Thornton 2004 (GRIT).

Methods	Multi‐centre randomised controlled trial Setting: 69 hospitals in 13 European countries
Participants	547 women (588 babies) recruited, outcomes were available on 547 mothers (587 babies) Inclusion criteria: pregnant women with fetal compromise between 24 and 36 weeks, an umbilical artery Doppler waveform recorded, and clinical uncertainty whether immediate delivery was indicated
Interventions	Immediate delivery (N = 273; IPD N = 141): deliver now, within 48 hrs to permit completion of a steroid course Delayed delivery (N = 274; IPD N = 121): defer delivery, meaning until delivery could safely be delayed no longer
Outcomes	Infant survival to hospital discharge and the Griffith's development quotient at 2 years of age The trial was for compromised preterm fetus: a subset of women within this trial had severe PE.  IPD were available for this subset and these were the data which were extracted and analysed for this review. The outcomes for this subset were as follows. For the woman: CS For the baby: intraventricular haemorrhage, hypoxic ischaemic encephalopathy, or both Apgar score at 5 minutes neonatal seizures NEC ventilation measures of long‐term growth and development (e.g. CP diagnosis, Griffiths score) gestational age at birth
Notes	Only a subset of IPD data were included and analysed in this review (women with hypertension plus either proteinuria or IUGR). Dates of the study: November 1993 to March 2001 Funding sources: UK Medical Research Council, European Union Concerted Action, and the Dutch Princess Beatrix Foundation Declarations of interest: none mentioned
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“A paper‐based number sequence with balanced blocks of 8 to 12 weeks used except during office hours, when a computer‐generated sequence was used”
Allocation concealment (selection bias)	Low risk	“An independent programmer organised allocation, using both randomisation and minimisation.” “The process was designed to mask allocation from participating clinicians, including those with access to the central trial office.”
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Not possible to blind – but most of the outcomes not likely to be influenced by lack of blinding.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Some blinding of outcome assessment for long‐term outcomes, such as Griffiths assessment: “Assessors were masked to the child’s group allocations.”
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	A subset of IPD data was provided of women with severe PE (N = 262 – hypertension, IUGR, proteinuria, or a combination). It was not possible to tell how complete this dataset was, as it was provided by the authors of the original study.
Selective reporting (reporting bias)	Low risk	All expected outcomes appeared to have been reported upon.
Other bias	Unclear risk	Unclear.  Other bias may have been introduced, as only a subset of the original randomised sample provided data for analysis.