Arowojolu 2002

Methods	Randomised, double‐blind, multicentre trial. Random number generation done centrally Similar looking placebos were used
Participants	1160 healthy women recruited into the study from family planning clinics, University College Hospital, Ibadan, and Planned Parenthood Federation of Nigeria, Ikolaba, Ibadan Included women with regular menstrual periods (21‐35 days), who had attended the clinic within 72 h of a single act of unprotected intercourse Excluded women who were not available for follow‐up, were pregnant, on hormonal contraception in the current cycle and those who had contraindications to the use of hormonal contraceptive pills. 1118 into efficacy analysis, 1062 into safety analysis
Interventions	LNG 0.75 mg, 2 doses, 12 h apart orally (split dose) vs LNG 1.5 mg (single dose)
Outcomes	Observed number of pregnancies, side effects and changes in menstrual pattern
Notes	Loss to follow‐up: split‐dose 15/560 and single dose 27/600 Observed pregnancy/total number of women: split‐dose LNG 7/545, single‐dose LNG 4/573 Of the failed cases 3 women in split‐dose group and 1 in single‐dose group continued with their pregnancies and delivered live, healthy babies, while the others were lost to follow‐up
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"participants were randomised into two groups (A and B) using a computer generated random table."
Allocation concealment (selection bias)	Unclear risk	No explanation of allocation concealment
Blinding (performance bias and detection bias) All outcomes	Low risk	"These were administered by a family planning nurse who was blind to the contents in the boxes." "The medications were packed in similar boxes, each tagged with the users name, and containing two tablets. Group A took the box containing one 0.75 mg levonorgestrel tablet and one similarly looking, inactive placebo tablet" Double‐blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	Explained loss to follow‐up
Selective reporting (reporting bias)	Low risk	Reported planned outcomes
Other bias	Low risk	None detected