Kjotrod 2011.

Methods	RCT Setting: 8 infertility centres in Denmark, Finland, Norway and Sweden Method of randomisation: hospital pharmacy used a computer‐generated list Blinding: double Number randomised: 150
Participants	Summary: metformin vs placebo in non‐obese PCOS women Inclusion criteria: PCOS (Rotterdam criteria), aged < 38 years, BMI < 28 kg/m2, scheduled to undergo fertility treatment, minimum of 1 year infertility Exclusion criteria: contraindicated for rFSH, FSH > 10 IU/L, liver or kidney disease, diabetes, fasting blood glucose ≥ 7.0 mmol/L, alcoholism, drug abuse, hyperprolactinaemia, abnormal thyroid function tests, congenital adrenal hyperplasia, androgen‐secreting tumours, Cushing's syndrome; had received oral steroid hormones, cimetidine, anticoagulants, erythromycin or other macrolides. A 1‐month washout if woman previously had metformin. Baseline characteristics of each group: metformin (n = 74) vs placebo (n = 76) Age mean (SD) 29.6 (3.4) vs 29.5 (3.8) BMI mean (SD) 24.0 (2.7) vs 23.6 (2.8) Dropouts: 1 person withdrew from placebo group
Interventions	Main intervention: metformin 500 mg/d gradually increased to 2000 mg/d within first 2 weeks of treatment vs placebo Duration: 12 weeks Co‐interventions: diet and lifestyle advice were given to all patients
Outcomes	Primary: gastrointestinal side effects Secondary: clinical pregnancy rate
Notes	Did not exclude alternative causes of infertility such as male factor, tubal disease, endometriosis The study continued with fertility treatment for those women who did not conceive spontaneously with metformin or placebo. Endocrine parameters measured but results not recorded Women who did not conceive spontaneously after metformin vs placebo, went on to have IVF. We have included up to the IVF stage for analysis in this review.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer random‐number generator
Allocation concealment (selection bias)	Low risk	Pharmacy generated number and packaging of medications was identical
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants and personnel were blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Investigators were blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	Dropouts stated with reasoning where one participant withdrew consent after randomisation
Selective reporting (reporting bias)	Low risk	The primary outcome relevant to this review, spontaneous pregnancy, was clearly reported.
Other bias	Low risk	No evidence of other bias