Methods	RCT with randomisation of individual women. Used a centralised computer-based enrolment and randomisation service. The Coordinating Centre used the biased coin method of maintaining balance between study arms, and was stratified by hospital and gestation (< 7 weeks; 8-10 weeks; 11-13 weeks)
Participants	Inclusion criteria Women with incomplete or inevitable miscarriage at < 13 weeks’ gestation assessed clinically. Bleeding not excessive, haemodynamic system stable, temperature < 37.5 °C, no history of current serious systemic medical or surgical condition, use of prostaglandins not contraindicated (allergy, mitral stenosis, diabetes, blood dyscrasia, haemolytic disease, glaucoma, sickle cell anaemia, hypertension, epilepsy or severe asthma), 18 years or older, not taking anticoagulants or oral corticosteroids, singleton pregnancy, no intrauterine device in situ, and sufficient familiarity with English to complete written questionnaires. N = 40 women. Exclusion criteria A non-viable intrauterine pregnancy diagnosed on ultrasound but with no vaginal bleeding.
Interventions	Intervention 1: vaginal misoprostol. 400u g with repeat dose 4-6 hours later if needed (= 400 ug or 800 ug). N = 13 but 1 woman withdrew immediately after randomisation leaving. N = 12. Intervention 2: surgical management. Aspiration curettage or D&C under GA. N = 12. Comparison: expectant care. N = 15.
Outcomes	Successful evacuation; infection; haemorrhage; pain; bleeding; physical and emotional recovery; anxiety and depression Assessed clinically at 10-14 days and 8 weeks.
Notes	Setting: 5 metropolitan hospitals, Melbourne, Australia. Additional outcomes assessed but not pre-specified in the review: pain; return to usual activities after 2 and 6 days; HADS anxiety score at 2 and 6 days; would choose this method again.
Risk of bias
Bias	Authors’ judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“…a centralised computer-based enrolment and randomisation service…using the biased coin method of maintaining balance between study arms, and was stratified by hospital and gestation (7 weeks or less, 8 - 10 weeks, 11 - 13 weeks).”
Allocation concealment (selection bias)	Low risk	“…a centralised computer-based enrolment and randomisation service, available by telephone 24 hours a day.”
Blinding (performance bias and detection bias) All outcomes	High risk	“Participants and clinicians could not be blind. Unclear of outcome assessor blind or not, though not for outcomes assessed by women. Reports that “The data analyst had access to unblinded data but no contact with any study participant”
Incomplete outcome data (attrition bias) All outcomes	Low risk	1 woman randomised to medical Rx (misoprostol) withdrew following randomisation and was not included in the analyses Medical group: 1 woman was lost to follow up at 10 to 14 days; 1 woman was lost to follow up at 8 weeks Surgical group: 1 woman was lost to follow up at 8 weeks. Expectant group: 1 woman was lost to follow up at 8 weeks.
Selective reporting (reporting bias)	Unclear risk	Outcome measures are listed in the methods section and are those reported in the results section. We did not assess the trial protocol
Other bias	Low risk	Study was planned to recruit 831 women from power calculation 80% power to detect of 5% (99% to 91%) at 0.05 level, but staff were recruiting < 50% eligible women and of these only 22% agreed. So, in effect stopped early but not because of benefit, so probably no bias, just underpowered. No data provided on base line balance, but reported that: “there were no marked or systematic differences between the groups at trial entry with regards to gestation, women’s age, reproductive history, methods of diagnosis, days of bleeding, pain, haemoglobin or white cell count”. There seemed to be no differential diagnosis.