Carson 1998.

Study characteristics
Methods	Design: RCT, parallel 2‐arm, multicentre, international trial Setting: 4 university‐affiliated hospitals (3 in the USA; 1 in the UK) Recruitment: March 1996 to March 1997 Maximum follow‐up: 60 days
Participants	84 hip fracture participants undergoing surgical repair who had postoperative Hb < 10.0 g/dL were randomly assigned to 1 of 2 groups: Liberal group: n = 42; M/F = 9/33; mean (SD) age = 81.3 (8.1) years Restrictive group: n = 42; M/F = 11/31; mean (SD) age = 83.3 (10.8) years
Interventions	Liberal group: received 1 unit of packed RBCs at the time of random assignment and as much blood as necessary to keep Hb > 10.0 g/dL Restrictive group: received RBC transfusion for symptoms of anaemia or when Hb dropped to < 8.0 g/dL
Outcomes	Mortality, length of hospital stay, blood usage (units), complications, pneumonia, stroke, thromboembolism
Notes	Trial registration: not reported  Trial funding: supported by Richard C Reynolds Chair COI statement by investigators: none identified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation schedules were stratified by clinical site and cardiovascular disease state. Randomisation was designed in blocks of 2 to 8 participants to avoid imbalance within a site (Carson 1998 p 524)
Allocation concealment (selection bias)	Low risk	Study personnel at clinical sites randomly assigned participants by contacting the data co‐ordinating centre's 24‐hour automated telephone service (Carson 1998 p 523)
Blinding of participants and personnel (performance bias) Objective outcomes	Low risk	Blinding of personnel for this intervention was not feasible, but in our view, for objective outcomes such as mortality (the primary outcome used within this review), we graded risk of bias as 'low'
Blinding of outcome assessment (detection bias) Objective measures	Low risk	There was no blinding of participants or personnel. The primary outcome of mortality allowed a judgement of low risk of bias
Blinding of outcome assessment (detection bias) Subjective measures	Low risk	"Study nurses, blind to the transfusion status of the patient, obtained information from patients or proxies on survival, place of residence, and functional status" (Carson 1998 p 524)
Incomplete outcome data (attrition bias) All outcomes	Low risk	"Of the 84 enrolled patients, 3 refused to comply with the study protocol after randomization, although follow‐up was completed at 60 days. The assigned transfusion strategy was successfully implemented in 93.8 percent (76/81) of the remaining patients. One patient in the threshold transfusion group did not receive a transfusion, and four patients in the symptomatic transfusion group received a transfusion in violation of the protocol (i.e., they did not have symptoms of anemia or a Hb of 8 gldL). Sixty‐day follow‐up was obtained in all patients" (Carson 1998 p 525) Data were analysed by the intention‐to‐treat principle
Selective reporting (reporting bias)	Unclear risk	No reporting bias was apparent, but in the absence of prospective registration or a trial protocol, assessment must remain 'unclear'
Other bias	Low risk	No other biases identified