Where are the Systematic Reviews in Transfusion Medicine?A Study of the Transfusion Evidence Base

Carolyn Dorée, Simon Stanworth, Susan J. Brunskill, Sally Hopewell, Chris J. Hyde, and Mike F. Murphy

Transfusion medicine has become a large and complexspecialty. Although there are now systematic reviewscovering many aspects of transfusion, these span alarge number of clinical areas and are published acrossmore than a hundred different medical journals,making it difficult for transfusion medicine practi-tioners and researchers to keep abreast of the currenthigh-level evidence. In response to this problem, NHSBlood and Transplant's Systematic Review Initiative(SRI) has produced a comprehensive overview ofsystematic reviews in transfusion medicine. A sys-tematic search (to December 2009) and screeningprocedure were followed by the appraisal of systema-tic reviews according to predefined inclusion criteria.The 340 eligible systematic reviews were mapped to

From the Systematic Review Initiative, NHS Blood andTransplant (Oxford Centre), John Radcliffe Hospital, Oxford;NHS Blood & Transplant, Oxford Radcliffe Hospitals NHSTrust, and University of Oxford, Oxford; UK Cochrane Centre,Oxford, and Peninsula Technology Assessment Group(PenTAG), Peninsula College of Medicine and Dentistry,University of Exeter.Address reprint requests to Carolyn Dorée, Systematic Review

Initiative, NHS Blood and Transplant (Oxford Centre), JohnRadcliffe Hospital, Oxford, UK.E-mail: carolyn.doree@nhsbt.nhs.uk0887-7963/$ - see front matterCrown Copyright © 2010 Published by Elsevier Inc. All rights

reserved.doi:10.1016/j.tmrv.2010.05.003

Transfusion286

10 transfusion intervention groups and 14 topicgroups within clinical medicine. Trends in the sys-tematic review literature were examined and gaps inthe literature described. The spread of systematicreviews across clinical areas was found to be veryuneven, with some areas underreviewed and otherswith multiple systematic reviews on the same topic,making the identification of the best evidence forcurrent transfusion practice a continuing challenge.References and links to all systematic reviews includedin this overview can be freely accessed via the SRI'snew online database, the Transfusion Evidence Library(www.transfusionguidelines.org).Crown Copyright © 2010 Published by Elsevier Inc. Allrights reserved.

RANSFUSION MEDICINE HAS become a less biased in their approach to reviewing a subject

T huge specialty, and the traditional adage thatthis field is largely confined to blood componenttransfusion and massive transfusion is now outdat-ed, particularly in view of the continued develop-ments in such areas as alternatives to transfusion,cellular and tissue therapy, and organ transplanta-tion. A result of the expansion of transfusion-related therapies over the past 10 to 15 years, alongwith the concurrent move toward evidence-basedpractice, has been an increased imperative tounderstand both the efficacy and effects of themany different interventions within this specialty.It is now generally accepted that the best way for a

reader to understand the totality of the evidence basefor a topic is to identify systematic reviews.1

Systematic reviews aim to be more explicit and

than traditional literature reviews and can make theresults of different clinical trials much moreaccessible to clinicians. The aim of this project wasto describe the systematic review literature across allareas relevant to transfusion medicine and to mapthem to topic groups within clinical medicine. Anequally important objective was to identify areaswhere there are gaps or deficiencies in the evidencefor the clinical use of blood and its alternatives.

METHODS

Searching

A comprehensive list of terms and synonyms wasdeveloped through discussion between SystematicReview Initiative (SRI) transfusion clinicians, theSenior Information Scientist, and the InformationSpecialist to ensure that all areas of interest wouldbe covered by the search. A draft search strategywas devised for MEDLINE in the first instance,combining MeSH and free-text terms, which wasthen tested in combination with a revised andupdated version of the SIGN systematic reviewfilter (http://www.sign.ac.uk). Once the MEDLINEsearch strategy was finalized, it was translated foruse in EMBASE and The Cochrane Library.

Searches were first run in October 2008 andupdated in December 2009 in the followingdatabases: The Cochrane Library, Issue 4 2009(Cochrane Database of Systematic Reviews,DARE, HTA, NHSEED), MEDLINE (1950 on-wards), EMBASE (1980 onwards), and the SRI

Medicine Reviews, Vol 24, No 4 (October), 2010: pp 286-294

Fig 1. Characteristics of transfusion intervention groups.

287SYSTEMATIC REVIEWS IN TRANSFUSION MEDICINE

Systematic Review Handsearch Database. Thislatter database contains all systematic reviewsfound via the SRI's ongoing prospective andretrospective hand-searching project.2 (Furtherdetails are available from the authors).

Table 1. Systematic Reviews by

Transfusion interventionNo. of systematic

reviewsNo. ofRCTs

No. ofobservational

studiesMedian noper syst

Alternatives totransfusion

131 2583 468 12

Fractionated bloodproducts

78 1053 205 6.5

Transplant 43 278 263 6Volume replacement 20 563 94 16Red cells 20 167 159 4.5Exchange therapy 11 117 8 3Platelets 12 103 34 6.5Transfusion practice 16 72 124 3White cells 4 22 66 6Plasma 5 77 4 6Totals 340 5035 1425 7

Systematic Review Eligibility

Systematic reviews were considered eligible forinclusion in this overview if the intervention wasrelevant to transfusion medicine (see Fig 1 for

Transfusion Intervention

. (IQR) of RCTsematic review

Total sample size(% from RCTs)

Publicationyear range

No. (%)reporting appraisa

(6-23) 632,515 (55) 1995-2009 104 (79)

(2-9) 1,011,888 (96) 1995-2009 65 (83)

(0.5-9.5) 80,974 (56) 1998-2009 33 (77)(7-34) 78,783 (59) 1998-2009 14 (70)(2-9) 372,100 (11) 1998-2009 15 (75)(1-6) 7298 (96) 2002-2007 10(91)(2-17) 11,105 (68) 1998-2009 8 (67)(0-9) 1,513,133 (0.5) 1999-2009 13(81)(3-8.5) 1400 (78) 2003-2009 3 (75)(1-13) 5911 (74) 2004-2009 5 (100)(3-18) 3,715,107 (40) 1995-2009 270 (79)

l

Tab

le2.Sys

tematic

Rev

iewsby

Clin

ical

Area

Trans

fusion

interven

tion

Cardiac

surgery

andca

rdiology

Critica

lca

rean

dtrau

ma

Neu

rology

Obs

tetrics

Other

surgery⁎

Infectious

dise

ases

Other

med

icine†

Neo

nates

Ped

iatrics

Non

maligna

nthe

matolog

yMaligna

nthe

matolog

yOnc

olog

ySolid

orga

ntran

splant

Ove

rviews‡

Totals

Red

cells

13

12

12

41

23

20

Platelets

12

21

15

12

White

cells

11

24

Plasm

a1

11

11

5Frac

tion

ated

bloo

dprod

ucts

412

13

10

22

10

510

43

378

Exch

ange

therap

y4

13

11

111

Volum

ereplac

emen

t2

10

13

22

20

Alterna

tive

sto

tran

sfus

ion

26

64

31

112

35

54

25

27

131

Trans

plan

t7

41

44

18

32

43

Trans

fusion

prac

tice

84

12

Totals

42

32

17

15

41

616

23

17

34

35

30

523

336

⁎Includ

esalls

urge

rywiththeex

ceptionof

cardiacsu

rgery.

†Includ

esrena

ldisea

se,diab

etes

,an

dup

perga

strointestinal

blee

ding

.‡Rev

iewsof

interven

tion

across

ava

rietyof

clinical

settings

.

288 DORÉE ET AL

project scope) and if the authors had performed asystematic search of PubMed or MEDLINE plus atleast one other electronic database or suitablealternative (for example, systematic handsearchingof multiple conference abstracts). Systematicreviews that searched only PubMed or MEDLINEwere therefore excluded, a decision based on thefact that a comprehensive search is generallyconsidered a prerequisite of a high-quality system-atic review.3 No systematic reviews were excludedon the grounds of publication date, language,patient group, clinical setting, or type of includedstudies. Full text copies were obtained wherenecessary, and lead authors contacted if searcheswere not sufficiently detailed.

Data Extraction

The data extracted from each included systematicreview were as follows: the type of intervention andcomparator(s), the clinical setting, the patientpopulation, the types and number of includedstudies (ie, randomized controlled trials [RCTs]and/or observational studies), and the number ofpatients included in each trial (the sample size). Wealso recorded whether each systematic reviewconducted any kind of quality assessment of theirincluded studies (although the rigor of these qualityassessments was not evaluated).

Data Analysis

Data analysis was descriptive and limited to anexploration of the epidemiological nature of theincluded systematic reviews. To assist in the analysis,matrices of 10 transfusion intervention groups and14 clinical areas were created to which the eligiblesystematic reviews were mapped, based on consul-tation with SRI transfusion clinicians (Tables 1and 2). The characteristics of the 10 transfusionintervention groups are presented in Figure 1.

RESULTS

Characteristics of Included Systematic Reviews

The searches yielded a total of 10,926 references,of which 9701 (89%) were excluded on firstscreening. The remaining 1225 references wereappraised according to study type and the quality ofthe search, which led to a further 841 beingexcluded. Of these 841 excluded references, 176reviews (published between 1992 and 2009) wererejected because only PubMed or MEDLINE was

289SYSTEMATIC REVIEWS IN TRANSFUSION MEDICINE

searched by the authors. Our eligibility criterionthus effectively reduced the number of includedsystematic reviews by one third. During the processof data extraction, 44 of the remaining 384systematic reviews proved to be duplicate publica-tions (often conference abstracts) and were exclud-ed, leaving a core set of 340 primary systematicreviews for inclusion in the project (Fig 2).

The 340 eligible systematic reviews werepublished across 130 journals, with 94 of thesejournals (72%) publishing only 1 systematicreview. The journal with the largest number ofsystematic reviews was The Cochrane Library with92 (27%), followed by Transfusion (n = 12) andBlood (n = 11). Other types of publications includedhealth technology assessments (n = 9), guidelines(n = 2), websites (n = 2), and 1 book chapter.

Of the 340 systematic reviews, 257 (75%)included only RCTs, 57 included both RCTs andobservational studies, and 27 included only obser-vational studies. The number of RCTs and thenumber of observational studies included within anindividual systematic review ranged from 0 to 211

Fig 2. Searching flow diag

and from 0 to 102, respectively, and the samplesizes ranged from 0 to 567,476 (RCTs) and from 0to 1,504,198 (observational studies).

The included systematic reviews were publishedbetween 1995 and 2009, with a significant increasein the number published during the last 5 or 6 years(Fig 3). This is particularly the case in certainclinical areas—for example, the use of exchangetherapy, white cells, and plasma—which saw alarge increase in the number of systematic reviewspublished since 2002 (Table 1).

Systematic reviews generally showed a similarlevel of quality assessment, with 270 of the 340(79%) reporting some kind of appraisal of theirincluded studies (Table 1). It should be noted,however, that time constraints meant that it was notpossible to make a full quality assessment of the340 included systematic reviews.

Systematic Reviews Mapped by TransfusionIntervention Group

Table 1 presents the 340 eligible systematicreviews grouped according to type of transfusion

ram, December 2009.

Fig 3. Systematic reviews in transfusion medicine by year.

290 DORÉE ET AL

intervention. The intervention groups with thehighest number of systematic reviews were Alter-natives to Transfusion with 131 systematic reviews(RCTs, n = 2583; observational studies, n = 468) andFractionated Blood Products with 78 systematicreviews (RCTs, n = 1053; observational studies, n =205). The areas of least activity were Plasma with 5systematic reviews (RCTs, n = 77; observationalstudies, n = 4) andWhite Cells with only 4 publishedsystematic reviews (RCTs, n = 22; observationalstudies, n = 66) (Fig 4). For almost all transfusionintervention groups, the number of RCTs exceededthat of observational studies, with the exception ofTransfusion Practice and White Cells.Of the 2 largest intervention groups, Alternatives

to Transfusion (n = 131) was dominated by thesubgroups Erythropoietin with 53 systematicreviews (RCTs, n = 999; observational studies, n =193) and Antifibrinolytic Agents with 38 systematicreviews (RCTs, n = 1059; observational studies, n =24), which together accounted for 69% of the 131systematic reviews and 80% of the total number ofRCTs. Within Fractionated Blood Products, by far,the largest subgroup was Immunoglobulins, with 51systematic reviews (RCTs, n = 690; observationalstudies, n = 115). (Data for the subgroups are notpresented here but are available from the authors).The median number of RCTs per systematic

review was 7 (interquartile range [IQR], 3-18). Twointervention groups had a much larger median

number of RCTs per systematic review than theoverall median: Volume Replacement, with amedian number of 16 RCTs per systematic review(IQR, 7-34), and Alternatives to Transfusion with amedian of 12 RCTs per systematic review (IQR, 6-23). The median number of RCTs per systematicreview did not differ significantly from the overallmedian number of RCTs per systematic review inthe remaining 8 transfusion intervention groups.

The total number of participants per interventiongroup ranged from 1400 patients (White Cells) to1,513,133 patients (Transfusion Practice), with theoverall total exceeding 3,700,000 patients. In 2areas (Fractionated Blood Products and ExchangeTherapy), the proportion of RCTs far exceeded thatof observational studies (96% RCTs in bothgroups), and for most of the intervention groups(83%), RCT data accounted for 55% or more of thesample size. The 2 exceptions, and those with thelowest proportion of RCTs, were Red Cells (11%RCTs) and Transfusion Practice (0.5% RCTs), thelatter having by far the largest sample size of allgroups (n = 1,513,133 participants), 99.5% of whichconsisted of observational data from 1 systematicreview.4 For the Red Cell systematic reviews, only11% of the sample size came fromRCTs, despite thegroup containing more RCTs (n = 167) thanobservational studies (n = 159), a result of the factthat a large proportion (73%) of its large total samplesize (n = 372,100) originated from 45 observational

Fig 4. Systematic reviews by transfusion intervention.

291SYSTEMATIC REVIEWS IN TRANSFUSION MEDICINE

studies assessed in 1 systematic review.5 The factthat only 40% of the overall total sample sizeoriginated from RCTs, a surprisingly small propor-tion, was again due to the large amount ofobservational data included in the Red Cells andTransfusion Practice systematic reviews.

Systematic Reviews Mapped by Clinical Area

The matrix in Table 2 illustrates the distributionof 336 of the included systematic reviews across 14clinical areas; 4 systematic reviews relating toblood and stem cell donation were removed fromthis section because they were not applicable to aspecific clinical area.

The clinical areas containing the highest numberof systematic reviews were Cardiac (n = 42),Surgery (n = 41), Malignant Hematology (n = 35),Nonmalignant Hematology (n = 34), and CriticalCare + Trauma (n = 32). The 2 clinical areas withthe fewest systematic reviews were InfectiousDiseases (n = 6) and Solid Organ Transplant (n =5). Other noteworthy areas were Obstetrics, with arelatively small number of systematic reviews (n =15), and Oncology (n = 30), with 24 (80%) of itssystematic reviews on the use of erythropoietin.

The final clinical group in the matrix is Overviews(n = 23), which covered transfusion interventionsacross a number of clinical settings and containedsystematic reviews spread evenly across mostrelevant intervention groups.

Viewed from the perspective of the transfusionintervention groups, Alternatives to Transfusion andFractionated Blood Products were the most prevalent(covering 13 and 12 clinical areas, respectively), theformer with systematic review numbers highest inCardiac (n = 26), Surgery (n = 31), andOncology (n =25) and the latter with systematic reviews fairlyevenly spread. Systematic reviews in Red Cellscovered 10 areas (notable exceptions: Obstetrics,Pediatrics, and Solid Organ Transplant), followed byTransplant (covering 8 areas). The groups coveringthe fewest clinical areas were Exchange Therapy (6clinical areas), Plasma (5 clinical areas), White Cells(3 clinical areas), and Transfusion Practice (2 clinicalareas). However, whereas Exchange Therapy, Plas-ma, and Transfusion Practice have all been thesubject of overviews (fresh frozen plasma, forexample, has been systematically reviewed acrossall relevant clinical areas), only 4 aspects of bedsideand behavioral Transfusion Practice (published

292 DORÉE ET AL

between 2005 and 2009) have so far been system-atically reviewed.6

DISCUSSION

This project has identified a large number ofsystematic reviews covering a broad range of topicswithin transfusion medicine and reveals a signifi-cant growth in the numbers of systematic reviewspublished during the past 10 years. However,despite the quantity of systematic reviews, whichappears to indicate a wide evidence base, thefindings raise a number of important limitationsabout the strength of any conclusions. It is clear thatthe spread of systematic reviews is very uneven,with certain areas characterized by a large number ofsystematic reviews and others by a relative paucity.As a generalization, those areas with multiplesystematic reviews tend to be those associatedwith pharmaceutical drugs—for example, erythro-poietin (53 systematic reviews) and antifibrinolyticagents (38 systematic reviews)—or with fractionat-ed blood components, which arguably have someoverlap with pharmaceutical drugs. Some system-atic reviews appeared to follow a very narrow timeframe; for example, 50% of systematic reviews oferythropoietin have been published since 2007, and80% since 2003, and 83% of systematic reviews ofrecombinant factor VIIa have been published since2007. Some clinical areas were characterized by afew or no systematic reviews; for example, therewere no red cells systematic reviews in Obstetrics,Pediatrics, and Solid Organ Transplant; no Plasmasystematic reviews in Neonates; no Volume Re-placement systematic reviews in Pediatrics andSolid Organ Transplant; and no Platelets systematicreviews in Critical Care + Trauma, Other Medicine,Neurology, Oncology, and Solid Organ Transplant.Generally, blood components with short half-liveswere characterized by much smaller numbers ofsystematic reviews, and for certain blood compo-nents, such as cryoprecipitate, no systematicreviews were identified at all.When evaluating the blood components system-

atic reviews, it is interesting to make a comparisonwith data from studies informing us where bloodcomponents go and which patient groups receivemost blood. More recent data from a number ofcenters indicate greater blood usage among certainmedical specialties, but when one views thenumbers of systematic reviews by blood usagecategory, it is clear that some of these medical areas

are very underrepresented by numbers of system-atic reviews. For example, the Epidemiology andSurvival of Transfusion Recipients (EASTR)Study7 found that cardiac patients receive 17% ofplatelets issued per year in England and Wales, butonly 1 systematic review (on platelet-rich plasma-pheresis) has been published in this area.8

This project has focused on the numbers ofsystematic reviews, mapping them to differentintervention groups or clinical areas. The analysishas not taken into consideration the numbers ofduplicate systematic reviews covering the samearea or addressing the same question, and it has notevaluated the numbers of RCTs shared/repeatedbetween systematic reviews, which might beexpected to inflate the median numbers of RCTsper systematic review across all groups. In the caseof systematic reviews of autologous transfusion, forexample, the high level of RCTs (n = 115) and largeRCT sample size (8480 patients, 97% of the total)of the Cell Salvage systematic reviews was partiallyaccounted for by the fact that 1 systematic review9

contained all 28 RCTs (2195 patients) from anotherearlier systematic review,10 as well as a numberfrom a later systematic review.11 These pointssuggest a need for greater clarity about how newpublished systematic reviews relate to earlierpublished systematic reviews in the same topic area.

The large number of systematic reviews alsoraises questions for the reader as to whichsystematic review(s) should be consulted forspecific areas or questions. There is an increasingrecognition of the need for “systematic reviews ofsystematic reviews” to provide a summary of wheredifferent systematic reviews fit within specificclinical areas. There is also a need to investigatethe “added value” of multiple systematic reviews ofspecific interventions; for example, it would beinteresting to establish if all the systematic reviewson a certain topic came to the same conclusion and,if not, to establish the reason(s) for this—inparticular, whether there is any relationship be-tween a systematic review's quality and/or potentialconflicts of interest and its results. Multiplepublication of the same systematic review is alsoa potential issue; whereas Cochrane replaces theprevious version of a systematic review with itsupdate, this does not happen with systematicreviews published elsewhere, and it can be difficultto establish if a particular systematic review is themost recent or definitive version. Each new

293SYSTEMATIC REVIEWS IN TRANSFUSION MEDICINE

systematic review should ideally provide a sum-mary of all previous topic-specific systematicreviews and explain how this systematic reviewrelates to earlier secondary research.12 This wouldbe similar to how proposals for new clinical trialsare now expected to include a systematic review oftrials in the relevant field.

There is more interest now in understanding thequality of systematic reviews, and checklists existas aide memoirs. Most systematic reviews (n = 270)in this project reported a quality appraisal of theirincluded studies, which is to be expected andlargely the result of our methodology: our eligibilitycriterion (a search of MEDLINE plus 1 otherdatabase), which is based on the recently developedAMSTAR methodological instrument for assessingsystematic reviews,3 means that included system-atic reviews are more likely to be both of highmethodological quality and more recent—all werepublished from 1995 onward (93% in the last 10years, 78% in the last 5 years). Nevertheless, asurprising number of systematic reviews (n = 185)had to be excluded because only 1 database wassearched; this is perhaps less surprising for thosepublished during the 1990s (26 systematic reviews)but very surprising for the 159 systematic reviewspublished since 2000, as there have been severalstudies during the past 2 decades researching andemphasizing the importance of a high-quality andcomprehensive search when carrying out system-atic reviews, beginning with the early work of TheCochrane Collaboration.13

Implications of This Overview

This project has raised a number of opportunitiesfor new areas of research, in addition to emphasiz-ing the need for all new systematic reviews toprovide information both on existing systematicreviews within their topic area and on how a newsystematic review will improve the current evi-dence base—for example, by an updated search orby more stringent quality assessment. One poten-tially fruitful line of new research is the inclusion of

more observational studies within systematicreviews of clinically important areas with relativelylow numbers of RCTs; the number of observationalstudies in Table 1 does not reflect the number ofobservational studies in each area, but only thoseincluded within systematic reviews, which tend toshy away from observational studies as they aredifficult to identify, appraise, and analyze. Howev-er, the data suggest that for some areas the numbersof observational studies will be greater than forRCTs, and that the evidence base might be greatlyincreased if these studies were systematicallyreviewed alongside RCTs.

The wide spread of our 340 included systematicreviews across 130 journals from many differentclinical specialties has also highlighted anotherproblem for transfusion medicine researchers andpractitioners: the great difficulty in keeping abreast ofcurrent secondary research in this field. Out of thisproject, the SRI has, however, created a solution tothis problem—its recently launched TransfusionEvidence Library. This is a comprehensive onlinecollection of high-quality systematic reviews rele-vant to transfusion medicine, made freely availablewithin a fully searchable database. Simple andadvanced search options are provided, with fullcitation details given for each systematic review, andlinks to its entry (and abstract) in PubMed and to thefull text (where available—for recent papers, journalsubscriptions are usually required). The TransfusionEvidence Library is updated monthly and currently(as of March 2010) contains 446 systematic reviewsand 39 economic evaluations, including the SRI'sown completed systematic reviews in transfusionmedicine (21 to date). It also contains over 1300RCTs relevant to transfusionmedicine that have beenhandsearched by the SRI.

The UK Blood Transfusion and Tissue Trans-plantation Services (UKBTS) SRI TransfusionEvidence Library contains references and links toall 340 systematic reviews included in this overviewand can be freely accessed via the UKBTS websiteat http://www.transfusionguidelines.org.

REFERENCES

1. Mulrow CD: Rationale for systematic reviews. BMJ309:597-599, 1994

2. Brunskill SJ, Hyde CJ, Stanworth SJ, et al: Improvingthe evidence base for transfusion medicine: The work of the

UK Systematic Review Initiative. Transfus Med 19:59-65,2009, pp. 62

3. Shea BJ, Grimshaw JM, Wells GA, et al: Developmentof AMSTAR: A measurement tool to assess the methodological

294 DORÉE ET AL

quality of systematic reviews. BMC Med Res Methodol 7:10,2007

4. Stainsby D, Brunskill S, Chapman CE, et al: Safety ofblood donation from individuals with treated hypertension ornon–insulin dependent type 2 diabetes—A systematic review.Vox Sang 98(3 Pt 2):431-440, 2010

5. Marik PE, Corwin HL: Efficacy of red blood celltransfusion in the critically ill: A systematic review of theliterature. Crit Care Med 36:2667-2674, 2008

6. Stanworth SJ, Brunskill SJ, Hyde CJ, et al: Is freshfrozen plasma clinically effective? A systematic review ofrandomized controlled trials. Br J Haematol 126:139-152,2004

7. Wells AW, Lewellyn CA, Casbard A, et al: The EASTRStudy: Indications for transfusion and estimates of transfusionrecipient numbers in hospitals supplied by the National BloodService. Transfus Med 19:315-328, 2009

8. Rubens FD, Fergusson D, Wells PS, et al: Platelet-richplasmapheresis in cardiac surgery: A meta-analysis of the effect

on transfusion requirements. J Thorac Cardiovasc Surg 116(4):641-647, 1998

9. Carless PA, Henry DA, Moxey AJ, et al: Cell salvage forminimising perioperative allogeneic blood transfusion. CochraneDatabase Syst Rev 4:CD001888, 2006

10. Huet C, Salmi LR, Fergusson D, et al: A meta-analysis ofthe effectiveness of cell salvage to minimize perioperativeallogeneic blood transfusion in cardiac and orthopedic surgery.International Study of Perioperative Transfusion (ISPOT)Investigators. Anesth Analg 89:861-869, 1999

11. Wang G, Bainbridge D, Martin J, et al: The efficacy of anintraoperative cell saver during cardiac surgery: A meta-analysisof randomized trials. Anesth Analg 109:320-330, 2009

12. Hyde C, Stanworth S, Brunskill S, et al: Putting clinicaltrials into context. Lancet 266:980-981, 2005

13. Chalmers I: The Cochrane Collaboration. Preparing,maintaining and disseminating systematic reviews of the effectsof health care. Ann N Y Acad Sci 703:156-163, 1993 [discussion163-165]