Subjects

Between February 2003 and May 2004, a total of 50 primary infertile anovulatory CC-resistant women with PCOS and seeking pregnancy were enrolled. All patients were referred to 1 of the 2 Academic Departments of Gynecology.

The diagnosis of PCOS was made initially according to the National Institutes of Health criteria by the presence of both clinical and/or biochemical hyperandrogenism and oligoanovulation, although all patients also fulfilled the European Society for Human Reproduction and the American Society of Reproductive Medicine criteria. Chronic anovulation was diagnosed by a serum luteal progesterone assay (value <2 ng/mL for at least 2 previous consecutive cycles); clinical hyperandrogenism was defined as a Ferriman-Gallwey score ≥8, and biochemical hyperandrogenism was defined as serum total testosterone levels >2 SDs above our reference mean values.

CC-resistance was defined as failure to ovulate at a maximal dosage of 250 mg daily with the use of a classic incremental regimen. In particular, CC was given for 5 days, beginning cycle day 3 after a progesterone-induced withdrawal bleeding with a starting dose of 50 mg daily. If ovulation did not occur, the dose was increased by 50 mg in successive cycles until ovulation was achieved or up to a maximal dose of 250 mg daily.

The following exclusion criteria was used for all subjects: ages <18 or >35 years; body mass index (BMI), >35 kg/m ² ; neoplastic, metabolic (which included diabetes mellitus as excluded by fasting glucose level >126 mg/dL [SI, >6.99 mmol/L], or 2-hour oral glucose tolerance test (value, >200 mg/dL [SI, >11.10 mmol/L]); endocrine, hepatic, renal, and cardiovascular disorders or other concurrent medical illnesses; and current or previous use of any drug that affected hormone levels, metabolism, or appetite (a wash-out period of at least 3 months was considered appropriate before enrolment).

Other exclusion criteria were organic pelvic diseases, previous pelvic surgery, suspected peritoneal factor infertility/subfertility, and tubal or male factor infertility or subfertility that was excluded by hysterosalpingogram and semen analysis, respectively.

We also excluded those women who wanted to start a diet or a specific program of physical activity, cigarette smokers, or alcoholic beverage abusers.

Protocol and interventions

At study entry, all subjects underwent clinical evaluations that included venous blood drawing to evaluate a complete hormonal assay and fasting serum glucose and insulin levels.

Clinical evaluation consisted of anthropometric measurements (which included height, weight, BMI and waist-to-hip ratio) and Ferriman-Gallwey score. Body height and weight were measured without shoes and clothes, respectively. BMI was measured as the ratio between weight and the square of the height, whereas waist-to-hip ratio was calculated as the ratio between the smallest torso circumference and hip circumference. All measurements were performed with the patients in a standing position with relaxed abdomen, arms at their sides, and joined feet. The Ferriman-Gallwey score was calculated by the standard method.

During the same visit, the patient’s daily physical activity, job, and daily activities were evaluated by means of a semiquantitative questionnaire, and a transvaginal ultrasonography was performed.

Blood samples were obtained in the morning between 8:00 and 9:00 am after 12-hour overnight fasting and bed rest during the early proliferative phase (day 2-3) of the progesterone-induced withdrawal uterine bleeding. In anovulatory women, uterine bleeding was induced with a single dose of 100 mg natural progesterone intramuscularly.

Complete hormonal assays that consisted of follicle-stimulating hormone, luteinizing hormone, thyroid-stimulating hormone, prolactin, progesterone, 17-β-estradiol, 17-OH-progesterone, testosterone, androstenedione, dehydroepiandrosterone sulfate, and sex-hormone binding globulin were evaluated as previously reported. The homeostasis model of assessment (fasting glucose [millimoles per liter] × fasting insulin (microunits per milliliter) 22.5], the fasting glucose-to-insulin ratio (milligrams per 10 ^–4 units), and the free androgen index (testosterone [nanomoles per liter]/sex-hormone binding globulin × 100]) were calculated in each subject.

The subjects were allocated randomly into 2 treatment arms of 25 patients each (groups A and B). The randomization was achieved with online software ( www.randomization.it ) to generate a random allocation sequence in single block as the method of restriction. The random allocation sequence was concealed in sealed dark envelopes until the interventions were assigned.

Group A underwent LOD followed by 6 cycles of observation; group B received CC plus metformin for 6 cycles.

LOD procedures were performed by experienced operators. Briefly, according to ovary size, 3-6 punctures of approximately 3 mm diameter and approximately 4-5 mm in depth were performed on each ovary by insertion of an insulated needle cautery of 36 mm as perpendicularly as possible to the ovarian surface with a cutting current of 100-W power. Then, the needle was activated for 2-3 seconds on each point with a coagulating current of 40-W power. At completion of the procedure, the ovarian surface was washed with a crystalloid solution, and all injured areas were covered completely with hyaluronic acid gel to avoid ovarian adhesion formation. Based on clinical conditions, patients were discharged either the same day of the surgical procedure or postoperative day 1.

CC was administered with a classic incremental regimen that was similar to those used in the prestudy phase (detailed earlier).

Metformin was started from day 1 of a progesterone-induced withdrawal bleeding with a dose of 850 mg (1 tablet daily), and the dosage was increased after 1 week up to 1700 mg/day (2 tablets daily). Metformin was taken before lunch during the first week and thereafter before lunch and dinner. All subjects who became pregnant throughout the study suspended metformin.

In this study, no drugs to trigger ovulation were used. All subjects, in the absence of spontaneous withdrawal bleeding after 35 days from last progesterone-induced uterine bleeding and after exclusion of a pregnancy with a serum β-human chorionic gonadotropin assay, received a further dose of 100 mg natural progesterone intramuscularly.

Throughout the study, all subjects were advised not to diet or modify physical activity; to the contrary, they were instructed to follow their usual diet and physical activity.

Follicle growth was not monitored. However, each patient was motivated and instructed to have sexual intercourse regularly. Specifically, it was suggested to have intercourse at least 4 times per cycle (intercourse once every 3 days) starting on day 9 after the progesterone-induced or spontaneous uterine bleedings.

The duration of the study was 15 months for each patient. All women were observed for 6 months. Those who conceived were observed until the end of the pregnancy (for up to further 9 months) to obtain live-birth data for each treatment arm.

Reproductive outcomes

During the study, all reproductive events were recorded for each patient. At the end of the study, spontaneous cycle, pregnancy and live-birth rates per cycle, cumulative pregnancy, multiple pregnancy, miscarriage, and live-birth rates were calculated in each treatment group.

The ovulation was assessed by plasma progesterone assay (>10 ng/mL [SI: 32 nmol/L]) that was performed 21 days after spontaneous or progesterone-induced bleeding (7 days before the expected menses); rising β-human chorionic gonadotropin and sonographic evidence of intrauterine gestational sac were considered to be criteria to define pregnancy.

The ovulation, pregnancy, and live-birth rates per cycle were calculated as the percentage of ovulatory cycles, pregnancies, and live-births per total observed cycles, respectively.

Cumulative pregnancy and live-birth rates were defined as percentage of pregnant patients and patients who had a live baby, respectively, per total patients. The cumulative pregnancy and live-birth rates were also calculated according to time to the first event.

Multiple pregnancy rate was defined as the percentage of patients with a multiple pregnancy per pregnant patients. Miscarriage rate was defined as the percentage of miscarriages during the first 12 weeks of gestation per total pregnancies.

Other outcomes

During the study, the adverse events and patient compliance were noted for each subject. In particular, patients were instructed to report on a daily diary the onset of any adverse event and to specify the severity, duration, and a possible cause-effect relationship with treatment that had been received. To evaluate compliance with the treatment and with the protocol, the number of tablets forgotten, the changes in diet, physical activity, and weight, and the timing of the intercourse were also recorded in the same personal diary.

The total cost of the 2 strategies were also evaluated according to the Italian Diagnosis Related Groups and to the average costs of 6 cycles of medical treatment for groups A and B, respectively.

Statistical analysis

The primary endpoint of our pilot study was the live-birth rate. In January 2003, when the study was designed, no robust clinical data were available in literature regarding the efficacy of CC plus metformin in patients with CC-resistant PCOS; data regarding LOD were extremely heterogeneous. Thus, the sample size was defined arbitrarily. On the other hand, a poststudy power and sample size for live-birth rate were calculated to design a well-powered (>80%) RCT. The poststudy power analysis and the sample size calculation were performed with the use of SamplePower software (release 2.0; SPSS Inc, Chicago, IL).

Data were analyzed with the intention-to-treat method on the basis of treatment assignment and not on treatment received; only those subjects who had not undergone at least 1 follow-up visit after randomization were excluded from the final analysis.

For categoric variables, the Pearson χ ² test was performed; conversely, the Fisher’s exact test was required for the frequency tables when >20% of the expected values were <5.

The normal distribution of continuous variables data was evaluated with the use of the Kolmogrov-Smirnov test. Thus, our data were expressed as mean ± SD and were analyzed with the independent-samples t test.

Cumulative events (pregnancy and live-birth) rate was calculated by the Kaplan-Meier method, with the time to a first event as the outcome variable; the differences between groups were tested with the use of log-rank test. In addition, Cox proportional-hazards model was used to calculate the hazard ratio and its 95% confidence interval [CI] for pregnancy in patients who received experimental or control treatment. The hazard ratio also represented the relative risk because it was calculated for a dichotomous variable, in which there are 2 levels.

Statistical significance was set at a probability value of < .05. The Statistics Package for Social Science (version 14.0.1; SPSS Inc) was used for all statistical analyses.