Methods	RCT with randomisation of individual women. Sequentially numbered sealed envelopes, using a computer-generated number allocation
Participants	Womenwith spontaneous incomplete miscarriage after up to 13 weeks’ gestation assessed by ultrasound N = 94 women.
Interventions	Intervention: vaginal misoprostol - 600 ug (plus a second dose 24 hours later if miscarriage still not complete). N = 47 Comparison: surgical ERPC by sharp curettage following 20 U of oxytocin per litre of normal saline under GA with no prophylactic antibiotics but oral analgaesics were prescribed. N = 47
Outcomes	Women requiring ERPC after failed medical management; number of doses of misoprostol required; duration of bleeding; adverse effect profile (nausea, vomiting and/or diarrhoea); time spent away from work; use of analgesia
Notes	1. Setting: Gynaecology Outpatient Dept, Durban, South Africa. 2. Additional outcomes assessed but not pre-specified in the review: Hb at 4 days; pain (VAS); duration of analgaesia; days of sick leave; satisfaction (VAS); would use same treatment again; would recommend treatment to friend.
Risk of bias
Bias	Authors’ judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“Computer-generated patient number allocation.”
Allocation concealment (selection bias)	Low risk	“The number was sealed consecutively numbered envelopes by staff not involved in the study. Sealed envelopes were opened and consecutively enrolled women had their allocated treatment. It is not clear, however, whether the enveloped were opaque or not.”
Blinding (performance bias and detection bias) All outcomes	High risk	It was not possible to blind women nor clinicians, and it was unclear whether assessors were blinded for some of the outcomes
Incomplete outcome data (attrition bias) All outcomes	Low risk	No loss of participants nor exclusions reported.
Selective reporting (reporting bias)	Unclear risk	It would appear so though the pre-specified outcomes do not match the reported outcomes fully. Also we did not assess the trial protocol
Other bias	Unclear risk	No figures given on baseline data, only reported as “those who were randomised were well matched for demographic and clinical data”. Study not stopped early for benefit and no other apparent biases