A postsurgical gastrointestinal (GI) leak is one of the most dreaded surgical complications with significant morbidity and mortality.^1–3 Leakage of luminal contents leads to the formation of infected collections with rapid development of systemic sepsis. Those who survive the initial event may end up with chronic tenacious fistulas strongly compromising quality of life and associated with prolonged hospitalization. The incidence of post-surgical leaks varies widely in different reports depending on several factors; mainly type, location and indication of surgery and surgeon experience.^2–6 The reported incidence in upper GI surgeries is in the range of 4% to 17% with a mortality of about 10% to 65%.^1,3,7 The incidence in bariatric surgeries in particular may be lower (1% in sleeve gastrectomy and up to 5% in Roux-en-Y gastric bypass [RYGB]) but the number of bariatric surgeries being performed has been rising exponentially over the last two decades making bariatric surgeries a commonly encountered cause of leaks.^2,4–6 The risk of leakage after colorectal anastomoses is about 3% to 15% with a mortality of about 10%.^8–10 The mainstay of treatment for GI leaks used to depend on a choice between either conservative management or surgical repair, any experienced surgeon would know that this is being “caught between the devil and the deep blue sea”. Conservative management involves prolonged total parenteral nutrition (TPN) and use of multiple antibiotics until spontaneous healing occurs. This approach entails prolonged hospitalization, recurrent infections, numerous complications of TPN and a low rate of success.⁷ The option of surgical repair is also far from optimal, anatomy is markedly altered by the recent surgery, the developing adhesions and presence of pus, repair very frequently fails and leaks recur due to tissue friability and the presence of infection.¹¹ Morbidity is usually very high with mortalities up 50%.^12,13 The exception to this of course is repair performed very early (few hours) after the leak occurs, a situation seldom encountered.

Over the last 2 decades endoscopic alternatives have surfaced as an exciting alternative in the management of leaks. The endoscopic approaches have the advantages of being minimally invasive, many are not affected by the condition of the leak edges and they reduce the need for prolonged parenteral nutrition and hospital stay. Studies on endoscopic approaches have generally reported favourable results, yet there is still a lack of large prospective and randomized studies. Another caveat in many of these studies is that they pooled perforations occurring iatrogenically during endoscopy with leaks occurring postsurgically. Iatrogenic endoscopic perforations have a much better prognosis mainly due to the very early diagnosis, many heal spontaneously, and the vast majority respond well to endoscopic intervention during the same endoscopic session. For the purpose of this review we will be discussing only the management of the more tenacious post-surgical leaks.

A patient with a leak is not the patient that will just be done within your routine endoscopy list, nor is there a routine approach that will be applied to all patients with leaks. It is vital for an endoscopist approaching such a case to understand that there is more to it than just “plugging the hole”, what is more important is maintaining this closure. This depends on several factors which necessitate a wider angle view of the patient, an understanding of some basic surgical principles and a very tight collaboration with the managing team. It is beyond the scope of this article to list all principles of leak management, but below are a few main principles that may be frequently overlooked by endoscopists.

Drainage

This may go without saying for most surgeons but it may be a principle long forgotten by many endoscopists. Drainage is arguably the single most important step in the management of a GI leak. The first question on the endoscopist’s checklist should be: are all the collections well-drained? Without proper drainage sepsis is inevitable, apart from being life threatening, it renders the patient in a poor general status impairing the healing process, and without proper drainage the formed pus will be continuously damaging the leakage site precluding any chance of healing. Inserting a stent blocking the leak site with no drains inserted could just block the only possible exit of the collected pus, seriously aggravating the infection process. Options of drainage include ultrasound- or computed tomography-guided drainage or surgical drainage. Drains should be inserted close to the leakage site and in any other collections even if apparently remote. Drains inserted too close have the risk of migrating into the lumen through the leak, or impairing healing due to their continuous vacuum effect. Repeated contrast studies and the gradual withdrawal of the drains prevent this complication from occurring.

Drainage can also be performed endoscopically. In cases with large leaks the endoscope can be passed through the leak into the surrounding collections. All fluids can be aspirated, and more importantly, necrotic debris, solid material and foreign bodies such as sutures can be extracted. Passing the endoscope through the leak used to be considered an absolute “faux pas”, but now I believe it should be considered an opportunity not to be missed. The aspiration and cleansing of the leak cavities by the endoscope is much more comprehensive and effective than radiologic-guided drainage and much less invasive than surgical drainage.

Endoscopy unit setting

I find it essential to perform all procedures with fluoroscopic guidance. Although theoretically possible, performing such procedures with only an endoscopic view severely narrows the assessment of the leak and its extent and does not allow confirmation of technique success. Another vital issue in the setting is the availability of different endoscopic solutions for leak closure as it is almost impossible to decide on which technique will be used prior to performing the endoscopy. The endoscopist has to know that these patients may only have one chance to undergo the endoscopic procedure, compromising this chance by using suboptimal equipment might actually make things worse for the patient. A proper inventory would include at least: Fully covered stents with lengths between 12 to 23 cm and diameters between 20 to 28 mm, flexible and rigid guidewires, through-the-scope (TTS) clips, over-the-scope (OTSC) clips of variable diameters, fibrin glue, histoacryl, balloon dilators 12 to 20 mm, cytology brushes, endoscopic scissors and an argon plasma coagulation (APC) unit. Availability of a small calibre (pediatric) endoscope is very useful in situations needing tight angulation or when passing through a small leak is necessary to aspirate collections. Double channel endoscopes with wide channels are also valuable in the presence of thick pus/necrotic tissue and the need for strong suction while using the instrument channel.

Expertise

Endoscopic management of GI leaks is highly demanding technically, but more importantly is the necessary expertise of managing these cases as a whole: decision making such as to intervene or not, when to intervene, choice of endoscopic methods, assessment of radiologic findings before and after endoscopic procedures, detailed knowledge of different post-surgical anatomies etc. Unless the endoscopist is actually a surgeon, the optimal situation is having a “hybrid” endoscopist/surgeon which can merge the different views and knowledge of each. This requires dedication and following a rather long learning curve. For example I have found it extremely usefull as an endoscopist to actually attend surgeries performed to drain patients with leaks. To be able to feel the friable tissue at the leak site, to feel the texture and consistency of an eroding band and thus knowing how to cut it endoscopically, to see the nature and extent of damage on the outside of the leak in the peritoneal cavity, have all drastically changed my assessment and understanding of leaks and their management.

Multidisciplinary approach

A team managing such cases should comprise at least: a surgeon experienced in managing GI leaks, the operating surgeon (his knowledge of what actually happened during the surgery could be invaluable to the endoscopist), a dedicated endoscopist, an intensivist, a nutritionist and an experienced radiologist. Psychiatric help is also frequently necessary as many of these patients suffer depression which may be severe, nutritional deficiencies may also be associated with neurological and psychiatric disturbances. A tight collaboration between all team members is vital, it is really hard to point out which one is more important for the survival of the patient.

Confirmation of leak closure

A normal oral contrast study is not sufficient evidence of an adequate sealing of the leak whether by a stent, clip or any other technique. It is the composite result of drain output, clinical status and contrast studies that leads to this conclusion. Therefore, surgical drains should be removed only after at least a week of normalization of all these parameters.

De-epithelialization

One of the main obstacles to leak healing is epithelialization of the leak site and tract. This occurs in chronic fistulas and can render any endoscopic therapy futile. De-epithelialization or “roughening” of the leak edges can be achieved by a cytology brush, standard biopsies or forceps, but our preferred method is APC. APC at low settings (30 W) can be easily applied to the leak edges and the tract walls.^14,15 Care should be taken because a large part of these tracts are composed of thin granulation tissue. In our practice we routinely de-epithelialize any leaks older than 2 weeks.

Antibiotics

Once a leak is suspected an aggressive approach including multiple antibiotics is essential. Our empirical regimen includes intravenous amikacin, vancomycin, and clindamycin, which is then changed according to culture results. We recommend the continuation of antibiotics for at least 2 weeks after presumed leak closure (cessation of fever and drain output and normalization infection markers such as C-reactive protein and leucocytic count). Leaks are associated with prolonged use of multiple antibiotics and a poor nutritional status, the perfect conditions for fungal infection. In our practice an antifungal (fluconazole) is empirically administered in patients receiving antibiotics for over 10 days especially if markers of infection do not completely resolve.

Stents

The purpose of stenting is to divert the route of alimentation, saliva and gastric/intestinal secretions away from the site of leak thus preventing infection and creating a more stable environment for tissue healing. Initially most studies addressed the use of self-expandable plastic stents (SEPS), but these stents had a high rate of migration (up to 50%), an arguably cumbersome assembly technique and a large calibre delivery system.^16–18 Recently self-expandable metallic stents (SEMS) have become the first choice in the treatment of leaks. Uncovered metallic stents are not used as tissue hyperplasia develops rapidly and the stent becomes embedded within the mucosa rendering it unextractable. Partially covered SEMS (PCSEMS) have a small uncovered segment (about 1–2 cm in length) which renders them less liable to migration, but this segment can also become completely embedded within the hyperplastic mucosal tissue rendering stent extraction very difficult if at all possible. Fully covered SEMS (FCSEMS) are currently the preferred option by many experts as their extraction is simple, but this comes on the expense of a higher rate of migration.¹⁷ In a pooled analysis by van Halsema and van Hooft¹⁹ the migration rate was 21.8% (53/243) and 10.6% (23/218) for FCSEMS and PC-SEMS, respectively.

Stents are probably the most studied endoscopic technique in leak management and currently have the most robust evidence of efficacy. However, most studies have pooled results of post-surgical leaks and perforations despite their different nature, many have also pooled different stents together.^20–23 The largest and most recent studies including only post-surgical leaks or providing subgroup analysis of post-surgical leaks are listed in Table 1.^18,23–27 These studies show a pooled success rate of 76%, a very encouraging result putting in consideration the low rate of serious adverse events, and a great improvement in comparison to conventional non-endoscopic approaches. It is to be noted though that many of these patients eventually healed after the insertion of several stents, thus the failure of healing after removing the first stent should not be considered an endoscopic failure, further attempts should be sought unless there is frank deterioration of the patient’s general condition and need for urgent surgical intervention.

Factors associated with non-healing seem to be a larger leak size and delayed endoscopic intervention.^24,28 In the study by El Hajj et al,²⁴ 54 patients who underwent stent therapy were grouped into responders (40 patients) and non-responders (14 patients). Responders had a significantly shorter delay to endoscopic intervention than non-responders (9 vs 23 days;P = 0.003). A smaller leak size in responders was the only other significant difference between the 2 groups.²⁴ Similarly, the study by Bège et al²⁸ included 27 patients with post-bariatric surgery leaks who underwent stent therapy, the group who had earlier endoscopic intervention (< 30 days from time of diagnosis, n = 13) healed in a significantly shorter time (95 vs 210 days; P = 0.005) and required fewer endoscopic sessions (average of 3.15 vs 5.57 sessions) than the group presenting > 30 days after diagnosis of leakage (n = 14).²⁸ Surgeons used to be (and many still are) sceptical about early endoscopic interventions. This is partly due to the fear of damage to fragile staple lines by endoscopic manoeuvres or insufflation. There is also always this hope that the leak would heal spontaneously with watchful waiting. It is now obvious that the risks of early endoscopic intervention are minimal, especially in expert hands and with the use of CO₂ insufflation. Several experts now opt for an early endoscopy as the first step when a leak is suspected even before confirmation by contrast studies.²³

Bariatric surgery leaks in particular seem to respond well to stent therapy. A systematic review by Puli et al²⁹ included 67 patients diagnosed with post-bariatric surgery leaks who underwent treatment by expandable stents (plastic and metallic). The success rate was 88% with a migration rate of 17%. Some stents have been designed with specifications tailored to bariatric surgery leaks such as the ultra-long, wide and flexible Mega stent (Taewoong Medical, Gimpo, Korea) for sleeve gastrectomy (Fig. 1) and the segmented Beta stent for RYGB (Taewoong Medical). So far no large trials have confirmed superiority of these stents to conventional ones.

No consensus is yet present for the duration of stenting. In the study by van Heel et al³⁰ extraction of all stents within 6 weeks of insertion was successful without complications while 50% of those with an indwelling time of 6 to 84 weeks had major complications such as bleeding and stent fracture. Most experts agree that 6 weeks is usually appropriate. Shorter periods may be sufficient for small non-infected leaks.

The Achilles tendon of stents is the high rate of migration, studies assessing FCSEMS report migration in 20% to 84%, with the largest series reporting 53%.^{18,23,25–28,31–33} Migration necessitates re-intervention for stent positioning or exchange, and may be associated with serious adverse events such as perforation or necessity for surgical extraction.^23,25,26 Many attempts have been made to overcome this problem. The use of PCSEMS and a “stent-in-stent technique” for extraction has been proposed by some authors.³⁴ The technique entails the insertion of a PCSEMS, and after 4 to 6 weeks a FCSEMS or SEPS of the same size and length is inserted inside the older one for a week. The second stent compresses against the hyperplastic tissue causing pressure necrosis, both stents are then extracted together in a third session. In their experience of 76 patients, Swinnen et al³⁴ reported a healing rate of 84%, migration occurred in 11%, and failure to extract the stents occurred in 4% of patients. The results are tempting but caveats still exist with this technique; firstly, the technique by default adds a second stent and a third endoscopic procedure (for insertion of the SEPS inside the first stent), adding to the costs and risks. Secondly, tissue hyperplasia inside the PCSEMS is associated with luminal obstruction and significant dysphagia necessitating repeated dilatations or insertion of other stents in about 20% of cases. Thirdly, failures do still occur even after inserting a stent inside a stent, resulting in an irremovable stent necessitating a highly morbid surgical extraction. Some endoscopists attempted using TTS clips to attach the upper edge of the stent to the luminal wall, this technique has been repeatedly shown to be ineffective.^28,35,36 Stent modifications have been also attempted with data still lacking to confirm their effectiveness. The Choo stent (M.I. Tech, Seoul, Korea) has a proximal lasso that is brought out of the nostril and looped around the ear lobe thus preventing migration. The recently developed Mega stents (Taewoong Medical) have an ultra large diameter (28 mm shaft, 36 mm flared end) and significantly more flexibility thus proposedly reducing migration. Another technique that we have developed is the use of overlapping “stapled” stents. In this technique the migrated stent is left just a few centimetres below the leak and another stent is inserted inside the migrating stent in an overlapping manner with the upper edge lying above the leak site. The 2 stents are then fixed in this position by 3 endoscopic clips fixing the lower edge of the new stent to the shaft of the migrated stent (Fig. 2), this allows reaching stent lengths not attainable by any available stent. We have used this technique so far in 3 patients with gastroesophageal junction leaks, all were successful.

Complications have been reported with the use of stents including perforation, stricture formation, intolerance, bleeding and tracheal compression.¹⁹ The risk of serious adverse events is low but significant (3%–7%), requiring vigilance throughout the period of stenting, repeated contrast studies and frequent clinical assessments.^19,23–28 With the exception of long stent indwelling time (> 6 weeks), no analysis so far has deduced other risk factors associated with the occurrence of complications.

Endoscopic clips

TTS clips have been greatly successful in the management of GI bleeding. Many earlier reports have initially shown successful use of clips in sealing post-surgical leaks.^37–39 These reports, however, were mostly case reports or very small case series and were probably strongly affected by publication selection bias. TTS clips are hampered by their weak superficial grasp attaching only to the mucosal layer, thus not providing sufficient tissue apposition. These clips also need sufficient space for deployment, a condition seldom present at anastomotic sites in the immediate post-operative period. Recently, the OTSC clip has been developed (Ovesco, Tubingen, Germany) providing a clip that has a strong grasp including more than the mucosal layer (Fig. 3). The OTSC is made of super-elastic nitinol, it is loaded over a cap and deployed in a fashion similar to that of endoscopic bands. The tissue around the leak is drawn into the cap by suction and then the clip is released. Two accessories are available to help draw more tissue into the cap: the tissue anchor and the twin-grasper (Ovesco). Apart from being large in diameter, the elastic nature of the OTSC means that there is continuous pressure on the grasped tissue, so even when edema resolves and the grasped tissue shrinks, sufficient pressure will still be present to maintain tissue apposition.

After a few initial small case series, recently larger studies have emerged all reporting encouraging results, however, all still retrospective in nature.^40–42 Haito-Chavez et al⁴³ recently reported the largest series of postoperative leaks treated by OTSC clips, their study included 32 patients, 15 patients (47%) presenting after colorectal surgery and 17 patients (53%) after upper GI surgeries. One clip was used in 25 patients while 2 clips were used in 2 patients. Technical success was achieved in 90% of patients while long-term healing occurred in 22 patients (73%).⁴³ Mercky et al⁴⁴ reported their similar experience with OTSC in 30 patients, 27 patients of which had postsurgical leaks while 2 patients were post-mucosectomy and 1 patient had a Crohn’s rectovaginal fistula. Technical success was eventually achieved in all cases (4 patients needed 2 attempts). Successful healing after the first attempt alone occurred in 53% (16/30) while eventual overall success (after additional endoscopic techniques) was 70% (21/30).⁴⁴

Stents vs clips?

In the study by Farnik et al,⁴⁵ the authors performed a retrospective comparison of OTSC and stents in the management of leaks and perforations. In the postoperative leaks subgroup OTSC’s were successful in 56% (9/16) in comparison to 39% (22/57) in patients treated with stents. Post-interventional leaks (endoscopic perforations) healed significantly better (OTSC 87% [13/15], stents 44% [4/9]). The authors, however, reasonably did not conclude the superiority of OTSC’s to stents. They found that patients in the stent group had significantly larger leaks (12.6 vs 7.1 mm), more frequently had local abscesses (71% vs 44%) and had a significantly longer intensive care unit stay (7 vs 0 days).⁴⁵ I highly concur with the authors’ conclusion that the two techniques are complementary rather than competing, patients with larger leaks or evidence of infection may benefit better with stents, while OTSC’s might be considered as a first choice in those with smaller non-infected leaks and a favourable position for clip application.

At our institution we have adopted a strategy for bariatric surgery leaks in particular favouring the combination of the 2 techniques whenever possible. When approaching a postoperative leak we believe the optimal situation would be a applying a clip to seal the leak and simultaneously inserting a stent (Fig. 4). The stent offers 3 main advantages: firstly, provides additional sealing to the leak site thus promoting healing while allowing early enteral nutrition. Secondly, it eliminates any distal stenosis to the leak, whether a frank stricture or subtle stenosis (as in sleeve gastrectomies). This distal stenosis leads to a reservoir effect at the site of leakage impairing its healing. Thirdly, presence of the stent will reduce the risk of development of stenosis at the site of leakage, a common complication occurring after leak healing. Applying a clip, however, should not be attempted if there is any evidence of infection or if the leak is not in a favourable position for clip application (tangential or requiring acute retroflexion of endoscope). So in our practice a stent may be inserted alone but application of a clip alone is much less common, only in cases where stent application is not suitable (e.g., colonic leaks) and under the condition of healthy leak walls and absence of any evidence of infection. These conditions are seldom present in postoperative leaks but typical of post-endoscopic perforations. Our recently published series included 22 patients with post-bariatric surgery leaks treated with Mega stents with or without OTSC clips. Our algorithm includes putting a stent in all patients in the first session with or without an OTSC sealing of the leak, in further sessions however clips are favoured as infection has resolved and tissues are less friable. Clips were applied in 12 patients (55%); 5 simultaneously with stents and 7 after stent removal. An average of 1.4 stents and 2.8 endoscopic procedures were required per patient. Primary success after 1 endoscopic procedure occurred in 59% while long-term success was achieved in 18 patients (82%). Of the 4 failures 2 had persistent leaks only detected radiologically but with no symptoms or clinical sequelae, including those would raise the success rate to 91% (20/22).⁴⁶

Endoscopic internal drainage

In some patients treated by stents, a leak cavity persists after removal of the stent, this persistence of leakage can be seen endoscopically and confirmed by contrast injection. It has been noted; however, that some of these patients do not suffer clinically from these persistent cavities, this has been noted also by surgeons in some patients undergoing conservative management.^22,27 It seems that granulation tissue eventually forms a sealed cavity communicating with the lumen but not with the abdominal cavity. Based on this observation Donatelli et al⁴⁷ have devised an approach based on promoting the formation of such cavities. The technique involves inserting 1 or 2 double pigtail stents through the leak and exchanging them regularly over several weeks. These stents allow a constant internal drainage of any leaked fluid and eventually a sealed granulation tissue cavity is formed. Nasojejunal tubes are usually inserted to allow enteral feeding until healing occurs.

In their ongoing series of 67 patients, Donatelli et al⁴⁸ reported a success rate of 74% (50/67) so far. At the time of publication 9 patients were still undergoing treatment, there were 5 documented failures and 2 mortalities, this suggests that the eventual success rate will fall anywhere between 74% and 92%. The average number of endoscopies was 3.14 and the time to healing was 57 days. The technique seems to have several advantages including low cost and absence of stent-induced complications. The disadvantages include the need for repeated stent exchanges, use of nasojejunal feeding for several weeks and possibly a high incidence of stricture formation as 6 patients (9%) in this series developed stenosis. This may be one advantage supporting the use of SEMS as they simultaneously dilate any stenosis while allowing the leak to heal.

Suturing devices

Since the advent of endoscopy, a device that can perform surgical sutures endoscopically has been regarded as the holy grail of endoscopists. The Overstitch device (Apollo Endosurgery, Austin, TX, USA) is currently the closest device to achieve this target (Fig. 5). It is a disposable device able to place both absorbable (2-0 and 3-0 polydioxanone) and non-absorbable (2-0 and 3-0 polypropylene) sutures. The system allows continuous or interrupted stitches to be made of various lengths.⁴⁹ An in vivo study on human colons has showed safe subserosal positioning of all the sutures with no risk of injury to surrounding organs.⁵⁰

Two initial case reports concluded success of the Overstitch in healing a chronic esophagopleural and a gastrocutaneous fistula.^49,51 Cai et al⁵² reported successful closure of 2 sleeve gastrectomy leaks using the Overstitch; however, one had a stent inserted simultaneously. Recently, Sharaiha et al⁵³ reported results of a large multicentre series, indications for Overstitch were as follows: stent anchorage (n = 47), fistulas (n = 40), leaks (n = 15), and perforations (n = 20). Technical success was achieved in 97.5%. Clinical success was 91.4% in stent anchorage, 93% in perforations, 80% in fistulas, but only 27% in anastomotic leak closure.⁵³ The results highlight the different nature of iatrogenic endoscopic perofrations and postsurgical leaks, moreover, they show how poorly sutures in particular perform in postsurgical leaks.

For postsurgical leaks, we as endoscopists should reach the conclusion reached by surgeons decades ago that simply putting a good suture on a surgical leak is not the solution. With infected, friable, devascularized leak edges it might be better to opt for stenting or the use of the elastic OTSC’s with their constant compression or use endoscopic internal drainage, rather than putting a perfectly placed standard surgeon’s suture. Suturing devices also remain to date costly and very technically demanding limiting their widespread use.

Vacuum therapy

Vacuum assisted closure (VAC) is a well-established technique to promote healing of superficial wounds.⁵⁴ A vacuum-sealed sponge is inserted in the wound cavity and connected to a suction device. This allows the constant removal of infected fluids, promotes rapid resolution of tissue edema and improves the micro-circulation in that area, leading to the rapid formation of granulation tissue.⁵⁴ The sponge has to be replaced every few days until wound healing occurs. The endoscopic version of this technique (EVAC) was probably first well-described by Nagell and Holte,⁵⁵ who described their experience with 4 patients with rectal anastomotic leaks. A sponge was inserted into the leak cavities and was connected to a suction device through the anus. Leaks healed by forming a granulation tissue lined cavity in all 4 patients at a mean of 51 days.⁵⁵ Larger series followed such as the one by Weidenhagen et al⁵⁶ where 29 patients with leaks after anterior rectal resection were treated by EVAC. Success was achieved in 28 patients (96%) with a mean of 11.4 endoscopies per patient. It should be noted, however, that 21 patients (72%) had a protective stoma.⁵⁶ The procedure seems to be well tolerated by patients, in the study by Riss et al⁵⁷ 9 patients treated by EVAC for rectal anastomotic leaks had a median visual analogue pain score of 3 (0–6) on a scale of up to 10. Six out of the 9 patients said they were willing to be retreated with EVAC.

The technique has been recently used for post-esophagectomy leaks. Wedemeyer et al⁵⁸ treated 8 patients with post-esophagectomy leaks using EVAC. Leaks healed in 7 patients (88%) in a mean of 23 days and 7 endoscopies per patient. Recent studies have compared EVAC to stent therapy for the management if leaks.^20,59 In one study, 39 patients were treated by SEMS or SEPS while 32 were treated by EVAC. Leak healing occurred in 84% of EVAC group vs 54% in the stent group with no difference in hospitalization or mortality. Strictures were also seen more commonly with stents (28% vs 9%; P < 0.05).²⁰

The results are intriguing but selection bias for the method of treatment has to be considered as these patients were not randomized, the initial surgical procedures in both groups were also significantly different. The results of EVAC are certainly impressive but the enthusiasm has to be taken with caution. Complications have been reported such as fistula formation with small intestinal loops and the aorta.⁶⁰ The major drawback seems to be the need for numerous endoscopic procedures and probably the need for prolonged hospitalization. The technique also by default needs a cavity large enough to accommodate the sponge and prolonged TPN is usually necessary for those with esophageal leaks. Nevertheless, the technique is a very welcome addition to the armamentarium of techniques for leak closure, and may be especially useful in large refractory leaks.

Tissue sealants

The use of tissue sealants and adhesives is a well-established surgical concept.⁶¹ Fibrin glue (Tissucol Duo, Baxter, Germany) is composed of fibrinogen and thrombin which when injected simultaneously into the fistulous tract form a three-dimensional gel within 10 to 60 seconds, an “acellular clot”. This gel acts as a scaffold for regenerating tissues thus promoting healing.⁶² Some authors have advocated the injection of fibrin submucosaly at the leak edges rather than inside the fistula.⁶³ N-butyl-2-cyanoacrylate (Histoacryl; B. Braun, Tuttlingen, Germany) on the other hand is a pure adhesive, after contact with liquids it almost instantly solidifies and occludes the fistulous cavity.⁶⁴ Cyanoacrylate is also irritant to the surrounding tissues promoting an inflammatory reaction which is thought to improve vascularity and healing.

Eleftheriadis et al⁶⁵ published one of the earliest reports on endoscopic application of fibrin glue to treat post-operative enterocutaneous fistulas. All seven patients included were treated successfully. Several case reports and a few small case series then followed with success rates ranging between 50% to 86%.^64,66,67 Recently, some authors combined fibrin glue with a Vicryl mesh (Ethicon, Hamburg, Germany).⁶⁸ The mesh is inserted inside the leak cavity and then fibrin glue is injected onto the mesh and submucosaly at the leak edges. Using this method Böhm et al⁶⁹ reported success in 13 out of 15 patients in 1 to 4 sessions.

In general, sealants and adhesives may be usefull in small leaks with long tracts that can accommodate a significant amount of the sealant. Larger leaks and acute perforations with no tracts will usually not be suitable. For post-surgical leaks the main role of sealants and glue is currently that of an adjunctive therapy to clips or stents.

Postsurgical GI leaks are serious complications with significant morbidity and mortality. Several endoscopic techniques have shown efficacy and a favourable safety profile. Nevertheless, to date, most evidence comes from retrospective case series. Choice of technique depends on several factors including site, size and duration of leak, operator familiarity with the techniques and availability. Different endoscopic techniques should be considered complementary or interchangeable rather than competing. The combination of techniques is frequently effective, endoscopists should not be afraid to be innovative and to combine the techniques when appropriate, assuming the operator has sufficient experience in managing such cases.