Validity of Clinical Criteria in the Management of Endoscopic Retrograde Cholangiopancreatography–Related Duodenal Perforations

Objective  To assess the validity of predetermined clinical and radiologic criteria in the management of endoscopic retrograde cholangiopancreatography (ERCP)–related duodenal perforations (DPs).

Design  Prospective case series.

Setting  Tertiary medical facility.

Patients  Twenty-two individuals with ERCP-related DPs.

Interventions  The guidelines advocated operative management for patients with free contrast leak and those with positive peritoneal signs or other indicators suggesting sepsis irrespective of the mechanism or location of injury.

Main Outcome Measures  Thirty-day mortality rates, success of nonoperative management, and complications related to nonoperative and operative policies.

Results  Diagnosis of DP was accomplished early (within 6 hours of ERCP) in 20 of 22 patients (91%). Three patients with early positive peritoneal signs were treated surgically; 2 of them sustained injury from the endoscope, and the third by papillotomy. All 3 patients had significant findings justifying immediate surgery. Nineteen patients with retroperitoneal DPs due to papillotomy or guidewire insertion were treated nonoperatively. In patients diagnosed early (n = 17), only 1 failure (6%) occurred. In the 2 patients with delayed diagnosis, there was 1 failure, which culminated in death.

Conclusions  Our results might validate the role of clinical criteria for the selective management of ERCP-related DPs. These criteria were found to correlate well with radiologic findings and the mechanism of injury.

Duodenal perforations (DPs) after endoscopic retrograde cholangiopancreatography (ERCP) and endoscopic sphincterotomy are rare, with a reported incidence of 0.3% to 1.3%.^1-3 However, DPs are well-recognized and potentially hazardous complications. Retroperitoneal DPs, mainly from papillotomy or precut papillotomy, represent almost 90% of cases, whereas intraperitoneal injuries are less common and occur mainly as a result of injury from the endoscope itself.^3-5 Because of the rarity of this complication, management recommendations have largely been based on small series,^6-14 expert opinion,¹ and literature reviews.^2,15 Although nonoperative management has been recognized as a viable option for retroperitoneal DPs since the early era of ERCP and endoscopic sphincterotomy,^6,7,10 multiple studies^8,9,11,13,16 advocating primary surgical management have been published.

Recently, several researchers^4,5,17 have introduced classifications of ERCP-related DPs based on anatomical location and mechanisms of injury (Table 1) and accordingly have suggested management guidelines. Therefore, it seems that the current consensus favors selective operative management. However, existent guidelines still differ as to the validity of clinical criteria for stratifying patients into surgical and nonsurgical cohorts and the appropriate imaging modality for the diagnosis and monitoring of treatment. Furthermore, the role of endoscopic management of these complications remains controversial.¹⁸ In this study, we prospectively evaluated the applicability of preset clinical and radiologic criteria for the selective management of ERCP-related DPs.

This study was approved by the ethics committee, Rambam Health Care Campus, Haifa, Israel. Between January 1, 1996, and June 30, 2003, 3104 ERCP procedures with (n = 1250, 40%) or without (n = 1854, 60%) sphincterotomy were performed at the Rambam Medical Center. Precut papillotomy was performed in 306 patients (10%). Twenty-two patients (0.7%) had documented DPs. The ERCP was diagnostic in 2 patients and therapeutic in 20.

Characteristics of the patients are given in Table 2. Data were prospectively collected for ERCP findings, method of detection of perforation, clinical presentation of patients, imaging findings, time to diagnosis and treatment, type of management, surgical procedures, subsequent endoscopic or surgical interventions, length of hospital stay, and general outcome.

Before conducting the present study, we applied a set of criteria for the selective management of ERCP-related DPs based on previous experience and pertinent literature. These guidelines advocated immediate operative management of documented DPs for the following criteria:     • Clinical criteria: Patients with positive peritoneal signs, abdominal pain of increasing intensity, or other signs suggestive of sepsis irrespective of the type, mechanism, and location of injury. These signs included any combination of 2 or more of the following: fever (temperature > 100.4°C), tachycardia (heart rate > 100/min), systolic blood pressure less than 100 mm Hg, dyspnea (respiratory rate > 20/min), and leukocytosis (white blood cell count > 12 × 10⁹/μL [to convert to ×10⁹ per liter, multiply by 0.001]).     • Radiologic criteria: Any free (intraperitoneal or retroperitoneal) leakage of contrast material documented during ERCP, upper gastrointestinal (UGI) study, or abdominal computed tomography (CT). Contained minor leaks were not considered an indication for surgical intervention.All other patients underwent initial nonoperative treatment, which included nasogastric drainage, intravenous fluids, ceftriaxone disodium (a third-generation cephalosporin), and close clinical monitoring in the surgical service for possible deterioration. The nasogastric tube was removed when patients started to demonstrate clear improvement, and diet was usually resumed 1 to 2 days later. Patients were discharged when they were completely asymptomatic and tolerating food intake. An immediate UGI sodium amidotrizoate (Gastrografin; Schering AG, Berlin, Germany) study was used in the first 4 patients; however, because it missed the diagnosis in 3 of them, it was not used further. Routine early abdominal CT was performed in all patients with no immediate indication for surgery (including cases in which a UGI study was initially used). The purpose was to confirm the suspected diagnosis and to assess free leaks, large fluid collections, or any intraperitoneal component of the perforation and the presence of possible pancreatitis. Failure to clearly improve within 24 to 48 hours, development of peritonitis, hemodynamic deterioration, and septic signs were considered indications for prompt surgical intervention.

The indications for ERCP were as follows: obstructing periampullary tumors (8 patients), choledocholithiasis before elective cholecystectomy (8 patients), choledocholithiasis associated with cholangitis (2 patients), biliary leaks and retained stones after cholecystectomy (2 patients), dilatation of bile ducts with impaired liver function test results (1 patient), and papillary stenosis (1 patient).

Difficulties in the procedures were noted in 16 of 22 patients (73%). Possible risk factors for ERCP complications included periampullary diverticulum of the duodenum in 8 patients and Billroth II gastrectomy and papillary stenosis in 1 patient each. A biliary stent was introduced in 2 patients, one with a bile leak after laparoscopic cholecystectomy and the other with choledocholithiasis and cholangitis. A nasobiliary drain was used in 1 patient with choledocholithiasis and cholangitis after precut papillotomy. The procedure was stopped after precut papillotomy and before selective cannulation of the bile ducts in 4 patients.

The clinical presentation of patients who were initially treated nonoperatively is depicted in Table 3, and the methods of diagnosis of DPs are given in Table 4. Abdominal CT was performed within 6 to 8 hours of ERCP in all patients except 1 who was diagnosed 24 hours after ERCP and 2 with perforation of the lateral duodenal wall caused by the endoscope that necessitated immediate surgery. The findings in these 19 patients are given in Table 5.

Patients with lateral DP caused by the endoscope (n = 2) underwent surgery immediately (Figure). Suture closure with omentopexy was performed. In 1 patient with inoperable carcinoma of the head of the pancreas, a cholecystostomy was added. In a third patient who had sustained an injury from papillotomy performed for retained choledochal stones after cholecystectomy, a 3-cm perforation was found in the retroperitoneal second part of the duodenum along with injury of the intraperitoneal distal common bile duct and subhepatic bile collection. Suture closure of the duodenum, bile duct exploration, and retroduodenal area and T-tube drainage were performed. All immediately operated-on patients recovered and were discharged after a mean ± SD hospital stay of 8.5 ± 2.2 days.

Early Diagnosis (≤ 6 Hours After ERCP). Seventeen patients who were immediately diagnosed after ERCP were initially treated nonoperatively (Figure). These patients met the predetermined criteria for nonoperative treatment. Thirteen patients (76%) improved significantly within 12 to 24 hours. Ten patients (59%) were nearly asymptomatic 2 to 4 days after the perforation. The nasogastric tube was removed after 48 hours in 13 patients (76%). In the remaining patients, it was removed 3 to 6 days after the procedure.

Three patients were treated endoscopically during ERCP with biliary stents (2 patients) and a nasobiliary drain (1 patient). One other patient with inoperable pancreatic malignancy and deep jaundice underwent percutaneous biliary drainage on day 2 after ERCP as a definitive palliation. Five patients agreed to undergo a second therapeutic ERCP, which was performed successfully 7 to 14 days after the perforation. In the remaining patients, surgery was performed for disease management.

Three patients demonstrated a slow course of improvement. Repeated CT of the abdomen after 3 to 4 days showed mild pancreatitis in 2 patients and a small retroduodenal fluid collection in the third that was successfully drained percutaneously. The mean ± SD hospital stay in this successfully managed group was 6.2 ± 2.3 days. One patient who was initially treated nonoperatively deteriorated within 6 hours of initiation of nonoperative management. Tachycardia (heart rate, 120/min), dyspnea, and signs of peritonitis in the right upper abdomen mandated urgent surgery. A 2-cm retroperitoneal tear in the duodenum was primarily sutured by means of omentopexy. The retroduodenal space (in which air and a minimal amount of bile were present) was drained; cholecystectomy and bile duct exploration with T-tube drainage were performed as well. The patient recovered uneventfully and was discharged 9 days after surgery.

Delayed Diagnosis (> 6 Hours After ERCP). One patient with retroperitoneal DP after removal of a biliary stent was successfully treated nonsurgically and was discharged 6 days after ERCP. In the other patient the perforation was diagnosed 24 hours after the procedure. The CT showed retroperitoneal and mediastinal air, 1 small retroperitoneal fluid collection, and a slight amount of subhepatic fluid. Because of advanced age, comorbidities, and borderline presentation, the patient was initially treated nonoperatively. After clinical deterioration, a salvage operation performed on day 3 showed a retroduodenal abscess and bilious perihepatic fluid, and the site of perforation was not found. Cholecystectomy and retroduodenal and T-tube drainage were undertaken. Subsequently, 1 repeated operation was necessary for uncontrolled duodenal leakage, and the patient died of sepsis on day 18.

Since the introduction of ERCP and endoscopic sphincterotomy, the management of DPs has been controversial. The debate has mainly focused on the management of retroperitoneal perforations because many intraperitoneal injuries were unquestionably managed surgically. It was difficult to draw clear guidelines owing to the limited experience in individual centers. Although endoscopists have provided early accounts of successful nonoperative management,^1,6,7,10 surgeons were reluctant to adopt this policy. Their recommendations varied from early surgery in most cases^8,9,11,13 to a policy of nonoperative management supported by a few.⁴ Some studies^9,12,17,19 stated nonsignificant findings or inability to identify the perforation site during exploration. On the other hand, a high incidence of morbidity and mortality was claimed to be associated with failed nonoperative management.^2,9,17 These data, however, should be viewed with scrutiny because it was unclear whether these operations were primary or salvage procedures. The indications for and timing of nonsurgical and surgical treatment were not reported either. Furthermore, either diagnosis was late in many cases^8,11,13,16 or salvage surgery was performed with considerable delay.^{8,9,11,13,16,17} As the body of data grew, it became clear that many patients could be treated nonsurgically and that the main clinical challenge was to distinguish patients who could be treated nonoperatively from those who would require early surgery.

Similar to other perforations of the gastrointestinal tract, the importance of early diagnosis cannot be overemphasized. This allows the immediate institution of treatment, with more patients being selected for nonoperative management. This might potentially ensure better outcomes.^4,6,10,20,21 The present results in patients diagnosed early are comparable with those in other recently published series.^4,5 These results contrast those reported in some other series^8,11,13,16 describing delayed diagnosis in 70% to 75% of cases, in which most of them were managed operatively with unfavorable outcomes.

In accordance with other studies,^{4,6,7,10,17,20} our experience showed that the clear demonstration of free retroperitoneal air or leakage during ERCP allowed immediate diagnosis in most patients (80%). In the remaining 4 cases, equivocal findings were confirmed by CT. Although routine CT was not used in several series,^4,5,17 we suppose that this modality should be used in every symptomatic case or in those with retroperitoneal air demonstrated in plain abdominal films as well as in cases with uncertain or delayed diagnosis and in nonsurgically treated patients displaying an unsatisfactory course. In these instances, it is considered to be the most accurate modality for the diagnosis of these perforations and their complications.^4,9,22-26 If used routinely after ERCP with papillotomy, CT may detect retroperitoneal air in 13% to 29% of patients.^24,25 In the absence of symptoms, the mere presence of retroperitoneal air has no clinical significance, and these patients do not require any special treatment. The recommendation for the routine use of UGI studies¹⁷ is not supported by our limited experience and does not seem to be substantiated in the literature. Furthermore, UGI studies are considered to be inferior to CT regarding sensitivity to detect sealed-off or minimal leaks, and they do not demonstrate intraperitoneal findings or the presence of other complications, such as pancreatitis or fluid collections.

This study differs from previously reported series in that patients were treated according to predetermined guidelines and data were collected prospectively. In agreement with us, several researchers^5,12,15 advocated that the clinical condition of the patient should be the key factor determining the mode of treatment. We believe that clinical and laboratory variables should be correctly interpreted in the overall clinical context. Combined with correlation with mechanism of injury and radiologic findings, this seems to guarantee the best guidance for selective management. Radiologic findings alone should not determine the mode of therapy because small amounts of intraperitoneal air or fluid in the setting of benign clinical presentation (2 patients in this series) might be successfully treated nonoperatively.^5,17 Furthermore, the amount of retroperitoneal or subcutaneous air correlates more with the degree and length of endoscopic manipulation and insufflation after the injury than with the actual size of the perforation.^1,5,23 Provided that these cases are silent clinically, our experience and that of others⁴ lends further support to a policy of nonoperative management. We tend to disagree with the conclusions of Stapfer et al¹⁷ and Howard et al⁴ regarding the unreliability of clinical findings in directing the treatment of these patients. In fact, it was surprising to realize that in the series by Howard et al,⁴ all patients who were operatively salvaged had harbored early clinical findings suggestive of peritonitis. In addition, in the study by Stapfer et al,¹⁷ at least several reported failures of initially nonoperative cases were, in retrospect, not suitable for this policy.

Another major issue is the role of immediate endoscopic intervention in periampullary DP. The rationale behind this is to divert the bile and pancreatic juice to ensure adequate drainage into the duodenum instead of the perforation site. Dunham et al¹⁰ pioneered this concept in 1982, and since then several proponents have advocated immediate endoscopic drainage, either by completing the sphincterotomy or by placing a stent or a nasobiliary catheter.^4,5,7 Healing problems were not observed, and the general outcome was successful. No comparisons with patients who did not undergo drainage were performed, thus making this mode of management difficult to assess. Our policy advocated a trial of immediate drainage only in those with obstructed and infected biliary systems (mainly cholangitis due to stones). With obstructing tumors, bile is not infected in most cases, and perforations usually occur distal to the tumor, thus obviating the immediate need for drainage. Nevertheless, endoscopic management seems to be a logical and useful option provided that an experienced endoscopist is available and there is a real need for immediate drainage of the biliary system. Delayed therapeutic ERCP constitutes a viable option for the management of disease, as was shown by us and recently by others.²⁷

The recent classification systems introduced by several researchers^4,5,17 attempted to improve our understanding of the issue of ERCP-related DPs and to better direct management policy based on mechanisms of injury or anatomical locations. These essentially similar systems (Table 1) are further supported by the present results (Figure). It is our opinion that these classifications should be prudently viewed just as a general predictor of the mode of treatment that a specific patient will probably need, and they should be cautiously used only in conjunction with clinical and radiologic findings for proper decision making.

We conclude that most ERCP-related DPs in symptomatic patients should be diagnosed during or immediately after the procedure by the demonstration of retroperitoneal free air or contrast leakage. A high index of suspicion and early recognition are key to a successful outcome. In doubtful cases or those with delayed presentation, CT is the modality of choice for establishing a diagnosis and monitoring conservatively treated patients. The clinical utility of UGI studies is doubtful. The decision for stratifying patients into surgical and nonsurgical cohorts should be based mainly on clinical grounds, which were found to correlate well with radiologic findings. In conjunction with mechanism of injury, this constitutes the best available predictor of the necessity of surgery. In the absence of evidence-based data, endoscopic management may also be considered, especially in cases with an urgent need for biliary decompression.

Correspondence: Ahmad Assalia, MD, Laparoscopic Service, Department of Surgery B, Rambam Medical Center, Haifa 31096, Israel (assaliaa@aol.com).

Accepted for Publication: April 25, 2006.

Author Contributions: Dr Assalia had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Assalia, Suissa, Ilivitzki, Mahajna, Hashmonai, and Krausz. Acquisition of data: Assalia, Suissa, Ilivitzki, Mahajna, and Yassin. Analysis and interpretation of data: Assalia, Suissa, and Krausz. Drafting of the manuscript: Assalia, Suissa, Ilivitzki, and Hashmonai. Critical revision of the manuscript for important intellectual content: Assalia, Suissa, Mahajna, Yassin, Hashmonai, and Krausz. Statistical analysis: Assalia. Obtained funding: Assalia and Suissa. Administrative, technical, and material support: Assalia, Suissa, Ilivitzki, Mahajna, Yassin, and Krausz. Study supervision: Assalia, Suissa, and Hashmonai.

Financial Disclosure: None reported.