BACKGROUND: Uncertainty about optimal red blood cell transfusion thresholds in cardiac surgery is reflected in widely varying transfusion rates between surgeons and cardiac centres. OBJECTIVE: To test the hypothesis that a restrictive compared with a liberal threshold for red blood cell transfusion after cardiac surgery reduces post-operative morbidity and health-care costs. DESIGN: Multicentre, parallel randomised controlled trial and within-trial cost-utility analysis from a UK NHS and Personal Social Services perspective. We could not blind health-care staff but tried to blind participants. Random allocations were generated by computer and minimised by centre and operation. SETTING: Seventeen specialist cardiac surgery centres in UK NHS hospitals. PARTICIPANTS: Patients aged > 16 years undergoing non-emergency cardiac surgery with post-operative haemoglobin < 9 g/dl. Exclusion criteria were: unwilling to have transfusion owing to beliefs; platelet, red blood cell or clotting disorder; ongoing or recurrent sepsis; and critical limb ischaemia. INTERVENTIONS: Participants in the liberal group were eligible for transfusion immediately after randomisation (post-operative haemoglobin < 9 g/dl); participants in the restrictive group were eligible for transfusion if their post-operative haemoglobin fell to < 7.5 g/dl during the index hospital stay. MAIN OUTCOME MEASURES: The primary outcome was a composite outcome of any serious infectious (sepsis or wound infection) or ischaemic event (permanent stroke, myocardial infarction, gut infarction or acute kidney injury) during the 3 months after randomisation. Events were verified or adjudicated by blinded personnel. Secondary outcomes included blood products transfused; infectious events; ischaemic events; quality of life (European Quality of Life-5 Dimensions); duration of intensive care or high-dependency unit stay; duration of hospital stay; significant pulmonary morbidity; all-cause mortality; resource use, costs and cost-effectiveness. RESULTS: We randomised 2007 participants between 15 July 2009 and 18 February 2013; four withdrew, leaving 1000 and 1003 in the restrictive and liberal groups, respectively. Transfusion rates after randomisation were 53.4% (534/1000) and 92.2% (925/1003). The primary outcome occurred in 35.1% (331/944) and 33.0% (317/962) of participants in the restrictive and liberal groups [odds ratio (OR) 1.11, 95% confidence interval (CI) 0.91 to 1.34; p = 0.30], respectively. There were no subgroup effects for the primary outcome, although some sensitivity analyses substantially altered the estimated OR. There were no differences for secondary clinical outcomes except for mortality, with more deaths in the restrictive group (4.2%, 42/1000 vs. 2.6%, 26/1003; hazard ratio 1.64, 95% CI 1.00 to 2.67; p = 0.045). Serious post-operative complications excluding primary outcome events occurred in 35.7% (354/991) and 34.2% (339/991) of participants in the restrictive and liberal groups, respectively. The total cost per participant from surgery to 3 months postoperatively differed little by group, just £182 less (standard error £488) in the restrictive group, largely owing to the difference in red blood cells cost. In the base-case cost-effectiveness results, the point estimate suggested that the restrictive threshold was cost-effective; however, this result was very uncertain partly owing to the negligible difference in quality-adjusted life-years gained. CONCLUSIONS: A restrictive transfusion threshold is not superior to a liberal threshold after cardiac surgery. This finding supports restrictive transfusion due to reduced consumption and costs of red blood cells. However, secondary findings create uncertainty about recommending restrictive transfusion and prompt a new hypothesis that liberal transfusion may be superior after cardiac surgery. Reanalyses of existing trial datasets, excluding all participants who did not breach the liberal threshold, followed by a meta-analysis of the reanalysed results are the most obvious research steps to address the new hypothesis about the possible harm of red blood cell transfusion. TRIAL REGISTRATION: Current Controlled Trials ISRCTN70923932. FUNDING: This project was funded by the National Institute for Health Research (NIHR) Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 20, No. 60. See the NIHR Journals Library website for further project information.