safe and successful restriction of transfusion

Safe and Successful Restriction of Transfusionin Burn Patients

Peter Kwan, BSc (Eng), MD, Manuel Gomez, MD, MSc, Robert Cartotto, MD, FRCS(C)

An acceptable strategy for transfusion of burn patients has not been specifically identified.In 1999, we empirically adopted a hemoglobin (Hb) transfusion trigger of 7.0 g/dl orgreater in our burn center. The purpose of this study was to evaluate the effects of this re-strictive transfusion strategy. Retrospective comparison of adults with 20% or greater TBSAtreated from 1999 to 2004 (restrictive group; REST) with patients treated before ouradoption of the restrictive transfusion strategy (1997–1998: liberal group; LIB). The RESTgroup (n � 135, age 42 � 17 years, %TBSA burn 37 � 14, and 26% incidence of inhalationinjury) did not differ significantly from the LIB group (n � 37, age 42 � 16 years, %TBSAburn 38 � 17, and 35% inhalation injury). The Hb triggering a transfusion was 7.1 � 1.2g/dl in the REST group, compared with 9.2 � 2.1 g/dl in the LIB group (P < .001). TheREST group received significantly fewer units of blood per day than the LIB Group. Pa-tients in the REST group appeared to have significantly better organ function, and therewere no differences between the groups in the incidence of acute myocardial infarction.Mortality at 30 days was significantly lower in the REST group (19% vs 38%; P � .03), aswas overall in-hospital mortality (22% vs 46%; P � .003). Transfusion restriction appears tobe safe and resulted in fewer transfusions among this group of burn patients. Prospectivestudies are needed before broadly recommending a transfusion trigger of 7.0 g/dl. (J BurnCare Res 2006;27:826–834)

Anemia frequently develops in thermally injuredpatients who have sustained burns to more than10% TBSA.1–3 Postburn anemia has a wide varietyof causes. First, acute erythrocyte destruction fromthermal and inflammatory insults, combined withsequestration of erythrocytes within the throm-bosed microcirculation of the burn wound,4–6 re-sults in a loss of up to 18% of erythrocytes within 24hours of a 15% to 40% TBSA full-thickness burn.2

Second, despite supranormal erythropoietin levels,the marrow response to erythropoietin appears tobe suppressed after burn injury, resulting in slug-gish erythrocyte production.7,8 Third, burn pa-tients are subjected to regular blood loss as theyundergo staged excisions of the burn wound andharvesting of skin grafts. Although the use of epi-

nephrine tumescence and tourniquets have signifi-cantly reduced operative blood loss,9,10 the factremains that 100 ml to 250 ml of blood (or 2% to5% of the circulating blood volume) are shed forevery 1% BSA that is excised and grafted.9,11 Fi-nally, burn patients, especially those that are criti-cally ill, are repetitively phlebotomized for labora-tory studies and monitoring. Although the amountof blood drawn from burn patients has not beenspecifically studied, a recent prospective observa-tional study of more than 1000 critically ill patientsin the intensive care unit found that a mean of 41ml of blood was drawn in a 24-hour period purelyfor laboratory investigations.12 It would be reason-able to assume that critically ill burn patients lose atleast this much blood on a daily basis, solely fordiagnostic purposes.

Consequently, patients with significant burns willlikely develop anemia, which almost certainly will betreated with one or more blood transfusions. Histor-ically, the decision as to when to transfuse a patientwas largely based on an arbitrary hemoglobin or he-matocrit level (or “trigger”), and was based to a lesserextent on the presence of a clinical hemodynamic

From the The Ross Tilley Burn Centre, Sunnybrook and Women’sCollege Health Sciences Centre, Toronto, Canada.

Address correspondence to Dr. R. Cartotto, Room D710, RossTilley Burn Centre, Sunnybrook Health Sciences Centre, 2075Bayview Avenue, Toronto, Ontario, Canada M4N 3M5.

Copyright © 2006 by the American Burn Association.1559-047X/2006

DOI: 10.1097/01.BCR.0000245494.45125.3E

826

indication, such as myocardial ischemia or acuteblood loss during surgery. The “classic” trigger formany years was a hemoglobin level less than 10 g/dlor a hematocrit less than 30%. Recently, Palmieriet al13,14 conducted multicenter surveys of NorthAmerican burn centers and found that burn physi-cians reported a mean transfusion threshold of 8.1g/dl.13 However, on closer examination of data from666 patients with burns 20% TBSA or greater,14 thehemoglobin levels at first and last transfusions were9.4 g/dl and 9.1 d/l, respectively, suggesting a trans-fusion trigger higher than 9 g/dl.

These transfusion triggers for burn patients arehigher and at odds with transfusion thresholds nowrecommended for other critically ill patients. TheTransfusion Requirements in Critical Care (TRICC)Trial15 prospectively randomized 838 patients in theintensive care unit (ICU) to either a restrictive trans-fusion strategy using a trigger of 7 g/dl with a goalhemoglobin level of 7 to 9 g/dl or to a liberal trans-fusion strategy using a trigger of 10 g/dl with a goalhemoglobin level of 10 to 12 g/dl. Patients in therestricted transfusion group received less blood, ex-perienced slightly less organ dysfunction, and hadcomparable mortality with those in the liberal group,indicating that the restrictive strategy was at least aseffective as the liberal strategy.

The appropriate transfusion trigger for a burn pa-tient is not known. In 1999, after the publication ofthe TRICC trial, we adopted a transfusion trigger of7 g/dl for patients in our burn center. This decisionwas done empirically, although there was some lim-ited evidence from two small studies to suggest thattransfusion thresholds significantly lower than the 10g/dl or 30% hematocrit level were well tolerated byburn patients.16,17 The purpose of the present studywas to evaluate the effects of implementing a restric-tive transfusion strategy with a transfusion trigger of7.0 g/dl among adult burn patients treated at ourfacility. Specifically, we were interested in whetherthe intended transfusion threshold was achieved,whether less blood was transfused, and whether therestrictive approach affected outcomes such as organfunction, length of stay, duration of mechanical ven-tilation, and survival.

METHODS

Study Design, Inclusion, andExclusion CriteriaThis study was an Ethics Review Board-approved,retrospective chart review by one researcher at anadult regional burn center. Eligible study patients

were identified using our burn center’s computerizeddatabase. Charts of eligible subjects were then re-trieved from the hospital’s health data records ser-vices. Transfusion records were retrieved from thehospital blood services laboratory patient database.Patients with %TBSA burns 20% or greater who sur-vived at least 24 hours were eligible for inclusion inthe study. Exclusion criteria included age youngerthan 16 years, presence of associated trauma, knowndiagnosis of chronic anemia at admission (hemoglo-bin �13.5 g/dl in male patients, �12.0 g/dl in fe-male patients), missing or incomplete records, and apatient’s declaration of personal or religious reasonfor refusing blood transfusions.

Study GroupsEligible subjects were assigned to one of two studygroups based on our adoption of a transfusion triggerof 7.0 g/dl in 1999, following publication of theTRICC Trial.15 Patients treated in the 2 years before1999, during which time we were not using a trans-fusion trigger of 7.0 g/dl (ie, from January 1, 1997,to December 31, 1998) formed the liberal transfu-sion group (LIB). Patients treated after adoption ofthe 7.0 g/dl transfusion threshold (ie, from April 1,1999, to May 3, 2004), were assigned to the restric-tive transfusion group (REST). A 2-year period wasselected for the LIB group because significant changesin patient care took place in 1996 with the appointmentof a new burn director in our unit at that time. Becausewe wished to avoid the confounding influence ofchanges in practice patterns that took place in 1996, weselected only the 1997 to 1998 period for the LIBgroup.

Data CollectionData were collected and analyzed during the first 30days post burn or until discharge or death if eitheroccurred before 30 days after injury. Admission dataincluded age, etiology of burn, %TBSA burns,APACHE II score, and presence of inhalation injurydiagnosed by bronchoscopic examination.18 Physio-logic variables were obtained daily and included thePaO2/FiO2 ratio, serum creatinine concentration,serum bilirubin concentration, and the platelet count.If the variable had been measured more than once ona given day, the value closest to 8 AM on that day wasrecorded. Data pertaining to anemia and transfusionincluded the daily hemoglobin concentration (again,if more than one determination, the value closest to 8AM was selected), the number of diagnostic blooddraws per day, the number of units of blood trans-fused daily in all settings (burn unit and operatingroom), and the hemoglobin concentration before

Journal of Burn Care & ResearchVolume 27, Number 6 Kwan, Gomez, and Cartotto 827

transfusion of blood. If a transfusion involved morethan 1 unit of blood, the hemoglobin before the firstunit was recorded as representative of all units givenat that transfusion, unless the hemoglobin concentra-tion was specifically measured between units, inwhich case this value also was recorded for the re-maining units that were administered at that transfu-sion. For blood transfused in the operating room,the preoperative hemoglobin concentration was re-corded as the pretransfusion hemoglobin, unless thehemoglobin was measured intraoperatively, in whichcase this value also was recorded as a hemoglobinbefore transfusion.

Transfusion ProtocolBecause this study was not prospective, patients werenot subjected to a rigid transfusion protocol. In theLIB group, which was treated before 1999, blood wastransfused at an undefined hemoglobin threshold. Inthe REST group, which was treated after 1999, a trans-fusion threshold of 7 g/dl was adopted by the physi-cians in our burn center. This threshold was not rigidlyenforced but was progressively adopted on an empiricbasis after publication of the TRICC trial.15 Single ormultiple unit transfusions were allowed, but, unlikethe TRICC trial, hemoglobin concentrations werenot measured before and after each and every unit ofblood transfused. Our blood bank supplies units ofblood, which range in volume between 240 and 340ml, with a hematocrit of approximately 80%. Begin-ning in April of 1999, all blood supplied to our facilityby the Canadian Blood Service was in the form ofleuko-reduced packed red blood cells.

Outcome MeasuresOutcome measures included mortality from all causesat 30 days, the in-hospital mortality rate, hospitallength of stay, and the duration of mechanical venti-lation. The duration of mechanical ventilation wasdefined as the time from intubation to the time wherethere had been 48 consecutive hours without anymechanical ventilation support, not including oxygenadministration by face mask, tracheostomy mask, ornasal prongs.

Organ dysfunction was measured by the raw valuesof the PaO2/FiO2 ratio (respiratory system), serumcreatinine concentration (renal system), serum biliru-bin concentration (hepatic system), and the plateletcount (hematologic system). The Multiple OrganDysfunction Score, defined by Marshall et al,19 as-signs a score of 0 (best) to 4 (worst) to each of 6 organsystems (Appendix 1). A daily MOD score was deter-mined for each patient as described by Marshall etal,19 and then patients who died were assigned a max-

imum (worst) score of 24 as a means of adjusting themeasured organ dysfunction for death (“the adjustedMOD score”). The Delta Multiple Organ Dysfunc-tion score (� MOD score) is the difference betweenthe MOD score on admission and the overall worstMOD score during hospitalization, and was used tomeasure the change in organ dysfunction over time.

The presence of sepsis, as defined by the ACCP/SCCM Consensus Criteria20 (ie, 2 or more positiveSIRS signs: temperature �36°C or �38°C, heart rate�90 beats/min, white blood cell count �4000/mm3

or �11,000/mm3, and respiratory rate �20 breaths/min or PaCO2 �32 mm Hg and the presence of adocumented infection), was assessed daily over the en-tire hospital stay for each patient.

Statistical AnalysisAll values are reported as the mean � SD unless oth-erwise stated. The means of continuous variables be-tween groups were compared using a two-tailed Stu-dent’s t-test, whereas categorical variables werecompared with the �2 test. A P value �.05 was con-sidered statistically significant.

RESULTS

Study PopulationsBetween January 1 1997 and December 1998, therewere 59 patients who were potentially eligible forinclusion in the LIB group. Of these, 6 were excludedas a result of death within 24 hours of injury, and 16were excluded because of missing or incompleterecords. This left 37 subjects who formed the LIBgroup. Between April 1, 1999, and May 3, 2004,there were 143 patients who potentially were eligiblefor inclusion in the REST group. Of these, five wereexcluded because of death within 24 hours of injury,two were excluded because of presence of associatedpolytrauma, and one was excluded because of missingrecords. This left 135 patients who formed the RESTgroup. The baseline characteristics of the two groupsare shown in Table 1. There were no significant differ-ences between the groups with respect to age, total orfull-thickness burn size, presence of inhalation injury,APACHE II score, timing of first wound excision, ortotal number of operative procedures (Table 1).

Anemia and TransfusionThe mean daily hemoglobin level over the first 30days after burn was 10.2 � 2.3 g/dl in the LIB groupand 8.6 � 2.2 g/dl in the REST group (P � .001).Also, the mean hemoglobin concentration was signif-icantly lower in the REST group on each day after the

Journal of Burn Care & Research828 Kwan, Gomez, and Cartotto November/December 2006

fourth day after burn injury (P � .01; Figure 1). Thepretransfusion hemoglobin (ie, the “transfusion trig-ger”) was 9.2 � 2.1 g/dl in the LIB group and 7.1 �1.0 in the REST group (P � .001). The %TBSA burndid not affect the transfusion threshold in the RESTgroup. Patients with 40% or greater TBSA (n � 51)had a pretransfusion hemoglobin of 7.0 � 10.0 g/dl,whereas those with burns less than 40% TBSA (n �84) had a pretransfusion hemoglobin of 7.1 � 10.3g/dl (P � .350). Conversely, patient age did influ-ence the pre-transfusion hemoglobin threshold in theREST Group: 7.4 � 8.7 g/dl in patients older than60 years of age (n � 23) vs. 6.9 � 10.0 in patientsyounger than 60 years of age (n � 112; P � .001).

The mean daily transfusion rate over the first 30days after burn was significantly greater in the LIBgroup than in the REST group (0.6 � 0.6 units/patient/day vs 0.4 � 0.4 units/patient/day, P �.03). The postburn day of the first transfusion was 4.9 �2.8 days in the LIB group and 5.3 � 3.1 days in the

REST group (P � .539). Within each group, signif-icantly more blood was transfused in the burn unit ascompared with in the operating room (LIB group:6.3 � 7.5 units vs 3.1 � 4.3 units, respectively, P �.03; REST group: 7.1 � 7.4 units vs 1.6 � 2.9 units,respectively, P � .001). The difference between theLIB and REST groups in the number of burn unittransfusions was not significant (P � .583), but therewere significantly fewer operating room transfusionsin the REST group (P � .02). A post-hoc subgroupanalysis was performed in the REST group to exam-ine only patients who always had a pretransfusionhemoglobin less than 7.5 g/dl (ie, those that mostclosely followed the restrictive strategy). There were80 patients who always met the criteria of a pretrans-fusion hemoglobin less than 7.5 g/dl. This subgroupreceived a mean of 3.9 � 5.6 units of blood over thefirst 30 days after burn, which was significantly lowerthan the patients in the LIB group (n � 37), whoreceived 6.3 � 7.3 units of blood in the first 30 days

Figure 1. Mean � SD daily hemoglobin concentration over the first 30 days after burn in the restrictive group (closed triangles)and the liberal group (closed circles). The mean daily hemoglobin concentration is significantly lower on all days in therestrictive group after postburn day 4 (P � .01).

Table 1. Baseline comparison of the liberal transfusion group (LIB) and the restrictive transfusion group (REST) withrespect to patient age, %TBSA burns, %TBSA full-thickness burns, admission APACHE II score, postburn day of the firstburn wound excision, and total number of operations performed in the first 30 days after burn

LIB REST P Value

N 37 135

Age � SD, years 42 � 16 42 � 17 .914

%TBSA burn � SD 38 � 17 37 � 14 .899

%TBSA full thickness � SD 19 � 18 15 � 18 .323

Admission APACHE II � SD 14 � 9 16 � 8 .219

Inhalation injury 13/37 (35%) 35/135 (26%) .211

Day first wound excision 4.4 � 3.6 4.4 � 4.4 .936

Total trips to operating room per 30 days 2.6 � 2.6 3.2 � 2.3 .15


(P �.05). Finally, among those patients who had notransfusions during the first 30 days after burn, theLIB group had significantly fewer transfusion-freedays than the REST group: in the LIB group, 8 of 37patients (22%) received no blood whatsoever over12.6 � 6.9 days while alive and in hospital, whereas inthe REST group, 35 of 135 patients (26%) receivedno blood whatsoever over 19.6 � 8.3 days while aliveand in-hospital (P � .03).

Single vs Multiple Unit Transfusions,Age of BloodIn the LIB group, 18% of all transfusions were single-unit transfusions, whereas 82% were multiple unit(�2 unit) transfusions. The mean number of unitstransfused at a multiunit transfusion was 2.2 � 0.4units. In the REST group, 34% of transfusions weresingle unit, and 66% were multiple-unit transfusions.The mean number of units transfused at a multiple-unit transfusion in the REST group was 2.1 � 0.2units. The difference in proportion of single vs multiple-unit transfusions between the LIB and REST groupswas statistically significant (P � .01). Data for thestored age of blood units transfused in the LIB groupwere not available, but the mean storage age of all

units transfused to patients in the REST group was18 � 8 days.

OutcomesThe 30-day mortality rate was 38% in the LIB group and19% in the REST group (P � .03). Overall in-hospitalmortality also was significantly greater in the LIB groupthan in the REST group (46% vs 22 %, P � .003). Thedaily % survival rates in each group over the first 30 daysafter burn are shown in Figure 2. Comparison of demo-graphic data between survivors and nonsurvivors in eachgroup is provided in Table 2.

The adjusted MOD score was significantly higher(worse) in the LIB Group than in the REST Group(7.9 � 2.5 vs 6.4 � 1.0, respectively, P � .004),(Figure 3). However, the � MOD score, where ahigher value represents a greater deterioration in or-gan function over time, did not differ significantlybetween the LIB and REST groups (3.5 � 3.3 vs3.7 � 2.9, respectively, P � .614). It was observedthat despite a 3-fold difference in sample size betweenthe LIB and REST groups, there was a 6-fold differ-ence in the number of laboratory measures of vari-ables used to determine the MOD score (PaO2/FiO2ratio, creatinine, platelet count, bilirubin), with 1372measures in the LIB group vs 8000 measures in theREST group. In the LIB group, there were 5.3 � 8.6separate episodes of consensus criteria defined sepsisover the entire hospital stay, compared to 4.7 � 5.4episodes in the REST group (P � .624).

The incidence of mechanical ventilation was 51% inthe LIB group and was 74% in the REST group (P �.004). The duration of mechanical ventilation was14 � 11 days in the LIB group and 20 � 18 days inthe REST group (P � .151). Overall hospital lengthof stay was not significantly different between the LIBand REST groups (45 � 88 days vs 44 � 38 days,respectively, P � .942). The incidence of acute myo-cardial infarction over the entire hospital stay was 2.7%in the LIB group and 2.9 % in the REST group.

Figure 2. Plot of the percent survival on each day post burnin the restrictive group (closed triangles) and the liberalgroup (closed circles).

Table 2. Comparison of mean � SD age, %TBSA, %BSA full thickness burn (%BSA full), and incidence of smoke inhalationinjury in survivors and nonsurvivors in the liberal transfusion group (LIB) and the restrictive transfusion group (REST)

LIB Group REST Group

Survivors Nonsurvivors Survivors Nonsurvivors

Age, years 37 � 12 49 � 17 40 � 15 50 � 18

%TBSA burn 37 � 10 44 � 20 36 � 14 45 � 16

%BSA full 11 � 13 27 � 18 12 � 16 26 � 21

Inhalation injury 5/20 (25%) 8/17 (47%) 22/105 (21%) 13/30 (43%)


DISCUSSION

In 1999, after the publication of the TRICC trial,15

which recommended a transfusion threshold of 7.0g/dl for critically ill patients in the intensive care unit,we empirically adopted a transfusion trigger of 7.0g/dl in our adult regional burn center. The results ofthe present study indicate that among adult burn pa-tients with burns 20% or greater TBSA, we success-fully achieved a transfusion trigger very close to 7.0g/dl and that over the first 30 days after burn, thisresulted in a significantly lower daily hemoglobinconcentration and significantly less transfusion ofblood than in the 2-year period before 1999, duringwhich we did not use a 7.0 g/dl trigger. Further-more, burn patients treated during the period whenwe used a restrictive transfusion protocol demon-strated less organ dysfunction and improved survivalcompared with their counterparts who were treatedwith a more liberal transfusion approach.

We were encouraged to find that the transfusiontrigger of 7.0 g/dl was successfully implemented de-spite the absence of a rigid protocol. This finding is incontrast to results obtained from two very large pro-spective multicenter studies on transfusion in the crit-ically ill that were published subsequent to theTRICC trial. In the Anemia and Blood Transfusion inCritically Ill Patients (ABC) Ttrial,12 3534 patients inWestern Europe were transfused for a mean (�SD)pretransfusion hemoglobin concentration of 8.5 �1.1 g/dl. In the CRIT study,21 4892 critically ill pa-tients in the United States were transfused blood forat a mean (�SD) pretransfusion hemoglobin concen-

tration of 8.6 � 1.7 g/dl, suggesting that althoughtransfusion triggers in the critically ill are lower thanthe historical level of 10 g/dl, they have not beenuniformly or consistently reduced to the level of 7.0g/dl identified in the TRICC trial. This alone is aninteresting observation on the relative success or fail-ure of adopting evidence based guidelines in clinicalpractice. Furthermore, the significant reduction in in-cidence of multiple-unit transfusions from 82% in theLIB group to 66% in the REST group is further evi-dence of a profound shift in our clinical approach totransfusion of blood.

The appropriate transfusion trigger for a burn pa-tient is not known. The recent multicenter study oftransfusion among 666 patients with 20% or greaterTBSA treated in North American Burn Centers re-vealed that the mean hemoglobin concentrations be-fore the first and last transfusions were 9.4 g/dl and9.1 g/dl, respectively. Two small studies in burn pa-tients have evaluated the use of lower transfusionthresholds than these. Sittig and Deitch16 prospec-tively compared a restrictive approach to transfusion(transfusion for hemoglobin �6.0 g/dl) in 14 pa-tients (mean of 28% TBSA burns), with a historicalcontrol group of 38 patients (mean of 26% TBSAburns), where a liberal transfusion strategy was ap-plied (hemoglobin maintained �10 g/dl). The re-strictive strategy significantly reduced blood transfu-sions and did not cause any apparent adverse clinicaleffects. In a retrospective study by Mann et al,17 41patients with burns greater than 10% TBSA treated in1980 were compared with 38 patients with burns

Figure 3. Graph showing the daily adjusted Multiple Organ Dysfunction score in the restrictive group (closed Triangles) andthe liberal group (closed circles).


greater than 10% TBSA treated in 1990. There was asignificant reduction in blood transfused between theearlier and later time periods. Although transfusionthresholds were not reported in that study, the au-thors suggested that the reduction in transfusion waspartially related to a reduction in the hematocrittransfusion threshold from 30% in the earlier periodto 15% to 20% in the later period. Although both ofthese studies are limited by small size and their retro-spective design, they provide an important first step inconsidering the concept of restricted transfusionstrategies in burn patients.

The results of our study would suggest that forburn patients with burns �20% TBSA, a restrictivetransfusion approach using a trigger of 7.0 g/dl is safeand at least as effective as the previous trigger of 9.2g/dl, which we used before 1999. It is essential toemphasize, however, that this approach must betested in a randomized prospective fashion because ofthe limitations inherent in a retrospective analysissuch as this. In particular, the use of a historical con-trol group, with a relatively limited number of sub-jects and where several cases were excluded because ofmissing records, are all important potential sources oferror. In particular, the small sample size in the LIBgroup may have been partly to blame for the relativelyhigh reported mortality rate in this group. Anotherimportant consideration in the present study is thatHb was not measured before and after each and everyunit of blood transfused, as was done in the TRICCtrial.15 Thus, the transfusion threshold may actuallyhave been greater than that reported because someunits of blood could have been transfused at a higher(but unidentified) threshold. We would expect thisundermeasurement of the transfusion threshold tohave been more profound in the LIB group since thevast majority of transfusions in that group (82%) in-volved two or more units of blood, whereas in thelater REST group, we would expect less of an under-measurement of threshold because a significantly lowerproportion of transfusions (66%) were multiple-unittransfusions. However, the mean number of unitstransfused at a multiple-unit transfusion in bothgroups was just more than 2 units per transfusion, soit seems unlikely that recording of Hb before eachand every unit of blood would have produced sub-stantially different results.

Another limitation in this study was that we col-lected data closest to 8 AM hours when more than onerecorded variable was recorded in a 24-hour period.We did this to be as consistent and standardized aspossible, but the potential downside of this approachis that we may have not captured deterioration orimprovement in a variable in some patients. How-

ever, we felt that any clinically relevant change in thevariable would have persisted for more than 24 hoursand would have been captured at the next 8 AM re-cording. It should also be noted that this approachhas been used in other well validated studies such asthat used to develop the MOD score, by Marshall etal.19 Also, the REST group had a disproportionatelygreater number of laboratory values measured relativeto the LIB group. Although we cannot specificallyexplain the reason for this, it is possible that patientsin the REST group were being more closely moni-tored, which may have contributed to the improvedoutcomes seen in this group. Also, fewer operatingroom transfusions were given in the REST groupthan in the LIB group. Therefore, some of the ob-served reduced transfusion requirements in the RESTgroup may simply have been from less surgical bloodloss. However, because intraoperative transfusion inour institution is based on measured hemoglobin lev-els rather than on empiric grounds and because therewere no major changes in surgical technique (eg, useof tumescence and tourniquets) between the LIB andREST groups, we believe this is a relatively minorconsideration.

We cannot rule out the contributory effects ofchanges in ICU care that have taken place duringthe7-year span of this study. The difference in the inci-dence of mechanical ventilation, for example, mightsuggest that the REST group was being managedmore aggressively. Also, although there were no spe-cific changes in protocol over the study period, therehave been evolving trends in critical care such as use oflung protective ventilation, and more restricted se-rum glucose control, that may well have contributedto the reported differences in outcomes. Finally,blood transfused after 1999 (ie, that given to patientsin the REST group) was leukoreduced, and it is con-ceivable that this also played a role in the better out-come seen in the REST group.

The much-anticipated randomized prospective studyof restrictive vs liberal approaches to transfusion in burnpatients, currently being organized through the BurnMulti-Center Trials Group of the American Burn As-sociation, will address the limitations in our study andwill likely provide more definitive guidelines on trans-fusion thresholds for burn patients. However, we be-lieve that the findings of this study support the notionthat a more restrictive approach to transfusion is safeand well tolerated among adult burn patients.

The question of whether a higher transfusionthreshold should be used in an elderly burn patient orin a burn patient with a known history of coronaryartery disease has not been answered by this study. Ofinterest was that our transfusion threshold was signif-


icantly higher in patients older than 60 years of agecompared to those younger than age 60, despite thefact that we did not predefine any specific adjustmentsfor patient age to our restrictive strategy. Interest-ingly, in the TRICC trial,15 neither age older than 55years nor the presence of cardiac disease had any effecton mortality in patients treated with a restrictivetransfusion strategy.

Blood transfusion is not a benign therapy. The useof blood products, specifically whole blood andpacked red blood cells, introduces a variety of risksand morbidities. Risks include transfusion reactions,transfusion related bacterial sepsis, and infectious dis-ease transmission from donor to recipient.22–24 Themorbidity associated with blood transfusion mainlyrevolves around immunosuppressive effects on therecipient.25–29 Studies in animal burn models and hu-mans with burns demonstrate that blood transfusiondepresses immune function and increases the risks ofinfectious complications.30–34 There was a statisti-cally insignificant trend toward fewer episodes ofsepsis in the REST group, which may reflect lesstransfusion-related immunosuppression with the re-strictive transfusion strategy. The development oftransfusion-related acute lung injury, defined clini-cally as a syndrome of acute lung injury that developsduring or within 6 hours of transfusion, is also anemerging potential risk now recognized to be associ-ated with transfusion.35 The potential adverse effectsassociated with blood transfusion have been demon-strated in both the ABC trial12 and the CRIT study21

which, when combined, involved more than 8000ICU patients. In both of these studies, the amount oftransfused blood was independently associated withincreased mortality. Similarly, among trauma pa-tients, blood transfusion increases the risk of subse-quent multiorgan failure (MOF). In a study of 513consecutive trauma patients, Moore et al36 demon-strated that blood transfusion during the first 12

hours after injury was an independent predictor ofMOF and that there was a dose-response relationshipbetween early blood transfusion and the severity oflater MOF. Thus, transfusion of blood appears to beharmful in many respects and, therefore, the restric-tion of transfusion would be expected to be associ-ated with improved outcomes. The results of thepresent study support this concept. Although we can-not directly relate the improved survival and reducedorgan dysfunction to the lower transfusion rate in theREST group, our results indicate that a restrictedtransfusion policy is at least as safe and effective as aliberal approach to transfusion. The improvements inoutcome may well have been related to other im-provements in treatment and accumulation of expe-rience that occurred over the seven year span of thisstudy. Nevertheless, the 7.0 g/dl transfusion thresh-old identified in the present study provides an impor-tant preliminary step which must now be more care-fully scrutinized in a prospective randomized study inburn patients.

CONCLUSION

In this study, a restrictive transfusion strategy resultedin significant decreases in the number of blood trans-fusions given to burn patients accompanied by a sig-nificant decrease in 30-day and overall mortality. Theuse of a restrictive transfusion strategy for burn pa-tients is recommended. Although the results of thisstudy support the use of restrictive blood transfusionstrategies in burn patients, a large-scale randomizedcontrolled trial is needed to definitively study theproblem. The resulting outcomes could help deter-mine whether a restrictive strategy is truly beneficialfor burn patients, and potentially which subgroups ofburn patients would most benefit.

Appendix 1. The Multiple Organ Dysfunction Score19

SCORE 0 1 2 3 4

PaO2/FiO2 �300 226–300 151–225 76–150 �75

Creatinine (�mol/l) �100 101–200 201–350 351–500 �500

Bilirubin (�mol/l) �20 21–60 61–120 121–240 �240

Pulse-adjusted heart rate (PAR)(heart rate � centralvenous pressure/meanarterial pressure)

�10 10.1–15.0 15.1–20 20.1–30 �30

Platelets/ml, 10–3 �120 81–120 51–80 21–50 �20

Glasgow Coma Scale 15 13–14 10–12 7–9 �6


REFERENCES

1. Shankar R, Amin CS, Gamelli RL. Hematologic, hematopoi-etic, and acute phase response. In: Herndon DN, editor.Total burn care. London: Saunders Co.; 2002. p. 331–2.

2. Topley E, Jackson DM, Cason JS, Davies JWL. Assessment ofRed Cell Loss in the First Two Days After Severe Burns. AnnSurg 1962;155:581–90.

3. Loebl EC, Baxter CR, Curreri PW. Erythrocyte survival fol-lowing thermal injury J Surg Res 1974;16:96–101.

4. Robb HJ. Dynamics of the microcirculation during a burn.Arch Surg 1967;94:776–80.

5. Kimber RJ, Lander H. The effect of heat on human red cellmorphology, fragility, and subsequent survival in vivo. J LabClin Med 1964;64:922–3.

6. Heatherill RJ, Till GO, Burner LH, Ward PA. Thermal in-jury, intravascular hemolysis, and toxic oxygen products.J Clin Invest 1986;78:629–36.

7. Deitch EA, Sittig KM. A serial study of the erythropoeiticresponse to thermal injury. Ann Surg 1993;217:293–9.

8. Andes WA, Rogers PW, Beason JW, Pruitt BA. The erythro-poietin response to the anemia of thermal injury. J. Lab ClinMed 1976;88:584–92.

9. Cartotto R, Musgrave MA, Beveridge M, Fish J, Gomez M.Minimizing blood loss in burn surgery. J Trauma 2000;49:1034–9.

10. Sheridan RL, Szyfelbein SK. Trends in blood conservation inburn care. Burns 2001;27:272–6.

11. Budney PG, Regan PJ, Roberts AHN. The estimation ofblood loss during burns surgery. Burns 1993;19:134–7.

12. Vincent JL, Baron JF, Reinhart K, et al. Anemia and bloodtransfusion in critically ill patients. JAMA 2002;288:1499–507.

13. Palmieri TL, Greenhalgh DG. Blood transfusion in burns:what do we do? J Burn Care and Rehabilitation 2004;25:71–5.

14. Palmieri TL, Caruso DM, Foster KN, et al. Blood transfusionpractices in major burns: a multicenter study. J Burn CareRehabil 2004;25:S47(abstract).

15. Hebert PC, Wells G, Blajchman MA, et al. A multicenter,randomized, controlled clinical trial of transfusion require-ments in critical care. Transfusion Requirements in CriticalCare Investigators, Canadian Critical Care Trials Group.N Engl J Med 1999;340:409–17.

16. Sittig KM, Deitch EA. Blood transfusions: for the thermallyinjured or for the doctor? J Trauma 1994;36:369–72.

17. Mann R, Heimbach DM, Engrav LH, Foy H. Changes intransfusion practices in burn patients. J Trauma 1994;37:220–2.

18. Fitzpatrick JC, Cioffi WG. Diagnosis and treatment of inha-lation injury. In: Herndon DN, editor. Total burn care.London: Saunders Co.; 2002. p. 232–41.

19. Marshall JC, Cook DJ, Christou NV, Bernard GR, Sprung C,Sibbald WJ. The multiple organ dysfunction score: a reliabledescriptor of a complex clinical outcome. Crit Care Med1995;23:1638–52.

20. Bone RC, Balk RA, Cerra FB. American College of ChestPhysicians/Society of Critical Care Medicine ConsensusConference: Definitions for sepsis and organ failure andguidelines for the use of innovative therapies in sepsis. Chest1992;101:1644–55.

21. Corwin HL, Gettinger A, Pearl R, et al. The CRIT study:anemia and blood transfusion in the critically ill–current clin-ical practice in the United States. Crit Care Med 2004;32:39–52.

22. Pomper GJ, Wu Y, EL, Snyder, et al. Risks of transfusion-transmitted infections: 2003. Curr Opin Hematol 2003;10:412–8.

23. Kleinman S, Chan P, Robillard P. Risks associated with trans-fusion of cellular blood components in Canada. Trans MedRev 2003;17:120–62.

24. Reading FC, Brecher ME. Transfusion-related bacterial sep-sis. Curr Opin Hematol 2001; 8:380–6.

25. Chang H, Hall GA, Geerts WH, et al. Allogeneic red bloodcell transfusion is an independent risk factor for the develop-ment of postoperative bacterial infection. Vox Sanguinis2000;78:13–8.

26. Triulzi DJ, Blumberg N, Heal JM. Association of transfusionwith postoperative bacterial infection. Crit Rev Clin Lab Sci1990;28:95–107.

27. Kendall SJ, Weir J, Aspinall R, et al. Erythrocyte transfusioncauses immunosuppression after total hip replacement. ClinOrthop Rel Res 2000 381:145–55.

28. Salinas JC, Cabezali R, Torcal J, et al. Immune response andcytokines in septic rats undergoing blood transfusion. J SurgRes 1998;80:295–9.

29. Waymack JP, Alexander JW. Blood transfusions as an immu-nomodulator—a review. Comp Immunol Microbiol InfectDis 1986; 9: 177–83.

30. Waymack JP, Rapien J, Garnett D, et al. Effect of transfusionon immune function in a traumatized animal model. ArchSurg 1986;121:50–5.

31. Waymack JP, Robb E, Alexander JW. Effect of transfusion onimmune function in a traumatized animal model. II. Effect onmortality rate following septic challenge. Arch Surg 1987;122:935–9.

32. Winslow GA, Shelby J, Nelson EW, et al. Influence of allo-geneic blood transfusion on natural killer cell activity in burn-injured mice. J Burn Care Rehabil 1996;17:117–23.

33. Waymack JP, Gallon L, Barcelli U, et al. Effect of bloodtransfusions on macrophage function in a burned animalmodel. Curr Surg 1986;43:305–7.

34. Gianotti L, Pyles T, Alexander JW, et al. Impact of bloodtransfusion and burn injury on microbial translocation andbacterial survival. Transfusion, 1992;32:312–7.

35. Toy P, Gajic O. Transfusion related acute lung injury. AnesthAnalg 2004;99:1623–4.

36. Moore FA, Moore EE, Sauaia A. Blood transfusion: an inde-pendent risk factor for postinjury multiple organ failure. ArchSurg 1997;132:620–4.


safe and successful restriction of transfusion

Health & Medicine