Ward 1993.

Methods	A randomised controlled trial was conducted to study the effectiveness of autotransfusion of shed mediastinal blood in decreasing the need for allogeneic blood transfusion in routine cardiac surgery. Method of randomisation and allocation concealment were not described. The operative team was blinded to the randomisation until the patient arrived in the surgical intensive care unit. No patient in either group donated autologous blood.
Participants	35 consecutive male patients undergoing elective myocardial revascularisation or valve replacement were randomised to one of two groups: Group 1 (Autotransfusion group): n=18; mean (sd) age = 64 (8.5) years Group 2 (Control group): n=17; mean (sd) age = 63 (8.2) years
Interventions	Group 1: Autotransfusion group patients received autotransfusion of mediastinal shed blood for the first 12 hours post‐operatively. Autotransfusion involved reinfusion within 4 hours, a minimum of 100mls of chest drainage in the reservoir before initiation of autotransfusion, and discontinuation of autotransfusion for core temperatures greater than 39.5 degrees celsius. A two‐filter system was employed to minimise emboli. Group 2: Control group were treated with standard chest drainage and fluid replacement. NB: Mediastinal chest drainage tubes were placed in all patients and connected to an in‐line autotransfusion system. The chest drainage system was placed on suction (20cm H20), and the tubes were milked every 15 minutes. Haemodilution was tolerated to a haemoglobin level of 8.0g/dL.
Outcomes	Outcomes reported: amount of blood re‐transfused from the cell saver, amount of allogeneic blood transfused, number of patients transfused allogeneic blood, adverse events, re‐operation for bleeding, blood loss, mortality, myocardial infarction, wound infection.
Notes	Transfusion threshold: patients in both groups received transfusions intra‐operatively and post‐operatively with packed red blood cells when the haemoglobin level fell to less than 8.0g/dL.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Method used to generate allocation sequences was not described.
Allocation concealment (selection bias)	Unclear risk	Method used to conceal treatment allocation was unclear.
Blinding (performance bias and detection bias) All outcomes	High risk