Esophagogastroduodenoscopy- associated gastrointestinal perforations: A single-center experience Amit Merchea, MD,a Daniel C. Cullinane, MD,a Mark D. Sawyer, MD,a Corey W. Iqbal, MD,a Todd H. Baron, MD,b Dennis Wigle, MD,a Michael G. Sarr, MD,a and Martin D. Zielinski, MD,a Rochester, MN Background. Esophagogastroduodenoscopy (EGD) is commonly used in the diagnosis and treatment of gastrointestinal (GI) disorders. Our aim was to define the risk of perforation associated with EGD and identify patients who required operative intervention. Methods. We retrospectively reviewed 72 patients from our institution plus 5 transferred patients who sustained EGD-associated perforations from January 1996 through July 2008. Percutaneous endoscopic gastrostomy, endoscopic ultrasonography, endoscopic retrograde cholangiopancreatography, transtho- racic echocardiography, and concurrent colonoscopy procedures were excluded. Results. Perforations in 72 of 217,507 EGD procedures were identified (incidence, 0.033%); 124,844 EGDs included an interventional procedure and 92,663 were examination only. The incidence of perforation was similar whether an interventional procedure was performed or not (0.040% vs 0.029%; P = .181). The esophagus was injured most commonly (51%), followed by the duodenum (32%), jejunum (6%), stomach (3%), and common bile duct (3%). Overall mortality after perfora- tion was 17% with a morbidity rate of 40%. Thirty-eight patients (49%) were initially treated non- operatively, 7 of whom (18%) failed nonoperative management. The only factors we could determine that were associated with failure were free fluid or contrast extravasation on computed tomography (75% vs 23% [P < .005] and 33% vs 0% [P = .047], respectively). The morbidity of failures was equivalent to those who underwent initial operative management (63% vs 61%; P = .917), with mortality seeming to be greater (43% vs 21%; P = .09). Conclusion. EGD is safe in the majority of patients; however, iatrogenic perforation is associated with considerable morbidity and mortality. Nonoperative management of GI perforation can be successful if there is no evidence of contrast extravasation or free fluid on radiographic studies. If nonoperative management fails, the outcomes may be worse than those treated initially with operative repair. (Surgery 2010;148:876-82.) From the Departments of Surgerya and Medicine,b Mayo Clinic, Rochester, MN ESOPHAGOGASTRODUODENOSCOPY (EGD) is a widely esophagus has previously been reported at our in- used tool in the diagnosis and treatment of upper stitution to range between 0.2% and 0.4%.3,4 As gastrointestinal (GI) diseases. The utility of this the use of rigid endoscopy waned and flexible en- procedure has seen a steady increase in its use, doscopy became more pervasive, the perforation and while considered a safe procedure, complica- rate improved to 0.008--0.11%.1,2,5-8 Although tions can occur.1,2 The perforation rate of the rare, perforation secondary to EGD is associated with substantive morbidity.2,9,10 Patients presenting after perforation with peritonitis or septic shock Presented in part at the Central Surgical Association Annual Meeting, March 13, 2010, Chicago, Illinois. would generally be managed operatively; however, there may be a subset of patients in whom a non- Reprint requests: Martin D. Zielinski, MD, Division of Trauma, Critical Care and General Surgery, The Mayo Clinic, 200 First operative approach may be employed. Street SW, Rochester, MN 55905. E-mail: zielinski.martin@ We aimed to define the risk of GI perforation mayo.edu. associated with EGD at a single, large-volume, 0039-6060/$ - see front matter tertiary care center and to identify the risk factors Ó 2010 Mosby, Inc. All rights reserved. that, if present, would require operative interven- doi:10.1016/j.surg.2010.07.010 tion. We hypothesized that the incidence of 876 SURGERY Surgery Merchea et al 877 Volume 148, Number 4 Procedures Performed (January 1996 - December 2007) Table I. Distribution of EGD procedures and 21500 incidence of perforation, January 1996--July 2008 19500 Incidence of 17500 Procedure Number perforation (%) 15500 Total EGDs performed 217,507 0.03 13500 Examination only 92,663 0.03 11500 Interventional procedure 124,844 0.04 Biopsy 108,053 9500 Dilatation 10,455 7500 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Coagulation/sclerotherapy 6,336 Year EGD, Esophagogastroduodenoscopy. Figure. Number of EGD procedures performed, January 1996--December 2007. Table II. Distribution of perforations by location EGD-associated GI perforation was greater if an within the gastrointestinal tract, January 1996--July interventional procedure was performed (biopsy, 2008 dilatation, coagulation, or sclerotherapy) and Proportion that patients without contrast extravasation on Location n of total (%) radiographic studies can be safely managed Esophagus 39 51 nonoperatively. Cervical 9 Proximal thoracic 3 METHODS Distal thoracic 15 Institutional review board approval was ob- Gastroesophageal junction 10 tained. The procedure database at our institution Unknown 2 was searched to identify all patients undergoing Stomach 2 3 EGD from January 1996 through July 2008 exclu- Duodenum 25 32 Jejunum 5 6 sive of percutaneous endoscopic gastrostomy, Common bile duct 2 3 endoscopic ultrasonography, endoscopic retro- Unknown 4 5 grade cholangiopancreatography, transthoracic Total 77 100 echocardiography, and concurrent colonoscopy procedures. This subset was further narrowed down by 2 methods: (1) all patients diagnosed Indications for exploration in those patients man- with ‘‘perforation’’ within 7 days of the EGD to aged nonoperatively included worsening clinical capture nonoperative management; and (2) those status, sepsis, and increased pain. Nonoperative with a perforation who underwent laparotomy, failure was defined as operative intervention or laparoscopy, thoracotomy, thoracoscopy, or neck death from an EGD-associated perforation compli- exploration within the same time frame. The cation. Overall survival was calculated from the database search yielded 77 patients with iatrogenic date of diagnosis of perforation to the date of perforation. Data were collected regarding indica- death from all causes, including postoperative tions, procedures performed, diagnostic modali- deaths. Univariate analysis of categorical variables ties, management, and overall outcomes. was performed using Pearson’s Chi-square test or Endoscopy reports were reviewed to determine Fisher’s exact test, as indicated. Continuous varia- procedural details, difficulty, complicated anatomy bles were analyzed by Student t test or Wilcoxon (from surgical alteration or abnormal anatomic rank sum test and are presented as mean or me- variant), and extent of intervention. An interven- dian as indicated (range). All tests were 2 sided tion was defined as any procedure in addition to with significance determined at P < .05. Statistical simple EGD and included dilatation, coagulation, analysis was performed utilizing JMP (Version 7; sclerotherapy, or any manipulation of normal tis- SAS Institute Inc., Cary, NC). sue planes (ie, biopsy). Postperforation courses were reviewed to determine outcomes. Nonopera- RESULTS tive management consisted of nil per os status, We performed 217,507 EGD procedures be- broad-spectrum intravenous antibiotics, and tween January 1996 and July 2008. The number percutaneous drainage of abscesses if present. of EGDs performed increased steadily from 12,183 878 Merchea et al Surgery October 2010 Table III. Complication by location of perforation within the GI tract, January 1996--July 2008 Wound Anastomotic EC MI/ Anoxic Death Pneumonia infection Abscess ARF leak fistula PE arrhythmia Obstruction brain injury Other Esophagus 4 4 1 1 0 1 0 0 2 0 1 2 Stomach 0 0 1 0 0 0 0 0 0 0 0 0 Duodenum 7 1 1 1 3 1 2 1 0 1 0 2 Jejunum 1 1 1 0 0 0 0 0 0 1 0 0 Common 00 0100 000 0 01 bile duct Unknown 1 0 0 0 0 0 0 0 0 0 0 0 in 1996 to >19,000 in 2008 (Figure). We included secondary to massive upper GI hemorrhage. Of 72 patients from our institution and 5 patients the 25 duodenal perforations, 12 (48%) were transferred to our institution (46 women, 31 difficult and 8 of these 12 perforations were noted men), with a median age of 70 years (range, 20-- to have complicated anatomy. Twenty-six of 39 95) who had an EGD-associated GI perforation. (67%) esophageal perforations and 19 of 25 The overall incidence of perforation for EGDs (76%) duodenal perforations underwent an inter- performed at our institution was 0.03%. Of these vention with biopsy being the most frequent procedures, 124,844 EGDs (57%) included an (11 patients and 8 patients, respectively). Of all intervention (biopsy, dilatation, coagulation, perforated patients, 7 had altered surgical anatomy and/or sclerotherapy). EGD with biopsy was the resulting in 3 jejunal perforations and a common most frequent intervention conducted (n = bile duct perforation secondary to misinterpreta- 108,053; 57%). The incidence of perforation was tion of a hepaticojejunostomy as a small bowel similar whether or not an interventional proce- stricture that was dilated. A second perforation of dure was performed (0.04% vs 0.03%; P = .181; the common bile duct occurred in a patient who Table I). The most common indication for EGD in underwent emergent EGD owing to massive GI perforated patients was evaluation of a possible up- bleeding 24 hours after endoscopic retrograde per GI cancer (26%) followed by dysphagia (22%). cholangiopancreatography with sphincterotomy.