See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/266144050

Efficacy of Metformin in Obese and Non-obese Women With Polycystic Ovary

Syndrome: A Randomized, Double-Blinded, Placebo-Controlled Cross-Over
Trial

Article  in  Obstetrical and Gynecological Survey · February 2008

DOI: 10.1097/01.ogx.0000300471.31456.ea

CITATIONS READS

20 110

4 authors:

Birgitta Trolle Allan Flyvbjerg

Aarhus University Aarhus University
26 PUBLICATIONS   628 CITATIONS    655 PUBLICATIONS   29,620 CITATIONS   

SEE PROFILE SEE PROFILE

Ulrik Schiøler Kesmodel Finn Friis Lauszus

University of Copenhagen Herlev Hospital Sygehus Sonderjylland
226 PUBLICATIONS   7,213 CITATIONS    264 PUBLICATIONS   1,502 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Granulosa cell tumor of the ovary View project

Birth Environment of the future - A randomized controlled trial View project

All content following this page was uploaded by Finn Friis Lauszus on 13 October 2017.

The user has requested enhancement of the downloaded file.

Human Reproduction Vol.22, No.11 pp. 2967–2973, 2007 doi:10.1093/humrep/dem271
Advance Access publication on August 31, 2007

Efficacy of metformin in obese and non-obese women with

polycystic ovary syndrome: a randomized, double-blinded,
placebo-controlled cross-over trial

B. Trolle1,5, A. Flyvbjerg2, U. Kesmodel3 and F.F. Lauszus4

1
Department of Obstetrics and Gynaecology, Aarhus University Hospital, Skejby, DK-8200 Aarhus N, Denmark; 2The Medical
Research Laboratories, Clinical Institute and Medical Department M (Diabetes and Endocrinology), Aarhus University Hospital,
DK-8000 Aarhus C, Denmark; 3Institute of Public Health, Department of Epidemiology, University of Aarhus, DK-8000 Aarhus C,
Denmark; 4Department of Obstetrics and Gynaecology, Herning Hospital, DK-7400 Herning, Denmark

Downloaded from http://humrep.oxfordjournals.org/ at Fagbiblioteket, Sygehus Viborg on January 9, 2014

5
Correspondence address. Tel: þ45-8949-6304; Fax: þ45-8949-6022; E-mail:ditte.trolle@dadlnet.dk

BACKGROUND: Our aim was to assess the effects of metformin on menstrual frequency, fasting plasma glucose
(FPG), insulin resistance assessed as HOMA-index, weight, waist/hip ratio, blood pressure (BP), serum lipids, and tes-
tosterone levels in women with polycystic ovary syndrome (PCOS) METHODS: In a randomized, controlled, double-
blinded setup, 56 women aged 18 –45 with PCOS were treated with either metformin 850 mg or placebo twice daily for
6 months. After a wash-out period of 3 months participants received the alternate treatment for 6 months. The
changes in the measured parameters were analysed by intention-to-treat and per protocol RESULTS: There were
no changes in menstrual frequency. In the intention-to-treat analysis, weight and systolic BP were reduced on metfor-
min treatment (p50.009 and 0.047, respectively), while high-density lipoprotein (HDL) increased (p50.001). On
placebo, weight and FPG increased (p<0.05). Post-hoc subgrouping according to BMI revealed reductions in testos-
terone (p50.013), FPG (p50.018), insulin (p50.045) and HOMA-index (p50.022) in obese women. Per protocol
analysis showed the following differences between the changes on placebo and metformin (mean (5 - 95 % percentiles):
weight (-4.2 (-7.0, -1.9) kg, p<0.001), FPG (-0.23 (-0.44, -0.01) mmol/l, p50.041), insulin (-4.17 (-8.10, -0.23) mIU/l,
p50.039) and HOMA index (-1.50 (-2.53, -0.47) mIU/l*mmol/l, p50.006). Weight, FPG and HOMA index were
lower after metformin than after placebo. CONCLUSIONS: Metformin treatment lowered weight and systolic
blood pressure and increased HDL in women with PCOS. In post-hoc analysis it increased insulin sensitivity and
lowered testosterone in obese women. Non-obese women did not benefit from metformin.

Keywords: polycystic ovary syndrome; metformin; obesity; randomized controlled trial

Introduction
Polycystic ovary syndrome (PCOS) is characterized by
oligo- or anovulation, clinical and/or biochemical signs of
hyperandrogenism and polycystic ovaries. The severity of the
symptoms is weight dependent (Vanky et al., 2004). Women
with PCOS show features of the metabolic syndrome (Holte
et al., 1998; Legro et al., 1999; Glueck et al., 2003), a constel-
lation of metabolic abnormalities associated with an increased
risk of type 2 diabetes mellitus and cardiovascular disease.
A disturbed balance between insulin sensitivity and b-cell
activity has been described in both obese and non-obese
women with PCOS (Dunaif et al., 1989) and .50% have
hyperinsulinemia and are at increased risk of developing type
2 diabetes.
Metformin reduces the incidence of diabetes in pre-diabetic
subjects (Diabetes Prevention Program Research Group, 2002)
and lowers body weight in patients with and without type 2 dia-
betes. When metformin was introduced as treatment for PCOS,
observational studies showed encouraging results on menstrual
pattern and infertility (Glueck et al., 1999). In randomized con-
trolled studies, the effect of metformin is less convincing, but
still positive on ovulation (Nestler et al., 1998; Moghetti
et al., 2000), body weight (Fleming et al., 2002; Gambineri
et al., 2004), insulin sensitivity, androgen levels and hirsutism
(Kelly and Gordon, 2002; Harborne et al., 2003b; Gambineri
et al., 2004; Ganie et al., 2004). A recent meta-analysis con-
cluded that metformin is an effective treatment for anovulation

# The Author 2007. Published by Oxford University Press on behalf of the European Society of Human Reproduction and Embryology. 2967
All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org
Trolle et al.
in women with PCOS, and its choice as first line agent seemed
justified (Lord et al., 2003). However, conception rate and live
birth rate are lower when women are treated with metformin
alone than with clomiphene alone or a combination of metfor-
min and clomiphene (Legro et al., 2007). Thus, metformin is
not effective in all women, and caution has been raised
against indiscriminate use in PCOS (Harborne et al., 2003a).
We wanted to further elucidate the effects of metformin in
PCOS by performing a randomized, double-blinded, placebo-
controlled cross-over study. The primary end-points were the
changes in menstrual frequency and androgens, notably free
testosterone, whereas the secondary end-points were changes
in insulin resistance, weight, lipids and blood pressure. We
report here the effect of metformin on these parameters.
Materials and Methods
Study population
The study was performed at the Department of Gynaecology and
Obstetrics, Holstebro Hospital, Holstebro, Denmark. Women 18–45
years of age, referred to the outpatient clinic in Holstebro from
September 2001 to December 2002 with symptoms indicating PCOS
were invited to participate in the study. Blood tests were performed
for fasting venous plasma glucose (FPG), gonadotrophins, total testo-
sterone, dehydroepiandrosterone sulphate, androstenedione, prolactin,
and for thyroid, renal and hepatic function. Women were considered
eligible for the trial if they had a testosterone value above the upper
normal limit and oligo- or amenorrhea, and did not meet any of the
exclusion criteria. Oligomenorrhoea was defined as irregular bleeding
periods with an interval varying between 5 weeks and 6 months and
amenorrhea as absent bleedings for at least 6 months. Ovarian mor-
phology or clinical hyperandrogenism were not diagnostic criteria
and were not registered in all cases. Exclusion criteria were pericli-
macteric gonadotrophin values, hyperprolactinaemia, diabetes melli-
tus, impaired thyroid, renal or hepatic function, hormonal treatment,
pregnancy, lactation or a wish for fertility treatment. Women taking
antihypertensive agents were included and their medication was
unchanged throughout the study.
Protocol
Eligible women who gave written informed consent were assigned to
6 months of treatment with either 850 mg of metformin or placebo twice
daily, followed by a wash-out period of 3 months before cross-over to
the alternate treatment for another 6 months. Randomization defining
treatment sequence was done at inclusion by random number tables
(Dien, 1963). The appearance of the tablets was identical, and patients
and investigators were blinded to treatment allocation.
The randomization code was stored in a closed envelope until all
participants had finished the treatment.
Participants were seen by one of the investigators before inclusion
and every second month during treatment periods, always in the
morning after an overnight fast of at least 8 h. They were weighed
wearing light clothing. Waist circumference was measured at the
umbilical level and hip circumference at the trochanter region. Systo-
lic (SBP) and diastolic blood pressure (DBP) was measured with a
semiautomatic blood pressure monitor (NA-777, A&D Instruments
Ltd., Abingdon, Oxford, UK), and a blood sample was drawn for
immediate analysis without respect to bleeding periods. All partici-
pants registered their bleeding periods in a calendar during both
study periods and the 3 months wash-out period.
2968
The participants were informed to contact the investigators in case
of adverse events or pregnancy. Sexually active women were asked to
use barrier contraception during the study. They were asked to do a
home ovulation test in case of abdominal discomfort and a pregnancy
test if they suspected pregnancy.
The efficiency of blinding was evaluated by telephone interview
of a group of study participants after the study, asking them
during which treatment period they assumed that they had received
metformin.
Statistics
Calculation of sample size was based on the assumption that at least
50% of the women would experience at least 30% more menstrual
periods on metformin than on placebo. Based on a power of 90
(b ¼ 0.10) to detect a significant difference [two-sided P-value (a)
of 0.05], the minimum sampling size was calculated to 44 subjects.
We also assumed a 10% drop-out rate, and thus aimed at including
Downloaded from http://humrep.oxfordjournals.org/ at Fagbiblioteket, Sygehus Viborg on January 9, 2014
50 women. As drop-out rate quickly rose higher than expected, we
decided to include 60 women.
In the intention-to-treat analysis, the values of each participant after
6 months of metformin or placebo were compared with the baseline
values. Linear regression analysis with the changes in testosterone
and homeostasis model assessment (HOMA) index as dependent vari-
ables was performed to examine potential relations between the
changes.
The per protocol analysis included data from participants completing
both study periods, i.e. the difference between the values of each par-
ticipant after placebo and metformin, respectively, was calculated, and
a significance test performed on the differences.
A paired Student’s t-test was performed if data followed Gaussian
distribution evaluated by the Kolmogorov–Smirnov test; otherwise
a Wilcoxon paired sample test was applied. An unpaired t-test was
used for testing differences between means.
A post-hoc analysis was performed by subgrouping the data according
to body mass index (BMI), as BMI was considered the most important
potential confounder.
As the data are not complete for all women at all times, the number
of analysed pairs varies. Some controls were cancelled because the
women failed to turn up or wanted to skip a control for personal
reasons. At some occasions, blood tests were not done because the
woman had not been fasting for at least 8 h.
The statistical software program Stata, version 9.2 (StataCorp 2005)
was used for the statistical evaluation.
The tests used in this study have the following normal values:
FPG , 6.1 mmol/l, fasting insulin ,40 mIU/l, total cholesterol
3.4–7.0 mmol/l, high-density lipoprotein (HDL)-cholesterol 0.8–
1.7 mmol/l, low-density lipoprotein (LDL)-cholesterol 1.5–
4.9 mmol/l, triglycerides 0.5–2.3 mmol/l, testosterone 0.6–1.8
nmol/l and sex hormone-binding globulin (SHBG) 41–169 nmol/l.
Insulin sensitivity was evaluated by HOMA, calculated as fasting
serum insulin (mIU/l)  FPG (mmol/l)/22.5. Obesity was defined
as BMI  30 kg/m2. Adherence to therapy was assessed by count-
ing the tablets returned by the participants after each treatment
period.
The study was approved by the regional Ethics Committee, the
Danish Medicines Agency, the Danish Data Protection Agency, and
conducted in accordance with the Helsinki Declaration and the guide-
lines for Good Clinical Practice.
Role of the pharmaceutical companies providing the study
medication
Metformin and placebo was donated by GEA A/S, Hvidovre,
Denmark. The pregnancy and ovulation tests were donated by

Unipath Limited, UK. The companies did not participate in or
influence study design, collection, analysis or interpretation of data;
in the writing of the report or the decision to submit the paper for pub-
lication. The corresponding author had full access to all data in the
study and had final responsibility for the decision to submit for
publication.
Results
Sixty women were included. All had both elevated testosterone
levels and oligo- or amenorrhoea and thus had PCOS according
to the Rotterdam definition. Fifty-six remained for the
intention-to-treat analysis and 38 for the per protocol analysis
(Fig. 1). There was no difference in drop-out rate between
obese and non-obese women.
Subjects included in the per protocol analysis did not differ
from the intention-to-treat group in any of the studied para-
meters (Table 1).
There was no difference in compliance between the two
periods (data not shown). Of 50 women taking metformin, 29
(58%) reported side effects, mostly gastrointestinal. Two
women reported side effects on placebo. There were no
serious adverse events. Thirty-one women were interviewed
after the study and 24 (77%) guessed correctly the sequence
of metformin and placebo. Their guesses were made mainly
on the basis of side effects.
Figure 1: Flow chart of participants during the study
Metformin, obesity and polycystic ovary syndrome
Effect of treatment
Bleeding calendars were available for 36 women who
completed both study periods. There were more bleeding
periods on metformin than on placebo, but the difference did not
reach statistical significance (P ¼ 0.063, Wilcoxon signed-rank
test). Nineteen had more frequent periods on metformin than on
placebo (4.8 against 3.2), while 11 had more frequent periods on
placebo (5.6 against 4.5) and 7 did not register any change.
There was no difference between obese and non-obese women.
The intention-to-treat analysis showed significant results for
both metformin and placebo (Table 2). All changes occurred in
the total group of participants or the obese group, while there
were no significant changes in non-obese women. The decrease
in testosterone levels was close to significance during the
placebo as well as the metformin periods. DBP, waist-hip
Downloaded from http://humrep.oxfordjournals.org/ at Fagbiblioteket, Sygehus Viborg on January 9, 2014
ratio, triglyceride, LDL, cholesterol and SHBG did not change.
The changes during metformin and placebo periods,
respectively, were compared and significant differences were
found in the total group of participants and the obese group.
The results are shown in Table 3. Only the differences in
change in HOMA index reached significance, when obese
and non-obese women were compared.
The per protocol analysis showed that after 6 months, body
weight, FPG and HOMA index were significantly lower on
metformin than on placebo (P , 0.001, P ¼ 0.022 and
P ¼ 0.032, respectively). Obese women also had lower
2969
Trolle et al.

Table 1: Baseline characteristics of the study subjects

All (n ¼ 56) Per protocol Non-obese Obese (n ¼ 40) P-value for obese
(n ¼ 38) (n ¼ 16) versus non-obese

Age (year)a 32 (21–42) 31 (30–33) 30 (18– 39) 32 (25–42) 0.238

BMI (kg/m2)a 33.8 (22.2– 46.0) 34.1 (22.5–46.0) 25.3 (21.7–29.8) 37.2 (30.5– 47.0) 0.000
WHRa 0.86 (0.73– 0.95) 0.86 (0.73–1.00) 0.84 (0.63–0.94) 0.87 (0.78– 0.98) 0.094
FPG (mmol/l)a 5.3 (4.4– 6.5) 5.3 (4.6– 6.5) 5.0 (4.3– 5.7) 5.4 (4.6–6.6) 0.006
Testosterone (nmol/l)a 2.89 (1.84– 4.58) 2.88 (1.78–4.22) 2.95 (1.78–4.58) 2.87 (1.91– 4.74) 0.753
Cholesterol (mmol/l)a 5.01 (3.80– 6.43) 5.09 (3.80–6.43) 5.11 (4.42–6.43) 4.98 (3.80– 6.45) 0.616
HDL (mmol/l)a 1.27 (0.94– 1.89) 1.28 (0.94–1.91) 1.36 (0.88–2.00) 1.23 (0.95– 1.52) 0.080
LDL (mmol/l)a 3.02 (2.10– 4.32) 2.99 (2.10–4.32) 2.98 (2.10–4.32) 3.04 (2.10– 4.44) 0.788
TG (mmol/l) 1.27 (0.56– 2.89) 1.36 (0.56–3.30) 1.16 (0.56–2.89) 1.36 (0.62– 3.33) 0.340
SHBG (nmol/l)a 35 (16–58) 34 (14–56) 39 (20– 71) 33 (15–57) 0.105
Insulin (mIU/l) 9.50 (2.39– 36.29) 9.70 (3.51–36.29) 5.13 (2.19–18.66) 12.78 (3.70– 40.75) 0.000
SBP (mmHg)a 133 (110– 170) 134 (110– 170) 121 (100 –140) 137 (115–175) 0.003
DBP (mmHg)a 84 (60–110) 86 (60–110) 75 (60– 95) 87 (64–113) 0.007
Weight (kg)a 95 (63–132) 96 (66–132) 72 (58– 82) 105 (83–133) 0.000
HOMA index (mmol/l*mIU/l) 2.25 (0.55– 9.60) 2.30 (0.72–9.60) 1.25 (0.43–4.23) 3.20 (0.72– 10.81) 0.000

Downloaded from http://humrep.oxfordjournals.org/ at Fagbiblioteket, Sygehus Viborg on January 9, 2014

a
Gaussian distributed parameters. Values are given as mean (5– 95% percentiles), P-value for obese versus non-obese was evaluated by two-sided t-test.
Non-normally distributed parameters are given as median (5–95% percentiles) and P-value evaluated by Mann– Whitney test. WHR, waist-hip ratio; TG,
triglycerides.

values of insulin (P ¼ 0.012) and testosterone (P ¼ 0.039).
No difference between metformin and placebo was seen in
non-obese women.
When regression analysis was performed with baseline
values of testosterone as the dependent variable and insulin,
glucose and weight as independent variables, insulin was the
only significant predictor [b ¼ 0.04, P , 0.001, 95% confi-
dence interval (CI) 0.019 – 0.060]. The change in insulin was
also the only significant predictor of the change in testosterone
after the metformin period (b ¼ 0.08, P , 0.001, 95% CI
0.048 – 0.118), while the change during the placebo period
was not associated with any of the other parameters. At
baseline, the HOMA-index was significantly associated with
weight (b ¼ 0.045, P ¼ 0.035, 95% CI ¼ 0.003 – 0.087) and
with testosterone (b ¼ 2.21, P , 0.001, 95% CI ¼ 1.21 –
3.21). The change on metformin was associated with the
change in testosterone (b ¼ 1.22, P ¼ 0.001, 95% CI 0.58 –
1.87), but not significantly with the change in weight.
Discussion
The present study shows a clear effect of metformin on weight,
testosterone and several metabolic factors in obese women with
PCOS. Further, the data indicate that there may not be an effect

Table 2: Changes from baseline to 6 months (intention-to-treat analysis)

All participants Obese Non-obese

Change P-value (n) Change P-value (n) Change P-value (n)

a
Weight (kg)
Metformin 22.3 (23.9, 20.6) 0.009 (42) 22.7 (25.0, 20.4) 0.02 (30) 21.0 (22.4, 0.2) 0.101 (12)
Placebo 1.5 (0.2, 2.7) 0.019 (45) 2.1 (0.6, 3.7) 0.008 (32) 20.2 (21.7, 1.4) 0.781 (13)
SBP (mmHg)a
Metformin 25.4 (210.8, 20.1) 0.047 (40) 25 (212, 1) 0.126 (29) 25.9 (215.3, 3.5) 0.193 (11)
Placebo 1 (23, 5) 0.529 (43) 0 (26, 5) 0.921 (31) 5.7 (23.2, 14.5) 0.185 (12)
HDL (mmol/l)a
Metformin 0.11 (0.05, 0.18) 0.001 (42) 0.14 (0.06, 0.22) 0.002 (30) 0.04 (20.06, 0.13) 0.386 (12)
Placebo 0.04 (0.01, 0.08) 0.116 (43) 0.02 (20.03, 0.08) 0.434 (31) 0.08 (20.02, 0.17) 0.093 (12)
Testosterone
(nmol/l)a
Metformin 20.30 (20.61, 0.01) 0.055 (42) 20.46 (20.82, 20.1) 0.013 (30) 20.11 (20.49, 0.71) 0.698 (12)
Placebo 20.24 (20.46, 0.00) 0.054 (45) 20.2 (20.5, 0.1) 0.178 (32) 20.28 (20.65, 0.08) 0.283 (13)
a
FPG (mmol/l)
Metformin 20.13 (20.26, 0.00) 0.058 (38) 20.18 (20.33, 20.03) 0.018 (27) 0.00 (20.3, 0.3) 1.000 (11)
Placebo 0.16 (0.01, 0.30) 0.031 (41) 0.22 (0.03, 0.40) 0.024 (29) 0.01 (20.2, 0.2) 0.924 (12)
Insulin (mIU/l)
Metformin 21.48 (218.38, 9.52) 0.101 (35) 22.09 (218.38, 8.45) 0.045 (25) 0.38 (210.49, 23.9) 0.721 (10)
Placebo 0.45 (27.77, 12.96) 0.379 (34) 0.77 (27.77, 12.96) 0.412 (23) 20.14 (23, 10.02) 0.790 (11)
HOMA index
(mIU/l*mmol/l)
Metformin 20.48 (25.96, 1.89) 0.070 (33) 20.66 (25.96, 21.54) 0.022 (23) 0.16 (22.48, 4.27) 0.575 (10)
Placebo 0.19 (22.10, 3.62) 0.092 (32) 0.38 (22.10, 3.62) 0.068 (21) 0 (20.63, 2.17) 0.929 (11)
a
Gaussian distributed parameters (*) are given as mean (5–95% percentiles) and tested by paired t-test, otherwise values are given as median (5– 95%
percentiles) and tested by Wilcoxon signed rank-sum test. n: number analysed.

2970
 Metformin, obesity and polycystic ovary syndrome

Table 3: Mean differences between changes during metformin and placebo periods (intention-to-treat analysis), paired t-test (5 –95% percentiles)

All participants Obese Non-obese Obese versus

non-obese
Difference P-value (n) Difference P-value Difference P-value P-value
(n) (n)

Weight (kg) 24.2 (27.0, 21.9) ,0.001 (38) 25.2 (28.2, 22.2) 0.001 (28) 21.6 (24.2, 1.1) 0.222 (10) 0.162
SBP (mmHg) 25.0 (211.2, 1.3) 0.116 (36) 24.4 (212.5, 3.6) 0.265 (27) 26.6 (216.3, 3.2) 0.158 (9) 0.772
HDL (mmol/l) 0.07 (20.01, 0.16) 0.095 (36) 0.11 (0.01, 0.20) 0.033 (27) 20.03 (20.23, 0.17) 0.747 (9) 0.168
Testosterone (nmol/l) 20.09 (20.50, 0.32) 0.656 (38) 20.24 (20.73, 0.24) 0.315 (28) 0.34 (20.48, 1.15) 0.378 (10) 0.208
FPG (mmol/l) 20.23 (20.44, 20.01) 0.041 (30) 20.34 (20.59, 20.09) 0.010 (22) 0.09 (20.36, 0.54) 0.660 (8) 0.072
Insulin (mIU/l) 24.17 (28.10, 20.23) 0.039 (26) 26.16 (211.18, 21.14) 0.019 (18) 0.32 (26.08, 6.72) 0.908 (8) 0.119
HOMA index (mIU/ 21.50 (22.53, 20.47) 0.006 (23) 22.28 (23.61, 20.95) 0.003 (15) 20.04 (21.35, 1.27) 0.943 (8) 0.028
l mmol/l)

n: value number analysed. An unpaired t-test was used for obese vs. non-obese.

Downloaded from http://humrep.oxfordjournals.org/ at Fagbiblioteket, Sygehus Viborg on January 9, 2014

in non-obese subjects. This finding must be interpreted with
caution because of the small number of non-obese patients
completing both study periods. It is still striking, however,
that metformin had no effect on any of the studied parameters
in this group, and it emphasizes the heterogeneity of PCOS and
the importance of testing treatments on clearly defined
subgroups of patients. It must also be taken into account that
obese and non-obese women differ at baseline. The obese
group generally have values more distant from the median of
the normal area, which makes the room for improvement and
the chance of a significant change greater. The motivation for
losing weight may also have been higher in the obese subgroup,
thus influencing eating habits and life style. However, the fact
that the obese women actually have a significant weight gain
during the placebo period does not support this assumption.
Our study demonstrated an improvement in bleeding frequency
on metformin, but this was not different from the improvement
on placebo. This is in contrast to Moghetti et al. (2000) who
found no effect on placebo, but this may be explained by the
fact that there were important differences between placebo
and metformin groups at baseline.
It is still under debate whether non-obese women with
PCOS are insulin resistant. While there is no difference
between non-obese and obese patients in some studies
(Dunaif et al., 1989), others have not been able to demonstrate
insulin resistance in lean PCOS women (Dale et al., 1992;
Ovesen et al., 1993; Vrbikova et al., 2004). Obesity clearly
aggravates the symptoms of PCOS (Dunaif et al., 1989;
Cresswell et al., 2003; Vanky et al., 2004), and weight loss is
central in the treatment. Weight loss has been shown to
enhance ovulation frequency and improve menstrual cyclicity
and endocrine profile (Kiddy et al., 1992; Crosignani et al.,
2003; Tang et al., 2006). It is also considered of fundamental
importance in reducing the cardiovascular risk factors
included in the metabolic syndrome. Metformin decreases
feeling of hunger during hypoglycaemia (Schultes et al.,
2003), and moderate weight loss is common during metformin
treatment, but the conclusions of randomized studies differ
with regard to the effect of metformin on weight in PCOS.
Fleming et al. (2002) reported results similar to ours on
both metformin and placebo, while others could not demon-
strate a weight reduction in patients treated with metformin
(Moghetti et al., 2000; Bridger et al., 2006; Lord et al.,
2006). Variations in study groups in pretreatment BMI, met-
formin dose, concomitant life-style changes and patient adher-
ence to treatment may account for these differences.
There is also no definite answer to whether metformin per se
has an effect of any of the symptoms of PCOS. We were unable
to demonstrate an effect of metformin on bleeding frequency.
The effect of metformin on anovulation, conception and live
birth rate in PCOS is well documented, but its efficacy is ques-
tioned compared with other options (Lord et al., 2003; Legro
et al., 2007). Tang et al. (2006) were unable to demonstrate
that metformin had an effect on menstrual cyclicity over
weight loss through life-style modification, and metformin
did not induce weight loss. There was no change in insulin sen-
sitivity in the metformin group, but testosterone levels were
reduced, presumably by a direct effect on ovarian steroid
genesis. The reduction in HOMA index and testosterone in
obese participants on metformin found in our study was depen-
dent on the reduction in insulin, but independent of changes in
weight, which leads us to conclude that the effect of metformin
is not mediated through weight loss.
In the intention-to-treat analysis of our results, we could not
demonstrate any significant difference between the changes in
testosterone levels during metformin and placebo periods. In
fact, the trend towards a significant reduction in testosterone
was surprisingly seen both in metformin and placebo treated
subjects. We speculate that the decline in testosterone levels
may be a result of care itself. Most participants had never
received any information or treatment of their condition
before entering the study, and they were generally pleased to
be subject to care. There are reports of a decreased quality of
life in PCOS (Elsenbruch et al., 2003) and a higher frequency
than expected of metabolic and hormonal changes in women
suffering from fear and anxiety (Rosmond et al., 2001). The
metabolic syndrome shows an association with chronic stress
and dysregulation of the hypothalamic – pituitary –adrenal
axis (Bjorntorp and Rosmond, 2000; Tsigos and Chrousos,
2002). Thus, psychosocial stress may influence the manifes-
tations of PCOS and render the condition susceptible to the
effect of care. The participants may also have changed their
life style beyond our knowledge as a result of the information
about their condition, but the fact that the testosterone changes
were not associated with changes in BMI, insulin or
HOMA-index makes this explanation less plausible.
2971
Trolle et al.
The blood tests were done without respect to bleeding pattern.
We aimed to include only anovulatory women with a steady-state
hormonal pattern in our study, but there may have been temporary
hormonal fluctuations interfering with the results.
In obese participants, HOMA-index and FPG levels were
significantly reduced on metformin treatment, whereas in the
placebo FPG increased significantly. FPG levels tend to
increase over time in PCOS women, and in our Danish popu-
lation 38% of the PCOS patients 35 years or older have
impaired FPG values (Trolle and Lauszus, 2005). The effect
of metformin on FPG is well known and consistent with the
observation that metformin suppresses endogenous glucose
production (DeFronzo, 1999). We speculate that the lowering
of FPG may postpone or abolish deterioration of glucose
metabolism in obese PCOS subjects, but long-term studies
are necessary to prove this. The same holds true for the develop-
ment of cardiovascular disease. So far, there is no clinical evi-
dence that metformin treatment reduces the development of
cardiovascular disease in PCOS patients.
The number of patients included in this study and the
cross-over design gives this study a considerable strength in
spite of the high percentage of drop-outs (32%). Thirty-eight
subjects were still available for per protocol analysis and
acted as their own controls, thus minimizing influence of pos-
sible confounders such as age, exercise and smoking habits.
We did not demonstrate a significant carry-over effect. An
order effect cannot be ruled out considering the frequent side
effects of metformin, which hampered the blinding to the
patients. The supposition of receiving active or inactive treat-
ment may have influenced eating pattern, interfered with adher-
ence or otherwise affected test results. However, adherence to
treatment was high despite the re-called insufficient blinding.
Very few studies on metformin have been paired studies.
The unpaired study design has a larger risk of potential bias
in the randomization procedure given the heterogeneity of
the PCOS phenotype. Similarly, the paired design minimizes
the effect of potential confounders better than unpaired
studies and the effect of regression to the mean when subgroups
of the population are studied. Thus, the design of this study
makes it possible to estimate the likely magnitude of positive
effects of metformin although population characteristics such
as weight, age, race and possibly other factors must be taken
into account. Re-calculation of test strength by the significant
results in the study suggests that the risk of type 2 error was
,99% with a , 0.01 in testosterone, HOMA-index and
SBP. The preliminary estimates mentioned in the method
section are still valid for body weight, while FPG due to the
variation shows a power of ,80 at a ¼ 0.05.
In conclusion, in obese PCOS women, metformin treatment
for 6 months lowers body weight, fasting glucose, HOMA-
index and testosterone levels, while there seems to be no
effect on these parameters in non-obese PCOS women.
Acknowledgement
This study was supported by grants from The Danish Research
Council for Health and Disease, The Novo Nordisk Foundation, the
2972
Danish Diabetes Association and Clinical Institute, University of
Aarhus. We are indebted to Kirsten Nyborg Rasmussen for excellent
technical assistance.
References
Dien K. Documenta Geigy: Scientific Tables. J.R. Geigy Pharmaceuticals,
1963.
Bjorntorp P, Rosmond R. The metabolic syndrome– a neuroendocrine
disorder? Br J Nutr 2000;83:49– 57.
Bridger T, MacDonald S, Baltzer F, Rodd C. Randomized placebo-controlled
trial of metformin for adolescents with polycystic ovary syndrome. Arch
Pediatr Adolesc Med 2006;160:241– 246.
Cresswell J, Fraser RB, Bruce C, Egger P, Phillips D, Barker DJP. Relationship
between polycystic ovaries, body mass index and insulin resistance. Acta
Obstet Gynecol Scand 2003;82:61– 64.
Crosignani PG, Colombo M, Vegetti W, Somigliana E, Gessati A, Ragni G.
Overweight and obese anovulatory patients with polycystic ovaries:
Downloaded from http://humrep.oxfordjournals.org/ at Fagbiblioteket, Sygehus Viborg on January 9, 2014
parallel improvements in anthropometric indices, ovarian physiology and
fertility rate induced by diet. Hum Reprod 2003;18:1928– 1932.
Dale PO, Tanbo T, Vaaler S, Åbyholm T. Body weight, hyperinsulinemia, and
gonadotrophin levels in the polycystic ovary syndrome: evidence of two
distinct populations. Fertil Steril 1992;58:487– 491.
DeFronzo RA. Pharmacologic therapy for type 2 diabetes mellitus. Ann Intern
Med 1999;131:281– 303.
Diabetes Prevention Program Research Group. Reduction in the incidence of
type 2 diabetes with lifestyle intervention or metformin. N Engl J Med
2002;346:393– 403.
Dunaif A, Segal KR, Futterweit W, Dobrjansky A. Profound peripheral insulin
resistance, independent of obesity, in polycystic ovary syndrome. Diabetes
1989;38:1165– 1174.
Elsenbruch S, Hahn S, Kowalsky D, Offner AH, Schedlowski M, Mann K,
Janssen OE. Quality of life, psychosocial well-being, and sexual
satisfaction in women with polycystic ovary syndrome. J Clin Endocrinol
Metab 2003;88:5801– 5807.
Fleming R, Hopkinson ZE, Wallace AM, Greer IA, Sattar N. Ovarian function
and metabolic factors in women with oligomenorrhea treated with metformin
in a randomized double blind placebo-controlled trial. J Clin Endocrinol
Metab 2002;87:569–574.
Gambineri A, Pelusi C, Genghini S, Morselli-Labate AM, Cacciari M, Pagotto U,
Pasquali R. Effect of flutamide and metformin administered alone or in
combination in dieting obese women with polycystic ovary syndrome.
Clin Endocrinol 2004;60:241– 249.
Ganie MA, Khurana ML, Eunice M, Gulati M, Dwivedi SN, Ammini AC.
Comparison of efficacy of spironolactone with metformin in the
management of polycystic ovary syndrome: an open-labeled study. J Clin
Endocrinol Metab 2004;89:2756–2762.
Glueck CJ, Papanna R, Wang P, Goldenberg N, Sieve-Smith L. Incidence and
treatment of metabolic syndrome in newly referred women with confirmed
polycystic ovarian syndrome. Metabolism 2003;52:908–915.
Glueck CJ, Wang P, Fontaine R, Tracy T, Sieve-Smith L. Metformin-induced
resumption of normal menses in 39 of 43 (91%) previously amenorrheic
women with the polycystic ovary syndrome. Metabolism 1999;48:511–519.
Harborne L, Fleming R, Lyall H, Norman J, Sattar N. Descriptive review of the
evidence for the use of metformin in polycystic ovary syndrome. Lancet
2003a;361:1894–1901.
Harborne L, Fleming R, Lyall H, Sattar N, Norman J. Metformin or
antiandrogen in the treatment of hirsutism in polycystic ovary syndrome.
J Clin Endocrinol Metab 2003b;88:4116– 4123.
Holte J, Gennarelli G, Wide L, Lithell H, Berne C. High prevalence of
polycystic ovaries and associated clinical, endocrine, and metabolic
features in women with previous gestational diabetes mellitus. J Clin
Endocrinol Metab 1998;83:1143–1150.
Kelly CJG, Gordon D. The effect of Metformin on hirsutism in polycystic
ovary syndrome. Eur J Endocrinol 2002;147:217–221.
Kiddy D, Hamilton-Fairley D, Bush A, Short F, Anyaoku V, Reed MJ, Franks S.
Improvement in endocrine and ovarian function during dietary treatment of
obese women with polycystic ovary syndrome. Clin Endocrinol
1992;36:105–111.
Legro RS, Barnhart HX, Schlaff WD, Carr BR, Diamond MP, Carson SA,
Steinkampf MP, Coutifaris C, McGovern PG, Cataldo NA et al.

Clomiphene, metformin, or both for infertility in the polycystic ovary
syndrome. N Engl J Med 2007;356:551– 566.
Legro RS, Kunselman AR, Dodson WC, Dunaif A. Prevalence and predictors
of risk for type 2 diabetes mellitus and impaired glucose tolerance in
polycystic ovary syndrome: a prospective, controlled study in 254 affected
women. J Clin Endocrinol Metab 1999;84:165– 169.
Lord J, Thomas R, Fox B, Acharya U, Wilkin T. The effect of metformin on fat
distribution and the metabolic syndrome in women with polycystic ovary
syndrome-a randomised, double-blind, placebo-controlled trial. BJOG
2006;113:817–824.
Lord JM, Flight IHK, Norman RJ. Metformin in polycystic ovary syndrome:
systematic review and meta-analysis. BMJ 2003;327:951–953.
Moghetti P, Castello R, Negri C, Tosi F, Perrone F, Caputo M, Zanolin E,
Muggeo M. Metformin effects on clinical features, endocrine and metabolic
profiles, and insulin sensitivity in polycystic ovary syndrome: a randomized,
double-blind, placebo-controlled 6-month trial, followed by open, long-term
clinical evaluation. J Clin Endocrinol Metab 2000;85:139–146.
Nestler JE, Jakubowicz DJ, Evans WS, Pasquali R. Effects of metformin on
spontaneous and clomiphene-induced ovulation in the polycystic ovary
syndrome. N Engl J Med 1998;338:1876–1880.
Ovesen P, Moller J, Ingerslev HJ, Jorgensen JO, Mengel A, Schmitz O,
Alberti KG, Moller N. Normal basal and insulin-stimulated fuel
metabolism in lean women with the polycystic ovary syndrome. J Clin
Endocrinol Metab 1993;77:1636–1640.
View publication stats
Metformin, obesity and polycystic ovary syndrome
Rosmond R, Baghaei F, Holm G, Bjorntorp P. Relationships between
personality disorders and anthropometry, hormones and metabolism in
women. J Endocrinol Invest 2001;24:159–165.
Schultes B, Oltmanns KM, Kern W, Fehm HL, Born J, Peters A. Modulation of
hunger by plasma glucose and metformin. J Clin Endocrinol Metab
2003;88:1133–1141.
Tang T, Glanville J, Hayden CJ, White D, Barth JH, Balen AH. Combined
lifestyle modification and metformin in obese patients with polycystic
ovary syndrome. A randomized, placebo-controlled, double-blind
multicentre study. Hum Reprod 2006;21:80–89.
Trolle B, Lauszus FF. Risk factors for glucose intolerance in Danish women
with polycystic ovary syndrome. Acta Obstet Gynecol Scand 2005;
84:1192–1196.
Tsigos C, Chrousos GP. Hypothalamic-pituitary-adrenal axis, neuroendocrine
factors and stress. J Psychosom Res 2002;53:865–871.
Vanky E, Kjotrod S, Salvesen KA, Romundstad P, Moen MH, Carlsen SM.
Clinical, biochemical and ultrasonographic characteristics of Scandinavian
women with PCOS. Acta Obstet Gynecol Scand 2004;83:482–486.
Vrbikova J, Cibula D, Dvorakova K, Stanicka S, Sindelka G, Hill M, Fanta M,
Vondra K, Skrha J. Insulin sensitivity in women with polycystic ovary
Downloaded from http://humrep.oxfordjournals.org/ at Fagbiblioteket, Sygehus Viborg on January 9, 2014
syndrome. J Clin Endocrinol Metab 2004;89:2942–2945.
Submitted on January 24, 2007; resubmitted on July 4, 2007; accepted on
July 19, 2007
2973