Chung 1996.

Methods	RCT, stratified by parity.
Participants	Inclusion criteria: women with singleton breech presentation, as confirmed by US, at 36 to 38 weeks' gestation. N = 51 recruited but 50 analysed. Exclusion criteria: contraindication to tocolytic therapy, scarred uterus, antepartum haemorrhage, hypertension, impaired fetal growth, oligohydramnios, vaginal delivery contraindicated, abnormal umbilical artery Doppler flow pattern.
Interventions	Intervention:tocolytic: ritodrine ‐ beta stimulant (A1) ‐ parenteral IVI of ritodrine 0.4 mg/mL in 5% dextrose at 1.5 mL/min via an infusion pump, for 15 minutes before and during ECV attempt. If uterine contractions appeared to be preventing successful version, the infusion rate was increased in steps of 0.75 mL/min. Compared with matching 5% dextrose infusion. ECV attempted by 2 investigators, followed by repeat US scan and CTG. N = 25. Comparison:placebo: N = 25.
Outcomes	Failed ECV attempt. Other data presented according to successful or failed ECV attempt: non‐cephalic presentation at birth (1/24 vs 23/26); CS (5/24 vs 19/26). 1 intrauterine death occurred 4 weeks after successful ECV (group not stated). Subgroup analysis showed that statistically significant benefit was limited to nulliparous women.
Notes	Paired sequential analysis reached significance after 10 pairs. Trial was continued because of erroneous statistical calculations. Thereafter little benefit was seen from tocolysis. Study authors suggest that tocolysis is helpful only during the learning phase of the technique. A subsequent trial (published earlier) from the same group showed no benefit of tocolysis (Stock 1993). Nulliparous and parous women randomly assigned separately.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated random numbers: sequential paired design.
Allocation concealment (selection bias)	High risk	Randomisation code was known to 1 of the authors who attended each woman throughout the procedure and for 20 minutes thereafter. He did not take part in version attempts. It is not clear whether allocation in pairs may have enabled the unblinded study author to know the next allocation in some cases, which could introduce selection bias, as could the study author knowing the code even if he did not undertake the procedure.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Participant was blinded; 2 doctors who attempted the version were blind to randomisation throughout, but the code was known to a third review author, who was in attendance throughout.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Participant was blinded; 2 doctors who attempted the version were blind to randomisation throughout, but the code was known to a third review author, who was in attendance throughout.
Incomplete outcome data (attrition bias) All outcomes	Low risk	1 woman was excluded because of absent end‐diastolic flow in the umbilical artery before commencement of the procedure (unclear whether they were excluded before randomisation). Unclear whether the analysis was intention‐to‐treat.
Selective reporting (reporting bias)	Unclear risk	We did not assess the trial protocol.
Other bias	Unclear risk	Groups judged by study authors to be similar in age, parity and mean gestation (Table 1, page 721).