Article Text

Download PDFPDF

Endorsing single-unit transfusion combined with a restrictive haemoglobin transfusion threshold after knee arthroplasty
  1. J M Naylor1,2,3,
  2. S Adie2,
  3. M Fransen4,
  4. S Dietsch1,
  5. I Harris2
  1. 1Whitlam Joint Replacement Centre, Fairfield Hospital, Sydney, New South Wales, Australia
  2. 2South West Sydney Clinical School, University of New South Wales, Sydney, New South Wales, Australia
  3. 3The University of Sydney, Sydney, New South Wales, Australia
  4. 4The George Institute for International Health, University of Sydney, Sydney, New South Wales, Australia
  1. Correspondence to Justine M Naylor, Orthopaedic Department, Liverpool Hospital, Locked Bag 7103, Liverpool, BC NSW 1871, Australia; justine.naylor{at}


Problem The utility of single-unit transfusions in the presence of restrictive haemoglobin transfusion thresholds is unknown.

Design A prospective, pre–post intervention study was undertaken to evaluate a new transfusion strategy designed to reduce the rate of allogeneic transfusion and promote single-unit transfusion.

Setting Joint replacement centre within a public hospital.

Participants Patients undergoing primary unilateral knee arthroplasty (baseline, n=93; postintervention, n=347).

Objectives of the intervention Decrease the use of donor blood by (1) reducing the rate of donor transfusion and (2) endorsing the use of single-unit transfusion.

Strategies for change A restrictive transfusion protocol was introduced, which included assessment of the need for transfusion based on haemoglobin value, and presence of signs, symptoms and comorbidity. Single or multiple units of blood were endorsed depending on the indication.

Key measures for improvement Primary outcomes were transfusion rate; frequencies of attempted and successful single-unit transfusions. Secondary outcomes included 6-week haemoglobin and complications within 6 months postsurgery.

Effects of change Transfusion rate significantly improved (41% (38/93) to 18% (64/347), χ2 21.3, p<0.001). The prescription of single units of blood (24% (9/38) to 33% (21/64), χ2 1, p=0.33) and successful single-unit transfusion (24% (9/38) vs 24% (15/64), χ2<0.01, p=1.0) were unchanged as were most secondary outcomes.

Lessons learnt Restrictive haemoglobin thresholds are a safe, potent frontline strategy for decreasing the rate of blood transfusion. Judicious endorsement of single units is a secondary strategy for reducing the consumption of donor blood when the transfusion haemoglobin trigger is strict.

  • Blood transfusion
  • arthroplasty, knee
  • single-unit transfusion
  • continuous quality improvement

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

The threat to organ oxygenation secondary to acute blood loss primarily underlies the need for blood transfusion after major surgery. However, blood products are limited in supply and allogeneic (donor) transfusion is not without risk.1–3 Consequently, the avoidance of unnecessary transfusion in some patients is as important as the need to provide transfusion in others. Since the seminal randomised trial conducted by Hebert et al4 observing that a restrictive haemoglobin trigger (haemoglobin <70 g/l), compared with a more liberal haemoglobin trigger (<100 g/l), was not associated with worse 30-day morbidity and mortality outcomes in critically ill patients, interest in the use of restrictive transfusion practices to reduce unnecessary donor blood transfusion has grown.5 6

In addition to restrictive haemoglobin triggers, single-unit transfusions are also purported to be viable strategies to restrict donor blood transfusion.2 6 7 The administration of one unit of blood with subsequent evaluation of signs and symptoms potentially reduces the volume of donor blood to which each transfused patient is exposed.2 8 9 It follows then that a reduction in the volume of exposure reduces the overall risks associated with the transfusion episode for the individual as well as decreases total blood usage by the health service.10 Consequently, the administration of single-unit transfusions is increasingly being recognised as an appropriate approach to donor blood transfusion.2 6 7 10

Several recent orthopaedic trials8 9 11–14 have reported the use of comparatively strict haemoglobin transfusion thresholds (<85 g/l) after total knee arthroplasty (TKA). Transfusion rates ranging from 12% to 22% postintroduction of restrictive practices have been reported. On the other hand, the utility of single-unit transfusions when combined with restrictive haemoglobin transfusion triggers has been subject to very little prospective investigation, both in the orthopaedic literature and elsewhere. The available literature is arguably outdated,10 15 and, with few exceptions,9 16 limited to retrospective chart10 15 17–19 or qualitative review.20 Consequently, there is very little evidence in the literature to guide the incorporation of single-unit transfusion into a restrictive transfusion protocol.

Description of context

Setting and function

The Whitlam Joint Replacement Centre is within an orthopaedic unit in a metropolitan public hospital, specialising in lower limb arthroplasty, servicing >300 patients per year. Eight orthopaedic surgeons provide surgery with junior doctors (interns and registrars) rotating through the centre every 3 to 6 months. At the time of the study, the centre did not engage in autologous predonation programmes, and when red blood cells were transfused, each unit averaged 250 to 300 ml in volume and was not leucodepleted.

Problem and purpose of change

Historically, transfusion practices within the centre were clinician-dependent, contrary to current best practice guidelines.1 2 The general transfusion trigger used was a haemoglobin <90 g/l, and single-units were not specifically endorsed. A preliminary snapshot audit revealed the centre's transfusion rate after TKA to be high compared to the prevailing literature. In view of the need to contemporise the centre's transfusion practices—with the ultimate aim of reducing the use of donor blood—a quality improvement study was undertaken in two parts. The first part aimed to understand the centre's transfusion practices and associated outcomes and to develop a new transfusion strategy based on the findings. The second part aimed to evaluate transfusion outcomes after the introduction of the new transfusion protocol. Specifically, the new protocol aimed to (1) reduce the rate of donor blood transfusion after TKA, (2) increase the frequency with which single-unit transfusions were prescribed and (3) increase the frequency with which single-unit transfusions were given successfully.

Design, patient population, situation analysis and strategies for change


A prospective baseline audit was initially conducted to enable a comprehensive situation analysis of the preintervention period (October 2003 to July 2004). This period was subsequently compared with a period immediately after the introduction of the new transfusion protocol (January 2005 to December 2006). The intervening period provided time to design the new protocol based on findings of the baseline audit and recommended guidelines.1 2 Approval for the study was obtained separately for the two study stages. Ethical approval for part 1 was obtained from the Sydney South West Area Health Service Human Research Ethics Committee. Fairfield Hospital's Quality Improvement Committee then provided approval for part 2.

Patient population

All consecutive patients undergoing primary, unilateral TKA during the baseline and postintervention period were included. Patients undergoing additional surgical procedures (such as removal of femoral plates) were excluded.

Situation analysis at baseline and subsequent strategies for change

The baseline period (n=93 patients) confirmed the finding of the preliminary audit—that the observed transfusion rate after TKA was comparatively high (41%) (38/93). Review of our patient and surgical characteristics indicated that these were not atypical compared with published cohorts. On the other hand, several concerns were identified in relation to our ward-based transfusion practices. The documentation concerning the indication for transfusion was poor, 2 units were commonly prescribed without recommendations for a review between units and single units appeared to be prescribed incidentally. In terms of the occurrence of inappropriate transfusions, although no transfusions were associated with a pretransfusion haemoglobin exceeding 100 g/l, 34% (n=13) of transfusions were given in the absence of associated signs and symptoms, and 29% (n=11) were given in the absence of a documented indication. The baseline audit also revealed that most patients were anaemic on discharge (mean haemoglobin 106 g/l), the prescription of oral iron supplementation postsurgery was clinician-dependent and a significant proportion of patients (25%) remained anaemic (ie, had a haemoglobin below the sex-normal range) 6 weeks after surgery.

Informed by these observations and published guidelines,1 2 the new ward-based restrictive transfusion protocol included both a restrictive transfusion haemoglobin trigger and judicious endorsement of single-unit transfusion (box 1) and required clinicians to document the associated indication.

Box 1 Red blood cell transfusion criteria and protocol

  • Transfusion criteria and protocol

  • Pre haemoglobin <90 g/l with either signs and symptoms or significant heart, respiratory or renal disease. Units transfused -----------------clinician dependent

  • Post haemoglobin <75 g/l -------------------×2 units

  • Haemoglobin 75 to 90 g/l and signs or symptoms-------------------------×1 unit

  • Haemoglobin 75 to 100 g/l and significant cardiac, respiratory or renal disease-----------------×2 units

  • Large amount blood loss (usually intraoperatively)-----------------------×2 units

Single or multiple units were prescribed, depending on the indication for transfusion and the presence of signs, symptoms and comorbidity. Further units were prescribed after the initial transfusion episode if signs and symptoms persisted. The protocol was novel as it specifically endorsed the use of single units for low-risk symptomatic patients (ie, postoperative haemoglobin between 75 and 90 g/l and the absence of significant comorbidity) and because documentation of the transfusion episode was via a stamp containing the protocol (box 2).

Box 2 Blood transfusion stamp

Blood transfusion stamp

(circle appropriate response)

A) [Haemoglobin] ≤75

B) [Haemoglobin] >75<90 and signs or symptoms

C) [Haemoglobin] >75<100 and cardiovascular/respiratory/renal history

D) Large amount blood loss (theatre); specify loss:_____


1) A,C,D—×2 units; reevaluate. OR 2) B—×1 unit; support with oral/intravenous fluids; reevaluate signs and symptoms/(haemoglobin).

Note: The stamp was changed in 2006 to a sticker as the ink affected the legibility of the medical record.

Judgement concerning the risk associated with comorbidity remained the responsibility of the assessing doctor. Compliance was facilitated through the education of clinicians with each change in orthopaedic staff rotation. To offset the risk that patients may be discharged with lower haemoglobin levels subsequent to the more restrictive criteria, oral iron supplementation for 6 weeks was routinely prescribed on discharge.


Key measures for improvement

Primary outcomes in the postintervention period remained the same as those collected during the baseline audit. These measures included the rate of donor blood transfusion, the frequency with which single-unit transfusions were prescribed (% attempted), the frequency with which a single-unit was successful in alleviating signs and symptoms related to anaemia (% successful), and the average number of units transfused per transfused patient. Secondary outcomes included discharge haemoglobin, length of stay, discharge destination and complications up to 6 months after surgery. Patient follow-up after discharge was through an outpatient hospital clinic, surgeon consultation or via a phone call depending on where the patient resided and the availability of patient transport. In all cases, a standard complication pro-forma was followed. From a subset of patients (those attending the hospital for follow-up), a 6-week haemoglobin measure was obtained to identify the incidence of anaemia at that time.

Compliance with the protocol was evaluated by monitoring the haemoglobin trigger for transfusion (pretransfusion haemoglobin), the documentation pertaining to the indication for transfusion and the units of blood prescribed according to the set criteria. Demographic and surgical profiles were also monitored for stability across time.

Statistical analyses

The baseline audit was used to perform a power analysis for estimating the minimum sample size required to detect a significant change after intervention. The literature using restrictive haemoglobin thresholds provided the target transfusion rate of approximately 20% for the postintervention period.8 9 11 Thus, based on baseline data, a postintervention sample of 74 patients would detect an absolute 20% reduction (41% to 20%) at the 5% significance level. Beyond the need to satisfy statistical power however, the definitive sample was determined by the need to demonstrate sustainability of the protocol across a lengthy time period (up to 2 years).

Descriptive statistics were computed for contextual demographic and surgical data. Between-period differences were analysed using independent Student t tests for continuous variables or the χ2 test of independence (Pearson χ2 or Fisher's exact test as appropriate) for categorical data. No adjustment was made for multiple comparisons and a p value <0.05 was considered significant. Evaluation of the new protocol was conducted every 3 to 6 months; only the 2-year data are reported here (cyclical evaluations are reported in the e-Addendum).


Excluding one patient who had a concurrent hip replacement, 347 patients were included in the postintervention period compared with the 93 in the baseline period. Table 1 summarises the demographic, surgical and care profiles across the two periods. With few exceptions, profiles remained stable across the two study periods. The type of anaesthetic and prosthesis were clinician-dependent but were essentially unchanged across time. Tourniquets were used routinely (unless contraindicated) from skin incision to skin closure; thus, intraoperative blood loss was not routinely measured. Across both periods, most of the transfusions were given on the orthopaedic ward (baseline, 89% (34/38); postintervention, 81% (52/64)), with few given elsewhere (theatre and recovery (n=4); high-dependency unit (n=11)), and the highest pretransfusion haemoglobin in the postintervention period was 94 g/l. No patient received autologous reinfused blood.

Table 1

Demographic, surgical and care profiles across the study periods


Transfusion rate

A large and significant reduction in transfusion rate was observed in the postintervention period (18% (64/347) compared with 41% (38/93), χ2 21.3, p<0.001).

Prescription and success of single-unit transfusions

There was a non-significant increase in the frequency of prescribing single-unit transfusions from baseline to the postintervention period (24% (9/38) vs 33% (21/64), χ2 1, p=0.33). The frequency with which single-unit transfusion was successful in alleviating signs and symptoms of anaemia did not change from the baseline audit (24% (9/38) to 24% (15/64)).

Number of units given per transfused patient

The average number of units given per transfused patient did not change (1.95 (0.74) to 1.82 (0.68) units, p=0.37).

Secondary outcomes

There was a small reduction in discharge haemoglobin in the postintervention period (106.4 (10.2) to 103.3 (12.6) g/l, p=0.037). Length of stay was unchanged (6.4 (4.3) to 6.2 (3.4) days, p=0.56), but the proportion of patients discharged to inpatient rehabilitation decreased (16% to 9%, χ2=3.8, p=0.05). In the subset of patients who attended the outpatient hospital clinic for follow-up (baseline, 34/93 (36%); postintervention, 159/347 (46%)), no difference in 6-week haemoglobin was observed after commencing the restrictive protocol (124.2 (11.3) to 126.4 (11.3) g/l, p=0.53) nor was there an increase in the incidence of below sex-normal anaemia (24% to 22%).

Most of the patients across both periods were followed up at 6 months (baseline, 84/93 (90%); postintervention, 315/347 (91%)). No significant changes in complications were observed across the two time periods (table 2).

Table 2

Complications up to 6 months post surgery

Protocol compliance

Consistent with the restrictive protocol, pretransfusion haemoglobin in transfused patients was significantly lower during the postintervention period (87.2 (12.4) vs 77.5 (6.1) g/l, p<0.001). Compliance with documenting an indication for transfusion significantly increased from the baseline audit: 71% (27/38 transfusion episodes had a documented indication vs 88% 56/64 (88%) transfusion episodes, χ2 4.3, p=0.04). Compliance with prescribing the number of units according to the documented indication was high (45/56 (80%) cases with a documented indication). Compliance for prescribing 2 units when indicated (indication 1, 3 or 4 on the protocol, box 2) (24/28 occasions) was similar to the compliance for prescribing a single unit when indicated (21/28 occasions).

Post hoc analysis for determinants of transfusion outcome

Unanticipated changes in cement fixation, tourniquet time and postoperative blood losses (table 1) necessitated further scrutiny of possible influence of these changes with respect to the timing of the improvement in the transfusion rate. A breakdown of the postintervention period into two periods (2005 and 2006) revealed that the transfusion rate had decreased to 20% (37/182, χ2 13.5, p<0.001) by the end of 2005 with no further decrease evident in 2006 (16% (27/165), χ2 0.91, p=0.34). In contrast, cement fixation increased to 82% in 2005 (from 60%) (p<0.01) and continued to increase throughout 2006 to 96% (p<0.01). The significant decreases in tourniquet time and postoperative blood loss did not occur until 2006. The timing of these surgical procedural changes was therefore not closely synchronised with improvements in transfusion rate. The influence of these surgical changes on transfusion rate was further explored using post hoc logistic regression modelling (backward regression). Variables in the regression models included categorical variables (the transfusion protocol, sex, anaesthetic type (spinal or epidural, other), and cement fixation) and continuous variables (preoperative haemoglobin, age, body mass index, postoperative blood loss, tourniquet time). The final model revealed that the type of transfusion protocol (p<0.0001) was the strongest predictor of whether a transfusion was given. Lower preoperative haemoglobin (p<0.0001), higher postoperative blood losses (p<0.01) and increasing age (p=0.02) were also predictors, whereas all other variables in the model were not significant.


Summary, context and interpretation

Consistent with the prevailing orthopaedic literature,8 9 11–14 a restrictive haemoglobin transfusion threshold as applied here was shown to be a potent method for lowering transfusion rate. Importantly, this study demonstrated that the lower transfusion rate was sustainable over an extended period and that the incidence of anaemia in the first six postoperative weeks (determined from patient subsets of similar proportions in both periods) was unaffected by the restrictive practices. As our contextual demographic and surgical data appear typical for TKA patients,8 9 12 13 our results are applicable to other arthroplasty centres where transfusions predominantly occur at ward level.

Despite specific endorsement of single-unit transfusion, we observed neither a reduction in the average number of units administered per transfused patient nor an increase in the number of patients successfully receiving single-unit transfusion. That approximately half the transfused patients were indicated 2 units, together with imperfect compliance in prescribing practices for single units, partly explains these observations. There are additional contributory factors that should be illuminated, however. An earlier study16 observed a small increase in single-unit transfusion consequent to a transfusion education programme for hip and knee arthroplasty patients. The indication for the administration of a single unit was not defined in this earlier study, however. A more recent study9 reported a decrease in the use of single-unit transfusion after introducing a maximum allowable blood loss algorithm to assist with transfusion decisions after lower limb arthroplasty. In the context of the latter study, the decrease in administration of single units could be explained by the fact that those patients who would have (perhaps inappropriately) received a single unit before the introduction of the new protocol were consequently (now appropriately) not receiving any donor blood at all. This explanation is also likely to apply to some cases in our study. Moreover, our aim to increase the number of single-unit transfusions—compared with multiple unit transfusions—may have been clinically unrealistic in some cases given that patients in the postintervention period were required to manifest considerably lower haemoglobin values before transfusions were administered. A single unit could then be considered unlikely to sufficiently relieve symptoms in all cases when attempted; hence, our observation that a single unit was not uniformly successful. Historically, controversy has surrounded single-unit transfusion,9 10 15 16 17 and this has been attributed to the long-held belief that if a transfusion is required, at least 2 units are needed. Our study supports this claim in part. Nevertheless, we also contend that for approximately 25% of transfused patients, a single unit did suffice; thus, endorsement of a single unit as documented in our restrictive protocol remains worthwhile.

Although not a primary aim of the study, it is noteworthy that the restrictive practices would have procured substantial savings in the cost of donor blood for the hospital. This is particularly important for the sustainability of TKA surgery in light of the increasing volume of surgery witnessed over the last decade.21


The unanticipated increase in the use of cement fixation may have contributed to a reduction in blood loss from bone. Previous research has shown that cement fixation in TKA patients decreased mean blood losses by a small volume (127 ml compared to no cementing).22 All patients in this latter study were transfused so it was unclear whether cement influenced transfusion risk. The increased cement use observed in the current study could potentially have affected transfusion rate, but post hoc analyses did not support this. We do acknowledge, however, that the increased use of cement fixation may partly explain why the decrease in discharge haemoglobin was modest. Imperfect compliance with the protocol (such as administering 2 units when 1 unit would suffice) may also have buffered the reduction in discharge haemoglobin. These two factors together may help explain why the restrictive practices appeared to be well tolerated both acutely and after discharge from hospital. Although no significant increases in adverse effects manifested up to 6 months after surgery consequent to the restrictive strategies, the study was underpowered to detect small but important changes in complication rates. Furthermore, fatigue questionnaires and return to work rates (although not relevant here for our older population) may have captured more subtle deleterious effects of the restrictive strategies.


The restrictive blood transfusion protocol applied here is simple and appears to be a safe way to avoid unnecessary blood transfusion among patients undergoing TKA. These factors, together with its potency, likely explain the good (though imperfect) clinician compliance observed and its apparent sustainability. Cost savings are inevitable secondary to reduced demand for donor blood. Single-unit transfusion appears to play a secondary role when the transfusion trigger is strict but remains a relevant strategy for many transfused patients.


The authors thank the orthopaedic team within the Whitlam Joint Replacement Centre for endorsing recommended practice.



  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the Sydney South West Area Health Service Human Research Ethics Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Linked Articles