DR FAHIMEH RAMEZANI TEHRANI( Orcid ID)9790-1204-0000-0000 :

Accepted Article
Article type: 5 Unsolicited Review

Association between biochemical hyperandrogenism parameters

and FerrimanGallwey score in patients with polycystic ovary

syndrome: a systematic review and meta-regression analysis

Mina Amiri 1, 2, Fahimeh Ramezani Tehrani *2, Fatemeh Nahidi 3, Razieh Bidhendi

Yarandi 4, 2, Samira behboudi-Gandevani2, Fereidoun Azizi5

1- Department of Midwifery and Reproductive Health, School of Nursing and

Midwifery, Shahid Beheshti University of Medical Sciences, Tehran, Iran.

2- Reproductive Endocrinology Research Center, Research Institute for Endocrine

Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran.

3- Department of Midwifery and Reproductive Health, Faculty of Nursing and

Midwifery, Shahid Beheshti University of Medical Sciences, Tehran, Iran.

4- Department of Epidemiology and biostatistics, School of public health, Tehran

University of Medical Sciences, Tehran, Iran.

5- Endocrine Research Center, Research Institute for Endocrine Sciences, Shahid

Beheshti University of Medical Sciences, Tehran, Iran.

This article has been accepted for publication and undergone full peer review but has not
been through the copyediting, typesetting, pagination and proofreading process, which may
lead to differences between this version and the Version of Record. Please cite this article as
doi: 10.1111/cen.13389
This article is protected by copyright. All rights reserved.
 *Correspondence Author: Fahimeh Ramezani Tehrani, MD, PhD,

Professor of Obstetrics and Gynecology

Accepted Article
Address: Research Institute for Endocrine Sciences,

No 24, Parvane Street, Yaman Street, Velenjak, Tehran, Iran.

Tel: +98 22432500; Fax: +98 22416264

Email: ramezani@endocrine.ac.ir; framezan@post.harvard.edu

Association between biochemical hyperandrogenism parameters

and FerrimanGallwey score in patients with polycystic ovary

syndrome: a systematic review and meta-regression analysis

Abstract

Background and Objective: A limited number of studies have evaluated the relationship

between clinical and biochemical hyperandrogenism (HA). This study aimed to evaluate the

association between biochemical hyperandrogenism parameters (BHPs) and Ferriman

Gallwey (FG) score in patients with polycystic ovary syndrome (PCOS).

Methods: We searched PubMed, Scopus, Google Scholar, ScienceDirect and Web of

Sciences databases (2000 to 2015) to identify studies investigating clinical and biochemical

parameters of HA in PCOS patients. In this meta-analysis, both fixed and random effect

models were applied to estimate pooled effect size. To assess the relationships between BHPs

and FG score, meta-regression analysis was used.

This article is protected by copyright. All rights reserved.

 Results: Fifty-five study groups with a total of 6593 PCOS patients were analyzed. Meta-

regression analysis of pooled data from all eligible studies showed significant positive
Accepted Article
relationships of FG score with androstenedione (A4) (p= 0.034) and dehydroepiandrosterone

sulfate (DHEAS) (p=0.012), whereas it showed no association with total testosterone (tT),

free testosterone (fT), sex hormone bonding globulin (SHBG) and free androgen index (FAI).

The results did not change after adjusting for quality assessment or method of assay. Nor did

the associations between A4 and FG score remain after adjusting for age and BMI, diagnostic

criteria for PCOS and study design or the association between DHEAS and FG score remain

after adjusting for ethnicity.

Conclusion: This meta-analysis confirmed the associations of some BHPs, including A4 and

DHEAS with FG score, indicating that measurement of these parameters can be useful for

managing PCOS patients with hirsutism.

Keywords: Meta-regression, polycystic ovary syndrome, biochemical hyperandrogenism,

Ferriman-Gallwey score.

Introduction

Polycystic ovary syndrome (PCOS) is one of the most common endocrine disorders in

women of reproductive age 1, characterised by clinical and/or biochemical hyperandogenism

(HA) and chronic anovulation. The criteria of the Androgen Excess Society (AES) for

diagnosis of PCOS require clinical or biochemical evidence of HA, the cornerstone of PCOS.
2
Excess synthesis of ovarian androgens plays an essential role in the clinical and biochemical

manifestations of HA in patients with PCOS. 3, 4 Clinical HA is identified by hirsutism, acne,

This article is protected by copyright. All rights reserved.

 and alopecia. 5 High levels of serum total testosterone (tT), free testosterone (fT), free

androgen index (FAI), androstenedione (A4) and dehydroepiandrosterone sulfate (DHEAS)

Accepted Article
also indicate biochemical HA 3. There are many pitfalls in the selection of biochemical tests.

Although the most frequently measured laboratory parameters are fT and FAI, increased

levels of A4 or other androgen precursors are the only findings in many patients. 6, 7

Hirsutism, a common complaint in women with PCOS, refers to abnormal growth of the

terminal hair in a male-like pattern. 2, 5 The modified Ferriman-Gallwey (FG) score has now

become the gold standard for determining the density of terminal hairs at 9 different body

sites, (i.e., upper lip, chin, chest, upper back, lower back, upper abdomen, lower abdomen,

arm and thigh); a total score of 8 is considered as hirsutism. 8, 9 Many hirsute women are

found to have increased circulating androgen levels, although in some of them, androgen

levels are in normal ranges. 3, 9, 10 Some studies show that the severity of hirsutism correlates

poorly with the severity of androgen excess; it has not yet been clarified, which of the

androgenic parameters have the most significant relationship with FG score. 9, 11 This obvious

paradox raises the question of the role of other factors including peripheral metabolism of

androgens, sensitivity of target tissues to androgens, key enzymes for biosynthesis and

metabolism of androgens and hormonal assay methods in the development of hirsutism. 9

Although androgens modulate the biological mechanism regulating the hair cycle, limited

studies have evaluated the relationship between clinical and biochemical HA and these have

reported conflicting findings. 3, 5, 6, 9, 12-14 Confirmed associations between endocrine and

clinical characteristics in hirsute patients with PCOS have important implications for follow-

up and management of this disorder. 6, 10 Accordingly, we aimed to investigate the association

This article is protected by copyright. All rights reserved.

 between BHPs and FG score for better understanding the pathophysiology of excess hair in

PCOS women; this information will potentially be helpful for selection of treatment options.
Accepted Article
Methods

We conducted this systematic review and meta-analysis based on the Preferred Reporting

Items for Systematic Reviews and Meta-Analyses (PRISMA) 15 to investigate the association

between BHPs and FG score. The ethics committee of the Research Institute for Endocrine

Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran approved this study.

Registration number: IR.SBMU.RIES.RES.1394.90.

Eligibility criteria

We searched randomized controlled trials (RCTs), non-comparative clinical trials,

prospective and retrospective cohort studies, and case-control studies restricted to papers

published in English between 2000 and 2015. Studies were eligible if they had: 1. a study

population, including PCOS women, diagnosed by Rotterdam, National Institutes of Health

(NIH) or the Androgen Excess Society (AES) criteria, with or without hirsutism, 2. assessed

FG score and at least one androgen marker, including tT, fT, A4, DHEAS, FAI and SHBG,

and 3. used a modified FG scoring system for diagnosis of hirsutism. We also excluded non-

human studies, reviews, commentaries, editorials, letters, meeting abstracts, case reports, and

studies that did not measure FG score or any hormonal markers, or those that had a study

population with idiopathic hirsutism or the other types of HA, or did not provide accurate and

clear data or methods.

This article is protected by copyright. All rights reserved.

 Search strategy

PubMed, Scopus, Google Scholar, and ScienceDirect and Web of Sciences were searched for
Accepted Article
studies documented from 2000 to 2015 investigating androgen levels and FG scores in

patients with PCOS, using MeSH terms including the following keyword combinations:

"Polycystic ovary syndrome" AND "hirsutism" OR "hirsute" OR "androgen" OR "hormone"

OR "androstenedione" OR "dehydroepiandrosterone".

Study selection

Two reviewers, experts in reproductive endocrinology (MA and SB), independently

conducted a primary selection using the titles, abstracts, and keywords to exclude studies that

did not fulfill the inclusion criteria. They independently read the full text of remaining papers

to identify eligible articles. Disagreements between reviewers were resolved by discussion or,

if necessary, by a third reviewer (FRT). Journals and authors were not blinded during study

selection.

Data extraction

The following data were collected from each study: First author and publication year, country

of study, study design and settings, total number of patients, diagnostic criteria for PCOS,

patients characteristics including age and BMI, outcome measurements including FG score,

tT, fT, A4, DHEAS and SHBG. Data were extracted from full text articles by one reviewer

(MA) in close consultation with the second (FRT). Data of all studies were extracted with

mean and standard deviation values. 16 To prevent extraction errors, a control check between

the final data used in the meta-analysis and the original publications was performed by all

authors. Baseline data of all studies such as clinical trials were extracted.

This article is protected by copyright. All rights reserved.

 Quality assessment

All studies included for the meta-analysis were appraised for the quality of their
Accepted Article
methodological and results presentation. Two reviewers (MA and SB) who were blinded to

study author, institution and journal name, assessed the quality of the studies separately.

Disagreement was resolved and adjusted by the senior reviewer (FRT).

The modified Consolidated Standards of Reporting Trials (CONSORT) was used as a

validated quality assessment checklist for clinical trials. 17 All clinical trial papers were

categorized in four groups including high, moderate, low and very low quality; studies with

scores 70 % of the highest level of the CONSORT checklist were considered as high, 40 to

70 % as moderate, 20 to 40 % as low and < 20 % as very low quality.

Moreover, the quality assessment for case control studies was based on the modified

Strengthening the Reporting of Observational Studies in Epidemiology (STROBE). 18 The

highest level of STROBE score was considered to be 25; if a paper obtained 70 % of the

highest level of STROBE score, it was considered as high quality, 40 to 70% as moderate, 20

to 40% as low quality and < 20 % as very low quality.

In addition, cohort studies were appraised according to the NewcastleOttawa scale for

observational cohort studies. 19 In this respect, three domains were scored concerning

selection and comparability of study cohorts, and outcome of interest was ascertained using a

score range of 0 to 9. If a study obtained 70 % of the highest level of NewcastleOttawa

scale, it was considered as high quality, 40-70% as moderate, 20 to 40% as moderate and <

20 % as very low quality.

This article is protected by copyright. All rights reserved.

 Statistical analysis

This meta-analysis was conducted to obtain pooled effect size of FG score and BHPs
Accepted Article
including tT, SHBG, fT, A4, DHEAS and FAI. Chi-squared and I-squared statistics were

estimated as the measures of heterogeneity. Random effect meta-analysis was applied for

significant results (I2 greater than 50% or significant Chi2 test), otherwise fixed-effect model.

Forest plots summarize the effect sizes, by the Mantel-Haenszel method of pooled estimation,

which used variance of between studies to weight each study and the estimated pooled mean

of FG score means (mean of means). Hence, the Forest plot of FG score presents pooled

effect size (95% CI) of FG score (Figure 2) and the other forest plots show the pooled effect

size (CI) of BHPs of included studies (Figures S1-S3). To illustrate associations, we depicted

Bubble plots, which show associations between androgens and FG score by a linear

regression model (Figure 3); each study has its own contribution to estimate this linear

association, which is illustrated by bubbles, where the size of bubble is proportional to the

inverse variance. 20, 21

22 23
To assess publication bias, Begg and Eggers tests , and Funnel plot were used.

Publication bias was found to be significant for p-values < 0.05, which indicated significant

asymmetry in funnel plot, analogously. For significant results or an asymmetric Funnel plot,

the trim and fill method was used to identify and correct for publication bias by adding some

study measures. 24, 25

Furthermore, we used random effect meta-regression models to assess the observational

associations between BHPs and mean FG score across studies. First, the associations between

BHPs and FG score were estimated separately without adjusting for confounders. Then,

meta- regression models were fitted to investigate the associations between BHPs and FG

score adjusted for age and BMI, diagnostic criteria for PCOS (Rotterdam, NIH and AES),

quality assessment (moderate and high), ethnic groups (Western, Middle East, Eastern,

This article is protected by copyright. All rights reserved.

 African), study design (RCT, cohort, case-control, cross-sectional, before and after design),

and methods of hormonal assays (chemical /electrochemical luminescence,

Accepted Article
radioimmunoassay, enzyme and unknown).

In addition, we performed yet another meta-regression analysis to investigate the association

between BHPs and FG score in PCOS patients with clinical and/or HA signs, in which

studies with unknown phenotypes of Rotterdam criteria were excluded (Supplementary Data

5). A subgroup analysis was also done to evaluate the association between BHPs and FG

score based on the different methods of hormonal assay (Supplementary Data 6).

An overall test of all the covariates in the random effects model was assessed by t statistics

and p-values as well as 95% CIs for the regression coefficients (). For all results, p-values

<0.05 were considered significant. Statistical analysis was done using STATA software

(version 10; STATA, INC., college station, TX, USA).

Results

Search results and study selection

The selection process and reasons for exclusions are depicted in Figure 1. Thirty-five studies

were included in the meta-analysis, of which 19 had no comparative groups, 13 had two

comparative groups, two had three comparative groups and one study had four comparative

groups. Of the 35 studies selected, 5 had cross sectional design, 5 case-control, 6 cohort, 7

clinical trial and 12 RCT.

Quality assessment

Regarding quality, 15 studies were classified as high and 20 as moderate, while none were

identified as being of low or very low quality (Table 1). Therefore, all studies included in the

meta-analysis ranged from moderate to high quality.

This article is protected by copyright. All rights reserved.

 Study characteristics

Characteristics of the included studies are reported in Table 1. Sixteen study groups were
Accepted Article
performed in the West region, 11 in the Middle East, 7 in the East and one in Africa. All

studies were published between 2000 and 2015, and assessed a total number of 6593

participants with PCOS.

Results of meta-regression analysis

Results of univariate, weighted, linear meta-regression analysis showed significant positive

observational relationships between A4 ( = 2.2, 95% CI: 0.2-4.2; p = 0.034), DHEAS ( =

4.1, 95% CI: 1-7.3; p = 0.012) and FG scores across studies. The associations of FG score

with tT, fT, SHBG and FAI were not significant (as model 1).

Model 2 represented the relationships between BHPs and FG score adjusted for age and BMI;

there were significant positive observational relationships between DHEAS and FG score (

= 3.8, 95% CI: 1.1-6.45; p = 0.008). Other associations of FG score with tT, fT, SHBG, A4

and FAI were not significant.

Model 3 detected the associations adjusted for diagnostic criteria for PCOS; there were

significant positive associations between DHEAS and FG score ( = 4.1, 95% CI: 1.1-7.1; p

= 0.009). We found no significant associations for FG score with tT, fT, SHBG, A4, FAI

after adjusting for diagnostic criteria.

Model 4 fitted meta-regression models were adjusted for quality assessment; we detected

associations of FG score with A4 ( = 2.5, 95% CI: 0.5-4.4; p = 0.017) and DHEAS ( = 3.5,

95% CI: 0.3-6.7; p = 0.035). The association between tT, fT, SHBG, FAI and FG score were

not significant.

This article is protected by copyright. All rights reserved.

 Model 5 regressed the associations between BHPs and FG score adjusted for study design,

similar to results found for model 1. We observed significant positive relationships between
Accepted Article
DHEAS and FG score ( = 4.0, 95% CI: 0.8-7.2; p = 0.016). Other relationships between

BHPs and FG score were not significant.

Model 6, fitted BHPs and FG score were adjusted for ethnic groups; we detected a positive

significant association only between A4 and FG score ( = 2.6, 95% CI: 0.4-4.8; p = 0.023).

Finally, in model 7, associations of FG score with BHPs adjusted for the hormonal assay

methods were obtained; we detected significant positive observational relationships of FG

score with A4 ( = 2.3, 95% CI: 0.2-4.5; p = 0.032), DHEAS ( = 4.8, 95% CI: 1.1-8.4; p =

0.012). There were no significant associations between other BHPs and FG score after

adjusting for this potential confounder.

The results of meta-regression analysis models are summarized in Table 3. In addition,

Figure 3 shows the association between BHPs and FG score, which fitted meta-regression

line for FG score by BHPs using bubble plot. All included studies assessed the FG score; as

shown in the forest plot, there was significant heterogeneity in mean FG score among studies

(I-squared= 99.4%, p< 0.001) (Figure 2). In addition, we observed significant heterogeneity

for BHPs including tT (I-squared= 99.4%, p< 0.001), fT (I-squared= 99.0 %, p< 0.001),

SHBG (I-squared= 95.6%, p< 0.001), A4 (I-squared= 98.4%, p< 0.001), DHEAS (I-squared=

95.8%, p< 0.001) and FAI (I-squared= 98.2%, p< 0.001) (see forest plots in the supplemental

figures S1-S3).

This article is protected by copyright. All rights reserved.

 The results of meta-regression analysis of subgroups of studies, which assessed only PCOS

patients with clinical and/or biochemical HA showed changes in regression models including
Accepted Article
a positive significant association between SHBG and FG score in model 1 ( = 0.17, 95% CI:

0.01-0.33; p = 0.041), a positive significant association between DHEAS and FG score in

model 2 ( = 3.75, 95% CI: 0.44-7.07; p = 0.029), no significant association between BHPs

and FG score in models 3 and 5, a reverse significant association between fT and FG score (

= -0.71, 95% CI: -1.29- -0.13; p = 0.022) and a positive significant association between

SHBG and FG score ( = 0.17, 95% CI: 0.01-0.34; p = 0.041) in model 4, a reverse

significant association between FG score FAI ( = -1.22, 95% CI: -2.03- -0.41; p = 0.009) in

model 6 and a positive significant association between FG score and A4 ( = 1.90, 95% CI:

0.33-3.48; p = 0.025) in model 7 (see detailed results in the supplemental Data 5).

Results of subgroup analysis

There was no association for FG score with tT, fT, SHBG and FAI in any of the subgroups of

hormonal assay methods (chemical/electrochemical luminescence, radioimmunoassay and

enzyme). However subgroup analysis based on the methods of hormonal assay showed that

when measurements were done by radioimmunoassay, associations of FG score with A4 and

DHEAS were significant (see detailed results in the supplemental Data 6).

Publication bias

We found statistically significant publication bias only for fT; the results of Egger and

Beggs tests were not concordant in this case (Table 2); therefore, trim and fill correction

were applied, and showed that non-significant publication bias was reliable (Figure S4).

Discussion

This article is protected by copyright. All rights reserved.

 Our review demonstrated significant positive observational associations of FG score with A4

and DHEAS, whereas associations of FG score with tT, fT, SHBG and FAI were not
Accepted Article
significant. Adjustment for potential confounders including diagnostic criteria for PCOS or

method for assessment of androgens demonstrated similar results.

Although the pathogenesis of hirsutism is complex and has not been clarified 9, it is well

known that the intensity and extent of hirsutism is influenced by several factors including the

individual reaction of pilosebaceous units to androgens 6, increased levels of androgens,

androgen receptor activity, defects in 5- reductase production and other key enzymes. 26 In

addition, the severity of hirsutism can differ by age 27, race, ethnicity and familial

background. 9, 26 Although the main source of androgen synthesis is the ovary, increased

adrenal androgen production has been documented in 25% of PCOS women 28, as result of a

genetic disorder or secondary to ovarian androgen secretion. 29

The major androgens in the serum of normal women are DHEAS, dehydroepiandrosterone

(DHEA), A4, T and dihydrotestosterone (DHT) 30; T and DHT bind to the androgen receptor

to promote changes in gene transcription, whereas DHEA, DHEAS and A4 do not bind to the

androgen receptor and are considered to be pro-hormones. 9 Polymorphisms in the androgen

receptor gene also influence the activity of the receptor. 31 Circulating T in women is a result

of peripheral conversion of other steroids, produced in the ovaries and adrenals in equal parts.

DHT is the most active androgen, and is localised in androgen target tissues where the

enzyme 5- reductase catalyses T to DHT. 32 Key enzymes such as 5- reductase are

necessary for biosynthesis and metabolism of androgens; they influence circulating

androgens for growth of human sebaceous glands and hair follicles. 33 Therefore, circulating

androgen levels may not reflect local androgen concentrations at the pilosebaceous unit. In

This article is protected by copyright. All rights reserved.

 addition, local androgen effects depend on the expression of the androgen receptor in the

pilosebaceous unit. 34
Accepted Article
Based on the mechanisms mentioned, hirsutism is considered to be the result of the

interaction between circulating androgen levels, local androgen levels and the sensitivity of

hair follicles to androgens. Hence, the severity of hirsutism does not exactly correlate with

circulating androgen levels. Moreover, the hair follicle response to circulating androgens

varies significantly within and between individuals; therefore some women with elevated

androgen levels do not show hirsutism, or they may have seborrhea, acne or alopecia without

clinical hirsutism. 9

In this meta-analysis, we did not observe significant associations of FG scores with tT, fT,

FAI and SHBG. Some investigators have reported that FG score is not correlated with serum

tT; they also suggest that tT may not be helpful in reflecting biochemical HA. 5, 10 Most

physicians consider testosterone and FAI in the diagnosis of biochemical HA, whereas

increased levels of DHEAS, A4 or other androgen precursors are the only findings in many

patients. 6, 35

Some previous studies have shown that compared to testosterone, pro-hormones have more

correlation with hirsutism. Rittmaster et al. (1990), in a study of PCOS patients, found that

the change in hair growth correlated with changes in A4, but not significantly with changes in

testosterone. They concluded that hair growth rates correlated more significantly with

changes in serum A4 levels. 36 Carmina et al. (1993) showed evidence for increased

androsterone metabolism in some normoandrogenic women with clinical HA. 37

As shown in the results, FG score had positive relationships with A4 and DHEAS. A4, a

This article is protected by copyright. All rights reserved.

 weaker androgen than T, has both ovarian and adrenal origin. Although elevated A4 levels

are a frequent finding among women with PCOS, it is not often considered a main diagnostic
Accepted Article
criterion of biochemical HA. DHEAS, DHEA, and A4 are all precursors of testosterone. 38 In

severe PCOS phenotypes, circulating A4 levels may be increased, making it a valuable

marker to assess biochemical HA 39; A4 and DHEAS are of both adrenal and ovarian origin;

therefore, the existence any positive associations between FG score and these parameters can

represent an adrenal disorder. Indeed, an adrenal contribution has been reported in about 50%

of hyperandrogenaemic women with acne and/or hirsutism; a heterogeneous adrenal

androgen hyper-responsiveness to ACTH, which resembles an exaggeration of adrenarche, is

the most commonly found abnormality. 40

Since hirsutism and / or BHPs are influenced by several factors such as age, BMI, methods of

hormonal assay, diagnosis criteria, study design and method of assay, we adjusted for these

factors. The prevalence of hirsutism may decline with age, which might be due to the

progressive decline in circulating androgens in women with PCOS during their reproductive

years. 27 In addition, hormonal profiles, especially DHEAS, may change with age.7 However,

this study showed that after adjusting for age, FG score was only associated with DHEAS.

Obesity, theoretically, can exert an additive or synergistic impact on the clinical and / or

biochemical manifestations of PCOS. 5, 41 Balance probably exists between production and

clearance of androgens. 42 Obesity is a state of increased androgen production with

simultaneous accelerated clearance in hirsute women, which causes a decline in androgen

effects 43. Obesity can increase aromatase enzyme action, which converts T to E2. 44 In our

study, after adjusting for BMI, FG score was only associated with DHEAS. However, the

relationship of BMI with FG score seems weak and thus remains unclarified.

The prevalence of hirsutism differs with ethnicity 45, and after adjustment for this, only A4

This article is protected by copyright. All rights reserved.

 had a significantly positive association with FG score; a finding confirmed by the fact that in

severe PCOS phenotypes, circulating A4 levels may be increased. 39

Accepted Article
Current studies in this meta-analysis used different methods for hormonal assays. All

measurements were conducted based on 3 methods including

chemi/electrochemiluminescence, radioimmunoassay or enzyme, and variations in assays that

can influence androgen levels. As data were grouped by assay method, the between-study

heterogeneity in the subgroups was lower compared to the main analysis. However,

adjustment for methods of assay did not significantly affect associations between BHPs and

FG score. We also performed a subgroup analysis based on assay method to investigate the

association between androgens and FG score. The findings of subgroup analysis confirmed

the results, i.e. testosterone was not associated with FG score.

After adjustment for quality assessment, our results did not change; the associations of FG

score with A4 and DHEAS remained, indicating that this confounder had no significant effect

on results. The association between A4 and FG score did not remain after adjusting for

diagnostic criteria for PCOS and study design.

The majority of studies included in this analysis used the Rotterdam criteria for diagnosing

PCOS and they may have included PCOS patients with oligomenorrhoea and cystic ovaries

without clinical or biochemical HA. Hence, we conducted further meta-regression with

exclusion of studies with unknown phenotype of Rotterdam criteria to investigate the

relationship between BHPs and FG score in patients with HA phenotype. Although the results

of the subgroup analysis revealed subtle changes in models, including a significant

association of FG score with SHBG, fT and FAI, after adjustment for important confounders

such as BMI and assay method, significant associations were observed only for FG score

with A4 and DHEAS, confirming our previous results.

This article is protected by copyright. All rights reserved.

 We have not considered acne and alopecia for evaluating associations between biochemical

and clinical hyperanrogenism; there is no reliable tool for precise assessment of these
Accepted Article
parameters, there is no single scoring system, and their assessment is prone to major inter-

observer bias. In addition, various surveys have noted a relatively high prevalence of acne in

the general population, particularly among younger women. Consequently, the degree to

which PCOS increases the risk of acne to above the prevalence in the general population is

unclear. Androgen sensitivity of the pilosebaceous unit varies, and there is poor correlation

between clinical features and evidence of biochemical hyperandrogenism. 46 In addition to

androgen excess, other potential aetiologies of alopecia may be genetic, environmental, and

nutritional. Besides, limited studies have evaluated acne and alopecia as clinical

manifestations of HA and had different diagnostic methods. 7, 47, 48 Furthermore it has been

shown that women with hirsutism often have higher levels of A4, DHEAS and testosterone;

and lower levels of sex hormone-binding globulin (SHBG) than patients with alopecia and

acne. 12, 14, 49 Hence, in this meta-analysis, we considered hirsutism as the main clinical

indicator of HA.

The strengths of this meta-analysis include adjusting for potential confounders and

performing subgroup analysis based on assay methods and diagnostic criteria for PCOS. The

study, of course has its limitations that should be considered when interpreting the findings.

The majority of studies did not report the ethnicity of patients; however, we divided countries

into ethnicity groups (Western, Middle East, Eastern, and African) and adjusted for ethnicity

as a confounder. Most studies in this meta-analysis recruited participants from settings such

as hospitals and clinics, which may represent a population with more severe manifestations of

PCOS. In this study, we could not investigate the associations of FG score with BHPs,

considering all confounders simultaneously because of the limited number of included

studies.

This article is protected by copyright. All rights reserved.

 Significant heterogeneity was found in mean FG scores between studies included in the meta-

analysis, which may have resulted from clinical heterogeneity related to variability in PCOS
Accepted Article
diagnosis, race, ethnicity and/or interpreting hirsutism. Additionally, FG score is a visual and

subjective scale; inter-observer variation cannot be completely eliminated. We also observed

significant heterogeneity in mean BHPs between studies, which was predictable because of

variability in age, BMI, SHBG concentration and assay methods. 9 Moreover, we used

random effect analysis to deal with the problem. To minimize selection bias, we selected

studies based on the eligibility criteria that had been set just before the study. We also

performed a hand search on the references lists of all selected studies to prevent missing

studies.

In conclusion, the current study demonstrated positive associations of some BHPs such as A4

and DHEAS with FG score. In most analyses, DHEAS was associated with FG score. These

relationships were not causal; thus our findings confirmed a relationship between clinical and

biochemical HA. Since DHEAS and A4 levels were strongly associated with hirsutism, these

markers should be considered valuable to assess HA in patients with PCOS. The data suggest

that peripheral androgen metabolism is one of the major determinants of hirsutism in PCOS

patients.

Declaration of interest

Authors have no conflict of interest to declare.

This article is protected by copyright. All rights reserved.

 Acknowledgements

The authors would like to thank Dr. Maryam Tohidi for valuable suggestions on hormonal
Accepted Article
assay methods and Ms. Niloofar Shiva for critical editing of English grammar and syntax of

the manuscript. The authors also thank the Research Institute for Endocrine Sciences, Shahid

Beheshti University of Medical Sciences for approval of this project and its funding as the

research project.

Supporting Information

S1 Figure. Forest plots of tT and DHEAS (jpg, 118 KB).

S2 Figure. Forest plots of SHBG and A4 (jpg, 118 KB).

S3 Figure. Forest plots of FAI and fT (jpg, 91.6 KB).

S4 Figure. Funnel plots of publication bias and its corrections (jpg, 89.4 KB)

Supplementary Data 5. Meta-analysis and meta-regression for assessing the association

between BHP and FG score in patients with clinical and/or HA findings (docx, 76.1 KB).

Supplementary Data 6. Subgroup analysis for assessing the association between BHP and

FG score based on the hormonal assay methods (docx, 2.35 MB)

FIGURE CAPTIONS

Fig 1. PRISMA flow diagram of studies selection.

Fig 2. Forest plot of Ferriman Galwey score.

Fig 3. Bubble plots fitted meta-regression line for associations between biochemical

hyperandrogenism parameters and FG score. Bubble plots present the following data in order

from the top and left to right: association between SHBG and FG score, association between

This article is protected by copyright. All rights reserved.

 fT and FG score, association between tT and FG score, association between A4 and FG score,

association between DHEAS and FG score.

Accepted Article
References

1 Azziz, R., Woods, K.S., Reyna, R. et al. (2004) The prevalence and features of the

polycystic ovary syndrome in an unselected population. The Journal of Clinical

Endocrinology & Metabolism 89, 2745-2749.

2 Azziz, R., Carmina, E., Dewailly, D. et al. (2006) Criteria for defining polycystic

ovary syndrome as a predominantly hyperandrogenic syndrome: an androgen excess society

guideline. The Journal of Clinical Endocrinology & Metabolism 91, 4237-4245.

3 Coskun, A., Ercan, O., Arikan, D.C. et al. (2011) Modified FerrimanGallwey

hirsutism score and androgen levels in Turkish women. European Journal of Obstetrics &

Gynecology and Reproductive Biology 154, 167-171.

4 Buyalos, R.P., Pekonen, F., Halme, J.K. et al. (1995) The relationship between

circulating andorgens, obesity, and hyperinsulinemia on serum insulin-like growth factor

binding protein-1 in the polycystic ovarian syndrome. American journal of obstetrics and

gynecology 172, 932-939.

5 Panidis, D., Tziomalos, K., Papadakis, E. et al. (2013) The clinical significance and

primary determinants of hirsutism in patients with polycystic ovary syndrome. European

Journal of Endocrinology 168, 871-877.

6 Cibula, D., Hill, M. & Starka, L. (2000) The best correlation of the new index of

hyperandrogenism with the grade of increased body hair. European Journal of Endocrinology

143, 405-408.

This article is protected by copyright. All rights reserved.

 7 Azziz, R., Carmina, E., Dewailly, D., Diamanti-Kandarakis, E. et al. (2009) The

Androgen Excess and PCOS Society criteria for the polycystic ovary syndrome: the complete
Accepted Article
task force report. Fertility and sterility 91, 456-488.

8 Goodman, N., Bledsoe, M., Cobin, R. et al. (2001) American Association of Clinical

Endocrinologists medical guidelines for the clinical practice for the diagnosis and treatment

of hyperandrogenic disorders. Endocrine practice: official journal of the American College of

Endocrinology and the American Association of Clinical Endocrinologists 7, 120.

9 Escobar-Morreale, H., Carmina, E., Dewailly, D. et al. (2012) Epidemiology,

diagnosis and management of hirsutism: a consensus statement by the Androgen Excess and

Polycystic Ovary Syndrome Society. Human Reproduction Update 18, 146-170.

10 Amato, M.C., Galluzzo, A., Merlino, S. et al (2006) Lower insulin sensitivity

differentiates hirsute from non-hirsute Sicilian women with polycystic ovary syndrome.

European Journal of Endocrinology 155, 859-865.

11 Legro, R.S., Schlaff, W.D., Diamond, M.P. et al. (2010) Total testosterone assays in

women with polycystic ovary syndrome: precision and correlation with hirsutism. The

Journal of Clinical Endocrinology & Metabolism 95, 5305-5313.

12 Guzel, A.I., Kuyumcuolu, U. & elik, Y. (2012) Factors affecting the degree of

hirsutism in patients with polycystic ovary syndrome. Archives of gynecology and obstetrics

285, 767-770.

13 Sales, M.F., Ster, M.O., Cndido, A.L. et al. (2015) Ferriman-Gallwey Score

correlates with obesity and insulin levels in Polycystic Ovary Syndrome: an observational

study. Rev. Soc. Bras. Cln. Md 13.

14 Chanukvadze, D., Kristesashvili, J. & Kvashilava, N. (2012) Correlation of

biochemical markers and clinical signs of hyperandrogenism in women with polycystic ovary

syndrome (PCOS) and women with non-classic congenital adrenal hyperplasia (NCAH).

This article is protected by copyright. All rights reserved.

 Iranian journal of reproductive medicine 10, 307.

15 Moher, D., Liberati, A., Tetzlaff, J. et al. (2009) Preferred reporting items for
Accepted Article
systematic reviews and meta-analyses: the PRISMA statement. Annals of internal medicine

151, 264-269.

16 Welt, C., Gudmundsson, J., Arason, G. et al. (2006) Characterizing discrete subsets of

polycystic ovary syndrome as defined by the Rotterdam criteria: the impact of weight on

phenotype and metabolic features. The Journal of Clinical Endocrinology & Metabolism 91,

4842-4848.

17 Moher, D., Hopewell, S., Schulz, K.F. et al. (2010) CONSORT 2010 explanation and

elaboration: updated guidelines for reporting parallel group randomised trials. Journal of

clinical epidemiology 63, e1-e37.

18 Von Elm, E., Altman, D.G., Egger, M. et al. (2007) The Strengthening the Reporting

of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting

observational studies. Preventive medicine 45, 247-251.

19 Quigley, J., Thompson, J., Halfpenny, N. et al. (2014) Critical appraisal of nonn-

randomized controlled studies-a review of recommended and commonly used tools.

assessment 4, 4.

20 Higgins, J.P., Thompson, S.G., Deeks, J.J. et al. (2003) Measuring inconsistency in

meta-analyses. Bmj 327, 557-560.

21 MichaelBorenstein, L.V.H., Julian P.T. et al. (2009) Introduction to Meta-Analysis.

John Wiley & Sons, Ltd.

22 Begg, C.B. & Mazumdar, M. (1994) Operating characteristics of a rank correlation

test for publication bias. Biometrics, 1088-1101.

23 Egger, M., Smith, G.D., Schneider, M. et al. (1997) Bias in meta-analysis detected by

a simple, graphical test. Bmj 315, 629-634.

This article is protected by copyright. All rights reserved.

 24 Taylor, S. & TWeedie, R. (1998) Trim and fill: a simple funnel plot based method of

testing and adjusting for publication bias in meta-analyses. Fort Collins, CO: Colorado State
Accepted Article
University.

25 Duval, S. & Tweedie, R. (2000) Trim and fill: a simple funnelplotbased method of

testing and adjusting for publication bias in metaanalysis. Biometrics 56, 455-463.

26 Zhang, H.Y., Guo, C.X., Zhu, F.F. et al. (2013) Clinical characteristics, metabolic

features, and phenotype of Chinese women with polycystic ovary syndrome: a large-scale

casecontrol study. Archives of gynecology and obstetrics 287, 525-531.

27 Panidis, D., Tziomalos, K., Macut, D. et al. (2012) Cross-sectional analysis of the

effects of age on the hormonal, metabolic, and ultrasonographic features and the prevalence

of the different phenotypes of polycystic ovary syndrome. Fertility and sterility 97, 494-500.

28 Tsilchorozidou, T., Overton, C. & Conway, G.S. (2004) The pathophysiology of

polycystic ovary syndrome. Clinical endocrinology 60, 1-17.

29 Moran, C. & Azziz, R. (2001) The role of the adrenal cortex in polycystic ovary

syndrome. Obstetrics and gynecology clinics of North America 28, 63-75.

30 Burger, H.G. (2002) Androgen production in women. Fertility and sterility 77, 3-5.

31 Calvo, R.M., Asuncin, M., Sancho, J. et al. (2000) The Role of the CAG Repeat

Polymorphism in the Androgen Receptor Gene and of Skewed X-Chromosome Inactivation,

in the Pathogenesis of Hirsutism 1. The Journal of Clinical Endocrinology & Metabolism 85,

1735-1740.

32 Longcope, C. (1986) 1 Adrenal and gonadal androgen secretion in normal females.

Clinics in endocrinology and metabolism 15, 213-228.

33 Thiboutot, D., Gilliland, K., Cong, Z. et al. (2003) Human skin is a steroidogenic

tissue: steroidogenic enzymes and cofactors are expressed in epidermis, normal sebocytes,

and an immortalized sebocyte cell line (SEB-1). Journal of Investigative Dermatology 120,

This article is protected by copyright. All rights reserved.

 905-914.

34 Chen, W.-C. & Zouboulis, C.C. (2009) Hormones and the pilosebaceous unit.
Accepted Article
Dermato-endocrinology 1, 81-86.

35 Azziz, R., Sanchez, L., Knochenhauer, E. et al. (2004) Androgen excess in women:

experience with over 1000 consecutive patients. The Journal of Clinical Endocrinology &

Metabolism 89, 453-462.

36 Rittmaster, R.S. & Thompson, D.L. (1990) Effect of leuprolide and dexamethasone

on hair growth and hormone levels in hirsute women: the relative importance of the ovary

and the adrenal in the pathogenesis of hirsutism. The Journal of Clinical Endocrinology &

Metabolism 70, 1096-1102.

37 Carmina, E. & Lobo, R. (1993) Evidence for increased androsterone metabolism in

some normoandrogenic women with acne. The Journal of Clinical Endocrinology &

Metabolism 76, 1111-1114.

38 Doi, S.A. (2008) Neuroendocrine dysfunction in PCOS: a critique of recent reviews.

Clinical medicine & research 6, 47-53.

39 Georgopoulos, N.A., Papadakis, E., Armeni, A.K. et al. (2014) Elevated serum

androstenedione is associated with a more severe phenotype in women with polycystic ovary

syndrome (PCOS). Hormones (Athens) 13, 213-221.

40 Lucky, A.W., Rosenfield, R.L., Mcguire, J. et al. (1986) Adrenal Androgen

Hyperresponsiveness to Adrenocorticotropin in Women with Acne and/or Hirsutism: Adrenal

Enzyme Defects and Exaggerated Adrenarche*. The Journal of Clinical Endocrinology &

Metabolism 62, 840-848

This article is protected by copyright. All rights reserved.

 41 Liou, T.-H., Yang, J.-H., Hsieh, C.-H. et al. (2009) Clinical and biochemical

presentations of polycystic ovary syndrome among obese and nonobese women. Fertility and
Accepted Article
sterility 92, 1960-1965.

42 Kirschner, M.A., Samojlik, E. & Silber, D. (1983) A comparison of androgen

production and clearance in hirsute and obese women. Journal of steroid biochemistry 19,

607-614.

43 Samojlik, E., Kirschner, M., Silber, D. et al. (1984) Elevated production and

metabolic clearance rates of androgens in morbidly obese women. The Journal of Clinical

Endocrinology & Metabolism 59, 949-954.

44 Cohen, P. (2001) Aromatase, adiposity, aging and disease. The hypogonadal-

metabolic-atherogenic-disease and aging connection. Medical hypotheses 56, 702-708.

45 Tehrani, F.R., Simbar, M., Tohidi, M. et al. (2011) The prevalence of polycystic

ovary syndrome in a community sample of Iranian population: Iranian PCOS prevalence

study. Reproductive Biology and Endocrinology 9, 1.

46 Deplewski, D. & Rosenfield, R.L. (2000) Role of hormones in pilosebaceous unit

development. Endocrine reviews 21, 363-392.

47 Rosenfield, R. & Lucky, A. (1993) Acne, hirsutism, and alopecia in adolescent girls.

Clinical expressions of androgen excess. Endocrinology and metabolism clinics of North

America 22, 507-532.

48 Lucky, A.W., McGuire, J., Rosenfield, R.L., Lucky, P.A. et al. (1983) Plasma

androgens in women with acne vulgaris. Journal of Investigative Dermatology 81, 70-74.

This article is protected by copyright. All rights reserved.

 49 Karrer-Voegeli, S., Rey, F., Reymond, M.J., Meuwly, J.-Y. et al. (2009) Androgen

dependence of hirsutism, acne, and alopecia in women: retrospective analysis of 228 patients
Accepted Article
investigated for hyperandrogenism. Medicine 88, 32-45.

50 Elter, K., Imir, G. & Durmusoglu, F. (2002) Clinical, endocrine and metabolic effects

of metformin added to ethinyl estradiolcyproterone acetate in non-obese women with

polycystic ovarian syndrome: a randomized controlled study. Human Reproduction 17, 1729-

1737.

51 Mastorakos, G., Koliopoulos, C. & Creatsas, G. (2002) Androgen and lipid profiles in

adolescents with polycystic ovary syndrome who were treated with two forms of combined

oral contraceptives. Fertility and sterility 77, 919-927.

52 Guido, M., Romualdi, D., Giuliani, M. et al. (2004) Drospirenone for the treatment of

hirsute women with polycystic ovary syndrome: a clinical, endocrinological, metabolic pilot

study. The Journal of Clinical Endocrinology & Metabolism 89, 2817-2823.

53 Hahn, S., Janssen, O.E., Tan, S. et al. (2005) Clinical and psychological correlates of

quality-of-life in polycystic ovary syndrome. European Journal of Endocrinology 153, 853-

860.

54 Palep-Singh, M., Mook, K., Barth, J. et al. (2004) An observational study of Yasmin

in the management of women with polycystic ovary syndrome. Journal of Family Planning

and Reproductive Health Care 30, 163-165.

55 Pehlivanov, B. & Mitkov, M. (2007) Efficacy of an oral contraceptive containing

drospirenone in the treatment of women with polycystic ovary syndrome. The European

Journal of Contraception & Reproductive Health Care 12, 30-35.

This article is protected by copyright. All rights reserved.

 56 Chae, S.J., Kim, J.J., Choi, Y.M. et al. (2008) Clinical and biochemical characteristics

of polycystic ovary syndrome in Korean women. Human Reproduction 23, 1924-1931.

Accepted Article
57 Adali, E., Yildizhan, R., Kurdoglu, M. et al. (2008) The relationship between clinico-

biochemical characteristics and psychiatric distress in young women with polycystic ovary

syndrome. Journal of International Medical Research 36, 1188-1196.

58 Wu, J., Zhu, Y., Jiang, Y. & Cao, Y. (2008) Effects of metformin and ethinyl

estradiol-cyproterone acetate on clinical, endocrine and metabolic factors in women with

polycystic ovary syndrome. Gynecological Endocrinology 24, 392-398.

59 zdemir, S., Grkemli, H., Gezgin, K. et al. (2008) Clinical and metabolic effects of

medroxyprogesterone acetate and ethinyl estradiol plus drospirenone in women with

polycystic ovary syndrome. International Journal of Gynecology & Obstetrics 103, 44-49.

60 Landay, M., Huang, A. & Azziz, R. (2009) Degree of hyperinsulinemia, independent

of androgen levels, is an important determinant of the severity of hirsutism in PCOS. Fertility

and sterility 92, 643-647.

61 Mahmood, M., El-Kattan, E.A., EL-Aal, H.A. et al. (2009). Evaluation of the Clinical

and Biochemical Effects of Medical Therapy in Women with Polycystic Ovary Syndrome.

Med. J. Cairo Univ 77, 58-67.

62 Kriplani, A., Periyasamy, A.J., Agarwal, N. et al. (2010) Effect of oral contraceptive

containing ethinyl estradiol combined with drospirenone vs. desogestrel on clinical and

biochemical parameters in patients with polycystic ovary syndrome. Contraception 82, 139-

146.

63 Taheripanah, R., Sepahvandi, M., Entezari, A. et al. (2010) Evaluation of serum PSA

after cyproterone compound treatment compared with oral contraceptive pill in hirsute

polycystic ovary syndrome patients. Middle East Fertility Society Journal 15, 159-162.

This article is protected by copyright. All rights reserved.

 64 Colonna, L., Pacifico, V., Lello, S. et al. (2012) Skin improvement with two different

oestroprogestins in patients affected by acne and polycystic ovary syndrome: clinical and
Accepted Article
instrumental evaluation. Journal of the European Academy of Dermatology and Venereology

26, 1364-1371.

65 Naka, K.K., Kalantaridou, S.N., Bechlioulis, A. et al. (2011) Effect of

ethinylestradiol/cyproterone acetate on endothelial function in young non-obese women with

polycystic ovary syndrome: a pilot study. Gynecological Endocrinology 27, 615-621.

66 Liang, S.-J., Hsu, C.-S., Tzeng, C.-R. et al. (2011) Clinical and biochemical

presentation of polycystic ovary syndrome in women between the ages of 20 and 40. Human

Reproduction 26, 3443-3449.

67 Misichronis, G., Georgopoulos, N., Marioli, D. et al. (2012) The influence of obesity

on androstenedione to testosterone ratio in women with polycystic ovary syndrome (PCOS)

and hyperandrogenemia. Gynecological Endocrinology 28, 249-252.

68 Alanbay, I., Ercan, C.M., Sakinci, M. et al. (2012) A macrophage activation marker

chitotriosidase in women with PCOS: does low-grade chronic inflammation in PCOS relate

to PCOS itself or obesity? Archives of gynecology and obstetrics 286, 1065-1071.

69 Ilie, I.R., Marian, I., Mocan, T. et al. (2012) Ethinylestradiol 30g-drospirenone and

metformin: could this combination improve endothelial dysfunction in polycystic ovary

syndrome? BMC endocrine disorders 12, 9.

70 Karabulut, A., Demirlenk, S. & evket, O. (2012) Effects of ethinyl estradiol-

cyproterone acetate treatment on metabolic syndrome, fat distribution and carotid intima

media thickness in polycystic ovary syndrome. Gynecological Endocrinology 28, 245-248.

71 Bhattacharya, S.M. & Jha, A. (2012) Comparative study of the therapeutic effects of

oral contraceptive pills containing desogestrel, cyproterone acetate, and drospirenone in

patients with polycystic ovary syndrome. Fertility and sterility 98, 1053-1059.

This article is protected by copyright. All rights reserved.

 72 Romualdi, D., De Cicco, S., Busacca, M. et al. (2013) Clinical efficacy and metabolic

impact of two different dosages of ethinyl-estradiol in association with drospirenone in

Accepted Article
normal-weight women with polycystic ovary syndrome: a randomized study. Journal of

endocrinological investigation 36, 636-641.

73 Aydin, K., Cinar, N., Aksoy, D.Y. et al. (2013) Body composition in lean women with

polycystic ovary syndrome: effect of ethinyl estradiol and drospirenone combination.

Contraception 87, 358-362.

74 Quinn, M., Shinkai, K., Pasch, L. et al. (2014) Prevalence of androgenic alopecia in

patients with polycystic ovary syndrome and characterization of associated clinical and

biochemical features. Fertility and sterility 101, 1129-1134.

75 Kahraman, K., kr, Y.E., Atabekolu, C.S. et al. (2014) Comparison of two oral

contraceptive forms containing cyproterone acetate and drospirenone in the treatment of

patients with polycystic ovary syndrome: a randomized clinical trial. Archives of gynecology

and obstetrics 290, 321-328.

76 Yildizhan, R., Gokce, A.I., Yildizhan, B. et al. (2015) Comparison of the effects of

chlormadinone acetate versus drospirenone containing oral contraceptives on metabolic and

hormonal parameters in women with PCOS for a period of two-year follow-up.

Gynecological Endocrinology 31, 396-400.

This article is protected by copyright. All rights reserved.

ccepted Articl
Table 1. Characteristics of studies included in the meta-analysis.
First author, year Country Study design Sample
size (n)
Diagnostic
criteria for
Age (mean
SD)
BMI (mean
SD)
FG score
(mean SD)
BHPs Measured QA

PCOS

Elter, 2002 50 Turkey RCT 20 Rotterdam 23.456.07 21.831.40 12.065.25 tTa, SHBGa, fTb, A4b, High
DHEASa

Elter, 2002 50 Turkey RCT 20 Rotterdam 24.906.62 22.742.66 9.475.48 tTa, SHBGa, fTb, A4b, High
DHEASa

Mastorakos, 2002 Italy RCT 14 NIH 17.530.51 25.501.79 15.711.63 tTb, SHBGb, fTb, A4b, Moderate
51
DHEASb

Mastorakos, 2002 Italy RCT 14 NIH 17.460.43 24.841.09 16.781.15 tTb, SHBGb, fTb, A4b, Moderate
51
DHEASb

Guido, 2004 52 Italy Clinical trial 15 Rotterdam 25.303.51 24.685.95 16.205.81 tTb, SHBGb, A4b, DHEASb High

Hahn, 2005 53 Germany Case-control 120 NIH 24.005.40 31.009.30 9.20+5.80 tTa Moderate

Palep-singh, 2005 UK Clinical trial 17 Rotterdam 26.405.30 29.003.70 18.502.74 tTd, SHBGd Moderate
54

Amato, 2006 10 Italy R-Cohort 75 Rotterdam 24.725.83 31.197.36 17.949.27 tTc, fTc, A4a, DHEASa Moderate

Amato, 2006 10 Italy R-Cohort 55 Rotterdam 24.354.28 28.706.16 6.502.02 tTc, fTc, A4a, DHEASa Moderate

Pehlivanov, 2007 55 Bulgaria Clinical trial 20 Rotterdam 23.004.10 26.450.65 11.600.90 tTc, SHBGc, DHEASc Moderate

Chae, 2008 56 Korea Case-control 87 Rotterdam 25.405.70 22.805.10 8.703.10 tTb, SHBGb, fTb, DHEASb High

Chae, 2008 56 Korea Case-control 52 Rotterdam 25.704.90 20.904.40 3.103.00 tTb, SHBGb, fTb, DHEASb High

Chae, 2008 56 Korea Case-control 23 Rotterdam 23.406.90 23.104.80 10.503.00 tTb, SHBGb, fTb, DHEASb High

Adali, 2008 57 Turkey P-Cohort 42 Rotterdam 23.543.13 28.424.30 12.423.40 tTa, DHEASa High

This article is protected by copyright. All rights reserved.

ccepted Articl
Wu, 2008 58

Wu, 2008 58
China

China
RCT

RCT
7

12
Rotterdam

Rotterdam
25.004.30

26.104.60
25.300.80

21.401.60
8.301.00

7.801.30
tTa

tTa
High

High

Wu, 2008 58 China RCT 6 Rotterdam 24.502.40 25.201.00 8.201.40 tTa High

Wu, 2008 58 China RCT 10 Rotterdam 25.804.00 21.601.40 7.802.10 tTa High

Ozdemir, 2008 59 Turkey RCT 32 Rotterdam 22.703.80 22.703.80 10.403.60 tTa, SHBGa, fTa Moderate

Liou, 2009 41 Taiwan R-Cohort 115 Rotterdam 27.605.50 31.204.40 3.103.80 tTb High

Liou, 2009 41 Taiwan R-Cohort 180 Rotterdam 26.205.20 20.302.10 3.103.30 tTb High

Landay, 2009 60 USA Cross-sectional 749 NIH 27.507.40 33.609.30 8.234.92 tTb, fTb, DHEASb High

Mahmood, 2009 61 Egypt RCT 21 AES 26.052.44 29.422.02 9.711.85 fTa, DHEASa Moderate

Kriplani, 2010 62 India RCT 30 Rotterdam 22.504.70 22.702.30 12.604.50 tTa, SHBGa High

Taheripanah, 2010 Iran RCT 30 Rotterdam 22.900.50 21.102.06 10.782.40 fTc, DHEASc High
63

Taheripanah, 2010
63 Iran RCT 30 Rotterdam 23.900.6 21.702.76 11.502.30 High
fTc, DHEASc
1

Colonna, 2012 64 Italy RCT 32 Rotterdam 25.604.40 23.353.60 16.212.71 tTb, SHBGb, A4b, DHEASb Moderate

Colonna, 2012 64 Italy RCT 27 Rotterdam 25.403.40 23.273.63 16.092.54 tTb, SHBGb, A4b, DHEASb Moderate

Naka, 2011 65 Greece Clinical trial 13 Rotterdam 20.903.70 23.004.00 8.502.20 tTa, SHBGa Moderate

Guzel, 2011 12 Turkey Case-control 224 Rotterdam 24.194.44 26.674.22 12.523.28 tTa Moderate

Guzel, 2011 12 Turkey Case-control 80 Rotterdam 23.104.32 25.233.09 4.311.79 tTa Moderate

This article is protected by copyright. All rights reserved.

ccepted Articl
Liang, 2011 66

Coskun, 2011 3
Taiwan

Turkey
R-Cohort

P-Cohort
453

43
Rotterdam

Rotterdam
27.904.70

26.050.76
25.706.60

29.270.84
3.803.80

11.204.20
tTb, DHEASd

tTa, DHEASa
High

Moderate

Chanukvadze, 2012 Georgia P-Cohort 24 AES 17.303.50 26.306.80 16.405.60 tTc, SHBGc, fTc, DHEASc Moderate
14

Misichronis, 2012 Greece Case-control 1087 Rotterdam 23.405.60 26.707.00 8.504.70 tTb, SHBGb, A4b, DHEASb Moderate
67

Alanbay, 2012 68 Turkey Cross-sectional 25 Rotterdam 27.304.00 23.101.00 10.103.90 tTc, fTc High

Alanbay, 2012 68 Turkey Cross-sectional 54 Rotterdam 25.303.70 27.801.90 10.203.20 tTc, fTc High

Iiie, 2012 69 Romania Clinical trial 25 AES 22.760.83 28.441.24 9.721.06 tTc, SHBGc High

Karabulut, 2012 70 Turkey Clinical trial 6 Rotterdam 22.406.30 28.554.24 12.003.40 tTd, DHEASd Moderate

Bhattcharia, 2012 India RCT 58 Rotterdam 22.244.47 25.414.49 5.554.51 tTa, SHBGc Moderate
71

Bhattcharia, 2012 India RCT 56 Rotterdam 22.324.17 26.413.81 6.845.17 tTa, SHBGc Moderate
71

Bhattcharia, 2012 India RCT 57 Rotterdam 22.334.76 26.474.65 6.145.15 tTa, SHBGc Moderate
71

Romualdi, 2013 72 Italy RCT 13 Rotterdam 22.923.80 22.133.34 11.625.66 tTa, SHBGa, A4a, DHEASa Moderate

Romualdi, 2013 72 Italy RCT 13 Rotterdam 21.922.43 22.652.75 11.425.88 tTa, SHBGa, A4a, DHEASa Moderate

Aydin, 2013 73 Turkey Clinical trial 28 Rotterdam 21.404.20 21.803.40 9.258.03 tTd, SHBGd, A4d, DHEASd Moderate

Panidis, 2013 5 Greece Cross-sectional 780 Rotterdam 23.805.60 27.307.00 11.402.50 tTb, SHBGb, A4b, DHEASb High

Panidis, 2013 5 Greece Cross-sectional 517 Rotterdam 25.105.90 25.906.80 3.402.40 tTb, SHBGb, A4b, DHEASb High

Zhang, 2013 26 China Case-control 719 Rotterdam 25.543.28 23.673.57 7.304.10 tTa Moderate

This article is protected by copyright. All rights reserved.

ccepted Articl
Quinn, 2014 74

Quinn, 2014 74
USA

USA
Cross-sectional

Cross-sectional
56

198
Rotterdam

Rotterdam
28.805.67

27.806.10
30.908.45

30.108.17
9.276.43

8.136.04
tTb, SHBGb, A4b, DHEASb

tTb, SHBGb, A4b, DHEASb

High

High

Kahraman, 2014 75 Turkey RCT 19 AES 22.303.20 23.203.30 15.807.30 tTb, SHBGb, fTb, A4b, High
DHEASd

Kahraman, 2014 75 Turkey RCT 20 AES 22.402.70 23.205.50 15.307.30 tTb, SHBGb, fTb, A4b, High
DHEASd

Yildizhan, 2015 76 Turkey RCT 60 Rotterdam 25.362.91 24.821.08 9.984.86 tTa, SHBGa, DHEASb Moderate

Yildizhan, 2015 76 Turkey RCT 60 Rotterdam 24.823.20 23.563.32 5.563.44 tTa, SHBGa, DHEASb Moderate

Sales, 2015 13 Brazil Cross-sectional 50 Rotterdam 31.005.00 28.905.30 11.007.00 tTa Moderate

Abbreviations: PCOS, polycystic ovary syndrome; SD, standard deviation; BMI, body mass index; RCT, randomised controlled trial; P-cohort, prospective cohort; R-Cohort, retrospective
cohort; NIH, National Institutes of Health; AES, Androgen Excess Society; BHP, biochemical hyperandrogenism parameter; tT, total testosterone; fT, free testosterone; SHBG, sex hormone
bonding globulin; A4, androstenedione; DHEAS, dehydroepiandrosterone; QA, quality assessment.

a, measured by chemi/elecrtrochemiluminescence method

b, measured by radioimmunoassay method

c, measured by enzyme method

d, unknown method of assay

This article is protected by copyright. All rights reserved.

 Table 2. Results of publication bias estimated for clinical / Biochemical hyperandrogenism parameters.
& $
Parameters I squared Chi squared Z-value* Egger test Beggs test Pooled effect size
p-value
Accepted Article
Value (%) Value (df$) (p-value) P-value (95%CI)

(p-value)

FG score 99.40 8393 (54) 209.22 0.10 0.119 8.30

(0.000**) (0.000**) (8.20-8.40)

tT 99.20 6388 (51) 228.69 0.273 0.138 0.80

(0.000**)
(0.000**) (0.60-0.80)

fT 99.10 2116 (18) 90.15 2.70

(0.000**) (0.000**)
0.073 0.023* (2.70-2.80)

SHBG 95.60 700.40 (31) 119.58 38.90

(0.000**) (0.000**) 0.588 0.697 (38.20-39.50)

A4 98.40 1153.73 (18) 158.89 0.478 0.478 2.90

(0.000**) (0.000**) (2.80-2.90)

DHEAS 95.80 777.80 (33) 164.73 2.50

(0.000**) (0.000**) 0.346 0.346 (2.50-2.50)

FAI 98.20 984.02 (18) 86.98 0.998 0.998 8.20

(0.000**) (0.000**) (8.0-8.40)

Abbreviations: FG score, Feriman-Gallwey; tT, total testosterone; fT, free testosterone; SHBG, sex hormone bonding
globulin; A4, androstenedione; DHEAS, dehydroepiandrosterone sulfate and FAI, free androgen index.
$
Degree of freedom for chi squared test = number of included studies - 1
**Statistical significance level P < 0.05
* Overall mean equal to zero z- test.
&
Egger test of publication bias to assess small-study effects.
$
Beggs test: an adjusted rank correlation test to detect the association between estimated effect size and its variance the
based on Kendalls tau.

Table 3. Meta-regression results for univariate and multiple (adjusted effect) models assessing the observational
relationships between BHPs and FG score.

This article is protected by copyright. All rights reserved.

 1 2 3 4 5 6 7
Parameter (SE)

(P-value) (P-value) (P-value) (P-value) (P-value) (P-value) (P-value)

(95% CI for ) (95% CI for ) (95% CI for ) (95% CI for ) (95% CI for ) (95% CI for ) (95% CI for )
Accepted Article
tT mean 1.2 -0.3 0.8 1 1.2 -0.5 1.4

(0.613) (0.912) (0.732) (0.678) (0.638) )0.817( )0.565(

(-3.6-6.1) (-5.0-4.5) (-3.8-5.4) (-3.8-5.7) (-3.8-6.1) (-4.8-3.8) (-3.5-6.3)

fT mean -0.1 0.1 -0.1 -0.1 -0.1 -0.1 -0.1

)0.594( )0.598( )0.581( )0.779( )0.558( )0.733( )0.721(

(-0.6-0.4) (-0.3-0.5) (-0.6- 0.3) (-0.6-0.4) (-0.6-0.3) (-0.5-0.4) (-0.9-0.6)

SHBG 0.1 0.1 0.02 0.1 0.1 0.03 0.06

mean
)0.252( )0.263( )0.661( )0.175( )0.269( )0.666( )0.390(

(-0.1-0.2) (-0.1-0.2) (-0.1-0.2) (-0.04-0.2) (-0.1-0.2) (-0.1- 0.2) (-0.08-0.2)

A4 mean 2.2 2.2 1.7 2.5 1.9 2.6 2.3

)0.034*( )0.127( 0.169 )0.017*( )0.051( )0.023*( )0.032*(

(0.2-4.2) (-0.7- 5.2) (-0.8-4.2) (0.5-4.4) (-0.01-3.9) (0.4-4.8) (0.2-4.5)

DHEAS 4.1 3.8 4.1 3.5 4.0 3.2 4.8

mean
)0.012*( )0.008*( )0.009*( )0.035*( )0.016*( )0.067( )0.012*(
)-0.2-6.6(
(1-7.3) (1.1-6.45) (1.1-7.1) (0.3-6.7) (0.8-7.2) (1.1-8.4)

FAI mean 0.2 0.3 0.2 0.2 0.3 0.01 0.3

(0.518) (0.442) (0.540) (0.619) (0.422) (0.976) (0.374)

(-0.4-0.8) (-0.5-1.1) (-0.5-0.8) (-0.5-0.8) (-0.4-0.9) (-0.6-0.6) (-0.4-1.0)

* Statistical significance level P<0.05.

1
model1; Univariate Meta Regression of FG score mean and BHPs.
2
model2; Multiple Meta Regression of FG score mean and BHPs adjusted for age mean and BMI mean.
3
model3; Multiple Meta Regression of FG score mean and BHPs adjusted for diagnosis criteria for PCOS.
4
Model4; Multiple Meta Regression of FG score mean and BHPs adjusted for quality assessment
5
model5; Multiple Meta Regression of FG score mean and BHPs adjusted for study design.
6
model6; Multiple Meta Regression of FG score mean and BHPs adjusted for ethnicity.
7
model7; Multiple Meta Regression of FG score mean and BHPs adjusted for method of assay.

This article is protected by copyright. All rights reserved.

Accepted Article

This article is protected by copyright. All rights reserved.

Accepted Article

This article is protected by copyright. All rights reserved.

Accepted Article

This article is protected by copyright. All rights reserved.