Performance of an A1C and fasting capillary blood glucose test for screening newly diagnosed diabetes and pre-diabetes defined by an oral glucose tolerance test in Qingdao, China.

Abstract

OBJECTIVE The study's goal was to evaluate the performance of A1C and fasting capillary blood glucose (FCG) tests as mass screening tools for diabetes and pre-diabetes, as determined by the standard oral glucose tolerance test (OGTT). RESEARCH DESIGN AND METHODS Data from 2,332 individuals aged 35-74 years who participated in a population-based cross-sectional diabetes survey in Qingdao, China, were analyzed. A 2-h 75-g OGTT was used to diagnose diabetes. The performance of A1C and FCG was evaluated against the results of the OGTTs by using receiver operating characteristic curve (ROC) analysis. RESULTS The prevalence of newly diagnosed diabetes and pre-diabetes (impaired fasting glucose and/or impaired glucose tolerance) was 11.9 and 29.5%, respectively. For subjects with newly diagnosed diabetes, the area under the ROC curve was 0.67 for A1C and 0.77 for FCG (P < 0.01) in men and 0.67 and 0.75 (P < 0.01) in women, whereas for pre-diabetes, these values were 0.47 and 0.64 (P < 0.001) in men and 0.51 and 0.65 (P < 0.001) in women. At the optimal A1C cutoff point of > or =5.6% for newly diagnosed diabetes, sensitivities (specificities) were 64.4% (61.6%) for men and 62.3% (63.3%) for women. CONCLUSIONS As a screening tool for newly diagnosed diabetes and pre-diabetes, the FCG measurement performed better than A1C in this general Chinese population.

Full text

Page 1

Performance of an A1C and Fasting
Capillary Blood Glucose Test for Screening
Newly Diagnosed Diabetes and
Pre-Diabetes Defined by an Oral Glucose
Tolerance Test in Qingdao, China
XIANGHAI ZHOU, MD, PHD1,2,3
ZENGCHANG PANG, MS4
WEIGUO GAO, MD1,5
SHAOJIE WANG, MD4
LEI ZHANG, MD1,5
FENG NING, MS1
QING QIAO, MD, PHD1,5
OBJECTIVE — Thestudy’sgoalwastoevaluatetheperformanceofA1Candfastingcapillary
blood glucose (FCG) tests as mass screening tools for diabetes and pre-diabetes, as determined
by the standard oral glucose tolerance test (OGTT).
RESEARCH DESIGN AND METHODS — Data from 2,332 individuals aged 35–74
yearswhoparticipatedinapopulation-basedcross-sectionaldiabetessurveyinQingdao,China,
were analyzed. A 2-h 75-g OGTT was used to diagnose diabetes. The performance of A1C and
FCG was evaluated against the results of the OGTTs by using receiver operating characteristic
curve (ROC) analysis.
RESULTS — The prevalence of newly diagnosed diabetes and pre-diabetes (impaired fasting
glucose and/or impaired glucose tolerance) was 11.9 and 29.5%, respectively. For subjects with
newly diagnosed diabetes, the area under the ROC curve was 0.67 for A1C and 0.77 for FCG
(P ? 0.01) in men and 0.67 and 0.75 (P ? 0.01) in women, whereas for pre-diabetes, these
values were 0.47 and 0.64 (P ? 0.001) in men and 0.51 and 0.65 (P ? 0.001) in women. At the
optimal A1C cutoff point of ?5.6% for newly diagnosed diabetes, sensitivities (specificities)
were 64.4% (61.6%) for men and 62.3% (63.3%) for women.
CONCLUSIONS — As a screening tool for newly diagnosed diabetes and pre-diabetes, the
FCG measurement performed better than A1C in this general Chinese population.
Diabetes Care 33:545–550, 2010
T
tality and morbidity including blindness,
renal failure, amputation, and cardiovas-
cular disease. Diabetes and its complica-
tions may occur several years before a
clinicaldiagnosisismade.Massscreening
for diabetes can lead to an early diagnosis
and timely treatment or intervention,
whichhavebeenshowntoreducediabetes-
ype 2 diabetes has become a serious
public health threat worldwide, and
itleadstoincreasedprematuremor-
associated complications and to prevent
or delay the onset of diabetes per se. Cur-
rently, diabetes and pre-diabetes are clas-
sified according to the 2-h 75-g oral
glucose tolerance test (OGTT) (1), which
has been considered as a standard diag-
nostic criterion for diabetes for several
decades. However, there are several
limitations associated with the OGTT.
These limitations include the uncertainty
ofthefastingstage,thepoorreproducibil-
ity of the 2-h glucose tests (2), and the
poor concordance between the fasting
plasma glucose (FPG) and the 2-h plasma
glucose levels (3).
The capillary blood glucose test is a
point-of-care determination and involves
only one finger prick. Because it is easy to
use and cheap, it is applied as a first-step
screening test for mass screening of sub-
jects in either the fasting (4,5) or the ran-
dom (4) state. In contrast, A1C is the
mean of the long-term glucose level and
does not require the subject to be in a
fasting state with only one blood sample
drawn.Itisusedtomonitorglucoselevels
in diabetic individuals who have received
antihyperglycemictreatment.Becausethe
A1C test is less standard and relatively
expensive, it is not widely available. Con-
sequently, its use in mass screening has
been limited. Data on the performance of
the A1C test in mass screening for diabe-
tes are still lacking. In the light of the re-
cent development in standardizing the
method of A1C measurement (6), the
A1C test has been adopted as a diagnostic
criterion for diabetes (7). To evaluate the
performance of the A1C as a mass screen-
ing tool for diabetes as determined by
OGTTs in a general population and to
compare its performance with that of the
FCG test, data from a population-based
study in Qingdao, China, were analyzed.
RESEARCH DESIGN AND
METHODS— A population-based
cross-sectional diabetes survey was con-
ducted in Qingdao, China, in 2006. A
stratified, random cluster sampling
method was used to recruit a representa-
tive sample of those in the general popu-
lationwhohadlivedinQingdaocityforat
least5years.Thesurveywasconductedin
threeurbanandthreeruraladministrative
areas. Five resident communities from
each area with 200–250 individuals from
eachcommunitywererandomlyselected.
A total of 6,100 individuals aged 35–74
years were invited to participate in the
● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
From the1Department of Public Health, University of Helsinki, Helsinki, Finland; the2Department of
Endocrinology and Metabolism, Peking University People’s Hospital, Beijing, China; the3Peking Univer-
sity Diabetes Center, Beijing, China; the4Qingdao Centers for Disease Control and Prevention, Qingdao,
China; and the5Diabetes Prevention Unit, National Institute for Health and Welfare, Helsinki, Finland.
Corresponding author: Zengchang Pang, cdcpang@126.com.
Received 2 August 2009 and accepted 24 November 2009. Published ahead of print at http://care.
diabetesjournals.org on 10 December 2009. DOI: 10.2337/dc09-1410.
© 2010 by the American Diabetes Association. Readers may use this article as long as the work is properly
cited, the use is educational and not for profit, and the work is not altered. See http://creativecommons.
org/licenses/by-nc-nd/3.0/ for details.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby
marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
E p i d e m i o l o g y / H e a l t h S e r v i c e s R e s e a r c h
O R I G I N A L A R T I C L E
care.diabetesjournals.orgDIABETES CARE, VOLUME 33, NUMBER 3, MARCH 2010
545

Page 2

survey and 5,355 attended, which gave a
response rate of 87.8%. Among 5,355
participants, 4,808 individuals who did
nothaveapriorhistoryofdiabetesunder-
went either FPG and/or 2-h plasma glu-
cose tests, and 4,203 (87.4%) underwent
A1C and 4,371 (90.9%) FCG tests. For
data analysis strict inclusion criteria were
applied: 1) undiagnosed diabetes identi-
fied based on both FPG and 2-h plasma
glucose criteria; 2) data for both A1C and
FCG; and 3) no data missing for BMI,
waist circumference, and blood pressure.
A total of 2,332 (986 men and 1,346
women) individuals met the strict inclu-
sion criteria. By using the strict inclusion
criteria, we were able to correctly classify
diabetes status and to compare A1C with
FCG, but at the cost of reducing the sam-
ple size by 51.5% (2,476 of 4,808). To
check whether this strategy biased the re-
sults, a sensitivity analysis was performed
including all individuals with either A1C
(n ? 4,203) or FCG (n ? 4,371) test val-
ues regardless of missing data for either
FPG or 2-h plasma glucose. Undiagnosed
diabetes was, thus, classified according to
either FPG or 2-h plasma glucose criteria
in the sensitivity analysis. The survey was
approved by the Qingdao Municipal
Health Bureau and the local ethics com-
mittee, and informed consent was ob-
tained from all participants.
Participants were interviewed by
traineddoctorsornursesinthelocalcom-
munity clinics. Height and weight were
measured with the participants wearing
only light clothes and without shoes.
Waist circumference was measured at the
middlepointbetweentheribcageandtop
of the iliac crest to the nearest 0.1 cm.
Three consecutive blood pressure read-
ings were recorded at least 30 s apart.
Theyweretakenfromtheupperrightarm
of seated individuals, and the mean of the
three readings was used in the data anal-
ysis. BMI was calculated as weight in ki-
logramsdividedbythesquareofheightin
meters.
Afteranovernightfastofatleast10h,
surveyparticipantsintheirrespectivesur-
vey sites were given the FCG test over the
0700–0930 period, using a Bayer Ascen-
sia BRIO blood glucose monitoring sys-
tem that was calibrated to give capillary
plasma/serum glucose equivalent results
(Bayer HealthCare Company, Shanghai,
China). A standard OGTT was also per-
formed on the same day over the 0700–
1130 period, and blood samples for
glucose determinations were collected
from the antecubital vein into a vacuum
tube containing sodium fluoride. All
blood samples were analyzed in the cen-
tral laboratory of Hiser Medical Center.
An Olympus AU system (Olympus, To-
kyo, Japan) was used for plasma glucose,
A1C, serum total cholesterol, triglycer-
ides, and HDL cholesterol levels. Plasma
glucose levels were determined using the
glucose oxidase method. A1C was mea-
sured using an immunoturbidimetry
method (Tina-qu.a A1C HIT 917 large;
Roche Diagnostics). The A1C concentra-
tion was calculated by using the formula
provided by Roche Diagnostics: [calcu-
lated A1C (%) ? 0.81 ? A1C (test re-
sult) ? 2.39] to match the values with
those found in a conventional high-
performance liquid chromatography
method. The calculated A1C was subse-
quently used in the data analysis. The ref-
erence range for the calculated A1C was
4.3–5.8%. Fasting insulin concentration
was measured in 1,636 individuals using
the chemiluminescence immunoassay
method (Abbott AxSym). The homeosta-
sis model assessment of insulin resistance
(HOMA-IR) was calculated using the for-
mula [HOMA-IR ? fasting insulin (mi-
cro-international units per milliliter) ?
FPG (millimoles per liter)/22.5], and the
homeostasis model assessment of ?-cell
function (HOMA-B) was calculated using
the formula {HOMA-B ? 20 ? fasting
insulin [micro-international units per
milliliter]/[FPG (millimoles per liter) ?
3.5]} (8).
Previously diagnosed diabetes was
defined as self-report of diabetes diag-
nosed by a doctor and/or on information
on treatment, and subjects with diag-
nosed diabetes were excluded from the
analysis. According to the World Health
Organization definition (1), newly diag-
nosed diabetes is defined as having either
FPG ?7.0 mmol/l and/or 2-h plasma glu-
cose ?11.1 mmol/l. IGT is defined as
having FPG ?7.0 mmol/l and 2-h plasma
glucose?7.8mmol/land2-hplasmaglu-
cose ?11.1 mmol/l. Impaired fasting glu-
cose (IFG) is defined as having FPG ?6.1
mmol/l and FPG ?7.0 mmol/l and 2-h
plasma glucose ?7.8 mmol/l. Normal
glucose tolerance (NGT) is defined as
FPG ?6.1 mmol/l and 2-h plasma glu-
cose ?7.8 mmol/l.
Statistical analysis
Statistical analysis was performed using
SPSS for Windows 15.0 (SPSS, Chicago,
IL) and STATA 9.1 (StataCorp, College
Station, TX). Continuous variables are
presented as means (95% CI), and cate-
goricaldataaregivenasnumber(percent-
age). Correlations between variables were
assessed by Pearson correlation coeffi-
cient. The differences between the means
of the groups were tested using a univar-
iate general linear model with adjust-
ments for age, and categorical data were
analyzed by using the ?2test. The sen-
sitivity and specificity of the A1C and FCG
cutoff values for detecting newly diag-
noseddiabetes,pre-diabetes,ornewlydi-
agnosed diabetes plus pre-diabetes based
on the OGTT response were calculated.
The receiver operating characteristic
(ROC) curve was obtained by plotting
sensitivity against 1 ? specificity for each
cutoff value. The optimal cutoff point
was identified as the coordinate closest
to the y intercept (0,1) of the ROC
curve, and at this point, the sum of the
sensitivity and the specificity is maxi-
mal. Diagnostic accuracy was assessed
by the area under the curve (AUC) (9).
RESULTS— The respective preva-
lence of pre-diabetes and newly diag-
nosed diabetes among the participants
was 29.5 and 11.9%. Individuals of both
sexes with pre-diabetes and newly diag-
nosed diabetes were older, more obese,
hypertensive, dyslipidemic, and insulin
resistant and had poorer ?-cell function
(estimated by HOMA-B) compared with
individuals with NGT (Table 1). The cor-
relation coefficients of the A1C with FPG
(0.57) and 2-h plasma glucose (0.61)
were higher in diabetic individuals than
in nondiabetic individuals (?0.32 and
0.001, respectively).
As expected, diabetic individuals of
both sexes with A1C ?6.5% had higher
meanFPGor2-hplasmaglucoseandalso
poorer ?-cell function than diabetic indi-
viduals with A1C ?6.5%. However, this
was not the case in the nondiabetic indi-
viduals (Table 2). Among nondiabetic in-
dividuals, participants of both sexes with
A1C ?6.5% had a lower prevalence of
IGT and/or IFG and lower mean FPG and
triglycerides than those with A1C ?6.5%
(Table 2).
The AUC was lower for A1C than for
FCG for detecting newly diagnosed dia-
betes (0.67 vs. 0.77, P ? 0.01, in men;
and 0.67 vs. 0.75, P ? 0.01, in women)
and pre-diabetes (0.47 vs. 0.64, P ?
0.001, in men; and 0.51 vs. 0.65, P ?
0.001, in women) (Fig. 1). Moreover, for
newly diagnosed diabetes plus pre-
diabetes, the AUCs for A1C and FCG val-
ueswere0.53vs.0.69inmen(P?0.001)
and 0.55 vs. 0.68 in women (P ? 0.001).
Performance of A1C for detecting diabetes
546
DIABETES CARE, VOLUME 33, NUMBER 3, MARCH 2010care.diabetesjournals.org

Page 3

The A1C values did not distinguish peo-
ple with pre-diabetes from those with
NGT. The chance-corrected proportion
agreement (?) for classification of diabe-
tes between the criteria of the OGTT and
theA1Cof?6.5%thatarerecommended
by the International Expert Committee
(7) was 0.20 in men and 0.14 in women.
The optimal A1C cutoff point for newly
diagnosed diabetes in this study popula-
tion at 5.6% was lower than the recom-
mended value of 6.5%. At the optimal
A1C cutoff point of 5.6% for newly diag-
nosed diabetes, the sensitivity was 64.4%
in men and 62.3% in women, whereas it
was 72.0 and 65.1% at the optimal FCG
cutoffpointof6.3mmol/lformenand6.6
mmol/l for women (Table 3). Using the
recommended diagnostic value of 6.5%
for A1C (7), the sensitivity for newly di-
agnosed diabetes was ?30% in both men
and women. The optimal FCG cutoff
value and the corresponding sensitivity
for pre-diabetes and pre-diabetes plus
newly diagnosed diabetes are shown in
Table 3.
To check whether the results were bi-
ased by excluding half of the survey par-
ticipants from the data analyses, a
sensitivity analysis was made using the
maximum number of participants for ei-
ther A1C or FCG. For undiagnosed dia-
betes, an AUC of 0.67 (95% CI 0.63–
0.72) in men and 0.70 (0.67–0.73) in
women for A1C and 0.78 (0.74–0.81) in
men and 0.76 (0.73–0.79) in women for
FCG was obtained, which did not differ
from the results presented in Fig. 1. The
groupwhometthestrictinclusioncriteria
Table 1—Characteristics of male and female subjects stratified by glucose tolerance status defined according to both FPG and 2-h plasma
glucose level
All
Normal: FPG ?6.1
mmol/l and 2-h PG
?7.8 mmol/l
Pre-diabetes: FPG 6.1–6.9
mmol/l and/or 2-h PG
7.8–11.0 mmol/l
Undiagnosed diabetes: FPG ?7.0
mmol/l and/or 2-h PG ?11.1
mmol/l
Male subjects
n (%)
Age (years)
Waist circumference (cm)
BMI (kg/m2)
Blood pressure (mmHg)
Systolic
Diastolic
Plasma glucose (mmol/l)
Fasting
2-h
A1C (%)
A1C ?6.5%
FCG (mmol/l)
Cholesterol (mmol/l)
Triglycerides (mmol/l)
HDL cholesterol (mmol/l)
HOMA-IR§?
HOMA-B§?
Female subjects
n (%)
Age (years)
Waist circumference (cm)
BMI (kg/m2)
Blood pressure (mmHg)
Systolic
Diastolic
Plasma glucose (mmol/l)
Fasting
2-h
A1C (%)
A1C ?6.5%
FCG (mmol/l)
Cholesterol (mmol/l)
Triglycerides (mmol/l)§
HDL cholesterol (mmol/l)
HOMA-IR§?
HOMA-B§?
Dataareage-adjustedmeans(95%CI)orn(%).*P?0.001;†P?0.01;‡P?0.05,forthedifferencebetweentheindexedandthenormalcategoryusingaunivariate
general linear model for continuous data and ?2test for categorical data. §Geometric mean (95% CI). ?With missing data.
986 588 (59.6)
48.1 (47.3?48.9)
85.6 (84.9?86.4)
25.2 (24.9?25.4)
266 (27.0)
51.2 (50.0?52.4)*
89.6 (88.5?90.7)*
26.4 (26.1?26.8)*
132 (13.4)
52.6 (50.9?54.3)*
89.1 (87.5?90.7)*
26.3 (25.8?26.8)*
49.5 (48.9?50.2)
87.2 (86.6?87.8)
25.7 (25.5?25.8)
134 (133?135)
87 (87?88)
132 (130?133)
86 (86?87)
136 (134?138)†
88 (87?90)
138 (135?141)†
90 (87?92)‡
5.7 (5.6?5.8)
7.1 (6.9?7.3)
5.7 (5.7?5.8)
116 (11.8)
6.1 (6.0?6.1)
5.3 (5.2?5.3)
1.2 (1.2?1.2)
1.64 (1.61?1.67)
1.1 (1.0?1.1)
55.1 (51.7?58.7)
5.1 (5.0?5.2)
5.7 (5.5?5.9)
5.6 (5.6?5.7)
67 (11.4)
5.7 (5.6?5.8)
5.2 (5.1?5.2)
1.1 (1.0?1.1)
1.68 (1.64?1.72)
0.8 (0.8?0.9)
60.9 (56.2?65.9)
5.9 (5.8?6.1)*
7.5 (7.2?7.8)*
5.5 (5.4?5.6)
12 (4.5)†
6.2 (6.0?6.3)*
5.3 (5.2?5.5)
1.4 (1.3?1.5)*
1.55 (1.49?1.60)*
1.5 (1.3?1.7)*
53.9 (47.6?61.0)
7.9 (7.7?8.1)*
12.4 (11.9?12.8)*
6.4 (6.3?6.5)*
37 (28.0)*
7.4 (7.2?7.6)*
5.5 (5.4?5.7)*
1.5 (1.4?1.6)*
1.61 (1.53?1.69)
1.7 (1.4?2.0)*
34.6 (28.9?41.4)*
1,346 777 (57.7)
47.2 (46.6?47.9)
80.8 (80.2?81.4)
25.3 (25.1?25.6)
423 (31.4)
51.4 (50.5?52.3)*
83.1 (82.2?83.9)*
26.4 (26.0?26.7)*
146 (10.8)
54.4 (52.9?55.9)*
84.6 (83.2?86.1)*
26.8 (26.2?27.4)*
49.3 (48.8?49.8)
81.9 (81.5?82.4)
25.8 (25.6?26.0)
132 (131?133)
84 (84?85)
128 (127?130)
83 (82?83)
137 (135?139)*
86 (85?88)*
138 (135?141)*
87 (85?89)*
5.6 (5.6?5.7)
7.3 (7.2?7.5)
5.6 (5.6?5.6)
137 (10.2)
6.1 (6.0?6.1)
5.2 (5.2?5.3)
1.1 (1.0?1.1)
1.68 (1.65?1.70)
1.4 (1.3?1.5)
67.9 (64.7?71.3)
5.2 (5.1?5.2)
6.0 (5.9?6.1)
5.6 (5.5?5.6)
82 (10.6)
5.7 (5.6?5.8)
5.2 (5.1?5.2)
1.0 (1.0?1.0)
1.73 (1.70?1.77)
1.2 (1.1?1.3)
74.4 (69.9?79.2)
5.8 (5.8?5.9)*
7.9 (7.7?8.1)*
5.5 (5.4?5.6)
23 (5.4)†
6.2 (6.1?6.3)*
5.3 (5.2?5.3)
1.1 (1.0?1.1)‡
1.61 (1.57?1.65)*
1.7 (1.6?1.8)*
65.0 (59.5?70.9)‡
7.4 (7.3?7.6)*
12.8 (12.5?13.2)*
6.1 (6.0?6.2)*
32 (21.9)*
7.4 (7.2?7.6)*
5.5 (5.4?5.7)*
1.4 (1.3?1.5)*
1.57 (1.50?1.65)*
2.3 (2.0?2.6)*
44.7 (38.4?52.1)*
Zhou and Associates
care.diabetesjournals.orgDIABETES CARE, VOLUME 33, NUMBER 3, MARCH 2010
547

Page 4

was not different from those who did not
with regard to mean age, BMI, and blood
pressure, but the former consisted of
more men (42.3 vs. 35.6%, P ? 0.001)
and had a slightly higher prevalence of
undiagnosed diabetes (11.9 vs. 10.1%,
P ? 0.04) than the latter as expected.
CONCLUSIONS — In this popula-
tion-based study, the FCG test performed
better than A1C in ability to identify
newly diagnosed diabetes as determined
bythestandard2-h75-gOGTT.TheA1C
test did not distinguish between pre-
diabetic and normal subjects. Less than
30% of the newly diagnosed diabetes
couldbeidentifiedattheA1Ccutoffpoint
of ?6.5%, the diagnostic benchmark rec-
ommended by the International Expert
Committee (7).
In previous studies, the optimal A1C
cutoff points for detecting undiagnosed
diabetes ranged from 5.3 to 6.1% in pop-
ulation-based studies among Australian
and American populations (10–12). In
those studies the sensitivity varied from
63.2 to 86.0%, and the specificity ranged
from 82.8 to 97.4% at a given optimal
A1C cutoff point. In a study among Hong
KongChinesewhowerereferredtoahos-
pital for confirmative testing for glucose
intolerance, an optimal A1C cutoff point
of 6.1% for detecting 2-h plasma glucose
?11.1mmol/lwasreported,whichgavea
sensitivity of 77.5% and a specificity of
78.8% (13). This Hong Kong study was,
however, performed among a preselected
high-risk population of which close to
one-fourth of its participants were diag-
nosed as having diabetes. Moreover, the
diagnostic criteria for diabetes in the
HongKongstudywerealsodifferentfrom
those used in our study. The inconsis-
tency in performance of the A1C test
among studies may have been due to the
discrepancies in ethnicity, assay methods
for A1C, the gold standard for diagnosis
of diabetes, study methodology differ-
Table 2—Characteristics of male and female subjects according to A1C level and diabetes status defined by the OGTT
Undiagnosed diabetesNondiabetes
A1C ?6.5% A1C ?6.5% A1C ?6.5% A1C ?6.5%
Male subjects (n ? 986)
n (%)
Age (years)
Waist circumference (cm)
BMI (kg/m2)
Blood pressure (mmHg)
Systolic
Diastolic
Plasma glucose (mmol/l)
Fasting
2-h
IGT and/or IFG
FCG (mmol/l)
Cholesterol (mmol/l)
Triglycerides (mmol/l)§
HDL cholesterol (mmol/l)
HOMA-IR§?
HOMA-B§?
Female subjects (n ? 1,346)
n (%)
Age (years)
Waist circumference (cm)
BMI (kg/m2)
Blood pressure (mmHg)
Systolic
Diastolic
Plasma glucose (mmol/l)
Fasting
2-h
IGT and/or IFG
FCG (mmol/l)
Cholesterol (mmol/l)
Triglycerides (mmol/l)§
HDL cholesterol (mmol/l)
HOMA-IR§?
HOMA-B§?
Data are age-adjusted means (95% CI) or n (%). *P ? 0.001; †P ? 0.01; ‡P ? 0.05, for the difference between the indexed and the A1C ?6.5% category within the
same glucose tolerance class of undiagnosed diabetes or nondiabetes using a univariate general linear model for continuous data and ?2test for categorical data.
§Geometric mean (95% CI). ?With missing data.
37 (3.8)
53.5 (50.4?56.6)
89.3 (86.4?92.3)
26.3 (25.2?27.3)
95 (9.6)
52.3 (50.3?54.2)
88.5 (86.7?90.4)
26.2 (25.5?26.8)
79 (8.0)
47.2 (44.9?49.4)
86.8 (84.7?88.9)
25.7 (25.0?26.4)
775 (78.6)
49.2 (48.5?50.0)
86.9 (86.3?87.6)
25.6 (25.3?25.8)
138 (132?145)
91 (87?94)
140 (136?144)
89 (87?92)
135 (131?139)
88 (85?91)
133 (131?134)
87 (86?88)
9.8 (8.8?10.9)
17.3 (15.6?19.1)
7.2 (6.5?7.8)*
10.5 (9.5?11.6)*
4.8 (4.7?5.0)
6.1 (5.8?6.5)
12 (15.2)
6.0 (5.7?6.2)
5.4 (5.2?5.6)
0.9 (0.8?1.1)
1.75 (1.64?1.85)
0.8 (0.7?1.0)
70.0 (58.3?84.0)
5.4 (5.3?5.4)*
6.2 (6.1?6.4)
254 (32.8)†
5.8 (5.8?5.9)
5.2 (5.1?5.3)
1.2 (1.2?1.2)*
1.63 (1.59?1.66)‡
1.0 (0.9?1.1)
57.6 (53.8?61.7)
9.0 (8.3?9.7)
5.7 (5.4?6.1)
1.5 (1.2?1.8)
1.68 (1.53?1.84)
1.3 (0.8?1.9)
19.7 (13.3?28.9)
6.8 (6.4?7.2)*
5.5 (5.3?5.7)
1.4 (1.3?1.7)
1.59 (1.49?1.68)
1.8 (1.4?2.4)
41.7 (32.5?53.4)†
32 (2.4)
55.9 (52.6?59.2)
88.1 (85.2?91.0)
27.4 (26.3?28.5)
114 (8.5)
53.9 (52.2?55.7)
85.3 (83.8?86.8)
27.0 (26.4?27.6)
105 (7.8)
46.7 (44.9?48.5)
84.1 (82.4?85.8)
25.8 (25.2?26.5)
1,095 (81.4)
48.9 (48.3?49.4)‡
81.2 (80.7?81.7)†
25.7 (25.4?25.9)
138 (130?147)
86 (82?90)
143 (139?148)
88 (86?91)
133 (129?137)
85 (82?87)
131 (129?132)
84 (83?84)
9.2 (8.6?9.9)
17.8 (16.3?19.2)
7.0 (6.6?7.3)*
11.6 (10.8?12.3)*
4.8 (4.6?4.9)
6.7 (6.4?7.0)
23 (21.9)
6.0 (5.8?6.2)
5.0 (4.8?5.2)
0.9 (0.9?1.0)
1.77 (1.69?1.85)
1.0 (0.9?1.2)
98.3 (84.6?114.2)
5.5 (5.4?5.5)*
6.7 (6.6?6.7)
400 (36.5)†
5.9 (5.8?5.9)
5.2 (5.1?5.3)‡
1.0 (1.0?1.1)‡
1.68 (1.66?1.71)
1.4 (1.3?1.5)*
68.7 (65.2?72.4)*
9.8 (9.2?10.5)
6.0 (5.6?6.4)
1.5 (1.2?1.8)
1.58 (1.41?1.75)
2.3 (1.7?3.0)
25.6 (19.0?34.5)
6.8 (6.5?7.2)*
5.6 (5.4?5.8)
1.4 (1.3?1.6)
1.59 (1.49?1.68)
2.4 (2.0?2.8)
49.5 (42.2?58.1)*
Performance of A1C for detecting diabetes
548
DIABETES CARE, VOLUME 33, NUMBER 3, MARCH 2010 care.diabetesjournals.org

Page 5

ences (population-based versus clinical
based), and the prevalence of other risk
factors such as obesity. The extent to
which these factors affect the perfor-
mance of the A1C needs to be further in-
vestigated. However, our results are
consistent with other studies showing
that the A1C test did not detect IGT
(11,14,15).
Few studies have evaluated the per-
formance of the FCG test. In a randomly
selectedpopulation-basedstudyinBrazil,
the optimal FCG cutoff point for undiag-
nosed diabetes was 5.6 mmol/l, which
gave a sensitivity of 87.2% and a specific-
ity of 72.4% (16). Another study com-
pared FCG with A1C assays among
Caucasians who were referred to a hospi-
tal because they were suspected as having
diabetes (17). That study also showed
that the performance of the FCG test was
better than that of the A1C test in detect-
ing either diabetes or diabetes plus pre-
diabetes, which was similar to the
findings in our present study.
Prospective studies showed that ele-
vatedA1Cvaluesincreasetheriskofmor-
tality from all causes (18,19) and
cardiovasculardisease(19).Itwasalsore-
ported that A1C predicted mortality
(20,21)orincidentcardiovasculardisease
(22) depending largely on either the 2-h
plasma glucose (20–22) or the FPG (21)
levels. In our study, the blood pressure
and lipid levels in individuals with A1C
?6.5% did not differ from those with
A1C ?6.5%. However, both of these pa-
rameterswerehigherfordiabeticandpre-
diabetic individuals than they were for
normal subjects as defined by the OGTT.
The prognosis of those individuals with
A1C ?6.5% needs to be further investi-
gated in prospective studies.
The large number of individuals in-
cluded in this population-based study
Figure 1—ROC curve for A1C and FCG to assess the presence of newly diagnosed diabetes (NDM) and pre-diabetes in men (A) and women (B).
Table 3—Sensitivity and specificity corresponding to the optimal cutoff values of A1C and FCG for detecting newly diagnosed diabetes,
pre-diabetes, and newly diagnosed diabetes plus pre-diabetes in male and female subjects
Optimal cutoff point
NDMPre-diabetes NDM ? pre-diabetes
SensitivitySpecificity SensitivitySpecificity SensitivitySpecificity
Male subjects
A1C (%)
?5.6 for NDM
?6.5 (diagnostic criteria)*
FCG (mmol/l)
?6.3 for NDM
?6.0 for pre-diabetes
?6.1 for NDM ? pre-diabetes
Female subjects
A1C (%)
?5.6 for NDM
?6.5 (diagnostic criteria)*
FCG (mmol/l)
?6.6 for NDM
?6.1 for pre-diabetes and NDM ?
pre-diabetes
Data are %. *Adopted as the diagnostic criteria for diabetes (7). NDM, newly diagnosed diabetes.
64.4
28.0
61.6
90.5
33.5
4.5
59.4
88.3
43.7
12.3
59.4
88.3
72.0
78.8
75.0
68.4
55.5
61.0
46.6
60.5
55.6
75.2
62.8
68.5
55.0
66.6
62.1
75.2
62.8
68.5
62.3
21.9
63.3
91.2
36.2
5.7
62.9
89.4
42.9
9.8
62.9
89.4
65.177.034.8 83.4 42.583.4
73.359.2 56.767.861.067.8
Zhou and Associates
care.diabetesjournals.orgDIABETES CARE, VOLUME 33, NUMBER 3, MARCH 2010
549

Page 6

provided our study with high statistical
powerfordataanalyses.Diabetesandpre-
diabetes were defined according to the
currentlyacceptedstandardcriteriabased
on both FPG and 2-h plasma glucose lev-
els. All individuals included in the analy-
sis provided data on FCG and A1C in
addition to the OGTT. These individuals
also yielded data on lipid and insulin,
which enabled us to study further the
clinical phenotypes of individuals of dif-
ferent glycemic categories. Because of the
strict inclusion criteria for glucose and
other variables, only 55% of the popula-
tion who had A1C measurements were
included in the data analyses. To check
whether the results have been biased by
the strict inclusion criteria, a sensitivity
analysis using all data available for the
A1C was made. The results of the sensi-
tivity analysis showed that the study was
less likely to be biased by the strict inclu-
sion criteria applied in the study. The
high prevalence of diabetes observed in
this study reflects the recent increase in
the prevalence of diabetes in Qingdao,
which has been investigated in detail and
reported (23,24). The question as to
whether the result can be found in other
populationsremainstobeconfirmed.An-
otherlimitationofourstudyisthathemo-
globin concentrations were not available,
and, thus, a potential influence of condi-
tions such as hemolytic anemia and he-
moglobinopathies on A1C assays cannot
be excluded. Nevertheless, these condi-
tions are rare in the general population
from which our study population was
drawn. The prevalence of abnormal he-
moglobin in the Shandong province in
which Qingdao is located is only 0.04%
(25).
Insummary,asamassscreeningtool,
the FCG test performed better than the
A1CtestinthegeneralpopulationofChi-
nese. In consideration of its high cost and
poor performance, the A1C test is not a
suitable test for mass screening, particu-
larly with the purpose of detecting pre-
diabetes for early intervention.
Acknowledgments— This study was sup-
ported by the Finnish Academy (118492).
BayerHealthCareCompany,Shanghai,China,
providedunrestrictedfinancialsupportforthe
data collection.
No potential conflicts of interest relevant to
this article were reported.
We express our sincere thanks to the local
research teams from Qingdao Municipal
Health Bureau, Qingdao Centers for Disease
Control and Prevention, Qingdao, China, for
their contribution to the field survey (http://
www.qddiabetes.org/Organize-6.asp).
References
1. World Health Organization. Definition, Di-
agnosis and Classification of Diabetes Mellitus
and Its Complications: Report of a WHO Con-
sultation. Part 1. Diagnosis and Classification
of Diabetes Mellitus. Geneva, World Health
Org., 1999 (WHO/NCD/NCS/99.2)
2. Eschwe `ge E, Charles MA, Simon D, et al.
Reproducibility of the diagnosis of diabe-
tes over a 30-month follow-up: the Paris
Prospective Study. Diabetes Care 2001;
24:1941–1944
3. Drzewoski J, Czupryniak L. Concordance
between fasting and 2-h post-glucose
challenge criteria for the diagnosis of dia-
betes mellitus and glucose intolerance in
high risk individuals. Diabet Med 2001;
18:29–31
4. Toscano CM, Duncan BB, Mengue SS,
et al. Initial impact and cost of a nation-
wide population screening campaign for
diabetes in Brazil: a follow up study. BMC
Health Serv Res 2008;8:189
5. SpijkermanAM,AdriaanseMC,DekkerJM,
et al. Diabetic patients detected by popula-
tion-based stepwise screening already have
adiabeticcardiovascularriskprofile.Diabe-
tes Care 2002;25:1784–1789
6. Consensus statement on the worldwide
standardization of the hemoglobin A1C
measurement:theAmericanDiabetesAsso-
ciation,EuropeanAssociationfortheStudy
of Diabetes, International Federation of
Clinical Chemistry and Laboratory Medi-
cine, and the International Diabetes Feder-
ation. Diabetes Care 2007;30:2399–
2400
7. InternationalExpertCommitteereporton
the role of the A1C assay in the diagnosis
ofdiabetes.DiabetesCare2009;32:1327–
1334
8. Matthews DR, Hosker JP, Rudenski AS,
et al. Homeostasis model assessment: in-
sulin resistance and ?-cell function from
fasting plasma glucose and insulin con-
centrations in man. Diabetologia 1985;
28:412–419
9. Akobeng AK. Understanding diagnostic
tests 3: receiver operating characteristic
curves. Acta Paediatr 2007;96:644–647
10. Rohlfing CL, Little RR, Wiedmeyer HM,
etal.UseofGHb(HbA1c)inscreeningfor
undiagnosed diabetes in the U.S. popula-
tion. Diabetes Care 2000;23:187–191
11. Colagiuri S, Hussain Z, Zimmet P, et al.
Screening for type 2 diabetes and im-
pairedglucosemetabolism:theAustralian
experience. Diabetes Care 2004;27:367–
371
12. Buell C, Kermah D, Davidson MB. Utility
of A1C for diabetes screening in the
1999–2004NHANESpopulation.Diabe-
tes Care 2007;30:2233–2235
13. Ko GT, Chan JC, Yeung VT, et al. Com-
bineduseofafastingplasmaglucosecon-
centration and HbA1c or fructosamine
predicts the likelihood of having diabetes
inhigh-risksubjects.DiabetesCare1998;
21:1221–1225
14. Saydah SH, Byrd-Holt D, Harris MI. Pro-
jected impact of implementing the results
of the Diabetes Prevention Program in the
U.S. population. Diabetes Care 2002;25:
1940–1945
15. Mannucci E, Ognibene A, Sposato I, et al.
Fasting plasma glucose and glycated hae-
moglobin in the screening of diabetes and
impaired glucose tolerance. Acta Diabetol
2003;40:181–186
16. Bortheiry AL, Malerbi DA, Franco LJ. The
ROC curve in the evaluation of fasting
capillary blood glucose as a screening test
fordiabetesandIGT.DiabetesCare1994;
17:1269–1272
17. Herdzik E, Safranow K, Ciechanowski K.
Diagnostic value of fasting capillary glu-
cose, fructosamine and glycosylated hae-
moglobin in detecting diabetes and other
glucose tolerance abnormalities com-
pared to oral glucose tolerance test. Acta
Diabetol 2002;39:15–22
18. Khaw KT, Wareham N, Bingham S, et al.
Association of hemoglobin A1c with car-
diovascular disease and mortality in
adults: the European prospective investi-
gation into cancer in Norfolk. Ann Intern
Med 2004;141:413–420
19. Nakanishi S, Yamada M, Hattori N, et al.
Relationship between HbA1cand mortal-
ity in a Japanese population. Diabetologia
2005;48:230–234
20. Qiao Q, Dekker JM, de Vegt F, et al. Two
prospective studies found that elevated
2-hr glucose predicted male mortality in-
dependent of fasting glucose and HbA1c.
J Clin Epidemiol 2004;57:590–596
21. Barr EL, Boyko EJ, Zimmet PZ, et al.
Continuous relationships between non-
diabetic hyperglycaemia and both cardio-
vascular disease and all-cause mortality:
the Australian Diabetes, Obesity, and
Lifestyle (AusDiab) study. Diabetologia
2009;52:415–424
22. Meigs JB, Nathan DM, D’Agostino RB Sr,
et al. Fasting and postchallenge glycemia
and cardiovascular disease risk: the Fra-
mingham Offspring Study. Diabetes Care
2002;25:1845–1850
23. Ning F, Pang ZC, Dong YH, et al. Risk
factors associated with the dramatic in-
crease in the prevalence of diabetes in the
adult Chinese population in Qingdao,
China. Diabet Med 2009;26:855–863
24. Gao WG, Dong YH, Pang ZC, et al. In-
creasing trend in the prevalence of type
2 diabetes and pre-diabetes in the Chi-
nese rural and urban population in
Qingdao, China. Diabet Med 2009;26:
1220–1229
25. Zeng YT, Huang SZ. Disorders of haemo-
globin in China. J Med Genet 1987;24:
578–583
Performance of A1C for detecting diabetes
550
DIABETES CARE, VOLUME 33, NUMBER 3, MARCH 2010care.diabetesjournals.org