Loop ileostomy closure: a retrospective comparison of three techniques

Loop ileostomy (LI) formation is a common practice for patients undergoing low anterior resection or restorative ileo‐anal pouch surgery. Ileostomy closure can be performed using a stapled or hand‐sewn technique, with or without resection. If hand‐sewn, the closure can be one or two layers. Randomized controlled trials have not demonstrated one technique to be superior, and meta‐analyses are limited by the heterogeneity of published studies. Our primary aim is to compare stapled ileostomy closure with single‐ and two‐layer hand‐sewn closures.


Introduction
Defunctioning loop ileostomy (LI) formation is a common practice for patients undergoing low anterior resection and ileo-anal pouch surgery. The use of a covering ileostomy is shown to reduce the rate of clinically significant anastomotic leaks, thereby reducing morbidity and re-operation rates. 1,2 However, subsequent closure is not without morbidity, with complications reported in up to 33% of cases. 3,4 Complications include wound infection, anastomotic problems and ileus. Ileus has a reported incidence of 0-15%, 4 with proposed mechanisms including peri-anastomotic oedema resulting in luminal narrowing, and delay in return of function in the previously defunctioned distal limb.
Anastomoses can be sewn in one or two layers, or stapled. Stapled closure of a LI was first described by Kestenberg and Becker,5 and has gained favour for the perceived speed and the wider anastomosis it produces. Advocates argue that the wider side to side stapled anastomoses result in earlier return of small bowel function, and additional cost of the stapling devices is offset by reduced operating time. 6,7 There is also debate as to whether a single-or twolayer hand-sewn closure is superior. Proponents of single-layer techniques argue this results in a reduction in ischaemia of the suture line, less narrowing of the lumen, and shorter operating time. 8 This study reviews our 17-year dataset to compare operating time, length of stay (LOS), morbidity and readmission rates associated with stapled and hand-sewn ileostomy closure using either one or two layers.

Methods
This retrospective, single-centre cohort study utilized data obtained from the Otago Clinical Audit, 9 a prospective, multicentre database collecting details of surgical admissions, operations, complications and re-admissions. Institutional approval was given for interrogation of the Otago Clinical Audit database for all LI closures performed between January 1999 and April 2016 at Dunedin Public Hospital. Demographic data were collected for age and gender. Operative data were collected for American College of Anaesthesiology grade (ASA), ileostomy closure technique, length of operation, LOS and grade of the surgeon/assistant. The clinician's identified complications followed standard definitions. 10 For ileus, this included requiring a nasogastric tube or being nil by mouth on post-operative day three or more, and small bowel obstruction required radiological or clinical diagnosis of a mechanical blockage of the small bowel. Mortality was defined as in-hospital mortality and readmissions were defined as return to hospital within 30 days of discharge. All patients with LOS >1 week without a recorded complication had case notes reviewed, and any previously unidentified complication was included.
Patients were excluded from analysis if there was incomplete documentation of the anastomotic technique, if stoma closure was an ileo-colic anastomosis, reversal of LI was in conjunction with another operation, or the operation was performed acutely because of small bowel obstruction. Anastomoses were categorized as either single-or two-layer hand-sewn, or stapled. This work has been reported according to STROCCS criteria. 11

Statistical analysis
Appropriate summary statistics were used to describe the sample, overall and by procedure type. Medians alongside 25th and 75th percentiles are shown for skewed continuous variables. Age, sex and ASA were compared between procedure types using Kruskal-Wallis tests for continuous variables and chi-squared tests for categorical variables. For continuous outcomes, linear regression models were used to estimate differences in means between procedures, initially without adjustment, then adjusting for age, sex and ASA; then adding consultant versus registrar. For binary outcomes, Poisson regression models with robust standard errors were used in a similar manner to estimate relative risks. For count outcomes, truncated Poisson (or negative binomial where the likelihood ratio test indicated over-dispersion) regression was used, with zero values truncated, to model ratios of means. The number of covariates was limited using a minimum of 10 events per variable. 12 All models included Froot's clustered robust errors at the clinician level. 13 For linear regression models, residual normality and homoscedasticity were examined through appropriate plots and natural logarithmic transformations used where appropriate. All analyses were performed using Stata 15.1 (College Station, Texas, USA) and two-sided (one-sided for chi-squared tests) P < 0.05 was considered statistically significant. Pairwise comparisons were performed where overall Wald tests were statistically significant and no further adjustments for multiple comparisons were made. This audit was given institutional approval, and the need for ethical approval and individual consent was waived.

Results
We identified 280 consecutive ileostomy closures with 244 meeting inclusion criteria. Excluded cases included 13 performed acutely, nine with inadequate description of the closure technique, seven with an ileocolic anastomosis and seven with ileostomy closure combined with another operation. Patients consisted of 138 males and 106 females, with a median age of 67 years (ranging from 13 to 88). Of included cases, 40% were performed by five consultants. A hand-sewn ileostomy closure was performed in 165 cases, of which 148 were single layer closures (60.7%) and 17 were twolayer closures (7.0%). There were 79 stapled ileostomy closures (32.4%). Single-layered closures usually involved direct closure of half the ileal circumference, without an ileostomy resection. Patient characteristics are summarized in Table 1. Patients undergoing twolayered closures were older (Dunn's test P ≤ 0.031), although there was no evidence of differences in ASA. While the use of stapled closures increased from 16.2% in the first half of the study period to 45.9% in the second half, there was no change in the likelihood of two-layered sewn closures (6.3% and 7.5%). As shown in Table 2, the duration of surgery ranged between 20 and 210 min, with a geometric mean (95% confidence interval (CI)) of 73.5 (70.4; 76.7). There was no significant difference in operative time between stapled single-or two-layer ileostomy closure, with respective times of 71.5, 73.1 and 88.5 min (overall P = 0.25).
LOS also varied, ranging between 1 and 63 days, with an overall mean (95% CI) of 5.8 (4.9,6.6). When adjusting for differences in patient demographics, ASA and consultant surgeon between groups, there was evidence for differences in LOS (P = 0.034), with significantly longer LOS for patients who had two-layered closures compared to stapled closures, staying 8.3 and 4.2 days, respectively (ratio of means (95% CI) 1.96, (1.08; 3.56), pairwise P = 0.026). There was no significant difference in LOS between hand-sewn single-layered and stapled (P = 0.105) or single and two-layered anastomoses (P = 0.192).
Morbidity and rate of readmissions did not differ between closure techniques ( Table 2). Individual complications are listed in Table 3. Five patients (2%) required re-operation. Reasons included an anastomotic leak in the two-layered hand-sewn group, a missed enterotomy for one patient in both the stapled and single-layered closure groups, one acute fascial dehiscence, and one reformation of ileostomy because of a previously unrecognized rectovaginal fistula. Two patients in the single-layered group developed pelvic collections, in spite of a normal pre-closure gastrograffin enema, which resolved with percutaneous treatment.
When adding year to the most adjusted models to examine changes in the duration of surgery, LOS, complications or readmissions, there was no evidence for linear trends over time.
For two-layered hand-sewn anastomoses, three consultants performed or supervised 10 of 17 recorded cases. Two preferred a two-layered closure and one changed his preference to a onelayered closure. When these same consultants performed singlelayered anastomoses, the mean LOS (95% CI) was 5.0 (4.3-5.8)  days, similar to the overall LOS for these closures. The operation notes were reviewed for difficulty of procedure by two independent reviewers. Thirteen of 17 procedures were routine and four were difficult, which contributed to a decision to resect small bowel on two occasions. However, the choice to perform a two-layered closure in these four cases was due to surgeon preference.

Discussion
Our results showed no significant (P = 0.105) benefit for a stapled closure compared to a single-layered closure for LOS, in contrast to the 1-day decrease in LOS reported in a 2018 meta-analysis. 14 In addition to this, by separately analysing single-and two-layered anastomoses, we demonstrated significant increases in stay for those undergoing a two-layered, compared to a stapled, closure.
Although the cost of stapling may not be repaid by gains in operating time, it may potentially be covered by improvements in LOS.
The most common complication we identified was small bowel obstruction/post-operative ileus, at an overall rate of 7.4% with no evidence of differences between groups. Studies use different criteria to define this complication, resulting in significant heterogeneity for this outcome. [15][16][17][18][19] Given this heterogeneity, LOS can be used as a surrogate for return of gut function.
We only had a small number of two-layered closures, and perhaps two-layered is now almost historic. However, we found a significantly longer LOS for two-layered, hand-sewn anastomoses compared to stapled anastomoses (8.3 versus 4.2 days) when adjusted for age, sex, ASA, and consultant surgeon. The mechanism for this may be the narrower lumen associated with a two-layered anastomosis. While few studies compare one and two-layered anastomoses when closing a LI, 8 a 2017 randomized controlled trial comparing single and double-layered ileal and colonic anastomoses noted a significant delay in the return of bowel sounds, passing first bowel motion, and LOS in the two-layered group. 20 When we reviewed results for two-layered operations, there was no evidence that our longer LOS was due to consultant discharge planning, a difference in the rate of complications, or that the choice of anastomosis was related to the difficulty of surgery. The duration of time from index surgery to stoma closure is another potential factor affecting post-operative outcomes. There were no differences in stoma duration between the closure groups (P = 0.49), so this was not added to our model.
The key differences reviewed in the literature when comparing closure techniques are the operative time, complications, and LOS. The additional operative cost of a stapling device is usually justified based on a reduction in operating time and/or LOS. 14,21,22 In randomized controlled trials comparing closures, there is a significantly shorter operating time for stapled closures, ranging from 4 to 24 min. [21][22][23][24][25] With respect to comparing different methods of suturing, the Cochrane review of all gastrointestinal anastomoses showed a mean difference of 11 min in favour of a single-layered over a two-layered anastomosis. 18 Three recent meta-analyses combining resection or no resection and different suturing techniques showed a reduction in operating time of approximately 12 min in favour of a stapled technique. [14][15][16] While our results showed no significant differences in operating time, our findings were similar to the literature, with a single-layered closure being approximately 2 min longer, and a two-layer closure technique taking approximately 15 min longer than a stapled anastomosis. In our study, for a single-layered anastomoses, time alone would not have been sufficient to offset the additional cost of a stapler.
To our knowledge, this is the only study which separately compares stapled ileostomy closures to both single-and two-layered hand-sewn techniques. One of the challenges when reviewing the literature is that a number of different hand-sewn techniques are used. Most systemic reviews and studies, especially those comparing suture and stapling, pool one-and two-layer techniques. 14,[21][22][23][26][27][28][29] While some studies do compare direct suture against ileal resection and suturing, [14][15][16]18,[22][23][24][25][26][27][28][29][30] we were unable to identify a review comparing one-and two-layered techniques. Although our study did not find statistically significant evidence of differences between hand-sewn techniques (P = 0.192), the differences in estimated mean LOS, along with wide confidence intervals (0.81, 2.87) suggest caution in pooling data for one-and two-layered techniques.
A factor to consider in comparing techniques is whether results are influenced by changes in practice over time. While some cohort studies show significant benefits for stapled closures, 6,7 studies assessing the effects of enhanced recovery after surgery have also reduced LOS and complications. 31 Similarly, the use of laparoscopic surgery for rectal resection with fewer adhesions may reduce operating time and ileus at the time of ileostomy closure. 32 In spite of these changes, and our increase in stapled ileostomy closure, both our study and a 2018 meta-analysis 14 found no statistical evidence of trends over time for the investigated outcomes.
This retrospective analysis of a prospective database has a number of potential limitations. The retrospective nature of our study means that complications which frequently develop after discharge from the hospital may not be captured. We note that our rate of complications is similar to that reported in other studies. 33,34 While the number of two-layer closures was limited, the statistically significant difference in LOS with such a small number of cases suggests this is a clinically important difference. A larger study assessing the impact of a two-layered closure would be interesting, however, as there is a longer operating time and no identified advantages, this is unlikely to be performed. Additionally, our lack of a significant difference between stapled and single-layered closures may have been due to a limited sample size. Differences in groups might also be due to surgeons favouring a particular anastomosis. This was not found to be a factor when we reviewed all two-layered closures and when individual surgeons' results were assessed.
In conclusion, our results and review of the literature suggest that differences between stapled and single-layered closure techniques are small, and it is reasonable for the choice to be determined by surgeon preference. We also demonstrated a significantly increased LOS for a two-layered technique compared to a stapled closure. As a two-layered closure has not been shown to reduce morbidity, increases LOS, and takes longer to construct, we would not recommend its routine use.