SupportCancer

Accepte

ReprintSurgical195, Unaware Sumn.ed

0039-60

Publishe

http://d

344 S

Blood transfusion and cancer surgeryoutcomes: A continued reasonfor concernWaddah B. Al-Refaie, MD, FACS, Helen M. Parsons, MPH, PhD, Abraham Markin,Jerome Abrams, MD, FACS, and Elizabeth B. Habermann, MPH, PhD, Minneapolis, MN

Background. The adverse effects of blood transfusion after cancer surgery have been recently challengedin older anemic persons or those with substantial intraoperative blood loss. We hypothesized thatintraoperative blood transfusions continue to adversely impact short-term cancer surgery outcomesregardless of age or preoperative hematocrit levels.Methods. Using the 2005 to 2008 American College of Surgeons National Surgical QualityImprovement Program, we identified 38,926 patients who underwent cancer surgery. Pre-, intra-, andpostoperative factors were compared by units of blood transfusion a patient received. Stratifiedmultivariable analyses, by age and hematocrit level, were performed to assess the impact of bloodtransfusion on operative outcomes, adjusting for covariates.Results. Fourteen percent of patients received an intraoperative blood transfusion. Of those, >60%received only 1 to 2 units of blood. Receipt of intraoperative blood transfusion was associated with higherrates of 30-day operative mortality, major complications, total number of complications, and prolongedlength of stay across age groups and in persons with low to normal hematocrit levels.Conclusion. The present study shows that intraoperative blood transfusion adversely impacts short-termoperative cancer surgery outcomes across all age groups and in those with low to normal hematocritlevels. These findings provide insightful implications on the patterns of blood transfusion during cancersurgery that deserve further investigation. (Surgery 2012;152:344-54.)

From the University of Minnesota and Minneapolis Veteran Affairs Medical Center, Minneapolis, MN

SURGICAL RESECTION remains the critical element oftreatment for most solid malignant tumors. How-ever, up to 40% of these patients present withcancer-related anemia that requires a preoperativeblood transfusion.1,2 Patients may also receive intra-operative blood transfusions because of unexpectedintraoperative events or provider-related factors.3-5

Both single-center and randomized studies haveshown that recipients of blood transfusions experi-ence higher rates of operative mortality, cancerrecurrences, and long-term deaths.6-14

Despite the demonstrated adverse effects ofblood transfusion on cancer surgery outcomes,

ed by The 2008 Veterans of Foreign Wars and LadiesResearch Center Endowment Fund.

d for publication June 7, 2012.

requests: Waddah B. Al-Refaie, MD, FACS, Division ofOncology, Department of Surgery, Mayo Mail Codeiversity of Minnesota and Minneapolis VAMC, 420 Del-treet SE, Minneapolis, MN 55455. E-mail: alref003@u.

60/$ - see front matter

d by Mosby, Inc.

x.doi.org/10.1016/j.surg.2012.06.008

URGERY

several studies have challenged these associa-tions.15-17 Specifically, Wu et al15-17 observed anassociation between preoperative anemia or polycy-themia and poorer operative outcomes after majornoncardiac surgery. Indeed, Wu et al15-17 haverecently shown that blood transfusion is associatedwith a lower 30-day operative mortality among olderpersons with preoperative hematocrit (Hct) levels of<24% or substantial blood loss after major noncar-diac surgery. However, such results dismiss the possi-bility that the adverse impact of blood transfusionon cancer surgery outcomes is linked to the circum-stances that necessitate it, rather than the transfu-sion itself. For example, patients undergoingemergent surgery may be more likely to receiveintraoperative blood transfusions compared tothose undergoing elective procedures, potentiallyconfounding the relationship between blood trans-fusions and operative outcomes. In addition, under-standing the impact of blood transfusion onoperative outcomes after cancer surgery deservesadditional investigation beyond single-center expe-riences18-20 because of the rising costs of bloodbank operations and the growing demand for can-cer surgery in the aging population.

SurgeryVolume 152, Number 3

Al-Refaie et al 345

We hypothesized that intraoperative bloodtransfusions impact short-term cancer surgery out-comes across different age groups or preoperativeHct levels. We evaluated the patterns and impact ofblood transfusions on operative outcomes aftercancer surgery in the American College of Sur-geons National Surgical Quality Improvement Pro-gram (ACS NSQIP), which is one of the premiersurgical quality program in the United States.21-29

METHODS

Data source. We used the 2005 to 2008 ACSNSQIP Participant Use File for our study, a large,prospectively collected, multihospital database.ACS NSQIP collects detailed patient data for thoseundergoing both inpatient and outpatient surgicalprocedures in more than 250 participating univer-sity and private centers across theUnited States.21-29

Data collected by a dedicated Surgical Clinical Re-viewer include more than 136 variables, consistingof preoperative risk factors and laboratory values,intraoperative factors, and 30-day postoperativeoutcomes. Approval to conduct the study was ob-tained from both of the University of Minnesota In-stitutional Review Board and the ACS NSQIP.However, the ACS NSQIP is not responsible forthe statistical validity of the data analysis or the con-clusions derived by users of the data. The ACSNSQIP and the hospitals participating in it arethe source of the data used herein; they have notverified and are not responsible for the statistical va-lidity of the data analysis or the conclusions derivedby the authors.

Patients and cancer surgeries. We identified38,926 patients over 18 years of age who weresurgically treated for thoracic, abdominal, or pelvicneoplasms according to the International Classifica-tion of Diseases, 9th revision (ICD-9) diagnoses codeswithin the ACS NSQIP. We then limited our analy-sis to those who underwent both elective and non-elective major resections for their neoplasmaccording to Current Procedure Terminology(CPT) codes. Patients were categorized accordingto their level of intraoperative packed red bloodcell (RBC) transfusions (in units): 0 units, 1–2units, and >3 units. Cancer surgeries were catego-rized into intermediate (segmental and total ab-dominal colectomy), major (total gastrectomy,hepatectomy, bile duct resection, nephrectomy,urinary cystectomy, proctectomy, and abdomino-perineal resection), and complex (lung, esopha-geal, and pancreatic resections).

Adverse postoperative outcomes. Short-termoperative outcomes included 30-day mortality, the

development of major complications, the totalnumber of postoperative complications (both ma-jor and minor), and the proportion of patientswith a prolonged duration of stay (DOS).

Postoperative complications were classified aseither major or minor. Major complications in-cluded deep incisional or organ/space surgical siteinfection, wound disruptions, pneumonia, theneed for reintubation, pulmonary embolism, pro-gressive renal insufficiency, acute renal failure,cerebrovascular accidents, coma, peripheral nerveinjury, cardiac arrest, myocardial infarction, graft/flap failure, sepsis, septic shock, the need forreturn to the operating room, or >48 hours on aventilator. Minor complications included superfi-cial surgical site infections, urinary tract infections,and deep vein thrombosis/thrombophlebitis. Anindividual’s total number of complications wascreated by summing major and minor complica-tions over the course of the patient’s hospital stay.

DOS was defined as the number of days fromoperation to hospital discharge. We analyzed onlypatients discharged alive from the hospital onpostoperative days 4 through 30 because 95% ofpatients fell within this range and a shorter DOSmay indicate potential miscoding. Consistent withprevious NSQIP work, we defined prolonged DOSfor our study as a DOS beyond the 75th percentilefor all procedures. In our study, 75% of the cohortwas discharged before postoperative day 9.6,30,31

Statistical analyses. We compared unadjusteddifferences in preoperative factors, preoperativelaboratory values, intraoperative data, and short-term operative outcomes by intraoperative RBCreceipt using Chi square tests for categorical vari-ables and analysis of variance tests for continuousvariables. We also conducted a Cochrane–Armitagetest for trend to examine the relationship betweenRBC transfusions and an increased total number ofcomplications.

We sought to identify the independent contri-bution of RBC transfusions to the development ofadverse operative outcomes. However, we acknowl-edge that some preoperative factors associatedwith adverse operative outcomes may also influ-ence transfusions of intraoperative RBCs. First, wecreated a logistic regression model to identifythose factors associated with receiving intraoper-ative RBC. Adjusting for these factors, we thenperformed multivariate logistic regression to iden-tify the independent association between intra-operative RBC transfusion and the likelihood ofdeveloping adverse short-term operative outcomes.Specifically, we examined 30-day operative

SurgerySeptember 2012

346 Al-Refaie et al

mortality, the likelihood of developing major (ver-sus no) postoperative complications, and pro-longed DOS adjusting for covariates. Finally, toexamine the association between intraoperativeRBC transfusion and an increased total numberof postoperative complications (either minor ormajor), we used ordinal logistic regression.

In order to ensure that our findings were not aresult of our model assumptions, we performedseveral sensitivity and interactions analyses (see theResults section). For all analyses, we consideredP < .05 to be statistically significant. All analyseswere performed with SAS software (version 9.2;SAS Institute, Cary, NC).

RESULTS

Of the 38,926 patients who underwent a majorthoracic, abdominal, or pelvic resection for cancerin the 2005 to 2008 ACS NSQIP, 14.6% received atleast 1 unit of RBCs during their cancer surgery(Table I). Receipt of 1–2 units of blood transfusionwas the common trend for intraoperative bloodtransfusion during cancer surgery.

While the overall rate of intraoperative bloodtransfusion was low within the ACS NSQIP, olderpatients and those with a higher American Societyof Anaesthetists (ASA) score or comorbidities wereall more likely to receive RBCs. In addition,patients with abnormal preoperative laboratoryvalues, including Hct and albumin, were alsomore likely to receive intraoperative RBCs.

Short-term cancer surgery outcomes by receiptof blood transfusion. The overall 30-day mortalityrate in this cohort was 2.3%. However, more than19% of patients experienced complications. Un-adjusted rates of 30-day operative mortality, majorcomplications, and prolonged DOS were all signif-icantly higher in recipients of intraoperative RBCscompared to those with no transfusion (Table II).

Factors associated with intraoperative bloodtransfusion during cancer surgery. In our multivar-iate model, several factors were significantly associ-ated with RBC receipt, including older age,dependent functional status, higher ASA score,and cardiac comorbidities (Table III). However,the effect of these factors, particularly the relation-ship between older age, poorer functional capacity,and higher ASA score, were highly correlated. Testsfor interaction revealed that the effect of ASA score,cardiac comorbidities, smoking, and functional sta-tus on predicting the receipt of intraoperative RBCsvaried significantly by age. As a result, our final re-sults present a stratified analysis to show how thesefactors vary with increases in age (Table IV).

In the stratified analyses, the largest factorsassociated with RBC receipt included preoperativeHct level, higher ASA scores, comorbidities, andtype of operative resection.

Multivariable analysis of blood transfusion andcancer surgery outcomes. Multivariable analysiscontinued to show the adverse effect of RBCtransfusion on short-term operative outcomes,adjusting for covariates. In addition, we identifiedan interaction between age and RBCs, indicatingthat the effect of RBC on adverse operative out-comes varies by increasing age. As a result, ourfinal results present the independent effect ofRBCs on adverse operative outcomes stratified byage groups (Table IV).

In our age-stratified multivariate models, intra-operative RBC transfusion remains significantlyassociated with poorer 30-day mortality when ad-justing for covariates (Table IV). The effect ofhigher levels of intraoperative RBCs on mortalityis most pronounced in patients less than 55 yearsold (OR, 6.6 for >3 units vs 0 units; P < .001).

In addition to its adverse effect on 30-dayoperative mortality, patients receiving intraopera-tive RBCs were also more likely to experiencemajor complications, an increased total numberof complications, and extended DOS across all agegroups.

Because of the literature showing an associationbetween Hct and operative outcomes, we furtherstratified our short-term operative outcomes byHct level.15-17 The adverse effect of intraoperativeRBC on operative outcomes after cancer surgerycontinued in those with anemia and normal Hctlevels (Table V). However, because of the smallnumber of patients with high Hct level who also re-ceived intraoperative RBCs (0.004%), we were un-able to accurately verify this association in thisparticular group.

We performed a series of sensitivity analyses toensure that our results were not influenced by ourmodeling decisions. First, the impact of RBCtransfusion on poorer operative outcomes re-mained unchanged regardless of whether we mod-eled units received as a continuous or categoricalvariable. Second, we repeated our multivariateanalyses after removing patients with missing datato show that our estimates of the effect of RBCreceipt on operative outcomes remained un-changed (data not shown). Third, given the po-tential differences between patients who receiveonly intraoperative RBCs compared to those whoadditionally receive either pre- or postoperativeRBCs, we repeated our multivariable analysis after

Table I. Pre- and intraoperative factors profile stratified by receipt of intraoperative red blood cells(N = 38,926)*

No. of units

P value0 (85.4%; n = 33,242) 1–2 (9.7%; n = 3,763) >3 (4.9%; n = 1,921)

Age (years) <.001<55 25.0% 18.8% 26.0%56–65 24.7% 21.7% 26.1%66–74 23.2% 26.0% 24.6%>75 27.1% 33.5% 23.3%

Sex <.0001Female 48.5% 49.7% 44.0%Male 51.5% 50.3% 56.0%

Race <.0001White 75.3% 74.5% 72.9%Black 8.9% 11.1% 9.5%Hispanic 3.9% 4.7% 5.5%Other 11.9% 9.7% 12.1%

Smoking status .01No 82.7% 84.3% 84.0%Yes 17.3% 15.7% 16.0%

ASA classification <.00011 + 2 46.6% 24.5% 21.8%3 + 4 53.4% 75.5% 78.2%

Cardiac conditions <.0001No 87.9% 82.8% 85.8%Yes 12.1% 17.2% 14.2%

Hypertension requiring medication <.0001No 47.2% 40.0% 45.2%Yes 52.8% 60.0% 54.8%

Pulmonary conditions <.001No 85.2% 80.0% 83.0%Yes 14.8% 20.0% 17.0%

CNS conditions <.0001No 93.3% 92.6% 94.1%Yes 6.7% 7.4% 5.9%

Diabetes mellitus <.0001No 83.9% 76.2% 77.7%Yes 16.1% 23.8% 22.3%

Total operation time (mean ± SD) 188.2 ± 108.5 278.8 ± 141.9 368.9 ± 155.3 <.0001Creatinine (mg/dL) <.0001

#1.6 89.5% 89.9% 89.6%>1.6 4.0% 7.3% 8.0%Missing 6.5% 2.8% 2.4%

Albumin (g/dL) <.001>2.6 67.1% 74.3% 78.0%#2.5 2.1% 7.2% 9.5%Missing 30.8% 18.5% 12.5%

WBCs (k/mL) <.0001#4.5 8.9% 11.0% 11.8%4.5–11.0 79.1% 74.6% 73.5%$11.0 6.9% 12.1% 12.9%Missing 5.1% 2.3% 1.8%

Total bilirubin (mg/dL) <.0001#1.0 59.6% 64.5% 65.1%>1.0 11.0% 17.9% 21.8%Missing 29.4% 17.6% 13.1%

(continued)

SurgeryVolume 152, Number 3

Al-Refaie et al 347

Table I. (continued)

No. of units

P value0 (85.4%; n = 33,242) 1–2 (9.7%; n = 3,763) >3 (4.9%; n = 1,921)

Hematocrit (%) <.0001#38 40.9% 72.0% 68.2%38–44 45.9% 23.3% 26.8%$44 9.1% 2.7% 3.2%Missing 4.1% 2.0% 1.8%

*Fewer than 5% of patients had nonindependent functional status; had >4 units of red blood cells transfused 72 hours before surgery; were admittedfrom a nonhome setting; had preoperative systemic sepsis; had a previous operation within 30 days of surgery; or had an emergent procedure.ASA, American Society of Anaesthetists; CNS, central nervous system; SD, standard deviation; WBCs, white blood cells.

Table II. Short-term operative outcomes stratified by receipt of intraoperative red blood cells (N = 38,926)

No. of units

P value0 (85.4%;n = 33,242)

1–2 (9.7%;n = 3,763)

>3 (4.9%;n = 1,921)

30-day mortality 1.7% (568) 4.8% (179) 7.8% (150) <.001Duration of stay (days)

Mean 7.95 ± 7.28 11.8 ± 11.3 14.8 ± 13.6 <.0001Median 6 9 10 <.0001% of pts with prolonged DOS (9 days) 25.5% 50.8% 63.8% <.0001

Types of postoperative complicationsMinor 7.6% 10.0% 9.5% <.0001Major 16.4% 30.0% 43.9% <.0001

Return to the ORYes 5.4% 9.3% 14.8% <.0001

DOS, Duration of stay; OR, operating room.

SurgerySeptember 2012

348 Al-Refaie et al

removing them from our analyses. Fourth, weperformed our analyses adjusting for individualprocedures rather broader groupings of proce-dures (intermediate, major, or complex) to evalu-ate whether our groupings impacted the effect ofRBCs on adverse outcomes. Fifth, we performedour analysis with and without patients who under-went emergency procedures, because they maypresent with different clinical factors than thoseundergoing elective resections. We also includedemergent status as a factor in multivariate models.Finally, because of potential collinear relationshipbetween the receipt of intraoperative RBC transfu-sion and Hct level, the adverse effect of RBCtransfusion on operative outcomes remains un-changed after removing Hct level from the analy-ses. The continued adverse effect of RBCs on worseoperative outcomes remained unchanged acrossall sensitivity analyses.

DISCUSSION

The present study shows that intraoperativeblood transfusions adversely impact short-termoperative cancer surgery outcomes at ACS NSQIP

hospitals. These adverse effects remained consis-tent across all age groups and among those withanemia or normal Hct levels. To our knowledge,this is one of the largest multihospital studies ofintraoperative blood transfusion and cancer sur-gery outcomes controlling for clinically relevantfactors including comorbidities, functional status,and level of Hct.

In the present study, patients who received anyblood transfusion had 2- to 3- fold higher risk ofpoorer operative outcomes measures. The currentresults are consistent with previous single-hospitalobservational and experimental studies that foundthat blood transfusions were associated withpoorer short-term outcomes because of its immu-nosuppressive effects.6-10,13,14,31,32 Despite these as-sociations, previous studies did not control forother important and clinically relevant factors,such as comorbidities, functional capacity, andASA that also predict blood transfusion.7-10,13,14,32

The use of ACS NSQIP---a robust and establishedsurgical quality program in the United States---al-lows us to accurately assess both the patterns andimpact of blood transfusion on several important

Table III. Selected factors associated with red blood cell receipt, stratified by age categories (N = 38,926)

Factor

Age group (years)

<55 (N = 9,509) 56–65 (N = 9,518) 66–74 (N = 9,173) $75 (N = 10,726)

SexFemale Ref Ref Ref RefMale 0.96 0.91 1.06 1.02

RaceWhite Ref Ref Ref RefAfrican American 1.39* 0.99 1.24* 1.18Hispanic 1.47* 1.21 1.27 1.35*

SmokingNo Ref Ref Ref RefYes 0.78* 0.76* 0.84 0.93

DiabetesNo Ref Ref Ref RefYes 1.15 1.04 1.16* 0.98

ASA1 + 2 Ref Ref Ref Ref3 + 4 1.71* 1.59* 1.58* 1.66*

Cardiac comorbiditiesNo Ref Ref Ref RefYes 1.56* 1.35* 1.24* 1.08

Functional statusOther Ref Ref Ref RefIndependent 1.52 1.35 1.53* 1.42*

Pulmonary comorbiditiesNo Ref Ref Ref RefYes 1.54* 1.16 1.20* 1.12

CNS comorbiditiesNo Ref Ref Ref RefYes 0.76 1.08 1.04 0.77*

Creatinine (mg/dL)#1.6 Ref Ref Ref Ref>1.6 1.47* 1.68* 1.47* 1.51*

Albumin (g/dL)#2.6 Ref Ref Ref Ref>2.6 0.43* 0.36* 0.32* 0.72*

Bilirubin (mg/dL)#1.0 Ref Ref Ref Ref>1.0 0.99 0.93 0.99 0.96

WBC (k/mL)<4.5 Ref Ref Ref Ref4.5–11.0 0.94 0.94 1.27 1.08>11.0 1.57* 1.52* 1.53* 1.44*

Hematocrit (%)38–44 Ref Ref Ref Ref#37 3.00* 2.81* 3.25* 4.22*$45 0.76 0.48* 0.76 0.67

Operative time (min) 1.01* 1.01* 1.01* 1.01*Procedure

Intermediate Ref Ref Ref RefComplex 1.52* 1.67* 1.86* 1.69*Major 3.18* 2.98* 2.95* 2.56*

*P < .05.Values adjusted for platelet count. All values presented as odds ratios.ASA, American Society of Anaesthetists; CNS, central nervous system; SD, standard deviation; WBCs, white blood cells.

SurgeryVolume 152, Number 3

Al-Refaie et al 349

Table IV. Red blood cell transfusion and adverse operative outcomes, stratified by age categories(N = 38,926)

RBC units transfused per outcome

Age group (years)

<55 56–65 66–74 >75

30-day mortality0 Ref Ref Ref Ref1–2 2.96* 2.30* 1.25* 1.75*>3 6.60* 2.98* 3.00* 2.89*

Major complications0 Ref Ref Ref Ref1–2 1.44* 1.47* 1.34* 1.42*>3 1.83* 2.31* 2.03* 2.47*

Extended DOS0 Ref Ref Ref Ref1–2 1.44* 1.74* 1.50* 1.68*>3 1.95* 2.25* 2.19* 2.52*

Increased total no. of complications0 Ref Ref Ref Ref1–2 1.42* 1.40* 1.29* 1.43*>3 1.77* 1.95* 2.10* 2.23*

No. of patients 9,509 9,518 9,173 10,726

*P < .05.All values presented as odds ratios. Hematocrit level has been removed. Values were also adjusted for sex, race, diabetes, smoking status, American Societyof Anaesthetists score, cardiac comorbidities, functional status, pulmonary comorbidities, central nervous system comorbidities, creatinine levels, albuminlevels, bilirubin levels, white blood cell count, platelet count, and type of procedure (intermediate, complex, and major).DOS, Duration of stay.

SurgerySeptember 2012

350 Al-Refaie et al

measures of short-term operative outcomes aftercancer surgery.

Contrary to recent literature showing a benefi-cial effect of blood transfusion after major noncar-diac surgery in older patients with Hct <24% orsubstantial blood loss, the present study shows thatblood transfusion has a detrimental effect oncancer surgery outcomes in all age groups and inthose with low or normal Hct levels.15-17 Ourresults, and those of Wu et al,17 have alsoconfirmed the previous body of evidence thatrecipients of blood transfusion are likely to besicker, undergo more extensive procedures, orhave longer operative times.17 The present study,however, focuses on patients who were surgicallytreated for solid malignant tumors only, wherethe adverse effects of blood transfusion havebeen previously attributed its immunosuppressiveeffect, a patient’s associated comorbidities,advanced cancer stage, extensive surgical resec-tion, or even circumstances that surrounded theneed for blood transfusions.5,6,9,10

While some literature suggests that the negativeimpact of blood transfusions on operative outcomesmay be a result of the complex interaction betweenpreoperative Hct levels and underlying comorbid-ities, our robust analyses shows that intraoperativeRBC transfusion remains significantly associated

with poorer outcomes independent of these con-founding factors. In our sensitivity analysis, wefound that RBC transfusion was one of the factorsmost strongly associated with adverse events, evenafter adjusting for preoperative comorbidities un-available in other data sources, such as functionalstatus and ASA score. In addition, we show that thisnegative relationship continues even after patientsare stratified by their preoperative Hct level. How-ever, we were unable to confirm this relationship inthose with high preoperative Hct, because thenumber of patients receiving transfusions in thisgroup was small.

In our study, more than 60% of patients receiv-ing blood transfusions were administered only 1 to2 units of RBCs in the absence of intraoperativeevents. This pattern may suggest a trend adoptedby some providers to administer blood intraoper-atively as a volume expander in place of crystalloidor colloids. Such a pattern may also be influencedby the previous literature suggesting that bloodtransfusions have a protective effect in anemic andolder persons. While the ACS NSQIP capturesintraoperative events, it does not collect indica-tions for administering intraoperative blood trans-fusions. In the absence of NSQIP data about bloodtransfusions in the intraoperative period, studies ofpatients in situations of increased physiologic and

Table V. Red blood cell transfusion and adverse operative outcomes, stratified by age and hematocritcategories (N = 38,926)

Adverse outcome

Age group (years)

<55 56–65 66–74 >75

30-day mortalityLow Hct (N = 17,616)

RBC units transfused n = 3,600 n = 3,598 n = 4,085 n = 6,3330 Ref Ref Ref Ref1–2 1.74* 2.28* 1.49* 1.75*>3 5.41* 3.22* 3.06* 2.89*

Normal Hct (N = 16,634)RBC units transfused n = 4,549 n = 4,589 n = 3,978 n = 3,5180 Ref Ref Ref Ref1–2 1.95* 2.68* 0.58 1.63*>3 9.01* 3.80* 3.62* 3.04*

Major complicationsLow Hct (N = 17,616)

RBC units transfused n = 3,600 n = 3,598 n = 4,085 n = 6,3330 Ref Ref Ref Ref1–2 1.27* 1.74* 1.29* 1.39*>3 1.72* 2.49* 1.97* 2.51*

Normal Hct (N = 16,634)RBC units transfused n = 4,549 n = 4,589 n = 3,978 n = 3,518

0 Ref Ref Ref Ref1–2 1.40* 1.09 1.43* 1.50*>3 1.95* 2.09* 2.43* 2.49*

High Hct (N = 3,189)RBC units transfused n = 943 n = 949 n = 780 n = 517

0 Ref Ref Ref Ref1–2 2.91* 1.27 0.87 1.06>3 1.56 4.75* 0.89 2.01

Prolonged DOSLow Hct (N = 15,863)

RBC units transfused n = 3,253 n = 3,229 n = 3,705 n = 5,6760 Ref Ref Ref Ref1–2 1.41* 1.71* 1.47* 1.58*>3 1.82* 2.44* 1.84* 2.49*

Normal Hct (N = 14,404)RBC units transfused n = 3,943 n = 3,922 n = 3,439 n = 3,1000 Ref Ref Ref Ref1–2 1.06 1.75* 1.49* 1.66*>3 1.68* 1.67* 2.79* 2.27*

High Hct (N = 2,641)RBC units transfused n = 778 n = 643 n = 445 n = 775

0 Ref Ref Ref Ref1–2 2.22 0.68 0.67 2.43>3 1.56 6.02* 0.64 1.71

Increased total number of complicationsLow Hct (N = 17,616)

RBC units transfused n = 3,600 n = 3,598 n = 4,085 n = 6,3330 Ref Ref Ref Ref1–2 1.33* 1.58* 1.24* 1.42*>3 1.72* 2.06* 1.93* 2.29*

Normal Hct (N = 16,634)RBC units transfused n = 4,549 n = 4,589 n = 3,978 n = 3,5180 Ref Ref Ref Ref

(continued)

SurgeryVolume 152, Number 3

Al-Refaie et al 351

Table V. (continued)

Adverse outcome

Age group (years)

<55 56–65 66–74 >75

1–2 1.34 1.16* 1.39* 1.47*>3 1.82* 1.93* 2.69* 2.03*

High Hct (N = 3,189)RBC units transfused n = 943 n = 949 n = 780 n = 517

0 Ref Ref Ref Ref1–2 2.15 1.39 0.71 1.30>3 1.39 2.83* 0.91 2.89

No. of patients 9,509 9,518 9,173 10,726

*P < .05.Note that stratified models for high hematocrit levels are not presented because of the small number of mortality events in this subgroup. Values were alsoadjusted for sex, race, diabetes, smoking status, American Society of Anaesthetists score, cardiac comorbidities, functional status, pulmonary comorbid-ities, central nervous system comorbidities, creatinine levels, albumin levels, bilirubin levels, white blood cell count, platelet count, type of procedure(intermediate, complex, and major), and operative time.DOS, Duration of stay; Hct, hematocrit; RBC, red blood cell.

SurgerySeptember 2012

352 Al-Refaie et al

metabolic stress offer insight into the efficacy ofblood transfusion.33-39 The optimum red call massis that red cell mass that provides the greatest oxy-gen delivery to tissues at the lowest energy expenseto the individual.39 While earlier studies have con-cluded that a hemoglobin (Hb) concentration of10 gm/dL is optimum, later studies found thatoxygen delivery in euvolemic subjects remainedat acceptable levels for individuals with Hb concen-trations between 9 and 15 gm/dL.33,36-38 In addi-tion, in a study of hip fracture patientsexamining transfusion therapy in those $60 yearsof age, the authors concluded that transfusion ofRBCs did not change the risk of 30- or 90-day mor-tality.34 In addition, the Transfusion in CriticalCare trial revealed that euvolemic, hemodynami-cally stable patients likely benefit from RBCreplacement when Hb is <7 gm/dL.40 The benefitof transfusion is unlikely when Hb is >9 gm/dL.39

Therefore, our explanation for these transfusionpatterns remains speculative. Importantly, theseblood transfusion patterns must be distinguishedfrom scenarios where by patients are intuitivelytransfused for bleeding solid tumors or unex-pected intraoperative events, which ultimatelylead to poorer cancer surgery outcomes.

The overarching goal of ACS NSQIP is toimprove surgical quality within surgical services atits participating hospitals.21-29 However, there areseveral limitations. First, NSQIP does not collectinformation on receipt of other types of blood pro-ducts or estimated blood loss. Second, becauseNSQIP does not collect data on cancer stagebeyond disseminated cancer, the impact of bloodtransfusion may have been over- or underestimatedacross cancer stages. Finally, transfusions may stillrepresent a marker of illness, advanced cancer

stage, or extensive surgical procedure despite ourrobust multivariate analyses. Nevertheless, ourstudy stems from an important surgical qualityprogram from more than 250 participatinghospitals, thereby reducing the effects of selectionor referral biases encountered in single-hospitalexperiences.41

The current study provides several implicationsfor the practice of blood transfusion during cancersurgery. First, it supports previous practice guide-lines from the ASA task force for perioperativeblood transfusion and alerts members of the sur-gical team to likely recipients of intraoperativeblood transfusions and its additional risk ofadverse operative outcomes.42,43 For example,our results show that persons who are older orwith comorbidities will likely receive intraoperativeRBCs. Our results do not encourage transfusionunless clinically indicated, because no single mea-sure should replace clinical judgment. Second,because of the transfusion-related risks and costs,our findings validate previous recommendations,including careful preoperative patients assessmentand communication between surgical team mem-bers.42 Finally, the finding of more than 60% ofpatients receiving blood transfusions of only 1 to2 units with no documented intraoperative eventsshould encourage future assessments of trendsbehind blood transfusion during cancer surgeryand whether provider-level factors also contributeto these trends, especially when the notion totransfuse for Hct <30% has been refuted morethan 2 decades ago.42,43

In conclusion, the present study reveals thatintraoperative blood transfusions continue toadversely impacts cancer surgery outcomes acrossall age groups and in those with anemia or normal

SurgeryVolume 152, Number 3

Al-Refaie et al 353

Hct levels. While the current study supports previ-ous investigations, it provides insightful implica-tions on the practice of blood transfusion duringcancer surgery and uncovers potential trends inneed of further investigation.

REFERENCES

1. Birgeg�ard G, Aapro MS, Bokemeyer C, Dicato M, Drings P,Hornedo J, et al. Cancer-related anemia: pathogenesis,prevalence and treatment. Oncology 2005;68(Suppl 1):3-11.

2. Birgeg�ard G, Bokemeyer C. New guidelines on anaemiamanagement in patients with cancer: how do these affectclinical practice? Oncology 2005;69(Suppl 2):17-21.

3. Goodnough LT, Brecher ME, Kanter MH, AuBuchon JP.Transfusion medicine. Second of two parts---blood conserva-tion. N Engl J Med 1999;340:525-33.

4. Goodnough LT, Brecher ME, Kanter MH, AuBuchon JP.Transfusion medicine. First of two parts---blood transfusion.N Engl J Med 1999;340:438-47.

5. Sima CS, Jarnagin WR, Fong Y, Elkin E, Fischer M, Wuest D,et al. Predicting the risk of perioperative transfusion for pa-tients undergoing elective hepatectomy. Ann Surg 2009;250:914-21.

6. Al-Refaie WB, Parsons HM, Henderson WG, Jensen EH,Tuttle TM, Vickers SM, et al. Major cancer surgery in the el-derly: results from the American College of Surgeons Na-tional Surgical Quality Improvement Program. Ann Surg2010;251:311-8.

7. Busch OR, Hop WC, Hoynck van Papendrecht MA, Mar-quet RL, Jeekel J. Blood transfusions and prognosis in colo-rectal cancer. N Engl J Med 1993;328:1372-6.

8. Busch OR, Hop WC, Marquet RL, Jeekel J. Blood transfu-sions and local tumor recurrence in colorectal cancer. Evi-dence of a noncausal relationship. Ann Surg 1994;220:791-7.

9. Busch OR, Hop WC, Marquet RL, Jeekel J. The effect ofblood transfusions on survival after surgery for colorectalcancer. Eur J Cancer 1995;31A:1226-8.

10. Busch OR, Marquet RL, Hop WC, Jeekel J. Colorectal can-cer recurrence and perioperative blood transfusions: a crit-ical reappraisal. Semin Surg Oncol 1994;10:195-9.

11. Edna TH, Bjerkeset T. Association between transfusionof stored blood and infective bacterial complications af-ter resection for colorectal cancer. Eur J Surg 1998;164:449-56.

12. Edna TH, Bjerkeset T. Perioperative blood transfusionsreduce long-term survival following surgery for colorectalcancer. Dis Colon Rectum 1998;41:451-9.

13. Fong Y, Karpeh M, Mayer K, Brennan MF. Association ofperioperative transfusions with poor outcome in resectionof gastric adenocarcinoma. Am J Surg 1994;167:256-60.

14. Heiss MM, Mempel W, Delanoff C, Jauch KW, Gabka C,Mempel M, et al. Blood transfusion-modulated tumor recur-rence: first results of a randomized study of autologous ver-sus allogeneic blood transfusion in colorectal cancersurgery. J Clin Oncol 1994;12:1859-67.

15. Wu WC, Rathore SS, Wang Y, Radford MJ, Krumholz HM.Blood transfusion in elderly patients with acute myocardialinfarction. N Engl J Med 2001;345:1230-6.

16. Wu WC, Schifftner TL, Henderson WG, Eaton CB, PosesRM, Uttley G, et al. Preoperative hematocrit levels and post-operative outcomes in older patients undergoing noncar-diac surgery. JAMA 2007;297:2481-8.

17. Wu WC, Smith TS, Henderson WG, Eaton CB, Poses RM,Uttley G, et al. Operative blood loss, blood transfusion,and 30-day mortality in older patients after major noncar-diac surgery. Ann Surg 2010;252:11-7.

18. Hannon T. The contemporary economics of transfusions.In: Spiess BD, Spence RK, Shander A, editors. Periopera-tive transfusion medicine. Philadelphia, PA: LipponcottWilliams and Wilkins; 2005. p. 13-38.

19. Jemal A, Siegel R, Ward E, Hao Y, Xu J, Murray T, et al. Can-cer statistics, 2008. CA Cancer J Clin 2008;58:71-96.

20. Jemal A, Ward E, Hao Y, Thun M. Trends in the leadingcauses of death in the United States, 1970-2002. JAMA2005;294:1255-9.

21. Daley J, Forbes MG, Young GJ, Charns MP, Gibbs JO, Hur K,et al. Validating risk-adjusted surgical outcomes: site visit as-sessment of process and structure. National VA SurgicalRisk Study. J Am Coll Surg 1997;185:341-51.

22. Daley J, Henderson WG, Khuri SF. Risk-adjusted surgicaloutcomes. Annu Rev Med 2001;52:275-87.

23. Daley J, Khuri SF, Henderson W, Hur K, Gibbs JO, BarbourG, et al. Risk adjustment of the postoperative morbidity ratefor the comparative assessment of the quality of surgicalcare: results of the National Veterans Affairs Surgical RiskStudy. J Am Coll Surg 1997;185:328-40.

24. Henderson WG, Daley J. Design and statistical methodol-ogy of the National Surgical Quality Improvement Pro-gram: why is it what it is? Am J Surg 2009;198(5 Suppl):S19-27.

25. Khuri SF, Daley J, Henderson W, Hur K, Demakis J, AustJB, et al. The Department of Veterans Affairs’ NSQIP: thefirst national, validated, outcome-based, risk-adjusted, andpeer-controlled program for the measurement and en-hancement of the quality of surgical care. National VASurgical Quality Improvement Program. Ann Surg 1998;228:491-507.

26. Khuri SF, Daley J, Henderson W, Hur K, Gibbs JO, BarbourG, et al. Risk adjustment of the postoperative mortality ratefor the comparative assessment of the quality of surgicalcare: results of the National Veterans Affairs Surgical RiskStudy. J Am Coll Surg 1997;185:315-27.

27. Khuri SF, Daley J, Henderson W, Hur K, Hossain M, SoybelD, et al. Relation of surgical volume to outcome in eightcommon operations: results from the VA National SurgicalQuality Improvement Program. Ann Surg 1999;230:414-29.

28. Khuri SF, Daley J, Henderson WG. The measurement ofquality in surgery. Adv Surg 1999;33:113-40.

29. Khuri SF, Najjar SF, Daley J, Krasnicka B, Hossain M, Hen-derson WG, et al. Comparison of surgical outcomesbetween teaching and nonteaching hospitals in the Depart-ment of Veterans Affairs. Ann Surg 2001;234:370-82.

30. Al-Refaie WB, Parsons HM, Henderson WG, Jensen EH,Tuttle TM, Rothenberger DA, et al. Body mass index andmajor cancer surgery outcomes: lack of association orneed for alternative measurements of obesity? Ann SurgOncol 2010;17:2264-73.

31. Borja-Cacho D, Parsons HM, Habermann EB, Rothen-berger DA, Henderson WG, Al-Refaie WB. Assessment ofACS NSQIP’s Predictive Ability for Adverse Events AfterMajor Cancer Surgery. Ann Surg Oncol 2010;17:2274-82.

32. Yeh JJ, Gonen M, Tomlinson JS, Idrees K, Brennan MF,Fong Y. Effect of blood transfusion on outcome after pan-creaticoduodenectomy for exocrine tumour of the pan-creas. Br J Surg 2007;94:466-72.

33. Babineau TJ, Dzik WH, Borlase BC, Baxter JK, Bistrian BR,Benotti PN. Reevaluation of current transfusion practices in

SurgerySeptember 2012

354 Al-Refaie et al

patients in surgical intensive care units. Am J Surg 1992;164:22-5.

34. Carson JL, Duff A, Berlin JA, Lawrence VA, Poses RM,Huber EC, et al. Perioperative blood transfusion and post-operative mortality. JAMA 1998;279:199-205.

35. H�ebert PC, Wells G, Tweeddale M, Martin C, Marshall J,Pham B, et al. Does transfusion practice affect mortalityin critically ill patients? Transfusion Requirements in Criti-cal Care (TRICC) Investigators and the Canadian CriticalCare Trials Group. Am J Respir Crit Care Med 1997;155:1618-23.

36. Messmer K. Hemodilution---possibilities and safety aspects.Acta Anaesthesiol Scand Suppl 1988;89:49-53.

37. Messmer K, Lewis DH, Sunder-Plassmann L, Kl€ovekornWP, Mendler N, Holper K. Acute normovolemic hemodi-lution. Changes of central hemodynamics and microcir-culatory flow in skeletal muscle. Eur Surg Res 1972;4:55-70.

38. Shah DM, Gottlieb ME, Rahm RL, Stratton HH, Barie PS,Paloski WH, et al. Failure of red blood cell transfusion toincrease oxygen transport or mixed venous PO2 in injuredpatients. J Trauma 1982;22:741-6.

39. Stubbs J. Transfusion therapy. In: Abrams J, Druck P, CerraFB, editors. Surgical critical care. New York: Taylor andFrancis; 2005. p. 833-93.

40. H�ebert PC, Wells G, Blajchman MA, Marshall J, Martin C,Pagliarello G, et al. A multicenter, randomized, controlledclinical trial of transfusion requirements in critical care.Transfusion Requirements in Critical Care Investigators,Canadian Critical Care Trials Group. N Engl J Med 1999;340:409-17.

41. Hall BL, Hamilton BH, Richards K, Bilimoria KY, CohenME, Ko CY, et al. Does surgical quality improve in the Amer-ican College of Surgeons National Surgical QualityImprovement Program: an evaluation of all participatinghospitals. Ann Surg 2009;250:363-76.

42. American Society of Anesthesiologists Task Force on Periop-erative Blood Transfusion and Adjuvant Therapies. Practiceguidelines for perioperative blood transfusion and adjuvanttherapies: an updated report by the American Society of An-esthesiologists Task Force on Perioperative Blood Transfusionand Adjuvant Therapies. Anesthesiology 2006;105:198-208.

43. Consensus conference. Perioperative red blood cell transfu-sion. JAMA 1988;260:2700-3.