Clinical Review Stapled versus handsewn intestinal anastomosis in emergency laparotomy: A systemic review and meta-analysis

David N. Naumann, MRCS,a,b Aneel Bhangu, PhD,a Michael Kelly, MBChB,a and Douglas M. Bowley, FRCS,a,b Birmingham, UK

Background. The optimal technique for gastrointestinal anastomosis remains controversial in emergency laparotomy. The aim of this meta-analysis was to compare outcomes of stapled versus handsewn anastomosis after emergency bowel resection. Methods. A systematic review was performed for studies comparing outcomes after emergency laparotomy using stapled versus handsewn anastomosis until July 2014 (PROSPERO registry number: CRD42013006183). The primary endpoint was anastomotic failure, a composite measure of leak, abscess and fistula. Odds ratio (OR; with 95% CI) and weighted mean differences were calculated using meta-analytical techniques. Subgroup analysis was conducted for trauma surgery (TS) and emergency general surgery (EGS) cohorts. Risk of bias for each study was calculated using the Newcastle– Ottawa scale for cohort studies, and Cochrane Collaboration’s tool for randomized trials. Results. The final analysis included 7 studies of 1,120 patients, with a total of 1,205 anastomoses. There were 5 TS studies and 2 EGS studies. There were no differences in anastomotic failure between handsewn and stapled techniques on an individual anastomosis level (OR, 1.53; 95% CI, 0.97–2.43; P = .070), or on an individual patient level (OR, 1.44; 95% CI, 0.92–2.25; P = .110). There were no differences in the individual rates of anastomotic leak, abscess, fistulae, or postoperative deaths between techniques. Subgroup analysis of EGS and TS studies demonstrated no superior operative technique. Conclusion. Available evidence is sparse and at high risk of bias, and neither stapling nor handsewing is justifiably favored in emergency laparotomy. Surgeons might therefore select the technique of their own choice with caution owing to unresolved uncertainty. (Surgery 2015;157:609-18.)

From the Department of General Surgery,a Heart of England NHS Foundation Trust, Bordesley Green East; and Royal Centre for Defence Medicine,b Birmingham, UK

THE RECENT COCHRANE REVIEW that compared elective none of the individual studies favored handsewn stapled and handsewn colorectal anastomosis or stapled techniques, meta-analysis favored the sta- showed no differences in outcome between tech- pled technique in the cancer resection subgroup niques, but commented that during higher risk but not the Crohn’s disease resection subgroup.2 emergency laparotomy, this question remains unan- Such a finding suggests that there may be unique cir- swered.1 Another recent Cochrane review analyzing cumstances in which even small differences in out- elective ileocolic anastomosis found that, although comes between techniques have the potential to make a difference in specific populations. Choice Conflict of interests: None declared. of operative technique also depends on differences Presented at the Tripartite Colorectal Meeting 2014, Birming- in opinion, interpretation of evidence, as well as ham, UK both personal and institutional practice, making Accepted for publication September 26, 2014. this a topic of debate and controversy. There have Reprint requests: Aneel Bhangu, PhD, Heart of England NHS been no randomized clinical trials (RCTs) in the Foundation Trust, Bordesley Green East, Birmingham B9 5SS, last 10 years regarding elective handsewn versus sta- UK. E-mail: [email protected]. pled colorectal anastomoses, leading the Cochrane 0039-6060/$ - see front matter reviewers to stipulate that the research question has Ó 2015 Elsevier Inc. All rights reserved. lost its strength.1 However, emergency surgery dif- http://dx.doi.org/10.1016/j.surg.2014.09.030 fers from elective surgery in terms of pathology,

SURGERY 609 610 Naumann et al Surgery April 2015 perioperative physiology, and general management well as whether there were any differences among principles, which may mean that technical success in trauma and EGS study subgroups. the elective setting may not be transferable to emer- gency patients, and that these situations must be METHODS considered separately. There is particular concern Data sources and search strategy. This review was that edematous bowel is less suited to stapled anasto- performed according to a prespecified protocol, mosis than handsewn in an emergency situation.3-6 which was registered with the International Bowel becomes edematous after splanchnic hypo- Prospective Register of Systematic Reviews (PROS- perfusion and subsequent reperfusion; in addition, PERO, ID number: CRD42013006183). A systematic the inflammatory host response may be initiated in search of the OVID SP and PubMed versions of response to sepsis or trauma. We hypothesized that Medline, the Cochrane Database of Systematic Re- there would be a favorable trend toward handsewn views, Current Controlled Trials, and ClinicalTrials. anastomosis in an emergency setting, because this gov was performed for published studies comparing technique might be more suitable than stapling stapled and handsewn anastomoses for emergency for technical ‘‘finetuning’’by the operating surgeon bowel surgery. Only studies published after 1990 in to compensate for the more hostile conditions of English were included. The search was performed bowel. independently by 2 authors (M.K., A.B.). Medical We also hypothesized that there may be addi- subject heading terms were used to search Medline, tional differences between trauma surgery (TS) combining domains of the operation, anastomosis and emergency general surgery (EGS) patients, type and randomization, using the AND function because the physiologic conditions are typically (Supplementary Table I). A manual search of refer- dominated by hemodynamic instability from hem- ence lists in relevant review articles was undertaken orrhage in the former, whereas the latter is pre- to further identify studies of potential interest. Ab- dominantly complicated by sepsis. The 2 groups stracts and conference proceedings were excluded may also have different patient demographics, with because of the high probability of incomplete TS patients being younger and more likely to be data. Citations were collated with EndNote Refer- male than EGS patients. There have been no ence Manager (Version X4, Thomson Reuters), systematic reviews comparing handsewn and sta- and duplicates removed. The last search was per- pled anastomosis in the emergency laparotomy formed on July 25, 2014. setting; neither have there been any analyses of the Inclusion and exclusion criteria. RCTs, as well as subgroups of TS and EGS. We therefore sought to both prospective and retrospective cohort studies address this gap in the literature with a summation comparing $1 outcome of stapled versus hand- of the best available evidence. sewn anastomosis after emergency bowel resection RCTs in the emergency surgical context remain were included. All gastrointestinal anastomoses, challenging. Because there is no preoperative for any emergency indication, were eligible for certainty that anastomosis will be required, gaining inclusion. Studies with only elective patients or informed consent is difficult. Furthermore, insti- unclear anastomosis technique were excluded. tutional review boards may prohibit RCTs that seek Where there was overlap in the same patients to predetermine operative techniques among sur- between studies, the older study was excluded. geons, because it may not be ethical or practical to Patients who underwent concurrent handsewn and ‘‘force’’ surgeons to use techniques they are not stapled anastomoses during multiple anastomoses accustomed to, or do not wish to use. Therefore, were excluded from further analysis, because it when conducting this systematic review and meta- would be impossible to identify which outcome was analysis, we were aware that RCT data might be attributable to which anastomosis. sparse, and that a pragmatic approach of utilizing Data extraction. Two authors extracted data all other retrospective and cohort study data would independently (D.N.N., M.K.). Discrepancies in likely have to suffice. Summation of these data, outcome extraction were resolved by reexamination although less than scientifically ideal, will provide of the relevant study until consensus was achieved. the largest and most generalizable evidence to date Data extracted on study design included: Indication on this topic, which may help to assess the need for (TS or EGS), randomization technique, interven- more specifically targeted trials if potential tion arms, mechanism of anastomosis (using a avenues of investigation were to be highlighted. stapling device or handsewn). Details relating to We aimed to determine whether there was a the included patients were number, age, gender, difference in outcome between stapled and hand- mechanism of injury, injury severity score, and sewn anastomosis after emergency laparotomy, as anatomic level of bowel resection and anastomosis. Surgery Naumann et al 611 Volume 157, Number 4

Details relating to the operation were policy for blood product requirement, surgical site infec- anastomosis technique and type of stapler or sutures tions, and duration of operating time. used. Subgroup analysis. As well as meta-analysis for all Definitions. ‘‘Stapled’’ was defined as an anasto- emergency studies, the subgroups of trauma and mosis fashioned using a stapling device of any kind, EGS were also separately analyzed using the same with or without sutured closure of enterotomy or primary outcome (anastomotic failure). Secondary oversewing of staple lines. ‘‘Handsewn’’ was defined outcomes listed were also meta-analyzed where $2 as an anastomosis that was fashioned by the surgeon studies reported each outcome. using any suture material, and without the utiliza- Assessment of bias. The Newcastle–Ottawa score tion of a stapling device. Anastomotic leak, abscess, was applied to assess the quality of nonrandomized and fistula were defined as present if a statement of and cohort studies based on the risk of bias.7 Prede- definitive diagnosis was made in the text and/or fined criteria were used to determine this score, table of the study, including $1 of (a) radiologic or including assessment of the selection and compara- (b) intraoperative (celiotomy/laparotomy) diag- bility of the study groups, and the methods of nosis, regardless of whether subsequent percuta- measuring outcomes. A star system was used to neous or operative intervention was required. judge quality in 8 domains, with a maximum of 9 ‘‘Anastomotic failure’’ was defined as any anasto- stars. The authors defined studies with $7 stars as motic leak, abscess or fistula as defined above. ‘‘TS relatively high quality, and those with #6 stars as studies’’ were defined as studies that reported pa- relatively low quality. Additionally, risk of bias for tient outcomes after bowel resection as a result of RCTs was assessed using domains provided by the blunt or penetrating trauma to the abdomen. ‘‘EGS Cochrane Collaboration’s tool for assessing risk of studies’’ were defined as studies that reported bias in randomized trials,8 including assessment of patient outcomes after emergency laparotomy in sequence generation, allocation concealment, the absence of any traumatic injury to the abdomen. blinding, completeness of data, and method of Outcome measures. Outcomes in emergency outcome reporting. Two authors (D.N.N., A.B.) surgery studies differ slightly from elective surgery independently calculated the quality scores using in that they are typically presented with the de- these systems, verified by a third author in case of nominator being either the total number of discrepancy until consensus achieved. Funnel plots patients or the total number of anastomoses (or were also used to assess for publication bias, being both may be simultaneously reported), because judged by 2 authors independently (M.K., A.B.). some patients in this situation may receive multiple Statistical analysis. A meta-analysis was conduct- anastomoses. This is especially true for retrospec- ed according to guidelines from the Preferred tive studies in which the methodology does not Reporting Items for Systematic reviews and Meta- exclude multiple anastomoses. Data extraction was Analysis group (PRISMA).9 The odds ratio (OR) therefore performed at both the patient and indi- was used as the statistical measure for dichotomous vidual anastomosis levels. The authors agreed that, outcomes, and the weighted mean difference was where possible, final analysis would be presented used as the statistical measure to compare contin- on both of these levels because this most accurately uous data. ORs were calculated from the original reflects the available literature. Furthermore, pre- data and meta-analyzed using the Mantel–Haens- sentation of these data might further serve to zel method. The OR represents the odds of an illustrate the differences between elective and adverse event (such as anastomotic leak) occurring emergency surgery in terms of patterns of surgical in the intervention (stapled) versus control group practice. However, where a choice between these 2 (handsewn). An OR of >1.00 indicated a greater reporting methods was necessary, patient-level data risk of an adverse event occurring in the interven- are reported in preference to anastomotic-level tion group, but when the 95% CI crosses 1.00, then data, because they are more clinically relevant to the overall effect is not significant. Significance was patient-centered surgical practice. judged using the Chi-square test. The I2 method Because some studies did not contain all 3 of was used to quantify heterogeneity (between- (a) anastomotic leak, (b) abscess, and (c) fistula, studies variance). The authors decided that meta- the primary outcome assessed for meta-analysis was analysis would only performed for outcomes that a composite measure of ‘‘anastomotic failure’’ (as $3 studies reported to maximize reliability of defined). Secondary outcomes recorded were in- data. However, for subgroup analysis of TS and dividual rate of anastomotic leaks, abscesses, and EGS groups, this threshold was reduced to $2 fistulae, duration of stay in hospital, duration of studies because there were fewer studies available stay in intensive care, postoperative mortality, in each subgroup. 612 Naumann et al Surgery April 2015

Metaregression was used to test the relationship between year of study and OR for each outcome. Statistical analysis was performed using Review Manager 5.1 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2011). Heterogeneity and mixed effects modelling. The authors were aware that the studies included in this meta-analysis were likely to be at relatively high risk of bias owing to heterogeneity (owing to variations in study methodology and populations between studies). Random effects modelling was used to mitigate this risk, as described by DerSi- monian and Laird.10 Regardless of the calculation of heterogeneity (I2) for each individual compari- son, the fixed-effect model could not be used because it requires the assumption that the effect size for all the studies is identical, with the varia- tion in effect size between studies being owing to Figure 1. Search results for systematic review. sampling error only (ie, all studies are treated as if they essentially contain patients from the same population). By using the random effects model, the aim is not to estimate 1 true summary effect, South Africa, Colombia, and Italy between 2000 and but instead to measure the mean of a distribution 2013, with study periods ranging from 20 to of effects. This approach allows each study to 130 months. There were a total of 1,120 patients contribute its own effect size to the overall mean with 1,205 anastomoses included. None of the studies effect, without any 1 of the studies having over- included data from minimally invasive or laparo- whelming (or underwhelming) influence. scopic surgery. Study characteristics, as well as the OR This meta-analysis asks the question: ‘‘Is there a for the primary outcome (anastomotic failure) are generalizable difference between handsewn and summarized in Table I (patients with multiple com- stapled anastomosis in emergency laparotomy?’’ bined handsewn and stapled anastomoses are not Therefore, the null hypothesis is that the mean illustrated in this table because they were excluded outcome effect (for all studies) of handsewn versus from further analysis). Five studies included the stapled anastomosis techniques is equal. Such a gender of patients, with 579 of 867 being male null hypothesis is best tested by using the random (66.8%). Six studies included patient age with a effects model, because the fixed effects model mean (±SD) of 40.9 years (±17.3) for all studies. would only test the null hypothesis that the Study quality assessment. The included studies outcome effects in every study are equal. were all found to be at potentially high risk of bias according to the Newcastle–Ottawa score and Co- RESULTS chrane assessments (Supplementary Table II). All of Study selection. The initial literature search the nonrandomized studies lacked well-matched yielded 1825 potential studies of interest after stapled and handsewn groups, independent out- duplicates were removed. After abstract screening, comes assessment, or data regarding patient comor- there were 36 full texts assessed for eligibility. One bidities or risk factors for adverse outcomes. The study11 was excluded because it included overlap- studies were also deemed to be at high risk of clinical ping patients with a later study,3 and 28 further heterogeneity owing to variation in outcomes defi- studies did not adequately compare outcomes after nitions for anastomotic leak, abscess, and fistula, handsewn and stapled anastomosis. Final analysis including a combination of radiologic, operative therefore included 7 studies, consisting of 5 studies and clinical findings (Supplementary Table III). regarding TS laparotomy,3,4,12,13 and 2 studies There was some asymmetry in funnel plots, indi- regarding EGS laparotomy5,14 (Fig 1). Two of the cating the presence of a degree of publication studies were prospective (one each of trauma and bias, most likely arising from the pooling of results EGS), of which one (EGS) was randomized. The from low-quality and small numbered studies remainder were retrospective cohort studies. (Supplementary Figure). Study characteristics. The 7 analyzed studies Reporting of outcomes. One TS study12 and 1 included patients from the United States, Canada, EGS study5 only reported outcomes with total Surgery Naumann et al 613 Volume 157, Number 4

Table I. Study settings, design, and number of patients Study Total patients, n (%) Total anastomoses, n (%) period/ Study y months Nations Type Design Stapled Handsewn OR Stapled Handsewn OR Brundage 2001 48 USA TS R 111 (55.8) 63 (31.7) 3.2 175 (60.1) 114 (39.4) 3.6 Catena 2004 72 Italy EGS RCT 106 (52.7) 95 (47.3) 1.3 106 (52.7) 95 (47.3) 1.3 Demetriades 2002 20 Colombia, TS PNR 79 (38.2) 128 (61.8) 1.2 79 (38.2) 128 (61.8) 1.2 South Africa, USA Farrah 2013 55 USA EGS R — — — 133 (57.1) 100 (42.9) 2.8 Kashuk 2009 84 USA TS R 6 (20.70) 23 (79.3) 0.7 6 (20.7) 23 (79.3) 0.7 Kirkpatrick 2003 103 Canada, USA TS R 55 (43.3) 25 (19.7) 1 64 (40.5) 38 (24.1) 1.4 Witzke 2000 130 USA TS R 93 (73.8) 27 (21.4) — 110 (34) 34 (23.6) 0.7

EGS, Emergency general surgery; OR, odds ratio; PNR, prospective nonrandomized; R, retrospective; RCT, randomized controlled trial; TS, trauma surgery. anastomoses as the chosen denominator, with no failure at patient level (beta, À0.204; 95% CI, data reported with regard to total number of pa- À0.647, 0.240; P = .240), or at an anastomosis level tients. The remaining 5 studies (4 TS and 1 EGS) (beta, 0.035; 95% CI, À0.299, 0.299; P = .746). all reported outcomes with both total anastomoses Secondary outcomes. For the studies that re- and total patients as denominators. Of the 7 ported anastomotic leak, there was no difference studies reporting data on an individual anasto- at an anastomosis level between stapled (20/540; mosis level, 6 reported anastomotic leak, 4 re- 3.7%) and handsewn (21/432; 4.9%) groups (OR, ported abscesses, 4 reported inpatient duration 1.00; P = 0.99). Similarly, there was no difference of stay, and 3 reported enterocutaneous fistulae. on an individual patient level between the stapled Of the 5 studies reporting data on an individual pa- (19/450; 4.2%) and handsewn (20/361; 5.5%) tient level, all 5 reported anastomotic leak and groups (OR, 1.11; P = .80). For the studies report- postoperative mortality, 3 reported abscesses, and ing abscesses, there was no difference between 2 reported enterocutaneous fistulae. stapled and handsewn groups on a patient level Because only 2 studies reported injury severity (41/450 [9.1%] vs 27/361 [7.5%]; OR, 1.58; score,3,6 2 studies reported duration of stay in P = .12), or on an anastomosis level (53/540 intensive care,3,6 and 2 studies compared oper- [9.8%] vs 32/432 [7.4%]; OR 1.42; P = .24). ating times,5,14 a meta-analysis was not performed Fistulae could only be analyzed at an anastomosis for these outcomes. Similarly, because only 1 level, and there was no difference between the sta- study3 reported associated injury score, 1 reported pled (8/540; 1.5%) and handsewn (3/432; 0.70%) surgical site infections,14 and 1 reported perioper- groups (OR, 1.29; P = .69). There was no differ- ative blood product requirement,6 these variables ence in inpatient duration of stay between stapled could not be meta-analyzed. and handsewn groups (weighted mean difference, Primary outcome. Table II summarizes the meta- À1.37; P = .180). There was no difference between analysis of the outcomes in terms of (1) total anasto- stapled and handsewn groups with regard to post- moses and (2) total patients. When the primary operative death (OR, 1.87; P = .100). outcome of anastomotic failure was assessed on an Subgroup analysis. In a subgroup analysis of anastomosis level for all 7 studies, there was no dif- patient characteristics, those in the TS group were ference in outcome, with 101 of 673 anastomotic more likely to have multiple anastomoses (688 failures (15%) in the stapled group, and 62 of 532 patients with 827 anastomoses) than the EGS group in the handsewn group (11.7%; I2 28%; OR, 1.53; (432 patients with 434 anastomoses). Trauma pa- 95% CI, 0.97, 2.43; P = .070; Fig 2, A). When anasto- tients were more likely to be male than EGS patients motic failure was assessed on a patient level for all 5 (86.4% and 47.0%, respectively). Mean age was eligible studies, there was similarly no difference be- significantly younger for TS studies (29.9 years) tween groups, with 66 of 357 anastomotic failures than for EGS studies (62.9 years). (18.5%) in the stapled group and 50 of 334 The 2 EGS studies could only be analyzed on an (15.0%) in the handsewn groups (I2 0%; OR, 1.44; anastomosis level because 1 of them only reported 95% CI, 0.92, 2.25; P = .110; Fig 2, B). this denominator.5 There was no difference be- Using meta-regression analysis, there were no tween the groups with regard to anastomotic fail- associations between year of study and anastomotic ure (27/239 [11.3%] stapled vs 11/195 [5.64%] 614 Naumann et al Surgery April 2015

Table II. Summary table for all stapled versus handsewn anastomoses outcomes reported for individual anastomosis and individual patients Random effects analysis (stapled vs handsewn) Outcome Total (n) OR/WMD 95% CI I2 P value Individual anastomoses Anastomotic failure 1,205 1.53 0.97, 2.43 28 .070 Anastomotic leak 972 1.00 0.42, 2.40 20 .999 Abscess 742 1.42 0.79, 2.57 23 .240 Enterocutaneous fistula 535 1.29 0.36, 4.69 0 .690 Inpatient duration of stay (d) 815 À1.37 À3.37, 0.63 49 .180 Individual patients Anastomotic failure 691 1.44 0.92, 2.25 0 .110 Anastomotic leak 691 1.11 0.53, 2.31 0 .800 Abscess 461 1.58 0.89, 2.79 2 .120 Postoperative death 918 1.87 0.88, 3.96 0 .100

OR, Odds ratio; WMD, weighted mean difference.

Figure 2. Anastomotic failure for all studies on an (A) individual anastomosis level and (B) individual patient level. handsewn; P = .06; Fig 3, A). Furthermore, there The 5 trauma studies were analyzed on both an was no difference in mean duration of hospital individual patient and individual anastomosis stay between stapled and handsewn groups level. There was no difference between the groups (weighted mean difference, À2.2; 95% CI, À4.84, with regard to anastomotic failure on an individual 0.43; P = .10). There was no difference in mortality anastomosis level (74/434 [17.1%] stapled versus between the stapled (14/239 [5.9%]) and hand- 51/337 [15.1%] handsewn; P = .30; Fig 3, B). Simi- sewn (7/195 [3.6%]) groups (P = .29). larly, there was no difference on an individual Surgery Naumann et al 615 Volume 157, Number 4

Figure 3. Subgroup meta-analysis of anastomotic failure for emergency general surgery and trauma surgery studies. A: Anastomotic failure on an individual anastomosis level for emergency surgery studies; B: Anastomotic failure on an individual anastomosis level for trauma surgery studies; C: Anastomotic failure on an individual patient level for trauma surgery studies. patient level (59/251 [23.5%] stapled versus 45/ specifically described the criteria for handsewn 239 [18.8%] handsewn; P = .18; Fig 3, C). There anastomoses in terms of which suture and which was no difference in mortality between the stapled type of anastomosis (ie, single layer or double (17/269 [6.3%]) and handsewn (4/215 [1.8%]) layer; Supplementary Table III). groups (P = .19). Operative technique. The types of individual DISCUSSION anastomoses varied between studies; 2 trauma Main findings. This systematic review and meta- studies4,12 excluded large bowel anastomosis and analysis specifically investigates the differences in 1 trauma study6 excluded small bowel anastomosis. outcomes between handsewn and stapled anasto- Three studies specifically described the stapling mosis after emergency laparotomy. The results technique and product used for stapled anastomo- using a random effects model show that there ses5,12,14; however, no studies explicitly discussed are no differences between the handsewn and whether the staple line was oversewn. Four studies stapled groups in terms of anastomotic failure on 616 Naumann et al Surgery April 2015 either individual patient or anastomosis levels. Handsewn anastomoses have been reported as Similarly, there are no demonstrable differences more time consuming than stapled during elective in the rates of anastomotic leak, abscess, fistula, surgery.19,21 Such differences in time may be duration of hospital stay, or mortality between particularly important during emergency surgery, these groups. Rather than specifically favoring because duration of anesthetic and operative times either operative technique, these data give credi- may need to be kept to a minimum to curtail the bility to a surgeon selecting a technique of their adverse physiologic effects of prolonged surgery own choice in the knowledge that this is supported in a metabolically vulnerable group.22 Unfortu- by the currently available summation of evidence. nately, only 2 of the included studies in this review Subgroup analysis. When comparing the TS and compared duration of surgery between techniques, EGS subgroups there are some (perhaps obvious) both of which were EGS studies.5,14 However, they patterns that emerge: Trauma patients were both reported a significantly shorter operating younger and more likely to be male. They were time for stapled anastomosis than handsewn, also much more likely to have multiple anastomo- although neither demonstrated an effect on pa- ses during a single laparotomy. Similar to the meta- tient outcome. Comparison of overall cost for analysis of all the studies, when using random each of the operative techniques has previously effects analysis, neither of the 2 subgroups favored been reported in a systematic review regarding handsewn or stapled anastomosis in any of the elective surgery,23 but such considerations could outcome measures. However, the number of not be analyzed in this study owing to lack of studies in each subgroup was low (especially reporting by the individual studies. EGS). Because there are only 2 studies examining Strengths and limitations of this study. Statistical the outcomes after EGS, we recommend that approach. When analyzing data from different further evidence in this particular subgroup may studies with a high likelihood of differences in be required to more fully address the question in populations and methodology, it is important to this circumstance. incorporate this heterogeneity into the analysis of Comparison with other studies. There have the overall effects of the intervention.10 The au- been a number of systematic reviews comparing thors’ suspicions regarding the heterogeneity and stapled and handsewn techniques for elective quality of studies was borne out by quality assess- surgery in the last 15 years,15-18 the most recent ment of individual studies, confirming the superi- identifying 7 RCTs for ileocolic anastomoses2 and ority of the random effects to the fixed effects 9 for colorectal anastomoses.1 A recently published model. One potential problem when using this systematic review of systematic reviews and pano- model is that it is much more likely that variances, ramic meta-analysis of pooled estimates that CIs, and standard errors for the overall summary compare stapled and handsewn anastomoses for effects will be larger, making it more likely that all operative procedures analyzed 11 systematic re- the null hypothesis will be accepted (ie, there are views and reported no evidence of superiority for no differences in outcome between handsewn either stapled or handsewn anastomosis with and stapled techniques). Such an approach, respect to postoperative complications, or dura- despite being less likely to demonstrate superiority tion of stay, but a potential benefit of stapled anas- in an operative technique, is nevertheless superior tomosis in ileocolic anastomosis.19 However, none to the alternative of using fixed effects modelling, of these systematic reviews specifically analyzed which would require errant assumptions to be emergency surgery (or indeed separate trauma or made regarding interstudy homogeneity. What re- EGS subgroups) as a separate entity. There is a mains unknown, unfortunately, is whether there striking paucity of data regarding anastomosis in is a genuine yet undetected overall effect. emergency laparotomy, and this may warrant Individual studies. In terms of study availability, an further investigation through well-designed cohort obvious limitation to the current review is that there studies if RCTs are not feasible. Because such was only 1 RCT that directly compared handsewn future studies are unlikely to be RCTs, but instead and stapled anastomosis in the emergency setting,14 utilize prospective observational data regarding and another prospective study that was nonrando- operative techniques, propensity score matching mised6; the remainder were retrospective, cohort analysis may be used to measure treatment effects studies. Furthermore this single RCT did not stand while reducing the bias associated with nonran- up well to the Cochrane Collaboration’s tool for as- domization. We observed that none of the non- sessing risk of bias in randomized trials, with lack of randomized studies in this review utilized this clarity regarding randomization and allocation statistical test, although it was described in 1983.20 concealment. Surgery Naumann et al 617 Volume 157, Number 4

Definitions. Unfortunately, although anastomotic own preference. However, such practice must also leak is among the most commonly reported out- be undertaken with a cautious skepticism, because comes in colorectal surgery, its definition is vari- it is still possible that 1 technique is superior to able among studies,24 and this review was also another in certain circumstances that have not yet limited by some heterogeneity in this respect. In been borne out by evidence to date. Recent large- an attempt to compensate for this problem, the scale collaborations in the UK have demonstrated composite measure of ‘‘anastomotic failure’’ was the achievability of well-designed, multicenter, used as the primary outcome, which allowed the surgical studies through regional networks25 that authors to encompass all complications for a given could be used as a model in the design larger individual anastomosis or patient (ie, the sum of and more reliable studies to further investigate all reported anastomotic leaks, abscesses, and more specific question regarding bowel anasto- fistulae). We believe that anastomotic failure is mosis in trauma and emergency abdominal sur- also the most clinically relevant measure in surgical gery, even if not in the form of RCTs. practice because it is the key element in whether a patient requires intervention after operative resec- SUPPLEMENTARY DATA tion, regardless of the perceived etiology or Supplementary data related to this article can be presentation. found online at http://dx.doi.org/10.1016/j.surg.2014. Potential for missing data. Although the systematic 09.030. search for studies was thorough, it is possible that some data may have missed inclusion owing to being published in a language other than English, REFERENCES and the authors acknowledge that ideally all 1. Neutzling CB, Lustosa SA, Proenca IM, da Silva EM, Ma- tos D. Stapled versus handsewn methods for colorectal published literature might be used for analysis. anastomosis surgery. Cochrane Database Syst Rev 2012: Furthermore, unpublished data were not included, CD003144. which was considered sensible by the authors, who 2. Choy PY, Bissett IP, Docherty JG, Parry BR, Merrie A, believe that peer-reviewed data represent more Fitzgerald A. Stapled versus handsewn methods for ileo- reliable basis for analysis. colic anastomoses. Cochrane Database Syst Rev 2011: CD004320. There has been a trend toward increased utiliza- 3. Brundage SI, Jurkovich GJ, Hoyt DB, Patel NY, Ross SE, Mar- tion of minimally invasive techniques for emergency burger R, et al. WTA Multi-institutional Study Group. West- abdominal surgery, but it is likely that open laparot- ern Trauma Association. Stapled versus sutured omy will still be the most commonly performed gastrointestinal anastomoses in the trauma patient: a multi- technique for trauma and EGS. It is possible that the center trial. J Trauma 2001;51:1054-61. 4. Kirkpatrick AW, Baxter KA, Simons RK, Germann E, Lucas inclusion of data from laparoscopic surgery may CE, Ledgerwood AM. Intra-abdominal complications after have increased the selection bias for the stapled surgical repair of small bowel injuries: an international re- technique. However, this meta-analysis was unaf- view. J Trauma 2003;55:399-406. fected by such bias, with no minimally invasive data 5. Farrah JP, Lauer CW, Bray MS, McCartt JM, Chang MC, Mer- reported in any of the included studies. edith JW, et al. Stapled versus hand-sewn anastomoses in emergency general surgery: a retrospective review of out- Policy implications. Although the gold standard comes in a unique patient population. J Trauma Acute for best possible practice to guide surgeons would Care Surg 2013;74:1187-92. be through multicenter, randomized controlled 6. Demetriades D, Murray JA, Chan LS, Ordonez~ C, Bowley D, trial data, it is clear from the current review that Nagy KK, et al. Handsewn versus stapled anastomosis in there is a paucity of this standard of evidence, and penetrating colon injuries requiring resection: a multi- center study. J Trauma 2001;52:117-21. the studies available are at overall high risk of bias. 7. Wells GA, Shea B, O’Connell D, Peterson J, Welch V, Losos M, With such deficiencies, calling for further studies et al. The Newcastle-Ottawa scale (NOS) for assessing the qual- may seem intuitive to those with equipoise ity of nonrandomized studies in meta-analyses [cited 2014 May regarding optimal operative technique, but unfor- 10]. Ottawa Hospital Research Institute. Available from: http:// tunately the practicalities of such aspirations are www.ohri.ca/programs/clinical_epidemiology/oxford.asp. 8. Higgins JP, Altman DG, Gøtzsche PC, Juni€ P, Moher D, Ox- not straightforward. Ours is the first systematic man AD, et al. Cochrane Bias Methods Group. Cochrane review of the literature to formally summaries the Statistical Methods Group. The Cochrane Collaboration’s current available data regarding handsewn versus tool for assessing risk of bias in randomised trials. BMJ stapled anastomosis in trauma and EGS laparot- 2011;18:343. omy. Although the available meta-analysis is less 9. Moher D, Liberati A, Tetzlaff J, Altman DG, PRISMA Group. Preferred reporting items for systematic reviews and meta- than ideal, it suggests that until any further evi- analyses: the PRISMA statement. Int J Surg 2010;8:336-41. dence is presented, surgeons and institutions may 10. DerSimonian R, Laird N. Meta-analysis in clinical trials. opt for either operative technique based on their Control Clin Trials 1986;7:177-88. 618 Naumann et al Surgery April 2015

11. Brundage SI, Jurkovich GJ, Grossman DC, Tong WC, Mack 19. Hemming K, Pinkney T, Futaba K, Pennant M, Morton DG, CD, Maier RV. Stapled versus sutured gastrointestinal anas- Lilford RJ. A systematic review of systematic reviews and tomoses in the trauma patient. J Trauma 1999;47:500-7. panoramic meta-analysis: staples versus sutures for surgical 12. Witzke JD, Kraatz JJ, Morken JM, Ney AL, West MA, Van procedures. PLoS One 2013;8:e75132. Camp JM, et al. Stapled versus hand sewn anastomoses in 20. Rosenbaum PR, Rubin DB. The central role of the propen- patients with small bowel injury: a changing perspective. sity score in observational studies for causal effects. Bio- J Trauma 2000;49:660-5. metrika 1983;70:41-55. 13. Kashuk JL, Cothren CC, Moore EE, Johnson JL, Biffl WL, 21. McLeod RS, Wolff BG, Ross S, Parkes R, McKenzie M, Inves- Barnett CC. Primary repair of civilian colon injuries is safe tigators of the CAST Trial. Recurrence of Crohn’s disease in the damage control scenario. Surgery 2009;146:663-8. after ileocolic resection is not affected by anastomotic 14. Catena F, La Donna M, Gagliardi S, Avanzolini A, Taffurelli type: results of a multicenter, randomized, controlled trial. M. Stapled versus hand-sewn anastomoses in emergency in- Dis Colon Rectum 2009;52:919-27. testinal surgery: results of a prospective randomized study. 22. Khan A, Hsee L, Mathur S, Civil I. Damage-control laparot- Surg Today 2004;34:123-6. omy in non trauma patients: review of indications and out- 15. MacRae HM, McLeod RS. Handsewn vs. stapled anastomo- comes. J Trauma Acute Care Surg 2013;75:365-8. ses in colon and rectal surgery: a meta-analysis. Dis Colon 23. Gong J, Guo Z, Li Y, Gu L, Zhu W, Li J, et al. Stapled vs hand Rectum 1998;41:180-9. suture closure of loop ileostomy: a meta-analysis. Colorectal 16. Lustosa SA, Matos D, Atallah AN, Castro AA. Stapled versus Dis 2013;1:e561-8. handsewn methods for colorectal anastomosis surgery. Co- 24. Whistance RN, Forsythe RO, McNair AG, Brookes ST, Av- chrane Database Syst Rev 2001:CD003144. ery KN, Pullyblank AM, et al. Core Outcomes and iNfor- 17. Lustosa SA, Matos D, Atallah AN, Castro AA. Stapled versus mation SEts iN SUrgical Studies - ColoRectal Cancer handsewn methods for colorectal anastomosis surgery: a Working Group. A systematic review of outcome report- systematic review of randomized controlled trials. Sao Paulo ing in colorectal cancer surgery. Colorectal Dis 2013;15: Med J 2002;120:132-6. e548-60. 18. Choy PY, Bissett IP, Docherty JG, Parry BR, Merrie AE. Sta- 25. Bhangu A, Kolias AG, Pinkney T, Hall NJ, Fitzgerald JE. pled versus handsewn methods for ileocolic anastomoses. Surgical research collaboratives in the UK. Lancet 2013; Cochrane Database Syst Rev 2007:CD004320. 382:1091-2.