- Split View
-
Views
-
Cite
Cite
Stefano Palomba, Angela Falbo, Fulvio Zullo, Francesco Orio, Evidence-Based and Potential Benefits of Metformin in the Polycystic Ovary Syndrome: A Comprehensive Review, Endocrine Reviews, Volume 30, Issue 1, 1 February 2009, Pages 1–50, https://doi.org/10.1210/er.2008-0030
- Share Icon Share
Abstract
Metformin is an insulin sensitizer widely used for the treatment of patients affected by type 2 diabetes mellitus. Because many women with polycystic ovary syndrome (PCOS) are insulin resistant, metformin was introduced in clinical practice to treat these patients also. Moreover, metformin’s effect has other targets beside its insulin-sensitizing action. The present review was aimed at describing all evidence-based and potential uses of metformin in PCOS patients. In particular, we will analyze the uses of metformin not only for the treatment of all PCOS-related disturbances such as menstrual disorders, anovulatory infertility, increased abortion, or complicated pregnancy risk, hyperandrogenism, endometrial, metabolic and cardiovascular abnormalities, but also for the prevention of the syndrome.
- I.
Introduction
- A.
Background
- B.
Rationale and mechanism of action
- C.
Pharmacokinetics
- D.
Therapeutic regimens
- E.
Adverse effects
- F.
Potential predictors
- A.
- II.
Menstrual Disorders
- III.
Anovulation
- A.
Metformin as first-step treatment
- B.
Metformin as second-step treatment
- C.
Metformin in patients who receive gonadotropins
- A.
- IV.
Adverse Pregnancy Outcomes
- A.
Miscarriage
- B.
Gestational diabetes
- C.
Pregnancy-induced hypertension and preeclampsia
- D.
Poor infant outcomes
- A.
- V.
Endometrial Abnormalities
- A.
Fertility implications
- B.
Cancer implications
- A.
- VI.
Hyperandrogenism
- VII.
Malignancies
- VIII.
Quality of Life Impairment
- IX.
Obesity
- X.
Cardiovascular Risk
- A.
Cardiopulmonary impairment
- B.
Diabetes mellitus
- C.
Hypertension
- D.
Dyslipidemia
- E.
Impaired fibrinolysis
- F.
Chronic inflammation
- G.
Endothelial impairment
- H.
Syndrome X
- A.
- XI.
Other Organ Impairment
- A.
Liver impairment
- B.
Thyroid impairment
- C.
Cognitive function
- D.
Sleep disturbances
- A.
- XII.
High-Risk Patients for PCOS
- XIII.
Future Perspectives
- XIV.
Summary
I. Introduction
POLYCYSTIC OVARY SYNDROME (PCOS) is a heterogeneous disorder. Although, essentially characterized by hyperandrogenism, ovarian dysfunction, and polycystic ovarian morphology, its definition remains somewhat fluid and controversial. Until now, two main criteria have been proposed: the National Institute of Health (NIH)/National Institute of Child Health and Human Development (NICHD) criteria proposed in April 1990 (1), and the European Society for Human Reproduction and Embryology (ESHRE) and the American Society for Reproductive Medicine (ASRM) proposed in May 2003 (2).
The 1990 NIH criteria (1) indicated the presence of hyperandrogenism and/or hyperandrogenemia and chronic anovulation with exclusion of related ovulatory or other androgen excess disorders (e.g., thyroid dysfunction, hyperprolactinemia, androgen-secreting neoplasms, or nonclassic adrenal hyperplasia) as PCOS features. The presence of polycystic ovaries (PCO), with morphology assessed by ultrasound, was not included as a diagnostic criterion because it was considered by some authors to be a finding with a low specificity (3). PCO can be present in women with normal ovulation and sex hormone levels, whereas women with all the endocrine PCOS features occasionally can have normal ovarian morphology as assessed by ultrasound examination (3).
In 2003, an international conference in Rotterdam, The Netherlands (2), proposed to revise the PCOS and PCO criteria. In particular, the ESHRE/ASRM conference proceedings expanded the definition of PCOS, establishing that it could be diagnosed, even after the exclusion of related disorders, by at least two of the following features: 1) oligo- or anovulation; 2) clinical and/or biochemical signs of hyperandrogenism; and 3) PCO. All women considered to be affected by PCOS, according to the NIH criteria, would also be identified through the Rotterdam criteria, which included women with PCOS excluded by the NIH criteria: those with PCO affected by hyperandrogenism and ovulatory cycles or chronic anovulation and normal androgen levels. More recently, an expert panel of the Androgen Excess and PCOS Society (AEPS) recommended that PCOS should be considered an androgen excess disorder and that the NIH diagnostic criteria should be used with some modifications, taking into consideration the concerns expressed in the proceedings of the 2003 Rotterdam conference (4). In particular, the AEPS Task Force concluded that PCOS was above all a disorder of androgen biosynthesis, utilization, and/or metabolism in women. Thus, the AEPS criteria for PCOS are fulfilled by the combination of hyperandrogenism with at least one of the following: ovarian dysfunction (including infrequent or irregular ovulation or anovulation) and PCO (4). In addition, the Task Force noted that support for this criterion is based on the risk for metabolic morbidity in the disorder, not on whether hyperandrogenism per se is present or not (4).
The phenotypes considered to represent the disorder may vary according to the definition used, and several authors (5, 6) questioned the points of agreement between the new phenotypes according to the ESHRE/ASRM criteria and the features classically related to the syndrome. Ovulatory women with PCO seem to be less insulin-resistant than anovulatory women with PCO (7–9). Barber et al. (10) showed that patients with PCO, chronic anovulation, and normal androgen levels are not insulin-resistant. Moreover, only women who fulfil the NIH criteria have been demonstrated to be at high risk for metabolic sequelae of PCOS, i.e., glucose intolerance (11) and metabolic syndrome (12).
Notwithstanding the extreme variability of the diagnostic criteria adopted, the main clinical features related to the syndrome, albeit not pathognomonic, are menstrual disorders, anovulatory infertility, and signs of hyperandrogenism (2). Moreover, PCOS is characterized by a number of other clinical traits and long-term sequelae.
The reduced fertility observed in PCOS patients cannot be attributed to anovulation alone. Other factors may be operant in PCOS to lower fertility in these women, including reduced oocyte and/or embryo quality, defects in endometrial development, and implantation abnormalities (13). In addition, pregnant women with PCOS seem to have a significantly higher risk for miscarriage, gestational diabetes mellitus (DM), pregnancy-induced hypertension (PIH), preeclampsia (PE), and poor infant outcome (13). Finally, even if no well-controlled prospective data have demonstrated a higher mortality for cardiovascular disease (CVD) in PCOS patients, an increased prevalence of several surrogate end-points, mainly type 2 DM and metabolic syndrome, have been demonstrated in PCOS.
The pathogenesis of PCOS is still unclear, and a complex of genetic and environmental factors may be involved. In particular, family studies demonstrated that PCOS is significantly more prevalent among family members than in the general population (14, 15). Among first-degree female relatives of patients with PCOS, 35% of premenopausal mothers and 40% of sisters were affected by the disorder (16). These rates are significantly higher than the 6–7% observed in the general population (17). Other than PCOS itself, increased androgens, insulin secretion, and insulin resistance also appear to be under significant genetic control, as established by family studies (15, 18–24). Several genes have been logically chosen as candidates, but most of these studies were performed in small cohorts and confounded by several biases. Thus, no particular gene is universally recognized as importantly contributing to PCOS risk, and several environmental factors probably modulate the clinical expression of genetic predisposition. In particular, obesity, insulin resistance, prenatal androgen excess during early gestation, birth weight, precocious pubarche, dietary intake, lifestyle, and stress may act on the genetic substrate. Obesity is the most important factor contributing directly to insulin resistance and hyperinsulinemia. Low birth weight modifies the relationship between insulin resistance and PCOS and seems to increase the risk of developing PCOS. Diet composition, eating disorders, and psychological stress seem to be related to the syndrome, even if no causality link has been clearly demonstrated. However, insulin resistance with compensatory hyperinsulinemia is probably a cornerstone of the pathogenesis of the PCOS, and this recognition has led to a much better understanding of the syndrome and associated short- and long-term complications.
The molecular mechanisms of insulin resistance in PCOS differ from those in other common insulin-resistant states, such as obesity and type 2 DM (25). Despite the compensatory hyperinsulinemia that accompanies insulin resistance in nondiabetic individuals, insulin-mediated glucose uptake remains subnormal (26). Interestingly, insulin resistance occurring in PCOS patients is not evident in all tissues, but it seems to have a tissue peculiarity (26). Specifically, in skeletal and adipose tissue, insulin resistance led to a decreased glucose uptake and increased lipolysis, respectively, whereas, the ovary, adrenal, liver, and skin seem to remain insulin sensitive (26). Furthermore, recently a peripheral insulin resistance, which is related to ovarian (27) and uterine (28, 29) abnormalities detected in PCOS patients, has been demonstrated. Compensatory hyperinsulinemia exerts androgen-stimulating effects on the ovaries and adrenal glands and is responsible for the main clinical manifestations of the syndrome (see Section I.A).
Because insulin resistance with compensatory hyperinsulinemia is widely considered the pivotal feature of PCOS, insulin-sensitizing drugs (ISDs) are introduced for the treatment of these patients.
A meta-analysis of randomized controlled trials (RCTs) (30) published in 2003, having as its aim the reproductive and metabolic effects of ISDs in PCOS, demonstrated that only metformin had multiple placebo-controlled trials suitable for analysis. Metformin, in fact, is the most extensively studied ISD in the treatment of the short- and long-term sequelae related to the syndrome, and it is also used as a preventive treatment in girls at high-risk for PCOS development. In particular, if we perform a bibliographic search using as key words “ISDs” and “PCOS,” the rate of publications on metformin use in PCOS is approximately 90% of overall publications (Fig. 1). In particular, this figure is also sustained when we select only meta-analyses and RCTs, which are the cornerstones of evidence-based medicine.
The aim of the current review was to evaluate present evidence-based uses and potential future uses of metformin in PCOS patients. In particular, we analyzed the effects of metformin administration on all features and on all short- and/or long-term PCOS complications. The use of metformin as a preventive treatment in girls at high risk for PCOS will be also evaluated. For the bibliographic search, we used the U.S. National Library of Medicine and the NIH website to search abstracts having as a key word “metformin,” without limits for time of publication or language. The methodology used for the current comprehensive review consisted in searching all available meta-analyses for each specific issue, updating them with more recent RCTs until publication. In case of the absence of strongly evidence-based data, nonrandomized prospective, uncontrolled prospective, retrospective, and finally experimental studies were considered. When data regarding the use of metformin in PCOS patients were missing, we reported evidence from diabetic populations treated with metformin as an analogy.
A. Background
Metfomin (1,1-dimethylbuguanide hydrochloride) (Fig. 2) is a biguanide currently used as an oral antihyperglycemic agent. Metformin was introduced in 1957, but it only became available for use in the United States in 1995. To date, metformin administration is approved by the U.S. Food and Drug Administration (USFDA) to treat type 2 DM, and the safety profile is probably better than those observed with other ISDs (31).
In 1994 Velazquez et al. (32) first evaluated the effects of metformin administration in 26 obese PCOS patients to investigate the role of insulin resistance in the pathogenesis of the syndrome. After 6 months of metformin at a dose of 1500 mg/d, they reported a significant reduction in circulating androgen levels and body weight. In addition, the efficacy of metformin in inducing regular menstrual and ovulatory cycles in these patients was demonstrated (32).
Presently, even if more gynecologists and endocrinologists use metformin to treat PCOS patients, the AEPS (33) suggests that metformin could be used to treat and to prevent progression to impaired glucose tolerance (IGT) in PCOS patients, and the American Association of Clinical Endocrinologists’ guidelines (34) recommend metformin as an initial intervention in overweight and obese patients with PCOS; to date, neither in Europe nor in the United States has metformin been approved for this type of treatment in PCOS patients.
In March 2007, in Thessaloniki, Greece, a second international ESHRE/ASRM-sponsored PCOS Consensus Workshop Group (35) was held to address the therapeutic challenges raised in women with infertility and PCOS and to answer important questions regarding the efficacy and safety of various treatments available for these women, including ISDs. As in the Rotterdam meeting, a panel of international experts was invited to discuss treatment in women affected by PCOS to reach at a consensus regarding therapy. This workshop group concluded that ISDs should not be used as first-choice agents in ovulation induction of women with PCOS, and their use should be restricted to those patients with IGT. This is probably due to the fact that knowledge regarding its effects and regimens of administration in PCOS patients is still incomplete.
More recently, Nestler (36), in an interesting editorial, criticizes previous conclusions (35) maintaining the open debate on this issue.
B. Rationale and mechanism of action
The rationale for the use of an ISD, such as metformin, in treatment of patients with PCOS (as summarized in Figure 3) arises from the knowledge that insulin resistance with compensatory hyperinsulinemia has provided an insight into the pathogenesis of PCOS, although, not an essential criteria for the diagnosis of PCOS.
Insulin resistance and secondary hyperinsulinemia affect approximately 65 to 70% of PCOS women (37). PCOS patients are more insulin resistant and hyperinsulinemic than age- and body mass index (BMI)-matched non-PCOS controls, regardless of BMI (38). In addition, pancreatic β-cell dysfunction with an increased basal secretion of insulin and an inadequate postprandial response has been described in PCOS patients, also after weight loss, despite an improvement in glucose tolerance (39).
The major defect of insulin action in PCOS patients is probably a decrease in insulin sensitivity secondary to a postbinding abnormality in insulin receptor-mediated signal transduction, with a less substantial, but significant, alteration in receptor binding (25). The decreased insulin sensitivity in patients affected by PCOS is potentially an intrinsic defect in genetically predisposed subjects. In fact, independent of obesity, body fat distribution, sex hormone levels, and metabolic abnormalities, there may be genetic abnormalities in the regulation of insulin receptor phosphorylation, resulting in increased insulin-independent serine phosphorylation and decreased insulin-dependent tyrosine phosphorylation (25). Although insulin resistance is present independently from BMI, obesity may further exacerbate insulin resistance and exert a direct and synergistic or indirect deleterious impact on glucose metabolism with a possible worsening of the clinical and biochemical features of the PCOS.
Insulin resistance and secondary hyperinsulinemia might influence the spectrum of the disorders related to PCOS by means of various and still unclear mechanisms. Certainly, the most debated argument is the effect of insulin disorder on specific characteristics of the syndrome, i.e., ovulatory dysfunction and hyperandrogenism.
Central and peripheral mechanisms have been hypothesized to explain the role of insulin resistance on the pathogenesis of the PCOS.
At the central level, insulin was implicated in the regulation of LH secretion, but contrasting data rose from in vitro and in vivo studies (25). Particularly, rat pituitary cells preincubated with insulin showed an increased response of LH after GnRH administration. On the other hand, insulin infusion in PCOS women did not cause significant alterations in LH secretion or release after GnRH stimulation (25). In addition, the experimental reduction of hyperinsulinemia induced a reduction in serum LH levels (25), even if it is unclear whether decreased serum LH is a direct result of lowered circulating insulin levels or secondary to the feedback effect of increased ovarian estrogen production due to resumed folliculogenesis. Moreover, these data were not confirmed by other studies (25). Recently, the influence of insulin on gonadotropin secretion in PCOS women and healthy controls was evaluated, and no alteration, due to insulin infusion, was observed in both groups over an interval of 12 h in episodic gonadotropin secretion and LH response to multidose GnRH stimulation by use of hyperinsulinemic-euglycemic clamp technique (25).
At the peripheral level, insulin resistance influences hepatic, muscle, and ovarian function.
At the hepatic level, hyperinsulinemia inhibits the synthesis of SHBG, increasing the free androgen amount and consequently the peripheral androgen action; hyperinsulinemia also inhibits the hepatic secretion of the IGF binding protein (IGFBP)-1, leading to increased bioactivity of IGF-I and -II, two important regulators of ovarian follicular maturation and steroidogenesis (25). The IGF-I and -II systemic increase augments ovarian androgen production from theca cells by acting on IGF-I receptors (25).
At the muscular level, insulin resistance is associated with reduced expression of genes involved in mitochondrial oxidative metabolism (40). Moreover, experimental data indicated that insulin resistance induced transcriptional alterations in insulin signaling pathways, free fatty acid (FFA) metabolism, and calcium homeostasis in PCOS patients (40).
At the ovarian level, compensatory hyperinsulinemia might contribute to anovulation directly by interfering with follicular development, causing premature follicular atresia and antral follicular arrest (25), and indirectly by affecting gonadotropin effectiveness for abnormal intraovarian environment (25). On the other hand, increased insulin binds to the IGF-I receptors, again enhancing the androgen theca cell production in response to LH stimulation (25). Insulin stimulates the basal androgen secretion by cultured human theca cells from hyperandrogenic women, suggesting that the stimulatory effect of insulin on thecal steroidogenesis is augmented in PCOS (25). Insulin may also increase endogenous androgen concentrations by increasing the cytochrome P450c17α activity, a key enzyme in the biosynthesis of ovarian and adrenal androgens having both 17α-hydroxylase and 17,20-lyase activities. In ovarian theca cells, the P450c17α 17α-hydroxylase activity converts progesterone to 17α-hydroxyprogesterone, whereas its 17,20-lyase activity converts 17α-hydroxyprogesterone to androstenedione. Androstenedione is then converted to testosterone by the enzyme 17β-reductase. Thus, the increase in P450c17α activity is accompanied by an increase in the serum testosterone concentration. In this regard, in PCOS this abnormality may be responsible for exaggerating 17-hydroxyprogesterone response to stimulation by GnRH analog (41).
Intraovarian androgen levels may modulate follicular function by acting directly on the ovary. In particular, short exposition to androgens suppresses granulosa cell apoptosis prolonging survival of small antral follicles, which in normal-cycling women either grow rapidly to become dominant follicles or collapse in atresia (42, 43). On the other hand, a prolonged exposition to androgens, such as in PCOS, is responsible for the collapse of these follicles into the ovarian stroma, leading to stromal hypertrophy and promoting the process of ovarian atresia and anovulation (44).
Even if this rationale initially induced several researchers to use metformin in PCOS patients, no clear evidence has been provided yet regarding the effectiveness of metformin related to its insulin-sensitizing effects. On the other hand, other non-insulin-sensitizing mechanisms can be supposed or cannot be excluded.
In Figure 4, all main potential mechanisms of action of metformin in PCOS patients are shown. Other mechanisms of metformin action on specific districts will be detailed in specific paragraphs.
In an early meta-analysis on ISDs (30), metformin administration in PCOS patients, when compared with placebo, resulted in a significant decrease in fasting glucose and insulin levels, as well as insulin levels in the area under the curve (AUC) after oral glucose administration. Mechanistic studies have shown that metformin improves insulin-mediated glucose disposal in women with PCOS (45). Furthermore, to understand metformin’s main mechanisms of action better, several data obtained in patients with type 2 DM could be translated in PCOS patients.
Metformin has been shown to reduce basal hepatic glucose production by 9 to 30% in patients with type 2 DM through multiple effects (46). The stimulating effect of this drug on nonoxidative glucose metabolism, including glycogenosynthesis, conversion to lactate, and incorporation to triglycerides, is responsible for the insulin-stimulated glucose utilization estimated to be more than 50% in patients with type 2 DM (47) and first-degree relatives of patients with type 2 DM (48).
Experimental studies (49) evaluating glucose production from collagenase-isolated hepatocytes of starved rats demonstrated that therapeutic concentrations of metformin potentiate the antigluconeogenic effect of insulin by enhancing the suppression of gluconeogenesis and by reducing the glucagon-stimulating gluconeogenesis. Increased intestinal use of glucose and decreased FFA oxidation may also contribute to metformin action in reducing gluconeogenesis (50). An accelerated FFA oxidation promotes hepatic gluconeogenesis by providing acetyl-coenzyme A, ATP and a reduction in equivalents (50), and reduces glucose utilization in peripheral tissues secondarily to an inhibition of pyruvate dehydrogenase activity (50). On the other hand, the decreased FFA oxidation due to metformin treatment (50) decreases hepatic gluconeogenesis and increases glucose uptake and oxidation in skeletal muscle (50), improving insulin sensitivity.
However, the nature of the mechanism of metformin action on hepatic glucose production remains unclear. Data from in vitro studies suggest several effects of metformin on the reduction of hepatic gluconeogenesis through short- (metabolic) and long-term (gene expression) effects.
Among the short-term effects, a decrease in the uptake of gluconeogenic substrates by liver cells was shown (51, 52). In particular, in vitro studies demonstrated the inhibition by metformin of alanine uptake, related to a reduction in the Na+/l-alanine transport system (51), and of hepatic lactate uptake (52) in isolated rat liver hepatocytes. Other data showed that metformin decreased glucose production in the liver of diabetic rats by reducing flux through pyruvate carboxylase-phosphoenolpyruvate carboxykinase and perhaps also by an increased conversion of liver pyruvate into alanine (53).
Experimental data have demonstrated that one of the metformin mechanisms of action also involves the 5′-AMP-activated protein kinase (AMPK) pathway (54–56). AMPK is a pleiotropic serine/threonine kinase that acts as a fuel gauge in regulating energy metabolism, especially under stress conditions where biosynthetic pathways are blocked by phosphorylation of downstream AMPK substrates. In particular, AMPK activation restores cellular adenosine-5′-triphosphate (ATP) level by switching on the catabolic pathway and switching off catabolic pathways (57). At this regard, metformin enhances flux through pyruvate kinase (58) by activating the AMPK pathway (57) and inhibits hepatic gluconeogenesis through an AMPK-dependent regulation of the orphan nuclear receptor SHP (59).
Although the mechanism by which metformin activates the AMPK pathway is not clear, phosphorylation of threonine in AMPK is certainly necessary for its action as shown by in vitro study (60). In the liver of adult mice, the AMPK activation is mediated by action on a proximal kinase-serine-threonine protein kinase previously termed LKB1 (61). These molecular mechanisms probably explain the pleiotropic action of metformin resulting in decreased glucose production and increased FFA oxidation, not only in hepatocytes but also in skeletal muscle cells (60) and in ovarian tissue (62).
On the other hand, to support the hypothesis on long-term clinical effects of metformin, in vitro experiments on cultured starved rat hepatocytes indicated that metformin can regulate the expression of specific hepatic genes in an insulin-independent manner (63). This last study (63) showed that metformin affects the expression of genes encoding regulatory proteins of the phosphoenolpyruvate/pyruvate cycle. Indeed, metformin enhanced the expression of l-pyruvate kinase, whereas it decreased the phosphoenolpyruvate carboxykinase gene expression (63).
Experimental studies on human muscle cell cultures (64) and rat adipocytes (65) demonstrated that metformin acts by increasing glucose uptake through the glucose transport system. In fact, it facilitates the translocation of glucose transporters (GLUTs) from intracellular sites to the plasma membrane. In addition, metformin augmented the uptake of the nonmetabolizable sugar 3-O-methylglucose approximately, to the same extent, as the uptake of 2-deoxyglucose, suggesting a stimulation of hexose transport (64). The PCOS-related insulin-resistant state is associated both with decreased ability of insulin to stimulate glucose disposal into target cells and with reduced glucose response to a given amount of insulin. In fact, a reduction of the transporter GLUT-4 expression in the insulin target tissues has been demonstrated (64). An experimental study (66) showed that a 6-month metformin course induces a significant improvement of the GLUT-4 mRNA expression in the adipose tissue of PCOS patients, demonstrating that metformin action improves the peripheral insulin signal transduction.
Besides its effects on glucose metabolism, metformin may also act on the insulin levels and, specifically, on insulin receptor (46, 50, 67). In fact, significant reduction of both insulin and proinsulin levels in patients with type 2 DM has been observed.
Controversial data are available regarding the effect of metformin on the extent of insulin binding (68–72). In fact, after addition of metformin to tissue, some authors found no metformin effect on the insulin receptor (69, 70), whereas others reported that metformin elevated receptor tyrosine kinase activity (71, 72). In type 2 DM patients, metformin was demonstrated to have no effect in erythrocyte insulin receptor binding, but it increased both basal and insulin-stimulated insulin receptor tyrosine kinase activities of solubilized erythrocyte insulin receptors after 10 wk of treatment (73).
However, the increase of insulin binding in various cell types after metformin administration does not directly correlate with clinical and metabolic effects. In fact, Stith et al. (72) in an experimental model of Xenopus oocytes obtained interesting results regarding the intracellular mechanisms of action of metformin. Specifically, metformin must be internalized to act, and it stimulates both insulin and IGF-I receptors in oocytes (75) and mammalian tissue (74). Metformin can also act independently from insulin to stimulate the receptor tyrosine kinase and the γ-isoform phospholipase C activity and inositol 1, 4, 5-tris-phosphate production probably through inositol 1,4, 5-tris-phosphate and Ca++ elevation (74). The addition of metformin was found to mimic insulin action in cultured human skeletal muscle (76), human erythrocytes (77), rat cardiomyocytes (78), rat adipocytes (79), and mouse soleus muscle (80).
Part of the efficacy of metformin in PCOS might be related to direct actions on steroidogenesis such as their insulin-sensitizing effects, although controversy exists as to whether metformin has direct effects on steroidogenesis itself (81, 82). In fact, ovulation may be a result of a direct action of metformin on the ovary that leads to normal steroid production and likely steroid feedback effects that include a lowering of LH and androgen levels. On the other hand, metformin could act directly, improving abnormal steroidogenesis (systemic and/or local/ovarian) and, subsequently, the follicular function. Several data suggest that metformin could act on hyperandrogenism by interfering both with direct and specific mechanisms on peripheral androgen-secreting organs and with free androgen fraction-regulating systems. In fact, a reduced ovarian and adrenal secretion of androgens, a reduced pituitary secretion of LH, and an increased liver SHBG production seem to be the mechanisms mediating metformin effect on hyperandrogenism (50).
A direct effect of metformin on androgen production by thecal cells could be hypothesized. In vitro experiments demonstrated that metformin significantly inhibits androstenedione and testosterone production from theca cells through inhibition of the steroidogenic acute regulatory protein and 17α-hydroxylase expressions (83). Nestler and Jakubowicz (84) demonstrated in obese PCOS patients that metformin administration reduces the fasting and glucose-stimulated insulin levels and concomitantly decreases ovarian cytochrome P450c17α activity, inducing a reduction in the serum free testosterone concentration. These results were also confirmed in lean PCOS patients (85).
The reduction in insulin levels after metformin treatment in PCOS patients is associated with an increase in IGFBP-1 and a decrease in the IGF-I/IGFBP-1 ratio. IGF-I has autocrine and paracrine mechanisms in stimulating estrogen production by granulosa cells (86) and acts synergistically with FSH and LH in controlling granulosa cell aromatase levels. Thus, by reducing plasma insulin levels and IGF-I availability to the ovaries, metformin may modify the hyperandrogenic intrafollicular milieu recognized in PCOS. Recent in vitro studies have confirmed a significant effect of metformin on insulin and IGF-I pathway in granulosa cells (87). Moreover, metformin may inhibit ovarian gluconeogenesis through a direct effect, thus reducing ovarian steroidogenesis and, more specifically, androgen production (84). In particular, in bovine granulosa cells, metformin decreases steroidogenesis and MAPK3/MAPK1 phosphorylation through AMPK activation (55). Experimental evidence (54) indicates that AMPK could also be implicated in reproductive functions such as granulosa cell steroidogenesis and nuclear oocyte maturation in several species.
Keeping in mind that adrenal steroidogenic pathways are regulated much like those in the ovary, it is intuitive to hypothesize an effect of metformin on adrenal androgen production, too. In PCOS patients, 1 month of metformin administration-regulated adrenal enzyme activities by reducing the ratio of 17α-hydroxyprogesterone to progesterone, which indicates 17α-hydroxylase activity, and the ratio of androstenedione to 17α-hydroxyprogesterone, which indicates 17,20-lyase activity (88).
On the other hand, the effect of metformin on hyperandrogenism could also be due to a direct effect of metformin on LH secretion. In particular, metformin administration modulates LH secretion, decreasing LH pulse amplitude but not pulse frequency (89). In this regard, it has been suggested that metformin acts at the hypothalamic level on AMPK pathway. Hypothalamic AMPK is a key regulator of food intake in mammals, even if its role in reproduction at the central level is unclear (54). In the rat model, metformin increases in a dose-dependent and time-dependent way the AMPK activation by phosphorylation at Thr172 in GnRH neurones and, thus, the modulation of GnRH release (54).
Lastly, some metabolic hormones, such as adipokines (leptin, resistin, adiponectin) and ghrelin, which are involved in the control of the reproductive functions at the hypothalamus-pituitary-gonadal axis level (90) and of several processes of tumorogenesis (91), might act through AMPK signaling (92). Thus, AMPK could be not only one of the signaling pathways controlling the interactions between energy balance and reproduction and/or tumorogenesis, see also Section VII, but also a pivotal link explaining the effects of metformin at the central level. In other words, metformin could act on hypothalamic AMPK, exerting both a direct effect and an indirect effect via paracrine/endocrine action of adipokines and ghrelin.
In conclusion, even if no study has currently demonstrated a cause-effect relationship for metformin effectiveness or excluded its other mechanisms of action, the beneficial effects of this drug on reproductive and nonreproductive parameters are seen only in association with a reduction in circulating insulin levels, consistent with the hypothesis that this is the main mechanism by which metformin acts in patients affected by PCOS. Furthermore, it is possible that metformin might also act by other parallel mechanisms not necessarily implicated with insulin resistance. In fact, several experimental data suggest an indirect effect of metformin by ovarian (and adrenal) steroidogenesis suppression and/or a direct cellular action by AMPK signaling.
C. Pharmacokinetics
The pharmacokinetic properties of metformin have been evaluated in patients with type 2 DM and in healthy volunteers by using oral and iv preparations, although metformin is always administered orally in clinical practice. Metformin pharmacokinetics has been described as a two-compartment open model with first-order absorption (93). It has an incomplete gastrointestinal absorption ranging from 20 to 30% (93). Absorption is complete within 6 h from administration and is slower than the elimination; thus it represents the phase that determines the drug’s disposal rate (93). Metformin has been shown to have a dose-dependent absorption in humans (93). Furthermore, the absorption is proportionally greater for a dose of 0.5 g than for 1.5 g (93). In fact, transport occurs almost exclusively via the paracellular route (90%), and this transport is saturable (94).
The bioavailability of metformin is limited to 50–60% because the amount available may result from presystemic clearance or binding to the intestinal wall (93). After absorption of 1.5 g of metformin, a linear pharmacokinetics of metformin was reported in both diabetic and nondiabetic subjects (93).
Once absorbed, metformin is rapidly accumulated in the esophagus, stomach, duodenum, salivary glands, and kidneys. No binding with plasma proteins was reported; however, a slight association with blood cells was hypothesized (93).
Metformin is not metabolized but is excreted by the kidney with a mean 4- to 8-h half-life in healthy volunteers (93). Ranges of values for kidney and total clearance are reported to be 20.1–36.9 and 26.5–42.4 liters/h, respectively (93), indicating active tubular secretion of the drug.
Metformin freely passes the placenta by a carrier that transports cationic compounds bidirectionally, with a higher transfer rate from the fetal to the maternal compartment (95), resulting in the exposure of the fetus to metformin therapeutic concentrations (96). Furthermore, no effect on human placental glucose uptake or transport has been demonstrated (97).
The concentrations of metformin in breast milk are generally low. The mean infant exposure to metformin was reported to be in the range 0.28–1.08% of the weight-normalized maternal dose, well below the level of concern for breastfeeding (98).
The metformin pharmacokinetics can be influenced by several factors. In a rat model, metformin pharmacokinetics changes according to insulin resistance and/or presence/absence of streptozotocin-induced DM with significant variation in hepatic and renal clearance (93). On the contrary, no difference in metformin disposition seems to be present in patients with type 2 DM or by using different oral preparations (93). Coadministration of food was reported to decrease slightly the rate of metformin absorption (93). Drug interactions have also been reported: coadministration of guar gum was reported to interfere significantly with metformin absorption, α-glucosidase inhibitor acarbose significantly reduces the bioavailability of metformin (99), and histamine H2-receptor antagonist cimetidine has a competitive action on renal tubular secretion (100). Excretion is prolonged in patients with renal impairment and correlates with creatinine clearance (93). Metformin is transported by at least two organic cation transporters (OCTs), OCT1 and OCT2 (101). These OCTs are saturable in the rat model (102), and genetic polymorphism in these transporters was found to be associated with changes in pharmacokinetics and pharmacodynamics (103). In addition, genetic variants of the gene SLC22A2 encoding for the organic OCT2 protein result in a reduced renal clearance and, thus, in increased metformin plasma concentrations (103).
D. Therapeutic regimens
Metfomin is available in two formulations: immediate and extended-release. Metformin at immediate-release is commercialized as 500-, 850-, and 1000-mg tablets, whereas the drug at extended-release is available as 1000- and 2000-mg tablets.
Metformin is indicated in patients over the age of 10 yr, and the extended-release preparation is indicated in those over the age of 17 yr.
An extremely variable target dose of 1500 to 2550 mg/d was proposed. A dose-finding study (104) showed that metformin, at the upper limit of 2000 mg/d, had maximal benefit in lowering plasma glucose and glycated hemoglobin in diabetic patients. However, to date, no dose-finding study is available for a PCOS population, probably due to the clinical end-points used to test the effectiveness of metformin administration, which are various, and their intercorrelations are unknown.
Therapeutic regimens of metformin administration are not well standardized in clinical practice, and heterogeneous protocols were used in the various studies available in literature.
To minimize the drug-related adverse effects (see Section I.E), metformin is best taken on an empty stomach, starting with a low dosage and gradually increasing over a period of 4–6 wk. Nestler et al. (105, 106) suggested administering immediate-release metformin initially at a low dose at meals, beginning with 500 mg at dinner for 3–4 d, and then increasing by 500 mg every 3–4 d up to a maximal dosage of 1000 mg twice daily.
Extended-release metformin is usually taken with the evening meal and the only suggestion to minimize potential adverse affects is to divide the tablets into two administrations.
Previous data on patients with type 2 DM seemed to have demonstrated that the extended-release preparation may be less efficacious in women with PCOS than the immediate-release metformin (107). Furthermore, this result was not confirmed in a recent study (108) in which the extended-release preparation had similar efficacy at glycemic control to the immediate-release form. In addition, in a multicenter RCT (109) on 150 Chinese patients with type 2 DM, no significant differences were found between the extended- and immediate-release metformin formulation. In particular, the extended-release metformin administered at a daily dosage of 1500 mg was similarly effective as the immediate-release metformin at a 500-mg dosage three times daily (109). A significant difference between the two formulations was observed only for postprandial glycemia at 120 min, in favor of the immediate-release formulation (109).
Almost all published studies including PCOS patients used metformin in immediate-release preparation; however, in a recent RCT (110), PCOS patients were treated with the extended-release formulation (1 g twice daily). In that trial (110), there were significant decreases in BMI and total testosterone levels and significant increases in SHBG levels with extended-release metformin, whereas no significant change in fasting markers of insulin resistance was observed.
A dose-response effect of immediate-release metformin in obese women with PCOS was suggested (111). In fact, fasting markers of insulin resistance did not change when 500 mg of this formulation was administered three times daily but improved when 850 mg of immediate-release metformin was given three times daily.
Thus, different metformin preparations might have different efficacy in PCOS.
In addition, most RCTs of ovulation induction in PCOS patients used 1500–1700 mg/d, which could be a suboptimal dosage, rather than maximal doses of immediate-release metformin (30, 112).
The length of metformin treatment in PCOS patients is not standardized. Although, several evidences (113) support the efficacy at the metabolic level of metformin treatment for a limited time in insulin-resistant PCOS women, it is still unknown whether metformin administration should be considered as a symptomatic treatment or as a curative and definitive therapy. Thus, it is not clear for how long metformin should be administrated and which are the effects it may have on insulin sensitivity at treatment suspension after its short- or long-term administration.
A recent study (114) on infertile PCOS patients undergoing in vitro fertilization (IVF) demonstrated that metformin pretreatment did not affect androgen levels. Furthermore, within 36 h after metformin withdraw, an increase in androstenedione and free-testosterone index was observed.
Our previous data (115) on a non-insulin-resistant PCOS population showed that, after a long-term metformin treatment, drug suspension is related to a quick reversion of its beneficial effect on peripheral insulin sensitivity. In fact, a slightly, but significant, worsening of the insulin resistance and hyperandrogenic state was observed in comparison with baseline and with patients who received placebo, even if no PCOS patient developed IGT or DM at 12 months from metformin suspension. In the same study (115), the quick loss of beneficial effects due to metformin on insulin resistance and subsequently on hyperandrogenism after metformin suspension seemed to be related to a worsening of menstrual cyclicity. In agreement with these findings, previous data (116) showed that the withdrawal of metformin treatment seemed to be followed within 3 months by a significant reversal toward a pretreatment hyperandrogenic hyperinsulinemic state in 10 nonobese adolescents with hirsutism, oligomenorrhea, dyslipidemia, and a history of precocious pubarche. However, because time could be considered a confounding factor on the clinical evolution of PCOS, further data on a wide sample, a long-time follow-up, having as its primary aim the evaluation of the effect of the time on ovarian function, and the clinical features in PCOS patients are necessary to draw definitive conclusions.
Finally, because recent data seem to suggest that metformin is more effective in insulin-resistant PCOS patients with low BMI (see Section I.F), metformin dose should probably be adjusted according to the patient’s BMI and insulin resistance (117). Unfortunately, no model is currently validated to calculate the right dose according to these characteristics. From a clinical point of view and based on our clinical experience, we suggest that it is appropriate and useful to administer metformin in a slow and increasing manner up to a maximal tolerated dosage (117).
E. Adverse effects
For several decades, metformin has been used worldwide to treat type 2 DM, and, during the last few years, several studies have been performed on women with PCOS using metformin. Thus, the safety profile of metformin is well known.
Potential side effects of metformin are listed in Table 1.
Rare side effects |
Severe/serious |
Lactic acidosis |
Liver failure |
Slight/moderate |
Chest pain |
Cough |
Mouth metallic taste |
Nail changes |
Rash |
Runny nose |
Sneezing |
Skin flushing |
Alopecia |
Common side effects |
Abdominal discomfort |
Bloating |
Constipation |
Diarrhea |
Flatulence |
Heartburn |
Indigestion |
Nausea |
Vomiting |
Rare side effects |
Severe/serious |
Lactic acidosis |
Liver failure |
Slight/moderate |
Chest pain |
Cough |
Mouth metallic taste |
Nail changes |
Rash |
Runny nose |
Sneezing |
Skin flushing |
Alopecia |
Common side effects |
Abdominal discomfort |
Bloating |
Constipation |
Diarrhea |
Flatulence |
Heartburn |
Indigestion |
Nausea |
Vomiting |
Rare side effects |
Severe/serious |
Lactic acidosis |
Liver failure |
Slight/moderate |
Chest pain |
Cough |
Mouth metallic taste |
Nail changes |
Rash |
Runny nose |
Sneezing |
Skin flushing |
Alopecia |
Common side effects |
Abdominal discomfort |
Bloating |
Constipation |
Diarrhea |
Flatulence |
Heartburn |
Indigestion |
Nausea |
Vomiting |
Rare side effects |
Severe/serious |
Lactic acidosis |
Liver failure |
Slight/moderate |
Chest pain |
Cough |
Mouth metallic taste |
Nail changes |
Rash |
Runny nose |
Sneezing |
Skin flushing |
Alopecia |
Common side effects |
Abdominal discomfort |
Bloating |
Constipation |
Diarrhea |
Flatulence |
Heartburn |
Indigestion |
Nausea |
Vomiting |
In a recent meta-analysis (118), the rate of discontinuation for adverse events in therapy-naive PCOS patients receiving metformin, clomiphene citrate (CC), or both for anovulatory infertility was analyzed. Metformin and CC showed a similar effect on the discontinuation rate for adverse events [odds ratio (OR), 0.71; 95% confidence interval (CI), 0.22 to 2.25; P = 0.765] with homogeneity data (118). In fact, the discontinuation rate for adverse events was similar between treatments in studies of both Palomba et al. (119) (2.0% for each treatment arm) and Legro et al. (110) (1.9 vs. 2.9% for CC and metformin groups, respectively), whereas no dropout for drug-related side effects was observed in the study by Zain et al. (120). On the other hand, when data obtained from studies comparing the combination of metformin plus CC vs. CC alone (110, 119, 120) were pooled, no significant effect of metformin on discontinuation rate for adverse events (OR, 0.23; 95% CI, 0.04 to 1.24; P = 0.096) was observed, even if a significant heterogeneity was detected (118). No difference in the discontinuation rate for adverse events was recorded between treatments in studies by Legro et al. (110) (3.3 vs. 1.9%, respectively) and by Zain et al. (120) (no patients in both groups), whereas the study by Moll et al. (121) reported a significantly higher adverse events rate in patients treated with combination therapy (16.2 vs. 5.3%, respectively). Finally, no significant difference between metformin plus CC and metformin alone was observed in the discontinuation rate for adverse events (OR, 0.86; 95% CI, 0.23 to 3.04; P = 0.993) (110). These data were homogeneous because no difference in the discontinuation rate for adverse events was noted between treatments in both studies analyzed [3.3 vs. 1.9% for the Legro et al. study (110), and no case for the Zain et al. study (120)].
1. Common side effects.
Metformin is generally a well-tolerated drug. A significantly increased incidence of nausea, vomiting, and gastrointestinal distress was reported in women with PCOS under metformin treatment (30, 110). In particular, gastrointestinal symptoms are the most frequent drug-related adverse events occurring in about 30% of patients taking metformin, limiting the compliance to treatment.
The rate of gastrointestinal side effects seems to be lower with the use of the extended-release formulation (122), but clear data are still unknown because the studies in which this kind of formulation is used are very few. In the meta-analysis by Lord et al. (30), four trials (123–126) reporting details about side effects of metformin were analyzed. In all studies, metformin at immediate-release formulation was used, whereas the treatment duration varied from 6 wk (125) to 12 wk or more (123, 124, 126). Metformin caused a significantly higher incidence of nausea or vomiting (OR, 3.84; 95% CI, 1.07 to 13.81; P = 0.05) and other gastrointestinal disturbances (OR, 4.40; 95% CI, 1.82 to 10.66; P = 0.003). Successively, RCTs reported a rate of adverse events ranging from 7.9% (120) to 22.2% (119) in patients treated with immediate-release metformin. On the other hand, a recent trial (110) showed a good safety profile of the extended-release formulation with an incidence of serious adverse events of 1.0%, whereas other adverse events ranged from 3.3% (anovulatory bleeding) to 66% (nausea).
In all cases of drug-related side effects, the dose should be reduced until symptoms disappear. Patients with intolerance to metformin could benefit from the extended-release formulation, albeit given in divided doses.
2. Rare side effects.
Serious adverse events are rare with metformin administration.
Metfomin toxicity is manifested at a concentration of 100 μg/ml or higher, but in vivo data showed plasma levels of metformin less than 5 μg/ml, even at maximum dosage. Moreover, cases of metformin overdose have also been reported in literature (127). Aggressive supportive care can resolve, without sequelae, potentially lethal metabolic alteration also in patients with serum metformin levels higher than 100-fold in the therapeutic range (127).
Lactic acidosis is a rare complication of metformin administration (5.1 cases per 100,000 patient-years) (128), even if, when it occurred, a mortality of 50% was detected. This risk increases in patients with hepatic or renal impairment, cardiac or respiratory insufficiency, severe infection, or alcoholism, conditions that are, in themselves, associated with hypoxia and lactic acidosis (129, 130).
Almost all data on lactic acidosis in patients who received metformin pertained to type 2 DM; nonetheless, even if no published reports on lactic acidosis with metformin therapy in women with PCOS are currently available, it can be assumed that this complication represented a very rare drug-related side effect in this specific population because the PCOS subjects are generally young and without other major medical conditions.
A recent subanalysis of the Fremantle Diabetes Study, a wide longitudinal observational study, confirmed that lactic acidosis incidence in type 2 DM patients is very low, but it increases in older patients with a long duration of DM, those with the highest prevalence of cardiovascular and renal comorbidities (131). Unfortunately, current risk determinants for metformin-associated lactic acidosis are largely disregarded for patients with type 2 DM (132). Thus, appropriate caution must always be taken in patient selection to avoid this rare but potentially fatal side effect. A meta-analysis of RCTs on metformin use in patients with type 2 DM confirms that this adverse event does not occur if precise precautions are followed and if the drug is not administered to individuals with compromised renal or hepatic function (128). Furthermore, severe lactic acidosis can also be associated with high doses of metformin, i.e., higher than 3000 mg/d, in patients with no known contraindications for metformin prescription (133).
Symptoms of metformin-associated lactic acidosis are unspecific, and physicians should be aware that metformin, if prescribed in patients with renal impairment, can cause fatal lactic acidosis due to drug accumulation (134). Symptoms and signs of lactic acidosis are abdominal pain and psychomotor agitation; the physical examination always reveals signs of poor perfusion. Laboratory evaluation shows hyperkalemia, elevated creatinine and blood urea nitrogen, and mild leukocytosis, whereas arterial blood gases show severe lactic acidemia. More rarely, renal impairment could be the only alteration present in patients with metformin-induced lactic acidosis (134). The management of this severe adverse event consists of intensive care by use of vasopressor and ventilatory support and continuous venovenous hemodiafiltration.
The available data regarding the deleterious effect of metformin in patients with kidney disease are extremely limited, even if it is reasonable that metformin could be used at full dosage in patients with a glomerular filtration rate (GFR) of 60–90 ml/min (135). The doses should be reduced according to GFR with particular regard for patients with a GFR from 30–60 mg/min, even if using the same caution as with any kidney-excreted drug (135). Lastly, metformin should not be administered to women if serum creatinine is above 1.4 mg/dl for women (135).
Because metformin may cause malabsorption of vitamin B12, patients on metformin should be monitored for signs and symptoms of vitamin B12 deficiency, including numbness, parasthesia, macroglossia, memory loss, and behavioral changes. Vitamin B12 deficiency can also lead to the development of pernicious anemia (106).
Metallic taste in the mouth is less frequently observed. Great attention should always be given to those subjects undergoing treatment with different medications (136) or contrast media use (137) due to the risk of acute kidney dysfunction and lactic acidosis. In this regard, acute renal failure from contrast medium in patients taking metformin was described (138).
A case of hepatitis after metformin administration was described as a likely rare idiosyncratic reaction with high serum transaminases and intrahepatic cholestasis (139). In literature, case reports describing very rare drug-related side effects, such as autoimmune neutropenia in a patient with lymphoma, are also available (140).
Osteoporosis could be another potential long-term adverse effect of ISD administration in patients with type 2 DM, a population notoriously known to have poor bone mineral density and/or quality. A subanalysis from the Diabetes Outcome Progression Trial (141) showed an increased fracture rate at 5 yr after rosiglitazone. Furthermore, metformin seemed to be safer than rosiglitazone, with a fracture rate of about half (7.3 vs. 15.1% for metformin and rosiglitazone, respectively). In fact, metformin seemed to have a direct osteogenic effect on osteoblasts in culture, which could be mediated by activation/redistribution of phosphorylated ERK and induction of endothelial and inducible nitric oxide synthase (NOS) (142). Contrarily, a case-control study (143) tested the effect of oral hypoglycemic treatments on bone fractures in a population of diabetic patients demonstrating no significant association between fractures and metformin administration.
Among very rare adverse events cited in literature, a case of acute alopecia due to metformin treatment in a PCOS patient was also reported (144).
3. Adverse effects on pregnancy evolution.
Several authors studied pregnancy outcomes in PCOS patients who received metformin for ovulation induction (145). To date, metformin is still found in the B classification for USFDA pregnancy category (31). This means that no teratogenic effect was demonstrated in animal models, and human safety studies are not adequate. In fact, metformin has been tested in rat and rabbit models at doses up to 600 mg/kg·d, about 2- to 3-fold the maximum recommended in humans according to body surface (31). On the other hand, a recent wide clinical trial confirms that metformin alone or with supplemental insulin is as safe as insulin alone in patients with gestational DM (146).
To date, no teratogenic effects or adverse fetal outcomes were reported from metformin when administered to pregnant women with type 2 DM or gestational DM (147, 148), even if most treatments started after pregnancy had begun.
In 2006, a meta-analysis on metformin safety did not find any evidence for adverse pregnancy outcome in women undergoing treatment with metformin (149). More recently, another meta-analysis of eight studies focusing on pregnancy outcome after metformin use in PCOS women confirmed that there is no evidence of an increased risk for major malformations (OR, 0.70; 95% CI, 0.11 to 4.39) (150). Notwithstanding these safety data, in clinical practice metformin is usually discontinued during pregnancy in women with PCOS who conceived while receiving the drug.
Currently, preliminary data on the use of metformin during pregnancy in patients with PCOS (151) seem to be reassuring, suggesting also some beneficial effects on pregnancy loss reduction and pregnancy outcome improvement without affecting infant parameters (see Section IV). Also, a recent semirandomized study (152) demonstrated the safety of the metformin administration during the first trimester of pregnancy in regard to fetal growth and perinatal outcome in PCOS patients. In particular, no difference in major congenital malformations and in rate of babies admitted to the neonatal intensive care was observed between PCOS patients who received metformin and those who did not (152).
F. Potential predictors
The beneficial effect of metformin might vary according to the clinical characteristics of the patient to be treated. This concern is even more important and complex in consideration of the different PCOS phenotypes (see Section I). The knowledge of predictors for metformin response is crucial because their identification could lead clinicians to the best individualized treatment to improve metformin’s efficacy in optimizing the safety profile (153).
1. Body weight.
Some studies demonstrated that metformin appears to benefit PCOS subjects irrespective of their weight or degree of insulin resistance (154, 155). A study including lean, overweight, and obese PCOS patients observed that all three groups of patients demonstrated a significant decrease in fasting insulin and homeostasis model of assessment (HOMA) index after 6 months of metformin treatment, irrespective of their pretreatment degree of insulin resistance (154). Additionally, the overweight and obese groups demonstrated a significant decrease in AUC for insulin (AUCinsulin) response to an oral glucose challenge. In normal-weight PCOS patients, metformin significantly improved insulin sensitivity and the AUC for glucose (AUCglucose) to AUCinsulin ratio in comparison with both the baseline assessment and against a placebo group (115). Even in studies performed to examine the effects of metformin in PCOS patients with normal weight and with normal insulin sensitivity, significant decreases in fasting insulin, AUCinsulin, and HOMA were observed (156).
Several other data have demonstrated that metformin improves insulin sensitivity and is effective in decreasing ovarian androgen production and serum androgen levels with ovarian function restoration regardless BMI (84, 85, 157).
On the contrary, a recent subanalysis (117) of a previously published RCT (112) comparing metformin vs. CC was performed considering both ovulation and pregnancy as response markers to the treatment. These data showed significant differences in BMI and insulin RIs between ovulatory and anovulatory patients and between pregnant and nonpregnant patients treated with 1700 mg/d metformin. In particular, lower BMI was more likely associated with an ovulatory response.
This finding is consistent with previous data found in literature (124, 158–163). Specifically, Maciel et al. (158) first demonstrated in a randomized controlled fashion that a 6-month course of 1500 mg/d metformin decreased serum androgens and fasting insulin levels and improved menstrual cyclicity significantly better in nonobese than in obese PCOS patients. Subsequent studies agreed that metformin induces more pronounced endocrine effects and improves ovulation (159, 160, 162), particularly in nonobese PCOS patients. Legro et al. (162) confirmed that metformin efficacy in terms of ovulation was significantly and directly associated with BMI. In particular, the adjusted ORs for ovulation rate comparing patients with BMI lower than 30 Kg/m2 and those with BMI ranging between 30 and 34 kg/m2vs. patients with BMI equal to or higher than 35 kg/m2 were 2.36 (95% CI, 1.65 to 3.36) and 2.05 (95% CI, 1.46 to 2.88), respectively (162).
To date, it is plausible to consider patients with high BMI as subjects with a possible suboptimal response to metformin. Metformin dosage might be crucial for a good response and for treatment success (111, 164). PCOS patients who did not respond at a specific daily dose of metformin could probably respond clinically if higher doses were to be used. In this regard, it would be useful to administer metformin in PCOS patients at the maximal tolerated dosage according to their BMI. Regarding this issue, two RCTs (111, 164) tackling the problem of metformin dose were recently published. The first study (111) showed that obese PCOS women, but not morbidly obese women, responded to metformin reducing insulin sensitivity and body weight in a dose- and BMI-related manner, even if there was no significant difference between either the dose or obesity subgroups in the menstrual response to treatment. The latter study (164) confirmed that greater doses of metformin are more effective in reducing BMI and waist circumference in an overweight and obese PCOS population. Furthermore, no data presently evaluated the optimal dosage of metformin according to patient BMI with ovulation and/or pregnancy as primary outcomes.
2. Insulin sensitivity.
Insulin sensitivity indexes differed between ovulatory and anovulatory patients treated with metformin (117). This finding may be explained by hypothesizing that insulin resistance is a common but not an imperative feature in PCOS and might play, independently from BMI, a pathogenetic role only in a subset of women (25). Only these insulin-resistant subjects would substantially benefit from improved insulin sensitivity induced by metformin.
This suggestive hypothesis is supported by the results of other studies (126, 160, 165), even if some authors (124, 154, 159, 166) failed to demonstrate a predictive value of insulin RIs for ovulation under metformin. In these studies (124, 154, 159, 166), patients were probably insulin resistant, but the methods used for assessing insulin sensitivity were not very thorough. In addition, a peripheral insulin resistance that did not reflect in abnormal serum insulin sensitivity indexes should be hypothesized in these PCOS patients. On the other hand, it is possible to hypothesize that some insulin-resistant PCOS patients, notwithstanding the improvement of insulin sensitivity induced by metformin, have a poor clinical response because other mechanisms of ovarian dysfunction, independent from insulin resistance, could be involved.
3. Hyperandrogenism.
Some authors (124, 126) observed that responders to metformin were less hyperandrogenic than nonresponders. In particular, in a randomized, double-blind controlled study, Fleming et al. (124) compared the baseline characteristics of the subgroups that responded to metformin with those patients who did not respond with established normal ovarian function. Responders to metformin treatment showed significantly lower testosterone and higher SHBG levels. Another randomized, double-blind, placebo-controlled study (126) showed significant differences between responders and nonresponders to metformin treatment. Logistic regression analysis demonstrated that lower serum androstenedione was an independent predictor for clinical efficacy of metformin because responders had significantly lower serum androstenedione. A lower free androgen index (FAI) (OR, 1.59; 95% CI, 1.17 to 2.18 for FAI < 10 vs. FAI ≥ 10) was detected as a predictive factor for ovulation under metformin in another study (162).
These data were not recently confirmed (117), probably because all PCOS patients were hyperandrogenic at study entry.
Chang et al. (167) showed that the insulin levels and the degree of β-cell function measured by HOMA index are highest in oligoovulatory women with PCOS who are both hirsute and hyperandrogenic, compared with PCOS patients who are either hyperandrogenemic or hirsute only. In this regard, a greater improvement in (biochemical) efficacy was observed among PCOS women who were both hyperandrogenic and hyperinsulinemic (161).
4. Waist-hip ratio (WHR).
Moll et al. (168), in a subanalysis of a previous RCT (121), showed a significantly different chance of ongoing pregnancy between PCOS patients receiving metformin who were subgrouped according to age and WHR. In particular, metformin was demonstrated to have a positive effect in women older than 28 yr and with high WHR, a negative effect in women younger than 28 yr regardless of their WHR, and no effect in women older than 28 yr and with low WHR (168).
5. Genetic factors.
Genetic factors may also modulate the effectiveness of metformin in inducing ovulation, resulting in a nonsufficient response on the part of some patients. Although the molecular reason for the variability in response to metformin is not clear, the possibility for a pharmacogenomic approach to the use of metformin may emerge in the near future (101).
As previously discussed, genetic polymorphisms in OCT 1 and OCT2 were found to be associated with changes in pharmacokinetic/pharmacodynamic responses to substrate drugs. In addition, data from the Pregnancy in PCOS (PPCOS) trial indicated that a polymorphism of a serine-threonine kinase gene expressed in the liver and site of metformin action, STK11 (previously cited as LKB1), was associated with a significantly decreased chance of ovulation in PCOS patients treated with metformin (162). In particular, the C allele of a single nucleotide polymorphism in the STK11 gene was associated with a significantly decreased chance of ovulation in PCOS women treated with metformin (162). In fact, the percentage of patients who ovulated increased with the number of G alleles present (48 vs. 67 vs. 79% for women without alleles, with one allele, or with two alleles, respectively). In addition, the adjusted ORs for ovulation rate comparing the C/C and G/G genotypes to the G/G genotype were both 0.30 (95% CI, 0.14 to 0.66) (162).
Lastly, due to the lack of population-based studies in literature, it is not known whether ethnic background may modify the effectiveness of metformin. In a study conducted on Chinese women (123), no improvement was noted after metformin in the ovulation rate despite the significant reduction of BMI, serum testosterone, and fasting leptin concentrations. Moreover, these data were not confirmed (169), demonstrating the efficacy of metformin in the improvement of menstrual cyclicity, fertility, hirsutism, and PCO morphology.
II. Menstrual Disorders
Menstrual dysfunctions affect a high number of PCOS patients, ranging between 80 and 100% depending on the diagnostic criteria; up to 40% of women with PCOS-related oligoovulation were referred with a history of eumenorrhea (170).
The main complaints about menstrual disorders from PCOS patients are absence or infrequency of menstrual bleeding. Furthermore, other subclinical menstrual changes are frequent in PCOS patients, such as intermittent anovulation and related dysfunctional uterine bleeding (171–173).
Several pathogenetic mechanisms were proposed (174–176), although hyperinsulinemia could play a central role (see Section I. B) (25).
Based on these considerations, Velazquez et al. (32) first reported resumption of menstrual cyclicity in 21 of 22 amenorrheic PCOS women. Since then, several uncontrolled studies (45, 177, 179) and RCTs (180, 181) evaluated the effects of metformin on the restoration of regular menses.
Findings obtained from uncontrolled studies (45, 177–179) demonstrated that metformin was effective in restoring regular menses in approximately 62% of PCOS women with oligoamenorrhea. In addition, a significant improvement in the menses frequency under metformin was reported in two RCTs (180, 181), although in one study (181) the number or percentage of women reverting to regular menstrual cycles was not reported, whereas, in the other (180), insufficient data for this calculation were provided.
More recently, in a meta-analysis (30) on ISDs in PCOS patients, only one trial was reported to improve the menstrual pattern showing a significant effect for metformin (OR, 12.88; 95% CI, 1.85 to 89.61; P = 0.01). In particular, in a randomized, double-blind, placebo-controlled fashion, the effects on menstrual abnormalities of a 6-month course of metformin in a group of 23 PCOS patients with normal glucose tolerance (NGT) were evaluated (126). After treatment, subjects treated with metformin showed a significant improvement in menses frequency, whereas no change was observed in those receiving placebo. Five women in the metformin group vs. none in the placebo group had a substantial improvement in their menstrual pattern during treatment. In addition, in subjects experiencing regular menses after metformin, the improvement was observed within 3 months from the beginning of treatment. In an open, long-term follow-up of the same study (126), the frequency of menses per month per patient continued to improve significantly over time during the metformin administration period. Specifically, after a follow-up of about 1 yr [11.0 ± 1.3 months (mean ± sd); median, 8 months; range, 4–26 months) in 13 patients, a complete regularization of the menses was reported, whereas four other patients experienced a striking amelioration of their menstrual cycle abnormalities.
The beneficial effect of metformin on menstrual cycle is commonly attributed to its effectiveness on ovulatory function. Furthermore, it is not infrequent to observe discordance between menstrual and ovulatory cycles. For example, Moghetti et al. (126) assessed ovulatory function measuring serum progesterone in the luteal phase of 39 cycles in 10 PCOS women experiencing regular menses after treatment with metformin, which resulted in the confirmation of only 32 of these assessments (79%). These observations may indicate that the effectiveness of metformin on menstrual cyclicity is probably secondary not only to an indirect effect on the ovary but also to a direct effect on the endometrium (see Section V).
Finally, a meta-analysis (182) showed that metformin was significantly less effective in improving the menstrual pattern in comparison with oral contraceptives (OCs) (OR, 0.08; 95% CI, 0.01 to 0.45; P = 0.004). Furthermore, metformin and OCs were compared in only two trials by the same group of researchers (183, 184). The first study (183) showed a restoration of menstrual cyclicity in 75 and 100% of obese women with PCOS treated with metformin or OCs, respectively. In a successive study (184), the same authors confirmed the superiority in improving menstrual pattern of OCs in comparison with metformin in nonobese women with PCOS (50 vs. 100%, respectively).
Unfortunately, no trial assessed whether this more favorable OC effect leads to a reduction in the long-term risk of endometrial cancer compared with metformin and/or an improvement of the quality of life (QoL). On the other hand, controversial data (185–187) are available on the metabolic and cardiovascular risks in OCs users. In fact, clinical data suggested that OCs (188, 189) reduce insulin sensitivity in women with PCOS. These observations, combined with the known actions of OCs to reduce insulin sensitivity and increase triglyceride levels in unselected patients, have resulted in the concern that these drugs could have adverse metabolic effects in women with PCOS (185).
Presently, no prospective long-term studies of OC use in women with PCOS are available. A small observational study (189) did not find adverse metabolic outcomes in women with PCOS using OCs and followed for 6–18 yr.
In a recent meta-analysis, the metabolic effects of metformin were compared with OCs (190). Metformin was more effective at lowering insulin and triglyceride levels than OCs, whereas there was insufficient evidence to assess clinical end-points such as hirsutism, DM prevention, and weight loss. Unfortunately, the studies carried out in this meta-analysis (190) were obtained from small sample groups at short term, and few estroprogestin combinations were tested [frequently ethinyl estradiol (EE2) plus cyproterone acetate (CA) association].
Two further studies (186, 191) comparing metabolic findings in PCOS patients receiving either metformin or OCs have been published. Metformin improved fasting parameters of insulin sensitivity, although EE2 plus CA did not worsen these end-points (187). OCs containing 35 μg EE2 plus CA increased the AUCinsulin and arterial stiffness, compared with metformin and 20 μg EE2 plus levonorgestrel with 100 mg daily spironolactone (191). On the other hand, serum uric acid levels, a cardiovascular risk factor and marker, were significantly reduced in PCOS patients taking OCs (35 mg EE2 plus 2 mg CA) with a potential action on hyperandrogenic milieu, whereas no significant change was observed after 1700 mg metformin (192).
In conclusion, as well elucidated recently in a vignette by Nestler (106), metformin should be considered the first-choice drug for treating oligomenorrhea in PCOS patients in whom OCs are contraindicated.
III. Anovulation
Anovulation or oligoovulation is the cause of infertility in about 25% of cases (193). In turn, PCOS is the most common cause of anovulatory infertility (194). Thus, the treatment of infertile PCOS patients presently represents a social and economic burden.
Considering that insulin exerts several effects in regulating normal ovarian activity in all its compartments and insulin resistance seems to play a pivotal role in the pathogenesis of infertility in obese and nonobese women affected by PCOS, metformin administration can be assumed to have a beneficial effect. Metformin could act directly on the ovary. Moreover, very few experimental and clinical data are actually available. An interesting case report has shown the efficacy of metformin in an underweight young PCOS patient with oligomenorrhea and hirsutism (195). This case suggests that metformin could act exclusively at the ovarian level, probably by insulin sensitization in a way completely unrelated to weight loss or lifestyle modification (195). On the other hand, experimental data seemed to suggest further mechanisms. In fact, as already discussed, in rat model, metformin has been shown to inhibit GnRH release by activation of hypothalamic AMPK, a crucial regulator of food intake in mammals, in a dose-dependent and time-dependent fashion (54).
To clarify the exact role of metformin in the therapeutic strategy of ovulation induction, it is pivotal to distinguish several conditions in which PCOS patients could take advantage of ISDs.
The first issue regards the treatment of therapy-naive PCOS patients. CC was the first agent used in experiments for ovulation induction in oligomenorrheic women. It represented for many years the first therapeutic option for anovulatory infertility treatment for its effectiveness, low cost, limited dose-dependent side effects, and simplicity in administration and management (196, 197). Therefore, the efficacy of metformin compared with placebo, or no treatment, should be demonstrated first. After this, metformin could be considered as a valid first-step option for ovulatory infertility in PCOS, once efficacy and safety have been proven to be superior to CC. Moreover, metformin could have an additive effect when combined with CC as the first-step approach to PCOS patients.
The second issue is whether PCOS patients who remain anovulatory after CC (CC resistance) or even those who ovulate but do not achieve a pregnancy (CC failure) could benefit by metformin, administrated as a single agent, in combination with CC, or as pretreatment. In fact, metformin mechanisms are still not completely clear, and whether it could have beneficial effects on the ovary and endometrium, improving the response to CC, is yet to be understood.
Lastly, some PCOS patients treated with CC and metformin alone or in combination and not affected by concomitant infertility (tubal or male) factors do not achieve pregnancy and can only turn to IVF as a possible solution; for these patients, the administration of gonadotropins is the last therapeutic step (198). Unfortunately, it is well documented that the response of PCO to gonadotropin stimulation differs significantly from that of normal ovaries. This response was defined as “explosive” and is responsible for the significantly higher risks of cancelled cycles and/or for ovarian hyperstimulation syndrome (OHSS) (199). Clinical (200) and experimental (201) studies suggested that hyperinsulinemia and insulin resistance could be responsible for this abnormal response. Thus, PCOS patients receiving gonadotropins for controlled ovarian stimulation or IVF could benefit from the administration of ISDs.
For the above-mentioned reasons, we have separately analyzed the use of metformin as a first- and second-line drug in anovulatory infertile patients with PCOS and in PCOS patients scheduled for gonadotropin therapy.
A. Metformin as first-step treatment
Several systematic reviews and a meta-analyses (202–207) have evaluated the efficacy of metformin in the treatment of anovulation due to PCOS. The unanimous conclusion of these reviews (202–207) was that metformin monotherapy represents a safe and valid therapeutic option for the improvement of ovulation in PCOS patients.
Two meta-analyses (203, 204) demonstrated that metformin is an effective drug, when compared with placebo or no treatment, in the restoration of normal menstrual cycles and in inducing ovulatory cycles in oligoamenorrheic PCOS patients [OR, 3.88; 95% CI, 2.25 to 6.69; P < 0.00001 (203), and OR, 1.50; 95% CI, 1.13 to 1.99; P = 0.004 (204)]. However, in terms of pregnancies [OR, 2.76; 95% CI, 0.85 to 8.98; P = 0.09 (203), and OR, 1.07; 95% CI, 0.20 to 5.74; P = 0.9 (204)] and live births [OR, 1.00; 95% CI, 0.13 to 7.79; P = 1.00 (204)], no advantages resulted from metformin administration in comparison with placebo or no treatment. These data were successively confirmed showing an OR for a clinical pregnancy rate of 3.3 (95% CI, 0.92 to 11; P = 0.07), whereas no data were provided regarding ovulation and live-birth rates (206).
A recent meta-analysis of 17 relevant placebo-controlled RCTs for a total sample of 1639 PCOS patients (207) was aimed at updating the state of evidence on the efficacy of metformin, used alone or in combination with CC, on ovulation, early pregnancy, and live births in PCOS patients with unknown resistance to CC. Even if the P values were not reported for any meta-analytic comparison, the pooled statistical estimate for ovulation showed an overall benefit for PCOS patients receiving metformin over placebo (OR, 2.94; 95% CI, 1.43 to 6.02) with a number needed to treat using metformin of 4.0 (207). On the other hand, metformin did not have any effect on pregnancy achievement (OR, 1.56; 95% CI, 0.74 to 3.33) or live births (OR, 0.44; 95% CI, 0.03 to 5.88) (208). For both end-points, no significant heterogeneity was detected between studies. The reason metformin was not effective in improving fertility could be related to the short-term follow-ups and/or small population samples studied.
To define the role of metformin as first-line drug for anovulatory infertility in PCOS patients, CC, the traditional gold standard ovulation inductor agent for PCOS-related anovulation, was also compared with metformin.
A meta-analysis (206) aimed at evaluating the efficacy of metformin, CC, or both drugs in therapy-naive PCOS patients demonstrated no significant clinical benefit from metformin administration over CC as first-line therapy for ovulation induction in PCOS patients (OR, 0.88; 95% CI, 0.19 to 4.1; and OR, 0.96; 95% CI, 0.11 to 8.2, for clinical pregnancy and live-birth rates, respectively). However, the results of this last meta-analysis (206) were biased by two main factors (208): first, some studies not assessing the live-birth rate, which should have been the primary end-point, and others in which a metformin pretreatment was administrated, were included in the final analysis; and second, the conclusions regarding the comparison of metformin vs. CC were drawn using data analyzed with fixed model effects, although a significant data heterogeneity was detected in both pregnancy and live-birth rates.
More recently, a systematic review and meta-analysis of the head-to-head RCTs available in literature was conducted to better define the efficacy of CC and metformin, alone or in combination, as a first-step approach in treating anovulatory infertility in PCOS patients (118). In this regard, a total of four RCTs (110, 119–121) satisfied the exclusion and inclusion criteria and were analyzed (118). Furthermore, only one study (110) was powered to demonstrate a difference in live births, whereas the others were powered on pregnancy (119) and ovulation (120, 121) rates.
In the study by Palomba et al. (119), the cumulative ovulation rate was similar in women treated with CC or metformin (62.0 vs. 84.0%, respectively), whereas the pregnancy rate was significantly higher (32.0 vs. 62.0%, respectively) in women treated with metformin compared with those treated with CC. There was no statistical difference in live births (18.0 vs. 5.2%, respectively) in women treated with CC vs. metformin, although a trend favoring the metformin group was present. The subsequent study by Legro et al. (110) reported markedly different results. In this clinical trial, CC was superior to metformin in increasing cumulative ovulation (75.1 vs. 55.3%, respectively), pregnancy (29.7 vs. 12.0%, respectively), and live-birth (22.5 vs. 7.2%, respectively) rates. Lastly, in the study by Zain et al. (120), the cumulative ovulation rate was significantly higher in CC than metformin (59.0 vs. 23.7%, respectively), whereas no differences between groups were reported in pregnancies (15.4 vs. 7.9%, respectively) and live births (15.4 vs. 7.9%, respectively).
Pooling the data obtained from the studies comparing metformin to CC (110, 119, 120), metformin did not exert significant advantage over CC with respect to cumulative ovulation (OR, 1.55; 95% CI, 0.40 to 5.99; P = 0.527), pregnancy (OR, 1.22; 95% CI, 0.23 to 6.55; P = 0.815) or live-birth (OR, 1.17; 95% CI, 0.16 to 8.61; P = 0.881) rates. However, the most interesting findings were the significant heterogeneity detected for all reproductive end-points. Several potential explanations were given to justify the disagreement regarding the comparison between metformin and CC (209). However, the main concern is probably the great heterogeneity in the protocol used and in the populations studied.
Regarding treatment schedules and formulations, in the study by Palomba et al. (119), both treatments were administered at fixed dosages to simplify the double-blind design. Furthermore, a subsequent study (112) by the same authors demonstrated a higher safety and effectiveness of CC treatment at incremental doses in a similar population; thus, results could be biased in favor of metformin (119). On the other hand, in three RCTs (119–121) an immediate-release formulation of metformin was used, whereas the PPCOS Trial (110) used, for the first time in PCOS patients, the extended-release formulation of metformin.
The diagnosis of PCOS-related infertility was controversial between studies. Different criteria were adopted to assess eventual tubal and sperm patencies considered as potential subfertility factors. Fertility and subfertility tubal and male factors formally excluded by appropriate testing in one study (119) were not evaluated in others (120, 121), whereas Legro et al. (110) excluded only patients with bilateral tubal abnormalities and/or patients having a partner with a sperm concentration less than 20 million per milliliter, regardless of motility or morphology. On the other hand, Moll et al. (121) used the 2003 ESHRE/ASRM criteria (2) to diagnose PCOS, even if the enrollment started before the validation and publication of the criteria, suggesting that investigators performed a retrospective analysis on data prospectively collected.
The studied populations were very heterogeneous regarding the BMI distribution. In particular, Palomba et al. (119) excluded obese patients. It is well known that obesity adversely affects reproduction in PCOS patients (210) and metformin may be more effective in nonobese than obese subjects (158); thus, the selection of a nonobese population may have contributed to the metformin efficacy in this study. On the other hand, no limitation for BMI was defined in the other studies (110, 120, 121) and the sd values of the BMI varied widely, suggesting highly heterogeneous populations including a large proportion of obese and/or severely obese patients. Furthermore, no information was given regarding eventual cotreatments frequently used by obese patients and potentially influencing reproductive outcomes, such as ISDs, cholesterol-lowering drugs, dietary treatments, physical exercise programs, or bariatric surgery (211–213).
In addition, although the RCTs were designed to evaluate the best first-step treatment in inducing ovulation in anovulatory PCOS patients, more than 50% of the population studied by Legro et al. (110) was previously treated with ovulation inductors. This is critical because each specific treatment might influence response to the subsequent one as well known for CC, metformin, and OCs (214), and previous treatment potentially results in the selection of “resistant” patients to the treatment or patients who ovulated under treatment but did not conceive (treatment failure).
Whether the addition of metformin to CC could improve the efficacy of CC alone was analyzed in a meta-analysis by Moll et al. (206). A significantly higher clinical pregnancy rate (OR, 1.9; 95% CI, 1.2 to 3.3) was observed in patients treated with metformin plus CC compared with those treated with CC alone, although a significant heterogeneity in treatment effect across the trials included in the meta-analysis was reported (206). Seven RCTs comparing the combination CC and metformin with CC in 985 infertile women with PCOS were included in the final analysis (206). Only three of these RCTs reported data on live births. Furthermore, the meta-analysis of these last studies reported no significant benefit on the live-birth rate of the combined therapy (OR, 1.0; 95% CI, 0.82 to 1.3; P = 0.74) and no significant heterogeneity in treatment effect (206).
By using a more strict criteria and updating of data found in literature, three head-to-head RCTs (110, 120, 121) comparing reproductive efficacy of metformin plus CC combination vs. CC monotherapy in therapy-naive PCOS patients were available. Metformin plus CC was no more effective than CC alone in inducing ovulation in the study by Moll et al. (121) (64.0 vs. 71.9%, respectively) and by Zain et al. (120) (68.4 vs. 59.0%, respectively), whereas in the study by Legro et al. (110) the combination was significantly more effective than CC monotherapy (83.3 vs. 75.1%, respectively). The pregnancy (39.6 vs. 45.6%, 38.3 vs. 29.7%, and 21.1 vs. 15.4%) and live-birth (18.9 vs. 26.3%, 26.8 vs. 22.5%, and 18.4 vs. 15.4%) rates were similar between metformin plus CC combination and CC monotherapy in the studies by Moll et al. (121), Legro et al. (110), and Zain et al. (120), respectively. When the data from these three studies (110, 120, 121) were combined in a meta-analysis (118), the metformin plus CC association was shown to be no more effective than CC regarding rates of ovulation (OR, 0.84; 95% CI, 0.60 to 1.18; P = 0.360), pregnancy (OR, 0.85; 95% CI, 0.62 to 1.15; P = 0.326), or live births (OR, 0.99; 95% CI, 0.70 to 1.40; P = 0.982). Of note, no significant heterogeneity was detected for all three parameters.
Presently, only two RCTs (118, 120) compared metformin plus CC with metformin alone. Both studies agreed in showing a clear advantage of combination therapy over metformin alone in terms of ovulation rate (83.3 vs. 55.3% and 68.4 vs. 23.7%, respectively). Conversely, Legro et al. (110) alone showed that the combination therapy was significantly more effective than only metformin with regard to pregnancies (38.3 vs. 12.0%, respectively) and live births (26.8 vs. 7.2%, respectively), whereas no differences in pregnancies (21.1 vs. 7.9%, respectively) and live births (18.4 vs. 7.9%, respectively) were observed in the study by Zain et al. (120). Pooling data from these last two studies (110, 120), metformin plus CC was more effective than metformin alone in inducing ovulation (OR, 0.23; 95% CI, 0.15 to 0.34; P < 0.0001), and no significant heterogeneity was detected (118). Similar results were obtained regarding pregnancy (OR, 0.23; 95% CI, 0.14 to 0.37; P < 0.0001) and live-birth (OR, 0.23; 95% CI, 0.13 to 0.40; P < 0.0001) rates (118).
A main drawback in the design of the RCTs that assessed the efficacy of combined CC plus metformin vs. CC (110, 120, 121) or metformin (110, 120) alone was that metformin was started either concurrently with (110, 120) or after only 1 month (121) of CC initiation. Metformin has a slower onset of action than CC; hence, the studies as such were unintentionally biased against demonstrating the value of the CC plus metformin combination.
Of note, multiple pregnancies occurred more frequently under CC treatment (110), even if this result does not reach a statistical significance. In particular, 6.0% of patients treated with CC and 3.1% receiving the combination therapy had multiple births (110). On the other hand, only singleton births were observed under metformin alone (110, 120, 119). Assuming that the goal of infertility treatment is to achieve a singleton gestation, it may be argued that CC is not quite as successful as suggested by the data of the PPCOS trial (215). At the present time, only one case of multifetal pregnancy under metformin has been published (216). These data suggest that metformin may facilitate high-order pregnancies in predisposed patients who were previously anovulatory.
Finally, a recent subanalysis (217) of the PPCOS trial (110) found that the distribution of the first ovulations occurring with each CC dose was not different between patients under metformin and those under metformin plus CC. Contrary to previous studies (218, 219), this finding suggests that metformin does not benefit by the decrease in the CC dose needed to induce ovulation in anovulatory PCOS women. As stated by the same authors (217), the contrasting results found in the PPCOS study population could be due to insufficient power, intrinsic differences in the subject populations, use of extended-release formulation of metformin, and lack of a full effect of metformin due to an almost simultaneous initiation of the two drugs. Certainly, the adherence with metformin tables in the PPCOS trial was not the reason for the poor success rate observed in the metformin arm because it was within acceptable limits and unrelated to ovulation (220).
In conclusion, in anovulatory infertile therapy-naive PCOS patients, the combined approach of metformin plus CC is not better than CC or metformin monotherapy. On the other hand, the choice between CC and metformin as first-step treatment should be drawn considering also contingent circumstances because of the lack of clear evidence.
B. Metformin as second-step treatment
Several studies were led to evaluate whether metformin can be effectively used as a second-step approach in the treatment of anovulatory infertility in PCOS patients. In particular, metformin was used as single agent (treatment), combined agent (cotreatment), and/or before other treatments (pretreatment).
1. Metformin as single agent.
Only a few studies (123, 221, 222) addressed the potential role of metformin as a single agent in CC-resistant patients. In one RCT (123), 20 infertile CC-resistant patients with ultrasonographic evidence of PCO were randomized to metformin or placebo. In this study, no benefit of metformin over placebo was observed in ovulation, pregnancy, and live-birth rates. In particular, the risk for both clinical pregnancies and live births in patients treated with metformin was calculated to be 0.50 (95% CI, 0.05 to 4.6; P = 0.54).
Palomba et al. (221) compared metformin as a single treatment with laparoscopic ovarian diathermy (LOD) in 120 CC-resistant PCOS patients. No difference was present between metformin and LOD in the ovulation rate (54.8 vs. 53.2%, respectively), but significantly higher pregnancy (21.8 vs. 13.4%, respectively) and live-birth (86.0 vs. 64.5%) rates were observed. A successive meta-analysis (206) showed no evidence of difference in the clinical pregnancy rate (OR, 1.3; 95% CI, 0.96 to 1.7; P = 0.09), whereas the live-birth rate was still higher after metformin (OR, 1.6; 95% CI, 1.1 to 2.5; P = 0.02). However, metformin treatment was certainly not inferior to LOD and was about 20-fold less expensive than the surgical ovulation induction (221).
2. Metformin combined with other treatments.
The potential effects of metformin were assessed in PCOS patients receiving other treatments. In particular, a great amount of data regards the association of metformin plus CC, although several findings seem to show a role of metformin cotreatment also in patients that received aromatase inhibitors, other kinds of ISDs, or surgical ovulation induction.
The efficacy of metformin cotreatment in CC-resistant patients who received CC was evaluated in several studies (203–207). Two meta-analyses (203, 204) agreed in demonstrating a significant benefit of metformin cotreatment in comparison with CC alone, even if a significant heterogeneity was observed between studies included in both meta-analyses. In particular, in CC-resistant PCOS women, the addition of metformin was demonstrated to be effective in inducing ovulation in the two meta-analyses by Lord et al. (203) (OR, 4.41; 95% CI, 2.37 to 8.22; P < 0.0001) and Kashyap et al. (204) (OR, 3.04; 95% CI, 1.77 to 5.24; P = 0.0005). In the same manner, significant benefits of the combination therapy on pregnancy rate were demonstrated in both studies [OR, 4.40; 95% CI, 1.96 to 9.85; P = 0.0003 (203); and OR, 3.65; 95% CI, 1.11 to 11.99; P = 0.03 (204)].
A successive meta-analysis (205), designed to assess metformin coadministration as a second-step approach for CC-resistant PCOS patients, confirmed that metformin plus CC is significantly more effective than CC alone in terms of ovulations (OR, 6.82; 95% CI, 3.59 to 12.96; P < 0.00001), even if a significant heterogeneity was demonstrated across studies, and no data were provided regarding the effect on pregnancy and live-birth rates.
Moll et al. (206) also compared metformin plus CC to CC alone and confirmed that metformin plus CC led to significantly higher clinical pregnancy (OR, 5.6; 95% CI, 2.3 to 13; P < 0.0001) and live-birth rates (OR, 6.4; 95% CI, 1.2 to 34; P = 0.03) without significant heterogeneity in treatment effect across trials.
Recently, in the analysis of placebo-controlled RCTs only, Creanga et al. (207) showed a significant benefit of metformin over placebo in PCOS patients receiving CC in terms of ovulation (OR, 4.39; 95% CI, 1.94 to 9.96) and pregnancy achievement (OR, 2.67; 95% CI, 1.45 to 4.94) with a number needed to treat of 3.7 and 4.6, respectively. Furthermore, a significant heterogeneity was detected between all studies included in the analysis. Conversely, metformin did not have any effect on live births (OR, 1.74; 95% CI, 0.79 to 3.86), and no significant heterogeneity data were present. Of note, in subanalyzing studies according to CC resistance, obesity, and duration of treatment, metformin was found to be more beneficial than placebo among PCOS patients treated for short periods and not CC resistant, whereas the benefits of metformin plus CC over CC were significantly higher in CC-resistant and obese PCOS patients (207).
The length of metformin administration is also a major concern. In fact, in meta-analytic studies (202–207), the beneficial effect of metformin could be diluted in time with longer follow-up periods due to more time available for spontaneous ovulation in control group patients. On the other hand, short-course metformin (less than 4 wk) could be a suboptimal pretreatment period before beginning CC. Unfortunately, there presently is no RCT assessing this aim, and there is insufficient data to determine whether long-course metformin pretreatment, before initiation of CC for ovulatory infertility treatment, is more effective than short-course pretreatment (224).
The combination of metformin plus CC was demonstrated to be more effective than LOD in CC-resistant PCOS patients with anovulatory infertility (225). In particular, metformin plus CC association was related to higher ovulation rates than LOD, even if no difference in the rates of pregnancies, live births and miscarriages were detected between two procedures (225).
The association of metformin plus letrozole, an aromatase inhibitor widely studied as an ovulation inductor for PCOS patients, vs. metformin plus CC in CC-resistant PCOS patients has been tested by Sohrabvand et al. (226) in a randomized controlled fashion. Serum estradiol (E2) levels and E2 levels per mature follicle were significantly higher in CC patients without differences in mature follicles, ovulation, and pregnancy rates (226). However, endometrial thickness and full-term pregnancies were significantly higher in patients treated with metformin plus letrozole (226).
Metformin was also administrated in combination with rosiglitazone, an ISD afferent to the thiazolidinediones family, assuming that the combination of these two medications could have a greater effect on ovulation by modulating insulin sensitivity and insulin levels via different mechanisms. In a prospective study, 128 nonobese PCOS patients with normal fasting and glucose-stimulated insulin levels were randomized to receive metformin, rosiglitazone, or a metformin plus rosiglitazone combination for 6 months (159). No significant benefit of the rosiglitazone addition was detected, suggesting that the use of metformin plus rosiglitazone association is not presently supported by clinical evidence. Further studies are needed, given the selection biases of the enrolled population.
Finally, in one trial (222) 42 CC-resistant PCOS patients were randomized to LOD followed by metformin or LOD alone. Ovulation (86.1 vs. 44.6%) and pregnancy (47.6 vs. 19.1%) rates were significantly higher in patients who received metformin. Furthermore, a successive meta-analysis (206) demonstrated no significant benefit in clinical pregnancy rate (OR, 2.3; 95% CI, 0.82 to 6.2; P = 0.12) or live-birth rate (OR, 1.3; 95% CI, 0.39 to 4.0; P = 0.71) for the metformin administration after LOD.
3. Metformin pretreatment.
Several RCTs (113, 157, 219, 227–229) evaluated the efficacy of metformin pretreatment before CC in CC-resistant PCOS patients. Even if these studies were very heterogeneous and were performed on small populations, most of them (113, 157, 219, 227) seemed to suggest that metformin pretreatment improves the efficacy of CC in PCOS patients with CC resistance. In particular, one small RCT (228) showed no beneficial effect on ovulation and pregnancy rates after 3 months of metformin pretreatment, despite the improvement of insulin resistance and hyperandrogenemia. On the contrary, other studies (113, 157, 219, 227, 229) demonstrated that metformin pretreatment, also when given in ultrashort protocols (227, 229), improved CC response in terms of ovulation and pregnancy in CC-resistant PCOS patients. These findings support the idea that decreasing insulin secretion while administering metformin in PCOS patients facilitates the induction of ovulation by using CC (13).
C. Metformin in patients who receive gonadotropins
Metformin was also used in PCOS patients who received gonadotropins for inducing monoovulatory cycles or multiple follicular development in IVF cycles. The effects of metformin coadministration in patients who received gonadotropins was the object of evaluation by two meta-analyses (206, 230).
1. Monoovulatory cycles.
The first meta-analysis was published in 2006 by Costello et al. (230). Three RCTs evaluating the effect of metformin in patients who received gonadotropins for controlled ovarian stimulation were identified (230). Data obtained from the only available study showed that metformin does not improve ovulation rate during ovarian stimulation with gonadotropins (90 vs. 73.3%; OR, 3.27; 95% CI, 0.31 to 34.72; P = 0.33) (230). No significant metformin effect was observed regarding the improvement of pregnancy rate during ovarian stimulation with gonadotropins (28 vs. 10%; OR, 3.46; 95% CI, 0.98 to 12.2; P = 0.05), whereas no RCT reporting live births as an outcome measure was identified. On the other hand, metformin was shown to be significantly efficient in reducing the length of time for ovarian stimulation while using gonadotropin [weighted mean duration (WMD) = −4.14 d; 95% CI, −6.36 to −1.93; P = 0.0002] and the total dose of FSH used (WMD = −425.05 IU; 95% CI, −507.08 to −343.03; P < 0.00001), even if a significant heterogeneity was found between pooled studies. Finally, no RCT reporting OHSS as an outcome measure was identified (230).
In updated literature, two RCTs (231, 232) evaluating whether metformin changes ovarian responsiveness in controlled ovarian stimulation cycles were recently published. The first RCT (231) on 70 nonobese insulin-resistant PCOS patients who received a low-dose step-up gonadotropin stimulation protocol followed by timed intercourse or intrauterine insemination demonstrated a significant effect of metformin in increasing the rate of monoovulatory cycles and in reducing those of cancelled cycles. Furthermore, no effect of metformin pretreatment and coadministration was confirmed in ovulation, cycle cancellation, pregnancy, abortion, live births, multiple pregnancies, or OHSS. The latter study (232) showed that metformin improved the endocrine profile in insulin-resistant PCOS patients receiving gonadotropins in a step-up protocol and confirmed that it facilitated the monofollicular development during ovarian stimulation cycles. However, these findings also were included in the Costello meta-analysis (230) as abstract.
Although the exact mechanisms by which metformin exerts its beneficial action remains unknown, the results of previous studies allow us to hypothesize that metformin acts on the regulation of ovarian response to exogenous gonadotropins improving insulin resistance. In fact, a reduction in serum testosterone and insulin levels in follicular fluid was observed after metformin treatment (233). Thus, the improvement of the hyperinsulinemic and hyperandrogenic ovarian environment might be crucial for a normal folliculogenesis, homogeneous development, and responsiveness of follicles and atresia of the small cohort of follicles.
A more recent meta-analysis (206) on four RCTs demonstrated a pooled clinical pregnancy rate that was significantly higher when metformin was added to gonadotropins than with gonadotropins alone (OR, 1.7; 95% CI, 1.1 to 2.8; P = 0.03). Furthermore, no significant difference in live-birth rate could be proven (OR, 1.6; 95% CI, 1.0 to 2.9; P = 0.08). No heterogeneity in treatment effect across trials was reported for either pregnancy or live-birth rate. In addition, metformin was demonstrated to be effective in reducing multiple pregnancies (OR, 0.26; 95% CI, 0.07 to 0.96), whereas no difference in OHSS (OR, 0.59; 95% CI, 0.17 to 2.1) was reported.
The lack of an effect of metformin on this end-point was probably due to the low incidence of OHSS during ovarian stimulation with low-dose step-up gonadotropin protocols (234). On the other hand, the employment of a step-up low-dose protocol tailoring the gonadotropin administration seems to be a pivotal factor more and more in PCOS patients who receive metformin because metformin coadministration has been shown to increase the incidence of multiple pregnancies in patients who receive gonadotropins (235).
In conclusion, in patients who received gonadotropins as treatment for anovulation, metformin addition reduces the duration of gonadotropins administration and the doses of gonadotropins required, and increases the rate of monoovulations, reducing the risk of cancelled cycles.
2. Multiple-ovulatory cycles for IVF procedures.
Data obtained from two meta-analyses (206, 230) seemed to suggest that metformin did not improve the efficacy of gonadotropins when it was prescribed in the context of adjuvant treatment for IVF cycles. In particular, Costello et al. (230) analyzed four published RCTs and one trial published in abstract format only evaluating the effect of metformin addition to gonadotropins in women undergoing IVF cycles. Metformin was demonstrated to have no significant effect on the number of oocytes collected at IVF (WMD = 0.44; 95% CI, −0.98 to 1.86; P = 0.54), length of ovarian stimulation (WMD = −0.09 d; 95% CI, −0.49 to 0.31; P = 0.66), pregnancy (34 vs. 29%; OR, 1.29; 95% CI, 0.84 to 1.98; P = 0.25) and live-birth (36 vs. 22%; OR, 2.02; 95% CI, 0.98 to 4.14; P = 0.06) rates. No significant heterogeneity between studies for these last outcomes was reported. More recently, combining data from four RCTs, Moll et al. (206) confirmed previous results. Specifically, no significant differences in pregnancy (OR, 1.2; 95% CI, 0.85 to 1.6; P = 0.34) and live-birth (OR, 1.5; 95% CI, 0.92 to 2.5; P = 0.10) rates in PCOS patients undergoing IVF and treated with metformin were reported without significant heterogeneity between studies.
On the basis of these results, some authors concluded that to date no adjuvant pretreatment in IVF cycles should be suggested in the clinical practice (236). Furthermore, in a high-risk population such as PCOS patients, metformin seemed to have improved the safety of gonadotropin administration in IVF cycles. In fact, a significant decrease in the total dose of gonadotropins used during IVF cycles (WMD = −290.42 IU; 95% CI, −450.34 to −130.51; P = 0.0004) and in OHSS risk (5.6 vs. 21.0%; OR, 0.21; 95% CI, 0.11 to 0.41; P < 0.00001) was shown by Costello et al. (230), even if a significant heterogeneity across trials was reported. On the other hand, Moll et al. (206), pooling the data of the two RCTs who reported multiple pregnancy data, found no evidence of a significant benefit of metformin on multiple pregnancy rate (OR, 0.93; 95% CI, 0.42 to 2.1), even if a significant reduced risk of OHSS in favor of metformin (OR, 0.33; 95% CI, 0.13 to 0.80) was detected.
In conclusion, metformin administration in infertile PCOS patients scheduled for IVF cycles is useful to reduce the OHSS risk.
IV. Adverse Pregnancy Outcomes
Patients with PCOS have been shown to be at high risk of pregnancy and neonatal complications. The specific effects of metformin on miscarriage, gestational DM, PIH, PE, and poor infant outcomes have been analyzed in the following paragraphs.
A. Miscarriage
An increased risk of miscarriage in PCOS patients after spontaneous or assisted conception was reported up to a maximum of 30–50% (237), and the miscarriage rate seems to be 3-fold higher in PCOS subjects than in healthy women (238). In addition, the syndrome was diagnosed in about 40–80% of women with recurrent miscarriages (239, 240).
Unclear data explained the mechanisms underlying the increased risk of miscarriage in the PCOS population. High LH levels, hyperinsulinemia, hyperandrogenism, and hypofibrinolysis mediated by plasminogen activator inhibitor (PAI) activity could be involved alone or in combination.
Recent experimental findings (239, 240) showed that PCOS patients have serum glycodelin and IGFBP-1 concentrations significantly lower during the first trimester of pregnancy, suggesting a deficient endometrial environment for implantation and pregnancy safekeeping. In addition, PCOS patients who had miscarriages, when compared with patients who did not, showed serum glycodelin and IGFBP-1 levels significantly lower during the third to eighth weeks and ninth to 11th weeks of pregnancy, respectively (239). Glycodelin is secreted by endometrial glands and acts by reducing the endometrial immune response against embryo development (241). IGFBP-1 modulates adhesion processes at the feto/maternal interface (242) and hence may be important in the periimplantation period. Besides being synthesized by the endometrium, IGFBP-1 is primarily synthesized in the liver, and insulin is known to inhibit hepatic IGFBP-1 production (170).
On the other hand, a microarray and PCR analysis of oocyte cDNA revealed that normal and PCOS oocytes, which are morphologically indistinguishable and of high quality, exhibit different gene expression profiles (243). Cluster analysis suggested that androgens and other activators of nuclear receptors may play a role in differential gene expression in the PCOS oocyte (243). Several of these genes were associated with chromosome alignment and segregation during mitosis/meiosis and/or early embryonic development, suggesting a potential role in the reduced developmental competency of PCOS oocytes (243).
Considering the potential beneficial effects that metformin exerts at both systemic and local (endometrial, ovarian, embryonal) levels (see Section V.A), several authors (151, 244–248) suggested that metformin could play an important role in abortion prevention.
In particular, metformin exerts systemic actions by reducing body weight, insulin and PAI-1 levels (244, 248, 249), plasmatic endothelin-I (ET-1), androgen and LH concentrations (250, 251) and by increasing serum IGFBG-1 levels and glycodelin levels (240).
On the other hand, the success of implantation in human reproduction is the result of endometrial and oocyte/embryo factors. In this regard, metformin administration showed beneficial effects on both factors involved. Particularly in PCOS patients, metformin improved the uterine artery blood flow (28, 240) and several endometrial receptivity surrogate markers, as well as endometrial and subendometrial vascularization and endometrial pattern (28). Similarly, metformin normalized the ovarian artery impedance and perifollicular and pericorpus luteum vascularization (28). Thus, metformin could induce a follicular development similar to that observed in healthy women, minimizing hypoxic damage and cytoplasmatic/chromosomal disorders related to poor vascularization (252), as observed in CC-stimulated ovulatory cycles (29), and improving the oocyte and embryo quality. In this regard, an interesting study showed that a 15-d metformin pretreatment, at a dosage of 2000 mg/d, increased the high-quality embryos (37.8 vs. 24.3%), implantation (15.3 vs. 6.2%), and pregnancy (38.2 vs. 16.7%) rates in CC-resistant PCOS patients who underwent in vitro matured oocyte treatment, confirming a potential beneficial role of metformin on oocyte/embryo factors, even if an effect of endometrial receptivity cannot be excluded (254).
As already stated in Section I.B, experimental data demonstrated that metformin also acts by activating AMPK in granulosa cells. In both bovine (55) and rat (56) granulosa cells, metformin decreased steroidogenesis and MAPK3/MAPK1 phosphorylation through AMPK activation. Also in the human granulosa cell line HGL5, cell viability and phosphorylated protein kinase B/AKT and p44/42 MAPK expression were significantly modulated by metformin preincubation (255).
Lastly, a very interesting experimental study (256) demonstrated that metformin also induced AMPK activation within the blastocyst, leading to improved insulin signaling and pregnancy outcomes. In fact, the preimplantation blastocyst stage embryo was demonstrated to be an insulin-sensitive tissue, responsive to insulin or IGF-I via the IGF-I receptor/translocation of GLUT-4 with an increased glucose uptake (257). High insulin or IGF-I concentrations induced a down-regulation of IGF-I receptor (258) with consequent insulin-stimulated glucose uptake reduction, intraembryonic glucose levels dropping, and apoptosis triggering (257). Thus, the pregnancy outcome of these embryos transferred into surrogate mice was poor (259). In the experimental design, Eng et al. (256) exposed murine embryos to 200 nm IGF-I, by creating an experimental environment similar to that observed in PCOS patients, and compared them to embryos cocultured with excess IGF-I plus metformin and embryos cultured in control media. Resulting blastocysts exposed to high IGF-I concentrations showed a decrease in AMPK activation and insulin-stimulated glucose uptake and an increase in the number of apoptotic nuclei (256). Contrarily, blastocysts cocultured in metformin and excess IGF-I were similar to controls. In addition, the implantation rate and fetal size at d 14.5 were significantly lower among IGF-I-exposed embryos transferred into control mothers in comparison with control embryos transferred into control mothers. However, these parameters were reversed by coincubation with metformin and IGF-I before transfer (256).
Notwithstanding these hypothetical mechanisms by which metformin could reduce the abortion risk in PCOS women, the findings from RCTs do not seem to support the use of metformin in the preconceptional period concerning abortion risk reduction.
In particular, a recent meta-analysis (118) evaluating the effects of CC vs. metformin or the combination of both drugs as first-line therapy in anovulatory PCOS patients seeking pregnancy, reported no significant difference in abortion risk when metformin was compared to CC (OR, 1.58; 95% CI, 0.77 to 3.25; P = 0.219) and to CC plus metformin (OR, 0.47; 95% CI, 0.22 to 1.0; P = 0.069). No significant benefit of metformin addition to CC over CC alone was also reported (OR, 0.74; 95% CI, 0.43 to 1.26; P = 0.273).
More recently, a meta-analysis (260) of 17 RCTs, designed to clarify the role of preconceptional metformin administration in PCOS patients, showed no significant overall effect of metformin administration on abortion rate on the entire PCOS population (OR, 0.89; 95% CI, 0.65 to 1.21; P = 0.452). When data were distinguished on the basis of treatment, again no significant effect of metformin on abortion rate was observed when compared either with CC (OR, 1.02; 95% CI, 0.59 to 1.75; P = 0.941) or with no treatment or placebo in therapy-naive (OR, 0.76; 95% CI, 0.21 to 2.81; P = 0.683) and CC-resistant (OR, 1.43; 95% CI, 0.91 to 2.25; P = 0.125) PCOS patients receiving CC. In addition, no significant effect of metformin on abortion risk was detected in patients treated with gonadotropins for monoovulation induction (OR, 0.84; 95% CI, 0.24 to 2.95; P = 0.784) and for IVF cycles (OR, 0.96; 95% CI, 0.40 to 2.34; P = 0.935).
On the other hand, encouraging data derived from retrospective or prospective nonrandomized studies were observed. In fact, several authors experimented with the use of metformin during pregnancy showing a beneficial effect in reducing the incidence of pregnancy loss.
In a pilot study, Glueck et al. (244) firstly demonstrated that PCOS women treated with metformin during pregnancy had a drastic reduction of spontaneous abortion when compared with historical outcome recorded in the same group of women not receiving treatment in previous pregnancies (73 vs. 10%). The same report (244) demonstrated that PAI activity was inversely associated with having live births, and it was speculated that the ability of metformin in reducing PAI activity was related to an improvement in the insulin resistance state and, thus, in abortion rate. This hypothesis was subsequently confirmed (248). In particular, a subanalysis of a previous trial (221) showed a significant decrease in serum PAI-1 activity after metformin administration exclusively in women who had a favorable pregnancy outcome, whereas the serum PAI-1 activity was unchanged in women treated with LOD and women receiving metformin who miscarried (248).
A further retrospective study (246) showed a significantly lower rate of early pregnancy loss in PCOS women who received metformin during pregnancy in comparison with women who did not receive metformin (8.8 vs. 41.9%). The reduction in the abortion rate was remarkable if considering the historical miscarriage rate in the same population not treated with metformin in previous pregnancies (8.8 vs. 70.6%). Also in this study (246), it was proposed that the beneficial effect of metformin on androgen levels and insulin sensitivity was a mechanism to explain the reduced abortion rate.
Congruent with previous studies, Glueck et al. (261) showed in an observational report that metformin administration during pregnancy in PCOS patients was related to a significant reduction in early pregnancy loss. The abortion rates in women taking metformin during pregnancy were 5.4% in those who had never previously conceived, 24.0% in those having a previous abortion, and 30.0% in those who had previous live births without metformin. No difference in the abortion rate was observed between patients who continued metformin throughout pregnancy and those who interrupted treatment during the first trimester (261). This last study reached a power of 87%.
More recently, a prospective cohort study (262) was set up to determine the beneficial effects of metformin use during pregnancy in patients previously diagnosed with PCOS. An early pregnancy loss was experienced in 11.6% of women receiving metformin throughout pregnancy and in 36.3% of subjects who discontinued metformin at the time of conception or during pregnancy (OR, 0.23; 95% CI, 0.11 to 0.42; P = 0.0001) (262).
A role for metformin could also be proposed in PCOS women with recurrent miscarriages, as suggested by a recent case report (247) of an insulin-resistant PCOS woman who received metformin successfully before and during pregnancy. Recently, a placebo-controlled study on 29 PCOS women demonstrated that metformin administration, given before and during pregnancy beyond the first trimester, reduced significantly the abortion risk (15 vs. 55% for metformin and placebo, respectively) in patients with an abnormal oral glucose tolerance test (263).
Finally, regarding abortion prevention, no definitive conclusion on the efficacy of metformin administration during pregnancy may be reached because well-powered double-blind RCTs having abortion rate as their aim are lacking. Currently, two protocols designed to evaluate the effects of metformin administration on pregnant women with PCOS are in a recruiting phase (264, 265); thus, before drawing definitive conclusions on the abortion risk in PCOS patients treated with metformin, the results from these last RCTs (264, 265) are needed.
B. Gestational diabetes
PCOS patients are likely to develop gestational DM in 20 to 40% (266) of cases, whereas approximately 40% of women with gestational DM are likely to have underlying PCO morphology (267).
Meta-analytic data (268) demonstrated a significantly higher chance of gestational DM development in the PCOS population (OR, 2.94; 95% CI, 1.70 to 5.08; P = 0.00001), even if a significant heterogeneity between studies was detected. A subgroup analysis of five higher validity studies showed an OR of 3.66 (95% CI, 1.20 to 11.16; P = 0.02), but significant statistical heterogeneity remained (268).
In this regard, metformin use during pregnancy has also been advocated to reduce the risk of gestational DM in women with PCOS (269). The main mechanisms by which metformin could reduce the likelihood of developing gestational DM were: the reduction of preconceptional weight, insulin, insulin resistance, insulin secretion, and testosterone levels, and the persistence of these effects during pregnancy (270). However, the effect of metformin on maternal androgen reduction in pregnant women was not confirmed in a well-designed RCT (271).
A randomized, placebo-controlled trial on a population of 40 PCOS patients demonstrated that the rate of patients who fulfilled criteria for gestational DM at the eighth week of pregnancy increased significantly at the end of pregnancy (from 20% to 40%), regardless of metformin administration in this period (272).
In a combined prospective/retrospective, open-label, single center comparison (261), a significant reduction from 26 to 4% in development of gestational DM in PCOS pregnant women under metformin, in comparison with previous pregnancies not on metformin, was demonstrated. The rate of gestational DM of the women who received metformin during pregnancy was similar to that observed in the general population (261). These encouraging results were confirmed in a subsequent series (245, 273) that demonstrated a significant reduction in risk for gestational DM in PCOS women taking metformin compared with those not taking metformin.
In particular, in a prospective study Glueck et al. (245) showed that in PCOS subjects metformin use was associated with a 10-fold reduction in gestational DM. However, if these findings could be used to design a blinded RCT, 102 live births (51 randomized to metformin and 51 to placebo) stratified by age and BMI would be necessary to obtain a 95% chance for demonstrating that metformin significantly decreases the development of gestational DM in women with PCOS (245). In a subsequent study (273), metformin plus diet regimen during pregnancy was demonstrated to be effective for primary and secondary prevention of gestational DM. In fact, gestational DM developed in 12% of metformin-treated PCOS patients, in comparison with 30% of previous pregnancies without metformin (273). In addition, in PCOS patients affected by previous gestational DM, only 31% developed DM in subsequent pregnancies on metformin (273).
One study found an increase in perinatal mortality in diabetic women receiving metformin compared with those receiving insulin or sulfonylureas (274). Moreover, that study was biased by the presence of poorly controlled pregestational DM and other maternal comorbidities such as PE.
An overview (275) on the pharmacological management of gestational DM concluded that data are inadequate to support or refute the use of metformin in the treatment of gestational DM because metformin crosses the placenta and consequently could be helpful or harmful to the developing fetus. Nonetheless, Moore et al. (276) showed no difference in glycemic control and neonatal outcomes in 63 women with gestational DM, not controlled for diet and exercise and randomized to receive metformin or insulin. These findings have been confirmed in the Metformin in Gestational Diabetes (MiG) Trial (146). This is an open RCT on a wide population of women with gestational DM designed to evaluate safety and efficacy of metformin compared with insulin therapy for gestational DM and having as its primary end-point a composite neonatal outcome (146). A total of 751 pregnant patients between 20 and 33 wk gestation were randomized to metformin plus supplemental insulin whenever they required suboptimal glycemic control or insulin alone. This trial confirms that metformin alone or with supplemental insulin is as safe as insulin alone in patients with gestational DM, demonstrating a higher satisfaction rate for metformin treatment (146).
C. Pregnancy-induced hypertension and preeclampsia
Women with PCOS have a greater risk of developing PIH (OR, 3.67; 95% CI, 1.98 to 6.81; P < 0.0001; and OR, 3.71; 95% CI, 1.72 to 17.49; P = 0.004, for total and high quality studies, respectively) and PE (OR, 3.47; 95% CI, 1.95 to 6.17; P < 0.0001) than the general population (268).
The mechanisms underlying the increased risk for PIH and PE in PCOS patients remain unclear. However, our preliminary data showed that, similar to that observed in nonpregnant PCOS patients (28, 29), uterine artery resistance was significantly higher in PCOS women than in healthy controls during the first 12 wk of pregnancy. These findings could be interesting, considering that pregnancy complications are always characterized by a failure of normal trophoblastic invasion and lack of uterine spiral artery remodeling leading to a high-resistance uteroplacental circulation (277). In fact, a significant relationship between uterine artery Doppler indexes and histological features of placentation in normal and preeclamptic pregnancies was observed (277), and an experimental study (278) demonstrated an inverse relationship between vascular impedance of the uterine artery and percentage of trophoblastic vessels.
Our preliminary data focused attention on the role of metformin in first phases of the implantation, showing that metformin administration during the first trimester of pregnancy regularized the uterine artery RIs, and no difference between metformin-treated PCOS patients and healthy controls was observed. Thus, metformin in the first phases of the pregnancy might influence the trophoblastic invasion of the maternal decidua, myometrium, and blood vessels, allowing a successful placentation with consequent pregnancy outcome improvement. In agreement with our findings, Salvesen et al. (279) in a small RCT showed that metformin administration in PCOS patients during pregnancy reduced uterine artery impedance between 12 and 19 wk gestation, and this was associated with reduced complication rate.
Unfortunately, few clinical data regarding the preventive effect of metformin in development of PIH and/or PE are available.
In preconceptional diabetic PCOS women treated with metformin (270, 280), the incidence of PE was similar to the general population and untreated PCOS patients. Furthermore, women with pregestational type 2 DM treated with metformin showed a higher prevalence of PE in comparison with women treated with insulin or sulfonylurea (281).
A RCT (279) assessing the potential benefit of metformin on maternal and fetal outcomes in 40 pregnant women with PCOS was recently published. Metformin administration reduced, compared with placebo, pregnancy complications such as preterm delivery before 32 wk, severe PE, or serious postpartum events (279).
In addition, data from a retrospective study (274) on diabetic women suggested an increased rate of PE in women treated with metformin compared with those treated with sulfonylureas or insulin during their pregnancies. However, the groups in this study were not matched due to the retrospective nature of the study, and women who took metformin were more obese and had more poorly controlled DM, both of which are known as risk factors for PE. In addition, the excess perinatal mortality in the metformin group in this study occurred in two subjects who had poorly controlled DM and were noncompliant and therefore cannot be attributed to metformin treatment alone. In fact, these data were not subsequently confirmed (281), and metformin was shown not to be associated with PE in pregnant women with PCOS, appearing safe for mother and fetus.
D. Poor infant outcomes
Unclear data are available regarding worse neonatal outcomes assumed in infants born to women with PCOS.
Pooling the data of all RCTs, a significantly lower neonatal birth weight (WMD = −38.4 g; 95% CI, −62.2 to 14.6; P = 0.002) was detected (268). However, when only higher validity studies were analyzed, no significant difference was found (WMD = 26.5 g; 95% CI, −35.5 to 88.5; P = 0.40) (268). Similarly, no significant increase in the incidence of macrosomia (OR, 1.13; 95% CI, 0.73 to 1.75; and OR, 1.08; 95% CI, 0.6 to 1.96, for total and high-quality studies, respectively) and neonates small for gestational age (OR, 1.16; 95% CI, 0.31 to 5.12; and OR, 0.66; 95% CI, 0.05 to 7.91, for total and high-quality studies, respectively) was observed in PCOS patients (274).
More recently, Bolton et al. (282) showed that metformin administration throughout the first 12 wk of pregnancy is related to a lower incidence of neonates that are small or large for gestational age, a lower risk of neonatal hypoglycemia, and fewer requests for iv glucose therapy.
No significant difference in incidence of neonatal malformations was observed in PCOS patients (OR, 0.70; 95% CI, 0.11 to 4.39) (268). However, a significantly higher rate of admission to a neonatal intensive care unit (OR, 2.31; 95% CI, 1.25 to 4.26) and perinatal mortality (OR, 3.07; 95% CI, 1.03 to 9.21) was observed among women with PCOS vs. controls (268). An increased perinatal mortality in women with type 2 DM in treatment with metformin in comparison with those who received sulfonylurea or insulin was also reported (268).
The safety of metformin administration during pregnancy was evaluated by several authors, and all available studies were in agreement that metformin use was not related to any adverse fetal outcomes or congenital abnormalities.
These last data were confirmed by a meta-analysis (150) that showed no increased risk for major malformations when metformin is administered during the first trimester of pregnancy (OR, 0.50; 95% CI, 0.15 to 1.60; P = 0.24).
Babies conceived during metformin administration had gestational age, gestational length, head circumference, and birth weight similar to those of normal gender-specific infant populations (280, 283). In addition, at routine 3- and 6-month well-baby examinations, infants born to mothers with PCOS treated with metformin during pregnancy had normal motor and social development (280). These findings were confirmed over the first 18 months of life.
The same authors extended their studies to investigate growth and motor and social development in breast- and formula-fed infants born from metformin-treated PCOS women, demonstrating that metformin during lactation was safe in the first 6 months of life (280).
In conclusion, to date no clear evidence is available regarding a positive or negative effect of metformin administration on infant outcomes both during pregnancy and lactation.
V. Endometrial Abnormalities
Two main clinical concerns may be related to endometrial dysfunction in PCOS patients: first, endometrial receptivity might be affected and, thus, the implantation impaired with a consequential increase in abortion risk; and second, the risk for endometrial hyperplasia and cancer could be increased. In this regard, metformin administration may have, as detailed in Section IV.A, an impact on the endometrium, hypothetically both improving the potential for a successful pregnancy implantation and reducing the long-term risks of unopposed endometrial proliferation. Furthermore, before obtaining definitive and clinically valid conclusions, additional research should be carried out.
A. Fertility implications
In human reproduction, endometrial receptivity is, in concert with oocyte/embryo factor, a pivotal factor for the implantation success (284). Several data seem to show that PCOS patients have alterations of several surrogate end-points of endometrial receptivity. In particular, most PCOS patients have endometrial vascularization, pattern, and thickness abnormalities (28, 29). An alteration of the expression of various endometrial proteins has also been demonstrated (239, 240).
Anovulatory PCOS patients have an alteration in uterine vascularization (28, 29, 285, 286). In fact, at Doppler ultrasound, pulsatility index (PI) and RI values, two measures of blood impedance inversely related to blood flow, were not only unchanged throughout the cycle (287) but were also significantly higher in comparison with healthy controls (28, 29).
Elevated impedance to blood flow in the uterine arteries appears to be detrimental for endometrial receptivity (288–290). A high PI detected in the uterine arteries at the time of implantation predicts a poor chance of conception in spontaneous and gonadotropin-stimulated transfer cycles, suggesting that uterine perfusion may be a pivotal factor for a favorable reproductive outcome (288–290). During IVF cycles, the implantation rate is decreased when uterine artery PI is higher than 3.3–3.5 at the time of human chorionic gonadotropin administration (291). The reduction in uterine perfusion seems to be associated with a lower reproductive probability not only in patients affected by PCOS but also in patients with unexplained recurrent abortion (292).
Interesting data (240) have demonstrated that metformin treatment enhanced uterine vascularity and blood flow in women with PCOS. More recently, it was confirmed that metformin has beneficial effects not only on uterine artery blood flow (28, 240) but also on several surrogate markers of endometrial receptivity. In particular, Palomba et al. (28) showed a significant decrease in subendometrial and endometrial vascularity in PCOS women, assessed as RI and/or PI and as number and/or areas of colored spots at computerized image analysis. In this regard, the evaluation of subendometrial vascularity seems to be a more accurate and sensible method for studying uterine receptivity (293).
Metformin increased the vascular penetration of the uterus in zones 3 (vessels penetrating the inner hypoechogenic area of the uterine cavity) and 4 (vessels penetrating the layer surrounding the uterine cavity), zones corresponding to the endometrial layer (28).
Endometrial thickness was considered a marker for uterine receptivity and for pregnancy prediction with different results (294–296). Certainly, patients with a thick endometrium have a higher pregnancy rate compared with those having a thin endometrium, a minimum endometrial thickness being necessary for an effective endometrial receptivity and, thus, for pregnancy success (297). Furthermore, no effect of metformin monotherapy on endometrial thickness in comparison with healthy women has been demonstrated (240, 248). To the contrary, Kocak, et al. (223) demonstrated a significant effect of metformin on endometrial thickness in PCOS women treated with CC.
Endometrial pattern is another critical factor influencing the chance of pregnancy in spontaneous or drug-induced cycles, and a triple-line pattern is associated with a significantly higher pregnancy rate (298). Fanchin et al. (299) demonstrated that endometrial echogenicity, studied using objective methods as computer-assisted measurements, has a significant relationship with implantation and pregnancy rates in IVF cycles, confirming that this parameter can be useful to predict endometrial receptivity.
In this regard, PCOS women ovulating under metformin showed a triple-line endometrial pattern in a percentage of cases not different from those observed in healthy controls (28). Similarly, no difference in the high-grade pattern rate in the midluteal phase was also observed between ovulating PCOS patients under metformin and in healthy controls (28).
The mechanisms hypothesized for metformin action are various. Metformin can act directly and/or indirectly on endometrium.
A direct effect of metformin on endometrium via insulin action can be hypothesized. In fact, insulin stimulates glucose oxidation activity in the late luteal phase in human endometrium, suggesting its involvement in endometrial tissue metabolic activities (300, 301). In addition, insulin receptors are present at the endometrial level, reaching their maximal expression in the secretory phase, further supporting the hypothesis that insulin directly influences endometrial growth by means of its mitogenic and metabolic effects (302). Finally, GLUT-1, a non-insulin-regulated transponder, is expressed at endometrial levels in animals and humans (303, 304), indicating the importance of glucose transport in the endometrium (304, 305). At this level, another insulin-dependent transporter, GLUT-4, is also expressed and seems to be regulated by body weight and insulin. Specifically, a recent experimental study showed that GLUT-4 expression is impaired in PCOS patients, suggesting that in these subjects insulin resistance and hyperinsulinemia induce an inadequate GLUT-4 expression, and thus a reduced glucose supply. The consequences of this alteration could be responsible for the impaired endometrial receptivity, and metformin could be effective in endometrial functionality restoration through a direct effect.
The beneficial actions of metformin on the endometrium, and particularly on uterine vascularity, could be due to an indirect effect of the drug on serum androgen levels. In fact, it seems that insulin resistance did not play a key role in reducing uterine perfusion in a PCOS patient (307). On the contrary, it is possible to suppose that metformin acts on uterine perfusion, reducing androgen levels and, thus, their vasoconstrictive effect on vascular tissues (307).
In the endometrium of PCOS women, an altered modulation of the expression of the endometrial androgen receptors (ARs) is also present. Contrary to healthy women, an elevation in the endometrial expression of ARs during the late luteal phase has been detected in PCOS women, and this elevation seems to be strongly related to β3 integrin, a well-characterized biomarker of endometrial receptivity (308). The AR activation also seems to have a negative impact on the expression of a homeobox gene, i.e., HOXA10, a gene implicated in the endometrial plasticity during menstrual cycles and essential for endometrial receptivity (309). HOXA10, in fact, is significantly and inversely related to testosterone but is not regulated by insulin (310). Elevated levels of androgens, as observed in PCOS women, may have a detrimental effect on endometrial function via HOXA10 modulation (310).
Jakubowicz et al. (240) showed that metformin increased the midluteal phase concentrations of serum glycodelin and IGF-I, two putative biomarkers of endometrial receptivity, 3- and 4-fold, respectively, and that this effect was related to the decrease in androgen levels.
The beneficial effect of metformin on ET-1, a peptide synthesized in the vascular endothelial cells with vasoconstrictive action, should be considered as another indirect mechanism by which this drug acts on uterine vascularity. ET-1 is expressed in human endometrium (311) and inversely related to plasma E2 levels (312), suggesting that ET-1 and E2 could operate in concert to modulate endometrial blood flow. Clinical data demonstrated that in PCOS women there are increased concentrations of serum ET-1 (313) and that metformin administration improves this alteration (251).
B. Cancer implications
An association between PCOS and endometrial carcinoma was first suggested in 1949 (314). Although another study (315) quoted a 37% prevalence of endometrial cancer, successive studies (316, 317) obtained controversial data probably due to several confounding factors.
Coulam et al. (317) calculated a relative risk (RR) of 3.1 to develop endometrial cancer after a diagnosis of chronic anovulation, whereas women with anovulatory infertility were shown to have a RR of 4.2 for endometrial cancer. In addition, women with ovarian factor infertility are at risk for developing endometrial hyperplasia, and when atypical endometrial type is present, it represents a potentially premalignant form. Finally, obesity, frequently present in PCOS patients, seems to increase the risk of endometrial cancer 2.6- to 3-fold, suggesting that it is possibly a confounding factor or a pivotal cofactor in PCOS (318).
The main pathogenetic mechanism assumed to be responsible for increased risk for endometrial cancer in PCOS patients was hyperestrogenic stimulation of endometrial growth, either unopposed by progesterone or opposed to a less extent than in normally cycling women. In fact, estrogens act by genetic and epigenetic mechanisms on cancer cells, and a close relationship between estrogens, growth factors, and oncogenes is important in the development of several human cancers (319).
However, other mechanisms seem to be involved in the development of endometrial cancer in women with PCOS. In particular, LH hypersecretion could be associated with overexpression of LH/hCG receptors, and this condition could be a feature of endometrial hyperplasia and carcinoma in young anovulatory women, including PCOS subjects (320). In addition, Li et al. (321) showed that the expression of the progesterone receptor, of which dysregulation is involved in the promotion of endometrial neoplasia in the stroma of hyperplastic endometrium from PCOS women, is unevenly distributed and significantly less than seen in the normal endometrium found in non-PCOS women.
In the normal menstrual cycle, during the late secretory phase, active endometrial cell death occurs via apoptosis, and the Bcl-2/Bax system seems to play a role. Bcl-2 is an apoptotic inhibitor that promotes cell survival, whereas Bax induces apoptosis. Endometrial apoptosis is regulated by estrogen and progesterone, which are known to modify the expression of apoptotic proteins during the menstrual cycle. In women with PCOS, the endometrium displays increased Bcl-2 to Bax ratio in comparison with endometrium from women without PCOS (322). The greater expression of Bcl-2 activity in relation to Bax in endometrium of women affected by PCOS resulted in prolonged cell survival (322).
Lastly, increased concentrations of plasma insulin in patients with endometrial cancer (324) and insulin-binding sites have been found in the endometrial stroma of premenopausal women and women with endometrial cancer (323). Thus, compensatory hyperinsulinemia could exert a deleterious effect on endometrium acting on specific insulin-binding sites exerting a mitogenic effect. In particular, an isoform of the insulin receptor, predominantly expressed in malignant tissues, may lead this mitogenic effect.
Several additional mechanisms have been proposed for the mitogenic effect of insulin in endometrial cancer (325). In addition, insulin resistance may be associated with alterations of IGF and IGFBP expression or may inhibit the protective effect of progesterone (326). Binding sites for IGF-I and IGF-II have been confirmed in both normal and malignant endometrium. IGF-I binding is significantly higher in endometrial cancer compared with normal endometrium, and in the Ishikawa human endometrial cancer cell line IGF-I was a more potent mitogen than insulin or IGF-II (327). Interestingly, IGFBP-1 was undetectable or minimally expressed in endometrial cancers, and hyperinsulinemia can decrease IGFBP-1 levels significantly; because insulin up-regulates aromatase activity in endometrial glands and stroma, endogenous estrogen production is enhanced in women with high insulin levels as well as PCOS patients (325).
More recently, Pillay et al. (327) compared gene expression in the endometrial proliferative phase with or without simple hyperplasia derived from women with PCOS matched for age, parity, and biochemical profile, showing differences in gene expression related to hyperplasia. By microarray analysis, 24 genes were found to be up-regulated more than 2-fold in the hyperplastic group, for which three genes are known to have roles in endometrial carcinoma, namely human asialoglycoprotein receptor I, human-secreted phosphoprotein I (osteopontin), and cytochrome P450 XVIIAI (327).
Based on these considerations, metformin could theoretically have a beneficial effect in reducing the risk for endometrial cancer in PCOS patients. In fact, its administration may have the power to reduce the risk of unopposed endometrial proliferation, hyperplasia, or carcinoma by improving the regularity of ovulatory function and by reducing the effect of hyperinsulinemia on the endometrium.
Increased insulin sensitivity may improve the metabolic effect of insulin and decrease its mitogenic effect by tissue-specific mechanisms through the relative expression of the insulin receptor isoforms (328).
Notwithstanding these theoretical mechanisms, prospective studies providing clinical data are very few (329, 330).
Session et al. (329) first reported one case of atypical endometrial hyperplasia treated with metformin. In particular, the authors described a 37-yr-old patient treated with metformin after failed progestogen treatment for endometrial hyperplasia. One month after initiating metformin therapy, the patient’s endometrial biopsy demonstrated an atypical endometrial hyperplasia regression and the presence of proliferative endometrium (329).
A subsequent study (330) on a very small sample of seven PCOS patients confirmed these data. Sequential endometrial biopsies performed during metformin administration showed that endometrial histology normalized under treatment. In particular, patients with simple hyperplasia or adenocarcinoma at baseline showed no evidence of abnormal histology after 6 months of insulin-sensitizing therapy. Unfortunately, in this last study (330), the distinction between the beneficial effects of the drug due to the ovarian function improvement/restoration and the direct effects of the drug on the endometrium itself remained unclear.
VI. Hyperandrogenism
Hirsutism is the main clinical feature of hyperandrogenism in PCOS patients, and it is present in approximately 75% of these subjects (17). Moreover, there is no clear correlation between androgen levels and presence and grade of hirsutism in PCOS women (167) because the clinical presentation is related to genetic susceptibility of the pilosebaceous unit to serum androgen levels (331).
Other signs of hirsutism, such as acne and androgenic alopecia, are less frequent, involving approximately 40% of the population (17). Finally, frank virilization with deepening of the voice, clitoromegaly, severe androgenic alopecia, and/or masculine jaw is a very rare event.
Hyperandrogenemia may be affected by metformin. Preliminary in vitro data (83) indicated that metformin may directly decrease ovarian androgen production, and excellent experimental data (84, 85) demonstrated that metformin reduces the ovarian P450c17α activity with a consequent decline in the serum free testosterone concentration.
Ample evidence supports the beneficial effect of metformin on hyperandrogenemia in PCOS patients.
The meta-analysis by Lord et al. (30) showed a significant effect of metformin in comparison with placebo or no treatment regarding total testosterone (WMD = −0.34; 95% CI, −0.59 to −0.10; P = 0.005) and androstenedione (WMD = −1.21; 95% CI, −1.79 to −0.62; P = 0.00005), with and without significant heterogeneity, respectively. Decreases in free testosterone levels could hypothetically be observed in PCOS patients receiving metformin due to the up-regulation of circulating levels of SHBG caused by improved insulin sensitivity (30). Furthermore, clinical evidence is controversial. Meta-analytic data (30) reported a significant treatment effect of metformin (WMD = −4.40; 95% CI, −6.39 to −2.42; P = 0.00001) on free testosterone with a significant heterogeneity correlated to two of the studies analyzed. Therefore, excluding these studies, homogeneous data showing evidence of treatment effect (WMD = −1.79; 95% CI, −4.14 to 0.56; P = 0.14) were obtained (30). In addition, metformin was reported as not affecting SHBG levels (OR, 2.80; 95% CI, −1.11 to 6.71; P = 0.2), but a significant heterogeneity was reported (30). Only one RCT (30) showed no evidence of effect from both metformin and CC on total testosterone (OR, 0.0; 95% CI, −0.59 to 0.59; P = 1.0) or SHBG (OR, 10.0; 95% CI, −10.17 to 30.17; P = 0.3), but there was a significant treatment effect on free testosterone (WMD = −3.30; 95% CI, −5.07 to −1.53; P = 0.0003).
Whether metaformin is less effective than androgen-lowering drugs, is a debated issue since an RCT found that metformin had similar androgen-lowering efficacy in both obese and morbidly obese women affected by PCOS (111), whereas in another RCT nonobese women did not seem to benefit from metformin (332). A result that was difficult to understand rose from the meta-analysis by Lord et al. (30) regarding the significant increase of adrenal androgen dehydroepiandrosterone sulfate levels in metformin-treated patients (WMD = 0.66; 95% CI, 0.03 to 1.28; P = 0.04), but in this study a significant heterogeneity was found. In this regard, a recent double-blind, placebo-controlled, randomized study confirmed that the administration of 2000 mg metformin for 14 wk increased serum dehydroepiandrosterone sulfate levels alone (333). Of note, the suspension of metformin induced a potential “rebound” effect with an increase of serum androstenedione levels as well as of free testosterone index (333).
A comparison of metformin (2250 mg/d), rosiglitazone (4 mg/d), and the combination of both drugs to placebo demonstrated that the mean serum free testosterone levels in nonobese non-insulin-resistant PCOS patients on therapy were significantly lower than the levels found in subjects on placebo without significant advantage for the combination treatment (159). The similar efficacy of metformin and other thiazolinediones has also been confirmed in a further study comparing 2250 mg/d metformin to 30 mg/d pioglitazone randomly given for 6 months in 52 PCOS patients (334). In fact, both treatments demonstrated significant decreases in free testosterone, androstenedione, and LH levels with an approximately 30% decrease in the hirsutism score (334). On the other hand, a significant decrease in androgens was observed only under 4 mg/d rosiglitazone, whereas no significant change was detected in PCOS patients receiving 1700 mg/d metformin (155). The length of treatment was 12 wk, and the population was composed of both lean and obese PCOS patients (155).
To date, there is an increasing interest for combined therapies. A prospective study (335) designed to evaluate the effects of metformin plus flutamide on metabolic parameters and insulin resistance in nonobese women with PCOS showed a similar reduction in free testosterone after administration of both 2250 mg/d metformin and 250 mg/d flutamide. In addition, metformin plus OCs provided benefits on hyperandrogenism and insulin resistance in nonobese PCOS women not desiring pregnancy, but the effects on central adiposity and dyslipidemia were weaker (336). Recent studies showed that the combination of metformin and low-dose flutamide was related to endocrine and metabolic benefits in nonobese (337, 338) and obese (339) PCOS patients by acting on the upstream abnormalities of PCOS without merely masking downstream symptoms such as OCs. Lastly, 6-month low-dose dexamethasone administration further reduced androgen levels in PCOS patients treated with metformin and lifestyle interventions (340).
Besides the metformin effect on biochemical hyperandrogenism, its efficacy on clinical manifestations of hyperandrogenism is inconsistant. In this regard, the lack of significant benefit from metformin administration in hirsute PCOS patients could probably be explained with the need of longer follow-ups.
In an open-label study of 39 women with PCOS, metformin given at a dose of 500 mg three times daily significantly decreased the Ferriman-Gallwey score at the end of the 12-wk study period (341).
The meta-analysis by Lord et al. (30) analyzed two RCTs reporting hirsutism scores under metformin or metformin plus CC. Pooling data from these studies, a significant effect of metformin on hirsutism without significant heterogeneity (WMD = −5.12; 95% CI, −8.77 to −1.47; P = 0.006) was observed. Another double-blind, placebo-controlled, crossover study confirmed the effect of metformin in improving hirsutism, assessed by using the Ferriman-Gallwey score, patient self-assessment, and hair growth velocity (342).
Three RCTs compared the efficacy of metformin and OCs, consisting of 35 mg E2 and 2 mg CA (182, 183, 343). In particular, Harborne et al. (343) demonstrated a significant reduction in hirsutism with metformin, whereas two other clinical trials (182, 183) showed a trend in the opposite direction. A greater decrease in the hirsutism score with 1500 mg/d metformin than the OCs was observed in 37 patients after 12 months of treatment (343). On the other hand, in another study (182) a greater decrease in the hirsutism score with OCs was observed in 18 obese PCOS patients who received either metformin (1000 mg/d for 3 months followed by 2000 mg/d for 3 additional months) or OCs (182). The same group of researchers published another analysis (183) confirming their results in nonobese PCOS subjects treated with the same protocol. More recently, Luque-Ramírez et al. (186) demonstrated that OCs were more effective than 1700 mg/d of metformin in hyperandrogenism control and in restoration of menstrual cyclicity in 34 PCOS patients receiving the treatment for 24 wk. In addition, OCs were not associated with any clinically relevant worsening in the classic metabolic and cardiovascular risk profile in their population (186). No significant difference was seen in the development of type 2 DM between metformin and also OCs (OR, 0.17; 95% CI, 0.00 to 8.54; P = 0.37) (190).
A meta-analysis (181) of three RCTs comparing the effects of metformin and OCs on hirsutism demonstrated no difference in efficacy between these two drugs (standardized mean difference = −0.18; 95% CI, −0.67 to 0.32; P = 0.48). Furthermore, statistical heterogeneity, due to differences in the selection criteria for PCOS participants, the assessment method for hirsutism, and duration of treatment, was present.
The efficacy of metformin on hirsutism was also compared with antiandrogen therapy. A 6-month prospective, randomized, placebo-controlled trial (344) on 76 obese PCOS women demonstrated that PCOS patients who underwent a hypocaloric diet 1 month before treatment had benefits significantly higher from flutamide (500 mg/d) than from metformin (1700 mg/d) and that a combination therapy with metformin plus flutamide did not add any further benefit in comparison with flutamide alone. The combination of flutamide plus metformin could be efficient in normalizing the spectrum of the disorder in PCOS patients (338). However, addition of OCs is advocated to avoid the potential embryotoxicity; thus, it cannot be considered a standard therapy for widespread clinical practice (338).
Metformin seemed to be effective in improving achanthosis nigricans in PCOS patients after 6 months of treatment, but no standardized measure to assess clinical manifestation was used in the study (45). More recently, an open-label pilot study involving overweight or obese subjects with acanthosis nigricans randomized to 12 wk of either metformin or rosiglitazone showed no effect on the severity of acanthosis nigricans and a modest improvement of skin texture in both treatment groups (345).
Similarly, 3 months of metformin administration had poor effect on the acne score in young PCOS women (341). A single RCT (343) comparing metformin vs. OCs regarding acne demonstrated no significant difference in acne scores (WMD = 0.90; 95% CI, −0.40 to 2.20; P = 0.18) (181).
Briefly, metformin as well as OCs exert a relatively modest beneficial effect on excess hair growth that is significantly inferior to those observed with antiandrogen therapies. to date, other multi-drug regimens are under study. Current guidelines derived from meta-analyses (346, 347) of RCTs on medical treatments for hirsutism in women clearly stated that there is strong evidence that metformin is not a choice therapy for hirsutism and other strategies should be used (347). In fact, the random-effect meta-analyses of 16 RCTs showed a small decrease in the Ferriman-Gallwey score in hirsute patients receiving ISDs when compared with placebo (WMD= −1.5; 95% CI, −2.8 to −0.2) (347). ISDs improved hirsutism similarly to OCs (WMD= −0.5; 95% CI, −5.0 to 3.9), but less than with spironolactone (WMD=1.3; 95% CI, 0.03 to 2.6) and flutamide (WMD=5.0; 95% CI, 3.0 to 7.0) (347). Lastly, a recent Task Force behalf the Endocrine Society in clinical practice guidelines suggested against the use of ISDs as therapy for hirsutism, whereas recommended OCs or antiandrogens (348).
VII. Malignancies
The association between PCOS and endometrial cancer has been reported previously (see Section V.B). Anecdotal cases of low-grade endometrial stromal sarcoma and carcinosarcoma in women with PCOS have been described in the literature (349). Data regarding PCOS and epithelial ovarian and breast cancer derived from few studies, which seem to exclude any association between these conditions seems to be excluded by few available data (349).
Some epidemiological and experimental evidence suggests that insulin and IGFs influence the risk and the prognosis of cancer with particular regard for the breast cancer (350). In the same manner, elevated androgen levels have been proposed as factors involved in the oncogenetic processes with several mechanisms (349).
It is intuitive that drugs that can reduce both insulin resistance and hyperandrogenism, i.e., metformin, could be proposed as treatments with potential roles in the prevention and treatment of some types of cancers.
A recent study (351) suggests that metformin may reduce the risk of cancer, but its mechanism of action remains unclear. In vitro and in vivo experimental data have demonstrated that metformin led to a strong reduction of cyclin D1 protein level in tumors, suggesting evidence for an ulterior mechanism in the explanation of metformin’s antitumoral effect (351).
As detailed in Section I.B, in vitro and in vivo studies have demonstrated that metformin induces LKB1, a serine-threonine protein kinase that activates the AMPK by phosphorylation of the proximal site (60, 61), acting as tumor suppression. Recently, great attention has been given to the mammalian target of rapamycin (mTORC) pathway dysregulation in breast cancer because several experimental and epidemiological data have shown that this pathway influences cancer development directly or through AMPK (352).
Based on these considerations, metformin could exert a pleomorphic antiproliferative action on various malignancies. In fact, in diabetic patients treated with metformin, a low incidence of cancer has been demonstrated targeting AMPK pathway activation as the main antiproliferative effect of metformin (352).
The deficiency of phosphatase and tensin homolog tumor suppressor induces cell growth through an overactivation of the mTORC1 kinase, which is activated in the majority of human cancers. In mice deficient in phosphatase and tensin homolog tumor suppressor, the tumor onset was delayed with the use of metformin, which induces LKB1 and the consequential activation of AMPK pathway that acts to inhibit mTORC1 kinase (353).
Stimulation of AMPK by metformin resulted in a significant repression of cell proliferation and active MAPK1/2 in both estrogen receptor α (ERα)-negative and -positive human breast cancer cell lines (354). Furthermore, a different biological response to metformin was detected between ERα-negative and -positive human breast cancer cell lines. In fact, with metformin, only the first cell lines demonstrated an AMPK-dependent increase of the vascular endothelial growth factor expression (352, 354). These findings obtained in vitro have also been tested in the animal model (354). Athymic nude mice with ERα-negative breast cancer treated with metformin showed increased vascular endothelial growth factor expression, microvascular density, and reduced necrosis, even if the serum IGF-I levels and the tumor cell proliferation rate in vascularized regions were reduced (354). However, the overall tumorogenic progression was improved compared with untreated controls (354). Unfortunately, the metformin-induced neoangiogenesis could be a risk factor in breast cancer development (355).
In conclusion and in consideration of the available data, great caution should be given to the potential beneficial effects of metformin on tumorigenesis and to the translation of the in vitro findings to the in vivo model.
VIII. Quality of Life Impairment
Several data available in the literature (356) confirm that PCOS negatively affects the patients’ health-related QoL, a multidimensional concept that includes physical, emotional, and social aspects.
In two studies, QoL in PCOS patients was found to be worse than in the healthy control group (357, 358). This finding is understandable given several PCOS-related features. More interesting are the data comparing QoL in patients with PCOS and/or other diseases. One study (359) designed to evaluate the Short-Form Healthy Survey (SF-36), a well-validated generic instrument to assess the QoL, in UK people affected by several diseases was recently published. In this survey, the PCOS patient’s QoL was similar to those of patients with asthma, epilepsy, DM, and back pain, and better than patients with arthritis and coronary heart disease. Furthermore, the items related to the psychological aspects were significantly worse than in patients affected by other diseases (359).
As reported by a recent systematic review (356), obesity seems to exert the greatest negative influence on QoL in PCOS patients with a complex relationship with psychosocial well-being and sexual satisfaction. Infertility and menstrual disorders seemed to influence QoL assessed with both generic and PCOS disease-specific questionnaires (356). Clinical signs of hyperandrogenism on both PCOS and non-PCOS patients had a negative impact on self-image and esteem (356), and the treatment of these conditions improved the QoL and psychological well-being of the patients (360). In addition to somatic impairments, mood disturbances such as depression, limitations in emotional well-being and life satisfaction, diagnosis of PCOS also has a negative impact on sexual self-worth and sexual satisfaction (361).
In this regard, treatments used in PCOS patients by ameliorating the clinical aspects and preventing the long-term consequences of the disease could improve their QoL. Presently, few data (358, 362, 363) have evaluated the impact of specific treatment on QoL in PCOS patients.
Only one prospective observational study (358), not including placebo or control groups, evaluated the effect of metformin administration on QoL in a German PCOS population by using four standardized questionnaires: the Health-Related QoL (HRQL) and the SF-36 for generic measure of QoL, the Symptoms Check List 90 (SCL-90-R) for psychological disturbances, and the Visual Analog Scales (VAS) for sexual satisfaction. Metformin administration improved QoL, particularly for the psychological aspects rather than for physical ones, psychological disturbances, and sexual satisfaction. However, even if improved, these measurements remained worsened in PCOS patients compared with the normal German female population used as control. In addition, the authors (358) computed a correlation to address possible associations between metformin-induced clinical and/or biochemical improvements and psychological improvement. An interesting correlation between the clinical effect of metformin and psychological improvement was observed.
Presently, comparative data on several treatment options in PCOS using a disease-specific questionnaire are lacking. In fact, even if the PCOS Questionnaire (PCOSQ), an instrument consisting of five domains formed by 26 questions on emotions, body hair, weight, infertility, and menstrual problems, has been validated (364), no well-designed RCT evaluating the impact of treatments on QoL is available.
IX. Obesity
Obesity is observed in up to 60% of patients with PCOS in the United States (365), but it is less frequent in other populations (366).
More than a real presence of obesity or high BMI, the distribution of fat is the main significant feature in PCOS (367). Visceral adiposity is known to be metabolically active, is more highly associated with hyperinsulinemia than sc fat (368), and is more frequently noted in PCOS patients (369). In fact, regardless of ethnicity, most overweight women with PCOS have central obesity (android pattern) (369), and 70% of lean women with PCOS have an android distribution of fat (370).
Whether obesity is a promoting environmental factor or an intrinsic clinical sign of PCOS remains unclear. Certainly, obesity has a profound impact on the phenotypic expression of PCOS, being associated with more severe hyperandrogenism, insulin resistance, and fertility disorders.
Obesity, particularly central obesity, is related to and responsible for hyperinsulinemia that correlates with increased cardiovascular risk (371). Moreover, obesity is independently associated with intermediate cardiovascular risk factors, such as hypertension and type 2 DM. Lastly, obesity is also related to reduced fecundity (372). In fact, an increased rate of menstrual disorders and anovulatory infertility (373) and a reduced clinical response to the infertility treatments (158, 374–377) was observed in obese women. Higher risk for miscarriage, gestational DM, hypertension, and unfavorable obstetrics outcomes was also reported in obese subjects (378).
Previous studies demonstrated that weight loss induced by diet and/or physical activity causes endocrine, metabolic, and reproductive improvements in obese PCOS patients wishing to conceive (210, 211). In this regard, the UK guidelines for the management of obese women with PCOS recommend weight loss, preferably to a BMI of less than 30 kg/m2, before starting drugs for ovarian stimulation (379). Unfortunately, lifestyle modification programs are often not followed by weight loss maintenance, probably because of the low retention rate frequently related to lifestyle modifications (380). In fact, a high dropout rate from these programs often characterizes these interventions, even if performed for short periods (381). Although, both diet and exercise are effective in improving fertility in obese PCOS patients with anovulatory infertility, a structured exercise program has a higher compliance than diet alone (382).
Data obtained from a type 2 DM population receiving metformin seemed to clearly demonstrate that this drug, contrary to most antidiabetic agents, is not associated with weight gain (383). Thus, metformin can improve insulin sensitivity but does not reduce this effect with an increase in body weight (383). In this regard, a recent clinical study demonstrated that low-dose metformin pretreatment contrasted the increase in BMI due to pioglitazone administration, improved the glycemic control, and reduced the atherogenic index in type 2 DM patients (384). In addition, metformin, both as monotherapy (385) and combined with lifestyle intervention (386), has been demonstrated to reduce weight gain related to increased insulin resistance due to antipsychotic medication, and metformin alone was more effective in weight loss and insulin sensitivity improvement than lifestyle intervention alone (386). Thus, indirect evidence seems to relieve concern about a potential increase of body weight in PCOS patients already at risk for obesity.
On the other hand, data regarding the potential beneficial effects of metformin on obesity in PCOS patients are controversial.
The link between obesity and insulin resistance is strong (371), and the data showing a significant improvement in insulin resistance seem to be a key factor in determining that metformin administration could be effective in reducing body weight in several conditions in which insulin resistance could have a role in weight gain, such as in PCOS (383).
Metformin could act to improve body weight in obese PCOS patients both directly on the central nervous system and indirectly via adipokines modification. In fact, metformin has been shown not only to stimulate catabolic processes in the peripheral organs but also to modulate appetite in the hypothalamus (387). Indeed, a recent experimental study showed that acute and chronic administration of metformin increases anorectic gut hormone peptide YY fasting plasma levels in PCOS patients (388). Of note, this increase was significantly related to waist circumference reduction (388). Thus, metformin probably acts not only in the reduction of the total quantity of body fat, but also and especially in the visceral fat, which is the main endocrine organ producing metabolically active adipokines.
Not only insulin, ghrelin, neuropeptide Y, and leptin but also several other adipokines interact in the regulation of energy homeostasis in healthy women. On the other hand, an alteration of these metabolic signals was demonstrated in patients with PCOS (389). In these patients, metformin administration had no effect on leptin levels, whereas it improved the blunted neuropeptide Y release induced by ghrelin injection (390). However, in patients with type 2 DM, metformin decreased plasma ghrelin levels after glucose load (391) and prolonged the postprandial fall in ghrelin concentration (392). Recently, Tan et al. (393) reported a high mRNA and protein expression of vaspin in omental adipose tissue in overweight PCOS women and a significant effect of 6-month metformin administration on plasma vaspin levels.
Notwithstanding these intriguing experimental data on a potential effect of metformin on body weight, clinical data regarding the effectiveness of metformin on the improvement of body weight in PCOS patients, with particular regard for the obese, are very controversial.
Initial data (32) of metformin treatment reported a loss in body weight in women with PCOS. Successively, Tan et al. (154) analyzed data from three groups of PCOS patients according to their BMI, i.e., normal weight, overweight, and obese. The authors (154) found that metformin use was significantly associated with decreased body weight and BMI in the overweight and obese groups. On the other hand, the use of metformin was observed to decrease body weight, even in nonobese PCOS women (335). Other investigators reported a significant reduction in waist circumference, but no significant change was observed in weight in obese subjects treated with metformin (147). In another study, researchers observed that metformin reduced BMI in patients both with and without insulin resistance, but it had no influence on the WHR (394).
Lord et al. (30) showed no significant effect from metformin on body weight (OR, −5.83; 95% CI, −18.43 to 6.78; P = 0.4) or BMI (OR, −0.27; 95% CI, −0.98 to 0.45; P = 0.5). In addition, no significant difference in waist circumference (OR, −4.00; 95% CI, −16.95 to 8.95; P = 0.5) or WHR (OR, 0.0; 95% CI, −0.02 to 0.02; P = 0.7) was reported.
In a subsequent RCT (111) specifically designed to evaluate the effects of metformin therapy on body weight, BMI decreased significantly by approximately 4% in both obese and morbidly obese women with PCOS after metformin therapy (500 or 850 mg three times daily) without lifestyle intervention. In the same study (111), obese PCOS patients benefited from greater weight loss at the highest dose (2550 vs. 1500 mg/d) of metformin, whereas a similar degree of weight loss at both doses of metformin was detected in morbidly obese PCOS subjects.
Costello et al. (190), in a meta-analysis comparing the effects of metformin vs. OCs on several end-points, demonstrated no difference in BMI or WHR between the two treatments. Only one clinical trial comparing 6 months of OCs (35 μg E2 combined with 250 μg norgestimate in a cyclic regimen of 21 d of active pills followed by a 7-d pill-free interval) alone vs. OCs combined with 1500 mg/d metformin reported no significant difference in body weight (336). Finally, meta-analysis (190) of two studies investigating OCs alone vs. OCs combined with metformin again revealed no difference in BMI between groups.
A recent RCT (110) demonstrated a significant BMI reduction in PCOS patients receiving metformin alone or in combination with CC, although the treatment effect on BMI was not an end-point of the trial.
The additive effects of metformin in PCOS patients under lifestyle modification programs were also investigated.
Pasquali et al. (181) compared the efficacy of a low-calorie diet combined with metformin or placebo in PCOS patients and controls. Reduction of body weight, BMI, and visceral fat mass was greater in the metformin group than the placebo group. A study (395) on overweight PCOS patients confirmed that the combination of metformin plus lifestyle intervention was more effective in weight and androgen reduction than placebo plus lifestyle intervention. No difference in ovulation rate was reported between groups, but ovulation rate was significantly higher in patients who lost weight (395). By contrast, a RCT (396) evaluating metformin (850 mg twice daily) plus lifestyle modification in obese PCOS women did not find a significant improvement in body weight or menstrual frequency, even though that study was not powered to examine weight loss as an end-point. Nevertheless, the metformin group lost twice as much weight as the placebo group, and metformin therapy resulted in a significant reduction in waist circumference compared with lifestyle changes alone. In addition, a logistic regression analysis, used to analyze the independent predictors for menstrual restoration, showed that weight loss alone correlated with improvement in menses. These studies (395, 396) suggest that metformin therapy might facilitate modest weight loss and visceral fat reduction, particularly when combined with a hypocaloric diet, but some of these effects could be dose-related.
The effect of metformin on body weight was also investigated in subjects at high risk for PCOS development (see Section XII) or type 2 DM (397) and subjects with excess weight (398). In subjects at high risk for type 2 DM, metformin reduced BMI (OR, −5.3; 95% CI, −6.7 to −3.5; P < 0.00001) compared with placebo or no treatment (397). Statistically significant reductions in BMI were seen in women with PCOS (OR, −5.3; 95% CI, −7.2 to −3.4) and without known PCOS (OR, −5.4; 95% CI, −7.2 to −3.5), with obesity (OR, −5.1; 95% CI, −5.5 to −3.6) and without obesity (OR, −6.3; 95% CI, −9.4 to −3.3) (397). Lastly, as demonstrated in a recent 6-month clinical study (398) in a population of patients with excess weight that chose their own management, 2550 mg/d of metformin was significantly more effective than diet in reducing the incidence of overweight and obesity.
In conclusion, on the basis of the present data, metformin had a marginal effect on weight loss as monotherapy. Thus, both clinicians and patients should be advised against the use of metformin as an antiobesity drug. On the other hand, lifestyle modifications remain the cornerstone for weight loss in obese PCOS patients, although metformin cotreatment might improve the efficacy.
X. Cardiovascular Risk
Because of its association with insulin resistance, PCOS can lead to several metabolic complications. Although prospective controlled data on CVD morbidity and mortality in PCOS patients are lacking, various risk factors for CVD are detected in the PCOS population. In particular, PCOS subjects have an increased incidence of obesity, hyperinsulinemia, hypertension, dyslipidemia, elevated PAI-1 and ET-1, endothelial disarrangements, chronic inflammation, and impaired fibrinolysis.
The pathogenetic mechanisms accounting for these abnormalities are still controversial, but the most qualified data suggested a relationship with both hyperandrogenism (399) and impaired peripheral insulin sensitivity (400). However, the main risk factors for CVD in PCOS patients are probably IGT and type 2 DM.
In patients with type 2 DM, the UK Prospective Diabetes Study (UKPDS) (401) showed that metformin may be cardioprotective. In this large prospective trial on obese diabetic patients, metformin significantly decreased all-cause mortality and stroke end-points. Recent epidemiological findings on 1181 patients with type 2 DM showed that, controlling for confounders, i.e., glycemic control, metformin is as effective as insulin in terms of mortality and is probably more effective than insulin in terms of nonfatal myocardial infarction and stroke, suggesting a potential effect not mediated by insulin sensitivity (402). Also, other clinical data (396) have reported that metformin reduces cardiovascular end-points in subjects at risk for type 2 DM by actions that cannot be attributed solely to glucose-lowering effects. Based on these considerations, metformin might be used for long-term therapy designed to reduce the risk for CVD in normoglycemic patients, as frequently observed in PCOS.
Although the effects of ISDs on cardiovascular risk factors are favorable, currently there are no RCT examining the use of these drugs in the prevention of DM or cardiovascular events in women with PCOS. A recent experimental study (403) demonstrated that a very low dose of metformin exerted a cardioprotective effect in a nondiabetic murine model of myocardial ischemia-reperfusion injury, improving AMPK activation, already activated by myocardial ischemia as an endogenous protective signaling mechanism, and increasing endothelial NOS phosphorylation. Administrated after the start of reperfusion, metformin reduced infarct size in normoglycemic rats, reducing subsequent rigor contracture by inhibition of mitochondrial permeability transition pore opening with a phosphoinositide-3 kinase inhibitor-mediated mechanism (404).
A. Cardiopulmonary impairment
PCOS patients have been shown to have a subclinical cardiopulmonary impairment with a reduced cardiopulmonary functional capacity that could increase the risk for CVD, even at an early age (405).
The maximal oxygen consumption (VO2max), defined as the maximal capacity of an individual to perform aerobic work, is the product of cardiac output and arterial-venous oxygen difference at exhaustion, and it is probably the best predictor of functional capacity. This parameter is positively correlated with insulin sensitivity and is considered a strong determinant of insulin sensitivity index in both men and women (406).
Several data demonstrated that an increased VO2max is associated with a decreased cardiovascular mortality, even in subjects with CVD (407). In a recent study (405), there were significant differences in VO2max and in cardiovascular parameters between young PCOS women and healthy age- and BMI-matched controls. On the other hand, a 3-month structured exercise training program improved cardiopulmonary functional capacity in overweight PCOS patients (408), probably exerting a beneficial effect on mitochondrial biogenesis and insulin sensitivity.
It is unknown whether metformin could have a similar or additive effect of exercise training in insulin-resistant subjects, as well as in PCOS women. Metformin has recently been shown to inhibit complex I of the electron transport system, and thus it could exert some effect on aerobic capacity (409). In a recent double-blind, placebo-controlled study, Braun et al. (409) demonstrated that 2000 mg/d metformin exerts a significant effect on maximal exercise capacity in healthy patients without mitochondrial dysfunction.
B. Diabetes mellitus
Women with PCOS are at risk for developing IGT and type 2 DM (272). In fact, in PCOS patients the prevalence of IGT and type 2 DM is 30–40% and 5–10%, respectively (272). In addition, the conversion from IGT to type 2 DM is increased (272). Women with PCOS and baseline NGT have a 16% conversion rate per year to type 2 DM (410), whereas by the age of 30 yr, 30–50% of obese women with PCOS develop IGT or overt type 2 DM. This is a 3- to 7-fold greater risk than an age-comparable population (272).
The mechanisms underlying the association between PCOS and glucose metabolism impairment are still unknown.
Fetal growth restraint and low birth weight and/or size for gestational age followed by catch-up of weight during infancy may lead to hyperinsulinemia and insulin resistance, obesity, PCOS, and type 2 DM in later life (411, 412). In utero androgen exposure has been also proposed to influence the phenotypic expression of PCOS. In fact, in nonhuman primates, fetal exposure to high levels of androgen during early intrauterine development is associated with defects in insulin secretion and action in adult life (413), and prenatally androgenized female rhesus monkeys exhibit glucoregulatory deficits similar to those seen in adult women with PCOS (414). Of note, the timing of the androgen exposure seems to have different effects on glucose regulation: early androgen exposure has been associated with impaired pancreatic β-cell function, whereas exposure later in gestation appears primarily to alter insulin sensitivity. However, there is evidence that insulin resistance may play a major pathophysiological role in the development of glucose intolerance in PCOS women like that which occurs in the healthy population (415).
The Diabetes Prevention Program (DPP) Research Group recently conducted a large placebo-controlled RCT sponsored by the NIH (416) on 3234 subjects in the United States at high risk for developing type 2 DM (history of gestational DM or presence of IGT and a first-degree relative with DM). Subjects were randomized to standard management, intensive lifestyle intervention, metformin, or troglitazone. This original design changed because the treatment in the troglitazone arm was discontinued after 18 months due to the emerging risk of hepatic dysfunction (416). During the average follow-up period of 2.8 yr, intensive lifestyle intervention reduced the incidence of newly developed DM by 58%, whereas subjects treated with metformin showed a RR for progression to overt type 2 DM reduced by 31% in patients, and no difference based on gender was observed in any case (416). These results indicate that improvement in insulin sensitivity, through either intensive lifestyle modification or metformin, reduces the risk of developing DM in high-risk individuals (416).
At this point, three main concerns arise: 1) whether the beneficial effects of metformin are maintained over the time or disappeared at treatment suspension; 2) whether metformin reduces only DM (as intermediate end-point) or also the cardiovascular events, which are the main cause of death in DM patients; and 3) whether results from a heterogeneous population at high risk for DM can be translated in the PCOS population.
Data obtained from the same population studied in the DPP demonstrated that, after a short washout period, about 25% of the beneficial effect of metformin on type 2 DM prevention did not persist when treatment was withdrawn (417). However, a significant reduction in the incidence of DM was detected, supporting the hypothesis that metformin does not mask the development of DM but provides a curative effect in diabetic patients (417).
Different findings were obtained in a recent RCT on a small but well-selected population of normal-weight, non-insulin-resistant PCOS patients (115). In particular, metformin did not maintain its benefits at a biochemical and clinical level after 12-month treatment suspension (115). Conversely, a slight but significant worsening of the basal peripheral insulin sensitivity was observed (115).
The UKPDS (401) reported a significantly reduced incidence of myocardial infarction in overweight patients with type 2 DM treated with metformin. In addition, another study (418) demonstrated that metformin monotherapy or metformin combined with sulfonylurea was associated with reduced all-cause and CVD mortality compared with sulfonylurea monotherapy among new users of these agents.
Unfortunately, data on CVD mortality in PCOS patients treated with metformin are lacking at this time.
Although some evidences showed that metformin administration in PCOS patients improves intermediate end-points for DM, such as insulin sensitivity indexes, there are limited data on the long-term beneficial effects of metformin on the risk for type 2 DM in women with PCOS. One retrospective study of PCOS women treated with metformin for an average of 43 months found that metformin appeared to delay or prevent the development of IGT and type 2 DM (419). These investigators found an 11-fold decrease in the annual conversion rate from NGT to IGT, with 55% of IGT patients reverting to NGT.
Salpeter et al. (396) performed a meta-analysis of the pooled results of 31 clinical trials with 4570 participants followed for 8267 patient-years to assess the effect of metformin on metabolic risk in patients at high risk for DM. Fourteen trials included PCOS subjects, for a total of 620 subjects. Metformin treatment significantly (P < 0.01) reduced fasting glucose (OR, −4.5; 95% CI, −6.0 to −3.0), fasting insulin (OR, −14.4; 95% CI, −19.9 to −8.9), and HOMA index (OR, −23; 95% CI, −18.0 to −27.3), compared with placebo or no treatment. No statistically significant differences were found between subgroups, i.e., PCOS vs. non-PCOS and obese vs. nonobese. Metformin decreased new-onset DM by 40% (95% CI, 0.5 to 0.8) and reduced the absolute risk of DM by 6% (95% CI, 4 to 8); furthermore, no data on subgroups were provided.
Finally, a recent position statement from the AEPS recommended that women with PCOS, regardless of weight, be screened for IGT or type 2 DM by an oral glucose tolerance test at their initial presentation and every 2 yr thereafter (33). However, this statement noted that the use of metformin to treat or prevent the progression of IGT could be considered but should not be mandated at this point in time because well-designed RCTs demonstrating efficacy have yet to be conducted.
C. Hypertension
Controversial data are present in literature regarding the incidence of hypertension in PCOS women (420–422).
At a young age, PCOS patients generally do not manifest increased blood pressure values (420), even if an increased prevalence of labile daytime blood pressure, which is a predisposing factor to sustained hypertension later in life, has been reported (421). On the other hand, at menopause, women with PCOS have a risk of developing hypertension that is 2.5-fold higher than age-matched controls, and this may be related in part to the obesity associated with PCOS (423).
Some authors (424) suggested that a mechanism responsible for hypertension in women with PCOS could be mediated by ET-1. Diamanti-Kandarakis et al. (424) reported that serum ET-1 levels were higher in women with PCOS regardless BMI, and a positive correlation was found between ET-1 and androgen levels. In this regard, Chen et al. (422) showed an independent relation between higher androgen levels and systolic and diastolic blood pressure in nonobese PCOS women.
Notwithstanding these findings, the exact mechanisms by which androgens could influence the blood pressure in women with PCOS remain controversial.
Experimental data (425) support a role for the renin-angiotensin system in mediating androgen-stimulated increases in blood pressure. In particular, in women with PCOS, plasma renin activity was found to be elevated compared with a nonoligomenorrheic group of age-matched women (426). In fact, another possible mechanism inducing hypertension in PCOS could be the increased serum aldosterone concentration (427). Aldosterone is a well-recognized cardiovascular risk factor with an important role in the pathophysiology of hypertension, left ventricular hypertrophy, and heart failure. In PCOS women, serum aldosterone levels, even those within the normal range, were increased in comparison with healthy women (427); increased aldosterone levels have been demonstrated to predispose to the development of hypertension.
Metformin administration could prevent structural changes that precede hypertension. In an animal model, Wang et al. (428) demonstrated that fructose-fed induced advanced glycation end-products (AGEs), well-known atherogenetic molecules, significantly correlated with a reduction in endothelial NOS and a structural remodeling of the vessel wall in the aorta and mesenteric artery in the rat, and that metformin administration reduces these structural changes.
In women with PCOS, 6 months of metformin administration reduced ET-1 levels and ameliorated insulin resistance and hyperandrogenism, without changes in body weight (424) (see Section X.G). A recent RCT (429) comparing the effects of metformin vs. antiandrogenic OCs on PCOS patients showed that metformin treatment decreased daytime ambulatory blood pressure monitoring, whereas OCs exerted the opposite effect.
In a wide multicenter trial (397), metformin reduced ambulatory blood pressure in patients with type 2 DM and the white coat hypertension in an unselected population of overweight and obese patients. Lastly, recent studies (430, 431) suggested that the combination of thiazolinediones and metformin is associated with a slight but significant improvement in the long-term blood pressure control of patients with type 2 DM.
D. Dyslipidemia
Dyslipidemia is the most common metabolic abnormality in PCOS (415). Recent findings showed that women with PCOS have quantitative LDL alterations, with increased levels of atherogenic lipoprotein phenotype, associated with a greater cardiovascular risk (432).
In particular, serum total cholesterol, low-density lipoprotein (LDL)-cholesterol, and triglycerides were significantly higher in women with PCOS compared with controls (432), and obese women with PCOS usually have higher levels of triglycerides than normal-weight women with PCOS (432). Low concentrations of high-density lipoprotein (HDL)-cholesterol were also detected in this population (432).
Metformin action on lipoproteic metabolism can be direct or through a reduction of insulin concentration. Metformin improved hepatic fatty acid metabolism from lipogenesis toward oxidation and ketone body production in rats with diet-induced overweight and hypertriglyceridemia (433). In particular, the cumulative triglyceride output decreased by more than 60% and the total ketone body output increased by more than 100% from liver perfused with metformin (433). Moreover, AMPK plays a key role in the regulation of cellular lipid metabolism increasing the rate of fatty acid oxidation (FAO) (59, 60), and metformin could act as lipid-lowering agent activating AMPK and, thus, increasing FAO (59, 60). In fact, in a model liver cell line, metformin activated AMPK, increasing immediately and for a prolonged period the rate of constitutive exocytosis (triacylglycerol and apolipoprotein B) by about 2-fold (434). This action is a mechanism independent from FAO increase because the FAO inhibition with R-etomoxir does not produce a significant reduction of the lipid-lowering effects of metformin (434).
Although controversial data are available on the effect of metformin in reducing visceral fat, different beneficial effects are reported on dyslipidemia in PCOS women (115, 123, 124, 126, 203, 401, 432). Some evidence showed that metformin in PCOS women led to a decrease in plasma total and LDL cholesterol and in triglyceride levels (123, 124, 126), as well as an increase serum HDL-cholesterol concentrations (123, 124).
No significant effect of metformin on total cholesterol (WMD = −0.10; 95% CI, −0.41 to 0.21; P = 0.5), HDL-cholesterol (WMD = 0.04; 95% CI, −0.07 to 0.16; P = 0.4), and triglycerides (WMD = −0.01; 95% CI, −0.29 to 0.27; P = 1) was detected by Lord et al. (30), whereas LDL-cholesterol was significantly reduced during treatment (WMD = −0.44; 95% CI, −0.79 to −0.08; P = 0.02).
Other investigators studying normal-weight PCOS women confirmed that 12 months of metformin administration significantly improved LDL-cholesterol levels when compared with baseline levels and the placebo group, whereas no effect on total cholesterol, HDL-cholesterol, and triglycerides was found (115). In addition, at metformin suspension, LDL-cholesterol levels quickly worsened and returned to the baseline levels (115).
In a successive meta-analysis by Costello et al. (190), the lipid patterns of PCOS patients treated with metformin and OCs were compared, and no significant difference in total cholesterol levels between treatments (WMD = −0.11; 95% CI, −0.53 to 0.30; P = 0.60) was observed. Contrarily, metformin resulted in a significantly lower triglyceride level than OCs (WMD = −0.48; 95% CI, −0.86 to −0.09; P = 0.01).
Lastly, Rizzo et al. (432) showed that metformin therapy may have a significant impact on atherogenic lipoprotein phenotype, which is considered a risk factor for PCOS women. The ISDs, and particularly metformin, have been shown to improve this metabolic alteration, even if they cannot be considered as first-line therapy (432). In fact, as Wild et al. (435) commented, the combination of metformin and thiazolinediones, although mechanistically attractive, has the potential to do harm. In addition, in the absence of evidence from well-designed RCTs on PCOS subjects, we should use the best evidence available from other populations and the best clinical judgment to extrapolate (435). Thus, it is mandatory to use optimum tools to assess the burden of disease, and medication should be reserved only for PCOS patients with remarkable lipid abnormalities including low HDL-cholesterol, high triglycerides, and the presence of excess LDL-cholesterol (435).
Data obtained from a population at high risk for type 2 DM (396) and an unselected population of overweight and obese patients (397) showed the beneficial effect of metformin on lipoprotein pattern. In particular, metformin treatment was shown to significantly increase HDL-cholesterol (OR, 5.0; 95% CI, 1.6 to 8.3; P = 0.004) and to significantly reduce LDL-cholesterol (OR, −5.6; 95% CI, −8.3 to −3.0; P < 0.0001) and triglycerides (OR, −5.3; 95% CI, −10.5 to −0.03; P = 0.04) compared with placebo or no treatment, without significant differences between subgroups of PCOS/non-PCOS and obese/nonobese patients (396). Finally, in an unselected population of overweight and obese patients, metformin reduced the incidence of dyslipidemia significantly more than diet (397).
E. Impaired fibrinolysis
In PCOS patients, a significant increase in PAI-1, a 52-kDa glycoprotein whose main role is in the inhibition of plasmin formation during plasminogen activation and fibrinolysis, has been demonstrated (248).
In several conditions, including not only PCOS (248) but also type 2 DM and arterial/thrombotic diseases (436), PAI-1 activity was related to insulin resistance and it increased, irrespective of obesity. Thus, it can be speculated that elevated PAI-1 activity increased cardiovascular morbidity seen in PCOS (437).
Recent data have, in fact, demonstrated that PAI-1 synthesis, expression in vitro, and action in vivo are up-regulated by insulin (438). Adipose tissue is a major site of PAI-1 mRNA expression, and visceral fat tissue expresses more PAI-1 than sc tissue (438).
Metformin inhibits PAI-1 production in human adipose cells in vitro (439). In fact, metformin decreases PAI-1 production (and PAI-1 mRNA) under both basal and IL-1β-stimulated conditions. In particular, the metformin effect on PAI-1 levels was dose dependent, and the PAI-1 levels were inhibited by 47.8% at 1 mm metformin and by 100% at 10 mm (439).
The role of metformin in reducing the PAI-1 levels has been demonstrated in patients with type 2 DM (440).
Several authors (244, 247) also reported a significant reduction in PAI-1 activity after metformin administration in PCOS patients, hypothesizing beneficial effects on pregnancy outcomes. However, no clinical data are available regarding the influence of metformin-related PAI-1 reduction on cardiovascular risk in PCOS patients.
Finally, a significant reduction in circulating acylation-stimulating protein and in complement 3 was observed after metformin administration in 20 PCOS patients (441), suggesting further potential mechanisms by which metformin could act on impaired fibrinolysis in PCOS patients.
F. Chronic inflammation
An abnormal pattern of several markers of the chronic inflammation has been demonstrated in women with PCOS (442, 443).
Diamanti-Kandarakis et al. (442) evaluated the serum levels of cellular adhesion molecules, which reflect the degree of low-grade chronic inflammation and have been associated with several insulin-resistant states. In particular, women with PCOS had levels of high sensitivity C-reactive protein (CRP), soluble vascular cell adhesion molecule-1, and soluble endothelial leukocyte adhesion molecule-1 significantly higher in comparison to controls (442).
Metformin has been shown to exert a beneficial effect on inflammation markers with mediated and/or direct actions. It has been suggested that metformin acts on inflammatory markers by reducing insulin resistance (444). On the other hand, a recent double-masked, double-dummy study demonstrated that metformin probably also exerts a direct action on the inflammatory pattern because a similar improvement in insulin sensitivity in patients with type 2 DM with the use of metformin or the insulin secretagogue repaglinide was observed, even if the best benefits were observed under metformin (445). In fact, metformin can exert a direct vascular antiinflammatory effect by dose-dependently inhibiting IL-1β-induced release of the proinflammatory cytokines IL-6 and IL-8 in endothelial cells, human vascular smooth muscle cells, and macrophages. Investigation of potential signaling pathways demonstrated that metformin diminished IL-1β-induced activation and nuclear translocation of nuclear factor-κB in smooth muscle cells. Furthermore, metformin suppressed IL-1β-induced activation of the proinflammatory phosphokinases Akt, p38, and Erk but did not affect phosphoinositide-3 kinase activity (446).
In addition, it is well known that mitochondria have an important role in intracellular energy generation, and recent data have demonstrated their function in modulating cellular activation and tissue injury associated with acute inflammatory processes with particular regard for neutrophil activation. Metformin inhibits mitochondrial complex I, suggesting its effectiveness for the prevention or treatment of acute inflammatory processes in which activated neutrophils have a crucial role (447).
Lastly, it was observed that plasma migration inhibitor factor (MIF) concentrations and MIF mRNA expression in the mononuclear cell are elevated in the obese, consistent with a proinflammatory state in obesity (447). In this regard, metformin suppresses plasma MIF concentrations in the obese, suggestive of an antiinflammatory effect of this drug; thus, metformin may contribute to a potential antiatherogenic effect, which may have implications for the reduced cardiovascular mortality observed with metformin therapy in type 2 DM (448).
Clinical data confirmed the effects of metformin on chronic inflammatory markers observed in an experimental setting.
Metformin administration improved CRP and leukocyte count (442, 444) as well as serum high sensitivity CRP and soluble vascular cell adhesion molecule levels in PCOS patients (442).
Some reports of Ibanez et al. (449, 450) clearly demonstrated that a high leukocyte count was present in girls with hyperinsulinemic hyperandrogenism, that metformin therapy decreased leukocyte count and neutrophil to lymphocyte ratio, and also that a combination of metformin plus an oral or transdermal contraceptive could improve several markers of low-grade chronic inflammation (450). In agreement with the Spanish data, Italian data (444) confirmed the beneficial effect of metformin on the low-grade chronic inflammation, specifically on leukocyte counts and CRP levels, in adult PCOS women. The decrease in serum CRP levels during metformin therapy is in agreement with the known beneficial metabolic effects of this drug, and it suggests that CRP or other inflammatory parameters could be used as markers for the efficiency of therapy in PCOS. The reduction in circulating levels of CRP with metformin therapy has also been detected in obese women with PCOS (451). Moreover, when metformin is administered in combination with OCs containing E2 plus CA, the reduction seen in CRP was attenuated (451).
G. Endothelial impairment
Recent data support the hypothesis that PCOS is a condition associated with an increased vascular risk (452).
Our previous experience demonstrated that young women of normal weight with PCOS , without dyslipidemia and hypertension, when compared with age- and BMI-matched healthy subjects, had altered endothelial function, assessed by flow-mediated dilation of brachial arteries, and increased intima-media thickness at carotid artery and serum ET-1 values, suggesting early functional, structural, and biochemical preatherosclerotic vascular impairment (313).
Asymmetrical dimethylarginine (ADMA), a guanidino-substituted analog of l-arginine, is a potent endogenous competitive inhibitor of the endothelial NOS and has been associated with atherosclerosis, representing an independent marker for cardiovascular morbidity and mortality (453). Recent data have demonstrated that ADMA is significantly elevated in women with PCOS and that the degree of insulin resistance confers the greatest influence on ADMA level (453).
The effect of metformin on endothelium in PCOS patients has been investigated. The rationale for supposing a beneficial effect of metformin on the PCOS-related dysfunction consists of its indirect effects on insulin sensitivity and hormonal pattern and a direct action on cell proliferation.
Lund et al. (445) demonstrated, even if partially, a direct action of metformin on endothelium, whereas Romualdi et al. (454) showed a positive effect of metformin on the altered vascular reactivity in normoinsulinemic normal-weight PCOS subjects, suggesting that the improvement might be mediated through hormonal changes, i.e., improved hyperandrogenemia (454). On the other hand, in vitro studies demonstrated that metformin inhibited the proliferation of HeLa or KB cells (455) and the drug may reduce atherosclerosis through its ability to inhibit tyrosine kinase proliferation of smooth muscle cells (456).
A recent study (457) was designed to evaluate the effect of metformin on AGEs. This study involved 22 patients with PCOS who undertook metformin therapy for 6 months. At the end of the study, a significant reduction in AGE levels was detected. Several data (251, 458) evaluated the effects of metformin on subclinical vascular disorders as surrogate end-points of atherosclerosis underlying CVD. A clinical study demonstrated that metformin improved coronary microvascular function and coronary flow reserve in 16 insulin-resistant PCOS patients (458). Moreover, metaformin administration improved endothelial structure and function in PCOS patients (251).
In particular, a 6-month metformin administration in young, normal-weight women with PCOS significantly reduced the brachial artery diameter, flow-mediated dilation, diameter after reactive hyperemia, and intima-media thickness (251). The significant difference in insulin resistance observed after only 6 months of metformin suggests an important role of insulin excess in the precocious development of atherosclerosis in PCOS patients. In fact, insulin could promote atherogenesis by direct action on the arterial wall.
Finally, notwithstanding the generic notion of the beneficial effect of metformin on endothelium in PCOS patients provided by these previous sparse data, future studies are needed to establish the clinical relevance of these findings.
H. Syndrome X
The National Cholesterol Education Program Adult Treatment Panel (NCEPATP) defined the metabolic syndrome as the presence of three of the five following factors: waist circumference greater than 88 cm in females; fasting serum glucose 110 mg/dl or more; fasting serum triglycerides greater than 150 mg/dl; serum HDL-cholesterol less than 50 mg/dl; and blood pressure greater than 130/85 mm Hg (459). This clustering of abnormalities was also named dysmetabolic syndrome or, in the female, syndrome X. The NCEPATP III guidelines recognized syndrome X as a major cardiac risk factor and a specific ICD-9 code was assigned to the syndrome.
PCOS patients appear to have features of syndrome X. Syndrome X has been reported to occur at an increased overall prevalence rate of 43–47% in women with PCOS (12) compared with the 24% prevalence rate in U.S. women (460). In addition, some authors suggested that in women with PCOS, an albumin to creatinine ratio greater than 7 mg/g is strongly associated with the metabolic syndrome, high blood pressure, and elevated alanine aminotransferase levels (461).
As detailed in the previous sections, metformin administration in patients with PCOS could theoretically result in an improvement of the single features of the metabolic syndrome and a subsequent decreased risk for developing the syndrome.
Obesity, type 2 DM, and metabolic syndrome are all disorders of energy balance, which the AMPK regulates at both the cellular and whole body levels. As previously stated, AMPK switches cells from an anabolic state where nutrients are taken up and stored to a catabolic state where they are oxidized. Based on these considerations, AMPK activation due to metformin might be a potential mechanism for improving features of metabolic syndrome (462).
To date, no clinical data are available in literature regarding the use of metformin in PCOS patients as a treatment for metabolic syndrome or as a preventive measure for reducing the risk to develop metabolic syndrome. Thus, it is not currently possible to recommend metformin administration in PCOS patients to prevent the risk and the progression of metabolic syndrome. Thus, lifestyle modifications should always be considered the main approach for these patients.
XI. Other Organ Impairment
A. Liver impairment
Nonalcoholic fatty liver disease (NAFLD) is a spectrum of diseases ranging from simple steatosis to steatohepatitis and cirrhosis. NAFLD is closely associated with insulin resistance and can be a hepatic manifestation of the metabolic syndrome (463). Although insulin resistance and compensatory hyperinsulinemia are frequently found in obese patients with NAFLD, both are also noted in NAFLD patients without obesity and with NGT (463).
The incidence of NAFLD is significantly higher in PCOS populations, affecting up to 50% of patients, and these patients have greater insulin resistance and are more likely to be obese and have other features of the metabolic syndrome than PCOS patients without NAFLD (464–467).
Published data on the coexistence of PCOS and NAFLD are limited to only four retrospective studies (464–467), and it is unclear whether this clinical entity may be the result of the PCOS-related insulin resistance or whether NAFLD per se may contribute to insulin resistance frequently seen in patients with PCOS.
Recent findings showed that reduced plasma high molecular weight adiponectin levels, as observed in PCOS patients, are closely associated with the severity of NAFLD (464–467). Normal high molecular weight adiponectin levels act to inhibit hepatic stellate cells (HSCs), the key cells promoting liver fibrosis, by AMPK activation (468). Metformin might act indirectly increasing the adiponectin levels (469) but also with a direct effect of AMPK (468). In fact, metformin-mediated activation of the AMPK seems to prevent hepatic fibrosis, inhibiting HSC proliferation via suppression of reactive oxygen species production and subsequent inhibition of the AKT pathway (468). In addition, metformin-mediated activation of AMPK negatively modulates the platelet-derived growth factor-stimulated proliferation and migration of human HSCs, reducing also the secretion of monocyte chemoattractant protein-1 (470).
Clinical data on metformin administration demonstrated that 500 mg of the drug given three times daily resulted in a significant improvement in serum alanine transaminase levels in patients with NAFLD not responding to lifestyle interventions (471). Metformin therapy was also associated with a reduction in surrogate liver fat markers in obese women with PCOS (472). A significant benefit of 12 months of metformin plus N-acetylcysteine was detected in 20 patients with biopsy-proven NAFLD in terms of liver steatosis and fibrosis (473). Finally, ISDs can also lead to improvement in histological abnormalities of nonalcoholic steatohepatitis (474).
In conclusion, insulin resistance is a major feature of NAFLD that, in some patients, can progress to steatohepatitis. Treatments aimed at reducing insulin resistance have had some success, but larger placebo-controlled studies are needed to fully establish the efficacy of these interventions in reducing the deleterious effects of fat accumulation in the liver.
B. Thyroid impairment
Few data (475, 476) have shown a potential alteration of the thyroid function in PCOS patients.
Retrospective (475) and prospective (476) data suggested that metformin might modify thyroid hormone economy, causing suppression of TSH to subnormal levels. In particular, a recent study (476) on a very small sample of eight obese patients with DM and primary hypothyroidism under T4 replacement treatment demonstrated that 3-month metformin (1700 mg daily) administration reduced significantly the plasma TSH levels and increased those of free T4 (476). Furthermore, no data regarding metformin effects on thyroid impairment in PCOS population are available in literature.
C. Cognitive function
Presently, no published studies specifically evaluated cognitive function in PCOS.
Hyperandrogenism and hyperestrogenism have been hypothesized to differentially influence cognitive function across cognitive domains in the PCOS population. A recent large internet-based study (477) compared neuropsychological functioning in right-handed women with and without PCOS, and it demonstrated that PCOS is not associated with masculinized cognitive functioning and that the impaired performance on tasks is subtle and unlikely to affect daily functioning. A successive study by Schattmann et al. (478) showed that PCOS women had significantly worse performance on tests of verbal fluency, verbal memory, manual dexterity, and visuospatial working memory than the healthy control women.
The same authors (479) showed that 3 months of treatment either with anti-androgen CA plus estrogen or with placebo did not have a significant impact on cognitive performance in women with PCOS, although reductions in free testosterone may be beneficial for verbal fluency. However, no experimental or clinical data on metformin administration on this issue are actually available in the literature.
D. Sleep disturbances
PCOS is associated with poor sleep quality, excessive daytime sleepiness (EDS), and increased risk for obstructive sleep apnea (OSA) (480, 481).
To date, no data are available in literature on the potential role of metformin treatment in PCOS patients with sleep disturbances. Moreover, its potential role is suggestible. In fact, insulin levels and measures of glucose tolerance in PCOS are strongly correlated with the risk and severity of OSA. On the other hand, several data seem to indicate that the proinflammatory cytokines IL-6 and TNF-α were elevated, independently by obesity, in patients with disorders of EDS and that these cytokines could be the mediators of daytime sleepiness. Metformin could act on EDS (or OSA), reducing systemic inflammation (see Section X.F).
In obese patients, EDS and OSA may be manifestations of syndrome X, and an improvement of features of this syndrome by metformin treatment might also result in an amelioration of the sleep disturbances.
XII. High-Risk Patients for PCOS
In the human species, prenatal growth restriction was demonstrated in a high percentage of cases to be followed by a postnatal catch-up of growth related to an exaggerated fat gain, particularly visceral fat (411). Girls with growth restriction and precocious pubarche are recognized to be at risk for development of hyperinsulinemic androgen excess, leading up to PCOS (411).
Some authors proposed to use metformin in post-precocious pubarche adolescents (116, 179) and in postmenarcheal low birth weight (482) as a treatment to prevent progression to PCOS.
In particular, 6 months of metformin treatment in girls with precocious pubarche was demonstrated to induce less adipose body composition and proinflammatory state, to improve lipid profile, and to attenuate adrenarche (116).
Early metformin therapy in precocious pubarche girls prevented the progression to PCOS (116, 179). Successive studies (482, 483) confirmed that prepubertal metformin therapy has normalizing effects on PCOS features in high-risk girls with a combined history of low birth weight and precocious pubarche, but its beneficial effects are reversed as soon as metformin therapy is discontinued (483). However, in girls with similar characteristics, long-term metformin treatment reduced total and visceral fat and delayed menarche without attenuating linear growth (482, 483).
XIII. Future Perspectives
Notwithstanding so much clinical data on the uses of metformin in diabetic and PCOS patients, the specific mechanisms of metformin action still must be completely elucidated, and only when these mechanisms are clarified should it be possible to understand all potential uses of this drug. In particular, recent data suggested a pleiotropic effect of the drug across its action on AMPK, even if further investigations are necessary to elucidate the function of metformin-related activation of the AMPK in reproduction and tumorogenesis.
From a clinical point of view, the crucial issue for the next few years will be to define predictive factors, including possibly BMI and/or measures of insulin resistance, for a good response to metformin to optimize the treatment(s) and define the best protocol(s), especially in terms of timing, and the most effective dosages for different PCOS subpopulations and for different end-points. In this regard, there is also the potential for pharmacogenetic identification of women likely to respond to metformin.
These problems could probably be solved by using well-designed RCTs focused on select phenotypes of the PCOS population, instead of by conducting large-scale clinical trials on unselected patients, and successive wide clinical trials should be designed for evaluating the usefulness of metformin not only as monotherapy or as a drug to be associated with other treatments but also as a therapeutic step in an integrated strategy. In this regard, a wide multicenter Italian RCT comparing different strategies for metformin use is in the recruiting phase, and the data will not be available before 2010 (484). Finally, a well-validated and clinically applicable prediction model should be constructed to select the patients with the highest success chances, making the best patient-tailored therapeutic decision.
An interesting field of research will be the evaluation of the effects of metformin on endometrium. The future studies should establish whether metformin improves endometrial biology and histology, reducing miscarriages and premalignancy endometrial diseases. Certainly, to obtain significant data, a highly selected population must be studied.
Doubts existed until now regarding the use of metformin during pregnancy. Available data seem to be reassuring with regard to metformin administration in diabetic patients. Furthermore, well-designed RCTs are advocated before metformin would be safely prescribed in pregnant patients affected by PCOS as a preventive drug for reducing (potential) maternal and neonatal complications.
Even if metformin is not a cosmetic or anti-obesity drug, further research should clarify whether it has a role in the management of PCOS-related hirsutism and obesity and, specifically, its potential usefulness for optimizing the first-line treatments (e.g., antiandrogens and lifestyle modifications, respectively).
More and more cellular, animal, and human data have demonstrated the beneficial effects of metformin on several markers of CVD. Unfortunately, no wide clinical trial has demonstrated any preventive effect of metformin in PCOS patients. Thus, the long-term effect of metformin administration on morbidity and mortality in PCOS patients at high risk for metabolic derangements should be also evaluated in the future.
Finally, further studies on the clinical and metabolic effects of metformin suspension are also needed before the use of this drug become common in clinical practice as treatment of the PCOS features and, specifically, as a preventive measure for young girls at risk for developing PCOS.
XIV. Summary
In the clinical practice, metformin administration is less useful than OCs to regulate the menstrual cycle, but it should be considered as initial intervention in (overweight or obese) PCOS patients in whom OCs are contraindicated or with an initial metabolic derangement. In fact, clinical evidence shows that this drug is an effective first treatment for restoring ovulatory menstrual cycles in oligoamenorrheic PCOS patients, improving insulin sensitivity and, unlike OCs, lipoprotein pattern. On the contrary, OCs might worsen insulin sensitivity, and prospective long-term trials are needed to determine whether any of these changes translate into differences in clinical outcomes between women treated with metformin and those treated with OCs.
In PCOS patients with anovulatory infertility who were not previously treated, the administration of metformin plus CC is not better than monotherapy (metformin alone or CC alone); thus, the combined approach should be avoided as an initial treatment of anovulatory infertility. To date, no specific recommendation can be given regarding the use of CC or metformin because the best first-step option to induce ovulation in infertile anovulatory PCOS patients remains unclear and probably depends on the wishes of the patient. If time is of the essence, CC will work more quickly but could result in multiparity. If the woman is willing to wait for a few months, metformin could be equally effective. Metformin could also be a valid option in those PCOS patients who absolutely wish to avoid multiple gestations and/or in patients who do not tolerate CC (i.e., secondary to mood changes, visual disturbances, etc.). On the other hand, PCOS patients who have failed to ovulate with CC (i.e., CC-resistant) may benefit from the addition of metformin. It is possible to hypothesize that metformin therapy would augment the induction of ovulation in CC-resistant women because of its favorable change in androgens, gonadotropins, and insulin through mechanisms distinct from those of CC. It is plausible to assume that women with CC resistance receiving metformin have an increased response to CC secondary to an improved follicle steroidogenesis caused by the effect of metformin administration.
Certainly, metformin monotherapy seems to be more effective and cheaper than LOD for treating anovulatory infertile CC-resistant PCOS patients, and LOD should be reserved for inducing ovulation only in anovulatory PCOS patients with other suspected and/or diagnosed factors of subfertility (e.g., endometriosis, uterine leiomyomas). Currently, data on the efficacy of metformin pretreatment before CC in CC-resistant patients are controversial and inconclusive. In patients who received gonadotropins as treatment for anovulation, metformin addition increases the rate of monoovulations, reducing the risk of cancelled cycles, reduced the duration of gonadotropins administrations and the doses of gonadotropins required whereas in infertile PCOS patients scheduled for IVF cycles, metformin cotreatment reduces the OHSS risk and, thus, its use should be planned for patients at high-risk for OHSS.
Beneficial effects of metformin in reducing the rate of miscarriage in PCOS women have been widely suggested, but to date no advantage has been demonstrated when metformin is administered before pregnancy. On the other hand, considering the lack of RCTs it is not possible to either suggest or advise against the use of metformin during pregnancy for reducing the abortion risk. More relevant clinical data are needed regarding the potential benefits of metformin administration during pregnancy on miscarriage prevention.
Observational and anecdotal findings suggest that metformin may be beneficial in pregnancy by potentially reducing the risk of gestational DM or other insulin resistance-associated pregnancy complications, e.g., PIH and PE. Furthermore, at present the routine use of metformin during pregnancy in PCOS patients to prevent these morbidities is not recommended. In fact, well-powered RCTs are needed before general use of metformin during pregnancy in PCOS patients is advocated. Currently, there is no evidence suggesting a positive or negative effect of metformin administration on infant outcomes during both pregnancy and lactation. Furthermore, while waiting for definitive data regarding the continuation of metformin during pregnancy for women with documented insulin resistance to become available, the cessation of metformin after a positive pregnancy test is still a reasonable course of action.
Preliminary clinical and experimental data showed that metformin could exert beneficial effects on endometrium, suggesting a potential and future role for the prevention and treatment of endometrial degeneration due to PCOS-related chronic anovulation. Currently, it is not possible to draw definitive conclusions on these effects, and further research on a large sample and with long-term follow-up is needed to confirm these preliminary positive results. In addition, studies on the potential beneficial effects of metformin in reducing the long-term consequences of chronic anovulation are totally lacking.
The clinical effectiveness of metformin on hirsutism seems to be limited and inferior to that reported with OCs, which is also low. In hirsute PCOS patients, metformin could be coadministered with antiandrogens, i.e., flutamide, and OCs to reduce the potentially increased metabolic and cardiovascular risks related to OC use.
Controversial data are available on the effects of metformin on weight loss. Certainly, metformin should not be considered a treatment for weight loss, even in obese PCOS patients. However, metformin administration in PCOS patients who follow lifestyle modification programs seem related to the best benefits in terms of weight loss and metabolic pattern improvement.
In addition, the role of metformin as a preventive measure for PCOS development in girls at high risk for developing PCOS is strongly limited because the available data are still scarce and long-term effects on these specific populations are totally lacking.
Finally, metformin seems to be effective in ameliorating several intermediate markers of CVD in PCOS women during its administration. In particular, benefits may be seen in atherogenic profiles, including markers of subclinical inflammation, dyslipidemia, and insulin resistance. Enhanced endothelial function, coronary microvascular function, and coronary flow rate may also be seen. Notwithstanding these improvements in secondary CVD markers under metformin therapy in PCOS women, conclusive and prospective long-term studies have yet to be carried out, and the clinical efficacy of metformin in the prevention of the development of type 2 DM or in reducing the cardiovascular accidents in PCOS women is to be still proven.
Acknowledgment
Author Disclosure Statement: S.P., A.F., F.Z., and F.O. have nothing to declare.
Abbreviations:
- ADMA,
Asymmetrical dimethylarginine;
- AGE,
advanced glycation end-product;
- AMPK,
5′-AMP-activated protein kinase;
- AR,
androgen receptor;
- AUC,
area under the curve;
- BMI,
body mass index;
- CA,
cyproterone acetate;
- CC,
clomiphene citrate;
- CI,
confidence interval;
- CRP,
C-reactive protein;
- CVD,
cardiovascular disease;
- DM,
diabetes mellitus;
- E2,
estradiol;
- EDS,
excessive daytime sleepiness;
- EE2,
ethinyl E2;
- ERα,
estrogen receptor α;
- ET-1,
endothelin-I;
- FAO,
fatty acid oxidation;
- GFR,
glomerular filtration rate;
- GLUT,
glucose transporter;
- HDL,
high-density lipoprotein;
- HOMA,
homeostasis model of assessment;
- HSC,
hepatic stellate cell;
- IGFBP,
IGF binding protein;
- IGT,
impaired glucose tolerance;
- ISD,
insulin-sensitizing drug;
- IVF,
in vitro fertilization;
- LDL,
low-density lipoprotein;
- LOD,
laparoscopic ovarian diathermy;
- MIF,
migration inhibitor factor;
- mTORC,
mammalian target of rapamycin;
- NAFLD,
nonalcoholic fatty liver disease;
- NGT,
normal glucose tolerance;
- NOS,
nitric oxide synthase;
- OC,
oral contraceptive;
- OCT,
organic cation transporter;
- OHSS,
ovarian hyperstimulation syndrome;
- OR,
odds ratio;
- OSA,
obstructive sleep apnea;
- PAI,
plasminogen activator inhibitor;
- PCO,
polycystic ovaries;
- PCOS,
polycystic ovary syndrome;
- PE,
preeclampsia;
- PI,
pulsatility index;
- PIH,
pregnancy-induced hypertension;
- QoL,
quality of life;
- RCT,
randomized controlled trial;
- RI,
resistance index;
- RR,
relative risk;
- VO2max,
maximal oxygen consumption;
- WHR,
waist-hip ratio;
- WMD,
weighted mean duration.
Rotterdam ESHRE/ASRM-Sponsored PCOS Consensus Workshop Group 2004 Revised
Polycystic Ovary Syndrome Writing Committee
Thessaloniki ESHRE/ASRM-Sponsored PCOS Consensus Workshop Group
la
Dell'aglio DM, Perino LJ, Todino JD, Algren DA, Morgan BW 17 March 2008 Metformin overdose with a resultant serum pH of 6.59: survival without sequalae. J Emerg Med 10.1016/j.jemermed. 2007.09.034
World Health Organization
van
Metformin in pregnant PCOS women. ClinicalTrials.gov Identifier: NCT00159536
Uterine artery blood flow in pregnant women with polycystic ovary syndrome treated with metformin. ClinicalTrials.gov Identifier: NCT00466622
Kjøtrød SB, Sunde A, Düring VV, Carlsen SM 4 March 2008 Possible metformin effect on adrenal androgens during pretreatment and IVF cycle in women with polycystic ovary syndrome. Fertil Steril 10.1016/j.fertnstert.2007.11.069
van
National Institute for Clinical Excellence
UK Prospective Diabetes Study (UKPDS) Group
de
Diabetes Prevention Program Research Group
Luque-Ramírez M, Mendieta-Azcona C, Alvarez-Blasco F, Escobar- Morreale HF 18 June 2008 Effects of metformin versus ethinyl-estradiol plus cyproterone acetate on ambulatory blood pressure monitoring and carotid intima media thickness in women with the polycystic ovary syndrome. Fertil Steril 10.1016/j.fertnstert. 2008.03.082
Third Report of the National Cholesterol Education Program (NCEP)
de
Strategies for ovulation induction in anovulatory infertile patients with PCOS. ClinicalTrials.gov Identifier: NCT00461643