standard error of the weighted Kappa. For the 2 -hr OGTT, “Normal” was defined as a
plasma glucose level of < 140 mg/dL (< 7.8 mmol/L), “Prediabetes” was 140 mg/dL (7.8 mmol/L) to
199 mg/dL (11.0 mmol/L), and “Diabetes” was defined as a glucose le vel ≥ 200 mg/dL (11.1 mmol/L).
For the FPG (fasting plasma glucose) assessment, “Normal” was < 100 mg/dL (< 7.0 mmol/L),
“Prediabetes” was 100 mg/dL (5.6 mmol/L) to 125 mg/dL (6.9 mmol/L), and a diagnosis of “Diabetes”
was defined as a glucose level ≥ 126 mg/dL (≥ 7.0 mmol/L). Because NHANES sample weights were
applied to each participant, the sample size of each category should be interpreted using row
percentages, which have been adjusted based on individual sample weights, not “n”. Results using
“n” do n ot reflect the effect of the NHANES sample weights.
Table 5 shows the degree of concordance between the OGTT classifications and the FPG
outcomes by age group. Among young adults (n=2669), 20 -39 years old, those diagnosed as diabetic
using the OGTT were la beled diabetic using the FPG test about half the time, with 54.0% agreement.
Concordance between classifications of prediabetes for both tests was slightly lower, with 52.4%
concordance. Young adults labeled normal by the OGTT were frequently labeled norma l by the FPG
assessment, with 73.8% agreement. The weighted Kappa statistic (± SE) for individuals representing
U.S. young adults for the OGTT and the FPG was 0.206±0.025. Using a basic 2×2 matrix, with subjects
classified as non -diabetic or diabetic for t he OGTT and FPG test, sensitivity was 54.0% and specificity
was 99.5%, after taking into account individual sample weights.
Among middle -age adults (n=2617), 40 -59 years old, as shown in Table 5, those diagnosed as
diabetic using the OGTT were also identif ied as diabetic by the FPG test 56.1% of the time. Individuals
labeled prediabetic by the OGTT were also classified as prediabetic using the FPG assessment with
62.6% concordance. There was 60.4% overlap for those identified as normal using the OGTT and
normal according to the FPG test. The weighted Kappa (± SE) results for U.S. middle -age adults based
on the 3×3 matrix shown in Table 5 was 0.294±0.019. Displayed as a 2×2 matrix, with categories of
non-diabetic or diabetic, sensitivity was 56.1% and specifi city was 98.3%, based on calculations
performed using NHANES sample weights.
Table 5. Agreement between test results for the OGTT and FPG in U.S. adults analyzed by age group,
2009 -2016.
OGTT Classification Fasting Plasma Glucose Classification
Young Adults FPG: Normal FPG: Prediabetes FPG: Diabetes
OGTT: Normal
n=2400 n=1760
row %: 73.8 n=632
row %: 26.0 n=8
row %: 0.2
OGTT: Prediabetes
n=213 n=93
row %: 43.8 n=112
row %: 52.4 n=8
row %: 3.8
OGTT: Diabetes
n=56 n=6
row %: 14.6 n=16
row %: 31.4 n=34
row %: 54.0
Column: n=2669 Column: n=1859 Column: n=760 Column: n=50
Middle -age Adults FPG: Normal FPG: Prediabetes FPG: Diabetes
OGTT: Normal
n=2013 n=1186
row %: 60.4 n=808
row %: 38.8 n=19
row %: 0.8
OGTT: Prediabetes
n=439 n=146
row %: 31.3 n=260
row %: 62.6 n=33
row %: 6.1
OGTT: Diabetes
n=165 n=9
row %: 8.0 n=64
row %: 35.9 n=92
row %: 56.1
Column: n=2617 Column: n=1341 Column: n=1132 Column: n=144
Older Adults FPG: Normal FPG: Prediabetes FPG: Diabetes
OGTT: Normal
n=1265 n=649
row %: 51.9 n=603
row %: 47.2 n=13
row %: 0.9
OGTT: Prediabetes n=142 n=386 n=41

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 10 of 16
n=569 row %: 24.2 row %: 69.5 row %: 6.3
OGTT: Diabetes
n=292 n=22
row %: 8.2 n=168
row %: 58.4 n=102
row %: 33.5
Column: n=2126 Column: n=813 Column: n=1157 Column: n=156
Note: SE: standard error of the weighted Kappa. For the 2 -hr OGTT, “Normal” was defined as a
plasma glucose level of < 140 mg/dL (< 7.8 mmol/L), “Prediabetes” was 140 mg/dL (7.8 mmol/L) to
199 mg/dL (11.0 mmol/L), and “Diabetes” was defined as a glucose le vel ≥ 200 mg/dL (11.1 mmol/L).
For the FPG (fasting plasma glucose) assessment, “Normal” was < 100 mg/dL (< 7.0 mmol/L),
“Prediabetes” was 100 mg/dL (5.6 mmol/L) to 125 mg/dL (6.9 mmol/L), and a diagnosis of “Diabetes”
was defined as a glucose level ≥ 126 mg/dL (≥ 7.0 mmol/L). Because NHANES sample weights were
applied to each participant, the sample size of each category should be interpreted using row
percentages, which have been adjusted based on individual sample weights, not “n”. Results using
“n” do n ot reflect the effect of the NHANES sample weights.
As shown in Table 5, older U.S. adults (n=2126), 60 -80 years old, who were labeled diabetic by
the OGTT were also categorized as diabetic by the FPG test 33.5% of the time. Agreement for
classifications o f prediabetes was 69.5% and concordance for classifications of normal was 51.9%. For
older U.S. adults, the weighted Kappa (± SE) results comparing the OGTT and the FPG assessment
in the 3×3 matrix shown in Table 5 was 0.310±0.022. Based on a simple 2×2 ma trix, with participants
identified as either non -diabetic or diabetic, sensitivity was 33.5% in older U.S. adults and specificity
was 97.6%, after applying NHANES sample weights.
4. Discussion
The primary purpose of the present investigation was to determi ne the extent of agreement
between diabetes classifications of normal, prediabetes, and diabetes, based on the OGTT, A1c, and
FPG blood tests using standard cut -points. Another aim was to assess concordance between diabetes
classifications within women and men separately, and within young, middle -age, and older adults
separately. Because the NHANES sample was randomly selected using a sophisticated, multi -level
design, and each statistical analysis was performed accounting for strata, clusters, and individu al
sample weights, the results can be generalized to the non -institutionalized U.S. adult population.
Compared to the FPG and A1c tests, numerous investigations show that the OGTT is the best
single measure for predicting subsequent diabetes, disease, and/ or mortality [17-21]. Although more
time -consuming and difficult to administer for the diagnosis of diabetes, the OGTT is considered by
many to be the gold standard [22-28]. Conseque ntly, in the present study, the OGTT was used as the
referent in each comparison.
4.1. Comparing the OGTT and A1c
Findings showed that classification agreement between the OGTT and the A1c test was generally
poor, especially for the identification of diabe tes. When the entire sample of 7,412 U.S. adults was
analyzed and the OGTT resulted in a classification of diabetes, the A1c test agreed only 34.1% of the
time. In other words, approximately two -thirds of the time when the OGTT indicated diabetes, the
A1c did not. When the sample was delimited to women, the A1c assessment agreed with the OGTT
only 29.5% of the time. Clearly, the two blood tests are not in harmony when the outcome is a
diagnosis of diabetes. Similarly, there was low concordance when the OGTT resulted in a
classification of prediabetes. The A1c test agreed less than half the time (46.1%).
When the sample was confined to specific age groups, young adults, middle -age adults, and
older adults, agreement between the OGTT and A1c was not good withi n any of the age groups.
Alignment was especially poor within the older adult group, 60 -80 years old, when the OGTT
resulted in a classification of diabetes. The A1c assessment agreed only 25.2% of the time. In short,
75% of the time when the OGTT indicate d diabetes, the A1c test classified the individual as normal
or prediabetic, not diabetic.

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 11 of 16
A close look at the research used to validate the A1c as a measure of chronic plasma glucose may
help explain the discrepancies between the OGTT and A1c. The first m ajor study used data from the
Diabetes Control and Complications Trial (DCCT) [29]. Rohlfing et al. employed linear regression
weighted by the number of observations per subject to correlate mean plasma glucose and A1c
findings. The resulting regression equation derived from moni toring 1,439 individuals showed that
A1c accounted for 67% of the variance in mean chronic plasma glucose levels. Moreover, according
to the regression equation derived from the trial, the 95% prediction error for an average subject was
± 69 mg/dL ( ± 3.81 mmol/L) when participants had A1c levels between 6% and 9% [29]. Most would
consider this margin of error too large to be useful.
Because of the high level of prediction error associated with the DCCT equation, in 2008, Nathan
et al. used data from the A1c -Derived Average Glucos e (ADAG) study to develop a more precise
prediction equation for estimating average chronic glucose levels [30]. The equation resulting from
the ADAG investigation is the one primarily used today. The study showed that 84% of the variance
in mean glucose levels could be explained by the prediction equation based on A1c, a muc h stronger
relationship than the one derived from the DCCT [30]. Howeve r, although better, the ADAG equation
still resulted in substantial prediction error. For example, an A1c of 6.0% predicts an average glucose
level of 126 mg/dL (5.4 mmol/L), but the 95% confidence interval (CI) is 100 -152 mg/dL (5.5 -8.5
mmol/L). Similarly , an A1c of 8.0% results in an estimated average glucose level of 183 mg/dL (10.2
mmol/L), with a 95% CI of 147 -217 mg/dL (8.1 -12.1 mmol/L). Note that the predicted value of 183
mg/dL (10.2 mmol/L) is probably not the true value. The true value likely fall s somewhere between
147-217 mg/dL (8.1 -12.1 mmol/L), and there is roughly a 5% chance that the true value falls outside
of the 147 -217 mg/dL (8.1 -12.1 mmol/L) interval. Given a specific A1c value, few clinicians likely
consider the large range within which the true predicted glucose value might fall.
The A1c test can be performed any time of the day and without the need of fasting. These are
significant benefits. Moreover, the A1c assay is designed to produce results representing chronic
blood glucose leve ls over a 2 -3 month period. This is also a desirable quality. Consequently, many
medical practitioners use the A1c instead of the OGTT to diagnose diabetes [31]. However, given the
poor agreement between the OGTT and the A1c based on the present large national sample, use of
the A1c over the OGTT may not be a good choice. Clearly, diabetes classifications resulting from the
two assessments disagree fa r more than they agree.
So why do classifications resulting from the OGTT and the A1c disagree frequently? It is likely
because there are many factors that influence A1c results other than glucose metabolism [32,33] . The
World Health Organization (WHO) reports that A1c level s can be affected by genetic, hematologic,
and illness -related factors, especially anemia [32]. According to Gallagher, erythropoiesis, the
production of red blood cells, can be a significant factor affecting A1c [33]. Moreover, deficiencies in
iron or vitamin B12 can lead to decreased erythropoiesis an d increased A1c. On the other hand,
administration of iron, vitamin B12, or erythropoietin; reticulocytosis, and chronic liver disease can
decrease A1c [33]. Also, large amounts of aspirin, excess alcohol consumption, chronic opiate use,
hyperbilirubinemia, carbamylated hemoglobin, and increased lifespan of erythrocytes, can increase
A1c levels, and hypertriglyceridemia can decrease A1c [33]. Additionally, chemical or genetic
alterations of hemoglobin, liver failure, altered intra -erythrocyte pH, diseases of the spleen,
rheumatoid art hritis, the glycation gap, and a variety of prescription drugs, can influence A1c levels
[33-38]. In other words, although chronic blood glucose concentrations have an effect on A1c levels,
many other factors can increase or decrease A1c as well, which could lead to questionable A1c results
and poor alignment with the OGTT.
As seen in the present study of U.S. adults, diabetes classifications derived from the A1c assay
frequently do not agree with the OGTT. Despite this concern, the A1c test appears to be a good
predictor of disease and prema ture death. Research by Colagiuri et al, using a pooled analysis of 9
studies with over 28,000 individuals, reported that A1c was a significant predictor of diabetic
retinopathy [39]. A recen t study by Mancini et al demonstrated that A1c was a significant predictor
of survival within a sample of high -risk patients followed for 7 years with stable ischemic heart
disease and diabetes [40]. Furthermore, a paper by Mo et al indicated that A1c accounted for

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 12 of 16
significant differences in macrovascular complications within a sample of Chinese individuals with
Type 2 diabetes [41].
In a 15 -year prospective cohort study, Selvin et al s howed that elevated A1c levels increased risk
of all -cause mortality [42]. The investigation indicated that low le vels of A1c (< 5%) also increased
risk of death from any cause. Overall, the study found that the baseline A1c measure was a better
predictor of future disease and death than fasting glucose levels.
In a case -cohort analysis, research by Nathan et al show ed that A1c was related significantly
with retinopathy, nephropathy, and CVD [43]. Additionally, Forrest KY, Becker DJ 2000, found that
A1c increased risk of developing lower -extremity arterial disease (LEAD) over a six -year follow -up
period [44]. Lastly, A1c was related significantly with carotid intima -media wall thickness, after
adjusting for several covariates, in a large, multi -ethnic, cross -sectional investigation by McNeely et
al. Overall, it seems clear that elevated levels of A1c increase risk of numerous diseases.
Although many investigations indicate that elevated levels of A1c increase risk of disease, not
all studies support the link. Pruzin et al found no association between A1c and global or regional
Alzheimer disease pathology [45]. In a prospective cohort study by Orchard et al, after 10 years of
follow -up, baseline A1c showed no relationship with CAD [46]. Moreover, in a cross -sectional
investigation conducte d in 31 centers in 16 European countries, research by Koivisto et al identified
no association between A1c and CVD [47].
4.2. Comparing the OGTT and FPG
Concordance between the OGTT and fasting plasma glucose (FPG) was better than the
agreement between the OGTT and the A1c assay. This is probably because the OGTT and FPG are
the same test administered under different conditions. The A1c assay is an entirely different
assessment and does not measure glucose metabolism directly. Moreover, estimation of chronic
glucose levels from A1c valu es are based on a linear regression equation with significant prediction
error.
In the present study, when the OGTT classified an individual as diabetic with the entire sample
included in the analysis (n=7,412), the FPG test resulted in the same classifica tion 44.3% of the time —
worse than a coin -flip. Delimited to U.S. women only, agreement was lower, 33.7%. The FPG test was
more likely to classify a woman as prediabetic (52.8%) than diabetic (33.7%) when the OGTT labeled
her diabetic. Confined to U.S. men, concordance was about 60%. With the focus on older adults only,
the FPG test agreed with an OGTT classification of diabetes only 33.5% of the time.
Despite the lack of agreement between diabetes classifications based on the OGTT and fasting
plasma glucose , many investigations show that the FPG test is a good predictor of disease and
premature death. For example, in an investigation including 97 prospective studies, fasting glucose
levels were shown to be predictive of increased risk of cancer death (43 st udies), vascular mortality
(50 investigations), and noncancer, nonvascular death (42 studies) [48]. In another meta -analysis,
including 102 prospective investigations and approximately 280,000 individuals, Sarwar et al
reported that fasting blood glucose was non -linearly related to vascular risk and linearly associated
with CHD [49]. Additionally, Park et al studied almost 1.2 million Korean adults prospectiv ely for 16
years and reported that as fasting glucose levels increased, ischemic heart disease, myocardial
infarction, and thrombotic stroke increased in a J -shaped fashion [50].
4.3. Sensitivity and specificity
Sensitivity is the ability of a test (e.g., A1c) to correctly identify individuals with a specific disease
(diabetes). It is sometimes called the true positive rate. Specificity is the ability of a test to correctly
identify those who do not have the disease, the true negative rate. Sensitivity and specificity are both
import ant. In the present study, diagnosis of diabetes and prediabetes was based on results derived
from the OGTT.
Both the A1c assay and the FPG test had nearly perfect specificity or true negative rates. In other
words, both were able to accurately identify a dults who did not have diabetes or prediabetes. This
was partly because the vast majority of adults in the sample did not have diabetes or prediabetes.

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 13 of 16
One could have predicted that every adult in the present study had normal glucose metabolism and
the res ults would have been a true negative rate of better than 75%. However, the A1c and the FPG
tests both had low sensitivity or true positive rates, especially the A1c. With all 7,412 U.S. adults
included in the analysis, the A1c assay was able to detect thos e with diabetes only about one -third of
the time and the FPG test had a true positive rate of only 44%. High specificity with low sensitivity is
a dangerous combination in a clinical setting when the disease under consideration is diabetes.
4.4. Limitation s and strengths
The present study had several limitations. First, diagnosis of diabetes typically requires two
abnormal test outcomes from the same blood sample [51] or from two different test samples [52].
Results of the present investigation were based on single tests for the OGTT, A1c, and FPG. Increased
concordance may have occurred if a confirming test was adm inistered after a diagnosis of diabetes
or prediabetes. Additionally, fasting was self -reported. A fasting questionnaire was used to identify
individuals who failed to fast legitimately, but a partial fast could have been hidden from NHANES.
An illegitimat e fast would have affected the OGTT and the FPG results.
There were also multiple strengths associated with the present investigation. First, the study
sample was large (n=7,412), multiracial, and selected randomly. Also, participants were assigned
indivi dual sample weights by NHANES. The sample weights were included as part of each analysis,
allowing the results to be generalized to all civilian, non -institutionalized adults in the United States.
Second, the OGTT, A1c, and FPG assays were performed using high quality measurement methods.
Lastly, women and men were studied separately and combined. Moreover, young adults, middle –
age adults, and older adults were evaluated separately, also.
5. Conclusions
In conclusion, given current cut -points for classifying diabetes, prediabetes, and normal glucose
metabolism, the present investigation shows that the OGTT, A1c, and FPG test results are out of
alignment. With the OGTT as the standard, the A1c and FPG assays have low sensitivity. The odds
of a false negative diagnosis are high. The A1c and FPG assays frequently indicate that U.S. adults
have normal glucose levels when they are prediabetic or they classify individuals as prediabetic when
they are diabetic. This is concerning. Informing patients that t hey do not have diabetes or prediabetes
when they actually do could have catastrophic consequences. Clearly, more research is warranted.
Acknowledgments: The present study was not funded externally. The author is employed full -time as a
professor at Brigha m Young University. The author gives a heart -felt thanks to the women and men who
participated in the NHANES investigation. Without their participation, this study could not have been
conducted.
Author contributions: The author conceived the study, acquir ed the data, organized the data, analyzed the data,
and wrote the manuscript.
Conflicts of Interest : The author declares no conflict of interest.
Data availability: All NHANES data sets are available online for free to everyone [9].
Ethics statement: Written consent was obtained from each subject. Data collection was carried out following the
rules of the Declaration of Helsinki. The Ethics Review Board of the National Center for Health Statistics
approved data collection and posting the data online for public use [11].
References
1. International Diabetes Fede ration. Diabetes Atlas. www.diabetesatlas.org (Accessed June 12 : 2020).
2. American Diabetes As sociation. 2. Classification and Diagnosis of Diabetes: Standards of Medical Care in
Diabetes -2019. Diabetes Care 2019, 42, S13 -S28.
3. Meijnikman, A.S.; De Block, C.E.M.; Dirinck, E.; Verrijken, A.; Mertens, I.; Corthouts, B.; Van Gaal, L.F. Not
performin g an OGTT results in significant underdiagnosis of (pre)diabetes in a high risk adult Caucasian
population. Int J Obes (Lond) 2017, 41, 1615 -1620.

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 14 of 16
4. Saukkonen, T.; Cederberg, H.; Jokelainen, J.; Laakso, M.; Harkonen, P.; Keinanen -Kiukaanniemi, S.; Rajala,
U. Limited overlap between intermediate hyperglycemia as defined by A1C 5.7 -6.4%, impaired fasting
glucose, and impaired glucose tolerance. Diabetes Care 2011, 34, 2314 -2316.
5. Guo, F.; Moellering, D.R.; Garvey, W.T. Use of HbA1c for diagnoses of diabete s and prediabetes:
comparison with diagnoses based on fasting and 2 -hr glucose values and effects of gender, race, and age.
Metab Syndr Relat Disord 2014, 12, 258 -268.
6. Cowie, C.C.; Rust, K.F.; Byrd -Holt, D.D.; Gregg, E.W.; Ford, E.S.; Geiss, L.S.; Bainb ridge, K.E.; Fradkin, J.E.
Prevalence of diabetes and high risk for diabetes using A1C criteria in the U.S. population in 1988 -2006.
Diabetes Care 2010, 33, 562 -568.
7. Collaboration, N.C.D.R.F. Effects of diabetes definition on global surveillance of diab etes prevalence and
diagnosis: a pooled analysis of 96 population -based studies with 331,288 participants. Lancet Diabetes
Endocrinol 2015, 3, 624 -637.
8. Briker, S.M.; Aduwo, J.Y.; Mugeni, R.; Horlyck -Romanovsky, M.F.; DuBose, C.W.; Mabundo, L.S.;
Hormenu , T.; Chung, S.T.; Ha, J.; Sherman, A.; Sumner, A.E. A1C Underperforms as a Diagnostic Test in
Africans Even in the Absence of Nutritional Deficiencies, Anemia and Hemoglobinopathies: Insight From
the Africans in America Study. Front Endocrinol (Lausanne) 2019, 10, 533.
9. NHANES. Data Files: Questionnaires, Datasets, and Related Documentation.
https://wwwn.cdc.gov/nchs/nhanes/Default.aspx (accessed June 12, 2020).
10. NHANES. Fasting questionna ire: Data documentation, codebook, and frequencies.
https://wwwn.cdc.gov/Nchs/Nhanes/2015 -2016/FASTQX_I.htm (accessed June 12, 2020).
11. NHANES. National Center of Health Statistics Research Ethics Review Board (ERB) Approval.
http://www.cdc.gov/nchs/nhanes/irba98.htm (accessed June 12, 2020).
12. NHANES. Anthropometry Procedures Man ual. https://wwwn.cdc.gov/nchs/data/nhanes/2015 –
2016/manuals/2016_Anthropometry_Procedures_Manual.pdf (accessed June 12, 2020).
13. NHANES. NHA NES: Data Documentation, Codebook, and Frequencies: Two -hour Oral Glucose Tolerance
Test. https://wwwn.cdc.gov/Nchs/Nhanes/2015 -2016/OGTT_I.htm (accessed June 12, 2020).
14. NHANES. ME C Laboratory Procedures Manual. https://wwwn.cdc.gov/nchs/data/nhanes/2013 –
2014/manuals/2013_mec_laboratory_procedures_manual.pdf (accessed Ju ne 12, 2020).
15. NHANES. NHANES: Data Documentation, Codebook, and Frequencies: Glycohemoglobin.
https://wwwn.cdc.gov/Nchs/Nhanes/2013 -2014/GHB_H.htm (June 12, 2020).
16. Johnson, C.L.; Paulose -Ram, R.; Ogden, C.L.; Carroll, M.D.; Kruszon -Moran, D.; Dohrmann, S.M.; Curtin,
L.R. National health and nutrition examination survey: analytic guidelines, 1999 -2010. Vital Health Stat 2
2013, 1 -24.
17. Lu, J.; He, J.; Li, M.; Tang, X.; Hu, R.; Shi, L.; Su, Q.; Peng, K.; Xu, M.; Xu, Y.; Chen, Y.; Yu, X.; Yan, L.; Wang,
T.; Zhao, Z.; Qin, G.; Wan, Q.; Chen, G.; Dai, M.; Zhang, D.; Gao, Z.; Wang, G.; Shen, F.; Luo, Z.; Qin, Y.;
Chen, L.; Huo, Y.; Li, Q.; Ye , Z.; Zhang, Y.; Du, R.; Cheng, D.; Liu, C.; Wang, Y.; Wu, S.; Yang, T.; Deng, H.;
Li, D.; Lai, S.; Bloomgarden, Z.T.; Chen, L.; Zhao, J.; Mu, Y.; Ning, G.; Wang, W.; Bi, Y.; Group, C.S.
Predictive Value of Fasting Glucose, Postload Glucose, and Hemoglobin A1c on Risk of Diabetes and
Complications in Chinese Adults. Diabetes Care 2019, 42, 1539 -1548.
18. Qiao, Q.; Dekker, J.M.; de Vegt, F.; Nijpels, G.; Nissinen, A.; Stehouwer, C.D.; Bouter, L.M.; Heine, R.J.;
Tuomilehto, J. Two prospective studies found th at elevated 2 -hr glucose predicted male mortality
independent of fasting glucose and HbA1c. J Clin Epidemiol 2004, 57, 590 -596.
19. Meigs, J.B.; Nathan, D.M.; D'Agostino, R.B., Sr.; Wilson, P.W.; Framingham Offspring, S. Fasting and
postchallenge glycemia and cardiovascular disease risk: the Framingham Offspring Study. Diabetes Care
2002, 25, 1845 -1850.
20. Lind, M.; Tuomilehto, J.; Uusitupa, M.; Nerman, O.; Eriksson, J.; Ilanne -Parikka, P.; Keinanen –
Kiukaanniemi, S.; Peltonen, M.; Pivodic, A.; Lindstrom, J . The association between HbA1c, fasting glucose,
1-hour glucose and 2 -hour glucose during an oral glucose tolerance test and cardiovascular disease in
individuals with elevated risk for diabetes. PLoS One 2014, 9, e109506.
21. Shahim, B.; De Bacquer, D.; De Backer, G.; Gyberg, V.; Kotseva, K.; Mellbin, L.; Schnell, O.; Tuomilehto, J.;
Wood, D.; Ryden, L. The Prognostic Value of Fasting Plasma Glucose, Two -Hour Postload Glucose, and
HbA1c in Patients With Coronary Artery Disease: A Report From EUROASPIRE IV : A Survey From the
European Society of Cardiology. Diabetes Care 2017, 40, 1233 -1240.
22. Phillips, P.J. Oral glucose tolerance testing. Aust Fam Physician 2012, 41, 391 -393.

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 15 of 16
23. Jackson, S.L.; Safo, S.E.; Staimez, L.R.; Olson, D.E.; Narayan, K.M.V.; Long , Q.; Lipscomb, J.; Rhee, M.K.;
Wilson, P.W.F.; Tomolo, A.M.; Phillips, L.S. Glucose challenge test screening for prediabetes and early
diabetes. Diabet Med 2017, 34, 716 -724.
24. Shapiro, A.J.; Holden, E.C.; McGovern, P.G.; Alderson, D.; Morelli, S.S. Com parison of 2 -Hour Oral Glucose
Tolerance Test and Hemoglobin A1C in the Identification of Pre -Diabetes in Women with Infertility and
Recurrent Pregnancy Loss. Clin Med Insights Reprod Health 2019, 13, 1179558119831280.
25. Khan, S.H.; Ijaz, A.; Bokhari, S. A.; Hanif, M.S.; Azam, N. Frequency of impaired glucose tolerance and
diabetes mellitus in subjects with fasting blood glucose below 6.1 mmol/L (110 mg/dL). East Mediterr Health
J 2013, 19, 175 -180.
26. Sehrawat, T.; Jindal, A.; Kohli, P.; Thour, A.; Kaur, J.; Sachdev, A.; Gupta, Y. Utility and Limitations of
Glycated Hemoglobin (HbA1c) in Patients with Liver Cirrhosis as Compared with Oral Glucose Tolerance
Test for Diagnosis of Diabetes. Diabetes Ther 2018, 9, 243 -251.
27. Andersen, M.; Glintborg, D. Diag nosis and follow -up of type 2 diabetes in women with PCOS: a role for
OGTT? Eur J Endocrinol 2018, 179, D1-D14.
28. Picon, M.J.; Murri, M.; Munoz, A.; Fernandez -Garcia, J.C.; Gomez -Huelgas, R.; Tinahones, F.J. Hemoglobin
A1c versus oral glucose tolerance t est in postpartum diabetes screening. Diabetes Care 2012, 35, 1648 -1653.
29. Rohlfing, C.L.; Wiedmeyer, H.M.; Little, R.R.; England, J.D.; Tennill, A.; Goldstein, D.E. Defining the
relationship between plasma glucose and HbA(1c): analysis of glucose profil es and HbA(1c) in the Diabetes
Control and Complications Trial. Diabetes Care 2002, 25, 275 -278.
30. Nathan, D.M.; Kuenen, J.; Borg, R.; Zheng, H.; Schoenfeld, D.; Heine, R.J.; Group, A.c. -D.A.G.S. Translating
the A1C assay into estimated average glucose values. Diabetes Care 2008, 31, 1473 -1478.
31. World Health Organization. Use of Glycated Haemoglobin (HbA1c) in the Diagnosis of Diabetes Mellitus:
Abbreviated Report of a WHO Consultation ; Geneva, 2011.
32. International Expert, C. International Expert Committee report on the role of the A1C assay in the diagnosis
of diabetes. Diabetes Care 2009, 32, 132 7-1334.
33. Gallagher, E.J.; Le Roith, D.; Bloomgarden, Z. Review of hemoglobin A(1c) in the management of diabetes.
J Diabetes 2009, 1, 9-17.
34. Gould, B.J.; Davie, S.J.; Yudkin, J.S. Investigation of the mechanism underlying the variability of glycated
haemoglobin in non -diabetic subjects not related to glycaemia. Clin Chim Acta 1997, 260, 49-64.
35. Snieder, H.; Sawtell, P.A.; Ross, L.; Walker, J.; Spector, T.D.; Leslie, R.D. HbA(1c) levels are genetically
determined even in type 1 diabetes: evidence fr om healthy and diabetic twins. Diabetes 2001, 50, 2858 -2863.
36. Cohen, R.M.; Snieder, H.; Lindsell, C.J.; Beyan, H.; Hawa, M.I.; Blinko, S.; Edwards, R.; Spector, T.D.; Leslie,
R.D. Evidence for independent heritability of the glycation gap (glycosylation gap) fraction of HbA1c in
nondiabetic twins. Diabetes Care 2006, 29, 1739 -1743.
37. Cohen, R.M.; Franco, R.S.; Khera, P.K.; Smith, E.P.; Lindsell, C.J.; Ciraolo, P.J.; Palascak, M.B.; Joiner, C.H.
Red cell life span heterogeneity in hematologically normal people is sufficient to alter HbA1c. Blood 2008,
112, 4284 -4291.
38. Chalew, S.A.; McCarter, R.J.; Thomas, J.; Thomson, J.L.; Hempe, J.M. A comparison of the Glycosylation
Gap and Hemoglobin Glycation Index in patients with diabetes. J Diabetes Complicati ons 2005, 19, 218 -222.
39. Colagiuri, S.; Lee, C.M.; Wong, T.Y.; Balkau, B.; Shaw, J.E.; Borch -Johnsen, K.; Group, D. -C.W. Glycemic
thresholds for diabetes -specific retinopathy: implications for diagnostic criteria for diabetes. Diabetes Care
2011, 34, 145 -150.
40. Mancini, G.B.J.; Maron, D.J.; Hartigan, P.M.; Spertus, J.A.; Kostuk, W.J.; Berman, D.S.; Teo, K.K.; Weintraub,
W.S.; Boden, W.E.; Group, C.T.R. Lifestyle, Glycosylated Hemoglobin A1c, and Survival Among Patients
With Stable Ischemic Heart Disease and Diabetes. J Am Coll Cardiol 2019, 73, 2049 -2058.
41. Mo, Y.; Zhou, J.; Ma, X.; Zhu, W.; Zhang, L.; Li, J.; Lu, J.; Hu, C.; Bao, Y.; Jia, W. Haemoglobin A1c variability
as an independent correlate of atherosclerosis and cardiovascular disease in Chines e type 2 diabetes. Diab
Vasc Dis Res 2018, 15, 402 -408.
42. Selvin, E.; Steffes, M.W.; Zhu, H.; Matsushita, K.; Wagenknecht, L.; Pankow, J.; Coresh, J.; Brancati, F.L.
Glycated hemoglobin, diabetes, and cardiovascular risk in nondiabetic adults. N Engl J M ed 2010, 362, 800 –
811.
43. Nathan, D.M.; McGee, P.; Steffes, M.W.; Lachin, J.M.; Group, D.E.R. Relationship of glycated albumin to
blood glucose and HbA1c values and to retinopathy, nephropathy, and cardiovascular outcomes in the
DCCT/EDIC study. Diabetes 2014, 63, 282 -290.

J. Clin. Med. 2020, 9, x FOR PEER REVIEW 16 of 16
44. Forrest, K.Y.; Becker, D.J.; Kuller, L.H.; Wolfson, S.K.; Orchard, T.J. Are predictors of coronary heart disease
and lower -extremity arterial disease in type 1 diabetes the same? A prospective study. Atherosclerosis 2000,
148, 159 -169.
45. Pruzin, J.J.; Schneider, J.A.; Capuano, A.W.; Leurgans, S.E.; Barnes, L.L.; Ahima, R.S.; Arnold, S.E.; Bennett,
D.A.; Arvanitakis, Z. Diabetes, Hemoglobin A1C, and Regional Alzheimer Disease and Infarct Pathology.
Alzheimer Dis Assoc Disord 2017, 31, 41-47.
46. Orchard, T.J.; Olson, J.C.; Erbey, J.R.; Williams, K.; Forrest, K.Y.; Smithline Kinder, L.; Ellis, D.; Becker, D.J.
Insulin resistance -related factors, but not glycemia, predict coronary artery disease in type 1 diabetes: 10 –
year follow -up dat a from the Pittsburgh Epidemiology of Diabetes Complications Study. Diabetes Care 2003,
26, 1374 -1379.
47. Koivisto, V.A.; Stevens, L.K.; Mattock, M.; Ebeling, P.; Muggeo, M.; Stephenson, J.; Idzior -Walus, B.
Cardiovascular disease and its risk factors in IDDM in Europe. EURODIAB IDDM Complications Study
Group. Diabetes Care 1996, 19, 689 -697.
48. Rao Kondapally Seshasai, S.; Kaptoge, S.; Thompson, A.; Di Angelantonio, E.; Gao, P.; Sarwar, N.;
Whincup, P.H.; Mukamal, K.J.; Gillum, R.F.; Holme, I.; Njolstad, I.; Fletcher, A.; Nilsson, P.; Lewington, S.;
Collins, R.; Gudnason, V.; Thompson, S.G.; Sattar, N.; Selvin, E.; Hu, F.B.; Danesh, J.; Emerging Risk Factors,
C. Diabetes mellitus, fasting glucose, and risk of cause -specific death. N Engl J Med 2011, 364, 829-841.
49. Emerging Risk Factors, C.; Sarwar, N.; Gao, P.; Seshasai, S.R.; Gobin, R.; Kaptoge, S.; Di Angelantonio, E.;
Ingelsson, E.; Lawlor, D.A.; Selvin, E.; Stampfer, M.; Stehouwer, C.D.; Lewington, S.; Pennells, L.;
Thompson, A.; Sattar, N.; White, I.R.; Ray, K.K.; Danesh, J. Diabetes mellitus, fasting blood glucose
concentration, and risk of vascular disease: a collaborative meta -analysis of 102 prospective studies. Lancet
2010, 375, 2215 -2222.
50. Park, C.; Guallar, E.; Linton, J.A.; Lee, D.C.; Jan g, Y.; Son, D.K.; Han, E.J.; Baek, S.J.; Yun, Y.D.; Jee, S.H.;
Samet, J.M. Fasting glucose level and the risk of incident atherosclerotic cardiovascular diseases. Diabetes
Care 2013, 36, 1988 -1993.
51. Selvin, E.; Wang, D.; Matsushita, K.; Grams, M.E.; Cor esh, J. Prognostic Implications of Single -Sample
Confirmatory Testing for Undiagnosed Diabetes: A Prospective Cohort Study. Ann Intern Med 2018, 169,
156-164.
52. American Diabetes Association. Diagnosis and classification of diabetes mellitus. Diabetes Ca re 2014, 37
Suppl 1 , S81 -90.