Impact of Radiologic Intervention on Mortality in Necrotizing Pancreatitis: The Role of Organ Failure

Background  Our group previously reported that organ failure and mortality in necrotizing pancreatitis (NP) are not different between patients with infected and sterile necrosis. Since that report, management of this disease has evolved to include image-guided percutaneous catheter drainage (PCD) to improve morbidity and mortality. We evaluated the effect of PCD on mortality in NP.

Design  Retrospective analysis.

Setting  Tertiary care referral center.

Patients  A total of 689 consecutive patients treated for acute pancreatitis between 2001 and 2005, of whom 64 (9.3%) had pancreatic necrosis documented on contrast-enhanced computed tomography.

Main Outcome Measures  Mortality and organ failure.

Results  In the 64 patients with documented NP, overall mortality was 16%. Thirty-six patients (56%) had organ failure according to the Atlanta classification. Compared with patients with sterile necrosis, those with infected necrosis did not have an increased prevalence of organ failure or increased need for intubation, pressors, or dialysis but had an increased mortality. Mortality in patients treated conservatively was 1 of 29 (3%); in those with PCD alone, 6 of 11 (55%); in those with PCD and surgery, 2 of 17 (12%); and in those with surgery alone, 1 of 7 (14%). All patients treated with PCD alone had organ failure, whereas 10 (59%) of those with PCD and surgery had organ failure.

Conclusion  The use of PCD did not improve the mortality of NP among patients with organ failure.

Acute pancreatitis accounts for 210 000 admissions to US hospitals each year.¹ The clinical course can vary from a mild, self-limiting form in 80% of cases with less than 1% mortality to a more severe form of the disease that carries 15% mortality.² Severe acute pancreatitis is associated with organ failure and local complications such as necrosis and pseudocyst formation and is characterized by the development of a systemic inflammatory response syndrome.³

Our group previously published a series of 99 patients with necrotizing pancreatitis (NP) who were treated conservatively, with operative intervention reserved for patients with documented infected necrosis or complications of organized necrosis.⁴ Since that report, the use of image-guided percutaneous catheter drainage (PCD) has been demonstrated to be an effective alternative to open debridement in selected patients with infected pancreatic necrosis.^5-7 These small series have sparked considerable debate between those who believe that surgery is the only method of adequate drainage and others who advocate minimally invasive techniques to treat or temporize the disease in patients who are critically ill and unable to tolerate an operation. In 2002, a group of eminent pancreatologists developed evidence-based guidelines for the surgical management of acute pancreatitis.⁸ These guidelines included the avoidance of early surgery (within 14 days after onset of disease) and the recommendation that patients with sterile necrosis be treated conservatively whereas those with infected pancreatic necrosis and clinical signs of sepsis undergo surgical and/or image-guided drainage.⁸ To date, there have been no studies, to our knowledge, of the efficacy of open surgical treatment compared with or in combination with image-guided PCD. Because this modality of therapy has gained significant use at our institution, we sought to review its effect on patients with NP.

We reviewed the medical and electronic records of all patients admitted to Brigham and Women's Hospital from January 1, 2000, through December 31, 2004, with a diagnosis of acute pancreatitis. The search was conducted with the use of the International Classification of Disease, Ninth Revision code for acute pancreatitis (577.0) through the Research Depository of Patient Records, a database that stores each patient's clinical information abstracted from his or her medical record. Once identified, only the patients who had demonstrated pancreatic necrosis identified by contrast-enhanced computed tomography at our institution were included for further study. The study was approved by the institutional review board and Brigham and Women's Hospital Committee for the Protection of Human Subjects.

The records of all patients with NP were reviewed for demographic data, cause of pancreatitis, presence of organ failure as defined by the Atlanta classification (shock, systolic blood pressure <90 mm Hg; pulmonary insufficiency, PaO₂ <60 mm Hg; renal failure, creatinine level >2 mg/dL after rehydration; and gastrointestinal tract bleeding, >500 mL/24 hours), presence of infection documented as a positive Gram stain/culture on fine-needle aspiration or intraoperative specimens, antibiotic use, type and timing of intervention (conservative, PCD, surgery, or PCD and surgery), length of stay, complications, and death. Statistical analysis was performed with the Wilcoxon matched pairs test for continuous variables and χ² and Fisher exact test for categorical data. A 2-tailed P < .05 was considered significant. Stata 7.0 statistical software was used (StataCorp LP, College Station, Texas).

Patient demographic information is summarized in Table 1. Of 689 patients with a diagnosis of acute pancreatitis during this 5-year period, 64 had demonstrated pancreatic necrosis by contrast-enhanced computed tomography, representing 9.3% of the cohort. Because our institution is a tertiary referral center, 48 (75%) of the 64 patients were transferred from outside hospitals, whereas 12 (19%) were admitted through the emergency department and 4 (6%) were inpatients. The most common cause of pancreatitis was biliary (23 [36%]), followed by alcoholic (19 [30%]), other (drugs, metabolic, and idiopathic; 16 [25%]), and post–endoscopic retrograde cholangiopancreatography (6 [9%]).

Disease severity was stratified by means of the organ failure criteria from the Atlanta Symposium. Because most patients were transferred from outside institutions, lack of sufficient clinical information prevented the calculation of accurate Acute Physiology and Chronic Health Evaluation II scores. The overall prevalence of organ failure was 56%, with 15 patients (23%) having single-organ failure. Twenty-one patients demonstrated shock, 32 had pulmonary insufficiency, and 13 developed renal failure. Quantification of gastrointestinal tract bleeding could not be standardized from the medical records, and this factor was not used in the identification of organ failure.

Infected necrosis was defined as a positive culture from needle aspiration or specimen from the operating room. Patients who did not undergo sampling because of their benign course were considered to have sterile necrosis. On the basis of these criteria, 49 patients had sterile necrosis and 15 had infected necrosis. Microbiological data showed that staphylococcal species accounted for 60% of the isolates, followed by Enterococcus species (47%), Enterobacter species (26%), and Escherichia coli, Serratia species, Pseudomonas aeruginosa, Klebsiella species, and Haemophilus influenzae (all 13%). Cultures from 12 of 15 patients contained polymicrobial flora. Of note, there were no primary fungal infections. Twenty-eight patients were treated with carbapenem antibiotics, whereas 52 were treated with noncarbapenem antibiotics.

The relationship between infection status of the necrosis and organ failure is shown in Table 2. There was no difference in the prevalence of single or multisystem organ failure (MSOF) among patients with sterile or infected necrosis. In addition, patients with infected necrosis did not have an increased need for pressors, intubation, or dialysis compared with those with sterile necrosis. The presence of organ failure on admission was associated with increased mortality (8 of 28 patients [29%] vs 2 of 36 [6%]; P = .04), intubation (20 of 28 patients [71%] vs 6 of 36 [17%]; P = .005), pressor use (14 of 28 patients [50%] vs 5 of 36 [14%]; P = .04), and dialysis (8 of 28 patients [29%] vs 1 of 36 [3%]; P = .01). Overall mortality was higher in patients with infected necrosis than in those with sterile necrosis (33% vs 10%; P = .045); however, among patients with MSOF, mortality was not different between patients with infected or sterile necrosis (3 of 5 [60%] vs 5 of 16 [31%]; P = .32).

The treatment strategy and outcome for the 64 patients with necrotizing pancreatitis is outlined in Table 3. Conservative therapy alone was successfully used in 29 patients; surgical therapy alone was used in 7. The remaining 28 patients had either PCD alone (11 patients) or PCD followed by surgical intervention (17 patients). All but 6 patients in the conservative management group did not undergo fine-needle aspiration and were presumed to have sterile necrosis. The majority of these patients (17 [59%]) did not have organ failure, and only 4 patients (14%) had more than 1 organ fail. One patient had a positive fine-needle aspiration yielding Lactobacillus species, whereas the other 5 who underwent sampling had negative Gram stains and cultures. The 1 patient who died in this group developed severe MSOF, and care was withdrawn by the family's request. In the surgical group, 5 of 7 patients were taken to the operating room for infected necrosis, while 2 patients had symptoms of organized necrosis (persistent abdominal pain and inability to tolerate oral intake). The median interval to surgery was 31 days (range, 0-65 days) and there was 1 death in this group due to MSOF.

Percutaneous catheter drainage was used as primary therapy in 11 patients, of whom 3 had infected necrosis. All of these patients had organ failure, which was MSOF in 7 of 11 patients (64%); the mortality was 55% in this group. Seventeen patients had PCD followed by surgical intervention. The indications for surgery were infected necrosis in 10 patients, catheter complication in 2 patients (bleeding and gastrointestinal tract fistula), symptoms of organized necrosis (persistent abdominal pain, inability to eat, obstruction, and incomplete drainage) in 4 patients, and hemorrhagic pancreatitis in 1 patient. Although 10 of these 17 patients (59%) had some form of organ failure, 8 of them (47%) had MSOF resulting in 2 deaths (12% mortality). The median interval to surgery in this group was 109 days (range, 1-600 days). One patient experienced a laceration of the splenic artery from the catheter insertion procedure. After a failed attempt at angiographic embolization, he was brought to the operating room emergently. After splenectomy and abdominal packing, the patient remained persistently hypotensive and coagulopathic despite aggressive resuscitation and died on the following day. The other patient developed fulminant gram-negative sepsis from infected necrosis that was refractory to PCD, surgical debridement, and maximal supportive therapy, and care was withdrawn at the family's request. The use of PCD alone was associated with a significantly higher mortality than PCD followed by surgery (55% vs 12%; P = .03).

The purpose of this retrospective study was to examine the impact of image-guided PCD on patients with NP. Although previous reports have determined the feasibility of this modality to treat selected patients, our study is, to our knowledge, the first to evaluate the success of PCD when compared with the outcome of our entire cohort of patients with NP. In addition, we used organ failure as defined by the Atlanta criteria to stratify the severity of disease in our patients between different treatment modalities.⁹

The initial treatment of acute NP has shifted from aggressive surgical debridement to a more conservative strategy of supportive therapy during the first 2 weeks after onset of symptoms. In addition, intervention has been reserved for proven infected necrosis or for symptoms of organized necrosis, such as pain and the inability to tolerate feeding. In our previous cohort of 99 patients from 1995 to 2000 who were treated with a conservative strategy, 34 patients required intervention (31 surgical, 3 PCD). The prevalence of necrosis was 9%, that of infection was 34%, that of organ failure was 52%, and overall mortality was 14%.⁴ In the current update, comprising the next consecutive 5-year period from 2000 to 2005, although the overall number of patients presenting to our institution with acute pancreatitis has decreased (from 1110 to 689), neither the prevalence of necrosis (9%) nor that of organ failure (56%) has changed. In addition, the percentage of patients with NP transferred to our institution was 43% in 2000; it was 75% in the present series. This trend reflects the perceived requirement for specialized centers equipped to manage this complicated disease process. Although the progression to infection has decreased (from 37% to 23%), our overall mortality was still 16% compared with 14% previously (P = .82), and mortality with MSOF was 38% compared with 48% previously (P = .57). This was despite broader use of diagnostic modalities, improvements in supportive care in the intensive care unit, and a decrease in prophylactic antibiotic use at our institution.¹⁰ One of the most significant differences has been the role of PCD, which was used as either primary or adjunctive therapy in 28 of 64 patients (44%) in our present series.

The rationale for the introduction of image-guided PCD stems from the assumption that critically ill patients with infected necrosis and severe comorbid disease resulting in MSOF might not tolerate the stress of open surgery. Another proposed benefit is the ability to use PCD as a “bridge” treatment in selected patients who are too unstable for the operating room by using catheters to decompress the pus under pressure until surgery is deemed safe. This approach could also benefit patients whose condition continues to deteriorate during aggressive supportive therapy without clear positive culture data.

In 1998, Freeny et al⁵ published the first study of a series of 34 patients with acute NP whose sepsis was initially treated with PCD. Thirty-seven catheters were placed, vigorously irrigated every 8 hours, and exchanged 146 times for a mean of 9 days (range, 1-48 days). Sixteen of these patients were deemed to be cured, defined by the investigators as defervescence, reversal of leukocytosis, resolution of necrotic cavity, and avoidance of operative debridement by catheter drainage alone. Echenique et al⁶ performed transcatheter debridements in 20 patients with necrotizing pancreatitis who were hemodynamically stable, with a 100% success rate and without mortality. This approach required an average of 17 (range, 7-32) catheter debridements per patient, and half of the patients developed gastrointestinal tract fistulas (3 to the bowel and 7 to the pancreatic duct). Cheung et al⁷ proposed using percutaneous drainage later in the course of the disease when organized necrosis has developed. In their series of 8 patients, 5 had PCD between 3 and 5 weeks, whereas the remaining 3 had PCD at 7, 16, and 22 weeks. With this method, 3 patients had complete resolution of their necrotic cavity, and 3 required surgical intervention because of persistent sepsis, with 1 death. The remaining 2 patients required multiple sessions of PCD over several months before resolution was achieved. These studies included percutaneous drainage of pancreatic abscesses or infected fluid collections that usually have a more benign course, with less than half the mortality of infected pancreatic necrosis.¹¹ Another criticism of these reports has been the potential for selection bias because the outcome of only the patients who underwent the radiologic procedure were included in the study and little information regarding the severity of their illness was provided.

Several groups, including ours, have previously proposed organ failure as a significant determinant of mortality in NP.^12-15 This is supported by the fact that our mortality was 0% in patients without organ failure, 13% in patients with single-organ failure, and 38% in patients with MSOF. We did not observe a difference in the development of organ failure (single or multiple) between those with sterile vs infected necrosis. In addition, we did not detect an increased need for mechanical ventilation, hemodialysis, or pressor support in patients with infected necrosis as opposed to sterile necrosis. Although overall mortality was increased in patients with infected necrosis, patients with organ failure had similar death rates whether they had sterile or infected necrosis. This observation supports the hypothesis that organ failure is a more reliable marker of severity in NP.

Conservative management was successful in 29 patients, and more than half (59%) lacked any organ failure. Seven patients underwent surgical intervention alone, and 4 of these (57%) did not have organ failure. The 28 remaining patients were treated with PCD either alone (11 patients) or in conjunction with surgical intervention (17 patients). Between these 2 groups, the mortality was significantly higher in the patients treated with PCD alone (55% vs 12%). However, all of these patients had at least 1 failing organ, and 7 of 11 (64%) had MSOF. Conversely, of the 8 of 17 patients (47%) with MSOF who were initially treated with percutaneous catheters and then proceeded to surgical debridement, only 2 died. It is conceivable that PCD benefited this group of patients by decompressing their inflammatory focus in an effort to optimize surgical intervention. As with other retrospective series, a major limitation of this study is the inability to delineate transient vs persistent organ failure. Johnson and Abu-Hilal¹⁵ clearly demonstrated a survival benefit in patients with severe acute pancreatitis whose organ failure lasted less than 48 hours. The role of PCD may be expanded if its application could be shown to ameliorate either the onset or the duration of organ failure in NP.

In summary, although we did not see an improvement in overall mortality or mortality in MSOF with the addition of PCD, it is possible that a select patient population may benefit from this intervention. These might be patients with early infected necrosis and organ failure who need stabilization before definitive operative debridement. For now, it is still unclear how the evolution of PCD in the present management algorithm will translate into improved overall survival in this challenging patient population.

Correspondence: Stanley W. Ashley, MD, Department of Surgery, Brigham and Women's Hospital, 75 Francis St, Boston, MA 02115 (sashley@partners.org).

Accepted for Publication: February 5, 2008.

Author Contributions:Study concept and design: Rocha, Whang, Banks, Ashley, and Mortele. Acquisition of data: Rocha and Benoit. Analysis and interpretation of data: Rocha, Zinner, Whang, and Banks. Drafting of the manuscript: Rocha, Benoit, and Banks. Critical revision of the manuscript for important intellectual content: Rocha, Zinner, Whang, Banks, Ashley, and Mortele. Statistical analysis: Rocha and Benoit. Obtained funding: Not applicable. Administrative, technical, and material support: Not applicable. Study supervision: Zinner, Whang, Banks, Ashley, and Mortele.

Financial Disclosure: None reported.

Previous Presentation: This study was presented at the 86th New England Surgical Society Annual Meeting; October 1, 2005; Bretton Woods, New Hampshire.