Effects of low‐dose metformin in Japanese women with clomiphene‐resistant polycystic ovary syndrome

Abstract

Background and Aims:  The aim of the present prospective observational study was to evaluate the effects of low‐dose, short‐term metformin, in combination with clomiphene (CC), in CC‐resistant infertile Japanese women with polycystic ovary syndrome (PCOS).

Methods:  Metformin therapy was administered orally (one 250 mg tablet, twice daily) to 15 CC‐resistant infertile patients with PCOS, beginning on the third day of progestin‐induced withdrawal bleeding, and was continued for 14 days in the first cycle. In the event of anovulation, 100 mg/day of CC was given during subsequent cycles on days 5–9, in addition to the aforementioned dose of metformin. Hormonal and metabolic parameters were measured on the second or third days of the first cycle and also the fourth cycle, following an overnight fast.

Results:  None of the 15 women successfully ovulated during the first cycle with metformin treatment alone. After two subsequent cycles with the combination of CC and metformin, ovulation was confirmed in 17 of 29 cycles (61%) and in 13 of 15 patients (87%). Two women became pregnant within 2 months of therapy (13%). There were no cases of ovarian hyperstimulation syndrome. Following three cycles of metformin therapy, a slight reduction in serum levels of luteinizing hormone (LH), free testosterone, androstenedione, dehydroepiandrosterone sulfate, hemoglobin A1c and total cholesterol was seen, while serum LH/follicle‐stimulating hormone ratio and serum level of low‐density lipoprotein cholesterol were significantly decreased. Although there were no significant differences between the responder (n = 11) and non‐responder (n = 2) groups at baseline, the levels of plasma fasting insulin was significantly higher and fasting glucose/insulin ratio was significantly lower in the non‐responder group compared with the responder group after three cycles.

Conclusion:  Low‐dose, short‐term metformin, combined with CC, can improve ovulation rates in CC‐resistant infertile Japanese women with PCOS. (Reprod Med Biol 2004; 3: 19–26)

INTRODUCTION

POLYCYSTIC OVARY SYNDROME (PCOS) is a common endocrine disorder characterized by anovulation, infertility, obesity and hyperandrogenism in women of reproducting age, affecting 5–10% of premenopausal women. ¹ Insulin resistance and compensatory hyper‐insulinemia are prominent features of PCOS. ² High insulin concentrations cause hyperandrogenism because of increased production of ovarian androgen and decreased synthesis of sex hormone‐binding globulin (SHBG).

Weight loss in obese women with PCOS leads to a reduction of hyperinsulinemia ³ and recently insulin‐sensitizing agents have been tried in the management of PCOS. ⁴ , ⁵ The most extensively studied insulin‐sensitizing drug in the treatment of PCOS is metformin. ⁶ , ⁷ Metformin is an orally administered drug, routinely used to lower blood glucose concentrations in patients with diabetes mellitus type 2. Metformin enhances insulin sensitivity both in the liver, where it inhibits hepatic glucose production, and peripherally, where it increases glucose uptake and use by muscle tissue. By increasing insulin sensitivity, metformin reduces insulin secretion and consequently hyperinsulinemia. By reducing insulin levels, metformin also reduces the activity of cytochrome P450C‐17α in both obese and lean women with PCOS. ⁸ This leads to a decrease in intraovarian and plasma levels of androgen and stimulates hepatic SHBG synthesis, which in turn favors a reduction in estradiol levels, allowing orderly follicular growth. Although some studies have shown significant improvements in insulin sensitivity and hyperinsulinemia in obese women with PCOS following treatment with metformin, ⁴ others failed to show any benefit in terms of insulin concentrations or hormonal and metabolic profiles. ⁹ , ¹⁰

Clomiphene citrate (CC) is the usual first‐line drug for inducing ovulation in women with PCOS. Successful ovulation is achieved in 70–85% of women and 40–50% will conceive. ¹¹ Women with PCOS who remain anovulatory following treatment with CC show greater insulin resistance than those who successfully respond. ¹² Gonadotrophin treatment can be offered to those women who fail to respond to CC, although the use of gonadotrophin is more expensive and associated with a higher risk of multiple pregnancy and of developing ovarian hyperstimulation syndrome (OHSS). At present, there are very few studies evaluating the use of metformin in women with CC‐resistant PCOS. The drug information guidelines of Japan recommend a starting dose of metformin for patients with diabetes mellitus type 2 of 250 mg twice daily and a maximum dose of 250 mg thrice daily.

The aims of the current prospective observational uncontrolled study were: (i) to determine the rates of ovulation and pregnancy; and (ii) to explore the differences in the endocrine and metabolic profiles between responders and non‐responders, among Japanese women with CC‐resistant PCOS, following low‐dose, short‐term metformin treatment for 3 months.

MATERIALS AND METHODS

Patients

A TOTAL OF 15 infertile patients with a diagnosis of PCOS and a documented history of resistance to CC were recruited for the study. Polycystic ovary syndrome was diagnosed when the following three criteria were fulfilled: (i) the presence of chronic ovulatory disorders such as oligomenorrhea, anovulatory cycles, or secondary amenorrhea; (ii) endocrinologic disorders, such as elevated luteinizing hormone (LH) and/or the presence of hyperandrogenism (elevated testosterone, androstenedione, or dehydroepiandrosterone sulfate (DHEAS) levels); and (iii) the presence of polycystic ovaries on transvaginal ultrasound examination. Clomiphene citrate resistance was defined as the failure to have an ovarian response for two consecutive cycles on transvaginal ultrasonographic examination with concomitant failure of follicle maturation (>10 mm), following 5 days of treatment with CC, at a dose of 100 mg daily. Adrenal and thyroid disorders, as well as hyperprolactinemia, were specifically excluded by means of hormone assays. Male‐related infertility was excluded with semen analysis and a hysterosalpingogram was used to exclude tubal causes of infertility. Patients who had taken any medications in the 3 months prior to the study that might influence carbohydrate metabolism were excluded. Diabetics were excluded using a fasting glucose cut‐off level of 126 mg/dL and a 2 h glucose level of 200 mg/dL following a 75 g oral glucose tolerance test (OGTT). The patients with impaired glucose were examined by an endocrinologist. Each patient gave written informed consent prior to participating in the study, which was approved by the Ethics Committee, Faculty of Medicine, Niigata University.

Study protocol

Metformin (one 250 mg tablet, twice daily; Nihonshinyaku, Tokyo, Japan) was administered orally beginning on the third day of progestin‐induced withdrawal bleeding, and was continued for 14 days. On the 12th and 16th day of each cycle, measurement of follicular size by transvaginal ultrasound examination was performed. For follicles >18 mm in diameter, 5000 U of human chorionic gonadotropin (Mochida, Tokyo, Japan) was administered intramuscularly, and metformin therapy was discontinued until the third day of the next cycle. In this case, serum progesterone was measured after the follicle had ruptured and ovulation was deemed successful if the progesterone levels were >5 ng/mL. If no mature follicle was observed by the 25th day of each cycle, withdrawal bleeding was induced with a progestin. In the anovulatory cases, 100 mg/day of CC (Shionogi, Tokyo, Japan) was added on cycle days 5–9 in the subsequent second and third cycles. Hormonal and metabolic parameters were measured on the second or third days of the first cycle and also the fourth cycle, following an overnight fast.

Assays

All blood samples were centrifuged and stored at −20°C until assayed. Serum levels of LH, FSH and plasma levels of insulin were measured with enzyme immunoassays. Serum levels of estrone, free testosterone, androstenedione, DHEAS, 17‐α hydroxyprogesterone, SHBG and leptin were measured with radioimmunoassays. Serum levels of insulin‐like growth factor‐I were measured using an immunoradiometric assay, while plasma levels of glucose and serum levels of total cholesterol, triglyceride, high‐density lipoprotein cholesterol (HDL‐C), low‐density lipoprotein cholesterol (LDL‐C) and free‐fatty acid were measured using an enzymatic assay. The insulinemic and glycemic responses to the 75 g OGTT were expressed as the area under the curve (AUC), calculated 2 h after glucose ingestion by the trapezoidal rule using absolute values. ¹² The homeostasis model assessment for insulin resistance (HOMA‐IR) was calculated by the formula:

HOMA‐IR = Fasting blood sugar (mg/dL) × fasting insulin (µU/mL)/405. ¹³

The ratio of fasting glucose to insulin, which is a useful marker for insulin resistance, was also calculated. ¹⁴ Side‐effects such as nausea, vomiting and diarrhea were recorded at each monthly outpatient visit.

Statistical analysis

Data from the baseline and from the conclusion of the study after three cycles were compared using Wilcoxon's signed rank test. Mann–Whitney's U‐test was used to compare results between the responders and non‐responders. P‐values < 0.05 were considered statistically significant. Data are reported as the mean ± SD.

RESULTS

TABLE 1 SHOWS THE ENDOCRINE and metabolic characteristics of the 15 study participants. There were six patients with primary infertility and nine patients with secondary infertility. Their mean age was 30.6 ± 3.8 years (range: 24–36 years) and their mean body mass index (BMI) was 26.2 ± 5.8 (range: 16.8–37.7 kg/m²). Nine women were obese, that is BMI ≥25, which is the definition of obese in Japanese women. The mean serum free testosterone was 1.6 ± 0.9 pg/mL, DHEAS was 2160 ± 895 ng/mL and HOMA‐IR was 3.1 ± 2.0. Six women (40%) showed impaired glucose tolerance (IGT) following the OGTT.

Table 1.

 The endocrine and metabolic characteristics and ovulation rate of 15 study participants

Number	Age (years)	Gravida	BMI	LH/FSH	Free‐T	DHEAS	HOMA‐IR	F‐G/I	75 g OGTT	Number of mature follicles at			Ovulation rate
										1	2	3
1	26	1‐0	25.3	1.5	1.4	2100	3.6	5.6	N	×	1	1	2/3
2	25	2‐0	26.5	3.0	0.8	1670	3.3	6.8	N	×	1	1	2/3
3	36	1?1	30.8	0.7	1.4	3928	2.0	10.1	N	×	1	1	2/3
4	24	1‐0	27.0	2.4	1.1	2750	7.4	2.5	N	×	1	1	2/3
5	32	0‐0	37.7	1.0	3.3	3730	3.5	7.2	IGT	×	1	×	1/3
6	28	0‐0	19.4	1.2	1.6	2230	1.1	17.0	IGT	×	1	×	1/3
7	33	0‐0	16.8	1.6	1.3	1380	0.8	20.0	IGT	×	1	×	1/3
8	33	0‐0	19.5	1.3	1	2350	0.4	44.5	N	×	1	×	1/3
9	29	1?1	21.3	1.0	1.9	2540	1.2	13.2	N	×	×	2	1/3
10	34	0‐0	22.7	1.3	0.5	598	3.0	5.5	N	×	×	1	1/3
11	29	0‐0	32.5	1.1	1.5	1322	6.3	4.1	IGT	×	×	1	1/3
12	34	1?1	30.8	2.5	3.7	1670	3.8	6.0	IGT	×	×	×	0/3
13	31	2‐0	28.3	1.3	1.4	1960	3.8	6.9	IGT	×	×	×	0/3
14	34	2?1	28.6	0.9	1.8	2009	3.8	4.7	N	×	2 Preg		1/2
15	29	1?1	24.2	2.5	1.6	1710	2.7	6.5	N	×	1 Preg		1/2

BMI, body mass index; DHEAS, dehydroepiandrosterone sulfate; F‐G/I, fasting glucose/insulin ratio; free T, free testosterone; FSH, follicle‐stimulating hormone; HOMA‐IR, homeostasis modelassessment for insulin resistance; OGTT, oral glucose tolerance test; LH, luteinizing hormone; N, normal type; IGT, impaired glucose tolerance; ×, no ovulation.

Ovulation failed in all 15 women following the first cycle of metformin alone. After the second and third cycles with combined metformin and CC treatment, ovulation was confirmed in 17 of 28 cycles (61%) and in 13 of 15 patients (87%). There were 15 cycles with one mature follicle and two cycles with two mature follicles, and there was no case of OHSS. The mean ovulatory date was 17.9 ± 2.5 days. Two women became pregnant within 2 months of therapy (13%), so that only 13 women were subject to repeated measurements of endocrine and metabolic profiles. The patients did not show significant weight loss during the three cycles.

Table 2 shows the changes of endocrine and metabolic parameters that occurred during treatment with metformin. A slight reduction in serum levels of LH, free testosterone, androstenedione, DHEAS, hemoglobin A1c (HbA1c) and total cholesterol was observed, while serum LH/follicle‐stimulating hormone (FSH) ratio and serum level of and LDL‐C were significantly decreased after three cycles of metformin therapy.

Table 2.

The change of endocrine and metabolic parameters during metformine therapy

Parameters	At baseline	After three cycles	P
Luteinizing hormone (mIU/mL)	9.4 ± 2.4	5.9 ± 3.9	0.08
Follicle‐stimulating hormone (mIU/mL)	6.8 ± 2.5	6.9 ± 1.9	0.33
Luteinizing hormone/follicle‐stimulating hormone	1.49 ± 0.68	0.91 ± 0.65	0.05
Estrone (pg/mL)	40.2 ± 18.7	45.3 ± 20.2	0.44
Free testosterone (ng/mL)	1.7 ± 1.0	1.3 ± 0.7	0.08
Androstenedione (ng/mL)	2.7 ± 0.6	2.3 ± 0.5	0.18
Dehydroepiandrosterone sulfate(ng/mL)	2196 ± 965	1847 ± 790	0.13
17α‐hydroxyprogesterone (ng/dL)	0.6 ± 0.3	0.6 ± 0.2	0.37
Sex hormone‐binding globulin (nMol/L)	38.6 ± 20.0	46.6 ± 35.0	0.37
Insulin‐like growth factor‐I (ng/mL)	205 ± 37	200 ± 64	0.64
Fasing blood sugar (mg/dL)	90 ± 8	89 ± 7	0.82
Fasting insulin (µU/mL)	10.5 ± 5.4	10.6 ± 5.9	0.87
Area under the curve glucose (mg/dL/min)	16 388 ± 3339	16 278 ± 2358	0.95
Area under the curve insulin (µU/mL/min)	10 428 ± 6599	8438 ± 5564	0.35
Hemoglobin A1c (%)	5.3 ± 0.3	5.1 ± 0.3	0.06
Homeostasis model assessment for insulin resistance	2.4 ± 1.3	2.4 ± 1.5	0.86
Fasting glucose/insulin	13.0 ± 11.6	12.0 ± 8.1	0.93
Total cholesterol (mg/dL)	180 ± 18	168 ± 20	0.06
Triglyceride (mg/dL)	117 ± 59	137 ± 79	0.39
High‐density lipoprotein cholesterol (mg/dL)	49 ± 18	51 ± 16	0.21
Low‐density lipoprotein cholesterol (mg/dL)	110 ± 25	100 ± 28	0.05
Free fatty acid (mEq/L)	0.49 ± 0.24	0.59 ± 0.33	0.53
Leptin (ng/mL)	12.4 ± 8.4	11.0 ± 7.9	0.31

n = 13.

We divided the 13 women into two groups: a responder group (n = 11), which had one or two ovulations in the current study; and a non‐responder group (n = 2), which produced no ovulation. We compared endocrine and metabolic parameters between responders and non‐responders at baseline and after three cycles of metformin therapy (Table 3). At baseline, there were no significant differences in the parameters between the two groups. After three cycles, the levels of plasma fasting insulin was significantly higher and fasting glucose/insulin ratio was significantly lower in the non‐responder group, compared with the responder group, while slight increases in HbA1c, HOMA‐IR and serum levels of triglyceride was observed in the non‐responder group. Non‐responders showed higher insulin resistance parameters, that is, fasting insulin, HOMA‐IR, fasting glucose/insulin ratio, compared with responders at baseline, and after three cycles of metformin therapy (Fig. 1). In the non‐responder group, fasting insulin was more than 15 µU/mL, HOMA‐IR was more than 3.8 and the fasting glucose/insulin ratio was less than 6 at baseline, and after three cycles of metformin therapy.

Table 3.

The change of endocrine and metabolic parameters between responder and nonresponder group during metformine therapy

Parameters	Responders	At baseline Non‐responders	P	Responders	After three cycles Non‐responders	P
n	11	2		11	2
Luteinizing hormone (mIU/mL)	9.3 ± 3.2	10.1 ± 1.1	0.46	6.0 ± 4.4	5.5 ± 0.7	0.81
Follicle‐stimulating hormone (mIU/mL)	6.8 ± 2.4	5.8 ± 2.1	0.58	7.1 ± 2.0	6.1 ± 1.6	0.81
Luteinizing hormone/follicle‐stimulating hormone	1.46 ± 0.65	1.90 ± 0.86	0.27	0.90 ± 0.73	0.92 ± 0.12	0.55
Estrone (pg/mL)	47.7 ± 38.6	37.6 ± 14.3	0.84	47.1 ± 22.1	37.9 ± 10.2	0.43
Free testosterone (ng/mL)	1.4 ± 0.7	2.6 ± 1.6	0.23	1.1 ± 0.7	1.9 ± 0.8	0.15
Androstenedione (ng/mL)	2.7 ± 0.7	2.5 ± 1.0	0.58	2.3 ± 0.6	2.3 ± 0.1	0.90
Dehydroepiandrosterone sulfate (ng/mL)	2236 ± 1002	1815 ± 205	0.52	1894 ± 872	1635 ± 219	0.81
17α‐hydroxyprogesterone (ng/dL)	0.53 ± 0.30	0.80 ± 0.0	0.17	0.55 ± 0.24	0.55 ± 0.07	0.70
Sex hormone‐binding globulin (nMol/L)	40.3 ± 20.2	23.5 ± 6.5	0.20	51.2 ± 37.3	26.0 ± 9.6	0.24
Insulin‐like growth factor‐I (ng/mL)	203 ± 42	190 ± 42	0.62	220 ± 63.8	145 ± 49	0.18
Fasing blood sugar (mg/dL)	89 ± 8	100 ± 5	0.07	88 ± 6	96 ± 8	0.12
Fasting insulin (µU/mL)	11.0 ± 7.0	15.5 ± 0.8	0.37	9.0 ± 5.0	18.0 ± 2.8	0.04
Area under the curve glucose (mg/dL/min)	16 422 ± 4018	19 913 ± 711	0.28	15 428 ± 2331	18 435 ± 1379	0.10
Area under the curve insulin (µU/mL/min)	9932 ± 5966	14 970 ± 750	0.60	8558 ± 5167	7980 ± 398	0.70
Hemoglobin A1c (%)	5.3 ± 0.4	5.5 ± 0.3	0.33	5.0 ± 0.2	5.5 ± 0.2	0.07
Homeostasis model assessment for insulin resistance	3.0 ± 2.3	3.8 ± 0.0	0.17	2.0 ± 1.2	4.3 ± 1.0	0.06
Fasting glucose/insulin	12.4 ± 12.0	6.5 ± 0.6	0.69	13.5 ± 8.3	5.4 ± 0.4	0.03
Total cholesterol (mg/dL)	177 ± 17	201 ± 23	0.09	165 ± 18	189 ± 0	0.10
Triglyceride (mg/dL)	117 ± 59	165 ± 11	0.28	107 ± 55	247 ± 56	0.06
High‐density lipoprotein cholesterol (mg/dL)	50 ± 17	35 ± 2	0.07	56 ± 17	39 ± 3	0.16
Low‐density lipoprotein cholesterol (mg/dL)	107 ± 21	144 ± 28	0.09	94 ± 25	126 ± 16	0.10
Free fatty acid (mEq/L)	0.42 ± 0.24	0.67 ± 0.04	0.24	0.61 ± 0.36	0.50 ± 0.23	0.81
Leptin (ng/mL)	12.6 ± 8.6	12.6 ± 4.1	0.55	10.7 ± 8.8	12.5 ± 3.1	0.29

Figure 1.

The change of insulin resistance parameters in individual polycystic ovary syndrome subjects, classed as responders (; n = 11) or non‐responders (; n = 2) to metformin. HOMA‐IR, homeostasis model assessment for insulin resistance.

DISCUSSION

THE CURRENT STUDY was designed so that Japanese women with CC‐resistant PCOS received low‐dose (250 mg twice daily), short‐term treatment (14 days in each cycle) with metformin. The protocol was based on the fact that the drug information guidelines of Japan recommend a starting dose of metformin for patients with diabetes mellitus type 2 of 250 mg twice daily and a maximum dose of 250 mg thrice daily. The doses of metformin were 1500 mg or 1700 mg daily for CC‐resistant PCOS patients in other studies. ¹⁵ , ¹⁶ , ¹⁷ In the present study, none of the 15 women ovulated during the first cycle of metformin alone, indicating that the duration and/or the dose of metformin may have been too small or that CC‐resistant PCOS may represent a more resistant form of anovulatory disorder. ¹⁸ Following the second and third cycles with combined metformin and CC treatment, ovulation was confirmed in 61% of cycles and 87% of patients, although only two of the 15 women (13%) fell pregnant. This indicates that the addition of metformin to CC in women with CC‐resistant PCOS improves the rate of ovulation, but not the overall fertility rate compared to other studies. ¹⁹ There have been two previous controlled studies evaluating the effectiveness of metformin in combination with CC on the restoration of ovulation and/or pregnancy in women with CC‐resistant PCOS. ¹⁵ , ¹⁶ However, another study in a similar group of Chinese women did not demonstrate improved ovulatory and pregnancy rates in spite of improved hormonal levels, although the total sample size was only 20 women, limiting the power of the study. ¹⁷ The combined data of the three studies showed that 31 of 46 (67%) women ovulated on metformin plus CC, 12 of 50 (24%) ovulated on CC alone, nine of 38 (24%) women conceived on metformin plus CC, and one of 41 (2.4%) conceived on CC alone. ¹⁹ For non‐conceived cases in the current study, higher doses and longer treatment times with metformin may be needed to induce ovulation.

The striking differences in clinical response to metformin among individual subjects with PCOS are not easily explained. We hypothesize that this phenomenon might reflect the heterogeneity in the pathogenesis of the syndrome. The differences in the findings from the present study, compared with other studies, may also be a result of the recruitment of women with CC‐resistant PCOS, diagnostic criteria for PCOS and the inclusion of non‐obese women in the current study. Most studies of metformin therapy were for obese patients with PCOS, but our study also evaluated the effectiveness of metformin therapy for lean patients with PCOS. The endocrine environment is different between obese and lean patients with PCOS. It is clear that obesity contributes to higher insulin resistance and hyperlipidemia. Overweight patients with PCOS have a greater adrenocorticotrophic hormone and cortisol response to opiate blockage than do lean women with PCOS. ²¹ Serum sex hormone binding globulin levels were lower, fasting insulin levels were higher and endometrial surface area was higher in obese PCOS women compared with lean women. ²²

We measured surrogate markers of insulin sensitivity, such as fasting plasma insulin, AUC of insulin, AUC of glucose and the plasma glucose/insulin ratio. The mean values for all patients did not change significantly with metformin treatment. Interestingly, although the improvements of insulin sensitivity in these women were not dramatic, the effects of metformin on menstrual abnormalities were striking, bacause ovulation was confirmed in 87% per CC‐resistant PCOS patients after two cycles with the combination of CC and metformin. The lack of effect on the markers of insulin sensitivity contrasts with some studies that have looked at metformin treatment in PCOS; ⁴ , ²² these reported significant decreases in one or more of these variables or increases in the glucose/insulin ratio. However, another study ²³ showed similar results to ours. In our study, none of these markers at baseline predicted the response to metformin. Characteristics of patients that have previously been shown to favor response to metformin include a higher BMI, higher fasting insulin and lipid concentrations, high blood pressure, lower androstenedione concentrations, and less severe menstrual irregularities. ²² Accordingly, we looked for differences in the parameters that could distinguish between responders and non‐responders after three cycles of metformin therapy. The non‐responder group showed higher insulin resistance parameters, that is fasting insulin, HOMA‐IR, and fasting glucose/insulin ratio, when compared with the responder group. Although there were only two non‐responders in this study, they still failed to ovulate after three cycles of metformin therapy. The familial clustering of anovulation and PCOS suggests an underlying genetic basis. ²⁴ In future, the responsiveness to insulin‐sensitizing drugs may be predicted by related gene analysis.

Recent studies have shown that sequential treatment with metformin and CC was as effective as human menopausal gonadotropin (hMG) in improving ovulation and pregnancy rates in women with PCOS, with a lower drop‐out rate and comparable cost effectiveness. ²³ In our study, the pregnancy rate was very low. Those women who did not conceive on metformin and CC were later advised to undergo hMG treatment. Metformin pretreatment in hMG cycles has been shown to result in fewer mature follicles, lower estradiol levels and lower cancellation rates. ²⁵

In addition to reproductive difficulties, the association of PCOS with insulin resistance and hyperinsulinism puts patients at risk for lifestyle‐related diseases such as type 2 diabetes mellitus, dyslipidemia and cardiovascular disease. ²⁶ Women with PCOS are also at significantly increased risk for IGT and diabetes mellitus type 2 ²⁷ and from our study six women (40%) showed IGT, although two of them were not obese. Polycystic ovary syndrome is also associated with a high risk of coronary heart disease because of dyslipidemia ²⁸ and endothelial dysfunction. ²⁹ We showed that serum levels of total cholesterol and LDL‐cholesterol were significantly decreased after three cycles of metformin therapy. Recent studies suggest that dyslipidemia is secondary to excess androgen action in concert with the hyperinsulinemia associated with insulin resistance, and that anovulation seems to be mainly attributable to insulin resistance and hyperinsulinemia. ³⁰ Women with PCOS also have an increased risk of endometrial cancer. ³¹

In summary, this is the first study that low‐dose, short‐term therapy with metformin restores the responsiveness to CC and increases the rates of ovulation in Japanese women with PCOS who were previously refractory to treatment. Thus, our results provide a rationale for treating CC‐resistant PCOS with metformin and CC, before resorting to the use of gonadotropins or in vitro fertilization and embryo transfer.