Endoscopic closure versus surgical revision in the management of gastrogastric fistula after Roux-en-Y gastric bypass.

This was a retrospective matched cohort study of patients with GGF who underwent endoscopic closure (ENDO) or surgical revision (SURG) at 2 tertiary referral centers. Patients were identified through a systematic patient search by use of a large research patient data registry to evaluate patients who received treatment for GGF that was deemed technically successful at the time of the intervention. Institutional review board approval was obtained for a retrospective review of data used in this study (approval number 2003P-001597, renewed approval on 5/8/2020). ENDO patients were matched 1:1 to SURG patients on the basis of age within 5 years, sex, body mass index within 5 kg/m 2 , and percentage of weight regain from that lost after initial RYGB within 10%. Patient demographics, GGF size, procedural details, symptoms related to GGF, and treatment-related adverse events were collected. Symptoms of onset or recurrence of diabetes, acid reflux, abdominal pain, and weight parameters were extracted. Inclusion criteria included adult patients (>18 years) with the diagnosis of GGF through prior imaging (upper GI series, CT of the abdomen) or EGD and having undergone GGF repair (endoscopic or surgical). Patients were excluded if they underwent endoscopic or surgical treatment for an alternative diagnosis to GGF. A comparison of symptom improvement between ENDO and SURG after GGF treatment at 12 months was evaluated. Specifically, resolution of diabetes, acid reflux, abdominal pain, and weight parameters were compared. Weight measurements included weight regain (defined as having gained >15% of maximum post-RYGB weight loss), total weight change (lb) and percentage of total weight loss between GGF treatment and end of the follow-up period. Additional comparisons of overall treatment-related adverse event rates and serious (severe) treatment-related adverse event rates were made between groups. Treatment-related adverse events were recorded per validated reporting standards. For treatment-related adverse events associated with endoscopy, the American Society for Gastrointestinal Endoscopy lexicon was referenced. 13 This reporting system was developed in collaboration with representatives of the National Surgical Quality Improvement Program, which has been endorsed by the American College of Surgeons as a reliable surgical adverse event reporting platform. 14 , 15 Standard statistical analyses were performed, including the Fisher exact test used to compare symptom profiles and treatment-related adverse event rates (categoric variables) and either Student t test or Wilcoxon rank sum test used for weight profile comparisons (continuous variables) based on data normality. A kurtosis value <1 was determined to show adequate normality for comparisons based on sample sizes. Comparisons were considered statistically significant when P values were <.05. All statistical analyses were performed with SAS software, version 9.4 (SAS Institute, Cary, NC, USA).

A total of 90 RYGB patients with GGF (45 ENDO, 45 matched SURG) were included ( Table 1 ). Baseline characteristics were similar between groups. Specifically, the mean (SD) age was 49.7 (10.3) years in the ENDO group compared with 49.2 (10.2) years in the SURG group, and both groups contained 37 (82.2%) women. The overall average GGF size was 15 ± 9 mm (14.6 ± 9.4 mm in the ENDO group and 15.9 ± 8.4 mm in the SURG group). The average time between RYGB and GGF treatment was 8.9 ± 5.3 years (9.7 ± 5.7 years in the ENDO group and 8.1 ± 4.8 years in the SURG group). GGF symptoms included new or recurrent onset of diabetes in 11 (25.6%) and 16 (36.4%), acid reflux in 34 (79.1%) and 33 (75%), abdominal pain in 23 (53.5%) and 33 (75%), and weight regain in 37 (82.2%) and 35 (77.8%) in the ENDO and SURG groups, respectively. Table 1 Baseline cohort characteristics of RYGB patients who underwent endoscopic closure or surgical revision of GGF Characteristics Endoscopy (n = 45) Surgery (n = 45) Sex, female, n (%) 37 (82.2) 37 (82.2) Age, y, mean (SD) 49.7 (10.3) 49.2 (10.2) Duration between RYGB and revision, y, mean (SD) 9.7 (5.7) 8.1 (4.8) Pre-RYGB weight, lb, mean (SD) 307.4 (63.7) 299.7 (54.0) Nadir weight, lb, mean (SD) 183.5 (45.4) 171.7 (44.7) Weight at surgical revision, lb, mean (SD) 227.2 (55.1) 222.0 (56.6) BMI at surgical revision, kg/m 2 , mean (SD) 36.9 (7.5) 36.8 (8.0) Weight regain (% of maximal weight loss), mean (SD) 36.4 (22.6) 39.6 (25.5) GGF size, mm, mean (SD) 14.6 (9.4) 15.9 (8.4) Prerevision symptoms, n (%)  Diabetes onset 11 (25.6%) 16 (36.4%)  Acid reflux 34 (79.1%) 33 (75%)  Abdominal pain 23 (53.5%) 33 (75%)  Weight regain 37 (82.2%) 35 (77.8%) BMI , Body mass index; GGF , gastrogastric fistula; RYGB , Roux-en-Y gastric bypass. Baseline cohort characteristics of RYGB patients who underwent endoscopic closure or surgical revision of GGF BMI , Body mass index; GGF , gastrogastric fistula; RYGB , Roux-en-Y gastric bypass. Endoscopic closure methods ( Table 2 ) included argon plasma coagulation (APC) with endoscopic suturing (n = 18), APC with endoscopic plication (n = 7), endoscopic submucosal dissection with endoscopic suturing (n = 6), APC with endoscopic plication, clips, and fibrin glue (n = 2), APC with endoscopic plication and clips (n = 2), endoscopic suturing alone (n = 2), APC with endoscopic plication and glue (n = 2), clips alone (n = 2), APC with endoscopic submucosal dissection and endoscopic suturing (n = 1), APC and clip (n = 1), clips and glue (n = 1), and stent placement alone (n = 1). Endoscopic suturing was performed with the Apollo Overstitch device (Apollo Endosurgery, Austin, Tex, USA) in an interrupted, running, or purse-string fashion depending on fistula size and location. Endoscopic plication was performed with the Bard Endocinch device (Bard, Warwick, RI, USA). Surgical revision methods included laparoscopic GGF resection (n = 31) or open resection of GGF (n = 14). Table 2 Endoscopic techniques used for fistula closure Method of endoscopic treatment N = 45 APC, suture 18 APC, endoscopic plication 7 ESD, suture 6 APC, endoscopic plication, clips, and fibrin glue 2 APC, endoscopic plication and clips 2 Endoscopic suturing alone 2 APC, endoscopic plication, glue 2 Clips alone 2 APC, ESD, suture 1 APC, clips 1 Clips, glue 1 Stent to cover fistula 1 APC , Argon plasma coagulation; ESD , endoscopic submucosal dissection. Endoscopic suturing was performed with the Apollo Overstitch device (Apollo Endosurgery, Austin, Tex, USA). Endoscopic plication was performed with the Bard Endocinch device (Bard, Warwick, RI, USA). Fibrin glue was used for glue. Endoscopic techniques used for fistula closure APC , Argon plasma coagulation; ESD , endoscopic submucosal dissection. Endoscopic suturing was performed with the Apollo Overstitch device (Apollo Endosurgery, Austin, Tex, USA). Endoscopic plication was performed with the Bard Endocinch device (Bard, Warwick, RI, USA). Fibrin glue was used for glue. At 12 months, both groups experienced similar resolution of diabetes mellitus and improvement in GERD; however, endoscopic revision was associated with more frequent improvement in abdominal pain ( Table 3 ). Specifically, resolution of diabetes (among those in whom diabetes was present) occurred in 1 (9.1%) and 1 (6.3%) of the ENDO and SURG groups, respectively ( P  = .0), and reflux symptoms (among those in whom they were present) were improved in 7 (20.6%) and 2 (6.1%) in the ENDO and SURG groups, respectively ( P  = .15). Abdominal pain improved in 12 (52.2%) ENDO and 5 (15.2%) SURG patients at 12 months ( P  = .007). Table 3 Presence of symptoms associated with gastrogastric fistula at 1 year Postprocedure characteristics Endoscopy (n = 45) Surgery (n = 45) P value Diabetes resolution, ∗ n (%) 1 (9.1) 1 (6.3) 1.0 GERD improvement, ∗ n (%) 7 (20.6) 2 (6.1%) .15 Abdominal pain improvement, ∗ n (%) 12 (52.2) 5 (15.2) .007 6-month weight change, lb, mean (SD) 2.8 (10.4) 10.5 (7.8) .0005 6-month %TWL, † % 0.59 (5.7) 5.5 (3.9) .0002 12-month weight change, † lb 4.2 (18.8) 14.1 (14.9) .02 12-month %TWL, † % 1.9 (6.9) 6.2 (5.2) .007 Presented as mean (standard deviation) for normal variables. %TWL , Percentage of total weight loss. ∗ Calculated at 12 months using number of patients with initial symptoms in each cohort as shown in Table 1 . † Median (interquartile range) for nonnormal variables. Fisher exact test performed for categoric variables and Wilcoxon rank sum test performed for nonnormal continuous variables (weight change). Presence of symptoms associated with gastrogastric fistula at 1 year Presented as mean (standard deviation) for normal variables. %TWL , Percentage of total weight loss. Calculated at 12 months using number of patients with initial symptoms in each cohort as shown in Table 1 . Median (interquartile range) for nonnormal variables. Fisher exact test performed for categoric variables and Wilcoxon rank sum test performed for nonnormal continuous variables (weight change). At 6 months, 35 (78%) of the ENDO group and 40 (89%) of the SURG group stopped gaining weight ( P  = .23), with the ENDO and SURG groups experiencing 0.59% and 5.5% TWL ( P  = .0002), respectively ( Table 3 ). At 12 months, 59% of the ENDO group and 93% of the SURG group stopped gaining weight ( P  = .007), with the ENDO and SURG groups experiencing 1.9% and 6.2% TWL ( P  = .007), respectively. Treatment-related adverse events occurred in 4 (8.9%) ENDO and 16 (35.6%) SURG patients ( P  = .005) ( Table 4 ). Among treatment-related adverse events, 0 and 8 (17.8%) were serious (severe) in the ENDO and SURG groups, respectively ( P  = .006). Treatment-related severe adverse events in the SURG group included leak (4), abdominal hernia (3), abscess (3), severe abdominal pain (3), GI bleeding (1), small-bowel obstruction (1), and gastrojejunal anastomotic stricture (1). Among adverse events, the open surgical technique was associated with 3 abdominal hernias, 2 leaks, 1 severe abdominal pain, and 1 small-bowel obstruction. The remainder of adverse events in the surgical group occurred when the laparoscopic technique was used. Table 4 Treatment-related adverse events over 12-month follow-up duration after treatment for gastrogastric fistula Treatment-related adverse events Endoscopy (n = 45) Surgery (n = 45) P value Overall treatment-related adverse events, n (%) 4 (8.9) 16 (35.6) .005 Marginal ulceration 3 (6.7) 0  Abdominal pain 1 (2.2) 3 (6.7)  GI leak 0 4 (8.9)  Abscess 0 3 (6.7)  GI bleeding 0 1 (2.2)  Small-bowel obstruction 0 1 (2.2)  Gastrojejunal anastomotic stricture 0 1 (2.2)  Abdominal hernia 0 3 (6.7) Serious treatment-related adverse events, n (%) 0 8 (17.8) .006 Treatment-related adverse events over 12-month follow-up duration after treatment for gastrogastric fistula

In conclusion, endoscopic repair of GGF results in clinically and statistically fewer overall and serious treatment-related adverse events and greater resolution of abdominal pain when compared with surgical repair. By contrast, surgical revision appears to yield greater weight loss at 1 year.

The following authors disclosed financial relationships: R. Dolan: Consultant for Centaur Laboratories. P. Jirapinyo, Consultant for ERBE, Endogastric Solutions, and Spatz; research support from Apollo Endosurgery, Boston Scientific, Fractyl, GI Dynamics, and USGI Medical. C. Thompson: Consultant for and research support from Apollo Endosurgery, Boston Scientific, GI Dynamics, Olympus/Spiration, Fujifilm, and Lumendi; research support from Aspire Bariatrics, and ERBE; general partner in BlueFlame and Healthcare Venture Fund; consultant for Covidien/Medtronic; founder and consultant board member Entarasese Ltd, EnVision Endoscopy, and GI Windows; consultant and advisory board member for Fractyl and USGI Medical; member of and research support from USGI Medical. All other authors disclosed no financial relationships.

GGF is an uncommon late adverse event after RYGB; however, it is associated with undesirable symptoms of recurrence or onset of diabetes, abdominal pain, acid reflux, and weight regain. 7 Subsequently, repair through either an endoscopic or a surgical approach is often required. The present study is currently the largest study directly comparing the efficacy and safety of endoscopic with surgical treatment of GGF in RYGB patients. The endoscopic approach was associated with a similar rate of diabetes and GERD improvement, and a higher rate of abdominal pain improvement, but with a lower adverse event rate compared with surgical revision. The difference in abdominal pain may at least be in part due to differences in adverse event rates, particularly given that only severe treatment-related adverse events occurred within the surgical group. Additionally, the endoscopic approach avoids the need for de novo incisions that are required in the surgical approach, reducing additional potential pain sources. Treatment-related adverse events occurred 4-fold more frequently in the surgical repair group (n = 16, 35.6%) than in the endoscopic group (n = 4, 8.9%), and serious (severe) treatment-related adverse events occurred only in the surgery group. The observed adverse events differed between groups and were likely related to the procedural approach, which may have contributed to the magnitude of weight loss in either group, particularly the surgical group based on the character of treatment-associated adverse events. Marginal ulceration (3) and severe abdominal pain (1) were observed in the endoscopic repair group, which may be the result of localized tissue ischemia from tissue approximation. An increased rate of surgical adverse events after surgical repair is expected, given the invasive nature of the procedure in comparison, particularly when the open surgical approach is used. Leak (4), abdominal hernia (3), abscess (3), severe abdominal pain (3), GI bleeding (1), small-bowel obstruction (1), and gastrojejunal anastomotic stricture (1) as observed were similar to those in previous studies. 9 Furthermore, these treatment-related adverse events may have contributed to the greater amount of weight loss seen in the surgical group at 6 and 12 months. The magnitude of weight loss achieved over 1 year after surgical GGF repair in this study (14.1 lb; 6.4 kg) is also similar to that in prior studies evaluating surgical GGF repair (13.7 lb; 6.0 kg). 16 This may be partially attributable to fistula size (15 ± 9 mm) being more amenable to surgical repair, inasmuch as prior studies have demonstrated that GGF size <10 mm in diameter is associated with better outcomes after endoscopic repair. 11 A notable limitation historically with endoscopic intervention is instrument size, which is restricted to allow passage through the transoral approach. This requires the use of a variety of smaller-scale methods to achieve fistula repair, such as endoscopic clipping, fibrin glue, or gastroplication. 11 , 17 Notably, although the present study incorporated a variety of repair methods, including combinations of APC, endoscopic submucosal dissection, gastroplication, endoscopic suturing, clips, fibrin glue, and stent placement, there were no statistically significant differences in symptom resolution or weight changes when endoscopic suturing and other repair methods were stratified. This lends credence to the individualized approach that should be pursued when GGF is treated endoscopically, because the lesions can vary in size and location. Although an algorithmic approach toward treating GGF is desired, the variable characteristics (ie, size, number, tissue quality, and location) of GGFs require an individualized approach to treatment, particularly as new devices emerge. By contrast, surgical repair allows the introduction of larger instruments to permit GGF repair with staple closure, 18 excision, 19 RYGB/gastrojejunal anastomosis revision, or partial gastrectomy, 19 where ultimately fistula resection may provide a more definitive treatment. There are a few limitations to the present study. This was a retrospective cohort study; therefore, although patient characteristics were controlled for, residual confounding remains possible. The patient population was also limited to 2 large tertiary referral centers within a single region, limiting generalizability of the results. However, this likely reflects routine practice because referral to expert centers is common. Additionally, longer duration of follow-up care would provide greater insight into the durability of GGF treatment in either group, most notably that of weight trends, given that GGF often leads to weight gain.