biochemical hyperandrogenism is associated with metabolic syndrome independently of adiposity and...
TRANSCRIPT
ORIGINAL ARTICLE
Biochemical hyperandrogenism is associated with metabolicsyndrome independently of adiposity and insulin resistancein Romanian polycystic ovary syndrome patients
Alice Albu • Serban Radian • Simona Fica •
Carmen Gabriela Barbu
Received: 23 December 2013 / Accepted: 12 June 2014
� Springer Science+Business Media New York 2014
Abstract The aim of the study was to determine whether
Romanian polycystic ovary syndrome (PCOS) patients
have an increased prevalence of metabolic syndrome
(MetS) and to study the involvement of adiposity, insulin
resistance and hyperandrogenism in the pathogenesis of
MetS in PCOS. A total of 398 PCOS patients and 126
controls were evaluated between January 2006 and
December 2012. MetS was defined by National Cholesterol
Education Program, Adult Treatment Panel III criteria.
Principal component analysis (PCA) was used to analyze
the correlations among variables of interest by grouping
them in few components, and principal component (PCs)
scores were saved and used as independent variables in
logistic regression. The prevalence of MetS was higher
among patients with PCOS (20.4 %) than in controls
(11.1 %, p \ 0.05). In PCOS patients, PCA extracted three
PCs from the analyzed variables. First PC aggregated
variables related to adiposity and insulin resistance, with
factor loadings showing strong relationship between these
parameters. The second PC included markers of hyperan-
drogenemia and was best represented by free androgen
index (FAI) which correlated strongly and exclusively with
this PC. The third component was best represented by
hirsutism. Logistic regression analysis revealed that in
PCOS patients, the first and the second PCs were inde-
pendently associated with MetS, whereas the third com-
ponent was not. Romanian PCOS patients have an
increased risk for MetS; adiposity, insulin resistance and
hyperandrogenemia, but not hirsutism, are independent
predictors of MetS presence. Our data also suggest that
insulin resistance is only secondary to increased adiposity
and FAI is a good marker of biochemical hyperandroge-
nism with little influences from the metabolic component.
Keywords Polycystic ovary syndrome � Insulin
resistance � Hyperinsulinemia � Hyperandrogenemia �Metabolic syndrome
Introduction
More than 75 years have passed from the first description
by Stein and Leventhal of the disease known today as
polycystic ovary syndrome (PCOS) [1]. Although at the
beginning PCOS was considered mainly a reproductive
disorder, studies of the last two decades showed that PCOS
also implies a predisposition to metabolic complications
and an association with cardiovascular and metabolic risk
factors [2, 3]. Metabolic syndrome (MetS) is among the
most studied cardio-metabolic risk factors in PCOS
patients, and a recent meta-analysis reported an increased
risk for MetS of PCOS patients [3]. Nevertheless, not all
studies of PCOS patients found an association with MetS
[4]. This could be due to heterogeneity of the study pop-
ulations with regard to various confounding factors such as
age, adiposity and ethnicity. For example, the prevalence
of obesity ranges from 25.9 % in Italian PCOS patients [5]
to 61 and 76 %, respectively, in the USA [6] and Australia
A. Albu � S. Radian � S. Fica � C. G. Barbu
‘‘Carol Davila’’ University of Medicine and Pharmacy, Dionisie
Lupu Street 37, Bucharest, Romania
A. Albu � S. Fica (&) � C. G. Barbu
Endocrinology and Diabetes Department, Elias University
Hospital, Marasti Street 17, Bucharest, Romania
e-mail: [email protected]
S. Radian
‘‘CI Parhon’’ National Institute of Endocrinology, Aviatorilor
Blvd 34-36, Bucharest, Romania
123
Endocrine
DOI 10.1007/s12020-014-0340-9
[7]. Ethnicity seems to influence the metabolic profile and
the predisposition to metabolic complications as suggested
by the high prevalence of hypertension and cardiovascular
complications among women of African descent, whereas
Hispanic women are more likely to present comorbidities
such as MetS and type 2 diabetes mellitus [8]. Few pre-
vious studies [9–11] highlighted the contribution of genetic
influences, lifestyle factors and cultural practices to the
variability of the metabolic and cardiovascular risk profiles
in PCOS. Therefore, characterizing the metabolic pheno-
type of a specific population is of great importance for
clinical practice.
The unifying factor between PCOS and MetS is
assumed to be insulin resistance (IR), a central patho-
genic mechanism in both conditions. Another possible
contributor is hyperandrogenism, although not all studies
agree on this aspect [12, 13]. Although IR is considered
an intrinsic characteristic of PCOS, with a different
pathogenic basis than the IR of obesity, the high preva-
lence of obesity and central adiposity among PCOS
patients could exacerbate the already existing IR. The
hyperinsulinism (HI) contributes to hyperandrogenism in
PCOS by stimulating ovarian androgen production and by
increasing androgen bioavailability through decreased
hepatic sex-hormone-binding globulin (SHBG) produc-
tion. Increased adiposity could contribute directly to hy-
perandrogenism, independently of IR, through
overactivation of 17beta-hydroxysteroid dehydrogenase
and 5alfa-reductase in adipose tissue, which results in
production of the potent androgens testosterone and
dihydrotestosterone [14]. The increased androgens may
amplify both IR and visceral adiposity, as suggested by
clinical and in vitro studies [15–18]. Since all these
factors presumably contributing to the high prevalence of
MetS in PCOS are subject to this complex interplay, it is
difficult to interpret their individual contributions to MetS
pathogenesis in PCOS.
The aims of our study were as follows: (1) to establish
whether Romanian PCOS patients have an increased
prevalence of MetS (previously not published) and (2) to
study the involvement of adiposity, IR/HI and hyperan-
drogenism in the pathogenesis of MetS in PCOS.
Due to assumed close relationship between adiposity,
IR/HI and hyperandrogenism, we decided to submit the
variables related to these parameters to principal compo-
nent analysis (PCA): a statistical method able to analyze
the relation between variables and to reduce the number of
correlated variables which capture the same information by
grouping them in a smaller number of uncorrelated vari-
ables called principal components (PCs). One of the major
advantages of PCA is that all the correlated variables can
be included in the analysis without the limitations due to
collinearity.
Subjects and methods
The study was performed in the academic setting of the
Endocrinology Department of Elias University Emergency
Hospital and C. I. Parhon National Institute of Endocri-
nology, both in Bucharest Romania. The study protocol
was approved by the local ethics committee. All subjects
gave written informed consent before evaluation. PCOS
subjects included in the study were recruited from PCOS
patients consecutively admitted in our departments
between January 2006 and December 2012. PCOS patients
were women referred by a wide range of specialist: der-
matologists, gynecologists, nutritionists, primary care
physicians. The presenting complaints were menstrual
irregularities, hirsutism, acne and infertility. Controls were
women with regular menstrual cycles and no clinical or
biochemical hyperandrogenism recruited during routine
evaluation for thyroid dysfunction (proven euthyroid),
obesity, fertile women who presented in the Gynecology
Department of Elias Hospital for routine gynecological
exam and hospital employees at periodical checkup. PCOS
was diagnosed according to the Rotterdam Consensus cri-
teria [19], i.e., two out of the following three criteria: (1)
chronic oligo/anovulation, (2) clinical and/or biochemical
hyperandrogenism and (3) polycystic appearance of the
ovaries on ultrasound and exclusion of hyperprolactinemia,
hypercortisolism, thyroid dysfunction, hypogonadism and
hyperandrogenism of other causes. PCOS and control
subjects had to be off oral contraceptives, insulin sensi-
tizers or antiandrogens for at least 3 months before eval-
uation, without any treatment that could significantly affect
hirsutism (cosmetic treatment or oral contraceptives/anti-
androgens for more than three months) and without any
significant associated disease. Only postmenarcheal and
premenopausal non-pregnant women were included in the
study.
Clinical evaluation
For all subjects, we recorded age, weight, height, waist
(WC) and hip circumference (HC) and menstrual history.
Weight was measured in light clothing without shoes and
height was measured using a stadiometer and the body
mass index (BMI) was calculated. Waist and hip circum-
ferences were measured with the patient standing, at the
level of umbilicus and pubic symphysis, respectively.
Blood pressure was measured with the patient seated after
at least 10 min of rest.
Clinical hyperandrogenism was considered in the pre-
sence of hirsutism, defined as a modified Ferriman–Gall-
wey (FG) score of at least 8 [20], evaluated by the
investigators through systematic use of the FG maps.
Chronic ovulatory dysfunction was diagnosed in the
Endocrine
123
presence of menstrual irregularities: oligomenorrhea
(menstrual cycles of 35–90 days) or amenorrhea (absence
of menses for more than 90 days). In patients with regular
menstrual cycles, chronic anovulation was considered if
absence of ovulation was documented by ultrasound and/or
mid-luteal phase serum progesterone in two consecutive
cycles.
Laboratory evaluation
Blood samples were collected after an overnight fast,
between days 3 and 5 of a spontaneous or progesterone-
induced menstrual cycle. Total testosterone (TT), SHBG,
thyroid stimulating hormone (TSH), glucose, insulin, total
cholesterol, high-density lipoprotein (HDL) cholesterol,
LDL cholesterol and triglycerides were measured in all
subjects (PCOS and controls). Luteinizing hormone (LH),
follicle-stimulating hormone (FSH), estradiol, prolactin
and oral glucose tolerance test with 75 g of glucose and
measurements of glucose and insulin fasting and 2 h after
glucose ingestion were performed only in PCOS patients.
Morning serum 17-hydroxyprogesterone and cortisol after
an overnight suppression test were evaluated only in
selected patients with a high clinical suspicion index.
Biochemical hyperandrogenism was defined in the pre-
sence of TT C 0.73 ng/mL and/or a free androgen index
(FAI) C 4.5. These cutoff values were determined as the
95th percentile of a group of 150 reproductive age women
without hirsutism/acne and with regular menstrual cycles,
from the same population (unpublished data).
HOMA-IR was calculated as [fasting glycemia (mg/
dL) 9 fasting insulinemia (lUI/mL)]/405. FAI was
obtained by dividing total testosterone (nmol/L) 9 100 to
SHBG (nmol/L).
The hormonal blood tests were performed using a
chemiluminescent immunometric assay (Immulite 2000,
Siemens Healthcare Diagnostics Products Ltd.).
Metabolic syndrome evaluation
Metabolic syndrome diagnosis was based on the National
Cholesterol Education Program, Adult Treatment Panel III
(NCEP ATP III) criteria (at least 3 of the following):
WC [ 88 cm, serum triglycerides C150 mg/dL, HDL
cholesterol \50 mg/dL, systolic blood pressure C130 and/
or diastolic blood pressure C85 mm Hg, serum glucose
C110 mg/dL [21].
Statistical analysis
Statistical analysis was performed using SPSS version 20.0
(SPSS Inc., Chicago, IL, USA). Results are reported as
medians and interquartile range or percentages as
appropriate. In order to test the normality of the distribu-
tion of variables Kolmogorov–Smirnov test was used. For
comparisons between groups, we used the Mann–Whitney
U test for numerical variables and chi-square test for cat-
egorical variables. Binary logistic regression was per-
formed in order to adjust for age or adiposity markers when
analyzing parameters associated with MetS.
Eleven parameters related to adiposity, HI/IR and hy-
perandrogenism were submitted to PCA.
PCA is a statistical method useful for both data reduc-
tion and structure detection. The purpose of data reduction
is to identify from a panel of data the variables which are
highly correlated and to reduce the entire data set to a
smaller number of uncorrelated variables PCs. PCA is also
able to perform structure detection, meaning to examine
the underlying relationship between variables. There are
many advantages when using PCA. For example, it could
be difficult to decide which marker of adiposity to choose
to investigate the link between MetS and adiposity,
knowing the differences in metabolic activity of specific
type of adiposity distribution. Using PCA, all the markers
of adiposity can be submitted to analysis, grouped in PCs
and then the PCs can be introduced in a model of multi-
variate regression to study their association with MetS. In
this way, we are able to take into consideration all the
variables representing adiposity, avoiding in the same time
collinearity. Moreover, PCA can identify the marker which
best represent the group by having the highest correlation
coefficient in this particular PC and weak correlation
coefficients with the other PCs, but also the markers which
represent a ‘‘bridge’’ between PCs by having high corre-
lation coefficients with many PCs.
Parameters with non-Gaussian distribution were log-
transformed before PCA. All factors with eigenvalues
exceeding one were retained for analysis. Varimax rotation
was applied in order to facilitate the PCs interpretation.
PCs loadings were interpreted as follow: between 0.3 and
0.5 demonstrated small correlations, and loadings above
0.5 identified strong correlations. The PCs scores were
saved and used to analyze their relationship with MetS in
binary logistic regression. p values \0.05 were considered
indicative of statistical significance.
Results
Clinical and paraclinical parameters of the study
subjects
We observed that the PCOS patients (n = 398) were sig-
nificantly younger than controls (n = 126) (p \ 0.001).
Age ranged between 17 and 40 years in PCOS subjects and
between 17 and 42 years in controls. BMI ranged between
Endocrine
123
18.5 and 45 kg/m2 in PCOS study group and between 19
and 49 kg/m2 in controls. The two groups were similar in
terms of BMI, waist and hip circumferences, but PCOS
subjects had significantly higher waist–hip ratio (WHR)
(Table 1).
Prevalence of the metabolic syndrome and individual
components
The prevalence of MetS was significantly higher among
patients with PCOS (20.4 %, n = 81/398) than in controls
(11.1 %, n = 14/126, p = 0.019). In PCOS patients,
among the individual criteria used to define MetS, low
HDL cholesterol was the most prevalent (50.33 %), fol-
lowed by increased waist circumference (43.5 %), high
blood pressure (19.25 %), hypertriglyceridemia (18.3 %)
and high blood glucose (4.53 %). In controls, low-HDL
cholesterol and WC had very close prevalences (40.27 %
and 41.46 %, respectively), followed in the same order by
high blood pressure (17.2 %), high triglycerides levels
(6.45 %) and high blood glucose (3.36 %).
After adjustment for age using binary logistic regression
with MetS as dependent variable and age and diagnosis
(PCOS/controls) as independent variables, we found that
PCOS patients still had an increased risk of MetS (OR
3.36, CI 1.71–6.5, p \ 0.0001). However, after including
WHR in the logistic regression model, PCOS was no
longer significantly associated with MetS, although age and
WHR continued to be independent predictors of MetS.
Comparison between PCOS patients
with and without MetS
In PCOS patients, the presence of MetS was associated
with increased BMI, WC, HC, WHR and higher insuline-
mia (fasting and 2 h) and insulin resistance (HOMA-IR)
(Table 2). Patients with both PCOS and MetS also had
more severe hyperandrogenism, both clinically (signifi-
cantly higher FG score of hirsutism) and biochemically, as
reflected by the FAI levels, but not by TT levels (Table 2).
Principal component analysis
Eleven variables representing adiposity (BMI, WHR, WC,
HC), insulinemia (fasting and 2 h insulin), insulin resis-
tance (HOMA-IR) and hyperandrogenism (TT, SHBG,
FAI, hirsutism) were used for PCA and three PCs were
extracted, accounting for 74.6 % of the total variance
(Table 3). First component (‘‘adiposity and insulin resis-
tance’’) aggregated variables related to adiposity, insuli-
nemia and insulin resistance and explained 41.6 % of the
Table 1 Clinical and biochemical data in PCOS and controls
PCOS
(n = 398)
Controls
(n = 126)
p
Age (years) 24 (7) 28 (10) \0.0001
BMI (kg/m2) 26.1 (10.9) 25.61 (12) NS
WC (cm) 83 (27.6) 80 (27.26) NS
HC (cm) 103(20) 107 (23) NS
WHR 0.81 (0.13) 0.77 (0.12) \0.0001
Fasting glycemia
(mg/dL)
84 (12) 84 (15) NS
2 h glycemia (mg/dL) 99 (32.7) NA
Fasting insulin
(lUI/mL)
11.53 (12.25) 7.42 (6.32) \0.0001
2 h insulin (lUI/mL) 54.9 (64.6) NA
HOMA-IR 2.41 (2.83) 1.64 (1.52) \0.0001
Total testosterone
(ng/mL)
0.72 (0.4) 0.45 (0.3) \0.0001
FAI 6.32 (6.81) 1.69 (2.64) \0.0001
SHBG (nmol/L) 39.69 (33.3) 79.69 (76.22) \0.0001
Data are expressed as median and interquartile range. For compari-
sons between groups, Mann–Whitney U est was used
BMI body mass index; WC waist circumference; HC hip circumfer-
ence; WHR waist–hip ratio; HOMA-IR homeostasis model of
assessment of insulin resistance; FAI free androgen index; SHBG sex-
hormone-binding globulin; NS not significant. NA not applicable
Table 2 Clinical and biochemical parameters in PCOS patients with
and without MetS
PCOS with
MetS (n = 81)
PCOS without
MetS (n = 317)
p
Age (yrs) 27 (11) 23 (7) \0.0001
Hirsutism
(FG score)
11 (10) 9 (7) 0.035
BMI (kg/m2) 34 (8.38) 23 (9.1) \0.0001
HC (cm) 116 (16.7) 100 (16) \0.0001
WHR 0.9 (0.09) 0.78 (0.11) \0.0001
Fasting insulin
(lUI/mL)
22.5 (18) 10.27 (9.52) \0.0001
2 h insulin
(lUI/mL)
113.4 (126) 49 (49.1) \0.0001
HOMA-IR 5.38 (4.8) 2.15 (2.03) \0.0001
Total
testosterone
(ng/mL)
0.76 (0.42) 0.72 (0.4) NS
FAI 10.26 (11.8) 5.64 (6.1) \0.0001
SHBG (nmol/L) 27.2 (21.5) 44.11 (40) \0.0001
Data are expressed as medians and interquartile range. For compari-
sons between groups, Mann–Whitney U test was used
BMI body mass index; WC waist circumference; HC hip circumfer-
ence; WHR waist–hip ratio; HOMA-IR homeostasis model of
assessment of insulin resistance; FAI free androgen index; SHBG sex-
hormone-binding globulin; HDL high-density lipoprotein; FG Ferri-
man–Gallway; NS not significant; NA not applicable
Endocrine
123
total variance. Factor loadings showed significant rela-
tionship between variables representing adiposity and IR/
HI. First PC was best represented by WC which had the
highest correlation coefficient (0.912) and did not correlate
with the other PCs. The second PC (‘‘biochemical hyper-
androgenism’’) explained 19.8 % of total variance and
included markers of hyperandrogenemia (TT, FAI) and
SHBG which correlated strongly with this PC. None of the
other parameters were correlated with the second PC,
except for the 2 h insulinemia. FAI was the main repre-
sentative of this PC due to highest correlation coefficients
with the second PC and weak correlation coefficients with
the other PCs. TT behaved like a bridge between compo-
nents having small loading with the first and the third PCs.
The third component (‘‘clinical hyperandrogenism,’’
explaining 13.2 % of the total variance) was best repre-
sented by hirsutism which correlated exclusively with this
PC, but also gathered adiposity markers (BMI and HC),
TT, SHBG which all had weak correlation coefficients and
explained 13.2 % of the total variance.
Logistic regression analysis
In order to avoid influences of latent factors and collin-
earity when studying the association between clinical and
biochemical parameters and MetS in PCOS patients, we
used logistic regression to construct a model with MetS as
dependent variable and age and the three PCs identified by
PCA as independent predictors. After adjustment for age,
the first PC, representing adiposity and insulin resistance,
and the second PC, representing hyperandrogenemia, were
independently associated with MetS (p \ 0.0001 and
p = 0.003, respectively), whereas the third PC represent-
ing clinical hyperandrogenism was not significantly related
to MetS (Table 4).
Discussion
The present study is to our knowledge the first specifically
assessing the association of MetS and PCOS in Romanian
patients and detailing the data structure by means of PCA.
The prevalence of MetS was similar to that reported by
previous studies of European PCOS populations [22, 23],
but visibly lower than in the US and Australian studies [24,
25]. The increased WHR seems to be responsible for the
high prevalence of MetS in our population of PCOS
patients, as the significant association between MetS and
PCOS was lost after adjustment for WHR in binary logistic
regression analysis. WHR was demonstrated to have a
good correlation with visceral adiposity measured by
Table 3 Loading matrix of the three PCs extracted by PCA in PCOS patients
First PC (41.6 %*) Second PC (19.8 %*) Third PC (13.2 %*)
‘‘Adiposity and
insulin resistance’’
‘‘Biochemical
hyperandrogenism’’
‘‘Clinical
hyperandrogenism’’
BMI 0.821 0.192 0.387
WHR 0.738 0.055 -0.047
WC 0.912 0.121 0.267
HC 0.778 0.125 0.418
Fasting insulin 0.832 0.222 -0.081
2 h insulin 0.604 0.303 -0.133
HOMA-IR 0.837 0.204 -0.072
TT 0.311 0.650 20.379
SHBG -0.071 20.768 20.468
FAI 0.240 0.950 0.163
Hirsutism (FG score) 0.069 0.100 0.801
The eleven variables showed in the table were submitted to PCA and the three PCs were extracted. Significant correlation coefficients ([0.3
showing small and [0.5 showing strong correlations) are in bold. * Percentage of the total variance explained by the corresponding PC
PC principal component; PCA principal component analysis; BMI body mass index; WHR waist–hip ratio; WC waist circumference; HC hip
circumference; HOMA-IR homeostasis assessment of insulin resistance; TT total testosterone; SHBG sex-hormone-binding globulin; FAI free
androgen index; FG Ferriman–Gallwey
Table 4 Logistic regression analysis with MetS as dependent vari-
able and age and the three PCs as independent variables
Coef SE OR 95 %CI of OR p
Age 0.042 0.059 1.04 0.9–1.17 NS
First PC 1.519 0.352 4.56 2.29–9.09 \0.0001
Second PC 0.631 0.225 1.87 1.2–2.92 0.003
Third PC 0.293 0.315 1.34 0.7–1.17 NS
PC principal component; Coef regression coefficient; SE standard
error; OR odds ratio; CI confidence interval; NS not significant
Endocrine
123
computed tomography in women [26]; therefore, we could
assume a role for visceral adiposity in increased prevalence
of MetS in Romanian PCOS patients. This finding is in
accord with recent publications highlighting the impor-
tance of central adiposity, rather than global adiposity in
the vicious circle of adiposity–insulin resistance–hyperan-
drogenism, as mediator of the metabolic effects in PCOS
[22]. Although many studies reported an increase in central
adiposity characterizing PCOS patients, this may not be
true for all PCOS populations, as suggested by a well-
designed study showing that PCOS and controls matched
for BMI and fat mass had similar distribution of adiposity
[27]. Indeed, the great metabolic heterogeneity of PCOS
patients is further supported by metabolomics studies
showing that hyperinsulinemia of lean patients with PCOS
is not necessarily secondary to insulin resistance [28].
Interestingly, all the markers of adiposity and HI/IR
aggregated in the same PC suggesting the strong relation-
ship between these two important contributors to PCOS
pathogenesis in our population. This PC explained 41.6 %
of the total variance of the data set, highlighting the
importance of metabolic component in PCOS pathogene-
sis. Similar data were published by Dewailly et al. [29]
who concluded that IR/HI of PCOS is just a state associ-
ated with obesity, but with a higher prevalence compared
with general population. Another recently published report
studying PCOS patients attending an in vitro fertilization
department demonstrated increased prevalence of IR only
in overweight/obese PCOS comparing to controls, but not
in lean PCOS [30]. These results challenge previous studies
reporting that insulin resistance of PCOS is intrinsic to the
syndrome and has a different mechanism of that of obesity.
Moreover, the PCOS patients in our study were not
severely obese (median BMI 26.1 kg/m2); therefore, we
can assume that regional adiposity rather than global adi-
posity is mainly involved in HI/IR in our group. Indeed, the
first PC was best represented by WC, a marker of central
adiposity which correlated strongly and exclusively with
this PC. The positive correlation of 2 h insulinemia with
the second PC challenge a previous report showing that
transient hyperinsulinemia produced by an oral glucose
load produce a decrease in circulating androgen levels [31].
The second PC gathered the androgenic variables: TT,
FAI and SHBG and was best represented by FAI which
correlated strongly and exclusively with this PC. It was
somehow surprising to see FAI and SHBG were not cor-
related with the first PC, knowing the great influence of IR
on SHBG hepatic production. A possible explanation not
tested yet in Romanian patients is the genetic influence on
SHBG levels and consecutively on FAI level reported by
other authors [32]. Therefore, FAI seems to be a specific
marker of hyperandrogenism, without the influence from
the metabolic component, at least in our patients. This
finding is contrasting with a previous report [29] showing
FAI has strong correlation with metabolic status and is not
specific for PCOS. Although TT loaded mainly on the
second PC, it had also weak correlation with the first and
third PC, unraveling the possible bidirectional link between
TT level and metabolic parameters. Interestingly, TT cor-
related weakly and negatively with the third PC mainly
represented by hirsutism. A possible explanation is the
observation that there is an inverse relationship between
clinical and biochemical hyperandrogenism with increas-
ing age, meaning that hirsutism aggravates, whereas hy-
perandrogenemia diminishes [33]. Hirsutism did not
aggregate with the biochemical markers of hyperandroge-
nism, indicating it is not a good surrogate of circulating
androgens. Moreover, it was not linked with the metabolic
component, possibly reflecting other influences. For
example, variants of the 5alpha-reductase type 1 and type 2
genes were shown to be associated with the severity of
hirsutism in PCOS women [34]. Therefore, the presence of
hirsutism and hyperandrogenemia is not equivalent, and
this type of information could be important for clinical
studies of PCOS patients, especially when taking into
consideration the metabolic end points. Indeed, previous
studies suggested that the exclusive presence of hirsutism
in PCOS patients is associated with milder metabolic dis-
turbances compared to those with hyperandrogenemia [35].
Logistic regression analysis confirmed this information
showing that the third PC represented by hirsutism was not
independently associated with the presence of MetS in
PCOS patients. Instead, the first PC representing adiposity
and IR/HI and the second PC representing hyperandro-
genemia were independently related with MetS.
The relationship between circulating androgens levels
and MetS in women with PCOS continues to be intensely
debated. A recently published meta-analysis [36] showed
that higher total and free testosterone are associated with
increased prevalence of MetS in PCOS patients (although
the link was weaker than in non-PCOS women). Never-
theless, those studies addressing the role of hyperandrog-
enism in the MetS risk in PCOS reported divergent data,
with some studies founding a positive association [12, 24,
25], whereas other authors reported similar androgens
levels in patients with and without MetS [13]. In fact, few
of these studies took into account the potential confounding
effect of adiposity on the relationship between MetS and
androgen levels, making it difficult to interpret the results.
Although not addressing directly the influences of andro-
gens levels on MetS, but rather on MetS components, the
study of Zanolin et al. [37] suggested also an independent
relationship between free testosterone and at least some
metabolic risk factors (arterial blood pressure). In our
study, the link between circulating androgens and MetS
independent of adiposity, insulin resistance and age is
Endocrine
123
clearly demonstrated using the exact mathematical tool of
PCA. Moreover, we included in our analysis all anthro-
pometric indices representing both global and abdominal
adiposity, trying to take into consideration the complex
relation between androgens and body fat.
The main reason for using the NCEP ATP III definition
for MetS in our study was to allow comparison with previous
studies on MetS in PCOS patients since most of them used
NCEP ATP III criteria. Moreover, there may be additional
benefits as suggested by existing reports showing that dif-
ferent definitions of MetS identify distinct categories of
patients in terms of cardio-metabolic risk. Therefore, it is
possible that using NCEP ATP III criteria to achieve a better
selection of patients with increased risk of cardiovascular
complication, as compared to the IDF and EGIR criteria [38].
In turn, the predictive value for incident diabetes seems to be
similar for IDF, NCEP ATP III and WHO definitions of MetS
as suggested by recent studies [39, 40].
One of the limitations of our study is the fact that the
PCOS and control groups were not matched for age.
However, we consider this did not have an impact on our
analysis taking into consideration that generally the prev-
alence of MetS is increasing with age, still the PCOS
patients had higher prevalence of MetS despite being
younger. Moreover, the association between MetS and
PCOS was maintained after adjustment for age.
In this study, we did not systematically measure serum
androstenedione and dehydroepiandrosterone sulfate
(DHEAS) levels. Nevertheless, the impact of this is rela-
tively limited given the fact that these two androgens add
little to the PCOS diagnosis, as only 10 % of patients with
biochemical hyperandrogenism have isolated increased
levels of androstenedione or DHEAS [41]. At the same
time, the lack of complete circulating androgen profile did
not allowed us to analyze the metabolic correlations of
DHEAS in our population, and this could be an interesting
aspect since previous studies suggested an association of
DHEAS with a more favorable metabolic profile [42].
In conclusion, our study demonstrated that Romanian
PCOS patients have an increased risk for cardio-metabolic
complications and unraveled some particularities of our
population, namely the increased adiposity as the main
contributor to the association of the two conditions and HI/
IR more likely as a satellite phenomenon of adiposity, than
an intrinsic characteristic of the disease. At the same time,
hyperandrogenemia was found to be an independent con-
tributor to the high prevalence of MetS in PCOS patients,
contrasting with hirsutism which was not involved in this
metabolic condition.
Acknowledgments We are grateful to all of the staff at both centers
and especially to Suzana Florea (Elias Hospital Laboratory) for
technical assistance. The study was partly funded by the CNCSIS
Grant No. 1333/2007 of the Romanian Ministry of Youth, Education
and Research.
Conflict of interest The authors declare that they have no conflict
of interest.
References
1. I.F. Stein, M.L. Leventhal, Amenorrhea associated with bilateral
polycystic ovaries. Am. J. Obstet. Gynecol. 29, 181–191 (1935)
2. P. Scicchitano, I. Dentamaro, R. Carbonara, G. Bulzis, A.
Dachille, P. Caputo, R. Riccardi, M. Locorotondo, C. Mandurino,
Ciccone MM, Cardiovascular risk in women with PCOS. Int.
J. Endocrinol. Metab. 10(4), 611–618 (2012)
3. L.J. Moran, M.L. Misso, R.A. Wild, R.J. Norman, Impaired
glucose tolerance, type 2 diabetes and metabolic syndrome in
polycystic ovary syndrome: a systematic review and meta-ana-
lysis. Hum. Reprod. Update 16(4), 347–363 (2010)
4. J. Vrbikova, K. Vondra, D. Cibula, K. Dvorakova, S. Stanicka, D.
Sramkova, G. Sindelka, M. Hill, B. Bendlova, J. Skrha, Meta-
bolic syndrome in young Czech women with polycystic ovary
syndrome. Hum. Reprod. 20(12), 3328–3332 (2005)
5. E. Carmina, N. Napoli, R.A. Longo, G.B. Rini, R.A. Lobo,
Metabolic syndrome in polycystic ovary syndrome (PCOS):
lower prevalence in southern Italy than in the USA and the
influence of criteria for the diagnosis of PCOS. Eur. J. Endocri-
nol. 154(1), 141–145 (2006)
6. C.J. Glueck, S. Dharashivkar, P. Wang, B. Zhu, P.S. Gartside, T.
Tracy, L. Sieve, Obesity and extreme obesity, manifest by ages
20–24 years, continuing through 32–41 years in women, should
alert physicians to the diagnostic likelihood of polycystic ovary
syndrome as a reversible underlying endocrinopathy. Eur.
J. Obstet. Gynecol. Reprod. Biol. 122, 206–212 (2005)
7. H.L. Ching, V. Burke, B.G. Stuckey, Quality of life and psy-
chological morbidity in women with polycystic ovary syndrome:
body mass index, age and the provision of patient information are
significant modifiers. Clin. Endocrinol. Oxf. 66, 373–379 (2007)
8. J.C. Lo, S.L. Feigenbaum, J. Yang, A.R. Pressman, J.V. Selby,
A.S. Go, Epidemiology and adverse cardiovascular risk profile of
diagnosed polycystic ovary syndrome. J. Clin. Endocrinol.
Metab. 91, 1357–1363 (2006)
9. N. Schweighofer, E. Lerchbaum, O. Trummer, V. Schwetz, S.
Pilz, T.R. Pieber, B. Obermayer-Pietsch, Androgen levels and
metabolic parameters are associated with a genetic variant of
F13A1 in women with polycystic ovary syndrome. Gene 504,
133–139 (2012)
10. E. Carmina, R.S. Legro, K. Stamets, J. Lowell, R.A. Lobo, Dif-
ference in body weight between American and Italian women
with polycystic ovary syndrome: influence of the diet. Hum.
Reprod. 18(11), 2289–2293 (2003)
11. S.F. Witchel, S.E. Recabarren, F. Gonzalez, E. Diamanti-Kand-
arakis, K.I. Cheang, A.J. Duleba, R.S. Legro, R. Homburg, R.
Pasquali, R. Lobo, C.C. Zouboulis, F. Kelestimur, F. Fruzzetti,
W. Futterweit, R.J. Norman, D.H. Abbott, Emerging concepts
about prenatal genesis, aberrant metabolism and treatment para-
digms in polycystic ovary syndrome. Endocrine 42(3), 526–534
(2012). doi:10.1007/s12020-012-9701-4
12. D.A. Ehrmann, D.R. Liljenquist, K. Kasza, R. Azziz, R.S. Legro,
M.N. Ghazzi, PCOS/Troglitazone Study Group, Prevalence and
predictors of the metabolic syndrome in women with polycystic
ovary syndrome. J. Clin. Endocrinol. Metab. 91, 48–53 (2006)
13. A. Dokras, M. Bochner, E. Hollinrake, S. Markham, B. Van-
voorhis, D.H. Jagasia, Screening women with polycystic ovary
Endocrine
123
syndrome for metabolic syndrome. Obstet. Gynecol. 106,
131–137 (2005)
14. A.M. Corbould, M.J. Bawden, T.C. Lavranos, R.J. Rodgers, S.J.
Judd, The effect of obesity on the ratio of type 3 17beta-
hydroxysteroid dehydrogenase mRNA to cytochrome P450 aro-
matase mRNA in subcutaneous abdominal and intra-abdominal
adipose tissue of women. Int. J. Obes. Relat. Metab. Disord.
26(2), 165–175 (2002)
15. J.C. Lovejoy, G.A. Bray, M.O. Bourgeois, R. Macchiavelli, J.C.
Rood, C. Greeson, C. Partington, Exogenous androgens influence
body composition and regional body fat distribution in obese
postmenopausal women—a clinical research center study. J. Clin.
Endocrinol. Metab. 81(6), 2198–2203 (1996)
16. A. Gambineri, C. Pelusi, S. Genghini, A.M. Morselli-Labate, M.
Cacciari, U. Pagotto, R. Pasquali, Effect of flutamide and met-
formin administered alone or in combination in dieting obese
women with polycystic ovary syndrome. Clin. Endocrinol. 60(2),
241–249 (2004)
17. M. Allemand, Y. Asmann, K. Klaus, K. Nair, An model for
PCOS related insulin resistance: the effects of testosterone on
phosphorylation of intracellular insulin signaling proteins in rat
skeletal muscle primary culture. Fertil. Steril. 84(Suppl 1), S30–
S31 (2005)
18. A. Corbould, Chronic testosterone treatment induces selective
insulin resistance in subcutaneous adipocytes of women.
J. Endocrinol. 192(3), 585–594 (2007)
19. Rotterdam ESHRE/ASRM-Sponsored Consensus Workshop
Group, Revised 2003 consensus on diagnostic criteria and long-
term health risks related to polycystic ovary syndrome (PCOS).
Hum. Reprod. 19, 41–47 (2004)
20. H.F. Escobar-Morreale, E. Carmina, D. Dewailly, A. Gambineri,
F. Kelestimur, P. Moghetti, M. Pugeat, J. Qiao, C.N. Wijeyaratne,
S.F. Witchel, R.J. Norman, Epidemiology, diagnosis and man-
agement of hirsutism: a consensus statement by the Androgen
Excess and Polycystic Ovary Syndrome Society. Hum Reprod
Update. 18(2), 146–170 (2012)
21. Expert panel on detection, evaluation, and treatment of high
blood cholesterol in adults: Executive summary of the third report
of the National Cholesterol Education Program (NCEP) expert
panel on detection, evaluation, and treatment of high blood
cholesterol in adults (Adult Treatment Panel III). J Am Med
Asoc. 285(19), 2486–2497 (2001)
22. D. Panidis, D. Macut, K. Tziomalos, E. Papadakis, K. Mikhaili-
dis, E.A. Kandaraki, E.A. Tsourdi, T. Tantanasis, G. Mavro-
matidis, I. Katsikis, Prevalence of metabolic syndrome in women
with polycystic ovary syndrome. Clin Endocrinol (Oxf). 78(4),
586–592 (2013)
23. J.J. Espinos-Gomez, J. Rodriguez-Espinosa, J. Ordonez-Llanos, J.
Calaf-Alsina, Metabolic syndrome in Mediterranean women with
polycystic ovary syndrome: when and how to predict its onset.
Gynecol. Endocrinol. 28(4), 264–268 (2011)
24. T. Apridonidze, P.A. Essah, M.J. Iuorno, J.E. Nestler, Prevalence
and characteristics of the metabolic syndrome in women with
polycystic ovary syndrome. J. Clin. Endocrinol. Metab. 90,
1929–1935 (2005)
25. A.J. Cussons, G.F. Watts, V. Burke, J.E. Shaw, P.Z. Zimmet,
B.G. Stuckey, Cardiometabolic risk in polycystic ovary syn-
drome: a comparison of different approaches to defining the
metabolic syndrome. Hum. Reprod. 23, 2352–2358 (2008)
26. D. Berker, S. Koparal, S. Isik, L. Pasaoglu, Y. Aydin, K. Erol, T.
Delibasi, S. Guler, Compatibility of different methods for the
measurement of visceral fat in different body mass index strata.
Diagn Interv Radiol. 16(2), 99–105 (2010)
27. T.M. Barber, S.J. Golding, C. Alvey, J.A. Wass, F. Karpe, S.
Franks, M.I. McCarthy, Global Adiposity Rather Than Abnormal
Regional Fat Distribution Characterizes Women with Polycystic
Ovary Syndrome. J. Clin. Endocrinol. Metab. 93, 999–1004
(2008)
28. H.F. Escobar-Morreale, S. Samino, M. Insenser, M. Vinaixa, M.
Luque-Ramırez, M.A. Lasuncion, X. Correig, Metabolic hetero-
geneity in polycystic ovary syndrome is determined by obesity:
plasma metabolomic approach using GC-MS. Clin. Chem. 58(6),
999–1009 (2009)
29. D. Dewailly, P. Pigny, B. Soudan, S. Catteau-Jonard, C. Decan-
ter, E. Poncelet, A. Duhamel, Reconciling the Definitions of
Polycystic Ovary Syndrome: The Ovarian Follicle Number and
Serum Anti-Mullerian Hormone Concentrations Aggregate with
the Markers of Hyperandrogenism. J. Clin. Endocrinol. Metab.
95, 4399–4405 (2010)
30. Alebic, M.S., Bulum, T., Stojanovic, N., Duvnjak, L.: Definition
of insulin resistance using the homeostasis model assessment
(HOMA-IR) in IVF patients diagnosed with polycystic ovary
syndrome (PCOS) according to the Rotterdam criteria. Endo-
crine. Feb 13. (2014) [Epub ahead of print]
31. J.H. Fox, T. Licholai, G. Green, A. Dunaif, Differential effects of
oral glucose-mediated versus intravenous hyperinsulinemia on
circulating androgen levels in women. Fertil. Steril. 60(6),
994–1000 (1993)
32. E.P. Wickham, K.G. Ewens, R.S. Legro, A. Dunaif, J.E. Nestler,
J.F. Strauss 3rd, Polymorphisms in the SHBG gene influence
serum SHBG levels in women with polycystic ovary syndrome.
J. Clin. Endocrinol. Metab. 96(4), 719–727 (2011)
33. D. Glintborg, H. Mumm, P. Ravn, M. Andersen, Age associated
differences in prevalence of individual Rotterdam criteria and
metabolic risk factors during reproductive age in 446 caucasian
women with polycystic ovary syndrome. Horm. Metab. Res.
44(9), 694–698 (2012)
34. M.O. Goodarzi, N.A. Shah, H.J. Antoine, M. Pall, X. Guo, R.
Azziz, Variants in the 5alpha-reductase type 1 and type 2 genes
are associated with polycystic ovary syndrome and the severity of
hirsutism in affected women. J. Clin. Endocrinol. Metab. 91(10),
4085–4091 (2006)
35. W.Y. Chang, E.S. Knochenhauer, A.A. Bartolucci, R. Azziz,
Phenotypic spectrum of polycystic ovary syndrome: clinical and
biochemical characterization of the three major clinical sub-
groups. Fertil. Steril. 83(6), 1717–1723 (2005)
36. J.S. Brand, I. van der Tweel, D.E. Grobbee, M.H. Emmelot-
Vonk, Y.T. van der Schouw, Testosterone, sex hormone-binding
globulin and the metabolic syndrome: a systematic review and
meta-analysis of observational studies. Int. J. Epidemiol. 40,
189–207 (2011)
37. M.E. Zanolin, F. Tosi, G. Zoppini, R. Castello, G. Spiazzi, R.
Dorizzi, M. Muggeo, P. Moghetti, Clustering of cardiovascular
risk factors associated with the insulin resistance syndrome:
assessment by principal component analysis in young hyperan-
drogenic women. Diabetes Care 29, 372–378 (2006)
38. P.M. Nilsson, G. Engstrom, B. Hedblad, The metabolic syndrome
and incidence of cardiovascular disease in non-diabetic sub-
jects—a population-based study comparing three different defi-
nitions. Diabet. Med. 24, 464–472 (2007)
39. C. Lorenzo, M. Okoloise, K. Williams, M.P. Stern, S.M. Haffner,
The metabolic syndrome as predictor of type 2 diabetes: the San
Antonio heart study. Diabetes Care 26, 3153–3159 (2003)
40. A.J. Hanley, A.J. Karter, K. Williams, A. Festa, R.B. D’Agostino
Jr, L.E. Wagenknecht, S.M. Haffner, Prediction of type 2 diabetes
mellitus with alternative definitions of the metabolic syndrome:
the insulin resistance atherosclerosis study. Circulation 112,
3713–3721 (2005)
41. R. Azziz, E. Carmina, D. Dewailly, E. Diamanti-Kandarakis, H.F.
Escobar-Morreale, W. Futterweit, O.E. Janssen, R.S. Legro, R.J.
Norman, A.E. Taylor, S.F. Witchel, Task Force on the Phenotype
of the Polycystic Ovary Syndrome of The Androgen Excess and
Endocrine
123
PCOS Society: The Androgen Excess and PCOS Society criteria
for the polycystic ovary syndrome: the complete task force report.
Fertil. Steril. 91, 456–488 (2009)
42. E. Lerchbaum, V. Schwetz, A. Giuliani, T.R. Pieber, B. Obe-
rmayer-Pietsch, Opposing effects of dehydroepiandrosterone
sulfate and free testosterone on metabolic phenotype in women
with polycystic ovary syndrome. Fertil. Steril. 98, 1318–1325
(2012)
Endocrine
123