Osborne 2011.

Methods	Multicentre RCT (US, Japan, Canada and South Africa) with central randomisation through the GOG Centre. Clinical trial registration ID: NCT00003702 Study duration:1999 to 2007. Follow‐up: 1 year.
Participants	Low‐risk GTN (see notes). Number randomised: 240. Number ineligible:24. Number evaluable: 216.
Interventions	Group 1 = MTX, IM, 30 mg/m², weekly. Group 2 = ACT, IV, 1.25 mg/m², every two weeks.
Outcomes	Primary: CR defined as a normal hCG sustained over four weekly measurements. Secondary: number of CT cycles to remission, treatment failure/need for second‐line CT. Adverse effects.
Notes	Baseline characteristics were similar between the two groups. Risk scoring: WHO/FIGO 2000. WHO risk score 0 to 4 was used between June 1999 to June 2002, then modified score 0 to 6 was used from July 2002 to February 2007. Twenty‐seven women had WHO scores of 0 (balanced between groups) which may have inflated the CR rate. Two women did not receive their allocated treatment, therefore, they were included in ITT analysis but not in analysis of toxicity. Twenty‐four women deemed ineligible: 13 did not meet the entry criteria for persistent disease, 4 did not have GTN, 4 had inadequate documentation of the disease and 3 had a centrally re‐calculated risk WHO risk score > 6. Non‐response (NR) defined as any set of three consecutive assay results that declined by < 10%. Eleven NR women (5 MTX and 6 Act D) continued to receive their allocated treatment and went on to achieve CR. No women had to have allocated treatment terminated because of toxicity. Alopecia coded as dermatological toxicity. 11 women continued on the allocated regimen after being assessed as non‐responders and attained CR. If these women had been included in the analyses of CR the percentage of responders would have been 63% for MTX and 79% for Act D (compared with 53% and 70% respectively).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated random sequence.
Allocation concealment (selection bias)	Low risk	Central randomisation and allocation of treatment.
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Neither participants nor treatment providers were blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	90% of participants were analysed : 107/120 (89%) of weekly MTX arm and 108/120 (90%) of "pulsed" ACT D arm.
Selective reporting (reporting bias)	Low risk	All pre‐specified and expected outcomes reported. Analysis by ITT.
Other bias	Low risk	No evidence of other bias.