Springer 2016.

Study characteristics
Methods	Design: RCT, parallel three‐arm, single‐centre study Setting: orthopaedic specialist centre, Charlotte, NC, USA Recruitment: May 2012 to October 2015 (study dates) Maximum follow‐up: duration of hospital stay
Participants	Participants undergoing primary unilateral total hip arthroplasty (THA) or total knee arthroplasty (TKA) were randomised to one of three groups: Reinfusion drain (Cell salvage/intervention group): N = 60 Hemovac drain (Control/no cell salvage group): N = 61 Tranexamic acid (TXA): N = 65 The average (range) age was 63.3 (33.1 to 81.8) years. No other demographic data are available.
Interventions	Reinfusion drain (Cell salvage/intervention group): the cell salvage group received a reinfusion drain postoperatively (OrthoPAT, Haemonetics, Braintree, USA). The reinfusion drain was used intra‐ and postoperatively for THA participants and postoperatively only for TKA participants as the procedure was performed under tourniquet and intraoperative blood loss would therefore be negligible. The drain was removed on postoperative day 1 at 06:00. The OrthoPAT drain collects, washes, and returns a highly concentrated volume of haematocrit to the patient for up to 12 hours postoperatively. Collected bags are re‐infused within 6 hours of collection. Hemovac drain (Control/no cell salvage group): the drain group received a Hemovac suction drain. This was removed at 06:00 on postoperative day 1. Tranexamic acid (TXA): TXA group received 20 mg/kg of tranexamic acid just prior to skin incision and a further dose 10 minutes prior to tourniquet release for TKA patients.
Outcomes	Outcomes reported: allogeneic blood transfusion, change in haemoglobin level (delta haemoglobin), autologous blood reinfusion, hospital costs
Notes	Transfusion protocol: standard and established transfusion guidelines were based on physiological need rather than a set haemoglobin level. Once hypovolaemia had been corrected, participants were transfused if there were ongoing clinical signs of anaemia. Symptomatic anaemic patients were transfused according to the physicians’ discretion. Prospective registration status: the study was retrospectively registered with a trials registry (NCT01636414). The study start date was before registration. Ethical approval: it is unclear whether the study received approval from a Research Ethics Committee or Institutional Review Board. Language of publication: English Study groups: for the purpose of our review, we have used data from theReinfusion drain (cell salvage/intervention) group and Hemovac drain (control/no cell salvage) group. Trial funding: Novant Health, Charlotte Orthopedic Hospital, Carolinas Medical Center, Mercy Hospital Conflicts of interest: reported but inaccessible
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A random number generator (random.org) was used to generate the stratified randomisation schedule.
Allocation concealment (selection bias)	Low risk	A 1:1:1 randomisation schedule was performed the day before surgery by the study co‐ordinator and the surgeon and hospital staff were notified of the treatment allocation by email.
Blinding of participants and personnel (performance bias) Objective outcome: mortality	Low risk	No objective outcomes reported (mortality unlikely to be affected by blinding if reported in future publications)
Blinding of participants and personnel (performance bias) Subjective: transfusion protocol	High risk	Transfusion protocol in place for the study; however, there is scope for significant between‐participant variability: "Standard and established transfusion guidelines were based on physiological need rather than a set haemoglobin level. Once hypovolaemia had been corrected, patients were transfused if there were ongoing clinical signs of anaemia. Symptomatic anaemic patients were transfused according to the physicians’ discretion".
Blinding of participants and personnel (performance bias) Subjective: all other outcomes	Unclear risk	Study participants and personnel were not blinded to treatment allocation, though no information on other outcomes
Blinding of outcome assessment (detection bias) Objective outcomes: mortality and transfusions	Low risk	No objective outcomes reported (mortality unlikely to be affected by blinding if reported in future publications)
Blinding of outcome assessment (detection bias) Subjective outcomes	Unclear risk	The blinding of outcome assessors is not described and so a judgement cannot be made, though no information on other outcomes
Incomplete outcome data (attrition bias) All outcomes	Low risk	No loss to follow‐up, all outcomes reported
Selective reporting (reporting bias)	Low risk	A study protocol is available and was registered prior to participant recruitment. All outcomes listed are described in the study.
Other bias	Unclear risk	Funding (non‐pharmaceutical) and conflicts of interest (inaccessible online) declared. Unable to assess baseline imbalance as these data are not presented.