epidural analgesia in hepatic resection

EAMK

Etsepntt

CTaPM

RbFGoSC1

©P

pidural Analgesia in Hepatic Resectionndrew Page, MD, Bradley Rostad, BS, Charles A Staley, MD, FACS, Jerold H Levy, MD, Jaemin Park, BS,ichael Goodman, MD, MPH, Juan M Sarmiento, MD, John Galloway, MD, FACS,eith A Delman, MD, FACS, David A Kooby, MD

BACKGROUND: Randomized trials show equivocal benefit of epidural analgesia (EA) for patients undergoingabdominal operations. Partial hepatectomy is often performed using low central venous pres-sure anesthesia to reduce intraoperative blood loss. We examined effects of pain managementstrategy on blood pressure, transfusion, and complications in patients undergoing hepaticresection with either EA or IV analgesia (IVA).

STUDY DESIGN: Data on patients undergoing hepatectomy from 2001 to 2004 at Emory University Hospitalwere analyzed according to route of perioperative pain management. Patient and treatmentfactors were analyzed for associations with transfusion and morbidity.

RESULTS: From 2001 through 2004, 367 patients underwent elective partial hepatectomy at EmoryUniversity Hospital. EA patients were more likely to be older, men, and with malignancy. Therewere no differences between the groups in extent of resection, operative time, blood loss, orstarting hematocrit level. The EA group had lower mean arterial pressure in recovery(86.6 � 14.0 mmHg versus 94.5 � 13.2 mmHg, p � 0.001) and were more likely to betransfused with packed red cells during the hospital course (44.5% versus 27.9%, p � 0.001).On multivariate analysis, age greater than 65 years, American Society of Anesthesiologistsgrade � 2, starting hematocrit � 38%, operative time � 300 minutes, blood loss � 1 L, anduse of EA were associated with increased numbers of patients receiving packed red blood cells.Complications and length of stay were similar for both groups.

CONCLUSIONS: Epidural analgesia was independently associated with increased risk of packed red blood celltransfusion after hepatectomy. EA did not appear to minimize complications or shorten hos-pital stay. Caution should be exercised when considering EA use in hepatic resection. (J Am Coll
Surg 2008;206:1184–1192. © 2008 by the American College of Surgeons)
msrtcpc

gimpapMnaac

pidural analgesia (EA) provides regional pain controlhrough neuroaxial blockade for many patients undergoingurgical procedures. Sensory nerves are bathed in local an-sthetic, narcotic, or a mixture to provide pain relief whilereserving motor function.1,2 Because EA reduces systemicarcotic requirements, it is associated with less postopera-ive sedation, better postoperative bowel function, and bet-er preservation of pulmonary function through easier pul-

ompeting Interests Declared: None.he Georgia Cancer Coalition provided financial support in preparing this

rticle.resented at the American Hepato-Pancreato-Biliary Association Annualeeting, Las Vegas, NV, April 2007.

eceived September 8, 2007; Revised November 5, 2007; Accepted Decem-er 4, 2007.rom the Departments of Surgery (Page, Rostad, Staley, Park, Sarmiento,alloway, Delman, Kooby) and Anesthesia (Levy), Emory University School

f Medicine, and the Department of Epidemiology, Emory University Rollinschool of Public Health (Goodman), Atlanta, GA.orrespondence address: David A Kooby, MD, Winship Cancer Institute,

m365C Clifton Rd, NE, 2nd Fl, Atlanta, GA 30322.

11842008 by the American College of Surgeons

ublished by Elsevier Inc.

onary toilet and earlier ambulation.3-5 Some reportsuggest EA may shorten stays in intensive care settings andeduce perioperative blood loss.6,7 Because of these poten-ial advantages, EA is used in many medical centers, espe-ially in centers with high-volume cardiothoracic surgeryrograms, because of the debilitating muscle and bone-utting chest incisions used in these procedures.8-10

Although EA use is routine for cardiac and thoracic sur-ical procedures, perioperative pain control practices varyn abdominal procedures for several reasons. First, EA is

ore difficult to administer and requires a coordinatedain management team to place the catheter appropriatelynd then monitor and manage variations in pain control,atient hemodynamic status, and potential adverse effects.any surgeons believe that midline abdominal incisions do

ot merit the hassles of EA. Second, abdominal surgery can bessociated with substantial fluid shifts, and patients may bedversely affected by the reduced sympathetic tone and in-reased venous capacitance associated with EA, ultimately
aking perioperative fluid management more complex.1
ISSN 1072-7515/08/$34.00doi:10.1016/j.jamcollsurg.2007.12.041

ltbpiaasftnvtfrth

MEWRtEcowf

tmodtsmail

mtopva

PPaDrioc

THtgtuswDd9wch

tt(tmaagwCw

RTapabpb

1185Vol. 206, No. 6, June 2008 Page et al Epidural Analgesia in Hepatic Resection

Partial hepatectomy is most often performed througharge, painful, muscle-cutting abdominal incisions and wasraditionally associated with high morbidity and mortalityecause of heavy perioperative blood loss.11 Although im-rovements have been made,12 hepatic resection remains

nherently more complex than other abdominal operationsnd is usually relegated to high-volume institutions. Hep-tectomy is often performed with low central venous pres-ure (CVP) anesthetic management to reduce blood lossrom hepatic venous injury during parenchymal transec-ion.13 This technique mandates maintaining a central ve-ous pressure of 0 to 5 mmHg, mainly by perioperativeolume restriction. Combining low CVP anesthesia withhe neuroaxial blockade of EA may make perioperativeluid management more complex. This study examines theole of epidural analgesia in hepatic resection with respecto transfusion requirements, complications, and length ofospital stay.

ETHODSpidural analgesiaith the approval of the Emory University Institutional

eview Board, inpatient and outpatient records of all pa-ients who underwent elective partial hepatectomy atmory University Hospital from January 1, 2001, to De-ember 31, 2004, were analyzed for demographics, disease,perative factors, and perioperative outcomes. Patientsere categorized according to method of pain control used

or postoperative analgesia.At Emory University Hospital, patients undergoing par-

ial hepatectomy are counseled about postoperative painanagement in the preoperative holding area. They are

ffered the choice of intravenous analgesia (IVA) with aemand pump or epidural analgesia (EA) with preopera-ive catheter placement, unless contraindicated. If EA iselected, informed consent is obtained, and a pain manage-ent team places the catheter at the T10-T12 region and

ssesses its function with a test dose of anesthetic agent. EAs usually started intraoperatively with a combination of

Abbreviations and Acronyms

ASA � American Society of AnesthesiologistsCVP � central venous pressureEA � epidural analgesiaINR � international normalized ratioIVA � intravenous analgesiaMAP � mean arterial pressureOR � odds ratioPRBC � packed red blood cells

ocal anesthetic (0.1% bupivacaine) and an opioid (0.1% p

eperidine). Catheters are assessed daily for proper func-ion and are removed when patients can be transitioned toral pain medication or by day 7 to prevent infection. Ifatients refuse EA, or if EA is contraindicated, IVA is pro-ided through a demand pump, and the most commongent is morphine.

atient factorsatient data included age, gender, body mass index (BMI),nd American Society of Anesthesiologists (ASA) score.iagnostic data were obtained from the final pathology

eport. Diagnoses were classified as benign (eg, bile ductnjury or obstruction, cyst, liver cell adenoma, hemangi-ma, focal nodular hyperplasia, other), or malignant (in-luding primary and secondary cancers).

reatment factorsepatic resections at Emory University Hospital during

he specified dates were performed primarily by five sur-eons with dedicated surgical oncology, hepatobiliary, orransplantation training. Most resections are performednder low CVP anesthetic technique, as previously de-cribed.14 Briefly, there is avoidance of preloading patientsith fluid in the preoperative setting and at induction.uring the resection, the patient is in a 15-degree Tren-

elenburg position, and systolic pressure is maintained at0 mmHg and urine output at a minimum of 20 mL/hour,ith fluid boluses as needed. Rehydration occurs after

ompletion of parenchymal transection and acquisition ofemostasis.Data collected pertaining to the operation included ex-

ent of hepatic resection, with a major resection includinghree or more Couinaud segments; length of operationminutes); blood loss (mL); intraoperative fluid adminis-ration (crystalloid, hetastarch, and albumin); mean CVP;ean recovery room heart rate in beats per minute (BPM);

nd average mean arterial pressure (MAP), both intraoper-tive and in the recovery room. MAP was used as a surro-ate for CVP, because many patients underwent resectionithout central venous lines. Additionally, recording ofVP in the recovery room was inconsistent; MAP readingsere always available.

eporting of transfusionsransfusion information was reported as number of unitsnd as number of patients transfused per hospital stay foracked red blood cells (PRBC), fresh frozen plasma (FFP),nd platelets (PLT). For PRBC, we also examined the num-er of patients transfused on the day of operation (24-houreriod) and the number of patients receiving only 1 or 2 U,ecause this is the group in which red cell transfusion is
otentially avoidable.

CPmApGimamtggwvc

cowtt

SPtudvtcipmiowwI

REFr1sppc(s

ad

PPaPwpesg

LPsfsa3otg(s

fcrtupw(tts4

Tttrto

caoP

1186 Page et al Epidural Analgesia in Hepatic Resection J Am Coll Surg

omplicationsostoperative complications were graded according toethods previously described by Martin and colleagues.15

score of 0 to 5 was assigned based on the most severeostoperative complication experienced by each patient.rade 0 indicates no complications were recorded. Grade 1

ndicates minor complications requiring observation orinor outpatient intervention (eg, warm compresses, oral

ntibiotics). Grade 2 indicates complications requiringoderate intervention (eg, wound debridement or IV an-

ibiotics). Grade 3 indicates major intervention (eg, CT-uided drainage of intraabdominal fluid collection or sur-ery). Grade 4 indicates permanent disability (eg, strokeith residual paresis or paralysis, loss of limb, or chronicentilator dependence) or discharge to a longterm care fa-ility. Grade 5 indicates death.

For the purposes of this study, all complications wereategorized as either “minor,” which included grades 1 to 2,r “major,” which included grades 3 to 5. Complicationsere also categorized by type as cardiac, pulmonary, infec-

ive, and other. Sixty-day mortality was also recorded forhe various treatment groups.

tatistical analysisatients who had epidural analgesia were compared withhose who received other types of pain management bysing chi-square tests for categorical variables and indepen-ent t-tests for continuous variables. For the t-test, equalariances were not assumed, unless significant on Levene’sest. Two-tailed p values � 0.05 were considered statisti-ally significant. Multivariate analyses were performed us-ng binary logistic regression models that controlled forossible confounders. All continuous data fields wereade dichotomous around the median value or accord-

ng to accepted cut-off points in the literature. The resultsf multivariate analyses were expressed as odds ratios (ORs)ith corresponding 95% confidence intervals. All analysesere performed using SPSS 14.0 statistical software (SPSS,

nc).

ESULTSpidural analgesiarom January 2001 through December 2004, 392 hepaticesections were performed at Emory University Hospital;5 of these occurred under urgent or emergent circum-tances and another 10 had inadequate data, leaving 367atients for analysis. Of these, EA was the main method ofain control for 256 (69.8%) patients and IVA was thehoice for 111 (30.2%). Of the 256 epidurals placed, 23692%) were started in the operating room and the rest were
tarted in the recovery room. Catheters were in place for an c
verage of 5.1 � 2.1 days (median 5 days, range 1 to 11ays).

atient and treatment factorsatient, disease, and operative variables for all 367 patientsccording to the method of pain control are shown inTable 1.atients treated with EA were more likely to be older menith a cancer diagnosis and a higher ASA class, as com-ared with patients receiving IVA. There were no differ-nces between the groups in various operative parameters,uch as extent of resection (60% major resection in eachroup), operative time, and operative blood loss (Table 1).

aboratory and transfusion resultsertinent laboratory values are shown in Table 2. As a mea-ure of nutritional status, preoperative mean albumin levelsor both groups were the same. Similarly, there was noignificant difference in preoperative total bilirubin or cre-tinine levels. Preoperative hematocrit was similar (EA,8.3 � 4.6 versus IVA, 38.0 � 5.7%, p � 0.6%), but pre-perative platelet counts and international normalized ra-ios deviated significantly more from the norm in the IVAroup. Postoperative hematocrit was lower in the EA group31.2 � 4.1% versus 33.6 � 4.5%, p � 0.01), despiteimilar blood loss between the groups.

The association between transfusions of PRBC, freshrozen plasma (FFP), and platelets, and method of painontrol, is shown in Table 2. Overall, 39.5% of patientseceived at least 1 unit of blood. There was no difference inhe mean number of units of packed red blood cell (PRBC)nits transfused between EA and IVA patients, but theercentage of EA patients receiving any PRBC transfusionas significantly greater than that for IVA patients

EA � 44.5%, no EA-30.6%, p � 0.01). Interestingly, ofhe patients who received PRBCs during their hospital stay,he percentage who received one or two units of blood wasignificantly higher in the EA groups (65.8% versus5.2%, p � 0.04).FFP was administered to 17.4% of patients in this study.

he mean number of units of FFP transfused was higher inhe IVA group (1.3 � 3.3 versus 0.6 � 0.7, p � 0.01), buthere was no statistical difference in percentage of patientseceiving FFP between the groups. Platelets were adminis-ered to only 16 patients (4.4%), and a higher proportionf these patients were in the IVA group (Table 2).

Results of the multivariate analyses evaluating the asso-iation between pertinent demographic and clinical factorsnd blood transfusion are shown in Table 3. Factors dem-nstrating a significant independent association with use ofRBC transfusion in this study were age � 65 years, ASA
lass � 2, hematocrit � 38%, use of epidural analgesia,

ol

CTmewqadmadmn(p

ip

mmbp

EoPmch3ppt1iHpt

TV

AG

BD

A

R

OOF

IIPP

AA minu


perative time � 300 minutes, and greater than 1 L bloodoss at operation.

omplications and length of stayable 4 shows postoperative complications according toethod of pain management. Overall, 146 patients (40%)

xperienced a postoperative complication; many of theseere minor complications; only 80 patients (21.8%) re-uired substantial intervention (ie, major complications) asresult of their complications. There was no significant

ifference in number of patients experiencing any major orinor complications between the EA and IVA groups. In

nalyzing specific types of event, there were no statisticalifferences between the groups: cardiac (p � 0.12), pul-onary (p � 0.84), or infective (p � 0.59).There was also

o difference in hospital length of stay between the groupsEA, 10.0 � 7.1 days versus IVA, 10.0 � 9.0 days,� 0.74).In the multivariate analyses, factors demonstrating an

ndependent association with development of major com-

able 1. Patient and Treatment Variablesariable Epidural, n � 25

ge, y 54 � 14ender, n (%)Female 116 (45.3)Male 140 (54.7)

ody mass index, kg/m2 27.3 � 6.2iagnosis, n (%)Benign 63 (24.6)Malignant 193 (74.6)

SA class, n (%)1 4 (1.6)2 95 (37.1)3 147 (57.4)4 10 (3.9)

esection type, n (%)Minor 102 (39.8)Major 154 (60.2)perative time, min 305 � 123perative blood loss, mL 709 � 694

luids given at operation, mLCrystalloid 4,054 � 2,294.1Hetastarch 348 � 456.2Albumin 562 � 712.3

ntraoperative CVP 8.2 � 3.1ntraoperative MAP 81.0 � 6.7ostoperative MAP 86.6 � 14.0ostoperative heart rate, BPM 88.0 � 5.1

ll results are expressed as either n (%) or mean � SD.SA, American Society of Anesthesiologists’ physical status; BPM, beats per

lications included patient age greater than 65 years, loss of 0

ore than 1 L of blood at operation, and undergoing aajor resection. Choice of perioperative analgesia did not

ear independent association with development of majorostoperative complications.

ffects of packed red blood cell transfusionn outcomesatients who received PRBC transfusions (n � 145) wereore likely to experience at least 1 postoperative compli-

ation than patients who were not transfused during theirospitalization (Table 5). Similarly, the frequency of gradeor higher (major) complications was higher in transfusedatients (34.7% versus 13.5%, p � 0.001). Infective com-lications were twice as common in transfused patientshan in their nontransfused counterparts (33.3% versus5.7%, p � 0.001). Similarly, 60-day mortality was highern transfused patients (6.9% versus 0.9%, p � 0.002).

ospital length of stay was nearly double among transfusedatients, when compared with that of nontransfused pa-ients (14.0 � 10.3 days versus 7.7 � 4.1 days, p �

No epidural, n � 111 p Value

50 � 16 0.010.02

74 (66.6)37 (33.3)

27.5 � 6.9 0.79� 0.001

58 (52.3)53 (47.7)

� 0.00115 (13.5)44 (39.6)42 (37.8)10 (9.0)

0.9045 (40.5)66 (59.5)

322 � 169 0.35780 � 1,037 0.51

3,909 � 2,505.3 0.61207 � 369.1 0.01527 � 783.3 0.699.0 � 3.0 0.03

83.9 � 8.2 0.00194.5 � 13.2 � 0.00191.2 � 6.2 0.07

te; CVP, central venous pressure; MAP, mean arterial pressure.

6

.001).

DTbtwdwa

tttcgdpb

o

nttrceit

rhe5iAtia

TV

P

T

A*P


ISCUSSIONhis retrospective analysis assessed perioperative outcomesy method of pain management for elective hepatic resec-ion. Analysis of our data showed EA use was associatedith a greater likelihood of perioperative blood transfusion,espite similar operative blood loss among patients managedith EA and IVA. Importantly, the only modifiable variable

ssociated with transfusion practices was EA use.Although early experiences with elective hepatic resec-

ion were associated with tremendous blood loss and mor-ality,16 improvements in anatomic understanding, surgicalechnique, anesthetic management, and transfusion medi-ine have resulted in better outcomes after hepatic sur-ery.12 Unfortunately, approximately 50% of patients un-ergoing partial hepatectomy will still receive some bloodroduct in the perioperative period, and most of these wille red cell transfusions.17

Transfusions are consistently associated with adverse

able 2. Perioperative Laboratory Values and Transfusion Rariable Timing

erioperative laboratory valuesAlbumin, g/dL PreoperativeHematocrit, %

PreoperativePostoperative

Platelet count, �103 mclPreoperativePostoperative

International normalized ratioPreoperativePostoperative

Total bilirubin, mg/dLPreoperativePostoperative

Creatinine, mg/dLPreoperativePostoperative

ransfusion resultsPRBC, units Hospital stayPatients receiving PRBC, n (%)

Operation daHospital stay1-2 U*

Fresh frozen plasma, U Hospital stayPatients receiving fresh frozen plasma, n (%) Hospital stayPlatelets, U Hospital stayPatients receiving platelets, n (%) Hospital stay

ll results are expressed as either n (%) or mean values.The denominator for this variable is the total number of transfused patientsRBC, packed red blood cells

utcomes in operations, especially in situations involving d

eoplastic disease.16,18-24 ABO incompatibility, hypersensi-ivity, and infection with pathogens are the most life-hreatening adverse reactions. More subtle concerns sur-ound resection for malignancy, where data suggest thatancer outcomes are worse in patients receiving periop-rative transfusions, perhaps because of compromisedmmunity associated with allogeneic blood productransfusion.25-27

One significant advance aimed at reducing transfusionequirements and improving outcomes in hepatic resectionas been incorporation of low central venous pressure an-sthetic technique.13,14 By maintaining low CVP (0 tommHg), hepatic venous bleeding can be minimized dur-

ng the parenchymal transection phase of the operation.28

fter resection, euvolemia is restored by the anesthesiaeam. A key component of the low CVP approach is avoid-ng overhydrating the patient in the preoperative holdingrea and at induction. The medical sympathectomy in-

tsEpidural, n � 256 No epidural, n � 111 p Value

53.5 � 0.5 3.5 � 0.6 0.57

38.3 � 4.6 38.0 � 5.7 0.6731.2 � 4.1 33.6 � 4.5 � 0.001

270.8 � 99.8 264.0 � 81.8 0.50196.2 � 73.8 216.4 � 72.3 0.02

1.03 � 0.1 1.06 � 0.1 � 0.0011.55 � 0.5 1.80 � 1.3 0.07

0.89 � 1.0 0.88 � 0.9 0.922.45 � 4.3 2.63 � 2.9 0.64

0.86 � 0.2 0.86 � 0.2 0.990.86 � 0.3 0.83 � 0.3 0.30

1.2 � 1.8 1.3 � 3.2 0.55

81 (31.6) 27 (24.3) 0.17114 (44.5) 31 (27.9) 0.00

75/114 (65.8) 14/31 (45.2) 0.040.6 � 01.7 1.3 � 3.3 � 0.01

40 (15.6) 24 (21.6) 0.180.0 � 0.2 0.2 � 0.7 � 0.01

7 (2.7) 9 (8.1) 0.02

roup.

esul

y

per g

uced by EA relaxes vascular smooth muscle, produces

vrhfahe

Itcaodt4Hbn

sIg

a4gttet

tTI3poshehpbust

T

V

ABAMPPMEOBR

A reope

TV

AMS

6L


enodilation, and increases venous capacitance.29 This mayesult in relative hypotension, which potentially induces aigher rate of fluid administration,30 a cycle that may beurther aggravated by venodilation at induction of generalnesthesia. Low CVP anesthetic technique, in the face ofypotension, may trigger more transfusions. This hypoth-sis was the motivation for this study.

To date, some randomized trials have compared EA withVA in patients undergoing abdominal operations. In 1rial, investigators randomly assigned 1,021 patients to re-eive both general anesthesia and postoperative parenteralnalgesia or epidural plus light general anesthesia and post-perative EA. The primary end points for this report wereeath at 30 days and postoperative complications. Overall,here were no differences in 30-day death (3.4% IVA versus.0% EA) or specific complications between the groups.epatectomy was not specifically addressed in the report,

ut the subset that underwent “biliary” operations showedo differences in mortality or complications.31

The Multicenter Australian Study of Epidural Anesthe-ia Trial (MASTER) provided essentially identical results.3

n this study, investigators randomized 915 patients under-oing major abdominal operations to EA versus IVA and

able 3. Multivariate Analysis for Factors Associated with R

ariable

Red cell transfusio

p Value Odds ratio

ge � 65 y 0.02 2.21MI � 30 kg/m2 0.33 0.14SA � 2 0.00 3.26alignant disease 0.13 1.76

reop albumin � 3.5 g/dL 0.22 1.47reop hematocrit � 38% 0.00 4.52ajor resection 0.96 1.01

pidural analgesia 0.00 3.64perative time � 300 min 0.00 3.05lood loss � 1 L 0.00 18.90ed cell transfusion — —

SA, American Society of Anesthesiologists; BMI, body mass index; preop, p

able 4. Postoperative Complication Profiles According to Mariable Epidural, n � 256

ny complication, n (%) 107 (41.8)ajor complications, n (%) 61 (23.8)

pecific complications, n (%)Cardiac 12 (4.7)Pulmonary 21 (8.2)Infective 60 (23.4)

0-d mortality, n (%) 7 (2.7)
ength of stay, d 10.0 � 7.1
gain found no overall differences in 30-day mortality (IVA.3% versus EA 5.1%) or complications between theroups. There was a 7% reduction in respiratory failure inhe EA arm. Again, hepatic resection was not specified inhis report. Existing data concerning hepatic resection andpidural analgesia are limited and do not address the ques-ions raised in this report.32-34

For patients undergoing hepatic resection at our institu-ion, epidural anesthesia was used in the majority (69.8%).here were several notable differences between the EA and

VA groups. Almost all the EA patients were ASA class 2 or(94.5%), with most being class 3 (57.4%); a substantial

ercentage of IVA patients were either ASA class 1 (13.5%)r class 2 (9%). This difference in distribution of healthtatus may reflect biases of both patients and treatingealth-care professionals. Healthy patients may have anxi-ty regarding the potential complications of EA (eg, spinaleadaches or epidural hematomas with resultant paresis oraralysis), and patients with several comorbid factors maye more likely to accept epidural catheter placement andse because they are more prone to pulmonary failure fromplinting and atelectasis secondary to pain-induced reduc-ion in respiratory effort. We acknowledge that there is

ell Transfusion and Major ComplicationsMajor complication

nfidencerval (95%) p Value Odds ratio

Confidenceinterval (95%)

2–4.35 0.05 1.84 1.01–3.382–2.70 0.14 1.62 0.86–3.055–6.42 0.32 1.38 0.73–2.584–3.67 0.89 0.95 0.48–1.900–2.73 0.75 0.91 0.50–1.660–8.55 0.71 0.89 0.49–1.635–1.87 0.02 2.02 1.10–3.713–7.66 0.35 1.36 0.71–2.616–5.62 0.29 1.38 0.76–2.495–41.47 0.03 2.15 1.09–4.25

— 0.15 1.65 0.83–3.27

rative.

d of Pain ManagementNo epidural, n � 111 p Value

39 (35.1) 0.2519 (17.1) 0.17

1 (0.9) 0.128 (7.2) 0.84

23 (20.7) 0.595 (4.5) 0.36

ed Cn

Cointe

1.10.71.60.80.82.40.51.71.68.6

etho

10.0 � 9.3 0.74

php

aswpiedmnt

stihblgiltwlpI

ttdwnhpsgtp

mctrgwtce

wImttpht

tebiesssTdraecbsisb

t

TV

AMS

6L


robably a lower threshold used to transfuse patients withigher ASA scores in general, although there is no officialolicy regarding this at our institution.Another interesting difference was the greater percent-

ge of men treated with EA. Perhaps the women in thistudy group had more anxiety over complications of EA orere more stoic with respect to anticipated postoperativeain. There were equal percentages of major hepatectomiesn both groups, but more EA patients had malignant dis-ase than did those treated with IVA. Having a malignantiagnosis correlated with having a higher ASA score, whichay explain this difference. Ultimately, the retrospective

ature of this study makes these differences difficult toease out.

Concerning the preoperative laboratory values, it is noturprising that IVA patients had a higher baseline interna-ional normalized ratio, because coagulopathy is a contra-ndication for epidural placement. Although preoperativeematocrit and intraoperative blood loss were no differentetween the two groups, postoperative hematocrit was

ower in the EA patients. We surmise this is because of areater volume of fluid administration in both the operat-ng room and recovery room in response to significantlyower MAPs in both locations. Our analysis failed to cap-ure recovery room fluid administration, because chartingas less meticulous in this setting. Intraoperative crystal-

oid administration was similar between the groups, but EAatients were given 26.6% more colloid on average thanVA patients.

The interplay of blood pressure, systemic vascular resis-ance, and fluid administration is complex. Because 75% ofhe systemic blood is in the venules, combined with theecreased amount of smooth muscle in venules comparedith arteries, the effect of epidural anesthesia is pro-ounced in the venous system.1,2 Patients with epiduralsad statistically lower MAPs in the operating room thanatients without epidurals, but the actual difference wasmall; the difference in recovery room MAPs between theroups was much greater. There are two possible explana-ions for this finding. First, in the operating room, blood

able 5. Postoperative Complication Profiles According to Tariable Transfused, n � 145

ny complication, n (%) 80 (55.6)ajor complications, n (%) 50 (34.7)

pecific complications, n (%)Cardiac 6 (4.2)Pulmonary 14 (9.7)Infective 48 (33.3)

0-d mortality, n (%) 10 (6.9)ength of stay, d 14.0 � 10.3

ressure control is extremely tight and can be manipulated t

inute to minute with pharmacologic intervention; in re-overy, the control is usually less precise. Additionally, al-hough most of the epidurals were dosed in the operatingoom, with intention of transitioning the patients fromeneral anesthesia to the awake state with less pain, theyere usually dosed toward the end of the operation, after

he blood-losing portion of the operation was presumedomplete, so the impact on blood pressure of EA was mostvident in recovery.

Because of the difference in MAPs between the groups,e also examined mean heart rate in the recovery room.

nterestingly, the mean heart rate was three beats perinute higher in the IVA group, a difference that trended

oward statistical significance. A possible explanation forhis observation could rest in quality of pain control. EAatients may have been more comfortable and, as a result,ad slightly lower heart rates when compared with those inhe IVA group.

Overall, the association beween EA use and red cellransfusion was significant, suggesting that EA is a periop-rative risk factor for transfusion. This finding is unique,ecause this is the first study to examine EA use exclusivelyn the context of hepatic resection. Our findings are differ-nt from results of other, more generalized studies that haveuggested that epidurals are protective against transfu-ions.35 We believe the hemodynamic issues of hepatic re-ection cannot be generalized to all abdominal operations.he total number of red blood cell units transfused was notifferent between the groups, but the number of patientseceiving blood during the hospital stay was 16.6% highermong EA patients. Even more impressive was the differ-nce in number of patients who received low-volume redell transfusions (1 or 2 U), which, again, may be explainedy hemodilution in response to lower systemic blood pres-ure in the postoperative setting. Among factors bearingndependent association with use of red blood cell transfu-ion in this study, epidural use was the only one that coulde changed in the perioperative period.Volume expansion without blood product administra-

ion has been examined. Acute normovolemic hemodilu-

fusion HistoryNot transfused, n � 222, n (%) p Value

66 (29.6) � 0.00130 (13.5) � 0.001

7 (3.1) 0.7715 (6.7) 0.3335 (15.7) � 0.001

2 (0.9) 0.0027.7 � 4.1 � 0.001

rans

ion has been evaluated as a potential tool for reducing

teplhtesOndfoAmh

hcpaotastmslcweds

htcmtgmfut

clsbm

cmmm

EfrpdpEecotctcpi

hcplbafctrc

A

SAA

DC

R


ransfusion requirements in potentially high blood loss op-rations.36 Preoperatively, whole blood is drawn off theatient and the acute volume loss is restored with crystal-

oid or colloid. Subsequent surgical blood loss is of reducedematocrit, and, after the blood loss phase of the operation,he original blood is reinfused. In practice, negligible ben-fit is provided by this technique, as assessed on metaanaly-is, and allogeneic transfusion is not reliably reduced.37

ne prospective report demonstrated a benefit of acuteormovolemic hemodilution in hepatic resection by stan-ardizing a hemodilution target hematocrit and thresholdor transfusion, reducing transfusion risk and improvingverall outcomes, but the study population was limited toSA I/II patients.38 The role of acute normovolemic he-odilution on higher-risk patients and regional anesthesia

as not been evaluated and deserves further examination.We did not observe any reduction in complications or

ospital length of stay in patients managed with epiduralatheters over those without them. Our data also supportrevious assertions that transfusion alone may negativelyffect outcomes when defined as complications and lengthf stay for patients undergoing hepatic resection. In addi-ion, our data suggest that patients receiving transfusionsre at increased risk of infective complications (eg, surgicalite infections, pneumonia, urinary tract infection, and in-raabdominal abscess). As expected, the factor that had theost influence on transfusion requirement was blood loss,

o red cell transfusion may be a surrogate marker for bloodoss to some extent. In the multivariate analysis for majoromplications, blood loss demonstrated direct associationith complications, but red cell transfusion did not. If

ither factor (red cell transfusion or blood loss) wereropped from the analysis, the remaining factor was highlyignificant (data not shown).

Limitations of this study include a retrospective design,eterogeneity in our patient populations, and an inconsis-ent selection for EA versus IVA. We did not capture painontrol data or patient satisfaction issues, which are im-ensely important. In addition, there was no standard

ransfusion trigger in this study. Even when in place, trig-ers are imperfect, because decision to transfuse can have soany variables involved. There were no standard protocols

or EA or IVA delivery with respect to timing or agent(s)sed, which, again, speaks to the retrospective nature ofhis evaluation.

It remains to be examined if manipulation of EA deliveryould decrease the need for blood transfusion. Avoidingocal anesthetic might reduce epidural-related hypoten-ion, and prompt holding of the epidural infusion com-ined with colloid infusion in response to hypotension
ight reduce the need for blood. Another measure to de-
rease transfusion requirements may rest in the develop-ent of transfusion consensus, supported by participatingembers of the surgical, anesthetic, and intensive careanagement team.In summary, this study is the first to examine the role of

A in hepatic resection, and it demonstrates the importantinding that EA is an independent factor associated withed cell transfusion in patients undergoing elective he-atic resection. In our study, EA did not appear to re-uce complications or hospital length of stay in thisatient population, when compared with IVA. AlthoughA can provide optimal pain management, the influ-nces of EA on transfusion and absence of benefit con-erning complications and length of stay in the contextf hepatic resection deserve critical analysis. Althoughhe interactive role of EA on pain, transfusions, compli-ations, and length of stay cannot be known with cer-ainty in a retrospective analysis, we believe the benefi-ial perioperative effects of EA are balanced and areotentially outweighed by the negative outcomes result-ng from increased transfusions.

Our results indicate that among patients undergoingepatic resection, epidural anesthesia is associated with in-reased frequency of blood transfusion in the perioperativeeriod, which cannot be attributed to intraoperative blood

oss. Any proposed benefit of epidurals on pain control maye potentially offset by the increased risk of transfusion. Ofll factors demonstrating association with increased trans-usion, the use of epidural analgesia is the only factor thatan be changed preoperatively in patients undergoing elec-ive partial hepatectomy. We recommend caution in theoutine use of epidurals in hepatectomy, despite the ac-epted benefits in pain control.

uthor Contributions

tudy conception and design: Page, Rostad, Koobycquisition of data: Page, Rostad, Park, Koobynalysis and interpretation of data: Page, Rostad, Staley, Levy,

Park, Goodman, Sarmiento, Galloway, Delman, Koobyrafting of manuscript: Page, Rostad, Koobyritical revision: Page, Rostad, Staley, Levy, Park, Goodman,

Sarmiento, Galloway, Delman, Kooby

EFERENCES

1. Liu S, Carpenter RL, Neal JM. Epidural anesthesia and analge-sia. Their role in postoperative outcome. Anesthesiology 1995;82:1474–1506.

2. Richman JM, Wu CL. Epidural analgesia for postoperative pain.
Anesthes Clin North Am 2005;23:125–140.

1

1

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

2

3

3

3

3

3

3

3

3

3


3. Rigg JR, Jamrozik K, Myles PS, et al. Epidural anaesthesia andanalgesia and outcome of major surgery: a randomised trial.Lancet 2002;359:1276–1282.

4. Brimioulle S, Vachiery JL, Brichant JF, et al. Sympathetic mod-ulation of hypoxic pulmonary vasoconstriction in intact dogs.Cardiovasc Res 1997;34:384–392.

5. Warner DO, Warner MA, Ritman EL. Human chest wall func-tion during epidural anesthesia. Anesthesiology 1996;85:761–773.

6. Rosberg B, Fredin H, Gustafson C. Anesthetic techniques andsurgical blood loss in total hip arthroplasty. Acta Anaesthesio-logica Scandinavica 1982;26:189–193.

7. Keith I. Anaesthesia and blood loss in total hip replacement.Anaesthesia 1977;32:444–450.

8. Lundstrom LH, Nygard E, Hviid LB, et al.The effect of thoracicepidural analgesia on the occurrence of late postoperative hypox-emia in patients undergoing elective coronary bypass surgery: arandomized controlled trial. Chest 2005;128:1564–1570.

9. van de Watering LM, Hermans J, Houbiers JG, et al. Beneficialeffects of leukocyte depletion of transfused blood on postoper-ative complications in patients undergoing cardiac surgery: arandomized clinical trial. Circulation 1998;97:562–568.

0. Veering BT, Cousins MJ. Cardiovascular and pulmonary effectsof epidural anaesthesia. Anaesthes Int Care 2000;28:620–35.

1. Foster JH, Berman MM. Solid liver tumors. Major Prob ClinSurg 1977;22:1–342.

2. Jarnagin WR, Gonen M, Fong Y, et al. Improvement in periop-erative outcome after hepatic resection: analysis of 1,803 con-secutive cases over the past decade. Ann Surg 236:397–406,2002; discussion 406–407.

3. Melendez JA, Arslan V, Fischer ME, et al. Perioperative outcomesof major hepatic resections under low central venous pressure anes-thesia: blood loss, blood transfusion, and the risk of postoperativerenal dysfunction. J Am Coll Surg 1998;187:620–625.

4. Cunningham JD, Fong Y, Shriver C, et al. One hundred con-secutive hepatic resections. Blood loss, transfusion, and opera-tive technique. Arch Surg 1994;129:1050–1056.

5. Martin RC, Brennan MF, Jaques DP. Quality of complication re-porting in the surgical literature. Ann Surg 2002;235:803–813.

6. Foster RS Jr, Costanza MC, Foster JC, et al. Adverse relationshipbetween blood transfusions and survival after colectomy for co-lon cancer. Cancer 1985;55:1195–1201.

7. Kooby DA, Stockman J, Ben-Porat L, et al. Influence of trans-fusions on perioperative and long-term outcome in patients fol-lowing hepatic resection for colorectal metastases. Ann Surg2003; 237:860–869; discussion 869–870.

8. Burrows L, Tartter P. Effect of blood transfusions on colonicmalignancy recurrent rate. Lancet 1982;2:662.

9. Rosenberg SA, Seipp CA, White DE, Wesley R. Perioperativeblood transfusions are associated with increased rates of recur-rence and decreased survival in patients with high-grade soft-tissue sarcomas of the extremities. J Clin Oncol 1985;3:698–709.

0. Hyman NH, Foster RS Jr, DeMeules JE, Costanza MC. Bloodtransfusions and survival after lung cancer resection. Am J Surg
1985;149:502–507.
1. Tartter PI, Burrows L, Papatestas AE, et al. Perioperative bloodtransfusion has prognostic significance for breast cancer. Surgery1985;97:225–230.

2. Tartter PI. Pretreatment prognostic factors in colorectal cancerpatients with synchronous liver metastases. Eur J Surg Oncol1987;13:485–491.

3. Francis DM, Judson RT. Blood transfusion and recurrence ofcancer of the colon and rectum. Br J Surg 1987;74:26–30.

4. Fong Y, Karpeh M, Mayer K, Brennan MF. Association of peri-operative transfusions with poor outcome in resection of gastricadenocarcinoma. Am J Surg 1994;167:256–260.

5. Schechter GP, Soehnlen F, McFarland W. Lymphocyte response toblood transfusion in man. N Engl J Med 1972;287:1169–1173.

6. Kaplan J, Sarnaik S, Gitlin J, Lusher J. Diminished helper/suppressor lymphocyte ratios and natural killer activity in recip-ients of repeated blood transfusions. Blood 1984;64:308–310.

7. Gafter U, Kalechman Y, Sredni B. Induction of a subpopulationof suppressor cells by a single blood transfusion. Kidney Int1992;41:143–148.

8. Wang WD, Liang LJ, Huang XQ, Yin XY. Low central venouspressure reduces blood loss in hepatectomy. World J Gastroen-terol 2006;12:935–939.

9. Shimosato S. Circulatory adjustments under spinal anesthesia.International Anesthesiol Clin 1974;12:127–144.

0. Donald JR. Epidural anaesthesia, induced hypotension andblood loss during surgery. Acta Anaesthesiologica Scandinavica1983;27:91–93.

1. Park WY, Thompson JS, Lee KK. Effect of epidural anesthesiaand analgesia on perioperative outcome: a randomized, con-trolled Veterans Affairs cooperative study. Ann Surg 2001;234:560–569; discussion 569–571.

2. Ho AM, Karmakar MK, Cheung M, Lam GC. Right thoracicparavertebral analgesia for hepatectomy. Br J Anaesthes 2004;93:458–461.

3. Schumann R, Zabala L, Angelis M, et al. Altered hematologicprofiles following donor right hepatectomy and implications forperioperative analgesic management. Liver Transplant 2004;10:363–368.

4. Cywinski JB, Parker BM, Xu M, Irefin SA. A comparison ofpostoperative pain control in patients after right lobe donorhepatectomy and major hepatic resection for tumor. Anesth An-alges 2004;99:1747–1752.

5. Richman JM, Rowlingson AJ, Maine DN, et al. Does neuraxialanesthesia reduce intraoperative blood loss? A meta-analysis.J Clin Anesth 2006;18:427–435.

6. Laks H, Pilon RN, Anderson W, O’Connor NE. Acute normo-volemic hemodilution with crystalloid vs colloid replacement.Surg Forum 1974;25:21–22.

7. Bryson GL, Laupacis A, Wells GA. Does acute normovolemichemodilution reduce perioperative allogeneic transfusion? Ameta-analysis. The international study of perioperative transfu-sion. Anesth Analges 1998;86:9–15.

8. Matot I, Scheinin O, Jurim O, Eid A. Effectiveness of acutenormovolemic hemodilution to minimize allogeneic bloodtransfusion in major liver resections. Anesthesiology 2002;97:
794–800.

epidural analgesia in hepatic resection

Documents