Methods	RCT with randomisation of individual women in blocks of 10.
Participants	Inclusion criteria Women with incomplete spontaneous or induced miscarriage, less than 12 weeks’ gestation. Women living within 1 hour of hospital; past or present history of bleeding during this pregnancy; cervical os open by visual/digital inspection; uterine size of no greater than 12 weeks since last menstrual period; generally in good health; willing to return for follow up. N = 300 women. Exclusion criteria Women with severe infection; known allergy to misoprostol; signs of severe infection (foul-smelling discharge, fever > 39 °C, or pulse > 110/minute) or a known allergy to misoprostol.
Interventions	Intervention: oral misoprostol. 600 ug single dose. N = 150. Comparison: surgery. MVA. N = 150.
Outcomes	Successful miscarriage; adverse effects; women’s satisfaction Study protocol did not call for routine ultrasonography either for initial diagnosis or for determination of treatment success. Assessment at 1 week.
Notes	Setting: Kagera Regional Hospital, Bukoba, Tanzania. All women observed for 3 hours after Rx before being allowed home and antibiotics were given as needed. If miscarriage still incomplete at 7 days, women offered additional week or MVA. Any woman still with incomplete miscarriage at 14 days was offered MVA. Additional outcomes assessed but not pre-specified in the review: bleeding; pain; fever; tolerability.
Risk of bias
Bias	Authors’ judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“….computer-generated random code, created in bocks of 10 at Genunity Health Projects’ Office in New York City.”
Allocation concealment (selection bias)	Low risk	“The code was used by a Genunity employee who was not part of the research team as a basis for sealing cards in consecutively numbered envelopes… staff would open the next envelope in the numbered series…”. Although not opaque enveloped, we think the numbered series should be alright
Blinding (performance bias and detection bias) All outcomes	High risk	Not able to blind participants or clinicians.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No loss to followup and no deviations from protocol allocation reported. ITT analysis
Selective reporting (reporting bias)	Unclear risk	Report on pre-specified outcomes, though we have not assessed the trial protocol
Other bias	Low risk	On most characteristics women did not differ significantly. But significantly more women in the misoprostol group had spontaneous miscarriage and were married. However, we considered that this probably will not have any impact on differences in outcome