In this study on women with PCOS, treatment with metformin resulted in a significant decrease in body mass index, improvement of insulin action with a parallel increase of SHBG concentrations. This study contributed to a few randomized double blind placebo controlled trials and assessed metformin effects on metabolic and endocrine profiles in polycystic ovary syndrome.

4.1  Body Weight And Body Mass Index


In the metformin-treated group, mean body mass index averaged from 40.6 before the trial to 38.9 after trial. The absolute difference in these two values was 1.7, which came to a 4.2% reduction in the mean BMI after metformin therapy. Decrease of BMI in the placebo-taking group was not significant. These findings are consistent with results of almost all observational studies, where a decrease in the BMI ranged from 1% to approximately 4.3%.

Velasquez et al. (1994); Glueck et al. (1999); Kolodziejczyk et al. (2000) additionally found an improvement in menstrual cyclicity during metformin therapy. The metformin-induced weight loss was supposed to be the real cause of improvement rather than direct actions of metformin. It is also well documented in the study by Kiddy and co-workers (1992) that more than 5% decrease in body weight can mitigate reproductive abnormalities. However, the following four double-blind placebo control studies of Moghetti et al. (2000); Morin-Papunen et al. (1998); Kocak et al. (2002) and Fleming et al. (2002) did not find a significant decrease in BMI despite the amelioration in hyperinsulinemia and hyperandrogenism. Furthermore, in the study by Moghetti et al., the metformin group had a significantly lower body mass index than the placebo group, which could explain the nonsignificant result upon completion.

To answer the question if weight loss with metformin is better than weight loss alone, three trials comparing effects of low calorie diet and metformin therapy were conducted. In a randomized control trial (RCT) including 18 obese women with PCOS, all subjects were given a low calorie diet for 1 month before and during 6 months of the study (Pasquali et al., 2000). Ten were treated with metformin and 8 with placebo. In a further trial by Casimiri and colleagues (1997), 24 patients were given a low calorie diet with placebo or metformin for 26 weeks. In both studies, compared with placebo, metformin improved insulin sensitivity, lowered testosterone levels and improved ovulatory and menstrual functions. One placebo-controlled 16-weeks study was not able to demonstrate any additional effect of metformin over-dieting (Crave et al., 1995). The subjects were all hirsute and very obese (mean BMI 35,2 kg/m²) and only 15 of the 24 women had polycystic ovaries on ultrasound. They may represent a particularly difficult group to treat successfully.


Lean PCOS subjects can experience similar hormonal and metabolic improvements during metformin treatment compared to obese women. Most of the beneficial changes were observed after 3 months of the treatment and at a dosage of 1 g/d. It was hypothesized that lower doses of metformin than those used in obese PCOS women could be sufficient in lean women (Ibanez et al. 2001; Morin-Papunen et al., 2003).

However, while metformin mainly affects central obesity and lipid metabolism, with minimal effects on hepatic clearance, its action seems to be focused mainly on hepatic clearance of insulin and steroid secretion in non-obese patients. This result suggests a particular characteristic of PCOS in non-obese women or the fact that obesity acts as a confounding factor of this disease (Morales et al.,1996).

In this study, significant weight loss was achieved only in the metformin-treated group, although all patients were advised to follow dietary restriction and to do regular physical exercises. Therefore, it was concluded, first, that metformin could contribute to weight loss, and, second, that weight loss can be significantly greater than observed in majority of RCTs (where metformin was administrated without dietary restriction) if metformin is combined with a hypocaloric diet or/and regular physical activity.

4.2 Body Fat Distribution and Waist-Hip Ratio


The majority of PCOS women are obese or overweight, however, obesity is not the obligatory feature of this syndrome. Of special importance is the amount of body fat and especially the sex specific kind of fat distribution, which can be indicators of the hormonal and reproductive status of woman. Up to now few studies analysed fat distribution patterns in obese and lean PCOS patients (Bringer et al.,1993; Douchi et al., 1997; Lefebvre et al., 1997; Kirchengast et al., 2001). The authors were able to tended to an android (abdominal) fat pattern in PCOS women in comparison with weight-matched control women.

The present study found a significant decrease in waist and hip circumferences in PCOS patients after metformin treatment. These data are in agreement with the results of all other studies that examined metformin effects on obese PCOS women. Pasquali and co-workers measured, by CT scan, additional anthropometric parameters, such as total adipose tissue (TAT), visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT). Whereas obese PCOS subjects lost significantly more visceral fat in the abdomen without any significant changes of SAT after metformin, the opposite was observed in the control obese group. These findings, therefore, suggested particular effect of metformin on visceral adipose tissue in PCOS.

In this study, metformin-treated women lost significantly more abdominal fat (waist circumference) compared to controls. A loss of hip circumference was obvious in both groups, however, a waist-to-hip ratio remains constant after 16 weeks of metformin or placebo treatment.


Available data seem to support the concept that PCOS and abdominal obesity have a negative impact on insulin sensitivity (Dunaif A., 1997). The study by Pasquali and co-workers (1993) on normal-weight and obese women with and without PCOS and different patterns of body fat distribution showed that hyperinsulinemia was more correlated with abdominal fat distribution, regardless of PCOS. In addition, Holte et al. (1995) indicated that hyperinsulinemia and insulin resistance might be significantly reversible through reduction or normalisation of abdominal fat depots. Therefore, it is hypothesized that abdominal (visceral) fatness may have a dominant role in determining these abnormalities in most women with PCOS, regardless of other factors, including genetic predisposition (Dunaif A., 1997).

4.3 Reproductive Hormones 

Sex Steroids And SHBG

The absence of change in total testosterone concentrations was accompanied by a rise in SHBG levels in this study. This has also been reported by Velasquez and co-workers. The increase in SHBG was not unexpected, due to the experimental data supporting the role of insulin in the regulation of hepatic SHBG production (Nestler et al., 1991). Patients demonstrated a significant increase in SHBG concentrations parallel to the decrease in total insulin response.

A weak inverse correlation between SHBG, fasting insulin levels and insulin levels in response to oGTT was revealed. This correlation was previously shown in both healthy subjects and hyperinsulinemic disease situations such as PCOS and type 2 diabetes (Sharp et al., 1991; Katsuki et al., 1996 and Jayagopal et al., 2003). The strong association prompted suggestions that a low level of SHBG could be used as a marker to identify individuals with insulin resistance.


In this study it was postulated that metformin would alter ovarian steroid production in women with PCOS. This hypothesis is based on finding that metformin treatment ameliorates the hyperinsulinemia and hyperandrogenemia (Velasquez et al., 1997; Moghetti et al., 2000) and that insulin appears to modulate the 17-hydroxylase and 17,20-lyase activities of the ovarian steroid-forming P450c17 alpha (Ehrmann et al., 1995). These enzymes are characteristically abnormally regulated in women with ovarian androgen excess in PCOS.

Contrary to postulates and results reported by most of other studies, except two (Ehrmann et al., 1997; Fleming et al., 2002), no significant changes in serum total and free testosterone levels after treatment with metformin were found.

On the other hand, a recent study by Maciel et al. (2004), that compared effects of metformin in obese and non-obese patients with PCOS, presented similar results: whereas lean women showed significant decrease in serum total testosterone levels after 6 months’ metformin therapy (1,5 g daily), obese patients demonstrated only an improvement in SHBG levels but not in serum total testosterone levels.


Hence, it could be suggested that the major effect of metformin in obese women on androgen secretion is mediated by changes in hyperinsulinemia and subsequently in SHBG rather than by direct inhibition of ovarian androgen production.


After 16 weeks of treatment, DHEAS concentrations significantly decreased in the placebo-treated group. DHEAS levels in the metformin-treated women showed a tendency to rise.

Nestler et al. (1994) demonstrated that metformin administration increased DHEAS levels by 80 % in non-obese women. Kolodziejczyk et al., (1999), Kazerooni et al., (2003) have reported the same results in obese women with normal DHEAS levels.


However, when effects of metformin were analysed separately in subgroups, a decrease in DHEAS levels was observed only in individuals with primarily elevated DHEAS concentrations (>10.3 μmol/l). The authors supposed that hyperinsulinemia inhibits adrenal 17,20-lyase and thus decreases DHEAS production.

Nevertheless, the interactions between DHEAS and insulin are still poorly understood. A negative correlation between these hormones has been reported (Schriock et al., 1988); furthermore, acute infusion of insulin or an increase of insulin induced by the oral glucose tolerance test led to a decrease in DHEAS levels in healthy subjects (Diamond et al., 1991). However, a decline in DHEAS was not observed after insulin infusion in women with hyperandrogenism and hyperinsulinemia (Falcone et al., 1990).

In the present study, the decrease of insulin levels was not accompanied by significant change in DHEAS in the metformin-treated group. It was not possible to conduct a subgroup analysis according to the baseline DHEAS levels since there were only four patients with elevated DHEAS concentrations in the entire treated group. Based on results of previous studies, one can speculate that the adrenal response to declining levels of insulin may be dependent on the baseline adrenal function.



Adrenal function can be modulated by the IGF system. Receptors and binding proteins for IGF have been identified in adrenocortical tissues; furthermore, both IGF-I and IGF-II have been shown to stimulate DHEAS production through type I IGF receptors (Fottner et al., 1998). Thus, it was expected that metformin treatment would result in a decline in DHEAS along with IGF-I. However, DHEAS declined only in the placebo-taking group and this decrease was not associated with any changes in IGF-I levels. This finding underscores the complexity of interactions between adrenal steroidogenesis and the insulin-IGF system; further studies should include determinations of IGF-II and relevant binding proteins.


The present study was not able to demonstrate any changes in FSH and LH concentrations after metformin therapy. These results are consistent with most of previous studies (Ehrmann et al., 1997; Morin-Papunen et al., 2003). We have to emphasize that the blood samples were obtained without regard to the menstrual cycle. The LH/FSH ratio was used to asses the hypothalamic function in our patients.


The prevalence of increased serum LH in PCOS ranges from 30% to 90% (Rebar et al., 1976; Franks, 1989). The reasons for these variations are unclear but probably include the heterogeneity of PCOS, a variable occurrence of associated obesity, different blood sampling frequencies, and the specificity of the gonadotropin assays.

Studies of rat pituitary cells in vitro suggested that hyperinsulinemia may potentiate both LH and FSH release to GnRH. However, these results have not been confirmed by all in vivo studies and only in two studies was a decrease in plasma insulin by the administration of metformin or troglitazone associated with a reduction in mean LH concentrations (Nestler et al., 1996; Dunaif, 1996). Additionally, obese women with PCOS have higher levels of plasma insulin, but the mean LH was unchanged or lower compared with the level in lean women (Morales et al., 1996).

Thus, the effects of insulin and consequently insulin sensitizers on gonadotropin secretion remain unclear and further research needs to be done.

4.4 Insulin And Glucose Metabolism 


According to Dunaif et al. (1987), up to 20% of obese PCOS women have impaired glucose tolerance or frank type 2 DM. This is in agreement with the results of our study, where seven (18%) women had increased 2-hours glucose concentrations in response to oGTT and one patient showed increased fasting glucose levels at the beginning of the study.

We observed no changes in glucose levels (both fasting and after glucose load) and glycosylated haemoglobin A1c concentrations in the metformin- or in the placebo- treated group. These data are consistent with results obtained in most studies in PCOS (Ehrmann et al., 1997; Fleming et al., 2002).

Before the study, we found a highly significant correlation between AUC insulin and androgenicity measured by the free androgen index (FAI). Several previous studies have also demonstrated this positive correlation (Burghen et al., 1980; Lobo et al., 1983). Furthermore, the severity of hyperinsulinemia correlates with the degree of clinical expression of PCOS syndrome (Robinson et al., 1993).


Whether hyperandrogenism results from hyperinsulinemia, or vice versa, has been debated since this correlation was demonstrated. The experiments in which hyperandrogenemia was diminished by bilateral oophorectomy (Nagamani et al., 1986) or the administration of GnRH-agonist (Dunaif et al., 1990) did not demonstrate changes in the hyperinsulinemic stage in women with PCOS. Diamanti-Kandarakis et al., (1995) reported that antiandrogen therapy did not alter insulin sensitivity in PCOS. Studies in which insulin levels have been lowered by agents that either decrease insulin secretion such as diazoxide (Nestler et al., 1989) or somatostatin (Prelevic et al., 1990); or improve insulin sensitivity such as metformin (Velasquez et al., 1994) or troglitazone (Dunaif et al., 1996) indicated a decrease of androgen production in PCOS. In summary, these findings demonstrate that disordered insulin action precedes the increase in androgens.

Most of studies demonstrated reduced insulin sensitivity in both obese and non-obese women with PCOS syndrome (Dunaif et al., 1992; Morales et al., 1996).

The effects of metformin therapy on insulin sensitivity in PCOS remain controversial. Some investigators, assessing insulin sensitivity by way of the intravenous insulin tolerance test (Unluhizarci et al., 1999) or the euglycemic clamp technique (Diamanti-Kandarakis et al., 1998; Moghetti et al., 2000), have shown significant improvement of insulin sensitivity after metformin use. Other authors have failed to confirm this (Acbay et al., 1996; Ehramann et al., 1997; Crave et al., 1995).


The possible reason for this discrepancy is that there is no consensus for the most accurate measurement of insulin sensitivity. The hyperinsulinemic euglycemic clamp technique is still regarded as the gold standard diagnostic test. Its routine use in the clinical setting is not practical for large randomized studies and was used only in three trials (see above). Other investigators used indices such as fasting glucose-to-insulin ratio, homeostasis assessment model (HOMA) and quantative insulin sensitivity check index (QUICKI). The Avignon sensitivity index (SiM) was used in this study to assess insulin sensitivity. It was shown by Chiampelli and co-workers (2005) that SiM has the best correlation with the hyperinsulinemic euglycemic clamp technique.

We did not directly measure insulin sensitivity by mean of clamp technique, however, we calculated the Avignon sensitivity index (SiM) and the measured substitute markers of insulin sensitivity, such as fasting serum insulin and serum insulin excursion after oral glucose ingestion.

There was a significant decrease in integrated insulin response to oGTT (AUC insulin) in the metformin group compared with controls. However, fasting insulin levels increased after both metformin and placebo treatment despite the amelioration of insulin action in response to oral glucose. These findings are partly in agreement with reports by Velasquez et al., (1997); Pasquali et al., (2000); Gambineri et al., (2004) who demonstrated considerable increase of insulin concentrations both fasting and after glucose load.


We analysed baseline predictors of clinical response to metformin according to efficacy of treatment on menstrual disturbances. Responders differed from nonresponders in several characteristics. Responders had significantly lower insulin levels in response to oGTT (AUC insulin) and significantly higher DHEAS levels. Fasting insulin levels showed a tendency to be lower in responders. The degree of insulin resistance (SIM index) tended to be more pronounced in nonresponders. It is important to notice that the majority of patients in both responder and nonresponder groups were insulin-resistant but the degree of insulin resistance was more severe in nonresponders. These data appear to suggest that metformin treatment may be effective in hyperinsulinemic and insulin-resistant patients with a mild degree of insulin resistance.

An unexpected finding of this study was the increase of fasting insulin levels in PCOS women after metformin treatment. Series of studies by Acbay et al., (1996); Ehrmann et al., (1997); Vandermolen et al., (2001); Fleming et al., (2002) revealed no changes in fasting insulin levels after metformin treatment. It is important to note that patients in these studies were manifestly obese (mean BMI ~ 37.6 –39.1) like women in the present study (mean BMI = 39.3).

Ehrmann and co-workers supposed that the ability of metformin to alter insulin secretion in obesity of this magnitude appears to be limited. This hypothesis could explain the results of subgroup analysis which revealed a significant decrease in insulin concentrations in response to oGTT in patients with BMI<37 after 16 weeks of metformin treatment. There were no changes in insulin concentrations in response to oGTT in patients with morbid obesity. At the same time, no changes in insulin fasting levels in women with BMI<37 were observed. Patients with BMI>37 demonstrated a significant increase of fasting insulin levels after 16 weeks of treatment with metformin.


In the present study, less obese women had higher fasting insulin concentrations and in response to oGTT insulin concentrations at baseline compared to more obese patients. This suggests that metformin treatment can be more efficient in hyperinsulinemic PCOS patients irrespective of their obesity.

A recent study of Maciel et al., (2004) comparing effects of metformin in obese and nonobese women with PCOS in a placebo-controlled trial demonstrated for the first time that non-obese women with PCOS respond better than obese women to metformin treatment at a dosage of 1500 mg/day for 6 months.

Thus, an alternative explanation for these findings and those of the present study include the possibilities that metformin was administrated in a dose or duration that was insufficient for very obese women.

4.5 Insulin Like Growth Factor-I 


The IGF-I levels remained unchanged during the study. Nevertheless, IGF-I concentrations in the metformin-treated group demonstrated a tendency to decline. This tendency was much stronger in less obese patients who demonstrated a 20% decrease (P not significant), whereas metformin treatment did not modify IGF-I levels in more obese women.

Similar results are reported by Kowalska et al. (2001) and De Leo et al. (2000). In addition, De Leo and colleagues observed a significant increase in IGF-I binding protein concentrations in women with PCOS after 30-32 days of metformin treatment. They also calculated the IGF-I/IGF-I binding protein ratio, which was significantly reduced too. Significant changes were also observed in IGF-I and IGF-I binding protein concentrations after therapy with another insulin-sensitizing drug, rosiglitazone (Belli et al., 2004).

IGF-I may contribute to ovarian hyperandrogenemia in PCOS. The in vitro-study by Nahum et al. (1995) documented that IGF-I stimulates androgen output by increasing DNA synthesis and expression of LH-receptors in theca cells. IGF-I has been shown to cause estrogen production by granulosa cells (Erikson et al., 1990) and to act synergistically with FSH and LH controlling aromatase levels in granulosa cells.


In PCOS, plasma IGF-I levels are within the normal range, whereas serum IGF-I binding protein (IGFBP-I) levels are reported to be significantly lower than in normal women. Insulin has been shown to reduce IGFBP-I concentrations, inhibiting its production in the liver. This leads to an increased bioavailability of IGF-I to the ovaries and subsequently to stimulation of androgen production.

Insulin in high concentrations can also mimic IGF-I effects by acting via IGF-I receptor (Leroith et al., 1995). Some authors suggested that this mechanism could be responsible for insulin-mediated hyperandrogenism. However, insulin has been shown to bind the IGF-I receptor with an affinity of 50-500 times lower than that of IGF-I. Moreover, Willis & Franks, 1995, using anti-insulin receptor and antitype-I IGF receptor antibodies, not only demonstrated that insulin effects on human granulosa cell steroidogenesis in vitro must be mediated via its own receptor, but also excluded both the insulin/type-I IGF hybrid receptor and the type-I IGF receptor as possible insulin action-mediated receptors.

The subgroup analysis revealed a strong positive correlation between changes in IGF-I concentrations and AUC insulin after metformin treatment in less obese women. Hence, based on the report of De Leo, Belli and colleagues, we could imagine that the lowering of insulin levels with metformin would be responsible for the increase in IGFBP-1 synthesis in the ovaries with a lowering of free IGF-1.

4.6 Leptin


Before treatment, a positive correlation between BMI and serum leptin levels was observed, as previously reported (Krotkiewski et al., 2003). The subgroup analysis revealed significant higher leptin concentrations in women with BMI>37 than in women with BMI<37. Consequently, circulating leptin appears to be a predictor for body mass and percentage of body fat.

In spite of significant weight reduction after metformin treatment, leptin levels did not decline. We could not find any explanation for this finding. It was hypothsized that the metformin effect on the leptin secretion may occur during the prolonged therapy. Another hypothesis is that changes in body mass are not accompanied by changes of leptin concentrations immediately and need some time to adapt.

Leptin is the hormone product of the obesity (ob) gene and is synthesized exclusively in adipose tissue. This hormone acts on the hypothalamus through a special receptor resulting in the suppression on food intake and increase of energy consumption. Leptin deficiency and high leptin levels are both associated with infertility (Barash et al., 1996).


A potential contribution of leptin to the pathogenesis of PCOS was suggested in a study by Brzechffa and colleagues (1996) in which a subgroup of women with PCOS appeared to have higher leptin levels than controls. However, the majority of research supported the view that serum leptin concentrations in women with PCOS are not significantly different from a control group with a similar BMI (Chapman et al., 1997; Laughlin et al., 1997; Maliqueo et al., 1999).

Several studies suggest that leptin modulates hypothalamic–pituitary–gonadal axis functions. Leptin may stimulate the release of gonadotropin releasing hormone (GnRH) from the hypothalamus and of gonadotropins from the pituitary. A synchronicity of LH and leptin pulses was demonstrated in the follicular phase of the menstrual cycle of healthy women (Licinio et al., 1998) and in patients with polycystic ovarian syndrome (Sir-Petermann et al., 1999), suggesting that leptin may regulate the episodic secretion of LH.

On the basis of our results, we could not demonstrate any correlation between leptin and LH levels. However, we measured a one-off secretion and not pulsatile hormone concentrations, as described.


Studies of insulin regulation of leptin in human yielded conflicting results. This relationship was investigated in a detailed study by Laughlin et al. (1997). They found that independently of body mass and percentage body fat, only 24-hour mean insulin concentrations contributed significantly to leptin levels. Despite this relationship and the 2-fold higher mean insulin concentrations in patients with PCOS compared with controls, serum leptin was not increased. The authors explained their results by the presence of a PCOS-specific form of insulin resistance in adipocytes, which impairs the stimulatory effect of insulin on leptin secretion. Additionally, it is considered that leptin secretion in women with PCOS is less than expected because of insulin resistance and accumulation of visceral fat that secrets less leptin than subcutaneous fat (Jacobs et al., 1999).

This hypothesis could explain our findings where the improvement of hyperinsulinemia did not alter leptin levels despite the significant weight reduction. The present results are in agreement with the studies by Mantzoros et al, (1997) and Belli et al., (2003), who also found no changes in leptin concentrations after treatment with thiazolidinedione, which is considered to be a more effective insulin-sensitizing agent.

4.7 Lipids

Triglycerides, cholesterol and lipoprotein concentrations approached the upper limit of the normal range and did not significantly change after treatment. Nevertheless, HDL cholesterol levels in the metformin-treated group showed a tendency to rise, but P=0.07 did not achieve a significant value. These findings are in line with the results obtained by Moghetti et al., (2000) who found modest improvement in HDL- cholesterol levels after six months metformin treatment. In most studies on women with PCOS, metformin and other insulin-sensitizing agent, like troglitazone produce no improvements in dyslipidemia in women with PCOS (Velasquez et al., 1997; Morin-Papunen et al., 1998; Legro et al., 2003). Interesting results have been presented in the study of Gambineri and colleagues (2004), who investigated the effects of long-term metformin and flutamide treatment, given alone and in combination.


Whereas flutamide administration showed a significant reduction of total and LDL cholesterol, the combined metformin + flutamide therapy tended additionally to increase HDL levels. The effects on total and LDL cholesterol appear to depend on the antiandrogenic effects of flutamide. Androgens are known to be directly involved in the regulation of cholesterol and lipoprotein metabolism (Anderson et al., 2002). The interaction between metformin and flutamide appear to confirm the role of insulin and androgens on HDL metabolism (Von Eckardsein, 1998).

Multiple studies have also reported that insulin resistance has been associated with decreased HDL-cholesterol and elevated triglyceride levels in PCOS women (Robinson et al., 1996; Rajkhowa et al., 1997; Mather et al., 2000). This finding is in agreement with results of correlation analysis in our study, which revealed a significant inverse correlation between fasting insulin and HDL cholesterol, and a significant positive correlation between 2-hour insulin and total cholesterol.

To summarize, HDL cholesterol concentrations demonstrated a noticeable trend to rise, and we can suppose that metformin treatment may have an impact on lipids as cardiovascular risk factors.

4.8 Menstrual Abnormalities 


After metformin treatment, in two of six patients with severe amenorrhea, the cycle returned to normal, one woman became oligomenorrheic and three patients remained amenorrheic. After placebo treatment, one of four amenorrheic patients became oligomenorrheic and three patients remained amenorrheic. There were no patients in the placebo treated group who returned to regular cycle. These results appear to suggest that metformin administration was associated with improvement of menstrual function in PCOS women with severe menstrual disorders.

Several studies have documented the restoration of regular menstrual cycles and supported that this effect is not limited to a single ethnic group or geographic area (Nestler et al., 1996; Pirwany et al., 1999; Glueck et al., 2001; Kazerooni et al., 2003). Nonetheless, most of these studies were small and uncontrolled. There are only two controlled and randomized trials demonstrating a significant improvement in regular menstrual cycles with metformin (Moghetti et al., 2000; Fleming et al., 2002).

In the present study, 50% of the women given metformin experienced no improvement of their menstrual cycle. The reasons for the differences in clinical response to metformin among the individual PCOS subjects are not easy to explain.


Kolodziejczyk et al. (2000) found that the improvement in the length of the menstrual cycle was significantly greater in women with high DHEAS levels (>400 μg/dL or >10,3 μmol/l) than in those with normal DHEAS.

Moghetti et al. (2000), based on multiple regression analysis, predicted clinical improvement for PCOS patients treated with metformin. These predictors were higher plasma insulin levels, less severe menstrual abnormalities and lower serum androstendione levels. These data strengthen the hypothesis that metformin may be effective only in the insulin-resistant subjects.

By contrast, subgroup analysis in the study by Fleming and co-workers (2002) revealed that body mass index and insulin resistance did not predict the ability to establish normal ovarian function and menstrual cycle. High SHBG concentrations and lower free index predicted the ability to establish normal ovarian function and menstrual cycle.


Pirwany et al. (1999) focused on the effect of metformin alone on menstrual cyclicity and ovulation rates. They also divided women into subgroups to see if various baseline parameters predicted a response to metformin. For example, patients with elevated testosterone levels had significant improvements in ovulation rates: 14. 8% to 38.9% (P = 0.005). Similarly, the subgroup of patients with fasting hyperinsulinemia (FI > 14 mIU/ml) at baseline had a significant increase in ovulation rate from 12.5% to 39.4% (P = 0.012). However, those with normal baseline insulin concentrations did not show an improvement.

We would draw a parallel between our own finding and those obtained by Douchi et al., (2002). The author and his colleagues examined eighty-three obese women (mean BMI = 31.9) who were divided into two groups according to their menstrual status: one with menstrual disorders and the other group with regular menstruation. They compared anthropometric characteristics between the two groups and found that women with irregular menses had higher trunk fat mass and trunk-leg fat ratio than controls. It was postulated that upper body obesity and not a degree of obesity, are associated with menstrual disorders.

We provided a subgroup analysis in order to test if body weight is an important factor for successful treatment with metformin. There was no difference apparent between women with BMI<37 and morbid obese patients in the menstrual response to treatment. This finding suggests that body weight possibly is not a predictive factor for menstrual response to metformin treatment.


The data of the present study are in agreement with results obtained by Fleming and colleagues (2005). The authors showed no difference between obese and morbid obese groups in the proportions attaining a normal menstrual rhythm.

In the present study, the subgroups contained small numbers of subjects (seven and twelve patients each). Large numbers of patients will be required to show convincing differences in many parameters.



Therefore, we concluded that until now there are not enough data to limit metformin treatment to a specific subgroup of women with PCOS. One of the factors to be determined is the relationship between dose and body mass index.

4.9 Metformin and Pregnancy Rates

There were three conceptions in three patients during the study, and one miscarriage in the first trimester. The distribution of pregnancies was: metformin, 1 of 23 patients; and placebo, 2 of 23 patients. This finding is a surprising observation, taking into account that menstrual cyclicity improved in the metformin-treated group. However, there were some limitations that could explain the results of the study. First, not all patients were tested for tubal factors of infertility. Second, spermiogramm analysis was done only in partners who had been already treated within the scope of in-vitro fertilisation. Oligozoospermia or azoospermia were diagnosed in twelve (32.4%) partners (both placebo and metformin) which contributed to low pregnancy rates.

On the other hand, we assume that the16 week period of metformin treatment was too short to influence the reproductive function in our patients more profoundly.


The clinical pregnancy rate reported by trials comparing metformin as a single agent with placebo did not show evidence of benefit (Ng et al., 2001; Fleming et al., 2002). However, the addition of metformin to clomiphene citrate (CC), results in an improved ovulation and pregnancy rate in both unselected and CC-resistant PCOS patients. In the trial by El-Biely and co-workers (2001), metformin in combination with CC was compared to CC alone in 90 women, with both ovulation and pregnancy as an end point. Patients in the CC plus metformin group had a higher ovulation rate (80% vs. 65%), higher pregnancy rate per patient (29% vs. 8%), and a lower rate of ovarian hyperstimulation syndrome (9% vs. 69%).

Taken together, our findings suggest that metformin treatment may restore menstrual abnormalities but the duration of the trial and the dose of metformin were possibly insufficient to influence fertility in obese women with PCOS.

© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.
DiML DTD Version 4.0Zertifizierter Dokumentenserver
der Humboldt-Universität zu Berlin
HTML generated: