Human Reproduction Update, Vol.15, No.3 pp.

297– 307, 2009

Advanced Access publication on March 4, 2009 doi:10.1093/humupd/dmp006

Adiponectin levels in women with

polycystic ovary syndrome: a
systematic review and a meta-analysis
K.A. Toulis 1,5, D.G. Goulis 1, D. Farmakiotis2, N.A. Georgopoulos3,
I. Katsikis 2, B.C. Tarlatzis 4, I. Papadimas 1, and D. Panidis2
1
First Department of Obstetrics and Gynecology, Unit of Reproductive Endocrinology, Papageorgiou General Hospital, Aristotle University,
Ring Road, Nea Efkarpia, 56403 Thessaloniki, Greece 2Second Department of Obstetrics and Gynecology, Unit of Reproductive
Endocrinology and Human Reproduction, Aristotle University, Thessaloniki, Greece 3Division of Reproductive Endocrinology, Department
of Obstetrics and Gynecology, University of Patras Medical School, Patras, Greece 4First Department of Obstetrics and Gynecology,

Downloaded from http://humupd.oxfordjournals.org/ by guest on March 27, 2015

Unit for Human Reproduction, Aristotle University, Thessaloniki, Greece
5
Correspondence address. Tel: þ30 2310-693131; Fax: þ30 2310-991510; E-mail:touliskos@gmail.com

table of contents
...........................................................................................................................
† Introduction
† Methods
† Results
Search results
Meta-analysis
Categorical analyses
Meta-regression
‘Nested’ meta analysis on BMI
† Discussion

background: Conflicting results regarding adiponectin levels in women with polycystic ovary syndrome (PCOS) have been reported.
To evaluate adiponectin levels in PCOS, a systematic review of all studies comparing adiponectin levels in women with PCOS with healthy
controls and a meta-analysis of those involving women with similar body mass index (BMI) were performed. The influence of possible effect
modifiers, such as insulin resistance (IR) and testosterone, was investigated. The influence of obesity was investigated through a ‘nested’
meta-analysis after within-study BMI stratification and appropriate pooling.
methods: Literature search was conducted through MEDLINE, EMBASE, Cochrane CENTRAL (through June 2008), references from
relevant studies and personal contact with the authors. Thirty-one studies, reporting data on 3469 subjects, were reviewed and 16 included
in the main meta-analysis.
results: Women with PCOS demonstrated significantly lower adiponectin values [weighted mean difference (95% confidence interval)
21.71 (22.82 to 20.6), P , 1024], yet with significant between-study heterogeneity. Lower adiponectin levels are associated with the IR
observed in women with PCOS, compared with controls. IR, but not total testosterone, was found significant among biological parameters
explored in the meta-regression model. Hypoadiponectinaemia was present in both lean and obese women with PCOS when compared with
non-PCOS counterparts. Data on high molecular weight (HMW) adiponectin are limited (three studies).
conclusions: After controlling for BMI-related effects, adiponectin levels seem to be lower in women with PCOS compared with non-
PCOS controls. Low levels of adiponectin in PCOS are probably related to IR but not to testosterone. Total adiponectin should not be used
as a biomarker of PCOS severity. Further investigation is needed for HMW adiponectin levels in PCOS.
Key words: polycystic ovary syndrome / adiponectin / insulin resistance / testosterone / high molecular weight adiponectin

& The Author 2009. Published by Oxford University Press on behalf of the European Society of Human Reproduction and Embryology. All rights reserved.
For Permissions, please email: journals.permissions@oxfordjournals.org
298
Introduction
The polycystic ovary syndrome (PCOS) is a common, heterogeneous
endocrine disorder in women of reproductive age. It is characterized
by a clustering of hyperandrogenism (either clinical or biochemical),
chronic anovulation and polycystic ovaries (Franks, 1995) and it is fre-
quently associated with insulin resistance (IR) and obesity (Ehrmann,
2005). Notwithstanding the known effects of excess adiposity on
insulin sensitivity, evidence suggests that women with PCOS demonstrate
an intrinsic form of IR that is unique to the disorder (Morales et al., 1996).
A possible link between adipose tissue and insulin sensitivity is pro-
vided through the mediation of adipokines, products of adipose
secretory activity. Among them, adiponectin, a protein produced
almost exclusively by adipocytes, is considered to exert insulin-
sensitizing, anti-atherogenic and anti-inflammatory actions (Weyer
et al., 2001; Kadowaki and Yamauchi, 2005). It has been suggested
that adiponectin expression is down-regulated by obesity (Stefan
and Stumvoll, 2002), whereas its plasma levels have been found
reduced in a variety of states frequently associated with IR, such as dia-
betes, cardiovascular disease and hypertension (Matsubara et al.,
2002; Trujillo and Scherer, 2005). Insulin-resistant states have also
been associated with a reduction in high molecular weight (HMW)
fraction of adiponectin (Hara et al., 2006; Lara-Castro et al., 2006),
which is considered a potent mediator of insulin sensitivity (Pajvani
et al., 2004). Furthermore, it has been reported that percentage com-
position of HMW adiponectin was decreased by testosterone in both
animal (mice) and human (hypogonadal men after the testosterone
replacement therapy) models (Xu et al., 2005).
Given the effects of excess adiposity and IR on adiponectin levels, a
marked decrease of adiponectin levels in women with PCOS could be
assumed. This assumption has been confirmed by the findings of
numerous studies (Ducluzeau et al., 2003; Panidis et al., 2004; Sie-
minska et al., 2004; Spranger et al., 2004; Ardawi and Rouzi, 2005;
Carmina et al., 2005, 2006, 2008a, b; Sepilian and Nagamani 2005;
Escobar-Morreale et al., 2006; Beckman et al., 2007; Aroda et al.,
2008; Glintborg et al., 2008; Thomann et al., 2008; Zhang et al.,
2008); however, this was not the case in several other studies,
where no significant difference was demonstrated (Orio et al., 2003,
2004; Panidis et al., 2003, 2004, 2005; Haap et al., 2005; Glintborg
et al., 2006; Gulcelik et al., 2006, 2008; Tan et al., 2006; Bik et al.,
2007; Moran et al., 2007; Shroff et al., 2007; Barber et al., 2008). In
an attempt to further investigate the role of adiponectin in PCOS,
many researchers examined the relationship between adiponectin
and PCOS characteristics, such as adiposity, IR and androgen levels,
yet with conflicting results.
Recently, low adiponectin levels have been associated with pro-
gression towards type 2 diabetes mellitus (Schwarz et al., 2006; Jalo-
vaara et al., 2008) and have been suggested to reflect increased risk
of cardiovascular disease in women (Zyriax et al., 2008). Both type
2 diabetes and cardiovascular disease are considered major long-term
health risks for women with PCOS (The Rotterdam ESHRE/ASRM-
Sponsored PCOS Consensus Workshop Group, 2004). Given the
fact that low adiponectin levels could serve a predictor to the pro-
gression to type 2 diabetes and cardiovascular disease in women, adi-
ponectin levels in PCOS might help identify high-risk PCOS patients or
even impose a stricter follow-up and possibly an early treatment
initiation. A definite answer to whether adiponectin levels are lower
Toulis et al.
(as expected), higher or the same in women with PCOS and controls
could provide useful implications for the pathogenesis of the syndrome
and potentially its prognosis and physical history. An answer to
whether hypoadiponectinaemia (if confirmed) is a result of the inter-
actions between IR and hyperandrogenaemia or an intrinsic feature of
PCOS is necessary for the interpretation of the former and might also
provide inferences for the diagnosis of the syndrome. These two
answers could help in deciding whether adiponectin measurement is
of any potential value as biomarker of severity for a clinician confront-
ing PCOS.
To investigate adiponectin levels in women with PCOS, a systematic
review of all studies reporting total and HMW adiponectin levels in
women with PCOS and a meta-analysis of the best available evidence
were performed. To gain insights into the influence of IR and testos-
terone on adiponectin levels, appropriate categorical analyses and
meta-regression modelling have been undertaken. To investigate the
influence of obesity on adiponectin levels in PCOS, a ‘nested’
meta-analysis was undertaken after body mass index (BMI) stratifica-
Downloaded from http://humupd.oxfordjournals.org/ by guest on March 27, 2015
tion of subjects into subgroups within studies and appropriate
pooling of them (Thompson and Higgins, 2005).
Methods
Search strategy
A preliminary search was conducted in electronic database MEDLINE on
various combinations of the terms ‘polycystic ovary syndrome’ (MeSH),
‘hyperandrogenism’ (MeSH), ‘insulin resistance’ (MeSH), ‘adiponectin’
(MeSH) and ‘adipokines’ (MeSH) in order to evaluate the size of the rel-
evant bibliography and to orientate to the keywords which would be used
in the main search. To identify eligible studies, the main search was con-
ducted in the electronic databases MEDLINE, EMBASE and Cochrane
Central Register of Controlled Trials (CENTRAL) from inception
through June 2008, using the terms (‘Adiponectin’ [MeSH] OR ‘Adipo-
kines’ [MeSH]) AND (‘Polycystic Ovary Syndrome’ [MeSH] OR ‘Hyperan-
drogenism’ [MeSH]) and restricted to English literature. The procedure
was concluded by the perusal of the reference sections of all relevant
studies or reviews, a manual search of key journals and abstracts from
the major annual meetings in the field of endocrinology and a contact
with experts on the subject, in an effort to identify relevant unpublished
data. Finally, unpublished studies were also sought in the web sites isrct-
n.org, clinicaltrials.gov, cihr.ca, action.org, UK clinical trials gateway and
Wellcome Trust.
The main search, as well as screening of titles, abstracts and full-text
articles, was completed independently by two reviewers (K.A.T. and
D.G.G.) with expertise in conducting systematic reviews. Chance-adjusted
inter-rater agreement was calculated using Cohen’s k-statistics and found
satisfactory. Any discrepancy was solved by consultation of a third
reviewer, not involved in the initial procedure (I.P.).
Eligibility of relevant studies
Suitable for the systematic review were studies of any design, which
reported adiponectin levels in women with PCOS [diagnosed consistently
to either Rotterdam (The Rotterdam ESHRE/ASRM-Sponsored PCOS
Consensus Workshop Group, 2004) or National Institute of Health
(NIH) criteria (Zawadzki and Dunaif, 1992)] compared with healthy con-
trols. Eligible for the main meta-analysis were those studies included in the
systematic review where BMI was similar between women with PCOS and
controls. Eligible for the ‘nested’ meta-analysis were those studies included
Adiponectin and PCOS: a meta-analysis
in the systematic review where adiponectin data on BMI subgroups were
available. Studies were excluded from the systematic review and the
meta-analyses if enrolled subjects had a disease other than PCOS, were
on any kind of medication, were pregnant or were genetically related.
Studies with no control group or control group including men were also
excluded, as well as reviews, letters to the editor and studies published
in language other than English.
Data extraction
Information from each study was extracted independently by two
reviewers (K.A.T. and D.G.G.), using a standardized data extraction
form. General characteristics of the study (author, journal, year of publi-
cation, design, ethnicity, study size and number of cases), characteristics
of the PCOS and control groups (criteria, selection, age, BMI, insulin sen-
sitivity status, androgen levels), methodology (PCOS definition, adiponec-
tin measurement method, study quality) and results [total and HMW
adiponectin means and standard deviation (SD), total adiponectin means
and SDs after BMI stratification, subgroup and regression analyses] were
recorded, where available, and double-checked. Where appropriate,
data set was completed through communication with the authors. Pre-
defined criteria [a modification of Newcastle– Ottawa Scaling for non-
randomized studies (Wells et al., 2000)] were used independently by
two reviewers to assess the quality of studies. The criteria were the com-
parability of subjects on the basis of BMI, the representativeness of sub-
jects (selection of PCOS and control group), the validity of definitions
and measurements and the lack of financial support from pharmaceutical
industry. Inter-rater agreement was evaluated by means of k-test, and dis-
agreement was resolved by consensus.
Institutional Review Board approval was not asked for this study, as no
experiment or intervention was held on humans.
Statistical analysis
Total adiponectin, HMW adiponectin and total testosterone levels in each
study were extracted as mean difference + SD. When not reported,
missing mean differences and SDs were estimated on the basis of reported
confidence intervals (CIs), where available, and confirmation was sought
through communication with the authors. Standard error of the mean
was transformed into SD. Homeostasis Model Assessment of Insulin
Resistance (HOMA2-IR) values were used as measures of insulin sensi-
tivity. When not reported, missing HOMA2-IR values were estimated
on the basis of reported mean glucose and insulin values, using the
Oxford Diabetes Trials Unit calculator (www.dtu.ox.ac.uk).
Weighted mean differences (WMDs) in total plasma adiponectin were
calculated for all eligible studies for the meta-analysis and combined using
fixed or random-effects model (DerSimonian and Laird, 1986), where
appropriate. Standardized mean differences (SMDs) in HMW adiponectin
were calculated, as they were measured in different units across studies.
Heterogeneity between the results of different studies was examined by
x 2 tests for significance (a P-value ,0.1 was considered statistically signifi-
cant) and I 2 test (I 2 . 50%: significant heterogeneity; I 2 , 25%: insignifi-
cant heterogeneity), which can be interpreted as the percentage of total
variation across several studies owing to heterogeneity (Higgins et al.,
2003). To assess the extent of publication bias, Egger test for publication
bias was used (Davey Smith et al., 1997). Visual inspection of publication
bias by a funnel plot of the data was not applied, as it may be incorrectly
inferred if studies are heterogeneous (Terrin et al., 2003). Sensitivity analy-
sis was undertaken with the exclusion of the studies with borderline
eligibility.
A priori hypotheses to explain potential heterogeneity across studies
included possible effect modifiers, such as IR and testosterone levels,
chosen on the basis of biological plausibility. As stated previously,
299
difference in IR and/or androgen status between PCOS and control
group could account for a difference in adiponectin levels and a percentage
of the potential variability of across-study results. The relative difference in
IR between groups of a study was expressed by the ratio of mean
HOMA2-IR value in PCOS women to controls, and correspondingly,
the relative difference in testosterone levels was expressed by the ratio
of mean total testosterone in PCOS women to controls. The influence
of study size and study quality was also investigated. A categorical
meta-analysis was performed by all four possible moderators after the
stratification of studies into relevant categories. Through this analysis, het-
erogeneity was partitioned into the components from the model (relevant
categories for each moderator) and the residual heterogeneity, an analysis
which is analogous to the partitioning of variance in ANOVA (Hedges and
Olkin, 1985). As a multiplicity adjustment procedure, the Bonferroni test
was used, where there are no assumptions required about the data distri-
butions and/or about the correlation structure among the end-points
(Sankoh et al., 1997). Adjusted P-values are calculated by k  unadjusted
P-values, where k is the number of planned comparisons, which are four in
this study. The effect of study-level characteristics on adiponectin levels
Downloaded from http://humupd.oxfordjournals.org/ by guest on March 27, 2015
was checked through restricted maximum likelihood (REML)-based
random-effects meta-regression analysis. Univariate meta-regression ana-
lyses were performed first. Variables that were significant at the 0.1
level were entered into the multivariable model. Non-significant covariates
(P , 0.05) in the multivariable model were deleted. The percentage of
between-study variance explained by the model (tauexp) was estimated
by the formula: tauexp ¼ 1 2 tau2 (without covariates)/tau2 (model).
To investigate the influence of obesity while securing reliable
interpretation, categories of subjects on the basis of BMI within every
study, included in the systematic review, were compared and then the
results combined appropriately over studies (Thompson and Higgins,
2005) by means of a ‘nested’ meta-analysis. In specific, data within
each study were stratified on the basis of BMI into three categories:
lean (BMI , 25 kg/m2), overweight (BMI  25 and ,30 kg/m2) and
obese (BMI  30 kg/m2). WMDs were calculated for all three categories
of each study. Thus, from each study, three additional WMDs were
derived, one regarding lean PCOS and control women (subjects’ sub-
group), one regarding overweight subgroup and one regarding obese
subgroup, given that relevant data were available. To investigate adipo-
nectin levels between lean women with PCOS and lean controls, a quan-
titative (meta-analysis) of the relevant WMDs was performed. This
statistical procedure was repeated for the overweight and obese subjects
as well.
Meta-analysis was conducted using Review Manager 4.2 software
[Review Manager (RevMan), version 4.2 for Windows, Oxford, UK: the
Cochrane Collaboration, 2002] and Stata/SE 9.0 for Windows (StataCorp
LP, 4905 College Station, TX 77845, USA). The report of the study was
complemented in adherence with the Meta-analysis Of Observational
Studies in Epidemiology (MOOSE) group standards for reporting
meta-analysis of observational studies (Stroup et al., 2000).
Results
Search results
The search strategy identified 158 potentially relevant studies. A flow
chart summarizing search results is provided in Fig. 1. Three unpub-
lished studies were identified, two of them through the web site clin-
icaltrials.gov under the registry numbers NCT00173043 and
NCT00667498 (currently active), and one by Panidis et al. (retrieved).
Thirty-seven publications were excluded because it was clear from the
title that they did not fulfil the selection criteria. Three of them were
300
Figure 1 Flow chart for the systematic review and the
meta-analysis
read in full (K.A.T.) to check the validity of this selection procedure.
From the remaining 121 publications, 64 were excluded on the basis
of the abstract. Fifty-seven articles were read in full independently
by two reviewers to assess their accordance with the selection criteria.
Eight studies were excluded because no control group of women
without PCOS was included (Lewandowski et al., 2005; Vrbikova
et al., 2005; Majuri et al., 2007; Xita et al., 2007; Elkind-Hirsch
et al., 2008; Jensterle et al., 2008; Weerakiet et al., 2007, 2008),
one study was excluded because adiponectin was not measured in
the control group (Xita et al., 2005) and one study was excluded
because recruited women were with PCOS and had precocious pub-
arche (Ibanez et al., 2004). In total, 26 studies were excluded because
they did not fulfil the predefined selection criteria. Finally, 31 studies
were included in the systematic review.
From the 31 studies included in the systematic review, 5 studies
were excluded from the meta-analysis, because PCOS women and
controls differed significantly in BMI (Panidis et al., 2004; Spranger
et al., 2004; Haap et al., 2005; Carmina et al., 2008a; Zhang et al.,
2008). Eight studies were excluded because no sufficient data were
available: in three of them, results were presented only as medians
or in figures (Sieminska et al., 2004; Bik et al., 2007; Carmina et al.,
2008b); in five of them, only in subgroups and not in total (Ducluzeau
Toulis et al.
et al., 2003; Orio et al., 2003; Ibanez and de Zegher, 2004; Ardawi
and Rouzi, 2005; Panidis et al., unpublished data). In one study, all
PCOS women had severe IR (Sepilian and Nagamani, 2005),
forming a rather distinct population than the target. This was con-
sidered to introduce bias in the analysis, and, therefore, the study
was excluded. Finally, a study was excluded (Carmina et al., 2006),
as mean adiponectin values and mean BMI coincide with those in a
previous study by the same authors (Carmina et al., 2005), suggesting
sample overlap. Finally, 16 studies were included in the main
meta-analysis.
Systematic review
The 31 studies were published between 2003 and 2008 and reported
data on 3469 subjects (1776 women with PCOS and 1693 controls).
In three of them, HMW adiponectin was also reported. Main data are
summarized in Supplementary Material, Tables S1 and S2. The
majority of studies was held in Europe (22 studies—70%), four in
Downloaded from http://humupd.oxfordjournals.org/ by guest on March 27, 2015
the USA, four in Asia and one in Australia. Eighteen studies (58%)
applied the NIH criteria for the diagnosis of PCOS, 12 (39%) the Rot-
terdam consensus criteria, whereas one applied modified NIH criteria
(Ibanez and de Zegher, 2004). Twenty-four studies (77%), using
diverse methodology, reported that PCOS women were more insulin-
resistant than controls, whereas seven reported no difference. Seven
studies (23%) were held in more than one centre and three were ran-
domized to an insulin-sensitizing agent (baseline adiponectin was
extracted). Overall, the reporting of studies was good: total
numbers, inclusion criteria and reporting on baseline characteristics
were generally clearly stated; however, methods of recruitment
were poorly reported and reporting on adiponectin was also incom-
plete. No substantial differences were observed in the methodology
applied to measure adiponectin. Almost half of them (15 out 31
studies) used radioimmunoassay (RIA) to quantify adiponectin levels,
13 studies (mostly more recent ones) applied enzyme-linked immuno-
sorbent assays or enzyme immunoassay (1 study), 2 studies (by the
same author) used time-resolved immunofluorometric assay, and in
1 study, no relevant information was reported. Many studies reported
quality control measurements, although intra-assay and inter-assay
coefficients of variation were high in some of them.
Meta-analysis
A meta-analysis of 16 studies, which included only subjects of similar
BMI, was performed. Included studies reported data on 1186 women
(660 PCOS and 526 controls), and their size ranged from 16 to 240
(median 49). In two of them, BMI was similar only on a subset of sub-
jects; only these data were used in the meta-analysis, and their influence
on the outcome was checked through sensitivity analysis.
Women with PCOS demonstrated significantly lower adiponectin
values, yet with significant between-study heterogeneity [16 studies,
random-effects WMD (95% CI) 21.71 (22.82 to 20.6); I 2 ¼ 80.7%;
Fig. 2]. Egger test for publication bias did not provide evidence of signifi-
cant effect (P ¼ 0.139). Sensitivity analysis, with the exclusion of the two
studies where BMI was similar between PCOS patients and controls only
in a subset of subjects, did not alter the result [14 studies, random-effects
WMD (95% CI) 21.71 (22.85 to 20.57); I 2 ¼ 82.6%].
Quantitative synthesis of the data from three studies where HMW
adiponectin was measured (Aroda et al., 2008; Barber et al., 2008;
Adiponectin and PCOS: a meta-analysis
Figure 2 Weighted mean adiponectin values in PCOS and control women and pooled estimate (random effects).
Glintborg et al., 2008) demonstrated no significant difference in HMW
adiponectin between women with PCOS and controls, yet with signifi-
cant between-study heterogeneity [three studies, random-effects SMD
(95% CI) 20.59 (21.33 to 0.16); I 2 ¼ 69.7%]. Unfortunately, this
analysis had limited power to detect a true difference, as only three
studies were included.
Categorical analyses
A priori hypotheses to explain potential heterogeneity between studies
included: HOMA2-IR ratio (mean values in women with PCOS to con-
trols in a single study), study size, total testosterone ratio (mean values
in women with PCOS to controls in a single study) and study quality.
Pooled WMDs in adiponectin between PCOS women and controls
were calculated in four predefined categories, each covering a quartile
interval (25% of studies) for the first two potential moderators
(HOMA2-IR ratio and study size). Quartile intervals for HOMA2-IR
ratio were ,1.36, 1.36 –1.7, 1.71 –2.12, .2.12. Quartile intervals
301
Downloaded from http://humupd.oxfordjournals.org/ by guest on March 27, 2015
for study size were ,38, 39 –49, 50 –112, .113. For the last two
moderators (testosterone ratio and study quality) and because of
their distribution, study-level WMDs were categorized in three
groups. For testosterone ratio, intervals were ,1.68, 1.69 –2.49,
.2.5, and for study quality, intervals were studies with lower quality
scores (,312), with medium quality scores (312) and with higher
quality scores (.312). When relevant information (HOMA or testos-
terone ratio) was not possible to be extracted from a study (missing
data), this study was classified into an additional separate category.
When HOMA2-IR ratio was used as a moderator in the categ-
orical meta-analysis, WMD in adiponectin across relevant cat-
egories reflected a trend of a direct correlation with it. In
studies with modest relative difference in IR between PCOS and
control groups, no significant difference in adiponectin was
observed. The P-value for between-groups heterogeneity was
0.001 (adjusted P-value ¼ k  0.001 ¼ 0.004 , 0.05, where k is
the number of designed comparisons, 23.75% of the total hetero-
geneity), indicating that the level of the relative difference in IR
302 Toulis et al.

Downloaded from http://humupd.oxfordjournals.org/ by guest on March 27, 2015

Figure 3 Summary of caterogical meta-analysis: categorical meta-analysis on HOMA-IR ratio; categorical meta-analysis on total testosterone ratio;
categorical meta-analysis on study size; categorical meta-analysis on study quality.
Adiponectin and PCOS: a meta-analysis
Table I Results of the multivariate meta-regression analysis
Fit of model without heterogeneity (tau2 5 0):
Proportion of variation due to heterogeneity
REML estimate of between-study variance:
WMD coef. std. error t P
.............................................................................................................................................................................................
HOMA ratio 22.50524 0.8842145 22.83
Study size 0.0278142 0.0087333
Cons 0.6025889 1.664828
between PCOS and control group (HOMA2-IR ratio) contributed
significantly to the heterogeneity across studies. Results are sum-
marized in Supplementary Material, Table S3A and visual inspection
of them in Fig. 3.
When study size was used as a moderator in the categorical
meta-analysis, WMD in adiponectin across relevant study size cat-
egories demonstrated evidence of inverse relation to it. In larger
study categories, women with PCOS did not demonstrate any signifi-
cant difference in adiponectin values compared with controls, in con-
trast to the smaller study categories, where adiponectin was found
significantly lower in women with PCOS. The P-value for
between-groups heterogeneity was significant (P-value , 0.001,
adjusted P-value , 0.004, 42.66% of the total heterogeneity).
Results are summarized in Supplementary Material, Table S3B and
visual inspection of them in Fig. 3.
No evident pattern of correlation arose when categorical analysis
was performed using total testosterone and study quality as modera-
tors. WMD in adiponectin across categories did not seem to be
explained by these two variables, as well as between-groups hetero-
geneity was not found to contribute significantly to total heterogeneity
in the case of T-ratio (P-value ¼ 0.014, adjusted P-value ¼ 0.056,
13.68% of the total heterogeneity). However, the P-value for
between-groups heterogeneity in the case of study quality was signifi-
cant (P-value ¼ 0.004, adjusted P-value ¼ 0.016, 14.53% of the total
heterogeneity). Results are summarized in Supplementary Material,
Table S3C and D and visual inspection of them in Fig. 3.
Meta-regression
To further investigate the impact of the predefined study-level charac-
teristics on WMD in adiponectin, an REML-based random-effects
meta-regression analysis was performed. WMD was used as the
dependent variable (logarithmic transformation was not possible for
WMD, as negative values were included), and HOMA2-IR ratio,
study size, total testosterone ratio and study quality were entered
as explanatory covariates. REML estimate of the between-study var-
iance (tau2) with no covariates in the model was 4.908. Univariate
meta-regression analyses were performed first. HOMA2-IR ratio (15
studies, P ¼ 0.086), study size (16 studies, P ¼ 0.078), total testoster-
one ratio (13 studies, P ¼ 0.278) and study quality (16 studies,
P ¼ 0.730) were assessed independently. Covariates whose
regression coefficients were significant at the level of 0.1 in the univari-
ate analysis, namely HOMA2-IR ratio and study size, were entered in
303
Number of studies 5 15
Q (12 df) 5 27.0027
Prob. > Q 5 0.008
I 2 5 0.556
tau2 5 2.1140
95% Conf. Interval
0.015 24.431778 20.5787022
3.18 0.008 0.0087861 0.0468424
0.36 0.724 23.02476 4.229938
the multivariate model. Regression coefficients for both HOMA2-IR
ratio and study size were found significant at the level of 0.05 in the
multivariate model. REML estimate of the between-study variance
Downloaded from http://humupd.oxfordjournals.org/ by guest on March 27, 2015
was reduced from 4.908 to 2.114, and the percentage explained by
the multivariate model (tauexp) was 56.92%. Results from the multi-
variate meta-regression analysis are presented in Table I. The possi-
bility of a quadratic, cubic and logarithmic relationship between
WMD and HOMA2-IR was explored in the multivariate model by rel-
evant transformation of the covariate, but yielded no improvement to
the model. Graphical display of the linear relationship between WMD
and HOMA2-IR ratio is presented in Fig. 4.
‘Nested’ meta-analysis on BMI
Data regarding differences in adiponectin levels between lean women
with PCOS and lean controls were available in 12 studies. WMDs
regarding this subset (lean) in each study were combined and a
pooled WMD derived [random-effects WMD (95% CI) 21.62
(23.20 to 20.04), I 2 ¼ 70.1%]. Data regarding differences in adipo-
nectin between overweight women with PCOS and overweight con-
trols were available in nine studies [fixed effects WMD (95% CI)
20.93 (22.14 to 0.27), I 2 ¼ 28.7%]. These subsets of studies had
wide CIs and small number of subjects (less than 20 in 5 out of 9
Figure 4 Regression fit graph of adiponectin weighted mean differ-
ence (WMD) and HOMA2-IR ratio (scatterplot and 95% confidence
intervals, 15 studies).
304
studies). Data regarding differences in adiponectin between obese
women with PCOS and obese controls were available in 12 studies
[random-effects WMD (95% CI) 21.22 (22.41 to 20.03),
I 2 ¼ 75%]. Adiponectin was found lower in both lean and obese
women with PCOS, when compared with healthy counterparts,
suggesting that relative hypoadiponectinaemia in women with PCOS com-
pared with healthy controls is present, regardless of the degree of obesity.
Discussion
Systematic review and meta-analysis of relevant studies suggest that
adiponectin is lower in women with PCOS compared with
non-PCOS controls of similar BMI. Lower adiponectin levels are
associated with the IR observed in women with PCOS compared
with controls. It has been demonstrated that the more insulin-resistant
women with PCOS recruited, the lower serum adiponectin levels
were found. Total testosterone levels failed to account for any differ-
ence in adiponectin observed between groups. Relative hypoadipo-
nectinaemia in women with PCOS compared with healthy controls
of similar BMI was present, regardless of the degree of obesity
examined.
Limitations of the study and alternative
explanation of findings
Certain limitations of the present meta-analysis might undermine the
generalization of the findings or impose an alternative context for
interpretation. First, the significant heterogeneity observed across
study results (I 2 ¼ 80.7%) reduces the reliability of the result,
suggesting different outcomes across studies with different design,
patients, methods and measurements. This inconsistency was only
partially explained (tauexp ¼ 56.92%) by predefined categorical and
meta-regression analyses, which pointed out that in smaller studies
and in studies where women with PCOS were markedly more insulin-
resistant than controls, adiponectin was found significantly lower in
women with PCOS. In contrast, in larger studies and in studies with
modest difference in IR between groups, no difference was observed.
Whereas there is a physiological rationale to explain the observed
influence of IR on adiponectin levels, scientific rationale for explaining
the influence of study size on adiponectin levels should be restricted to
methodological aspects and could be found in the limited precision of
smaller studies to evaluate differences in adiponectin levels.
Mean HOMA-IR values were chosen as a measure of IR in an effort
to obtain a uniform comparator of IR. Despite the relative inferiority of
HOMA to euglycaemic hyperinsulinaemic clamp and/or oral glucose
tolerance test, as it does not correlate with any aspect of insulin-
receptor binding (Matthews et al., 1985), HOMA2-IR values could
be easily approximated (where missing) on the basis of reported
mean glucose and insulin values (yet euglycaemic hyperinsulinaemic
clamp and/or oral glucose tolerance test results could not), providing
in this way this requisite uniform comparator of IR. Without that, the
investigation of the influence of IR on adiponectin levels could not be
feasible. On the other hand, this missing data manipulation integrated
a certain degree of systemic bias in the expression of IR, which should
be considered a limitation of the present analysis. In an effort to adjust
for it, significance level in the meta-regression models could be set at
0.025, instead of 0.05. Even in that case, regression coefficient
Toulis et al.
for HOMA2-IR in the multivariate model remains significant
(P ¼ 0.015), indicating a rather strong relationship, yet not establishing
causality. On the other hand, total testosterone levels, which were
used as a biochemical marker of hyperandrogenaemia on the basis
of the availability, failed to show significant influence on adiponectin
levels. An alternative explanation of this finding could be found in
the inferiority of total testosterone to free androgen index as a
marker of hyperandrogenaemia and the fact that in only 13 out of
16 studies testosterone data were available. It could also be attributed
to the failure of testosterone assays to accurately and precisely detect
levels of testosterone, especially below 100 ng/dl (3.47 nM) (Wang
et al., 2004). From the 16 studies included in the main meta-analysis,
6 applied specific RIAs to measure total testosterone levels, 4 applied
chemiluminescence immunoassays, and in 4, no data, no testosterone
measurement or other methodology was reported. Subgroup analyses
after BMI stratification failed to provide further insights, probably due
to the small numbers included. The influence of visceral adiposity
could not be checked, as waist circumference or waist-to-hip ratio
Downloaded from http://humupd.oxfordjournals.org/ by guest on March 27, 2015
was not available in the majority of studies. English language bias
could potentially have introduced bias in the present analysis (Egger
et al., 1997). Finally, insufficient data limit the generalization of
HMW adiponectin findings, as only three studies reported HMW
values, and in one of them, data extraction was implicated with
missing data manipulation.
Implications for clinical practice and research
Data analysis suggests a direct relationship between IR and differences
in adiponectin between PCOS patients and controls, which could be
extrapolated to the assumption that adiponectin is not an intrinsic
characteristic of PCOS. Interpreting the former in terms of clinical
practice, adiponectin measurement in women with PCOS could not
provide clinicians with additional information regarding the diagnosis,
prognosis and physical history of the syndrome to that provided by
an estimate of IR, and thus, it should not be used as a biomarker of
PCOS severity. The effect of testosterone on adiponectin levels and,
particularly, on HMW fraction should be further investigated. Anti-
atherogenic, anti-inflammatory and insulin-sensitizing actions of adipo-
nectin could be of potential research interest in PCOS.
Conclusions
After controlling for BMI-related effects, adiponectin levels seem to be
lower in women with PCOS compared with non-PCOS controls. Low
serum adiponectin levels in PCOS are probably related to IR, but not
to testosterone. Further investigation is needed to clarify whether
HMW adiponectin levels are also suppressed in PCOS.
Supplementary data
Supplementary data are available at http://humupd.oxfordjournals.
org/.
Acknowledgements
We are very grateful to Dr F. Orio (University ‘Federico II’, Naples,
Italy), Dr D. Glintborg (Odense University Hospital, Odense,
Denmark), Drs L. Moran and R. Norman (University of Adelaide,
Adiponectin and PCOS: a meta-analysis
Adelaide, Australia), Dr J. Puder (University of Lausanne, Lausanne,
Switzerland), Dr A. Dokras (University of Iowa, Iowa City, IA,
USA), Dr H. S. Randeva (University of Warwick, Coventry, UK),
who all provided additional data requested on their publications.
We also thank Dr Gulcelik (Ankara Oncology Hospital, Ankara,
Turkey), Drs J. Spranger and C. Schofl (Charité-University Medicine,
Berlin, and Hannover Medical School, Hannover, Germany) and Drs
T. Barber and M.I. McCarthy (Churchill Hospital, Oxford, UK) for
their willingness to provide further information on their publications.
References
Ardawi MS, Rouzi AA. Plasma adiponectin and insulin resistance in women
with polycystic ovary syndrome. Fertil Steril 2005;83:1708 – 1716.
Aroda V, Ciaraldi TP, Chang SA, Dahan MH, Chang RJ, Henry RR.
Circulating and cellular adiponectin in polycystic ovary syndrome:
relationship to glucose tolerance and insulin action. Fertil Steril 2008;
89:1200– 1208.
Barber TM, Hazell M, Christodoulides C, Golding SJ, Alvey C, Burling K,
Vidal-Puig A, Groome NP, Wass JAH, Franks S et al. Serum levels of
retinol-binding protein 4 and adiponectin in women with polycystic
ovary syndrome: associations with visceral fat but no evidence for fat
mass-independent effects on pathogenesis in this condition. J Clin
Endocrinol Metab 2008;93:2859 – 2865.
Beckman JA, Goldfine AB, Dunaif A, Gerhard-Herman M, Creager MA.
Endothelial function varies according to insulin resistance disease type.
Diabetes Care 2007;30:1226– 1232.
Bik W, Baranowska-Bik A, Wolinska-Witort E, Chmielowska M,
Martynska L, Baranowska B. The relationship between metabolic
status and levels of adiponectin and ghrelin in lean women with
polycystic ovary syndrome. Gynecol Endocrinol 2007;23:325 – 331.
Carmina E, Orio F, Palomba S, Cascella T, Longo RA, Colao AM,
Lombardi G, Lobo RA. Evidence for altered adipocyte function in
polycystic ovary syndrome. Eur J Endocrinol 2005;152:389 – 394.
Carmina E, Orio F, Palomba S, Longo RA, Cascella T, Colao A,
Lombardi G, Rini GB, Lobo RA. Endothelial dysfunction in PCOS: role
of obesity and adipose hormones. Am J Med 2006;119:356.e1 – 356.e6.
Carmina E, Bucchieri S, Mansueto P, Rini G, Ferin M, Lobo RA. Circulating
levels of adipose products and differences in fat distribution in the
ovulatory and anovulatory phenotypes of polycystic ovary syndrome.
Fertil Steril 2008a, doi:10.1016/j.fertnstert.2008.03.007, May 1 (Epub
ahead of print).
Carmina E, Chu MC, Moran C, Tortoriello D, Vardhana P, Tena G,
Preciado R, Lobo R. Subcutaneous and omental fat expression of
adiponectin and leptin in women with polycystic ovary syndrome.
Fertil Steril 2008b;89:642 – 648.
Davey Smith G, Egger M, Phillips AN. Meta-analysis. Beyond the grand
mean? BMJ 1997;315:1610 – 1614.
DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials
1986;7:177 – 188.
Ducluzeau PH, Cousin P, Malvoisin E, Bornet H, Vidal H, Laville M,
Pugeat M. Glucose-to-insulin ratio rather than sex hormone-binding
globulin and adiponectin levels is the best predictor of insulin
resistance in nonobese women with polycystic ovary syndrome. J Clin
Endocrinol Metab 2003;88:3626 – 3631.
Egger M, Zellweger-Zähner T, Schneider M, Junker C, Lengeler C,
Antes G. Language bias in randomised controlled trials published in
English and German. Lancet 1997;350:326– 329.
Ehrmann DA. Polycystic ovary syndrome. N Engl J Med 2005;
352:1223– 1236.
305
Elkind-Hirsch K, Marrioneaux O, Bhushan M, Vernor D, Bhushan R.
Comparison of single and combined treatment with exenatide and
metformin on menstrual cyclicity in overweight women with polycystic
ovary syndrome. J Clin Endocrinol Metab 2008;93:2670 – 2678.
Escobar-Morreale HF, Villuendas G, Botella-Carretero JI, Alvarez-Blasco F,
Sanchon R, Luque-Ramirez M, San Millan JL. Adiponectin and resistin in
PCOS: a clinical, biochemical and molecular genetic study. Hum Reprod
2006;21:2257 – 2265.
Franks S. Polycystic ovary syndrome. N Engl J Med 1995;333:853– 861.
Glintborg D, Andersen M, Hagen C, Frystyk J, Hulstrom V, Flyvbjerg A,
Hermann AP. Evaluation of metabolic risk markers in polycystic ovary
syndrome (PCOS). Adiponectin, ghrelin, leptin and body composition
in hirsute PCOS patients and controls. Eur J Endocrinol 2006;
155:337– 345.
Glintborg D, Frystyk J, Hojlund K, Andersen KK, Henriksen JE,
Hermann AP, Hagen C, Flyvbjerg A, Andersen M. Total and high
molecular weight (HMW) adiponectin levels and measures of glucose
and lipid metabolism following pioglitazone treatment in a randomized
placebo-controlled study in polycystic ovary syndrome. Clin Endocrinol
Downloaded from http://humupd.oxfordjournals.org/ by guest on March 27, 2015
(Oxf) 2008;68:165 – 174.
Gulcelik NE, Aral Y, Serter R, Demir Y, Culha C. Adiponectin is an
independent determinant of insulin resistance in women with
polycystic ovary syndrome. Gynecol Endocrinol 2006;22:511 – 515.
Gulcelik NE, Aral Y, Serter R, Koc G. Association of hypoadiponectinemia
with metabolic syndrome in patients with polycystic ovary syndrome.
J Natl Med Assoc 2008;100:64 – 68.
Haap M, Machicao F, Stefan N, Thamer C, Tschritter O, Schnuck F,
Wallwiener D, Stumvoll M, Haring HU, Fritsche A. Genetic
determinants of insulin action in polycystic ovary syndrome. Exp Clin
Endocrinol Diabetes 2005;113:275– 281.
Hara K, Horikoshi M, Yamauchi T, Yago H, Miyazaki O, Ebinuma H,
Imai Y, Nagai R, Kadowaki T. Measurement of the high-molecular
weight form of adiponectin in plasma is useful for the prediction of
insulin resistance and metabolic syndrome. Diabetes Care 2006;
29:1357– 1362.
Hedges LV, Olkin I. Statistical Methods for Meta-Analysis. Orlando:
Academic Press, 1985.
Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency
in meta-analyses. BMJ 2003;327:557 – 560.
Ibanez L, de Zegher F. Ethinylestradiol-drospirenone,
flutamide-metformin, or both for adolescents and women with
hyperinsulinemic hyperandrogenism: opposite effects on
adipocytokines and body adiposity. J Clin Endocrinol Metab 2004;89:
1592 – 1597.
Ibanez L, Valls C, Marcos MV, Ong K, Dunger DB, De Zegher F. Insulin
sensitization for girls with precocious pubarche and with risk for
polycystic ovary syndrome: effects of prepubertal initiation and
postpubertal discontinuation of metformin treatment. J Clin Endocrinol
Metab 2004;89:4331 – 4337.
Jalovaara K, Santaniemi M, Timonen M, Jokelainen J, Kesäniemi YA,
Ukkola O, Keinänen-Kiukaanniemi S, Rajala U. Low serum
adiponectin level as a predictor of impaired glucose regulation and
type 2 diabetes mellitus in a middle-aged Finnish population.
Metabolism 2008;57:1130 – 1134.
Jensterle M, Weber M, Pfeifer M, Prezelj J, Pfutzner A, Janez A.
Assessment of insulin resistance in young women with polycystic
ovary syndrome. Int J Gynaecol Obstet 2008;102:137 – 140.
Kadowaki T, Yamauchi T. Adiponectin and adiponectin receptors. Endocr
Rev 2005;26:439 – 451.
Lara-Castro C, Luo N, Wallace P, Klein RL, Garvey WT. Adiponectin
multimeric complexes and the metabolic syndrome trait cluster.
Diabetes 2006;55:249 – 259.
306
Lewandowski KC, Szosland K, O’Callaghan C, Tan BK, Randeva HS,
Lewinski A. Adiponectin and resistin serum levels in women with
polycystic ovary syndrome during oral glucose tolerance test: a
significant reciprocal correlation between adiponectin and resistin
independent of insulin resistance indices. Mol Genet Metab 2005;
85:61 – 69.
Majuri A, Santaniemi M, Rautio K, Vartiainen J, Ruokonen A,
Kesaniemi YA, Tapanainen JS, Ukkola O, Morin-Papunen L.
Rosiglitazone treatment increases plasma levels of adiponectin and
decreases levels of resistin in overweight women with PCOS: a
randomized placebo-controlled study. Eur J Endocrinol 2007;
156:263– 269.
Matsubara M, Maruoka S, Katayose S. Decreased plasma adiponectin
concentrations in women with dyslipidemia. J Clin Endocrinol Metab
2002;87:2764 – 2769.
Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF,
Turner RC. Homeostasis model assessment: insulin resistance and
beta-cell function from fasting plasma glucose and insulin
concentrations in man. Diabetologia 1985;28:412 – 419.
Morales AJ, Laughlin GA, Butzow T, Maheshwari H, Baumann G, Yen SS.
Insulin, somatotropic, and luteinizing hormone axes in lean and obese
women with polycystic ovary syndrome: common and distinct
features. J Clin Endocrinol Metab 1996;81:2854 – 2864.
Moran LJ, Noakes M, Clifton PM, Wittert GA, Belobrajdic DP, Norman RJ.
C-reactive protein before and after weight loss in overweight women
with and without polycystic ovary syndrome. J Clin Endocrinol Metab
2007;92:2944 – 2951.
Orio F Jr, Palomba S, Cascella T, Milan G, Mioni R, Pagano C, Zullo F,
Colao A, Lombardi G, Vettor R. Adiponectin levels in women with
polycystic ovary syndrome. J Clin Endocrinol Metab 2003;88:2619– 2623.
Orio F Jr, Palomba S, Cascella T, Di Biase S, Labella D, Russo T,
Savastano S, Zullo F, Colao A, Vettor R et al. Lack of an association
between peroxisome proliferator-activated receptor-gamma gene
Pro12Ala polymorphism and adiponectin levels in the polycystic ovary
syndrome. J Clin Endocrinol Metab 2004;89:5110 – 5115.
Pajvani UB, Hawkins M, Combs TP, Rajala MW, Doebber T, Berger JP,
Wagner JA, Wu M, Knopps A, Xiang AH et al. Complex distribution,
not absolute amount of adiponectin, correlates with
thiazolidinedione-mediated improvement in insulin sensitivity. J Biol
Chem 2004;279:12152 – 12162.
Panidis D, Kourtis A, Farmakiotis D, Mouslech T, Rousso D, Koliakos G.
Serum adiponectin levels in women with polycystic ovary syndrome.
Hum Reprod 2003;18:1790 – 1796.
Panidis D, Kourtis A, Kukuvitis A, Farmakiotis D, Xita N, Georgiou I,
Tsatsoulis A. Association of the T45G polymorphism in exon 2 of the
adiponectin gene with polycystic ovary syndrome: role of
Delta4-androstenedione. Hum Reprod 2004;19:1728 – 1733.
Panidis D, Farmakiotis D, Rousso D, Koliakos G, Kaltsas T, Krassas G.
Decrease in adiponectin levels in women with polycystic ovary
syndrome after an oral glucose tolerance test. Fertil Steril 2005;
83:232 – 234.
Sankoh AJ, Huque MF, Dubey SD. Some comments on frequently used
multiple endpoint adjustment methods in clinical trials. Stat Med 1997;
16:2529 – 2542.
Schwarz PE, Towers GW, Fischer S, Govindarajalu S, Schulze J,
Bornstein SR, Hanefeld M, Vasseur F. Hypoadiponectinemia is
associated with progression toward type 2 diabetes and genetic
variation in the ADIPOQ gene promoter. Diabetes Care 2006;
29:1645 – 1650.
Sepilian V, Nagamani M. Adiponectin levels in women with polycystic
ovary syndrome and severe insulin resistance. J Soc Gynecol Investig
2005;12:129 – 134.
Toulis et al.
Shroff R, Kerchner A, Maifeld M, Van Beek EJ, Jagasia D, Dokras A. Young
obese women with polycystic ovary syndrome have evidence of
early coronary atherosclerosis. J Clin Endocrinol Metab 2007;92:
4609 – 4614.
Sieminska L, Marek B, Kos-Kudla B, Niedziolka D, Kajdaniuk D, Nowak M,
Glogowska-Szelag J. Serum adiponectin in women with polycystic
ovarian syndrome and its relation to clinical, metabolic and endocrine
parameters. J Endocrinol Invest 2004;27:528– 534.
Spranger J, Mohlig M, Wegewitz U, Ristow M, Pfeiffer AF, Schill T,
Schlosser HW, Brabant G, Schofl C. Adiponectin is independently
associated with insulin sensitivity in women with polycystic ovary
syndrome. Clin Endocrinol (Oxf) 2004;61:738 – 746.
Stefan N, Stumvoll M. Adiponectin—its role in metabolism and beyond.
Horm Metab Res 2002;34:469 –474.
Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D,
Moher D, Becker BJ, Sipe TA, Thacker SB. Meta-analysis of
observational studies in epidemiology: a proposal for reporting.
Meta-analysis Of Observational Studies in Epidemiology (MOOSE)
group. JAMA 2000;283:2008 –2012.
Downloaded from http://humupd.oxfordjournals.org/ by guest on March 27, 2015
Tan BK, Chen J, Digby JE, Keay SD, Kennedy CR, Randeva HS.
Upregulation of adiponectin receptor 1 and 2 mRNA and protein in
adipose tissue and adipocytes in insulin-resistant women with
polycystic ovary syndrome. Diabetologia 2006;49:2723– 2728.
Terrin N, Schmid CH, Lau J, Olkin I. Adjusting for publication bias in the
presence of heterogeneity. Stat Med 2003;22:2113– 2126.
The Rotterdam ESHRE/ASRM-Sponsored PCOS Consensus Workshop
Group. Revised 2003 consensus on diagnostic criteria and long-term
health risks related to polycystic ovary syndrome (PCOS). Hum
Reprod 2004;19:41 – 47.
Thomann R, Rossinelli N, Keller U, Tirri BF, De Geyter C, Ruiz J,
Kranzlin M, Puder JJ. Differences in low-grade chronic inflammation
and insulin resistance in women with previous gestational diabetes
mellitus and women with polycystic ovary syndrome. Gynecol
Endocrinol 2008;24:199 – 206.
Thompson SG, Higgins JP. Treating individuals 4: can meta-analysis help
target interventions at individuals most likely to benefit? Lancet 2005;
365:341– 346.
Trujillo ME, Scherer PE. Adiponectin—journey from an adipocyte
secretory protein to biomarker of the metabolic syndrome. J Intern
Med 2005;257:167 – 175.
Vrbikova J, Dvorakova K, Hill M, Vcelak J, Stanicka S, Vankova M,
Sramkova D, Vondra K, Bendlova B, Starka L. Determinants of
circulating adiponectin in women with polycystic ovary syndrome.
Gynecol Obstet Invest 2005;60:155 – 161.
Wang C, Catlin DH, Demers LM, Starcevic B, Swerdloff RS. Measurement
of total testosterone in adult men; comparison of current laboratory
methods versus liquid chromatography-tandem mass spectrometry.
J Clin Endocrinol Metab 2004;89:534 – 543.
Weerakiet S, Bunnag P, Phakdeekitcharoen B, Wansumrith S,
Chanprasertyothin S, Jultanmas R, Thakkinstian A. Prevalence of the
metabolic syndrome in Asian women with polycystic ovary syndrome:
using the International Diabetes Federation criteria. Gynecol Endocrinol
2007;23:153 – 160.
Weerakiet S, Tingthanatikul Y, Boonnag P, Wansumrith S, Rattanasiri S,
Leelaphiwat S. Can adiponectin predict abnormal glucose tolerance in
Thai women with polycystic ovary syndrome? J Obstet Gynaecol Res
2008;34:55 – 61.
Wells G, Shea B, O’Connell D, Peterson J, Welch V, Losos M, Tugwell P.
The Newcastle– Ottawa Scale (NOS) for assessing the quality of
nonrandomized studies in meta-analyses. In: Proceedings of the 3rd
Symposium on Systematic Reviews. Beyond the Basics: Improving Quality
and Impact, Oxford, 2000, 3 – 5.
Adiponectin and PCOS: a meta-analysis
Weyer C, Funahashi T, Tanaka S, Hotta K, Matsuzawa Y, Pratley RE,
Tataranni PA. Hypoadiponectinemia in obesity and type 2 diabetes:
close association with insulin resistance and hyperinsulinemia. J Clin
Endocrinol Metab 2001;86:1930 – 1935.
Xita N, Georgiou I, Chatzikyriakidou A, Vounatsou M, Papassotiriou GP,
Papassotiriou I, Tsatsoulis A. Effect of adiponectin gene polymorphisms
on circulating adiponectin and insulin resistance indexes in women with
polycystic ovary syndrome. Clin Chem 2005;51:416–423.
Xita N, Papassotiriou I, Georgiou I, Vounatsou M, Margeli A, Tsatsoulis A.
The adiponectin-to-leptin ratio in women with polycystic ovary
syndrome: relation to insulin resistance and proinflammatory markers.
Metabolism 2007;56:766 – 771.
Xu A, Chan KW, Hoo RL, Wang Y, Tan KC, Zhang J, Chen B, Lam MC,
Tse C, Cooper GJ et al. Testosterone selectively reduces the high
molecular weight form of adiponectin by inhibiting its secretion from
adipocytes. J Biol Chem 2005;280:18073 – 18080.
307
Zhang N, Shi YH, Hao CF, Gu HF, Li Y, Zhao YR, Wang LC, Chen ZJ.
Association of þ45G15G(T/G) and þ276(G/T) polymorphisms in
the ADIPOQ gene with polycystic ovary syndrome among Han
Chinese women. Eur J Endocrinol 2008;158:255– 260.
Zawadzki JK, Dunaif A. Diagnostic criteria for polycystic ovary syndrome:
towards a rational approach. In: Dunaif A, Givens JR, Haseltine FP,
Merriam GR (eds). Polycystic Ovary Syndrome. Boston: Blackwell
Scientific Publications, 1992, 377 – 384.
Zyriax BC, Algenstaedt P, Hess UF, Schöffauer M, Bamberger C, Boeing H,
Windler E. Factors contributing to the risk of cardiovascular disease
reflected by plasma adiponectin: data from the coronary risk factors
for atherosclerosis in women (CORA) study. Atherosclerosis 2008;
200:403– 409.
Submitted on July 17, 2008; resubmitted on December 6, 2008; accepted on
February 3, 2009
Downloaded from http://humupd.oxfordjournals.org/ by guest on March 27, 2015