Acta Diabetol (2008) 45:157–165
DOI 10.1007/s00592-008-0037-8
ORIGINAL ARTICLE
Screening for biomarkers predictive of gestational
diabetes mellitus
Harry M. Georgiou Æ Martha Lappas Æ George M. Georgiou Æ
Adinda Marita Æ Valerie J. Bryant Æ Richard Hiscock Æ
Michael Permezel Æ Zeinab Khalil Æ Gregory E. Rice
Received: 5 October 2007 / Accepted: 25 March 2008 / Published online: 22 May 2008
Ó Springer-Verlag 2008
Abstract Screening for glucose intolerance during preg-
nancy provides an opportunity to offer management to
those women diagnosed with gestational diabetes mellitus.
However, there is a need to diagnose gestational diabetes
early to minimize exposure of the developing fetus to
suboptimal conditions and prevent perinatal complications
and their sequelae. The purpose of this study was to
identify potential biomarkers for impending gestational
diabetes that appear in the plasma before impaired glucose
tolerance. Pregnant women were prospectively recruited to
the study and blood was collected at the first antenatal visit
and at the time of routine oral glucose tolerance test.
Women diagnosed with gestational diabetes were matched
with an equal number of normal pregnant (control) women.
Biomarkers under investigation included endocrine and
H. M. Georgiou
M. Lappas
A. Marita
M. Permezel

G. E. Rice
Department of Obstetrics and Gynecology,
University of Melbourne, Parkville, VIC, Australia
H. M. Georgiou (&)
M. Lappas
V. J. Bryant

R. Hiscock
M. Permezel
Mercy Perinatal Research Centre, Mercy Hospital for Women,
P.O. Box 5027, Heidelberg West, VIC 3081, Australia
e-mail: harrymg@unimelb.edu.au
G. M. Georgiou
Z. Khalil
Department of Biochemistry and Molecular Biology,
University of Melbourne, Parkville, VIC, Australia
Z. Khalil
College of Medicine, University of Sharjah,
Sharjah, United Arab Emirates
G. E. Rice
Translational Proteomics,
Baker Medical Research Institute,
Melbourne, VIC, Australia
metabolic hormones, cytokines and chemokines, and sur-
rogate markers of oxidative stress. Compared to controls,
women with gestational diabetes exhibited elevated plasma
insulin and reduced plasma adiponectin concentrations at
28 weeks gestation. Significant differences in insulin and
adiponectin concentrations were also observed in plasma at
11 weeks gestation. Bivariate logistic regression analysis
showed that both insulin and adiponectin are associated
with subsequent development of gestational diabetes.
Plasma insulin and adiponectin concentrations, when
measured at 11 weeks, may be predictive of impending
gestational diabetes. Further studies are warranted to
determine the reliability of these biomarkers.
Keywords Gestational diabetes mellitus
Insulin

Adiponectin
Biomarkers
Prediction
Introduction
Gestational diabetes mellitus (GDM) usually manifests
itself in the latter half of pregnancy and is characterized by
carbohydrate intolerance of variable severity. Australian
centres report a frequency of GDM of 5.5–8.8% of preg-
nancies, but this can be considerably higher in Polynesian,
South Asian (Indian, Chinese, Vietnamese), African and
Indigenous Australians [1, 2]. Generally, the prevalence of
GDM is proportional to the frequency of Type 2 diabetes
within a given population [3]. Risk factors for GDM
include obesity, increased maternal age, family history of
Type 2 diabetes, past history of GDM, previous adverse
pregnancy outcome and belonging to a high-risk ethnic
group [2].
The presence of GDM has implications for both
the mother and the baby. Perinatal morbidity includes
123
158
macrosomia, hypoglycaemia, hyperbilirubinaemia and
respiratory distress syndrome, which in turn may generate
subsequent complications [4]. Longer term consequences
for the offspring may include obesity and diabetes inde-
pendent of genetic factors [5–7]. For the mother, there is an
increased risk of overt Type 2 diabetes later in life [8, 9].
Arguably, both GDM and Type 2 diabetes share common
pathogenic mechanisms where pregnancy serves to unmask
disease in those women who are destined to develop Type
2 diabetes later in life. It has become increasingly apparent
that endocrine/metabolic hormones (e.g., leptin, adipo-
nectin, resistin), proinflammatory mediators (e.g., TNF-a,
IL-1b, IL-6) and markers of oxidative stress (nitrotyrosine,
protein carbonyls, lipid hydroperoxides) are strongly
associated with abnormal lipid and carbohydrate metabo-
lism, obesity and cardiovascular disease [10–12].
Aside from the lack of international agreement on the
diagnosis of GDM [13], controversy abounds as to the
relevance and timing of screening, how this would influ-
ence management, and how management may affect
perinatal outcomes. The Australasian Diabetes in Preg-
nancy Society (ADIPS) recommends that screening for
glucose tolerance be offered to all pregnant women at
28 weeks gestation, but if resources are limited, then those
women with risk factors should be screened. The benefit of
treating GDM was recently highlighted in a multicenter
study comparing 1,000 pregnant women with GDM who
were randomized to either intensive diabetes management
or routine pregnancy care. Serious perinatal complications
were reduced in the intervention group (1% complications
with no perinatal deaths) compared to the women receiving
only routine care (4% complications including five peri-
natal deaths) [14]. Postnatal maternal quality of life was
also improved with significantly lower rates of depression.
Although a direct clinical benefit of the early diagnosis
of GDM remains to be established conclusively, identifi-
cation of women at greatest risk would allow triage of the
patients to an appropriate model of care and identify a
group who are at particular need of glucose tolerance
assessment. Numerous GDM risk-factor assessments have
been attempted during first trimester pregnancy such as
family history of GDM and/or diabetes [15], fasting plasma
glucose [16], 1-h glucose challenge test [17], oral glucose
tolerance test [18] and hemoglobin A1c [19], and although
some have been able to provide a good negative predictive
measure for subsequent GDM [20], all tests suffer from
poor positive predictive values and are of limited efficacy.
It is evident that other metabolic markers that precede
hyperglycaemia would need to be identified if GDM were
to be predicted from a test in early pregnancy. Recently, a
number of first trimester studies have associated various
biomarkers with subsequent development of GDM. These
include elevated serum or plasma C-reactive protein [21],
123
Acta Diabetol (2008) 45:157–165
lower sex hormone-binding globulin [22], increased pla-
cental growth factor [23] and elevated leptin [24]
concentrations. Although the associations are compelling,
when studied in isolation, none of the markers provide
adequate positive predictive values for subsequent GDM.
Recently, a strong association between body mass index
and low plasma adiponectin concentration with subsequent
GDM was described [25]. Thus far, no studies have
assessed the utility of multiple biomarker screening for
GDM. The aim of this study was to identify differentially
expressed plasma proteins that may be useful biomarkers in
a group of women in the first trimester of pregnancy.
Biomarkers under investigation included endocrine and
metabolic hormones, cytokines and chemokines, and sur-
rogate markers of oxidative stress.
Subjects, research design and methods
Patient recruitment
Pregnant women attending the outpatient clinic at the
Mercy Hospital for Women (East Melbourne, Australia)
were recruited to the study by a clinical research midwife.
Informed consent was obtained from all participants. A total
of 250 pregnant women were recruited to the study at their
first antenatal visit. The study was approved by the Mercy
Health and Aged Care Human Research Ethics Committee.
All pregnant women attending the Mercy Hospital for
women are routinely screened for GDM by an oral glucose
tolerance test (OGTT) at approximately 28 weeks gesta-
tion, according to the Australasian Diabetes in Pregnancy
Society (ADIPS) guidelines [2]. Of the 250 participants,
14 were diagnosed with GDM after overnight fasting
followed by a 75 g glucose load. GDM was diagnosed by a
fasting glucose concentration C5.5 mmol/l and/or a 2 h
glucose concentration C8.0 mmol/l. All women diagnosed
with GDM are advised to adopt appropriate lifestyle/
dietary management. Six of the 14 women diagnosed with
GDM required insulin in the third trimester of pregnancy
according to hospital guidelines for insulin therapy in
GDM. Women diagnosed with GDM had a further OGTT
at 6 weeks postnatally. For the purposes of this study, an
equal number of matched, normal (non-GDM) controls
were also screened postnatally. Normal controls were
matched with GDM women for maternal age, gravidity,
parity, ethnicity and body mass index (BMI). BMI was
determined at the first antenatal visit. Six of the 14 women
with GDM were of Asian descent with appropriate
matching of controls. All women delivered a live birth. The
control group comprised one preterm birth (\37 weeks)
and three Caesarean section deliveries, while the GDM
group comprised one preterm birth and five Caesarean
Acta Diabetol (2008) 45:157–165
deliveries. Exclusion criteria included pre-existing diabetes
or pregnancies complicated with pre-eclampsia, intrauter-
ine growth restriction or fetal macrosomia.
Blood collection
Blood was collected from all study patients on three sepa-
rate occasions during the pregnancy: at the first antenatal
attendance (approximately 11 weeks gestation); at the time
of the oral glucose tolerance test (approximately 28 weeks
gestation); and postnatally at the time of a follow-up OGTT
(approximately 8 weeks after delivery). Venous blood was
collected for routine pathology testing (including glucose
determination) and a further 5 ml of blood was collected
into a vacuum EDTA tube for research purposes. Blood
collected at the first antenatal visit was a nonfasting, random
sample. During the OGTT, 5 ml of blood was collected at
fasting, 1 and 2 h after a 75 g glucose load. Blood samples
were immediately centrifuged at 1,000g for 5 min and the
plasma aliquoted into 1-ml microfuge tubes. The plasma
was supplemented with 0.1 mM phenylmethylsulfonyl
fluoride (PMSF) protease inhibitor (USB, Cleveland, OH)
and samples were immediately stored at -80°C.
Screening assays
Blood glucose determination was performed by the hos-
pital pathology department using an automated glucose
oxidase/oxygen-rate method. All other determinants were
performed in our research laboratory using assay kits or in-
house assays. ELISA assays were read using a Benchmark
microplate reader (Bio-Rad Laboratories, Hercules, CA)
and results analyzed with Microplate Manager (v4.0)
software. Cytokine assays were read using the Bio-Plex
workstation (Bio-Rad Laboratories, Hercules, CA) and
results analyzed with Bio-Plex Manager (v3.0) software.
Endocrine and metabolic hormone assays
Standard ELISA assay kits for insulin (Diagnostic Systems
Laboratories, Webster, TX; limit of detection 0.26 lU/ml),
adiponectin (R&D Systems, Minneapolis, MN; limit of
detection 32 pg/ml), leptin (BioSource International,
Camarillo, CA; limit of detection 7.2 pg/ml) and resistin
(CytoLab, Rehovot, Israel; limit of detection 16 pg/ml)
were purchased and used according to the manufacturer’s
instructions. The adiponectin assay detects total adipo-
nectin protein including low, medium and high molecular
isoforms. For all assays, manufacturer specifications indi-
cate that intraassay and interassay coefficients of variation
are less than 10%. Plasma samples collected at 11 weeks
(random), 28 weeks (fasting and 2-h) and postnatally
(fasting and 2-h) were screened for each hormone.
159
Cytokine assays
Cytokine determination was performed using the Bio-Plex
suspension assay system and 17-plex human plasma cyto-
kine assay kits (Bio-Rad Laboratories, Hercules, CA). The
system uses fluorescent-coded polystyrene beads, each of
which is conjugated with a specific antibody and used to
create a sandwich immunoassay. Sample and fluoro-
chrome-conjugated antibody are allowed to react with the
antibody-conjugated beads in microplate wells. The assay
utilized 50 ll of plasma at ‘ dilution. The panel of
potential biomarkers assayed in this study included pro-
and anti-inflammatory cytokines as well as chemokines.
The limit of detection varied for each cytokine and ranged
from 0.2 to 3 pg/ml. For all assays, manufacturer specifi-
cations indicate that intraassay and interassay coefficients
of variation are less than 10%. Plasma samples collected at
11 weeks (random) and 28 weeks (fasting) were screened
for each cytokine.
Oxidative stress assays
Two surrogate markers of oxidative stress, plasma ni-
trotyrosine (NT) and lipid hydroperoxide (LPO), were
determined using in-house assays. NT is the major product
obtained from the nitration of tyrosine residues by per-
oxynitrite. NT was determined using a modified
competitive ELISA assay with nitrated bovine serum
albumin standards and rabbit anti-NT-BSA (Cayman
Chemical, Ann Arbor, MI), as previously published [26].
Briefly, plates were coated with nitrated-BSA overnight,
and an equal volume of sample (or standard) and rabbit
antinitrotyrosine antibody were incubated for 3 h. Bound
antibody was detected with sheep antirabbit HRP anti-
body. Reactants were visualized with TMB substrate and
read at 450 nm. Plasma NT values are expressed as NT-
BSA equivalents. The limit of detection of the assay was
400 ng/ml, and intraassay and interassay coefficients of
variation were less than 10%.
Peroxidation of both saturated and unsaturated lipids is
a consequence of oxidative injury in many pathophysio-
logical disorders. LPO was measured directly using a
modified method previously published [27]. The method
combines a deproteination procedure with the extraction
of lipid peroxides into chloroform. Hydroperoxides are
highly unstable and readily react with ferrous ions to
produce ferric ions. The resulting ions are detected using
thiocyanate ion as the chromogen and read at 500 nm.
The limit of detection of the assay was 0.5 nmol/ml and
intraassay and interassay coefficients of variation were
less than 10%. Plasma samples collected at 11 weeks
(random) and 28 weeks (fasting) were screened for each
marker.
123
160 Acta Diabetol (2008) 45:157–165

Statistical analysis Table 1 Summary of participant demographic data and blood glu-
cose values
Participant demographic data and assay data were statisti- Control GDM P value
cally analyzed using SPSS (v11.5 for Windows) and Stata (Mann–
(v9) software. Group data were initially analyzed for Whitney)
equality of variances and group means were compared Sample size (N) 14 14
using Student’s t test or Mann–Whitney U test. Values a
Maternal age (years) 32.6 ± 3.4 33.8 ± 5.0 NS
falling below the detectable limit of the assay were Graviditya 3.4 ± 1.7 4.1 ± 1.9 NS
recorded as zero, and nonparametric statistical analysis was Paritya 1.6 ± 0.9 2.1 ± 1.7 NS
performed. Similarly, for bivariate correlation coefficients, Body mass indexa 24.7 ± 5.1 28.2 ± 8.4 NS
parametric and nonparametric data was analyzed using
Gestation at delivery (weeks) 39.6 ± 1.8 38.0 ± 2.0 0.031
Pearson’s r or Spearman’s rho, respectively. Conditional
Gestation at first visit (weeks) 10.6 ± 2.6 11.1 ± 2.6 NS
univariate and bivariate logistic regression analyses were
Gestation at OGTT (weeks) 28.5 ± 1.1 26.6 ± 4.1 NS
used to determine the influence of significant independent
Post-partum visit (weeks) 6.7 ± 0.9 8.5 ± 3.7 NS
variables on GDM outcome. Receiver operating charac-
First visit Random glucose 4.7 ± 0.4 5.8 ± 1.1 0.003
teristics (ROC) curves were generated to evaluate the (mmol/l)
sensitivity and specificity of single and binary classifiers 28-week Fasting glucose 4.4 ± 0.3 5.4 ± 0.8 0.001
and to determine discrimination thresholds. (mmol/l)
28-week OGTT 2-h glucose 5.5 ± 0.7 8.9 ± 1.9 \0.001
(mmol/l)
Results Post-partum Fasting glucose 4.6 ± 0.3 5.1 ± 0.6 0.002
(mmol/l)
Patient demography Post-partum OGTT 2-h glucose 4.6 ± 0.8 6.1 ± 1.3 0.008
(mmol/l)

A summary of patient demographic data is presented in Data are expressed as mean ± SD

Table 1. The results of n = 14 GDM participants and
n = 14 matched controls are presented in this paper. Diag-
nosis of GDM was made on 28 week OGTT based on the
ADIPS criteria. Matched normal controls were selected after
women with GDM were identified. There were no signifi-
cant differences in maternal age, gravidity, parity or BMI
between GDM and control women at the time of recruit-
ment. Although the BMI for the GDM group falls within the
‘‘overweight’’ category, this was mainly due to one extreme
outlier that could not be closely matched. There was no
significant difference between groups in the timing of the
three visits (approximately 11 weeks at first antenatal visit,
28 weeks at OGTT and 8 weeks at postdelivery OGTT.)
Women with GDM were delivered significantly earlier than
controls (38.0 vs. 39.6 weeks, respectively). Five of the 14
women with GDM and five of 14 controls had a family
history of diabetes. As best could be ascertained none of the
women had preconception impaired glucose tolerance.
Blood glucose
Both the fasting glucose (control 4.4 ± 0.3 vs. GDM
5.4 ± 0.8 mmol/l, P = 0.001) and 2-h glucose (control
5.5 ± 0.7 vs. GDM 8.9 ± 1.9 mmol/l, P \ 0.001) con-
centrations were significantly higher in women with GDM
at the 28 week OGTT (Table 1). Also of note was the
fact that random blood glucose concentrations were sig-
nificantly different at first visit where pre-GDM women
a
Maternal age, gravidity, parity and BMI were determined at the
time of recruitment and control subjects were matched against sub-
jects with GDM
showed higher values (control 4.7 ± 0.4 vs. pre-GDM
5.8 ± 1.1 mmol/l, P = 0.003). At 8 weeks postpartum
OGTT, women previously with GDM had significantly
higher fasting and 2-h glucose concentrations (Table 1).
Three of the 14 postpartum women with GDM had
impaired glucose tolerance but none were overtly diabetic.
Endocrine and metabolic hormones
Insulin and HOMA-IR
These data are summarized in Table 2. At 28 week OGTT,
there was a significant difference between control and
GDM women in fasting (8.9 ± 3.7 vs. 14.3 ± 8.3 lU/ml,
P = 0.014) and 2-h (62.1 ± 30.2 vs. 101.6 ± 40.1 lU/ml,
P = 0.012) plasma insulin concentrations. These high
insulin values are consistent with pregnancy. There was
also a significant difference in random plasma insulin
between control and pre-GDM women at the first antenatal
visit (13.4 ± 16.3 vs. 45.9 ± 39.4 lU/ml, P \ 0.001).
Similarly, postpartum insulin concentrations were signifi-
cantly different between control and post-GDM women at
fasting (5.9 ± 2.1 vs. 10.0 ± 5.6 lU/ml, P = 0.005) and
at 2-h OGTT (20.4 ± 17.2 vs. 43.2 ± 23.9 lU/ml,

123
Acta Diabetol (2008) 45:157–165 161

Table 2 Summary of plasma hormone concentrations in control and GDM subjects at 11 and 28 weeks gestation and 8 weeks postpartum
Hormone Visit Sample Control GDM Statistic
(mean ± SD) (mean ± SD) (Mann–Whitney U test)

Insulin (lU/ml) Antenatal 11w Random 13.4 ± 16.3 45.9 ± 39.4 P \ 0.001
Antenatal 28w Fasting OGTT 8.9 ± 3.7 14.3 ± 8.3 P = 0.014
Antenatal 28w 2-h OGTT 62.1 ± 30.2 101.6 ± 40.1 P = 0.012
Post-partum 8w Fasting OGTT 5.9 ± 2.1 10.0 ± 5.6 P = 0.005
Post-partum 8w 2-h OGTT 20.4 ± 17.2 43.2 ± 23.9 P = 0.006
Adiponectin (lg/ml) Antenatal 11w Random 4.9 ± 0.6 3.0 ± 0.2 P = 0.001
Antenatal 28w Fasting OGTT 4.5 ± 1.9 2.9 ± 1.1 P = 0.008
Antenatal 28w 2-h OGTT 3.9 ± 1.6 3.2 ± 0.8 P = 0.039
Post-partum 8w Fasting OGTT 4.0 ± 1.7 2.8 ± 1.0 NS
Post-partum 8w 2-h OGTT 4.0 ± 1.9 3.0 ± 0.8 NS
Leptin (ng/ml) Antenatal 11w Random 21.2 ± 14.5 26.1 ± 16.5 NS
Antenatal 28w Fasting OGTT 30.7 ± 10.3 28.4 ± 19.0 NS
Antenatal 28w 2-h OGTT 18.2 ± 12.6 21.1 ± 17.3 NS
Post-partum 8w Fasting OGTT 24.4 ± 13.1 24.7 ± 17.5 NS
Post-partum 8w 2-h OGTT 18.4 ± 14.6 17.7 ± 15.2 NS
Resistin (ng/ml) Antenatal 11w Random 2.5 ± 1.1 2.7 ± 2.5 NS
Antenatal 28w Fasting OGTT 2.6 ± 1.0 2.7 ± 2.4 NS
Antenatal 28w 2-h OGTT 1.9 ± 0.6 2.6 ± 2.3 NS
Post-partum 8w Fasting OGTT 2.4 ± 0.8 2.3 ± 1.3 NS
Post-partum 8w 2-h OGTT 1.8 ± 0.6 2.3 ± 1.2 NS

P = 0.006). Plasma insulin concentrations at first antenatal
visit (11 weeks) significantly correlated with BMI (Pear-
son’s r = 0.412, P = 0.037).
It is generally accepted that pregnancy is associated with
increased insulin resistance. Our data confirmed that
HOMA-IR values in control women are significantly
higher during pregnancy (28 weeks antenatal) than those
after pregnancy (8 weeks postnatal) (mean 1.8 ± 0.8 vs.
1.1 ± 0.4, t test P = 0.005). These values were even more
elevated in women with GDM (mean 3.5 ± 2.4 vs.
2.3 ± 0.4, t test P = 0.043). There was a significant dif-
ference in HOMA-IR values between 28 week antenatal
control and GDM subjects (1.8 ± 0.8 vs. 3.5 ± 2.4,
Mann–Whitney U test P = 0.004). There was no differ-
ence in the antenatal HOMA-IR between Caucasian
(n = 8) and Asian (n = 6) women with GDM.
Adipokines
These data are summarized in Table 2. Plasma adiponectin
concentrations were significantly lower in women with
GDM compared to control women at both 11 weeks
(4.9 ± 0.6 vs. 3.0 ± 0.2 lg/ml, P = 0.001) and 28 weeks
gestation (fasting 4.5 ± 1.9 vs. 2.9 ± 1.1 lg/ml, P =
0.008; and 2-h 3.9 ± 1.6 vs. 3.2 ± 0.8 lg/ml, P = 0.039).
No difference was observed postpregnancy between GDM
and control groups. Leptin did not significantly differ
between women with GDM and control women either
during or after pregnancy. Leptin concentrations at first
antenatal visit (11 weeks) significantly correlated with
BMI (Pearson’s r = 0.684, P \ 0.001). Resistin did not
significantly differ between women with GDM and control
women either during or after pregnancy.
Correlation with glucose
Correlation analysis was made between fasting and 2-h
glucose measurements at 28 weeks and respective insulin,
adiponectin, leptin and resistin measurements. Pooled
(GDM and control) fasting plasma insulin significantly
correlated with fasting glucose concentrations (Pearson’s
r = 0.549, P = 0.002). Similarly, 2-h insulin significantly
correlated with 2-h glucose (Pearson’s r = 0.710,
P \ 0.001). Fasting adiponectin negatively correlated with
fasting glucose concentrations (Pearson’s r = - 0.383,
P = 0.044), while 2-h adiponectin also negatively corre-
lated with 2-h glucose concentrations (Pearson’s r =
- 0.376, P = 0.049). No correlation was evident between
glucose concentrations and leptin or resistin.
Cytokines and chemokines
Several of the cytokines/chemokines tested were unde-
tectable in all or most of the plasma samples screened.

123
162 Acta Diabetol (2008) 45:157–165

Table 3 Summary of plasma cytokine concentrations in control and GDM subjects at 11 and 28 weeks gestation
Cytokine Cytokine concentration (pg/ml)
Antenatal 11wRandom blood(mean ± SD) Antenatal 28wFasting blood(mean ± SD)
Control GDM Statistic Control GDM Statistic
(Student t test) (Student t test)

TNF-a 2.18 ± 1.11 2.17 ± 1.23 P = 0.971 6.02 ± 3.33 5.79 ± 3.22 P = 0.852
IFN-c 56.43 ± 84.89 36.90 ± 52.78 P = 0.471 18.78 ± 20.94 15.38 ± 13.71 P = 0.616
IL-2 5.84 ± 4.33 5.06 ± 3.81 P = 0.617 0.00 ± 0.00 0.00 ± 0.00 N/A
IL-5 2.08 ± 1.24 1.63 ± 0.50 P = 0.216 1.95 ± 1.88 1.39 ± 1.39 P = 0.379
IL-6 66.13 ± 42.05 74.38 ± 69.49 P = 0.707 31.87 ± 20.95 26.57 ± 27.50 P = 0.571
IL-7 14.18 ± 11.69 14.69 ± 4.24 P = 0.878 9.61 ± 5.11 11.35 ± 4.66 P = 0.354
IL-8 2.89 ± 0.94 3.88 ± 1.61 P = 0.058 4.72 ± 3.79 4.10 ± 2.09 P = 0.598
IL-10 4.91 ± 10.57 3.54 ± 5.87 P = 0.675 2.48 ± 5.25 1.00 ± 1.62 P = 0.323
IL-13 4.37 ± 6.07 1.65 ± 1.87 P = 0.120 2.87 ± 7.47 1.16 ± 4.36 P = 0.467
MCP-1 22.33 ± 5.54 24.91 ± 7.60 P = 0.314 60.91 ± 10.00 72.34 ± 24.69 P = 0.121
MIP-1b 32.58 ± 10.98 41.67 ± 14.89 P = 0.078 40.88 ± 11.63 42.44 ± 14.52 P = 0.756

These cytokines were disregarded, as meaningful analysis
of the data could not be made and included IL-1b, IL-4,
IL-12, IL-17, G-CSF and GM-CSF. Of the remaining
cytokines (IL-2, IL-5, IL-6, IL-7, IL-8, IL-10, IL-13,
IFN-c, TNF-a, MCP-1 and MIC-1b), none were found to
significantly differ between control and women with GDM
at either 11 weeks (random) or 28 weeks (fasting) gesta-
tion (Table 3). The only noteworthy observation was that
circulating IL-8 (t test, P = 0.058) and MIP-1b (t test,
P = 0.078) concentrations were elevated in pre-GDM
subjects compared to control at 11 weeks gestation, but the
differences were marginally insignificant.
Surrogate markers of oxidative stress
Nitrotyrosine
Plasma NT (NT-BSA equivalents) was significantly ele-
vated with advancing gestation (11 weeks, 2,258 ± 897
ng/ml vs. 28 weeks, 3,051 ± 812 ng/ml, paired t test,
P = 0.001). A significantly lower concentration of NT was
seen in women with GDM compared to controls at 28 week
fasting OGTT (2,711 ± 689 vs. 3,389 ± 804 ng/ml, t test
P = 0.024). This difference was not evident at 11 weeks
gestation (Fig. 1a).
Lipid hydroperoxide
The concentration of plasma LPO significantly decreased
with advancing gestation (11 weeks, 8.54 ± 7.03 nmol/ml
vs. 28 w, 3.12 ± 2.61 nmol/ml, paired t test, P = 0.001).
A significantly lower concentration of LPO was seen in
GDM women compared to controls at 28 week fasting
OGTT (1.97 ± 2.33 vs. 4.06 ± 2.56 nmol/ml, t test
P = 0.033). This difference was not evident at 11 weeks
gestation (Fig. 1b).
Regression analysis and GDM outcome
The association of putative predictive biomarkers of GDM
was analyzed using conditional univariate logistic regres-
sion and included 11 week values for insulin, adiponectin,
leptin, resistin, IFN-c, TNF-a, IL-2, IL-5, IL-6, IL-7, IL-8,
IL-10, IL-13, MCP-1, MIC-1b, NT and LPO. Of these, five
variables (insulin, adiponectin, IL-8, IL-13 and MIP-1b)
were associated with GDM outcome at a significance level
of 0.2. Given the sample size of 28, bivariate conditional
logistic regression (CLR) was performed to assess associ-
ation of each predictor variable with outcome when
adjusted for all possible pairings. The likelihood ratio chi2
statistic (LR Chi2) was used to test the null hypothesis
(regression coefficients equal zero) for both variables in the
model. Details are presented in Table 4. When unadjusted
for multiple comparisons, insulin and adiponectin were
associated with outcome. When adjusted for multiple
comparisons, using the conservative Bonferroni correction
(significance level 0.05/20 comparisons, adjusted signifi-
cance level = 0.0025), insulin was the only variable to
remain associated with outcome. Insulin’s regression
coefficient was 63.5% greater with inclusion of adiponectin
compared to that from univariable CLR, indicating sig-
nificant confounding by adiponectin. Adiponectin was
retained in the multivariable CLR model to adjust for this
potential confounding despite having only a weak associ-
ation with outcome and a small effect on size.
ROC curves (Fig. 2) for unconditional models contain-
ing insulin (panel a), adiponectin (panel b) and both insulin
and adiponectin (panel c) are presented from data derived

123
Acta Diabetol (2008) 45:157–165 163

Fig. 1 Nitrotyrosine (panel a) and lipid hydroperoxide (panel b)
evaluation at 11 weeks gestation (random blood) and at 28 weeks
gestation (2-h OGTT). A significant difference in both nitrotyrosine
and lipid hydroperoxide between GDM and control subjects was
evident at 28 weeks. There was no difference between controls and
pre-GDM subjects early in pregnancy. Overall, plasma nitrotyrosine
was significantly elevated with advancing gestation, while lipid
hydroperoxide was significantly diminished with advancing gestation.
The box represents the mean and interquartile range, while the
whiskers represent the 5th and 95th centiles

Table 4 Bivariate conditional logistic regression for all possible long-term risks to the fetus are to be minimized. Diagnosing
pairings of each predictor variable at 11 weeks gestation GDM by oral glucose tolerance test is inappropriate during
Primary Secondary LR Chi2 the first trimester of pregnancy as pregnancy-induced
variable variable hyperglycaemia is often not evident at this early stage.
Bivariable CLRa Exclusion of
primary Despite the relatively small sample sizes, this study was
variableb strengthened by the close matching of control subjects
against subjects diagnosed with GDM. Women were mat-
Insulin Adiponectin 0.04 0.004
ched for age, ethnicity, gravidity, parity and BMI, thus
IL-8 0.14 0.001
negating these parameters as confounders. Differences in
IL-13 0.11 0.001
circulating metabolic and inflammatory biomarkers at the
MIP-1b Nonconvergence time of GDM were not overtly discernible between groups
Adiponectin IL-8 0.69 0.04 and of the 23 biomarkers investigated at 28 weeks gesta-
IL-13 0.31 0.03 tion (four metabolic hormones, 17 cytokines and two
MIP-1b 0.22 0.02 surrogate markers of oxidative stress); only plasma insulin,
IL-8 IL-13 0.13 0.16 adiponectin, nitrotyrosine and lipid peroxidation were
MIP-1b 0.33 0.31 found to significantly differ between control women and
IL-13 MIP-1b 0.12 0.1 women with GDM.
a
Bivariable conditional logistic regression (CLR) analysis showing Pregnancy is often associated with insulin resistance.
the likelihood ratio chi2 statistic (LR Chi2)-associated P value This was demonstrated by elevated 28-week antenatal
between models always containing the primary variable HOMA-IR values compared to postpartum values. Plasma
b
LR Chi2 statistic-associated P value for the bivariable model insulin concentrations and HOMA-IR values are further
compared to that excluding the primary variable
increased in GDM. Both fasting and 2-h OGTT mean
insulin concentrations were approximately twofold higher
at 11 weeks gestation. Classification analysis using the in subjects with GDM compared to controls. A weakness of
unconditional model containing both insulin and adipo- the study was that first trimester blood samples were ran-
nectin with a probability cut-off of P = 0.6 correctly domly collected thus causing greater variation in insulin
classified 12 of 14 control subjects and 12 of 14 GDM concentrations. Nevertheless, differences in insulin
subjects (panel d). The model provides a sensitivity and between pre-GDM and matched controls were highly sig-
specificity of 85.7%, and positive and negative predictive nificant. The current study confirmed that GDM is
values of 85.7%. Based upon this data set, the predictive associated with reduced circulating adiponectin concen-
threshold values for GDM are [25 lU/ml insulin and tration [25, 28], but found no differences in circulating
\3.5 lg/ml adiponectin. leptin or resistin concentrations between matched control
and GDM subjects.
Our objective was to identify predictive biomarker(s) of
Discussion impending pregnancy-induced glucose-intolerance. We
postulated that hyperglycaemia associated with GDM at the
There is a growing need to diagnose and manage time of OGTT (*28 weeks gestation) is preceded by dif-
pregnancy-induced diabetes earlier if the short-term and ferential protein expression early in pregnancy. Of the

123
164
Fig. 2 Receiver operating
characteristic (ROC) curves
were generated for insulin
(panel a), adiponectin (panel b),
and both insulin and adiponectin
(panel c) from data derived at
11 weeks gestation.
Classification analysis using the
unconditional model containing
both insulin and adiponectin
with a probability cut-off of
P = 0.6 correctly classified 12
of 14 control subjects and 12 of
14 subjects with GDM (panel d)
23 biomarkers investigated at 11 weeks gestation, only
plasma insulin and adiponectin were found to significantly
differ between those of uncomplicated control women and
women destined to develop GDM. We confirmed the
observation of Williams et al. [25] that a low blood con-
centration of adiponectin early in pregnancy is associated
with increased risk of subsequent GDM. Based on our
dataset, predictive threshold values for GDM are[25 lU/ml
insulin and \3.5 lg/ml adiponectin. Utilizing these
thresholds, 12 of 14 control subjects and 12 of 14 GDM
subjects (based on ADIPS criteria) were correctly classified.
Both type 1 and type 2 diabetes are associated with
increased oxidative stress as measured by lipid peroxidation
and protein oxidation/nitration [29]. In this study, however,
circulating levels of both NT and LPH were significantly
lower in subjects with GDM. The reason for this is unclear,
but may be due to a relatively acute pregnancy-induced
increase in antioxidant activity, particularly superoxide
dismutase [30], that is further increased in GDM [31]. In the
longer term, however, as overt diabetes becomes a chronic
condition, oxidative stress is compounded by the formation
of advanced glycation endproducts leading to tissue damage
and diabetic complications.
Several circulating cytokines were found to alter with
advancing pregnancy, but none of the changes were asso-
ciated with GDM. We confirmed the observation of Kirwan
et al. [32] that circulating TNF-a increases with advancing
pregnancy, but found no difference between control and
GDM subjects (data not shown). While there was a strong
correlation between paired samples from the same indi-
vidual during pregnancy, considerable variation in cytokine
expression exists between individuals. This may reflect
occult infection, advanced maternal age or differences in
123
Acta Diabetol (2008) 45:157–165
BMI. The absence of a difference in plasma cytokine levels
between GDM and control subjects is consistent with our
previous tissue explant model, where there were no dif-
ferences in TNF-a, IL-6 and IL-8 secretion from placental,
adipose and skeletal muscle explants between control and
GDM subjects [33].
The diagnosis of GDM based on blood glucose values is
still controversial, with different international expert study
groups applying different diagnostic criteria. As such, the
prevalence of GDM can vary enormously either due to
‘‘under-diagnosis’’ (e.g. Canadian Diabetes Association
criteria) or ‘‘over-diagnosis’’ (e.g. ADIPS criteria) [13].
Attempting to diagnose GDM using any new biomarker(s)
is fraught with danger, as this will ultimately be compared
against one (or more) existing set of glucose-based criteria.
It is interesting to note that the two women with GDM,
‘‘incorrectly’’ classified as normal by our predictive mod-
elling, would have been excluded as GDM, had Canadian
or American diagnostic criteria been used. Alternatively,
the two control subjects ‘‘incorrectly’’ classified as GDM
were found to have high insulin concentrations probably
due to the fact that random, rather than fasting blood,
samples were utilized in this study.
Our data demonstrate that differential expression of
protein biomarkers preceding the onset of hyperglycaemia
can be used to predict the onset of GDM. The predictive
utility of plasma insulin in combination with adiponectin is
tantalizing, warranting a more extensive prospective phase
II diagnostic trial using first-trimester fasting plasma for
screening. A multianalyte first-trimester diagnostic incor-
porating these and other potential biomarkers such as sex
hormone-binding globulin and high sensitive C-reactive
protein [34] may provide improved predictive utility.
Acta Diabetol (2008) 45:157–165
Determining appropriate predictive threshold values for
insulin and adiponectin (or any other biomarker) requires
further analysis; however, the need for insulin treatment
may still be determined best by oral glucose tolerance test
later in pregnancy.
Acknowledgments The authors wish to acknowledge the assistance
of clinical research nurses Angela S. Denning, Lynette K. Tuttle,
Joanna L. McKay and Melissa A. Bolger for recruiting women and
obtaining their consent to the study; Sarah Holdsworth for expert
technical assistance; and Network Pathology for performing blood
glucose determinations. This work was funded by the Medical
Research Foundation for Women and Babies; and the Diabetes
Australia Research Trust.
References
1. Beischer NA, Oats JN, Henry OA, Sheedy MT, Walstab JE
(1991) Incidence and severity of gestational diabetes mellitus
according to country of birth in women living in Australia.
Diabetes 40(Suppl 2):35–38
2. Hoffman L, Nolan C, Wilson JD, Oats JJ, Simmons D (1998)
Gestational diabetes mellitus—management guidelines. The Aus-
tralasian Diabetes in Pregnancy Society. Med J Aust 169:93–97
3. American College of Obstetricians and Gynecologists Committee
on Practice Bulletins–Obstetrics (2001) ACOG practice bulletin.
Clinical management guidelines for obstetrician-gynecologists.
Number 30, September 2001 (replaces Technical Bulletin Num-
ber 200, December 1994). Gestational diabetes. Obstet Gynecol
98:525–538
4. Hod M, Merlob P, Friedman S, Schoenfeld A, Ovadia J (1991)
Gestational diabetes mellitus. A survey of perinatal complications
in the 1980s. Diabetes 40(Suppl 2):74–78
5. Van Assche FA, Aerts L, Holemans K (1992) Maternal diabetes
and the effect for the offspring. Verh K Acad Geneeskd Belg
54:95–106 discussion 107–108
6. Silverman BL, Metzger BE, Cho NH, Loeb CA (1995) Impaired
glucose tolerance in adolescent offspring of diabetic mothers.
Relationship to fetal hyperinsulinism. Diabetes Care 18:611–617
7. Van Assche FA, Holemans K, Aerts L (2001) Long-term con-
sequences for offspring of diabetes during pregnancy. Br Med
Bull 60:173–182
8. Mestman JH (1987) Follow-up studies in women with gestational
diabetes mellitus. Springer-Verlag, New York, pp 191–198
9. O’Sullivan JB (1989) The Boston gestational diabetes studies:
review and perspectives. Springer-Verlag, London, pp 287–294
10. Dandona P, Aljada A, Bandyopadhyay A (2004) Inflammation:
the link between insulin resistance, obesity and diabetes. Trends
Immunol 25:4–7
11. Evans JL, Maddux BA, Goldfine ID (2005) The molecular basis
for oxidative stress-induced insulin resistances. Antioxid Redox
Signal 7:1040–1052
12. Federici M, Lauro R (2005) Review article: diabetes and ath-
erosclerosis—running on a common road. Aliment Pharmacol
Ther 22(Suppl 2):11–15
13. Agarwal MM, Dhatt GS, Punnose J, Koster G (2005) Gestational
diabetes: dilemma caused by multiple international diagnostic
criteria. Diabet Med 22:1731–1736
14. Crowther CA, Hiller JE, Moss JR, McPhee AJ, Jeffries WS,
Robinson JS (2005) Effect of treatment of gestational diabetes
mellitus on pregnancy outcomes. N Engl J Med 352:2477–2486
15. Caliskan E, Kayikcioglu F, Ozturk N, Koc S, Haberal A (2004) A
population-based risk factor scoring will decrease unnecessary
165
testing for the diagnosis of gestational diabetes mellitus. Acta
Obstet Gynecol Scand 83:524–530
16. Sacks DA, Chen W, Wolde-Tsadik G, Buchanan TA (2003)
Fasting plasma glucose test at the first prenatal visit as a screen
for gestational diabetes. Obstet Gynecol 101:1197–1203
17. Maegawa Y, Sugiyama T, Kusaka H, Mitao M, Toyoda N (2003)
Screening tests for gestational diabetes in Japan in the 1st and 2nd
trimester of pregnancy. Diabetes Res Clin Pract 62:47–53
18. Bhattacharya SM (2004) Fasting or two-hour postprandial plasma
glucose levels in early months of pregnancy as screening tools for
gestational diabetes mellitus developing in later months of
pregnancy. J Obstet Gynaecol Res 30:333–336
19. Cypryk K, Czupryniak L, Wilczynski J, Lewinski A (2004)
Diabetes screening after gestational diabetes mellitus: poor per-
formance of fasting plasma glucose. Acta Diabetol 41:5–8
20. Nahum GG, Wilson SB, Stanislaw H (2002) Early-pregnancy
glucose screening for gestational diabetes mellitus. J Reprod Med
47:656–662
21. Wolf M, Sandler L, Hsu K, Vossen-Smirnakis K, Ecker JL,
Thadhani R (2003) First-trimester C-reactive protein and sub-
sequent gestational diabetes. Diabetes Care 26:819–824
22. Thadhani R, Wolf M, Hsu-Blatman K, Sandler L, Nathan D,
Ecker JL (2003) First-trimester sex hormone binding globulin and
subsequent gestational diabetes mellitus. Am J Obstet Gynecol
189:171–176
23. Ong CY, Lao TT, Spencer K, Nicolaides KH (2004) Maternal
serum level of placental growth factor in diabetic pregnancies.
J Reprod Med 49:477–480
24. Qiu C, Williams MA, Vadachkoria S, Frederick IO, Luthy DA
(2004) Increased maternal plasma leptin in early pregnancy and
risk of gestational diabetes mellitus. Obstet Gynecol 103:519–
525
25. Williams MA, Qiu C, Muy-Rivera M, Vadachkoria S, Song T,
Luthy DA (2004) Plasma adiponectin concentrations in early
pregnancy and subsequent risk of gestational diabetes mellitus.
J Clin Endocrinol Metab 89:2306–2311
26. Khan J, Brennand DM, Bradley N, Gao B, Bruckdorfer R, Jacobs
M (1998) 3-Nitrotyrosine in the proteins of human plasma
determined by an ELISA method. Biochem J 330:795–801
27. Mihaljevic B, Katusin-Razem B, Razem D (1996) The reevalu-
ation of the ferric thiocyanate assay for lipid hydroperoxides with
special considerations of the mechanistic aspects of the response.
Free Radic Biol Med 21:53–63
28. Retnakaran R, Hanley AJ, Raif N, Connelly PW, Sermer M,
Zinman B (2004) Reduced adiponectin concentration in women
with gestational diabetes: a potential factor in progression to type
2 diabetes. Diabetes Care 27:799–800
29. Atalay M, Laaksonen DE (2002) Diabetes, oxidative stree and
physical exercise. J Sport Sci Med 1:1–14
30. Perkins AV (2006) Endogenous anti-oxidants in pregnancy and
preeclampsia. Aust N Z J Obstet Gynaecol 46:77–83
31. Coughlan MT, Vervaart PP, Permezel M, Georgiou HM, Rice GE
(2004) Altered placental oxidative stress status in gestational
diabetes mellitus. Placenta 25:78–84
32. Kirwan JP, Hauguel De Mouzon S, Lepercq J, Challier JC,
Huston Presley L, Friedman JE, Kalhan SC, Catalano PM (2002)
TNF-alpha is a predictor of insulin resistance in human preg-
nancy. Diabetes 51:2207–2213
33. Lappas M, Permezel M, Rice GE (2004) Release of proinflam-
matory cytokines and 8-isoprostane from placenta, adipose tissue,
and skeletal muscle from normal pregnant women and women
with gestational diabetes mellitus. J Clin Endocrinol Metab
89:5627–5633
34. Smirnakis KV, Plati A, Wolf M, Thadhani R, Ecker JL (2007)
Predicting gestational diabetes: choosing the optimal early serum
marker. Am J Obstet Gynecol 196:410.e1–410.e7
123