Beruflich Dokumente
Kultur Dokumente
Submitted on September 29, 2009; resubmitted on March 10, 2010; accepted on March 23, 2010
background: Ovulation induction treatment with metformin, either alone or in combination with clomiphene citrate (CC), remains
controversial even though previous randomized trials have examined this.
methods: A double blinded multi-centre randomized trial was undertaken including 171 women with anovulatory or oligo-ovulatory
polycystic ovary syndrome. Women with high body mass index (BMI) . 32 kg/m2 received placebo (standard care) or metformin;
women with BMI 32 kg/m2 received CC (standard care), metformin or both. Treatment continued for 6 months or until pregnancy
was conrmed. Primary outcomes were clinical pregnancy and live birth.
results: For women with BMI . 32 kg/m2, clinical pregnancy and live birth rates were 22% (7/32) and 16% (5/32) with metformin, 15%
(5/33) and 6% (2/33) with placebo. For women with BMI 32 kg/m2, clinical pregnancy and live birth rates were 40% (14/35) and 29%
(10/35) with metformin, 39% (14/36) and 36% (13/36) with CC, 54% (19/35) and 43% (15/35) with combination metformin plus CC.
conclusions: There is no evidence that adding metformin to standard care is benecial. Pregnancy and live birth rates are low in
women with BMI . 32 kg/m2 whatever treatment is used, with no evidence of benet of metformin over placebo. For women with
BMI 32 kg/m2 there is no evidence of signicant differences in outcomes whether treated with metformin, CC or both.
ClinicalTrials.gov number NCT00795808; trial protocol accepted for publication November 2005: Johnson, Aust N Z Journal Obstet Gynaecol
2006;46:141 145.
Key words: polycystic ovary syndrome / anovulation / metformin / clomiphene citrate / randomized controlled trial
& The Author 2010. Published by Oxford University Press on behalf of the European Society of Human Reproduction and Embryology. All rights reserved.
For Permissions, please email: journals.permissions@oxfordjournals.org
1676 Johnson et al.
for women whose body mass index (BMI) is 32 kg/m2 or less. We whom any other important infertility factor was known to be present, includ-
designed a randomized trial to ascertain whether the addition of met- ing known tubal factor where at least one fallopian tube was blocked (although
fomin to standard treatment provided any clinically relevant benet. a tubal patency test was not a prerequisite for trial entry). However, women
We also studied the relative efcacy of CC, metformin and both known to have stage 1 or 2 endometriosis and men with very mild oligosper-
mia, with sperm count 15 million per ml, were included. Women with
among women with BMI 32 kg/m2; metformin versus placebo
important medical disorders were also excluded (Johnson, 2006).
among women with BMI . 32 kg/m2.
Our power calculation (Johnson, 2006) suggested a sample size of 160
Since initiating our trial, further RCTs have been published with con-
women distributed between the ve treatment arms in the study would be
icting results. The largest RCT, a high quality American multi-centre required to show an increase in the pregnancy rate from 30 to 55% by
trial, showed a signicant benet of CC over metformin (live birth rate adding metformin to standard therapy (to comfortably allow for attain-
22.5 versus 7.2%; Legro et al., 2007). This was widely accepted as ment of at least 122 participants required to have adequate power for
denitive evidence that CC should remain the rst line treatment this primary comparison of standard care versus standard care plus met-
for anovulatory PCOS, with apparently no place for metformin rst formin). The standard care group comprised women with BMI . 32 kg/m2
line (Al-Inany and Johnson, 2006), results largely conrmed by the receiving placebo and women with BMI 32 kg/m2 receiving CC; the stan-
most recent RCT from Malaysia (live birth rate 15.4% for CC versus dard care plus metformin group comprised women with BMI . 32 kg/m2
7.9% for metformin; Zain et al., 2008). However, a smaller Italian receiving metformin and women with BMI 32 kg/m2 receiving CC plus
metformin. A sample size of 61 participants per group was required for 80%
RCT had suggested the live birth rate for metformin (52.0%) was
power to detect an increase in pregnancy rate from 30% in the standard
higher than that for CC (18.0%) for non-obese women with anovula-
care group to 55% in the standard care plus metformin group. Thus,
histological evidence of trophoblastic tissue with spontaneous abortion or stopped taking study drugs owing to gastrointestinal side effects,
ectopic pregnancy) that occurred within 6 months of randomization and only one was actually taking metformin.
resultant live birth. Secondary outcomes were adverse events, ovulation
(conrmed if pregnancy occurred or if serum progesterone level was
Primary outcomes
25 nmol/l; strongly suggested if serum progesterone level was 15
25 nmol/l), spontaneous abortion, ectopic pregnancy, multiple pregnancy Primary and secondary outcomes are expressed in Table II. For
and other adverse perinatal or obstetric complications. women with BMI . 32 kg/m2, there were no signicant differences
All comparisons used the Pearson x2 test. For the combined data and in the live birth rate for metformin (5/32 16%) versus placebo
for the BMI . 32 kg/m2 subgroup, the comparison was between two pro- (2/33 6%) (P 0.21), nor for the clinical pregnancy rate for metfor-
portions; for the BMI 32 kg/m2 subgroup, the comparison was between min (7/32 22%) versus placebo (5/33 15%) (P 0.48). In the
the three treatments; for adverse events, the comparison was between metformin group, ve of the seven pregnancies occurred in the rst
the ve different treatment groups. This analysis was conducted on an 3 months and two in the second 3 months of therapy in association
intention-to-treat basis using worst case assumptions that women lost
with adjunct CC treatment; in the placebo group, three of the ve
to follow-up did not become pregnant, or that they did not have a live
pregnancies occurred in the rst 3 months and two in the second
birth if pregnant at the time of loss to follow-up.
3 months of therapy in association with adjunct CC treatment.
For women with BMI 32 kg/m2, the live birth rates were not sig-
nicantly different among women treated with metformin (10/35
29%), CC (13/36 36%) or combination CC plus metformin (15/
Results
*Biochemical androgen excess included women with total testosterone .2.5 nmol/l or free testosterone .50 pmol/l.
Hyperandrogenism included either biochemical androgen excess or symptoms of hyperandrogenaemia (including hirsutism and acne).
placebo and 41% in those receiving metformin; among women with diabetes, three developed a hypertensive disorder of pregnancy and
BMI 32 kg/m2, 31% with CC, 43% with metformin and 54% with three had a preterm delivery (ranging from gestation 3236 weeks).
CC plus metformin. The overall number of women experiencing side
effects did not differ signicantly between the ve different treatment
groups (P 0.51) and this was also true for gastrointestinal side
Discussion
effects (P 0.23). Only two women reported vasomotor side Our trial did not demonstrate any additional benet from metformin
effects. Of the 45 women having a live birth, one developed gestational added to standard treatment for ovulation dysfunction in PCOS-
Multicentre RCT metformin and clomiphene in PCOS 1679
related infertility. There was no evidence of differences in adverse and Gillett, 2006; Gillett et al., 2006) rather than the World Health
events between treatment groups. Another nding (although based Organization-dened BMI 30 kg/m2 cut-off, above which an individual
only on pilot data of small numbers) was the very similar clinical preg- is dened as obese. The main weakness was that the trial was insuf-
nancy and live birth rate among women with BMI 32 kg/m2 treated ciently powered to detect what many would regard as a clinically
with CC compared with those treated with metformin. meaningful differencein order to detect an increase in live birth
The strengths of this trial are that it was a multi-centre RCT with a from 20 to 30% by adding metformin to standard care (80% power
highly secure method of allocation concealment (third party computer at 95% condence level), 590 participants would have been required
randomization by a telephone call to the pharmacy) and a placebo- for this comparison (as opposed to the 136 participants contributing
controlled maintenance of blinding of all parties. Although it may to this comparison in our trial, which exceeded the power to detect
have been possible for participants to guess what treatment they an increase in pregnancy rate from 30 to 55%, established a priori
were receiving based on medication side effects, it was interesting by Johnson, 2006). Other points that may be considered weaknesses
to note that more women taking placebo stopped treatment owing include the BMI dichotomy of BMI 32 kg/m2 arising from a local policy
to gastrointestinal side effects than those taking metformin. A thus potentially limiting generalizability, the absence of data linking
further strength was the categorization according to BMI, as there is outcomes to womens weight loss whereas on the trial and the fact
a suggestion that these ovulation induction treatments have differential that the metformin only treatment arm in the BMI 32 kg/m2 sub-
effectiveness in women of different BMIin the case of this trial the population can provide only pilot data that need to be combined
cut-off was BMI 32 kg/m2, the threshold below which New Zealand with data from other trials to give sufcient power for meaningful
women qualify for government funded fertility treatment (Farquhar comparisons.
1680 Johnson et al.
Our results are in keeping with other RCTs that have not shown sig- pregnancies) are noteworthy, although based on small numbers.
nicant benet from use of combination CC plus metformin versus CC Legro et al. (2007) had similarly higher results for rst trimester losses
alone (Moll et al., 2006; Legro et al., 2007; Zain et al., 2008). It is worthy with metformin (40.0%) than CC (22.6%), but Palomba et al. (2005)
of debate whether detection smaller differences is important, as all had a lower rate of pregnancy loss in women treated with metformin
RCTs undertaken thus far have been underpowered to detect less (9.7%) than CC (37.5%). All these trials had a policy of stopping metfor-
than a 15% difference in live birth rate, yet other second line treatments min once pregnancy was diagnosed. Whether continuation of metfor-
(such as gonadotrophin injections) substantially increase costs and inva- min through the rst trimester might lower the pregnancy loss rate
siveness. We did not explore, in the lower BMI group, whether pre- among women becoming pregnant with metformin treatment is merit-
treatment with metformin might improve fertility outcomessome worthy of further investigation in robust RCTs.
authors have suggested that the optimal effect of metformin is achieved There is perhaps a paradox that metformin is known to give poor
after 46 months pretreatment (Baillargeon et al., 2004; Palomba et al., results for women with PCOS-related anovulatory infertility and
2005). It remains true that current RCT evidence suggests that CC is a obesity (Tang et al., 2006), where insulin resistance tends to be more
superior treatment to metformin for obese women (Legro et al., 2007; pronounced, yet we have found that the results for women with
Zain et al., 2008), although our RCT did not specically examine this PCOS and BMI 32 kg/m2 are as good for metformin as for CC.
comparison in the BMI . 32 kg/m2 group, but many authorities have Our results challenge the consensus statement that metformin should
questioned the appropriateness of administering ovulation induction be restricted to women with known glucose intolerance (The
treatment to women with obesity (Balen et al., 2006; Lord and Thessaloniki ESHRE/ASRM-Sponsored PCOS Consensus Workshop
Norman, 2006), the obstetric risks (both maternal and fetal) being par- Group, 2008; Thessaloniki ESHRE/ASRM-Sponsored PCOS Consensus
ticularly relevant in the context of gross obesity (Stotland, 2008). Workshop Group, 2008). The main disadvantage of metformin appears
However, it is unclear why the results of our New Zealand multi-centre to be a high incidence of gastrointestinal side effects, but other studies
RCT, the American multi-centre RCT data for the non-obese subpopu- have shown numerous potential advantages of metformin over CC.
lation of women (Table 1 of Supplementary Appendix of Legro et al., These include no known adverse endometrial effect whereas endo-
2007, available at www.nejm.org that showed a signicant advantage metrial thinning could reduce embryo receptivity for some women
of CC) and the Italian RCT that showed a signicant advantage of met- using CC (Gonen and Casper, 1990; Wu et al., 2007), no known
formin (Palomba et al., 2005) were so heterogeneous in the comparison increase in multiple pregnancy rate unlike that associated with CC and
metformin versus CC for non-obese women. Whether these dramati- thus no requirement for inconvenient and costly monitoring of ovulation
cally different results could relate to dose of metformin used, the failure induction cycles that many fertility clinics insist upon for CC, no concern
to adjust metformin dose according to BMI, the preparation used (for over long-term adverse effects on the ovaries such as the lingering
example, a sustained release metformin was used in the American concern over increased risk of ovarian cancer seen in some cohort
trial; Legro et al., 2007), where the outcomes from metformin treat- studies of women using CC, particularly serous ovarian cancer (Jensen
ment were particularly poor or fundamental differences in the popu- et al., 2009) and among those using long treatment courses (Rossing
lations studied in terms of their response to metformin, remains et al., 1994). Nonetheless CC has had remarkable longevity as the rst
unclear. line treatment for anovulatory PCOS and doubtless will continue to
Our results for early pregnancy loss for women with BMI 32 kg/m2 have an important place (Homburg, 2005). Such considerations
(four from 14 metformin pregnancies versus none from 14 CC suggest that metformin is a perfectly reasonable alternative choice to
Multicentre RCT metformin and clomiphene in PCOS
Table III Women experiencing adverse events.
*Women appearing in the table with severe side effects may also appear in the table as having moderate and mild side effects.
1681
Downloaded from http://humrep.oxfordjournals.org/ by guest on November 14, 2015
1682 Johnson et al.
CC as rst line for ovulation induction for women with ovulation dysfunc- Farquhar CM, Gillett W. Prioritising for fertility treatmentsthe New Zealand
tion related to PCOS in the absence of obesity and this debate should experience with limiting treatment only for women with body mass
not be closed. indices less than 32 kg/m2. Br J Obstet Gynaecol 2006;113:1107 1109.
Gillett W, Putt T, Farquhar CM. Prioritising for fertility treatmentsthe
Conclusion effect of excluding women with a high body mass index. Br J Obstet
Gynaecol 2006;113:1218 1221.
There is no evidence that adding metformin to standard care is ben-
Gonen Y, Casper RF. Sonographic determination of a possible adverse
ecial in improving pregnancy and live birth rates for women with ano-
effect of clomiphene citrate on endometrial growth. Hum Reprod
vulatory PCOS. 1990;5:670 674.
Homburg R. Clomiphene citrateend of an era? A mini-review. Hum
Authors roles Reprod 2005;20:2043 2051.
Hull MG. Epidemiology of infertility and polycystic ovarian disease:
N.J. designed the study, took principal investigator responsibility endocrinological and demographic studies. Gynecol Endocrinol 1987;
throughout, wrote the manuscript and is the guarantor of the 1:235 245.
research. A.S. was the trial statistician and was responsible for data Jensen A, Sharif H, Frederiksen K, Kruger Kjaer S. Use of fertility drugs and risk
analysis. J.F. was the co-ordinating research nurse for the study. of ovarian cancer: Danish population based study. Br Med J 2009;338:b249.
C.F., S.M., V.P.S. and K.B. commented on study design and were Johnson NP. No more surrogate end-points in randomised trialsthe
PCOSMIC trial protocol for women with polycystic ovary syndrome
involved in patient recruitment. Q.O. designed the trial database
using metformin for infertility with clomiphene. Aust N Z J Obstet
Tang T, Lord JM, Norman RJ, Yasmin E, Balen AH. Thessaloniki ESHRE/ASRM-Sponsored PCOS Consensus Workshop
Insulin-sensitising drugs (metformin, rosiglitazone, pioglitazone, Group. Consensus on infertility treatment related to polycystic ovary
D-chiro-inositol) for women with polycystic ovary syndrome, syndrome. Fertil Steril 2008;89:505 522.
oligo amenorrhoea and subfertility. Cochrane Database Syst Rev Wu HH, Wang NM, Cheng ML, Hsieh JN. A randomized comparison of
2010;1:CD003053. ovulation induction and hormone prole between the aromatase
The Rotterdam ESHRE/ASRM-Sponsored PCOS Consensus Workshop inhibitor anastrozole and clomiphene citrate in women with infertility.
Group. Revised 2003 consensus on diagnostic criteria and long-term Gynecol Endocrinol 2007;23:76 81.
health risks related to polycystic ovary syndrome (PCOS). Hum Zain MM, Jamaluddin R, Ibrahim A, Norman RJ. Comparison of clomiphene
Reprod 2004;19:41 47. citrate, metformin, or the combination of both for rst-line ovulation
The Thessaloniki ESHRE/ASRM-Sponsored PCOS Consensus Workshop induction, achievement of pregnancy, and live birth in Asian women
Group. Consensus on infertility treatment related to polycystic ovary with polycystic ovary syndrome: a randomized controlled trial. Fertil
syndrome. Hum Reprod 2008;23:462 477. Steril 2008;85:1448 1451.