So‐Osman 2013.

Study characteristics
Methods	Design: RCT, parallel, multiple‐arm, multiple‐site trial Setting: 3 hospitals in the Netherlands Recruitment: 2001 to 2003 Maximum follow‐up: 14 days
Participants	603 participants eligible for elective orthopaedic surgery Liberal group: n = 304; mean age (SD) = 70.7 (9.6) years Restrictive group: n = 299; mean age (SD) = 70.2 (10.3) years
Interventions	Restrictive transfusion was compared with liberal transfusion regimens, which varied among the 3 hospitals of the original trial ‐ So‐Osman 2010 ‐ to such an extent that amongst the 3 hospitals involved, investigators' planned so‐called "new" policy (protocol A (restrictive)) " .... was more restrictive than the standard policy (B) for two hospitals and the patients randomised to protocol A were labelled as 'restrictive' and those randomised to protocol B as 'liberal'"  However, in the third hospital, "the new transfusion trigger was in fact more liberal than the hospital's existing standard policy. As a consequence, patients randomised to the new policy (protocol A) actually received more RBC transfusions, and this group was now labelled as "liberal", whereas the group randomised to the standard policy (protocol B) was labelled as "restrictive"   In brief (and with emphasis added to show the overlap):   The protocol intended to be the 'liberal' arm (standard care aka Protocol B) varied between hospitals, thus: Hospital 1 ‐ perioperative period (day 0) trigger in range of 8.1 g/dL to 9.7 g/dL (1 to 2 units. Postoperatively (from day 1) if Hb < 9.7 g/dL, 2 RBC units were transfused, independent of age or other risk factors Hospital 2 ‐ perioperative period (day 0) target of no transfusion unless Hb below 6.4 g/dL in low risk (age < 60, ASA class I); no transfusion unless < Hb 8.1 g/dL if aged 60 or older and ASA class 1, 2, 3; no transfusion unless < 9.7 g/dL if ASA class IV or serious cardiopulmonary disease. Factors determined treatment from the first postoperative day (day 1), including age and cardiac history, leading to a range for no transfusion until Hb dropped to 7.2 g/dL (young, low risk) and 9.7 g/dL (older, high risk) Hospital 3 ‐ perioperative period (day 0) if Hb <  9.7 g⁄ dL and dependent on (expected) blood loss: 2 RBC units were transfused (for all patients). Postoperatively, on day 1, triggers were dependent on risk, but no transfusion until Hb was < 8.1 g/dL in low‐risk and 9.7 g/dL in high‐risk participants The protocol intended to be the 'restrictive' arm (Protocol A) entailed stratifying participants into 3 risk categories (low, intermediate, and high risk). Risk was determined based on age (> 50 years, 50 to 70 years, > 70 years) and/or any one of 11 conditions (e.g. unstable cardiac ischaemia, heart failure, insulin‐dependent diabetes). Within risk categories, triggers differed depending on length of time since surgery (within 4 hours of surgery, or later) At the lowest end possible of the range for Protocol A (low‐risk patients within 4 hours of surgery), patients received no transfusion if Hb remained at or above 6.4 g/dL At the highest end (high‐risk patients > 4 hours after surgery), patients received no transfusion if Hb remained at or above 9.7 g/dL   Hospital 2 had a more restrictive policy as standard care than that employed in the 'new', supposedly restrictive, Protocol A; therefore for the purposes of analysis, we used data from So‐Osman 2013 ‐ in which the arms of the latter trial were reversed and all data were thus appropriate for our comparison
Outcomes	Primary outcome variable: RBC use (originally) Secondary outcomes included: postoperative complications, quality of life
Notes	Review authors (JC and SS) re‐analysed the prior report (So‐Osman 2010), comparing restrictive vs liberal transfusion. It should be noted that whilst details of the thresholds applied in this trial are provided in the papers, there were challenges in implementing the protocol at all sites; significant differences between trial arms were not apparent Trial registration: none ascertainable Trial funding: this trial was fully supported by a grant from the LUMC, Leiden (Doelmatigheidsstudie P01.065) COI statement by investigators: "the authors declare no conflicts of interest" (So‐Osman 2013 p 6)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"For each stratum, a separate randomization list was created, using blocks of variable length to avoid predictability of the random treatment assignment towards the end of each block. Treatment allocation was random using a uniform distribution for a pregenerated list of sufficient length, based on the maximum expected sample size in each stratum. For each subject to be randomized, a sheet of paper with all relevant stratification and group‐allocation information was produced and placed in a sealed opaque envelope. Batches were created according to the stratification factors" (So‐Osman 2010 p 57)
Allocation concealment (selection bias)	Low risk	A research nurse opened sealed opaque envelopes
Blinding of participants and personnel (performance bias) Objective outcomes	Low risk	Blinding of personnel for this intervention is 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' Clinicians caring for participants were aware of allocation status. There was no blinding information on participants
Blinding of outcome assessment (detection bias) Objective measures	Low risk	Blinding of mortality (the primary outcome used within this review) is not relevant, and we graded risk of bias as 'low'. This also relates to transfusion requirements as the trial outcome
Blinding of outcome assessment (detection bias) Subjective measures	High risk	Quality of life measures were reported for an unblinded assessment
Incomplete outcome data (attrition bias) All outcomes	Low risk	No loss to follow‐up
Selective reporting (reporting bias)	Unclear risk	No trial registration identified
Other bias	Low risk	No other biases were apparent, apart from a slight baseline imbalance (history of COPD more prevalent in restrictive group than in liberal group)