Levis 2011.

Methods	Design: parallel‐group Number randomly assigned: 248 Number dropped out: 11 in soy group (but they completed all study visits) and 12 in placebo group (but they completed all study visits) Number lost to follow‐up: 23 in soy group (18 before 12 months, five between 12 and 24 months), 43 in placebo group (40 before 12 months and three between 12 and 24 months) Number analysed: 182 (99 in soy group and 83 in placebo group) Intention‐to‐treat analysis: no Power calculation: yes, 80% power to detect a 4% or greater difference in BMD of the lumbar spine, with the assumption that the control group will lose 4% to 5% of bone mass. Target total sample size was 306 with 15% attrition rate expected Duration: two years Timing: July 2004 to March 2009 Location: University of Miami Miller School of Medicine, South Florida, USA Funding: National Institute of Arthritis, Musculoskeletal and Skin Diseases
Participants	Inclusion: women 45 to 60 years of age, menopausal for longer than 12 months but less than five years; or absence of menses for six to 12 months and FSH > 40 IU/L Exclusion: osteoporotic fractures, a bone mineral density T score in the lumbar spine or total hip of < ‐2, BMI of 32 or higher, abnormal mammogram findings, cancer in previous 10 years (except skin cancer), taking bone active drugs, corticosteroids, or herbal products. Taking menopausal hormone therapy within six months before trial Mean age, years: 53 ± 3.3 in soy group; 52 ± 3.3 in placebo group Recruitment method: from South Florida area by direct mailings, posters and presentations to community organisations
Interventions	Isoflavones 200 mg daily from soy protein (4 × 50‐mg tablets) (n = 122). Each 200‐mg dose contained 91 mg genistein, 103 mg daidzein Placebo tablets (n = 126) Treatment for two years taken in the morning before breakfast. Additional calcium supplements were provided to participants as required. Followed up at baseline and at 12 and 24 months
Outcomes	Bone mineral density Bone collagen Menopausal symptoms Vaginal oestrogenisation Serum lipids Thyroid function
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation sequence in blocks of ten
Allocation concealment (selection bias)	Unclear risk	No details
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants and study personnel masked to treatment assignment
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Participants and study personnel masked to treatment assignment
Incomplete outcome data (attrition bias) All outcomes	High risk	Overall, 66 participants were lost to follow‐up (26.6%)—18.8% from the soy group and 34.1% in the placebo group
Selective reporting (reporting bias)	Unclear risk	Quality of life was indicated as a secondary outcome in the protocol but was not reported in the 2011 paper and was not a primary outcome of this review. Not all of the exclusion criteria were listed in the paper