Table 4

Risk of bias of the 11 included randomised controlled trials

StudySequence generationAllocation concealmentBlindingIncomplete outcome dataSelective reportingOther biases
Mangano (1996)Computer generated randomised listOnly pharmacy held the listAll blinded, list held by pharmacy2 patients did not complete the study protocol but were analysed as ITTOnly post discharge deaths are mentioned in the primary endpoint.No
Bayliff (1999)Blocks of 4Only one investigator knew the code kept on the patient's health record in a sealed envelope.Blinded1 patient did not undergo major resection and was not continued. 8 patients were withdrawn but were analysed as ITTNoNo
POBBLE (2005)Centrally at Blocks of size 2, 4 and 6 within 4 stratification factors (centre, age, sex and planned use of aortic cross clamping)4 digit trial number assignedAnaesthetists were unblinded. All other clinicians and trial coordinators were blinded1 death occurred after randomisation in a patient who was too ill to tolerate surgery which is not includedNo No
DIPOM (2006)Computer generated. Blocks of 8 stratified for sex, age, perioperative stress, history of coronary artery disease and malignant diseaseTelephone voice responseBlinded188 patients did not receive the allocated intervention but were analysed as ITTNoNo
MaVS (2006)Blocks of 4Not specifiedBlinded117 did not complete the study protocol but were analysed as ITTNoNo
Neary (2006)Packs containing medication or placebo were selected at random by the study investigatorSealed envelopeAllocation was available to the anaesthetic team in an emergency19 patients withdrew their consent and were excludedNoNo
BBSA (2007)Block randomisation in a 1 : 1 ratioNot specifiedBlinded design but β-blocker was titrated to heart rate, so likely effective unblinding5 patients who could not undergo spinal anaesthesia were excludedNoNo
POISE (2008)Computerised randomisation using block randomisation stratified by centre. Randomisation in a 1 : 1 ratio.Central phone randomisationParticipants, healthcare providers, data collectors and outcome adjudicators were blinded but analysts were not20 patients were lost to follow-up but were analysed as ITTNoNo
Yang (2008)Computer generated random tableNot specifiedYesNoNoNo
DECREASE IV (2009)Non-secure
DECREASE I (1999)Non-secure
  • ITT, intention to treat.