Systemic Leukofiltration Does Not Attenuate PulmonaryInjury after Cardiopulmonary Bypass

OLIVER J. WARREN, CAROLINE R. TUNNICLIFFE, RACHEL M. MASSEY, SOPHIE WALLACE, ANDREW J. SMITH,ELIZABETH M. H. ALCOCK, ARA DARZI, CHARLES A. VINCENT, AND THANOS ATHANASIOU

Pulmonary injury mediated by activated leukocytes is a rec-ognized complication of cardiopulmonary bypass. The aim ofthis paper is to systematically analyze the effects of systemicleukofiltration within the cardiopulmonary bypass circuit onpulmonary injury and related clinical outcomes. We per-formed a systematic search to identify randomized controlledtrials reporting on the effects of systemic leukofiltration onrespiratory parameters. Random effect meta-analytical tech-niques were applied to identify differences in outcomes be-tween the two groups. Sensitivity and subgroup analyses wereundertaken to evaluate study heterogeneity. Incorporating995 patients, 21 studies satisfied the inclusion criteria. Sys-temic leukofiltration significantly increased the PaO2/FiO2

ratio within 12 hours of bypass cessation, (weighted meandifference (WMD), 25.97; 95% confidence interval (CI),3.41–48.53; p � 0.02) but this effect was lost by 24 hours(WMD, 12.98; 95% CI, �7.93–33.89; p � 0.22). Leukofiltra-tion significantly reduced the duration of ventilatory supportpostoperatively (WMD, �2.11 hours; 95% CI, �0.65 to�3.58; p � 0.005), but had no impact on postoperative chestinfection, intensive care length of stay or hospital length ofstay. The heterogeneity of the included studies was high, dueto poor quality study design and failure to include patients athigh risk of pulmonary complications. Systemic leukofiltra-tion may attenuate bypass-related lung injury in the earlypostoperative period, but this does not seem to translate toclinically significant differences in outcomes. ASAIO Journal2008; 54:78-88.

Despite the development of off-pump coronary artery bypassgrafting, minimal access techniques and robotic surgery, car-diopulmonary bypass (CPB) remains an essential componentof most cardiac surgical procedures. Although the practice ofCPB has improved dramatically since its advent in the 1950s,1

it continues to be associated with an undesirable systemicinflammatory response,2 initiated through a variety of mecha-nisms including contact activation (exposure of blood to thenon-biological materials of the CPB circuit) and aortic crossclamping induced ischemia-reperfusion injury.3 These initiat-

ing factors result in the activation, generation or expression ofmultiple inflammatory pathways, including the cellular immunesystem, cytokines, complement and coagulation-fibrinolysis cas-cades, nitric oxide and adhesion molecules.4 The inherent redun-dancy within these varying inflammatory systems means thatprofound amplification may occur, with the potential for multi-organ dysfunction syndrome, manifesting clinically as coagu-lopathy, myocardial dysfunction, renal failure, neurocognitivedeficit and pulmonary injury.

The activation of leukocytes, in particular neutrophils, is animportant part of this inflammatory process,5–8 which has beendirectly implicated in the development of postoperative respi-ratory complications.9 CD18 and CD11b surface adhesionmolecules mediate neutrophil adherence to pulmonary endo-thelium, an essential step that when antagonized diminishespost-CPB pulmonary injury in an animal model.10,11 Onceattached, leukocytes transmigrate into the lung parenchymaunder the influence of IL-8, where they sequester and releaseoxygen free radicals and proteolytic enzymes (e.g., elastase,collagenase) causing cellular and parenchymal lung inju-ry.12–14 The resulting pulmonary dysfunction is an importantcause of morbidity and mortality in cardiac surgery patients,with 5.5% of patients requiring prolonged ventilation,15 and2% of cases developing adult respiratory distress syndrome,itself carrying a 50% mortality rate.10

Knowledge of this pathophysiological process led to thedevelopment of a variety of strategies to reduce leukocyteactivation and sequestration within the lung parenchyma.Pharmacological approaches include the perioperative admin-istration of aprotinin, a non-specific serine protease inhibitorshown to decrease CPB-induced leukocyte activation,16,17 andpentoxifylline, a non-specific phosphodiesterase inhibitor,demonstrated to decrease both neutrophil activation18 andpulmonary sequestration.19 Technical strategies, such as tem-perature manipulation and heparin coated circuits have alsohad mixed trial results.20 Over the last fifteen years, concurrentleukocyte removal from the perfusate using specialized filtershas been proposed as a means of reducing CPB-associatedinflammation. Developed in the early 1990s from the technol-ogy already available to filter allogeneic blood transfusions,leukocyte filtration seemed a logical technical intervention toameliorate pulmonary injury, by depleting one of the keycausative factors. Early animal studies supported this hypoth-esis,21–23 and thus clinical trials on leukodepleting filters havenow been performed at multiple sites within the CPB circuit.Although arterial line filtration is the most frequent,24,25 ve-nous line,26 blood cardioplegia line,27 and re-infused circuitblood filtration28 have all been tested. Beyond this, otherinvestigators have created more intricate leukodepletion strat-

From the Department of BioSurgery and Surgical Technology, Im-perial College, London, United Kingdom.

Submitted for consideration July 2007; accepted for publication inrevised form September 2007.

Reprint Requests: Oliver Warren, BSc (Hons), MRCS (Eng), ClinicalResearch Fellow, Department of BioSurgery and Surgical Technology,Imperial College, 10th Floor QEQM Building, St. Mary’s Hospital,London, W2 1NY, United Kingdom.

DOI: 10.1097/MAT.0b013e3181618e9b

ASAIO Journal 2008

78

egies by altering the timing of filtration,29,30 the temperature atwhich it occurs,31 or combining filtration techniques with eachother32 or various pharmacological interventions.33

Results thus far have been conflicting; the effect of leukode-pleting filters on CPB-associated lung injury remains unclear.Although many studies have reported no impact on postoper-ative pulmonary function,24,34–37 others have described signif-icant improvements in early postoperative indices, such asoxygenation index and time to extubation.38,39 Thus, any po-tential effect that leukofiltration may have on lung injury needselucidating. The aim of this systematic review and meta-anal-ysis was to determine the potential effect of systemic leukofil-tration on pulmonary injury and clinical outcomes in patientsundergoing CPB-assisted cardiac surgery.

Materials and Methods

Literature Search

A literature search was performed using Medline, Ovid, Em-base, Google Scholar, Web of Science, and Cochrane databases.All comparative studies published on leukocyte filtration withinthe CPB circuit, from database inception to early 2007, weresought. The following mesh headings were used: “LeukocyteReduction Filtration,” “Leukocyte Depletion,” “Cardiac Surgery,”“Cardiac,” “Leukocyte,” “Comparative Study,” “Respiratory In-fection,” and “Respiratory System.” The “related articles” functionwas utilized to broaden the search, and all abstracts, studies, andcitations scanned and reviewed. References of the articles ac-quired were also searched manually. No language restrictionswere made. The last date for this search was June 1, 2007.

Inclusion Criteria

Studies were included for meta-analysis if they met eachof the following criteria: 1) randomized controlled trials, 2)involved open-heart surgery, 3) reported on the effect of aleukocyte depleting filter within the arterial line of the cardio-pulmonary bypass circuit (“systemic leukofiltration”), 4) re-ported at least one clinical outcome relating to postoperativerespiratory function.

Exclusion Criteria

Studies were excluded from the review if 1) the trial wascarried out on animal models; 2) they published results sub-sequent to a previously published study (the smallest and theoldest were excluded); 3) systemic filtration had only occurredsimultaneous to filtration of the cardioplegic or cell-saver line,or administration of an anti-inflammatory medication, e.g.,aprotinin; 4) neither patients nor investigators were blinded tothe intervention; 5) a zero cell was displayed for the outcomesof interest for both groups; 6) inconsistency of data did notallow valid extraction; 7) the filter had been placed in thevenous line of the CPB circuit.

Data Extraction and Validation of Studies

Three reviewers (O.W., C.T., S.W.) identified trials for in-clusion and independently extracted the following data fromeach study, according to a prearranged protocol: first author,publication year, institution, study design, study population char-

acteristics, number of subjects in each subgroup, and outcomemeasures. Authors of included trials were contacted when nec-essary to clarify data and to identify multiple publications. Eachreviewer independently assigned each study a Jadad quality score(maximum score � 5) that evaluates the quality of reports ofrandomized clinical trials.40 In the case of any discrepancy be-tween reviewers, the decision was taken by consensus.

Outcomes and Definitions

Both physiological and clinical outcomes were investigated.The ratio of arterial oxygen to the fraction of inspired oxygen(PaO2/FiO2) was used as a physiological index of acute lunginjury. If articles did not report this parameter directly, wherepossible, the required data were extracted and a value calcu-lated. If pressures were reported as kilopascals (kPa), they wereconverted to mm Hg (1 kPa � 7.52 mm Hg). There wassignificant variability in the times at which these measure-ments were taken, and therefore to allow meta-analysis to beperformed, these were arbitrarily divided into “Early” PaO2/FiO2, defined as any measurement taken within 12 hours ofCPB cessation and “Late” PaO2/FiO2, defined as any measure-ment taken more than 12 hours post-CPB cessation.

The clinical outcomes investigated were a) duration of ven-tilatory support (hours), b) postoperative respiratory infectionrate, c) length of stay in the intensive care unit (days), d) lengthof stay in hospital (days). When length of stay was reported inhours, this was converted to days (time in hours/24).

Statistical Analysis

The group where no leukocyte filtration occurred was regardedas the reference group and that in which a leukocyte-depletingfilter was deployed, the treatment group. Meta-analysis wasperformed in line with recommendations from the CochraneCollaboration and the Quality of Reporting of Meta-analyses(QUORUM) guidelines.41,42 Where outcomes were continu-ous data, the effect measure estimated was weighted meandifference (WMD), reported with 95% confidence intervals(CI). A WMD �0 favored the leukofiltration group. The pointestimate of the WMD was considered statistically significant atthe p � 0.05 levels, if the 95% confidence interval did notinclude the value zero. Statistical analysis for categoric vari-ables was performed using the odds ratio (OR) as the summarystatistic. This ratio represents the odds of an adverse eventoccurring in the treatment (leukofiltration group) comparedwith the reference (standard filter) group. An OR of �1 favorsthe treatment group and the point estimate of the OR is con-sidered statistically significant at the p � 0.05 level, if the 95%confidence interval does not include the value one.

Pooled estimates, confidence intervals (CI), and tests forheterogeneity were calculated and visual evaluation of possi-ble publication bias was performed by the use of funnelplots.43 In the forest plots of the results, squares indicate pointestimates of treatment effect (WMD or OR), and horizontalbars indicate 95% confidence intervals. The diamond repre-sents the summary estimate of treatment effect from the pooledstudies with 95% confidence intervals. In a fixed-effect model,it is assumed that there is no heterogeneity in treatment effectbetween studies, whereas in a random-effect model, it is as-sumed that there is variation between studies and the calculated

79SYSTEMIC LEUKOCYTE FILTRATION AND PULMONARY INJURY AFTER CARDIAC SURGERY

Tab

le1.

Syn

op

sis

of

Stu

die

s

Pap

erS

tud

yD

esig

nA

dm

issi

onTy

pe

Sur

gica

lP

roce

dur

eE

xclu

sion

Crit

eria

Incl

usio

nC

riter

iaM

atch

ing

Crit

eria

Tota

l(n

)C

ontr

ol(n

)Fi

lter

(n)

Mea

nA

ge(y

rs)

Con

trol

Mea

nA

ge(y

rs)

Filte

r%

Mal

e

Pal

anzo

etal

.24

2N

A1,

2N

Aa

a,j,k

,e,b

,n,

o,p

3618

1866

.362

.5N

AJo

hnso

net

al.4

81

E1

NA

NA

a,d

,f,g,

j,o,p

3216

1664

6587

.5M

ihal

jevi

cet

al.3

72

E1

A,B

,C,D

a,b

a,d

,c,o

,p,

3218

1462

5710

0H

achi

da

etal

.50

2N

A3

Ca

b,k

,n,o

,p28

1414

5553

.6N

ALu

stet

al.5

21

NA

5N

AN

Aa,

h,i,d

,o,p

5025

25N

AN

AN

AD

iSal

voet

al.5

12

E,U

1N

Ac,

d,e

,fa,

d,f,

h,c,

g,l,p

2010

1065

.765

.890

Hur

stet

al.4

91

E3

HN

Aa,

d,h

,j,g,

m,o

,p24

1311

6163

45.8

Bak

sass

etal

.53

2E

1F,

H,K

,L,M

,Pg

a,b

,c,d

,g,h

,o,p

4020

2068

6782

.5M

air

etal

.36

2E

1C

,H,K

,N,

NA

a,c,

d,o

,p,r

4020

2061

6575

Sah

lman

etal

.54

2E

1A

,H,L

,P,Q

,SN

Aa,

f,d,o

,p,h

,j,a,

g,r,

m60

3030

6364

78.3

Fab

bri

etal

.35

2E

1C

,E,K

,H,J

ja,

h,j,k

,b,d

,q,o

,p20

1010

6971

70S

tefa

nou

etal

.58

2E

1H

NA

h,r,

g,d

,j,k,

m,c

,o,p

2010

10N

AN

A90

Efs

tath

iou

etal

.39

1E

1,2,

4,1�

2,1�

4C

,F,H

,L,P

,JN

Aa,

d,h

,k,e

,o,p

,q80

4040

62.6

58.9

73.8

She

pp

ard

etal

.57

1E

1C

,H,P

,JN

Aa,

q,j,

k,o,

p50

2525

59.6

61.4

100

Ale

xiou

etal

.56

2E

1C

,F,H

,M,N

,P,R

,S,J

ha,

j,o,p

,d,m

,h,r

,c,f,

s11

055

5564

6570

.9K

arai

skos

etal

.55

1E

5H

ka,

e,o,

p,j,

d,m

,h,k

,r,s

,t40

2020

6265

85C

hen

etal

.63

2E

1,3

C,F

,K,H

,L,M

,GN

Aa,

o,p

,h,c

,q,e

3216

1661

.260

.381

.3A

lexi

ouet

al.5

92

E1

C,F

,H,M

,N,P

,R,S

,Jh

a,o,

p,j,

d,h

,m,r

,b,c

,n,f

5025

2561

6676

She

pp

ard

etal

.60

1E

1C

,F,L

,S,J

ba,

o,p

,j,b

,k32

1616

62.2

62.4

NA

Leal

-Nov

alet

al.6

22

E1,

3I,H

,K,J

,O,C

ga,

o,p

,e,h

,m,b

,n,q

,u15

910

752

61.4

60.9

61K

oske

nkar

iet

al.6

11

E1

C,H

,K,L

,P,J

l,ia,

k,d

,r,m

,t,u

,o,p

,e40

2020

7273

.5N

Stu

dy

Des

ign:

1�

dou

ble

blin

dra

ndom

ized

cont

rolt

rial,

2�

sing

leb

lind

rand

omiz

edco

ntro

ltria

l.Ty

pe

ofA

dm

issi

on:

E�

Ele

ctiv

e,U

�U

rgen

t.S

urgi

calP

roce

dur

e:1

�C

AB

Gon

ly,

2�

Aor

ticV

alve

Rep

lace

men

t,3

�“V

alvu

lar

Sur

gery

,”4

�M

itral

Val

veR

epla

cem

ent,

5�

Any

open

hear

tp

roce

dur

e.E

xclu

sion

Crit

eria

:A

�V

alvu

lar

dis

ease

,B

�C

onge

nita

lH

eart

Dis

ease

,C

�A

cute

orch

roni

cp

ulm

onar

yd

ysfu

nctio

n,D

�P

ulm

onar

yH

yper

tens

ion,

E�

His

tory

ofA

llerg

y,F

�R

e-op

erat

ion,

G�

MI

inla

stm

onth

,H

�U

nab

leto

cons

ent,

I�

Par

sonn

etS

core

�10

,J

�H

emat

olog

ical

Dis

ease

/Ab

norm

aliti

es,

K�

Rec

ent

orcu

rren

tIn

fect

ion,

L�

Co-

Mor

bid

ity/

Mal

igna

ncy,

M�

Ste

roid

/NS

AID

user

s,N

�E

f�

50%

,O

�LM

Sst

enos

is,

P�

Ren

alD

isea

se,

Q�

EF

�30

%,

R�

Coa

gulo

pat

hy,

S�

Cer

ebra

lIns

uffic

ienc

y/S

trok

e.In

clus

ion

Crit

eria

:a�

Nor

mal

Pul

mon

ary

Func

tion,

b�

Age

�75

yr,c

�se

vere

Left

mai

nst

emd

isea

se�

70%

,d�

Poo

rLV

Func

tion

(EF

�40

%),

e�

faile

dP

TCA

,f�

unst

able

angi

na,

g�

LVE

F�

40%

,h

�A

ge�

70,

i�S

tab

leC

oron

ary

Art

ery

Dis

ease

j�C

PB

�60

min

s,k

�C

OP

D,

l�A

ge�

70yr

.M

atch

ing:

a�

age,

b�

Hem

oglo

bin

/Hem

atoc

rit,

c�

No.

ofG

raft

s,d

�P

reop

erat

ive

Car

dia

cFu

nctio

n,e

�P

roce

dur

e,f

�E

xten

tof

Cor

onar

yA

rter

yd

isea

se,

g�

Sym

pto

mS

ever

ity,

h�

sex,

i�R

ace,

j�B

ody

Wei

ght/

BM

I,k

�B

ody

Sur

face

Are

a,l�

Pre

viou

sC

AB

G,m

�R

isk

Fact

ors/

Co-

Mor

bid

ity,n

�P

reop

erat

ive

lab

.var

iab

les,

o�

Aor

ticcr

oss

clam

ptim

e,P

�C

PB

time,

q�

Low

est

Tem

per

atur

e(R

ecta

lor

nasa

l)r

�P

revi

ous

MI/

PTC

A,

s�

Dru

gD

osag

e(s

tero

ids/

nitr

ates

)t

�P

reop

erat

ive

lung

func

tion,

u�

Pre

oper

ativ

eris

kst

ratifi

catio

n.n,

num

ber

;N

A,

not

avai

lab

le.

80 WARREN ET AL.

WMD or OR therefore has a more conservative value.43,44 Insurgical research, and thus in our study, meta-analysis using therandom-effect model is preferable, particularly because patientsthat are operated on in different centers have varying risk profilesand selection criteria for each surgical technique.

Heterogeneity was explored using the chi-squared statistic,but the I2 value was calculated to quantify the degree ofheterogeneity across trials that could not be attributable tochance alone. I2 provides a superior measure of heterogeneitybetween trials, as it is independent of the number of trialsincluded in the analysis and is not limited by power.45,46

When I2 � 50, it indicates significant heterogeneity. Threefurther strategies were used to quantitatively assess heteroge-neity. First, data were reanalyzed by using both random-effectand fixed-effect models. Second, graphical exploration withfunnel plots was used to evaluate publication bias,47 andre-analysis performed once any outlying studies had beenremoved from the data. Third, sensitivity analysis was under-taken using subgroups. To do this, the following variables wereevaluated: i) All studies; ii) study size (�25 patients in eachstudy group); iii) studies with 8 or more demographic matchingcriteria, i.e., studies with minimal heterogeneity between con-trol and treatment groups; iv) higher-quality studies (JADADscore �2); v) studies including patients at high risk of respira-tory complications. These subgroups were identified in theprotocol before conducting the review and analysis.

Analysis was conducted by use of the statistical softwareSPSS version 12.0 for Windows (SPSS Inc., Chicago, IL), andReview Manager Version 4.2 for Windows (The CochraneCollaboration, Software Update, Oxford, UK).

Results

Eligible Studies

Using the above search keywords, 348 publications wereidentified, of which 278 were excluded following title andabstract review; 70 initial articles were investigated and theirreferences crosschecked, which yielded one further report forevaluation and allowed the exclusion of 43 studies based onthe evaluation of the study methodology (two studies hadinvestigated an experimental venous line filter26,34). Therefore,28 systemic leukofiltration randomized controlled trials werestudied in detail to identify their primary and secondary out-comes. Of these, 21 articles published between 1993 and2005 fulfilled all the inclusion and exclusion criteria, andyielded data suitable for this meta-analysis.24,35–37,39,48–63

These included a combined total of 995 patients, of which 528(53.1%) underwent CPB with a control filter, and 467 (46.9%)underwent systemic leukofiltration.

The characteristics of the included studies are shownin Table 1. Eight of the studies are double-blind,39,48,49,52,55,57,60,61 and the remaining ones are single-blind randomized controlled trials. Table 2 shows the numberof matching criteria and the JADAD score for the randomizedcontrol trials. Thirteen of the studies matched control andtreatment groups for eight or more demographic crite-ria.24,35,39,49,51,53–59,61,62 The mean JADAD score was 1.43(range, 1–3), but only eight of the studies scored higher than1,48,49,52,56,57,59,62,63 suggesting the overall quality of studies tobe relatively poor.

Table 3 demonstrates the different outcome measures re-ported in the 21 studies included in our analysis. Outcomescan be arbitrarily divided into three groups: basic scienceoutcomes, such as plasma levels of cytokines and the expressionof neutrophil adhesion molecules; clinical laboratory outcomes,such as leukocyte count and C-reactive protein concentration;and perioperative clinical outcomes, such as lung compliance,central venous pressure, and transfusion requirements. Most ofthe included articles studied at least two of these three out-come types. The variety of ways in which investigators havesought to investigate the potential impact of leukodepletion incardiac surgery patients is large; even with an element of datasynthesis, we have still reported 66 different outcome variablesacross 21 trials.

Results from Overall Meta-Analysis

Impact on PaO2/FiO2, Early and Late. Forest plots of themeta-analysis of the studies with regard to Early PaO2/FiO2

and Late PaO2/FiO2 for control vs. systemic leukofiltration areshown in Figures 1 and 2. Systemic leukofiltration significantlyincreased the PaO2/FiO2 ratio within 12 hours of CPB cessa-tion, [weighted mean difference (WMD), 25.97; 95% confi-dence interval (CI), 3.41, 48.53; p � 0.02], but the includedstudies were significantly heterogeneous (�2 � 16.93; I2 �70.5%; p � 0.005). This effect was lost after 12 hours, al-though a trend toward an increased PaO2/FiO2 ratio remained(WMD, 12.98; 95% CI, �7.93–33.98; p � 0.22), and hetero-geneity became insignificant (�2 � 0.47; I2 � 0%; p � 0.79).

Clinical Outcomes. Systemic leukofiltration significantly re-duced the duration of ventilatory support postoperatively (WMD,�2.11 hours; 95% CI, �3.58 to �0.65; p � 0.005) (Figure 3), butdata from the 13 studies reporting this outcome were highlyheterogeneous (�2 � 49.75; I2 � 75.9%; p � 0.00001). Systemicleukofiltration had no effect on postoperative respiratory infec-tion (OR, 1.06; 95% CI, 0.39–2.91; p � 0.90). There was a

Table 2. JADAD and Matching Criteria Scores for IncludedStudies

ReferenceJADAD Quality

ScoreDemographic Matching

Criteria (n)

Palanzo et al.24 1 8Johnson et al.48 2 7Mihaljevic et al.37 1 5Hachida et al.50 1 5Lust et al.52 3 6Di Salvo et al.51 1 8Hurst et al.49 2 8Baksass et al.53 1 8Mair et al.36 1 6Sahlman et al.54 1 11Fabbri et al.35 1 9Stefanou et al.58 1 10Efstathiou et al.39 1 8Sheppard et al.57 2 6Alexiou et al.56 2 11Karaiskos et al.55 1 12Chen et al.63 2 7Alexiou et al.59 2 12Sheppard et al.60 1 6Leal-Noval et al.62 2 10Koskenkari et al.61 1 10

81SYSTEMIC LEUKOCYTE FILTRATION AND PULMONARY INJURY AFTER CARDIAC SURGERY

Table 3. Outcome measures of the 21 included studies

Study Reference 24 48 37 50 52 51 49 53 36 54 35 58 39 57 56 55 63 59 60 62 61

Basic Science OutcomesMalondialdehyde Levels � � �Glutathione Reduction �Plasma Nitrotyrosine/Tyrosine Ratio � �TNF� � �IL-1, IL-2, IL-6, IL-8 � � � � � �CD11a/b/c/18/62DL/CD67/CD45Ro � � � �L-Selectin/P-Selectin �Thromboxane B2 �WCC chemiluminescence �Serum 6-keto-PGF1a �Lipid Peroxidation �Elastase � � � �Complement � �Myeloperoxidase � �Lactoferrin �

Clinical Laboratory OutcomesLeukocyte Count � � � � � � � � � � � � � � �Neutrophils � � � � � � � � � �Lymphocytes � � � � � �Monocytes �Platelets � � � � � � �CRP � �Bilirubin �Urea and Electrolytes (Na�/K�) �Osmolality �Serum CK/CK-MB � � � � �Troponin T �/or I � � � �Hemoglobin/Hematocrit � � � � �

Perioperative Clinical OutcomesPulmonary FunctionIntubation/Ventilatory Support Duration � � � � � � � � � � � � �Lung Compliance �Pulmonary Arterial Pressure � �Pulmonary Vascular Resistance � �Arterial Blood Gases � � � � � � �Pulmonary Capillary Wedge Pressure � � �pO2/FiO2 or Respiratory Index � � � � � � � � �Atelectasis/chest infection � � � � � �Intrapulmonary Shunt �FEV1/FVC � � �Exhaled Nitric Oxide � � �Pistolesi scoring for lung water �AaOI � �

Cardiac FunctionPerioperative Arrhythmia � � � �Perioperative Myocardial Infarction � �Cardiac Arrest �Inotropic Support � � � � � � �ECG Changes �Cardiac Output/Index � � � � �Mean Arterial Pressure � � �Right Atrial Pressure � �Central Venous Pressure � �Postoperative Vasodilators �Systemic Vascular Resistance Index �

Renal FunctionUrine output � � �Fluid Balance/Weight � �

Hematological/CoagulationBlood Loss/chest drainage � � � � � � � � �Re-Exploration � �Blood Transfusions � � � � � � � �Crystalloid Infusions � �FFP Requirements �

InfectionSepsis/Infections � �Wound Infection � �Fever in first 24 h �

82 WARREN ET AL.

trend toward a decrease in both length of intensive care unitstay (WMD, �0.11 days; 95% CI, �0.26–0.05; p � 0.17) andhospital stay (WMD, �0.41 days; 95% CI, �0.89–0.07; p �0.1) (Figure 4) within the treatment group although neither ofthese was statistically significant.

Sensitivity Analysis

The results of all the sensitivity analyses are presented inTable 4.

Studies With >25 Patients in Each Study Group. A sensi-tivity analysis was performed looking at studies that included�25 patients in both the control group and the leukofiltrationgroup. Only 7 of our 21 studies were large enough to beconsidered39,52,54,56,57,59,62; only one of these reported “Late”PaO2/FiO2 ratio,54 and thus, sensitivity analysis for this out-come was not possible. In this subgroup analysis, systemic

leukofiltration had no effect on any of the reported outcomes.Despite excluding the smaller studies, statistically significantheterogeneity was still reported in all outcomes except respi-ratory infection rates.

Studies With Eight or More Matching Criteria. Sensitivityanalysis of this group demonstrated a statistically significantmaintenance of early PaO2/FiO2 ratio (WMD, 33.04 mm Hg;95% CI, 9.04–57.03; p � 0.007), and a statistically significantreduction in the duration of ventilatory support postoperatively(WMD, �2.28 hours; 95% CI, �4.05 to �0.5; p � 0.01)within the systemic leukofiltration group. However, in bothinstances the sensitivity analysis failed to remove statisticallysignificant heterogeneity between the differing studies, it beingreported in all but two of the six studied outcomes.

Higher-Quality Studies (JADAD score >2). None of theeight studies in this sensitivity analysis48,49,52,56,57,59,62,63 re-

Figure 1. Forest Plot displaying the results of the meta-analysis of ‘Early’ PaO2/FiO2 ratio in patients undergoing cardiopulmonary bypasswith standard filter (controls) compared to systemic leukofiltration.

Table 3. Continued

Study Reference 24 48 37 50 52 51 49 53 36 54 35 58 39 57 56 55 63 59 60 62 61

Neurological OutcomesNeurological Complications �Stroke �

Other OutcomesTime in Intensive Care Unit � � � � � � � � � �Time in hospital � � � � � �Mortality � � � �

Figure 2. Forest Plot displaying the results of the meta-analysis of ‘Late’ PaO2/FiO2 ratio in patients undergoing cardiopulmonary bypasswith standard filter (controls) compared to systemic leukofiltration.

83SYSTEMIC LEUKOCYTE FILTRATION AND PULMONARY INJURY AFTER CARDIAC SURGERY

ported “Late” PaO2/FiO2 ratio, and thus sensitivity analysis forthis outcome was not possible. Sensitivity analysis of thissubgroup found systemic leukofiltration to have no significanteffect on any of the five predefined outcomes but importantlyeradicated any previous heterogeneity between studies.

Studies Including Patients at High Risk of Respiratory Com-plications. Eight studies included patients at high risk ofrespiratory complications;48,49,51–55,58 only one of these stud-ies reported postoperative respiratory infection rates,58 andthus, this outcome is not reported. Sensitivity analysis of thisgroup demonstrated a statistically significant reduction in theduration of ventilatory support postoperatively, although the CIdid included 0, (WMD, �1.85 hours; 95% CI, �3.72–0.01;p � 0.05) with no significant heterogeneity (�2 � 3.08; p �0.54; I2 � 0%) (Figure 5). However, leukocyte filtration had noeffect on any of the other four reported outcomes. Importantly,sensitivity analysis revealed heterogeneity between these stud-ies to be insignificant for all five reported outcomes.

Discussion

Meta-analysis assessing the effect of systemic leukofiltrationon respiratory injury and clinical outcomes has not been pre-

viously performed. We have utilized a systematic methodol-ogy to ensure all the available data were considered. Meta-analysis is most effective when randomized controlled trialsare analyzed, and in this article, all of the included studiesfulfilled these criteria.

Our results suggest systemic leukofiltration may attenuateearly acute lung injury in patients undergoing CPB, as dem-onstrated by increased PaO2/FiO2 ratio within 12 hours of CPBcessation and reduced duration of postoperative ventilatorysupport. However, for both these findings, there was signifi-cant heterogeneity across the reported studies. When studieswith well-matched study groups are analyzed as a subgroup,positive outcomes remain, but heterogeneity persists. Whenstudies including high-risk patients are analyzed, a significantreduction in the duration of postoperative ventilatory supportis found, and significant heterogeneity removed. Systemic leu-kofiltration did not affect PaO2/FiO2 ratio 12 hours after CPBcessation, incidence of postoperative respiratory infection,length of ITU stay or length of hospital stay.

Meta-analysis should place significant weight on explainingthe causes for heterogeneity within the literature, rather thanjust focus on calculating the overall estimates for several out-

Figure 3. Forest Plot displaying the results of the meta-analysis of the duration of ventilatory support in patients undergoing cardiopul-monary bypass with standard filter (controls) compared to systemic leukofiltration.

Figure 4. Forest Plot displaying the results of the meta-analysis of the length of hospital stay in patients undergoing cardiopulmonarybypass with standard filter (controls) compared to systemic leukofiltration.

84 WARREN ET AL.

comes of interest.64 The heterogeneity within this study is large,but we have identified two key reasons for this: 1) variability instudy quality (as measured by Jadad Score)40 and 2) exclusion ofpatients at high risk of pulmonary complications. When sensitiv-ity analysis was performed using these criteria, heterogeneity waseradicated for every outcome variable.

The clinical effectiveness of leukocyte depletion in amelio-rating lung injury has been extensively debated.2,10,65 Westudied six outcomes, and found systemic leukofiltration only

to have an effect on two: early PaO2/FiO2 ratio and duration ofintubation post-CPB cessation. Furthermore, the clinical sig-nificance of the effect is likely to be minimal in both cases.Interestingly, both outcomes are markers of early respiratoryfunction, and would imply that any effect of systemic leuko-filtration is transient, a finding that supports others’ work.Karaiskos et al. and Efstathiou et al. demonstrated reductions inthe respiratory index and intubation time in leukodepletedpatients, but demonstrated no effect on length of ITU stay.39,55

Table 4. Sensitivity Analysis Performed for Studies Comparing Controls vs. Systemic Leukofiltration

OutcomesNo.

PatientsNo.

Studies WMD OR 95% CI p

Heterogeneity

�2

Test pI2

(%)

Study size (>25 patients in each studygroup)

“Early” PaO2/FiO2 (mm Hg)39,52,62 289 3 18.11 — �12.13, 48.35 0.24 13.88 0.001 85.6Duration of ventilatory support

(hrs)39,52,59,62339 4 �1.70 — �4.35, 0.95 0.21 31.31 0.0001 90.4

Respiratory infection59,62 209 2 — 1.04 0.28, 3.88 0.95 1.17 0.28 14.7ITU length of stay (days)39,59,62 289 3 �0.07 — �0.31, 0.17 0.57 8.03 0.02 75.1Hospital length of stay (days)39,52,54,59 240 4 �0.18 — �1.10, 0.75 0.71 7.46 0.06 59.8

No. of matching criteria >8“Early” PaO2/FiO2 (mm Hg)24,39,55,62 315 4 33.04 — 9.04, 57.03 0.007 13.14 0.004 77.2“Late” PaO2/FiO2 (mm Hg)24,54,55 136 3 12.98 — �7.93, 33.89 0.22 0.47 0.79 0Duration of ventilatory support

(hrs)24,35,39,51,53,55,58,59,62465 9 �2.28 — �4.05, �0.5 0.01 37.05 0.0001 78.4

Respiratory infection58,59,62 229 3 — 1.06 0.39, 2.91 0.9 1.17 0.56 0ITU length of stay

(days)24,35,39,49,51,55,58,59,62449 9 �0.14 — �0.35, 0.08 0.21 16.92 0.03 52.7

Hospital length of stay(days)35,39,49,51,54,55,58,59

314 8 �0.48 — �1.05, 0.09 0.1 14.15 0.05 50.5

Higher quality studies (JADAD score >2)“Early” PaO2/FiO2 (mm Hg)52,62 209 2 5.85 — �10.8, 22.49 0.49 0.85 0.36 0Duration of ventilatory support

(hrs)52,59,62,63291 4 �0.55 — �1.76, 0.67 0.38 1.19 0.75 0

Respiratory infection59,62 209 2 — 1.04 0.28, 3.88 0.95 1.17 0.28 14.7ITU length of stay (days)48,49,59,62 265 4 0.06 — �0.22, 0.33 0.69 4.76 0.19 37.0Hospital length of stay (days)48,49,52,59 156 4 0.07 — �0.52, 0.66 0.82 0.91 0.82 0

Studies including patients at high riskof respiratory complications

“Early” PaO2/FiO2 (mm Hg)52,55 90 2 8.56 — �66.80, 83.91 0.82 1.86 0.17 46.3“Late” PaO2/FiO2 (mm Hg)54,55 100 2 9.61 — �13.85, 33.07 0.42 0.09 0.77 0Duration of ventilatory support

(hrs)51–53,55,58170 5 �1.85 — �3.72, 0.01 0.05 3.08 0.54 0

ITU length of stay (days)48,49,51,55,58 136 5 �0.20 — �0.69, 0.28 0.41 9.19 0.06 56.5Hospital length of stay

(days)48,49,51,52,54,55,58246 7 �0.20 — �0.73, 0.33 0.46 7.49 0.28 19.9

Figure 5. Forest Plot displaying the results of the meta-analysis sensitivity of duration of ventilatory support in patients undergoingcardiopulmonary bypass with systemic leukofiltration compared to controls, in studies including high-risk patients.

85SYSTEMIC LEUKOCYTE FILTRATION AND PULMONARY INJURY AFTER CARDIAC SURGERY

Johnson et al. reported increased PaO2 and lower levels ofintrapulmonary shunt in those in the leukocyte filter group, butthese findings were transient and disappeared by 24 hours.48

Alexiou et al. showed reduced exhaled NO in the study group,but no impact on respiratory complications or intubationtime.59 Furthermore, our sensitivity analysis of studies includ-ing high-risk patients showed a reduction in the duration ofventilatory support in the treated group, supporting those whohave suggested that high-risk patients may have the most togain from the attenuation of the inflammatory response thatleukodepleting filters potentially cause.55,66

This meta-analysis demonstrates that the efficacy of leu-kocyte depletion remains less impressive than might havebeen anticipated from our knowledge of the pathophysiol-ogy of the inflammatory response associated with CPB. Whythis should be the case remains unclear. One reason may be“filter exhaustion,” whereby filters progressively diminish inefficacy throughout the bypass cycle.29,36,37 Another reasonpostulated is that although leukodepleting filters certainlyremove leukocytes from a stream of blood in vitro, they donot seem to consistently lower leukocyte concentrations invivo, or they do it only transiently.67 This has led some tosuggest that leukodepleting filters may preferentially removeactivated leukocytes from the circulation, but given thesignificant body of evidence both for63,67 and against35,58

this, the issue remains controversial. Regardless of the rea-sons, we believe this article supports the conclusions madeelsewhere that leukocyte filtration has not been as effectiveas first hoped, and that alternative solutions to attenuate theinflammatory response should be sought.25,68 Despite itsefficacy in cardiopulmonary bypass never being fully estab-lished, a small number of centers employ systemic filtersroutinely in their CPB circuits, a practice we cannot cur-rently advocate on the basis of our results.

In recent years, investigators have sought to improve theimpact of systemic leukofiltration by using various strategies,such as total leukocyte control,32 altering the temperature atwhich filtration takes place,31 or combining filtration withpharmaceutical manipulation in the form of aprotinin.69 Thesehave met with varying success. The primary value of ourmanuscript is to give interim recommendations to those cur-rently practicing, highlight the deficiencies within the literatureand to guide future work. To this end, we have formulatedrecommendations using the “EPICOT” guidelines.70 Althougha moderate number of trials investigating systemic leukofiltra-tion exist, most are small with significant limitations in terms ofquality, and thus further research is likely warranted. Our datasupport further double-blind randomized controlled trials be-ing conducted, but these must incorporate strategies to preventfilter exhaustion at critical times, be of high quality, be well-powered and target (as opposed to exclude) high-risk, fragilepatients, as this cohort may have the most to gain. Outcomesmust be patient-focused, such as length of stay and haveobjective diagnostic criteria, such as acute respiratory distresssyndrome rates, atelectasis scoring on computed tomographyof the thorax or lung function tests. Beyond this, investigatorsshould consider incorporating other methods or strategies tomaximize leukofiltration’s impact, e.g., pharmaceuticalagents.

Limitations

This meta-analysis has several limitations that must be takeninto account when considering its results. The appropriatenessand place of meta-analytic techniques within the medicalliterature has been a source of intense debate among epide-miologists. Critics argue that the effect of meta-analysis is to“reinforce the inherent systematic biases of the studies, pro-duce spurious statistical stability, and discourage further re-search.”71 We do not seek to discourage further research, butbelieve that statistical quantification and pooling of previousresults provides a tool to explain variability, inconsistency, andheterogeneity within the literature and thus “set the stage forfurther research.”72

The key limitation with this meta-analysis is the severe heter-ogeneity present throughout the literature, which was only par-tially eradicated through sensitivity analysis. This meant thatwhen a significant difference was detected between the twogroups, frequently significant levels of heterogeneity accompa-nied it. Furthermore, despite there being 21 studies, the variabilitywith which they reported outcomes meant that most individualmeta-analysis involved �10 studies. Accompanied by relativelysmall sample size and low-quality trial design, interpretation ofthe data becomes increasingly difficult.

We used PaO2/FiO2 ratio and duration of ventilatory sup-port as markers of acute lung injury, as they are frequentlycited and where PaO2/FiO2 ratio was not cited it could fre-quently be calculated from other presented data. However, inthe majority of studies, these are surrogate outcomes, thestudies having been designed to answer other controversiesregarding leukofiltration. There may have therefore been aselection bias in the reporting of these outcomes. The numberof potential confounding variables not controlled for in theincluded studies is relatively high. Just 2 of the 21 studiesmatched patient groups for preoperative lung function.55,61

Furthermore, although some studies specified their criteria forwithdrawal of ventilatory support, e.g., fully awake, central tem-perature over 36°C, adequate spontaneous ventilation, and satis-factory arterial blood gases,54,55,61 these criteria are somewhatunavoidably subjective and variable between centers. Finally, wehad no choice but to set arbitrary definitions of “Early” and “Late”PaO2/FiO2 ratio, thus grouping measurements taken at varyingtimes in different studies, as the variability in study protocolsprohibited meaningful data synthesis for specific times.

Conclusion

Systemic leukofiltration may attenuate bypass-related lung in-jury in the early postoperative period, but this does not seem totranslate to differences in clinically significant outcomes. Thestudies performed in this area thus far are highly heterogeneous,because of poor-quality study design and the frequent exclusionof patients at high risk of postoperative pulmonary dysfunction, asubgroup who may have the most to benefit from this technology.Further trials of systemic leukofiltration are warranted but mustaddress these issues to ensure they contribute to the currentliterature.

References

1. Shann KG, Likosky DS, Murkin JM, et al: An evidence-basedreview of the practice of cardiopulmonary bypass in adults: a

86 WARREN ET AL.

focus on neurologic injury, glycemic control, hemodilution,and the inflammatory response. J Thorac Cardiovasc Surg 132:283–290, 2006.

2. Wan S, LeClerc JL, Vincent JL: Inflammatory response to cardio-pulmonary bypass: Mechanisms involved and possible thera-peutic strategies. Chest 112: 676–692, 1997.

3. Laffey JG, Boylan JF, Cheng DC: The systemic inflammatory re-sponse to cardiac surgery: Implications for the anesthesiologist.Anesthesiology 97: 215–252, 2002.

4. Levy JH, Tanaka KA: Inflammatory response to cardiopulmonarybypass. Ann Thorac Surg 75: S715–720, 2003.

5. Cremer J, Martin M, Redl H, et al: Systemic inflammatory responsesyndrome after cardiac operations. Ann Thorac Surg 61: 1714–1720, 1996.

6. Elgebaly SA, Houser SL, el Kerm AF, et al: Evidence of cardiacinflammation after open heart operations. Ann Thorac Surg 57:391–396, 1994.

7. Edmunds LH Jr: The evolution of cardiopulmonary bypass: Les-sons to be learned. Perfusion 17: 243–251, 2002.

8. Edmunds LH Jr: Inflammatory response to cardiopulmonary by-pass. Ann Thorac Surg 66: S12–S16; discussion S25–S18, 1998.

9. Asimakopoulos G, Smith PL, Ratnatunga CP, Taylor KM: Lunginjury and acute respiratory distress syndrome after cardiopul-monary bypass. Ann Thorac Surg 68: 1107–1115, 1999.

10. Clark SC: Lung injury after cardiopulmonary bypass. Perfusion 21:225–228, 2006.

11. Friedman M, Wang SY, Sellke FW, et al: Neutrophil adhesionblockade with NPC 15669 decreases pulmonary injury aftertotal cardiopulmonary bypass. J Thorac Cardiovasc Surg 111:460–468, 1996.

12. Westaby S, Fleming J, Royston D: Acute lung injury during car-diopulmonary bypass, the role of neutrophil sequestration andlipid peroxidation. Trans Am Soc Artif Intern Organs 31: 604–609, 1985.

13. Royston D, Fleming JS, Braude S, et al: Lung injury followingcardiopulmonary bypass; the potential role of oxidant-free rad-icals. Life Support Syst 4: 151–154, 1986.

14. Royston D, Fleming JS, Desai JB, et al: Increased production ofperoxidation products associated with cardiac operations. Evi-dence for free radical generation. J Thorac Cardiovasc Surg 91:759–766, 1986.

15. Reddy SL, Grayson AD, Griffiths EM, et al: Logistic risk model forprolonged ventilation after adult cardiac surgery. Ann ThoracSurg 84: 528–536, 2007.

16. Wachtfogel YT, Kucich U, Hack CE, et al: Aprotinin inhibits thecontact, neutrophil, and platelet activation systems during sim-ulated extracorporeal perfusion. J Thorac Cardiovasc Surg 106:1–9; discussion 9–10, 1993.

17. Soeparwata R, Hartman AR, Frerichmann U, et al: Aprotinindiminishes inflammatory processes. Int J Cardiol 53 (Suppl):S55–S63, 1996.

18. Sullivan GW, Carper HT, Novick WJ Jr, Mandell GL: Inhibition ofthe inflammatory action of interleukin-1 and tumor necrosisfactor (alpha) on neutrophil function by pentoxifylline. InfectImmun 56: 1722–1729, 1988.

19. Turkoz R, Yorukoglu K, Akcay A, et al: The effect of pentoxifyllineon the lung during cardiopulmonary bypass. Eur J CardiothoracSurg 10: 339–346, 1996.

20. Ishihara K, Oshida H, Endo Y, et al: Hemocompatibility of humanwhole blood on polymers with a phospholipid polar group andits mechanism. J Biomed Mater Res 26: 1543–1552, 1992.

21. Bando K, Schueler S, Cameron DE, et al: Twelve-hour cardiopul-monary preservation using donor core cooling, leukocyte de-pletion, and liposomal superoxide dismutase. J Heart LungTransplant 10: 304–309, 1991.

22. Bando K, Pillai R, Cameron DE, et al: Leukocyte depletion ame-liorates free radical-mediated lung injury after cardiopulmonarybypass. J Thorac Cardiovasc Surg 99: 873–877, 1990.

23. Schueler S, De Valeria PA, Hatanaka M, et al: Successful twenty-four-hour lung preservation with donor core cooling and leu-kocyte depletion in an orthotopic double lung transplantationmodel. J Thorac Cardiovasc Surg 104: 73–82, 1992.

24. Palanzo D, Manley N, Montesano R, et al: Clinical evaluation of

the LeukoGuard (LG-6) arterial line filter for routine open-heartsurgery. Perfusion 8: 489–496, 1993.

25. Warren O, Alexiou C, Massey R, et al: The effects of variousleukocyte filtration strategies in cardiac surgery. Eur J Cardio-thorac Surg 31: 665–676, 2007.

26. Gu YJ, de Vries AJ, Vos P, et al: Leukocyte depletion duringcardiac operation: A new approach through the venous bypasscircuit. Ann Thorac Surg 67: 604–609, 1999.

27. Sawa Y, Matsuda H, Shimazaki Y, et al: Evaluation of leukocyte-depleted terminal blood cardioplegic solution in patients under-going elective and emergency coronary artery bypass grafting.J Thorac Cardiovasc Surg 108: 1125–1131, 1994.

28. Gu YJ, de Vries AJ, Boonstra PW, van Oeveren W: Leukocytedepletion results in improved lung function and reduced in-flammatory response after cardiac surgery. J Thorac CardiovascSurg 112: 494–500, 1996.

29. Scholz M, Simon A, Matheis G, et al: Leukocyte filtration fails tolimit functional neutrophil activity during cardiac surgery. In-flamm Res 51: 363–368, 2002.

30. Matheis G, Scholz M, Simon A, et al: Timing of leukocyte filtrationduring cardiopulmonary bypass. Perfusion 16 (Suppl): 31–37,2001.

31. Alexiou C, Sheppard S, Smith D, et al: Effect of blood temperatureon the efficacy of systemic leucodepletion during cardiopulmo-nary bypass: A prospective randomized clinical study. ASAIO J51: 802–807, 2005.

32. Salamonsen RF, Anderson J, Anderson M, et al: Total leukocytecontrol for elective coronary bypass surgery does not improveshort-term outcome. Ann Thorac Surg 79: 2032–2038, 2005.

33. Gott JP: Leukodepletion and aprotinin improve clinical outcomeafter extracorporeal circulation. Perfusion 16 (Suppl): 5–9,2001.

34. Smit JJ, de Vries AJ, Gu YJ, van Oeveren W: Efficiency and safetyof leukocyte filtration during cardiopulmonary bypass for car-diac surgery. Transfus Sci 20: 151–165, 1999.

35. Fabbri A, Manfredi J, Piccin C, et al: Systemic leukocyte filtrationduring cardiopulmonary bypass. Perfusion 16 (Suppl): 11–18,2001.

36. Mair P, Hoermann C, Mair J, et al: Effects of a leucocytedepleting arterial line filter on perioperative proteolytic en-zyme and oxygen free radical release in patients undergoingaortocoronary bypass surgery. Acta Anaesthesiol Scand 43:452– 457, 1999.

37. Mihaljevic T, Tonz M, von Segesser LK, et al: The influence ofleukocyte filtration during cardiopulmonary bypass on postop-erative lung function. A clinical study. J Thorac Cardiovasc Surg109: 1138–1145, 1995.

38. Patel AN, Sutton SW, Livingston S, et al: Clinical benefits ofleukocyte filtration during valve surgery. Am J Surg 186: 636–639; discussion 639–640, 2003.

39. Efstathiou A, Vlachveis M, Tsonis G, et al: Does leukodepletionduring elective cardiac surgery really influence the overallclinical outcome? J Cardiovasc Surg (Torino) 44: 197–204,2003.

40. Jadad AR, Moore RA, Carroll D, et al: Assessing the quality ofreports of randomized clinical trials: Is blinding necessary?Control Clin Trials 17: 1–12, 1996.

41. Moher D, Cook DJ, Eastwood S, et al: Improving the quality ofreports of meta-analyses of randomised controlled trials: TheQUOROM statement. Quality of Reporting of Meta-analyses.Lancet 354: 1896–1900, 1999.

42. Cochrane Reviewers’ Handbook 4.2.1 [updated December2003].Chichester, John Wiley, 2004.

43. Egger M, Davey Smith G, Schneider M, Minder C: Bias in meta-analysis detected by a simple, graphical test. BMJ 315: 629–634, 1997.

44. DerSimonian R, Laird N: Meta-analysis in clinical trials. ControlClin Trials 7: 177–188, 1986.

45. Higgins JP, Thompson SG, Deeks JJ, Altman DG: Measuring in-consistency in meta-analyses. BMJ 327: 557–560, 2003.

87SYSTEMIC LEUKOCYTE FILTRATION AND PULMONARY INJURY AFTER CARDIAC SURGERY

46. Higgins JP, Thompson SG: Quantifying heterogeneity in a meta-analysis. Stat Med 21: 1539–1558, 2002.

47. Egger M, Smith GD: Misleading meta-analysis. BMJ 310: 752–754, 1995.

48. Johnson D, Thomson D, Mycyk T, et al: Depletion of neutrophilsby filter during aortocoronary bypass surgery transiently im-proves postoperative cardiorespiratory status. Chest 107: 1253–1259, 1995.

49. Hurst T, Johnson D, Cujec B, et al: Depletion of activated neutro-phils by a filter during cardiac valve surgery. Can J Anaesth 44:131–139, 1997.

50. Hachida M, Hanayama N, Okamura T, et al: The role of leukocytedepletion in reducing injury to myocardium and lung duringcardiopulmonary bypass. ASAIO J 41: M291–M294, 1995.

51. Di Salvo C, Louca LL, Pattichis K, et al: Does activated neutrophildepletion on bypass by leukocyte filtration reduce myocardialdamage? A preliminary report. J Cardiovasc Surg (Torino) 37:93–100, 1996.

52. Lust RM, Bode AP, Yang L, et al: In-line leukocyte filtration duringbypass. Clinical results from a randomized prospective trial.ASAIO J 42: M819–M822, 1996.

53. Baksaas ST, Videm V, Mollnes TE, et al: Leucocyte filtration duringcardiopulmonary bypass hardly changed leucocyte counts anddid not influence myeloperoxidase, complement, cytokines orplatelets. Perfusion 13: 429–436, 1998.

54. Sahlman A, Ahonen J, Salo JA, Ramo OJ: No impact of a leucocytedepleting arterial line filter on patient recovery after cardiopul-monary bypass. Acta Anaesthesiol Scand 45: 558–563, 2001.

55. Karaiskos TE, Palatianos GM, Triantafillou CD, et al: Clinicaleffectiveness of leukocyte filtration during cardiopulmonarybypass in patients with chronic obstructive pulmonary disease.Ann Thorac Surg 78: 1339–1344, 2004.

56. Alexiou C, Tang AT, Sheppard SV, et al: A prospective random-ized study to evaluate the effect of leukodepletion on the rate ofalveolar production of exhaled nitric oxide during cardiopul-monary bypass. Ann Thorac Surg 78: 2139–2145; discussion2145, 2004.

57. Sheppard SV, Gibbs RV, Smith DC: Does the use of leucocytedepletion during cardiopulmonary bypass affect exhaled nitricoxide production? Perfusion 19: 7–10, 2004.

58. Stefanou DC, Gourlay T, Asimakopoulos G, Taylor KM: Leu-codepletion during cardiopulmonary bypass reduces bloodtransfusion and crystalloid requirements. Perfusion 16: 51–58,2001.

59. Alexiou C, Tang AA, Sheppard SV, et al: The effect of leucodeple-

tion on leucocyte activation, pulmonary inflammation and re-spiratory index in surgery for coronary revascularisation: Aprospective randomised study. Eur J Cardiothorac Surg 26:294–300, 2004.

60. Sheppard SV, Gibbs RV, Smith DC: Does leucocyte depletionduring cardiopulmonary bypass improve oxygenation indicesin patients with mild lung dysfunction? Br J Anaesth 93: 789–792, 2004.

61. Koskenkari J, Rimpilainen J, Biancari F, et al: Leukocyte depletingfilter attenuates myocardial injury during elective coronary ar-tery bypass surgery. Scand Cardiovasc J 39: 358–368, 2005.

62. Leal-Noval SR, Amaya R, Herruzo A, et al: Effects of a leukocytedepleting arterial line filter on perioperative morbidity in pa-tients undergoing cardiac surgery: A controlled randomizedtrial. Ann Thorac Surg 80: 1394–1400, 2005.

63. Chen YF, Tsai WC, Lin CC, et al: Effect of leukocyte depletion onendothelial cell activation and transendothelial migration ofleukocytes during cardiopulmonary bypass. Ann Thorac Surg78: 634–642; discussion 642–633, 2004.

64. Blettner M, Sauerbrei W, Schlehofer B, et al: Traditional reviews,meta-analyses and pooled analyses in epidemiology. Int J Epi-demiol 28: 1–9, 1999.

65. Olivencia-Yurvati AH, Ferrara CA, Tierney N, et al: Strategicleukocyte depletion reduces pulmonary microvascular pressureand improves pulmonary status post-cardiopulmonary bypass.Perfusion 18 (Suppl 1): 23–31, 2003.

66. Lim HK, Anderson J, Leong JY, et al: What is the role of leukocytedepletion in cardiac surgery? Heart Lung Circ 16: 243–253,2007.

67. Samankatiwat P, Samartzis I, Lertsithichai P, et al: Leucocytedepletion in cardiopulmonary bypass: A comparison of fourstrategies. Perfusion 18: 95–105, 2003.

68. Warren O, Wallace S, Massey R, et al: Does systemic leukocytefiltration affect peri-operative haemorrhage in cardiac surgery?A systematic review and meta-analysis. ASAIO J 53: 514–521,2007.

69. Olivencia-Yurvati AH, Wallace N, Ford S, Mallet RT: Leukocytefiltration and aprotinin: Synergistic anti-inflammatory protec-tion. Perfusion 19 (Suppl 1): S13–S19, 2004.

70. Brown P, Brunnhuber K, Chalkidou K, et al: How to formulateresearch recommendations. BMJ 333: 804–806, 2006.

71. Shapiro S: Meta-analysis/Shmeta-analysis. Am J Epidemiol 140:771–778, 1994.

72. Petitti DB: Of babies and bathwater. Am J Epidemiol 140: 779–782, 1994.

88 WARREN ET AL.