meta-analysis of defunctioning stomas in low anterior resection for rectal cancer

Meta-analysis

Meta-analysis of defunctioning stomas in low anterior resectionfor rectal cancer

W. S. Tan1, C. L. Tang1, L. Shi2 and K. W. Eu1

1Department of Colorectal Surgery, Singapore General Hospital and 2Clinical Trials and Epidemiological Unit, SingaporeCorrespondence to: C. L. Tang, Department of Colorectal Surgery, Singapore General Hospital, Outram Road, Singapore 169608(e-mail: [email protected])

Background: A defunctioning stoma is frequently created to minimize the impact of any subsequentanastomotic leak after a low rectal anastomosis. This review evaluates the need for routine stomaformation.Methods: A meta-analysis was performed of randomized controlled trials (RCTs) and non-randomizedstudies with an interventional group evaluating the need for a defunctioning stoma after low anteriorresection for rectal cancer. Primary outcomes analysed included clinical anastomotic leak rate,reoperation rate and mortality related to leak.Results: Four RCTs and 21 non-randomized studies, with 11 429 patients in total, were analysed. Meta-analysis of the RCTs showed a lower clinical anastomotic leak rate (risk ratio (RR) 0·39 (95 per cent c.i.0·23 to 0·66); P < 0·001) and a lower reoperation rate (RR 0·29 (0·16 to 0·53); P < 0·001) in the stomagroup. Meta-analysis of the non-randomized studies showed a lower clinical anastomotic leak rate (RR0·74 (0·67 to 0·83); P < 0·001), lower reoperation rate (RR 0·28 (0·23 to 0·35); P < 0·001) and lowermortality rate (RR 0·42 (0·28 to 0·61); P < 0·001) in the stoma group.Conclusion: A defunctioning stoma decreases clinical anastomotic leak rate and reoperation rate. It isrecommended after low anterior resection for rectal cancer.

Paper accepted 20 January 2009Published online in Wiley InterScience (www.bjs.co.uk). DOI: 10.1002/bjs.6594

Introduction

Historically, patients with mid to distal rectal cancers had toundergo abdominoperineal resection to achieve adequateoncological clearance. However, with better equipmentand improved surgical techniques, low anterior resectionwith a low colorectal or coloanal anastomosis has becomethe technique of choice. This achieves sphincter preserva-tion without compromising the oncological outcome.

There is no universally accepted definition of lowanterior resection in the literature. In the authors’institution, it is generally defined as a resection inwhich the level of the anastomosis is below theperitoneal reflection. Studies have demonstrated that sucha low anastomosis carries a considerably higher risk ofanastomotic leakage1–3; the reported incidence of clinicalleakage varies from 2·6 to 26·2 per cent3–6. Other acceptedrisk factors for leakage include male sex and absence of adefunctioning stoma1,2,7–11. Even experienced surgeonssometimes find it difficult to predict which patient will

have an anastomotic leak, and know that leaks may occureven when the anastomosis is technically sound and therisk factors for leakage are absent.

An anastomotic leak is associated with increased mor-tality as well as morbidity in the form of reoperation,permanent stoma, septic complications, and poor func-tion as a result of stenosis and decreased compliance of theneorectum12,13. There is also evidence for an increased riskof local cancer recurrence and a decreased long-term sur-vival after pelvic sepsis14–16. With this background, a stomafor proximal faecal diversion has been suggested for lowanterior resection, but evidence of benefit has been conflict-ing. Some authors have recommended routine faecal diver-sion, believing that it results in a lower incidence of anasto-motic leakage and need for surgical intervention8–10,17–22.Others have found similar leakage and mortality ratesdespite the presence of a stoma. They recommend a selec-tive approach, with stoma formation only when there isconcern about the anastomosis owing to difficult surgery,

Copyright 2009 British Journal of Surgery Society Ltd British Journal of Surgery 2009; 96: 462–472Published by John Wiley & Sons Ltd

Defunctioning stomas in low anterior resection for rectal cancer 463

incomplete doughnuts or a positive air leak test1,6,23–26. Itshould not be forgotten that a stoma may itself cause signif-icant morbidity, including discomfort and inconvenience,high output with dehydration resulting, and anastomoticcomplications at the stoma closure site27–29.

This review evaluates the need for a routine stoma,by performing a meta-analysis of the available evidence.Specific outcomes studied include the incidence ofsymptomatic anastomotic leak, the need for surgicalreintervention and mortality rates, with and without astoma. The morbidity and mortality of stoma closureare also assessed. Finally, the risk factors for anastomoticleakage are discussed.

Methods

Literature search

Relevant studies published between January 1966 and April2008 were identified by searching Medline, Embase, TheCochrane Library and CINAHL databases. The followingsearch terms were used: low rectal cancer, low anastomosis,low anterior resection, anastomotic leak and stoma. Bothfree text search and MeSH search for keywords wereemployed. Two investigators independently performed theliterature search. The ‘related article’ function was usedto widen the search criteria. Relevant articles referencedin the publications were also searched for additionalstudies for potential inclusion. The search was restrictedto publications in English.

Study selection and data extraction

Randomized controlled trials (RCT) and non-randomizedstudies with an interventional group comparing surgeryfor low rectal tumours with and without a stoma wereselected. Publications were included if they comparedlow anterior resection with a defunctioning stoma to lowanterior resection without a stoma, in terms of anastomoticleak rate and/or sequelae of leakage (surgical interventionfor leak and mortality) in each group. If a study had bothhigh and low anterior resections, it was included only if abreakdown of data by level of anastomosis was available.In these studies, only the subgroup of patients with alow anastomosis was included. Studies were excluded ifall anterior resections (both high and low) were included,without any breakdown of level of anastomosis. Non-comparative studies, case series and case reports werenot included. Studies of laparoscopic or hand-assistedresections were not excluded.

Data on the following outcomes were sought: clinicalanastomotic leak rate, rate of surgical intervention for leak,

mortality rate related to leak, and morbidity and mortalityof stoma closure. Studies were included if they providedinformation on at least one of these outcome measures. Alleligible studies were assessed for quality of methodologyand relevance to the objective of the review. The Cochranemethods of assessment of the risk of bias and quality ofthe studies were employed30. No rating of the studies wasperformed; each was accepted or rejected based on thecriteria noted above. Each trial was critically appraised bytwo investigators. Data were extracted independently andcompared. Any discrepancies were resolved by discussion.When required information was not specified in the article,e-mails were sent to the authors requesting them to retrievethe relevant information, if possible. If data were notreported specifically for certain outcomes and if there wasno reply from the author via e-mail, the data were regardedas missing. No assumptions were made from these missingdata.

Clinical anastomotic leakage was defined as the presenceof clinical signs, such as peritonitis, pelvic abscess,faecal discharge from a drain and rectovaginal fistula.Radiologically demonstrated leakage with no clinicalsymptoms was not included as these leaks are not ofclinical significance. The guidelines in the QUOROMstatement31 were adhered to, from the literature search tothe presentation of the results and discussion.

Statistical analysis

Statistical analysis was using Review Manager (RevMan)software version 5.0 (The Nordic Cochrane Centre,The Cochrane Collaboration, Copenhagen, Denmark).Statistical heterogeneity was tested using χ2 and I2

tests. Data were pooled using a fixed-effect model ifheterogeneity was limited; a random-effect model wasused when there was significant heterogeneity among thetrials. The conventional 0·05 level of significance wasemployed. The RCTs were analysed separately from thenon-randomized studies.

Results

The search identified 26 publications comparing lowanterior resection with and without a defunction-ing stoma, four RCTs32–35 and 22 non-randomizedstudies1,3,6–10,17–26,36–40 . Six authors responded to thee-mail and were able to provide required information thatwas not specified clearly in the original article1,3,9,23,26,35.All but one study7 were included in the analysis. This studywas excluded as it described both high and low anteriorresections without providing specific data on the outcomes

Copyright 2009 British Journal of Surgery Society Ltd www.bjs.co.uk British Journal of Surgery 2009; 96: 462–472Published by John Wiley & Sons Ltd

464 W. S. Tan, C. L. Tang, L. Shi and K. W. Eu

Potentially relevant studiesidentified and screened for

retrievaln = 26

Studies excluded n = 0

Studies excluded n = 0

Studies withdrawn, by outcome n = 0

Studies excluded from meta-analysis n = 1 No breakdown of data by level of anastomosis n = 1

Studies retrieved for more detailedevaluation

n = 26

Potentially appropriate studies tobe included in meta-analysis

n = 26

Studies included in meta-analysisn = 25

Studies with usable data byoutcomen = 25

Fig. 1 QUORUM flow chart for study

of low anterior resections alone. The QUOROM flowchart is shown in Fig. 1.

The included studies were published between 1983 and2008, with a total study population of 11 429 patients.Sample size of studies varied from 36 to 2729 patients.There were four RCTs with a total of 358 patients, ofwhom 178 had a stoma and 180 did not. The other21 studies were non-randomized with a total of 11 071patients, of whom 4452 had a stoma and 6619 did not. Allstudies reported symptomatic anastomotic leak rates. Somealso included radiologically detected asymptomatic leaks,but these were not included in the analysis. A summary ofthe characteristics and validity of the included studies ispresented in Tables 1 and 2.

Meta-analysis of randomized trials

Four RCTs were included32–35. The study by Pimenteland colleagues35 has been published only in abstract form,

but it was included as the leak rate and sequelae of leakagewere available in the abstract.

The study populations were patients undergoing lowanterior resection for rectal cancer, in whom resection ofthe tumour and completion of a safe stapled anastomosishad been performed, as evidenced by complete anastomoticrings and a negative air leak test. Exclusion criteria variedbetween the studies and included patient refusal, surgeondecision to construct a stoma in view of an unreliableanastomosis and the presence of major intraoperativeadverse events. Definitions of low anterior resection in thestudies were fairly similar and are summarized in Table 1.

Randomization was performed by the sealed envelopetechnique in all four studies. Blinding of medical staff andpatients was not possible in all studies as it was obvi-ous whether or not a stoma was present. All patients inthe studies were analysed. Where there were violations ofprotocol, the data were analysed on an intention-to-treatbasis.



Table 1 Characteristics and validity of included randomized controlled trials

No. of patients

Reference Year StomaNo

stomaInclusioncriteria

Definition of lowanterior resection Definition of leak

Randomizationand allocation Blinding

Graffner et al.34 1983 25 25 Completion of safestapledanastomosis

Anastomosisbelow peritonealreflection

Defectiveanastomosis withsigns of abscessformation atproctoscopy

Sealed envelope Not possible

Pakkastie et al.33 1997 19 19 Negative air leak testand completeanastomotic rings

Anastomosisbelow peritonealreflection

Clinical andradiological signsof anastomoticleak


Pimentel et al.35 2003 18 18 Negative air leak testand completeanastomotic rings

Anastomosis≤ 4 cm fromanal verge



Matthiessen et al.32 2007 116 118 Negative air leaktest, intactanastomotic ringsand absence ofintraoperativeadverse events

Anastomosis≤ 7 cm fromanal verge



All patients were analysed in all studies.

Reference

Graffner et el.34

Pakkastie et el.33

Pimentel et el.35

Matthiessen et el.32

Total

Test for overall effect: Z = 3·56, P < 0·001Heterogeneity: χ2 = 0·25, 3 d.f., P = 0·970, I2 = 0%

Clinical leak rate

Stoma No stoma

1 of 253 of 191 of 18

12 of 116

3 of 25 6·913·74·6

74·8

6 of 192 of 18

33 of 118

17 of 178 44 of 180 100·0

Weight (%) Risk ratio Risk ratio

0·33 (0·04, 2·99)0·50 (0·15, 1·71)0·50 (0·05, 5·04)0·37 (0·20, 0·68)

0·39 (0·23, 0·66)

0·01 0·1 101 100Favours stoma Favours no stoma

Fig. 2 Meta-analysis of clinical leak rate in stoma versus non-stoma groups in randomized controlled trials. Risk ratios are shown with95 per cent confidence intervals

Reference

Graffner et el.34

Pakkastie et el.33

Pimentel et el.35


Total


Reoperation rate

Stoma No stoma

0 of 251 of 190 of 18

10 of 116

3 of 25 8·414·46·0

71·3

6 of 192 of 18

30 of 118

11 of 178 41 of 180 100·0


0·14 (0·01, 2·63)0·17 (0·02, 1·26)0·20 (0·01, 3·89)0·34 (0·17, 0·66)

0·29 (0·16, 0·53)


Fig. 3 Meta-analysis of reoperation rate in stoma versus non-stoma groups in randomized controlled trials. Risk ratios are shown with95 per cent confidence intervals



Table 2 Characteristics of included non-randomized studies

No. of patients



Definition of lowanterior resection Definition of leak Reason for stoma

Fielding et al.23 1984 214 407 All low anteriorresections

Anastomosis belowperitoneal reflection

Clinical, radiological,intraoperative andpost-mortem signsof leak

Surgeon decision

Karanjia et al.36 1991 125 75 All low anteriorresections

Anastomosis ≤ 12 cmfrom anal verge

Not mentioned Not mentioned

Kessler et al.18* 1993 85 272 All low anteriorresections


Clinical signs of leak Not mentioned

Karanjia et al.17 1994 157 62 All low anteriorresections with TME

Tumour ≤ 7 cm fromanal verge


Pakkastie et al.3 1994 10 51 All low anteriorresections


Clinical signs of leak Incomplete doughnuts

Grabham et al.6 1995 7 70 All low anteriorresections

Tumour belowperitoneal reflection

Clinical signs of leak Difficult dissection, tensionand incompletedoughnuts

Bokey et al.37* 1995 109 304 All low anteriorresections

Not mentioned Clinical signs of leak Surgeon decision

Rullier et al.8* 1998 86 45 All low anteriorresections


Clinical signs of leak Anastomotic tension, poorbowel preparation,incomplete doughnuts,leakage on testing,technical difficulties inperforming low rectalanastomosis, allcoloanal

anastomoses.Dehni et al.22 1998 152 106 All low anterior

resectionsInferior margin of

tumour 6–11 cmfrom anal verge

Clinical signs of leak Surgeon decision

Enker et al.38 1999 214 467 All low anteriorresections


Clinical signs of leakand leak seen oncontrast enema

Not mentioned

Poon et al.20 1999 61 87 All low anteriorresections withstapled colorectalanastomoses

Tumour ≤ 12 cm fromanal verge

Clinical signs of leak Presence of risk factors

Law et al.39 2000 103 93 All low anteriorresections with TME

TME performed Clinical signs of leak Poor bowel preparation,previous radiation,technical difficulties,positive air leak test,incomplete doughnuts,very low anastomosis

Marusch et al.25 2002 148 334 All low anteriorresections



Machado et al.24 2002 80 81 All low anteriorresections with TMEand colonic pouches

TME performed Clinical signs of leak Surgeon decision

Matthiessen et al.1 2004 28 61 All low anteriorresections



Peeters et al.10 2005 523 401 Patients in Dutch TMEtrial

TME performed Clinical signs of leak Surgeon decision

Shukla and Pandey40 2005 63 105 All low anteriorresections

Not mentioned Not mentioned Not mentioned

Wong and Eu26 2005 742 324 All low anteriorresections withoutprevious RT





Table 2 (Continued)

No. of patients



Definition of lowanterior resection

Definitionof leak Reason for stoma

Gastinger et al.21 2005 881 1848 All low anterior resections Anastomosis ≤ 8 cm fromanal verge


Eriksen et al.9 2005 622 1336 All low anterior resections Anastomosis ≤ 6 cm fromanal verge


Lefebure et al.19 2008 42 90 All low anterior resections Anastomosis ≤ 8 cm fromanal verge


*Only subset of low anterior resections included. TME, total mesorectal excision; RT, radiotherapy.

Reference

Graffner et el.34

Pakkastie et el.33

Pimentel et el.35


Total

Test for overall effect: Z = 0·67, P = 0·500Heterogeneity: χ2 = 0·91, 3 d.f., P = 0·820, I2 = 0%

Mortality rate

Stoma No stoma

0 of 250 of 190 of 180 of 116

0 of 25 18·245·518·218·1

2 of 190 of 180 of 118

0 of 178 0 of 180 100·0


1·00 (0·07, 15·12)0·20 (0·01, 3·91)

1·00 (0·07, 14·79)1·00 (0·06, 16·07)

0·64 (0·17, 2·38)


Fig. 4 Meta-analysis of mortality rate related to leakage in stoma versus non-stoma groups in randomized controlled trials. Risk ratiosare shown with 95 per cent confidence intervals

Clinical anastomotic leakAll four trials recorded a higher rate of anastomoticleakage in the non-stoma group, although this wasstatistically significant in only one trial32. Overall, among358 patients, the absence of a stoma was associated witha higher incidence of anastomotic leak (risk ratio (RR)0·39 (95 per cent confidence interval (c.i.) 0·23 to 0·66);P < 0·001). There was minimal heterogeneity among thetrials (3 d.f., P = 0·970, I2 = 0 per cent) (Fig. 2).

Reoperation for leakThere was a higher rate of reoperation in the non-stomagroup in all four studies, but this was statistically significantin only one trial32. Overall, the absence of a stoma wasassociated with a higher rate of reoperation (RR 0·29(95 per cent c.i. 0·16 to 0·53); P < 0·001). There wasminimal heterogeneity among the trials (3 d.f., P = 0·850,I2 = 0 per cent) (Fig. 3).

Mortality related to anastomotic leakDespite the increased incidence of anastomotic leaks andthe higher rate of reoperation, there was no statisticallysignificant difference in mortality rate between the twogroups in all four studies (RR 0·64 (95 per cent c.i. 0·17

to 2·38); P = 0·500). There was minimal heterogeneityamong the trials (3 d.f., P = 0·820, I2 = 0 per cent) (Fig. 4).

Meta-analysis of non-randomized studies

Both prospective and retrospective studies evaluatingthe use of a defunctioning stoma in anterior resectionswere included. Where both high and low anteriorresections were included in the study, only the low wereincluded in the analysis. Reasons for construction of astoma varied between the studies, and included surgeondecision, defective doughnuts, positive anastomotic leaktest, technical difficulties, poor bowel preparation andprevious irradiation (Table 2). Radiologically demonstratedleaks without clinical symptoms were not included.

Clinical anastomotic leakAll 21 studies reported on anastomotic leak rate. Fif-teen studies reported a lower rate of leakage witha stoma1,3,6,8–10,17–20,22,24,36,38,39 , four reported sim-ilar rates (less than 1 per cent difference) in bothgroups21,25,26,40 and two reported a higher leakage ratein the stoma group23,37. Overall, the meta-analysis showeda significantly lower risk of anastomotic leak in patientswith a stoma (RR 0·74 (95 per cent c.i. 0·67 to 0·83);



Reference

Fielding et el.23

Karanjia et el.36

Kessler et el.18

Karanjia et el.17

Pakkastie et el.3

Bokey et el.37

Grabham et el.6

Ruller et el.8

Dehni et el.22

Poon et el.20

Enker et el.38

Law et el.39

Machado et el.24

Marusch et el.25

Matthiessen et el.1

Peeters et el.10

Eriksen et el.9

Shukla and Pandey et el.40

Wong and Eu et el.26

Gastinger et el.21

Lefebure et el.19

Total

Test for overall effect: Z = 5·24, P < 0·001Heterogeneity: χ2 = 79·33, 20 d.f., P < 0·001, I2 = 75%

Clinical leak rate

Stoma No stoma

50 of 2141 of 125

10 of 8513 of 1571 of 10

21 of 1090 of 7

16 of 868 of 1522 of 616 of 2145 of 1037 of 80

16 of 1484 of 28

43 of 52344 of 5776 of 63

28 of 742128 of 881

3 of 42

65 of 40711 of 7552 of 27211 of 6215 of 5145 of 3042 of 709 of 45


103 of 4679 of 105

13 of 324262 of 184810 of 90

7·12·23·92·50·83·80·11·93·41·42·32·51·63·41·7

11·518·11·1

2·926·91·0

412 of 4407 800 of 5750 100·0


1·46 (1·05, 2·03)0·05 (0·01, 0·41)0·62 (0·33, 1·16)0·47 (0·22, 0·99)0·34 (0·05, 2·29)1·30 (0·81, 2·08)

1·77 (0·09, 33·81)0·93 (0·45, 1·94)0·31 (0·14, 0·69)0·26 (0·06, 1·13)0·57 (0·24, 1·38)0·30 (0·11, 0·80)0·71 (0·28, 1·77)1·03 (0·59, 1·80)0·51 (0·19, 1·38)0·52 (0·36, 0·74)0·35 (0·25, 0·48)1·11 (0·42, 2·97)

0·94 (0·49, 1·79)1·02 (0·84, 1·25)0·64 (0·19, 2·22)

0·74 (0·67, 0·83)

0·10·01 101 100Favours stoma Favours no stoma

Fig. 5 Meta-analysis of clinical leak rate in stoma versus non-stoma groups in non-randomized studies. Risk ratios are shown with95 per cent confidence intervals

Reference

Karanjia et el.36

Pakkastie et el.3

Grabham et el.6

Dehni et el.22

Poon et el.20

Law et el.39

Marusch et el.25

Machado et el.24

Peeters et el.10

Matthiessen et el.1

Shukla and Pandey et el.40

Wong and Eu et el.26

Eriksen et el.9

Gastinger et el.21

Lefebure et el.19

Total


Reoperation rate

Stoma No stoma


26 of 5233 of 283 of 637 of 742

14 of 57732 of 881

0 of 42

8 of 7511 of 512 of 70

17 of 1067 of 87

10 of 9323 of 33410 of 8160 of 40115 of 619 of 63

32 of 32448 of 467

186 of 18485 of 90

2·61·00·15·21·62·93·72·6

17·72·52·3

11·613·831·30·9

101 of 3542 443 of 4151 100·0


0·07 (0·01, 0·59)0·21 (0·01, 3·23)

1·77 (0·09, 33·81)0·25 (0·10, 0·60)0·09 (0·01, 1·63)0·04 (0·00, 0·72)0·29 (0·09, 0·97)0·61 (0·23, 1·59)0·33 (0·21, 0·52)0·44 (0·14, 1·38)0·33 (0·09, 1·17)0·10 (0·04, 0·21)0·24 (0·13, 0·42)0·36 (0·25, 0·52)0·19 (0·01, 3·40)

0·28 (0·23, 0·35)


Fig. 6 Meta-analysis of reoperation rate in stoma versus non-stoma groups in non-randomized studies. Risk ratios are shown with95 per cent confidence intervals



Reference

Fielding et el.23

Karanjia et el.36

Kessler et el.18

Karanjia et el.17

Pakkastie et el.3

Grabham et el.6

Dehni et el.22

Ruller et el.8

Poon et el.20

Law et el.39

Machado et el.24

Marusch et el.25

Matthiessen et el.1

Peeters et el.10

Gastinger et el.21

Total


Mortality rate

Stoma No stoma

4 of 2140 of 1251 of 850 of 1570 of 100 of 70 of 1521 of 860 of 610 of 1032 of 801 of 1481 of 286 of 5238 of 881

10 of 4073 of 75

10 of 2723 of 621 of 510 of 702 of 1060 of 450 of 870 of 933 of 813 of 3342 of 619 of 401

37 of 1848

8·25·25·65·90·6

3·50·8

3·52·21·5

12·028·3

28 of 4084 100 of 4979 100·0


0·76 (0·24, 2·40)0·09 (0·00, 1·65)0·32 (0·04, 2·46)0·06 (0·00, 1·09)

1·58 (0·07, 36·18)Not estmable

0·14 (0·01, 2·88)1·59 (0·07, 38·17)

Not estmableNot estmable

0·68 (0·12, 3·93)0·75 (0·08, 7·17)

1·09 (0·10, 11·52)0·51 (0·18, 1·42)0·45 (0·21, 0·97)

Wong and Eu et el.26 1 of 742 2 of 324 3·3 0·22 (0·02, 2·40)Shukla and Pandey et el.40 0 of 63 1 of 105 1·3 0·55 (0·02, 13·35)Eriksen et el.9 3 of 577 13 of 467 17·0 0·19 (0·05, 0·65)Lefebure et el.19 0 of 42 1 of 90 1·1 0·71 (0·03, 16·96)

0·42 (0·28, 0·61)


Fig. 7 Meta-analysis of mortality rate related to leakage in stoma versus non-stoma groups in non-randomized studies. Risk ratios areshown with 95 per cent confidence intervals

Table 3 Risk factors for anastomotic leak

Reference

Lowanasto-mosis

Typeof

anasto-mosis

Malesex

HighBMI

Preop.RT

Longoperatingtime

Increasedblood loss

and intraop.adverse events Age

Tumourstage

In-completedough-

nutsFaecalloading Surgeon

Use ofsigmoid

foranastomosis

Highversus

lowtie

Useof

drain

23 Yes Yes Yes18 Yes17 No No No No No Yes No No3 Yes No No No No No8 Yes No No No No No No No No38 No No No No No20 No Yes No No39 Yes Yes Yes25 Yes Yes10 No No No Yes No No Yes1 Yes Yes Yes Yes21 No9 Yes Yes Yes19 No No No No

Yes, statistically significant; no, not statistically significant. BMI, body mass index; preop., preoperative; RT, radiotherapy; intraop., intraoperative.

P < 0·001). For this outcome measure, the studies wererelatively heterogeneous (20 d.f., P < 0·001, I2 = 75 percent) (Fig. 5).

Reoperation for leakFifteen studies reported on the rate of reoperation afteranastomotic leakage, and this was higher in the group

without a stoma in all studies1,3,6,9,10,19–22,24–26,36,39,40 .After pooling all available data for the 7693 patients,the rate of reoperation was significantly lower in thestoma group (RR 0·28 (95 per cent c.i. 0·23 to 0·35);P < 0·001). For this outcome measure, the studies wererelatively homogeneous (14 d.f., P = 0·200, I2 = 23 percent) (Fig. 6).



Mortality related to anastomotic leak

Nineteen studies reported on this outcome. Mortalityrates were higher in the group without a stoma in113,9,10,17–19,21,22,24,35,40 , similar (less than 1 per centdifference) in seven1,6,20,23,25,26,39 and lower in one8.After pooling all available data, there was a statisticallysignificantly higher mortality rate in the group without astoma (RR 0·42 (95 per cent c.i. 0·28 to 0·61); P < 0·001).For this outcome measure, the studies were relativelyhomogeneous (15 d.f., P = 0·860, I2 = 0 per cent) (Fig. 7).

Other risk factors for anastomotic leakThe influence of other factors on anastomotic leakage wasexamined in 14 of the 21 studies (Table 3). The two mostfrequent factors that significantly affected leakage rate werelow anastomosis and male sex. Other less common factorsincluded surgeon factors, faecal loading, long operatingtime, increased blood loss and body mass index.

Morbidity and mortality after routine closure of adefunctioning stoma (randomized andnon-randomized studies)

Morbidity and/or mortality data related to stoma closurewere available in ten studies3,6,20,21,24,32–36. Routine stomaclosure was associated with a morbidity rate ranging from0 to 19·8 per cent, and a mortality rate ranging from0 per cent to 1·4 per cent (Tables 4 and 5).

Table 4 Morbidity and mortality of routine stoma closure inrandomized controlled trials

ReferenceMorbidity of

stoma closureMortality of

stoma closure

Graffner et al.34 0 of 23 (0) 0 of 23 (0)Pakkastie et al.33 1 of 17 (6) 0 of 17 (0)Pimentel et al.35 3 of 17 (18) 0 of 17 (0)Matthiessen et al.32 5 of 100 (5·0) 1 of 100 (1·0)

Values in parentheses are percentages.

Table 5 Morbidity and mortality of routine stoma closure innon-randomized studies

ReferenceMorbidity of

stoma closureMortality of

stoma closure

Karanjia et al.36 17 of 120 (14·2) 0 of 120 (0)Pakkastie et al.3 0 of 9 (0) 0 of 9 (0)Grabham et al.6 Not available 0 of 7 (0)Poon et al.20 4 of 46 (8·7) 0 of 46 (0)Machado et al.24 Not available 1 of 69 (1·4)Gastinger et al.21 126 of 636 (19·8) 2 of 636 (0·3)

Values in parentheses are percentages.

Discussion

The literature so far has appeared somewhat divided onwhether the use of a defunctioning stoma should be routineafter low anterior resection. The present meta-analysis ofboth randomized and non-randomized studies has shownthat a stoma significantly decreases both leak rate andreoperation rate. In addition, meta-analysis of the non-randomized studies showed a decrease in mortality raterelated to leakage, although this was not a significantfeature in the analysis of the RCTs. In the light of thesefindings, the authors recommend a stoma after low anteriorresection.

These results are similar to those published in an earliermeta-analysis by Huser and co-workers41. However, theselection of the study population was different in thetwo studies. The earlier analysis included all patients withanterior resections (both high and low), whereas the presentstudy included only patients with low anastomoses. Theauthors felt it was important to exclude high anteriorresections from the analysis as most surgeons routinelyavoid a stoma when the anastomosis is above the peritonealreflection. Hence, although the definition of low anteriorresection varied between the studies included in the presentanalysis, this variation was much less than that in the meta-analysis by Huser and colleagues. The present conclusionthat a stoma should be recommended seems more valid.

There was minimal heterogeneity between the RCTsfor all the outcomes measured, ensuring the quality ofthe meta-analysis. However, this analysis still has certainlimitations. First, there was only a small number of RCTsavailable for inclusion. This problem is common in surgeryand it is generally agreed that it is acceptable to performa meta-analysis with only a few studies, as long as thedegree of heterogeneity is minimal42, as was the case here.Second, one of the trials32 contributed a relatively largeproportion of the patients in the analysis. This may haveintroduced some bias as this trial would have contributeda significant amount of weight to the analysis. Third,the included studies spanned a relatively long periodof 25 years, from 1983 to 2008. The oldest study, byGraffner and colleagues34, was performed when circularstapling devices were first being introduced. Over the 25-year period, with ever improving technology and surgeonexpertise, it would not be possible to meaningfully comparethe studies. However, this is a limitation that has to beaccepted in reviews.

Pooling of the non-randomized studies is affected byconsiderable bias in view of differences in methodologybetween studies and the inclusion of patients likely to be athigher risk of anastomotic leak in the stoma group. The aimof meta-analysis of the non-randomized studies was just to



provide an overview of the general trend of outcomes inthese studies and it yielded similar results to the analysisof the RCTs. Despite the fact that the patients selected tohave a stoma in the non-randomized studies were probablythe ‘riskier’ patients, this group still fared better than thegroup with no stoma in terms of the primary outcomesmeasured.

A review of the factors associated with a significantlyhigher risk of anastomotic leak showed that the two mostcommon factors were anastomotic height, followed by malesex1,8–10,17–21,23,25,38,39 . Although it is hardly possible todecide on creating a stoma based on sex, a low anastomosis,based on all available evidence, should be protected with adefunctioning stoma.

The benefits of a stoma in decreasing the rateof anastomotic leak and in mitigating the potentialcomplications of a leak must, however, be balanced againstthe morbidity of the stoma itself, as well as the morbidityand mortality of its closure. A recent meta-analysis reportedthat up to a quarter of patients had complications associatedwith the presence of a stoma29. Morbidity rates of up to36·5 per cent and mortality rates of up to 1·4 per cent havealso been reported in patients who undergo elective stomaclosure21,24,28. The benefits conferred by a stoma may thusbe offset by the complications associated with the stomaand its subsequent closure. The balance of benefits andcomplications associated with a stoma is, however, outwiththe scope of this article.

Acknowledgements

The authors declare no conflict of interest.

References

1 Matthiessen P, Hallbook O, Andersson M, Rutegard J,Sjodahl R. Risk factors for anastomotic leakage after anteriorresection of the rectum. Colorectal Dis 2004; 6: 462–469.

2 Lipska MA, Bissett IP, Parry BR, Merrie AE. Anastomoticleakage after lower gastrointestinal anastomosis: men are at ahigher risk. ANZ J Surg 2006; 76: 579–585.

3 Pakkastie TE, Luukkonen PE, Jarvinen HJ. Anastomoticleakage after anterior resection of the rectum. Eur J Surg1994; 160: 293–297.

4 Nesbakken A, Nygaard K, Lunde OC, Blucher J,Gjertsen Ø, Dullerud R. Anastomotic leak followingmesorectal excision for rectal cancer: true incidence anddiagnostic challenges. Colorectal Dis 2005; 7: 576–581.

5 Eckmann C, Kujath P, Schiedeck TH, Shekarriz H,Bruch HP. Anastomotic leakage following low anteriorresection: results of a standardized diagnostic and therapeuticapproach. Int J Colorectal Dis 2004; 19: 128–133.

6 Grabham JA, Moran BJ, Lane RH. Defunctioning colostomyfor low anterior resection: a selective approach. Br J Surg1995; 82: 1331–1332.

7 Law WL, Chu KW. Anterior resection for rectal cancer withmesorectal excision: a prospective evaluation of 622 patients.Ann Surg 2004; 240: 260–268.

8 Rullier E, Laurent C, Garrelon JL, Michel P, Saric J,Parneix M. Risk factors for anastomotic leakage afterresection of rectal cancer. Br J Surg 1998; 85: 355–358.

9 Eriksen MT, Wibe A, Norstein J, Haffner J, Wiig JN,Norwegian Rectal Cancer Group. Anastomotic leakagefollowing routine mesorectal excision for rectal cancer in anational cohort of patients. Colorectal Dis 2005; 7: 51–57.

10 Peeters KC, Tollenaar RA, Marijnen CA, KleinKranenbarg E, Steup WH, Wiggers T et al. Risk factors foranastomotic failure after total mesorectal excision of rectalcancer. Br J Surg 2005; 92: 211–216.

11 Vignali A, Fazio VW, Lavery IC, Milsom JW, Church JM,Hull TL et al. Factors associated with the occurrence of leaksin stapled rectal anastomoses: a review of 1014 patients. J AmColl Surg 1997; 185: 105–113.

12 Nesbakken A, Nygaard K, Lunde OC. Outcome and latefunctional results after anastomotic leakage followingmesorectal excision for rectal cancer. Br J Surg 2001; 88:400–404.

13 Hallbook O, Sjodahl R. Anastomotic leakage and functionaloutcome after anterior resection of the rectum. Br J Surg1996; 83: 60–62.

14 McArdle CS, McMillan DC, Hole DJ. Impact of anastomoticleakage on long-term survival of patients undergoing curativeresection for colorectal cancer. Br J Surg 2005; 92:1150–1154.

15 Law WL, Choi HK, Lee YM, Ho JW, Seto CL.Anastomotic leakage is associated with poor long-termoutcome in patients after curative colorectal resection formalignancy. J Gastrointest Surg 2007; 11: 8–15.

16 Chang SC, Lin JK, Yang SH, Jiang JK, Chen WC, Lin TC.Long-term outcome of anastomosis leakage after curativeresection for mid and low rectal cancer. Hepatogastroenterology2003; 50: 1898–1902.

17 Karanjia ND, Corder AP, Bearn P, Heald RJ. Leakage fromstapled low anastomosis after total mesorectal excision forcarcinoma of the rectum. Br J Surg 1994; 81: 1224–1226.

18 Kessler H, Hermanek P JR, Wiebelt H. Operative mortalityin carcinoma of the rectum. Results of the GermanMulticentre Study. Int J Colorectal Dis 1993; 8: 158–166.

19 Lefebure B, Tuech JJ, Bridoux V, Costaglioli B, Scotte M,Teniere P et al. Evaluation of selective defunctioning stomaafter low anterior resection for rectal cancer. Int J ColorectalDis 2008; 23: 283–288.

20 Poon RT, Chu KW, Ho JW, Chan CW, Law WL, Wong J.Prospective evaluation of selective defunctioning stoma forlow anterior resection with total mesorectal excision. World JSurg 1999; 23: 463–467.

21 Gastinger I, Marusch F, Steinert R, Wolff S, Koeckerling F,Lippert H. Protective defunctioning stoma in low anterior



resection for rectal carcinoma. Br J Surg 2005; 92:1137–1142.

22 Dehni N, Schlegel RD, Cunningham C, Guiguet M,Tiret E, Parc R. Influence of a defunctioning stoma onleakage rates after low colorectal anastomosis and colonic Jpouch–anal anastomosis. Br J Surg 1998; 85: 1114–1117.

23 Fielding LP, Stewart-Brown S, Hittinger R, Blesovsky L.Covering stoma for elective anterior resection of the rectum:an outmoded operation? Am J Surg 1984; 147: 524–530.

24 Machado M, Hallbook O, Goldman S, Nystrom PO,Jarhult J, Sjodahl R. Defunctioning stoma in low anteriorresection with colonic pouch for rectal cancer: a comparisonbetween two hospitals with a different policy. Dis ColonRectum 2002; 45: 940–945.

25 Marusch F, Koch A, Schmidt U, Geibetaler S, Dralle H,Saeger HD et al. Value of a protective stoma in low anteriorresections for rectal cancer. Dis Colon Rectum 2002; 45:1164–1171.

26 Wong NY, Eu KW. A defunctioning ileostomy does notprevent clinical anastomotic leak after a low anteriorresection: a prospective, comparative study. Dis Colon Rectum2005; 48: 2076–2079.

27 Bakx R, Busch OR, Bemelman WA, Veldink GJ, Slors JF,van Lanschot JJ. Morbidity of temporary loop ileostomies.Dig Surg 2004; 21: 277–281.

28 Kaiser AM, Israelit S, Klaristenfeld D, Selvindoss P,Vukasin P, Ault G et al. Morbidity of ostomy takedown. JGastrointest Surg 2008; 12: 437–441.

29 Duchesne JC, Wang YZ, Weintraub SL, Boyle M, Hunt JP.Stoma complications: a multivariate analysis. Am Surg 2002;68: 961–966.

30 Higgins JPT, Green S (editors). Cochrane Handbook forSystematic Reviews of Interventions Version 5.0.1 [updatedSeptember 2008]. The Cochrane Collaboration, 2008.

31 Moher D, Cook DJ, Eastwood S, Olkin I, Rennie D,Stroup DF. Improving the quality of reports of meta-analysesof randomised controlled trials: the QUOROM statement.Quality of Reporting of Meta-analyses. Lancet 1999; 354:1896–1900.

32 Matthiessen P, Hallbook O, Rutegard J, Simert G,

Sjodahl R. Defunctioning stoma reduces symptomaticanastomotic leakage after low anterior resection of therectum for cancer: a randomized multicenter trial. Ann Surg2007; 246: 207–214.

33 Pakkastie TE, Ovaska JT, Pekkala ES, Luukkonen PE,Jarvinen HJ. A randomised study of colostomies in lowcolorectal anastomoses. Eur J Surg 1997; 163: 929–933.

34 Graffner H, Fredlund P, Olsson SA, Oscarson J,Petersson BG. Protective colostomy in low anterior resectionof the rectum using the EEA stapling instrument. Arandomized study. Dis Colon Rectum 1983; 26: 87–90.

35 Pimentel JM, Duarte A, Patricio J. The role of a protectingstoma in low anterior resection with TME and colonicJ-pouch for rectal cancer; results of a prospective randomizedtrial. Colorectal Dis 2003; 5(Suppl 2): 83 (Abstract).

36 Karanjia ND, Corder AP, Holdsworth PJ, Heald RJ.Risk of peritonitis and fatal septicaemia and the need todefunction the low anastomosis. Br J Surg 1991; 78:196–198.

37 Bokey EL, Chapuis PH, Fung C, Hughes WJ, Koorey SG,Brewer D et al. Postoperative morbidity and mortalityfollowing resection of the colon and rectum for cancer. DisColon Rectum 1995; 38: 480–486.

38 Enker WE, Merchant N, Cohen AM, Lanouette NM,Swallow C, Guillem J et al. Safety and efficacy of low anteriorresection for rectal cancer: 681 consecutive cases from aspecialty service. Ann Surg 1999; 230: 544–552.

39 Law WI, Chu KW, Ho JW, Chan CW. Risk factors foranastomotic leakage after low anterior resection with totalmesorectal excision. Am J Surg 2000; 179: 92–96.

40 Shukla PJ, Pandey D. Protective defunctioning stoma in lowanterior resection for rectal carcinoma (Br J Surg 2005; 92:1137–1142). Br J Surg 2005; 92: 1565–1566.

41 Huser N, Michalski CW, Erkan M, Schuster T,Rosenberg R, Kleeff J et al. Systematic review andmeta-analysis of the role of defunctioning stoma in low rectalcancer surgery. Ann Surg 2008; 248: 52–60.

42 Ng TT, McGory ML, Ko CY, Maggard MA. Meta-analysisin surgery: methods and limitations. Arch Surg 2006; 141:1125–1130.


meta-analysis of defunctioning stomas in low anterior resection for rectal cancer

Documents