Endoscopic treatment of colorectal perforations – a systematic review

Perforations may occur due to excessive mechanical force, either if the endoscope is directly inserted into a diverticulum or when passing a difficult bend as well as by retro-flexion in the rectum. Over-insufflation can cause barotrauma, and thermal injury caused by the use of electrocoagulation or argon plasma coagulation can lead to transmural burn and tissue necrosis [1, 4-6].

The most common site for perforations is the recto-sigmoid junction or the sigmoid colon. Risk factors include diverticulosis, severe inflammation, cancer and stenosis. Other risk factors are therapeutic interventions like endoscopic mucosal resection and endoscopic submucosal dissection, as well as size and number of polyps removed during intervention. The caecum with its thinner wall and larger diameter is more susceptible to thermal injuries and barotrauma. Defects seen during diagnostic procedures have been reported to be larger than those produced during therapeutic interventions [1, 4-7].

Perforations are detected either immediately during sigmoidoscopy/colonoscopy by visualisation of extra-intestinal tissue or post-procedure by visualisation of free air on abdominal computed tomography. The clinical signs are abdominal pain, distension, tachycardia, fever and nausea. The laboratory findings include leukocytosis and/or elevated C-reactive protein [8].

Treatment has traditionally been surgery either by open or laparoscopic suturing of the perforations. However, when perforations are not manageable by suturing, more extensive surgery with bowel resection and stoma may be required. Conservative non-surgical management has been described in small series based on highly selected patient groups [5, 9-11].

Endoscopic closure of perforations of the colon has previously been reported only sporadically in the literature since the first case report by Yoshikane et al in 1998 [12]. Subsequently, it has been reported that different types of endoscopic closure devices have been used to close perforations, most commonly endoscopic clips [4, 13].

The aim of this article was to assess the existing evidence on endoscopic closure of iatrogenic perforations of the colon and rectum through a systematic review of the literature.

METHODS

The literature search was performed by two authors (MAH, CØT). The strategy implemented was in accordance with the PRISMA guidelines.

Eligibility criteria and information sources

We searched the databases for clinical trials, case reports and small series and for retrospective, prospective and cohort studies. Studies published in English were included. Trials on animal models were excluded. Data were gathered through literature search in the following databases: MEDLINE, Embase and the Cochrane library. No limitations were used.

The search was made until 27 February 2015. The search criteria used were (“colonoscopy/adverse effects”[MeSH Terms]) AND clip, OR (((“colonoscopy/adverse effects”[MeSH Terms]) AND perforation)) AND clip, OR ((“colon”[MeSH Terms]) AND perforation) AND clip, OR ((clip) AND closure) AND colon, OR ((rectum) AND clip) AND perforation. Additionally, records were added after the authors’ reference lists had been reviewed manually for relevant articles. In case an article could not be retrieved, the author or publisher was contacted.

Study selection and data collection process

The search identified 275 records (Figure 1) including the manual reference list search. A total of 67 records were screened and 32 records were excluded. This included trials on animals, non-English language publications, perforations in the upper gastrointestinal (GI) tract, closure of fistulas or duplicates. A total of 35 full-text articles were retrieved for closer evaluation. Three of these were excluded due to conflicts of interest, other causes of perforation than sigmoidoscopy/colonoscopy or a combination of surgery and clipping. Thus, a total of 32 articles were finally included in the qualitative analysis of this review.

RESULTS

This review included 19 case reports and 13 studies, including three prospective studies, one case control study and nine retrospective studies.

Case reports

A total of 19 case reports [12, 14-31] with reports on 23 patients were included. Cases were published during the 1997-2015 period (Table 2). Perforations were associated with both diagnostic and therapeutic sigmoidoscopies/colonoscopies. Successful endoscopic closure was reported in all but one case [20].

Case-control studies

Won et al [32] included 22 patients between 2004 and 2009 in a single-centre case-control study. Patients were divided into a surgical and non-surgical group. The non-surgical group (two diagnostic- and three therapeutic colonoscopies) underwent conservative treatment with successful clip application and closure of the perforations. A total of 17 patients (nine diagnostic and eight therapeutic colonoscopies) were treated surgically. The authors found a significant difference in favour of the non-surgical group regarding fever and abdominal pain, but no significant difference in length of stay (LoS) between the two groups. Sample size is small in this study and procedures are not standardised; furthermore, pain is a subjective symptom.

Prospective studies

Voermans et al [33] included 36 patients (13 colon- and 23 upper-GI tract) with acute iatrogenic perforations of the GI tract from 2009 to 2010 in a multicentre study. The clip device used was the Over-The Scope-Clip (OTSC) (Ovesco Endoscopy, Tubingen Germany). Endoscopic repair was conducted according to a standardised operating procedure. Perforations of 3 cm or less were included. The primary endpoints were successful closure, absence of leakage on water-soluble contrast X-ray and absence of adverse events within 30 days after endoscopic closure. Out of 13 colon perforations, 12 were managed endoscopically (92% success rate). However, one patient died after 36 h of inclusion after abdominal surgery due to failure to close the perforation with clips.

Gubler et al [34] included 14 GI perforations, (nine colon- and five upper GI tract) to endoscopic closure between 2009 and 2011 in a single-centre study. The main outcome was defined as technically successful closure after subsequent clinical observation for 24 h. All nine colon perforations were closed endoscopically with a single OTSC. The authors reported a technical success rate of 100%. Three patients underwent laparoscopy due to pain where the lesion was confirmed to have been sealed.

Heldwein et al [35] included 2,257 patients and found 26 perforations in a multi-centre study with data collected in the course of 20 months. The endpoints were bleeding, bowel perforation and death during a 30-day follow-up period. Of the 26 perforations observed, nine were detected immediately, 11 within 24 h and six after more than 24 h. A total of 12 perforations were managed successfully non-surgically; five by clip application and seven conservatively.

Retrospective studies

Magdeburg et al [36] included 22,924 colonoscopies and detected 105 iatrogenic perforations (29 diagnostic and 76 therapeutic perforations) in the 2004-2011 period in a single-centre study. A total of 71 patients underwent immediate endoscopic repair with clips, and 59 of these were successful (83.1%). Clip application was possible in 62 of 76 (81.55%) cases in the therapeutic group, whereas clip application was possible only in nine of 29 (31.03%) patients in the diagnostic group. This is the largest study conducted to date. The definition of a successful closure was not detailed.

Kim et al [37] included 115,285 diagnostic sigmoidoscopies/colonoscopies and observed 26 perforations, from 2000 to 2011 in a multicentre study. Endoscopic closure was attempted in 16 patients, 13 were successful (81%). Successful closure was defined as complete closure of mucosal defect and improvement of clinical manifestations such as fever, leukocytosis and signs of peritoneal irritations. Perforations were significantly smaller in the endoscopic closure group. None of the patients in the endoscopic clipping group needed further surgical treatment.

Cho et al [38] investigated 51,738 diagnostic and therapeutic colonoscopies with 33 iatrogenic perforations detected between 2005 and 2009 at three centres. A total of 29 perforations were initially managed by endoscopic closure, with success in 17 patients (59%) defined as a hospital stay that was shorter than two weeks, no complications and complete resolution of the perforations. Complications, defined as a longer hospital stay than two weeks, were seen in five patients, including two with abscess formation. A total of 22 patients were treated non-surgically (76%). Seven patients underwent surgery.

In a study by Jovanovic et al [39], 8,601 colonoscopies were performed. Twelve iatrogenic perforations occurred (five diagnostic and seven therapeutic colonoscopies). Data were collected from a single centre between 2002 and 2008. Endoscopic closure was attempted in six patients of whom one failed (success rate 83%). The remaining six did not undergo an endoscopic attempt of closure due to the size of the perforations, technical difficulties, stool contamination or lack of experience.

Kang et al [40] included 20,660 sigmoidoscopies and 17,102 colonoscopies. Data on 53 procedure-related perforations from 2000 to 2007 were evaluated in a single-centre study. A total of 19 patients were treated conservatively and 34 underwent surgery. Conservative treatment was described as all means of non-operative medical management including fasting, hydration, intravenous antibiotics and serial abdominal examinations. The conservative group included seven successful endoscopic closures and 12 close observations. Endoscopic clipping was successful in seven out of nine attempted closures (78%). The remaining two underwent surgery. Analysing the conservative group, authors found that patients treated by clipping had significantly fewer days of fasting, a more limited use intravenous antibiotics and a shorter LoS than the patients in the non-clipping group.

Taku et al [41] presented 15,160 therapeutic colonoscopies with 23 colon perforations. Data were collected from the 1999-2003 period at four centres. Out of the 23 cases, 16 patients had immediate perforations, and the remaining seven had delayed perforations. Endoscopic clipping was attempted in 13 patients (56.5%) with immediate perforations. Successful closure was possible in nine patients (69.2%). The remaining patients in this group were managed surgically (five patients) and conservatively (two patients). In the group with delayed perforations, one was managed surgically and six conservatively. Successful closure was defined as clipping without any untreated closure or closure defects.

Yang et al [42] included 118,115 colonoscopies resulting in 38 perforations; 13 diagnostic and 25 therapeutic. Data were collected from 2001 to 2008 at a single centre. Perforations were categorised as endoscopically evident, suspected or radiologically proven. Successful closure was defined as a complete closure of the definite or suspected mural defect. Endoscopic closure after diagnostic perforations was attempted in six out of 13 perforations (42.6%), with successful outcome without surgery in five of six patients (83.3%); the last patient underwent surgery without further complications. In all, attempted closure was successful in 21 of 22 patients (95.5%). Perforations were smaller in the therapeutic than in the diagnostic group (Table 3).

Niyashama et al [43] had only two patients with perforations in the sigmoid colon after therapeutic colonoscopy and both were managed successfully with clips.

Seebach et al [44] presented a small study that included three cases with a colon perforation out of seven GI perforations. Three patients were treated successfully by a single OTSC. One of these patients underwent emergency laparotomy. Free air and a closed perforation were found. The further course was uneventful for all patients.

The total number of patients (apart from case reports) who have undergone endoscopic closure and been described in the literature in this study is 203. The size of perforation ranged from 2.9 to 40 mm and size was reported in seven out of 13 studies (Table 3). The mean clinical success was 87.8% (standard deviation: ± 13.0) with a median of 92.3% (58.6-100). The total number of patients needing surgery as a consequence of clip failure was 30 (14.8%). In two additional articles, four patients underwent surgery and sealed perforation was confirmed [34, 44]. One patient died after reported clip failure (0.5%) [33].