Retroperitoneal Approach and Endoscopic Management of Peripancreatic Necrosis Collections FREE

DESPITE NOTABLE progress during the past decade, acute necrotizing pancreatitis (ANP) still causes high mortality, ranging from 15% to 45% in recent reports.^1- 4 Except for the forms showing massive necrosis of the gland, which have a fulminant course from the beginning and are invariably lethal, the main risk and prognostic factor is the secondary infection of peripancreatic necrosis, provoking general sepsis and/or local complications.^5- 7 Therefore, the first management problem is to determine the bacterial contamination of the necrosis, this being aided by the use of computed tomography (CT)–guided fine-needle aspiration (FNA).⁸ The second problem is to achieve a good débridement and drainage of the peripancreatic infected necrosis for which a large panel of surgical methods is currently proposed.^9- 10 We have developed an endoscopic technique of débridement of peripancreatic fatty tissue necrosis that is based on a lumbar retroperitoneal approach. We have been using this technique in selected cases among a larger population of patients with ANP who otherwise had been treated by conventional methods. We reviewed the results of this method of retroperitoneal drainage and assessed its indications. Similarly, the clinical features and outcomes were analyzed for the patients who were managed with other techniques. We attempted to identify the elements of selection for each therapeutic method.

From January 1, 1990, to December 31, 1995, 53 patients with ANP were admitted to our institution (35 men and 18 women); the patient age ranged from 25 to 91 years (mean, 49 years). Twenty-five patients were transferred from another hospital to our intensive care unit because of the severity of their disease. The day of onset of the disease was designated day 0 and was recorded on admission to the hospital.

The causes of ANP were identified as follows: alcohol (24 patients), gallstones (17 patients), and miscellaneous (12 patients). The severity of the disease was assessed by the Ranson criteria score.¹¹ There were fewer than 3 criteria in 11 patients, 3 to 5 criteria in 31 patients, and more than 5 criteria in 11 patients (mean value for our global population, 3.3). The presence of peripancreatic necrosis was routinely evaluated by contrast-enhanced CT scan on admission to the hospital, the extent of peripancreatic necrosis being assessed by the Ranson and Balthazar grading system¹²: grade D (1 collection), 16 patients; grade E (>1 collection), 37 patients. The possible necrosis of the pancreatic gland was also evaluated as absent (n=34), one third of the pancreas (n=13), or more than one third (n=6). During the first 6 weeks, a CT scan was performed every 7 days to evaluate the morphologic changes of the collections and to detect the outbreak of gas in the peripancreatic area. The contrast medium bolus injection was omitted in patients with a risk of renal failure. Microbiologic examinations were made in patients with septic signs as follows: blood cell cultures (n=44), FNA of the collections (n=13), and intraoperative samples of necrosis (n=36).

All patients initially were hospitalized in the intensive care unit to ensure nasogastric suction, total parenteral nutrition, and cardiac monitoring. Intravenous antibioprophylaxy was systematically instituted provided that patients had evidence of necrosis on CT scan, the duration of ANP was less than 7 days, and no other antibiotherapy regimen had been previously prescribed (in another institution). Thirty-six patients met these criteria. The different drugs (cefotaxime sodium, ceftazidime, cefuroxime sodium, and imipenem) were used in monotherapy. During the first 7 days we also used mechanical ventilation (n=15), hemodialysis (n=10), vasopressive or inotropic drugs (n=13), and peritoneal dialysis (n=4). A laparotomy was performed only to check the possibility of an intraperitoneal complication such as gut perforation, digestive necrosis, or hemoperitoneum.

The criteria for a drainage procedure were the secondary outbreak or persistence (>7 days) of signs of sepsis unexplained by a source of infection other than an abdominal abcess; and/or organ failure despite medical treatment; and/or bacteriological proof of infection (blood cell cultures and FNA). Percutaneous drainage (PD) was reserved to collections predominantly liquid on CT scan (low density and homogeneous); lumboscopic drainage was proposed to patients having heterogeneous collections and/or rapid enlargement on 2 successive examinations and/or outbreak of gas in the peripancreatic area.

The patients could be retrospectively divided in 4 groups, depending on the methods of treatment: supportive therapy alone (group 1, n=14), percutaneous drainage (group 2, n=10), endoscopic retroperitoneal drainage (group 3, n=20), and laparotomy plus transperitoneal drainage (group 4, n=9).

A feeding jejunostomy was performed in 47 patients using a conventional (n=35) or a laparoscopic (n=12) technique. One or several biliary procedures were needed in 23 patients: endoscopic retrograde cholangiopancreatography and endoscopic sphincterotomy alone (n=4), endoscopic retrograde cholangiopancreatography combined with endoscopic sphincterotomy and cholecystectomy (n=8), cholecystectomy alone (n=6), and cholecystectomy and T tube (n=5).

The retroperitoneal approach to the pancreas was routinely conducted through a short lumbotomy (6 cm in length) centered on the 12th rib. The spleen and the descending colon were then mobilized anteriorly without opening the peritoneum, which provided direct access to the pancreas. Considering the depth of the operating field, we used a direct vision endoscope (23-cm mediastinoscope), which allowed a satisfactory exploration of the anterior pararenal space and prepancreatic area. For patients having collections in the right mesocolon and the right pararenal space, a similar right lumbotomy was used and the retropancreatic area was reached after mobilizing the ascending colon. The necrotic peripancreatic tissues and collections were removed by blunt dissection using a suction metal tube that also is equipped for electrocoagulation. After débridement, a tube drain allowing continuous irrigation was left in the retroperitoneal space and the wound was left open and packed to facilitate the subsequent procedures. The semilateral position of the patient allowed an abdominal extension of the incision in case of bleeding and feeding jejunostomy during the same operation.

The lumboscopic drainage was repeated under general anesthesia every 5 days until definitive elimination of all debris. Then a simple irrigation and lavage could be performed daily without anesthesia.

General or local complications were encountered, and reoperations or complementary procedures were done. The main outcome criteria in each group of patients were the mortality and the mean stay in our department from admission to the hospital to final release. Data were analyzed using the Fisher exact or the Wilcoxon test. Statistical significance was inferred for P<.05. Data are given as mean±SD.

Fourteen patients did not meet the criteria for percutaneous or surgical drainage and were treated exclusively by medical therapy. These were less severe cases (Table 1). None of these patients underwent FNA. Two early complications were encountered: 1 Staphylococcus aureus septicemia and 1 transient and mild renal failure with no need for hemodialysis. A spontaneous resolution was observed in all of the cases without mortality. The mean hospitalization stay was 23±9 days in this group (P<.05 vs groups 2, 3, and 4). One patient acquired diabetes as a definitive sequel.

In 10 patients, a PD was proposed after FNA. The preoperative clinical features of these patients are given in Table 1. The PD was performed between day 10 and day 25 (mean, day 17). One or 2 drain tubes were used, ranging in size from 5 to 10 mm. The mean duration of PD was 16 days (range, 4-22 days). The bacteriological features of these initial samples and the further events are given in Table 2 The patients who had a sterile collection underwent drainage because of the persistence of a large collection after day 20. The mean hospital stay was 89±24 days for the entire population of PD. There were 3 late pseudocysts persisting more than 6 weeks; 2 resolved within 3 months without specific treatment and 1 required a late PD. Two patients remained diabetic; 1 had to be reoperated on for a late hernia on the lumbotomy wound.

All patients had signs of uncontrolled sepsis before the lumboscopic drainage (Table 1). Three patients had previous proof of infection of a heterogeneous collection (FNA). Four patients had a positive blood cell culture for gram-negative rods. The other patients had indirect signs of infected necrosis on CT scan: presence of gas into the collection (n=4) and/or rapid enlargement of the collection in 2 successive examinations (n=7). The first lumboscopy procedure was performed between day 13 and day 26 (mean, day 18). A bacterial infection was observed on the intraoperative samples of the first lumboscopic procedure in 13 patients (Table 2). In the other 7 patients, the initial sample was sterile and a secondary superinfection occurred in 3 patients but may correspond to a contamination through the open wound. The mean number of lumboscopic procedures performed under general anesthesia was 5±4. Local complications occurred in 6 patients during the early postoperative phase (up to 4 weeks after the first lumboscopic drainage) (Table 2). Two of these complications (colonic fistula and splenic bleeding) obviously were caused by the technique of surgical drainage; the 4 other complications were more likely related to the location and severity of the pancreatitis (cephalic necrosis) and/or the persistence of an inflammatory process. In complement, a short laparotomy was needed in 2 patients to drain a persisting collection of the mesenteric root. A late complication (>6 weeks) occurred in 6 patients: 2 pseudocysts—1 treated by PD and 1 resolved spontaneously; 2 external pancreatic fistulas healed spontaneously without specific treatment; and 2 wound hernias on the lumbotomy required reoperation. Three patients remained diabetic. Mortality was 10% in this group. The mean hospital stay was 62±21 days (P<.05 vs groups 2 and 4).

All of the patients had signs of severe sepsis before the laparotomy with a suspected intraperitoneal complication (Table 1). The primary laparotomies were performed between day 4 and day 13. Two patients had a colonic perforation with peritonitis (1 death); 1 had a ruptured spleen with intraperitoneal hemorrhage. Two patients had a massive necrosis of the pancreatic gland (both patients died). The remaining 4 patients had no intraperitoneal complication but had multiple necrotic collections, with bacterial contamination in 2 patients. Bacteriological data and postoperative course are given in Table 2. Two late pseudocysts occurred and were treated by PD, 2 wound hernias required reoperation, and 3 patients remained diabetic. Mortality was 33% in this group (P<.05 vs group 1). The mean hospital stay was 86±32 days.

The overall mortality was 13.2% in this series. The average delay between the onset of the ANP and death was 47±44 days. Bacterial contamination was proved in 22 of 39 (56%) samples of peripancreatic necrosis. Mortality was significantly higher in patients having infected necrosis (7 of 22), compared with those with sterile collections (0 of 17) (P<.05).

Secondary infection of peripancreatic necrosis is regarded as the main prognostic factor of ANP.^5- 6,13 The role of gram-negative bacteria originating from the gastrointestinal tract has been shown in humans and experimental models.^14- 17 A major risk of bacterial contamination exists when a peripancreatic necrosis is present, depending on its extent and the duration of the disease.^18- 20 Nevertheless, an exact prediction of a further infection is difficult in the early phase of the ANP. In our series, CT signs of peripancreatic necrosis were constant on admission to the hospital and were the basis for the inclusion of the patients in this study. However, retrospective analyses of patients' courses show significant differences between the study groups despite minimal variations of average Ranson criteria scores. Obviously, patients in group 1 had a relatively benign disease and were not candidates for aggressive therapeutic methods. Conversely, patients in group 4 had a more fulminant course and had more local complications than groups 2 and 3. Even though this retrospective grouping allows no comparison between the different results, it gives information about selection of the patients for each therapeutic modality.

Bacterial contamination may be suspected on clinical signs of sepsis and/or positive blood cell cultures, but both elements lack specificity. Respiratory or renal failure or any worsening of the clinical state strongly supports the hypothesis of a peripancreatic infected necrosis when they occur after 7 days of evolution.^5,21- 22 In our experience, this has been an indication for drainage even without a previous bacteriological confirmation. In contrast, when organ failure is present from the beginning and persists despite supportive therapy, the use of FNA is of maximal value to show infection of necrotic tissues. We do not propose drainage in the absence of sepsis, organ failure, and/or bacteriological proof. In other words, we disagree with a policy of "prophylactic" débridement of sterile peripancreatic necrosis, which has been suggested by some authors.²³ Nevertheless, our cases in which initial bacteriological intraoperative samples were sterile indicate that a wider use of preoperative FNA should be recommended.

A debate persists about the reliability of PD of collections of peripancreatic necrosis.^24- 25 Our data show satisfactory results only when PD was performed for a late (>20 days) persistence of large fluid collections, and the sterility of these collections indicated that they could have resolved spontaneously without drainage. On the other hand, failures of PD were observed in group 2, with the subsequent need for other therapeutic modalities (lumboscopy or laparotomy). In cases of infected necrosis containing thick tissue debris, PD with conventional caliber tubes has a poor outcome and may not be adapted to good quality and prolonged drainage.

Surgical débridement and drainage is widely regarded as the only reliable therapeutic method for peripancreatic infected fat necrosis.^{1- 2,6,26- 28} In contrast, the benefit of excision of pancreatic parenchyma necrosis and/or pancreatic resections is controversial.^4,29 Accurate evaluation of the extent of gland necrosis is difficult and may result in excessive resections, especially when surgery is performed too early. Controversy also exists about the optimal time for débridement of peripancreatic fat necrosis.³⁰ In our experience, surgical débridement for infected necrosis is not performed before the seventh day of evolution. Nevertheless, because we did not systematically perform early FNA, we cannot assert that surgical drainage is useless during the first week of the disease.

The modalities of surgical treatment of infected peripancreatic necrosis are largely debated: surgical approach, closed or open drainage, and need for a continuous irrigation or iterative procedures.⁹ The only general agreement concerns the necessity of a complete débridement of all the infected peripancreatic necrosis.³¹ Before the CT scan, the use of the transperitoneal approach had the following justifications: it was the only way to provide an exhaustive assessment of multiple peripancreatic collections, it was the means to provide a total peripancreatic necrosectomy and drainage, it was the opportunity to inspect intraperitoneal organs, and it allowed biliary and/or "adjuvant" procedures, such as feeding jejunostomy and/or diversion loop ileostomy. In contrast, laparotomy had the following potential disadvantages: contamination of the peritoneal cavity, further intraoperative difficulties in case of relaparotomy, enhanced risk of multiple erosions of intestinal tract and splanchnic vessels due to the repeated procedures, wound mutilation with respiratory repercussions, and secondary occurrence of abdominal hernias.⁹ Subsequently, the concept of retroperitoneal drainage with a lumbotomy has been developed to avoid these drawbacks,^32- 33 an attitude that is aided by the systematic and repeated use of CT scan for detection of peripancreatic necrosis. The main drawback of this method is its narrowness and blindness about the potential risk of damage to visceral organs or vessels.³⁴

In this study, the use of endoscopic devices reduced the risk of visceral injury because the mobilization of intraperitoneal structures was minimal. When performed from the left, this method allows good access to the anterior side of the pancreas, beyond the midline (Figure 1 and Figure 2), down to the Toldt fascia and possibly to the pelvis. From the right, this method gives access to the posterior side of the duodenopancreas and also to the retrocolic area (Figure 3). The procedure allows the careful inspection of the operative field and the excision of large tissue debris (Figure 4). Even though the quality of débridement is good, we have a policy of second-look operations based on successive CT examinations (Figure 5). Leaving an open wound with an irrigation and lavage system makes the repetition of the procedures easier and avoids the need for general anesthesia after an average of 5 operations. This policy has avoided a laparotomy in 13 patients (65%) of this group.

It may be argued that this method of retroperitoneal drainage cannot evacuate collections located in the transverse mesocolon or the mesentery root.³⁵ In our experience, such collections usually had a progressive resolution after retroperitoneal drainage, probably because there are anatomical connections between the splanchnic and retroperitoneal cellular spaces. When this type of collection persists, we propose a débridement through a size-limited, "targeted" transperitoneal incision. In our series, the other indication for a laparotomy was the suspicion of, or the occurrence of, a complication such as peritonitis, enteric fistula, or intraperitoneal bleeding; or the requirement for surgical treatment of a concomitant biliary cause. An algorithm for management of peripancreatic necrosis collections is shown in Figure 6.

The overall mortality was 13.2% in this series of ANP, which compares with the results of recent studies.^1- 4 Mortality was 10% in patients treated by retroperitoneal drainage. This may indicate that lumboscopic drainage provides a quality of débridement as good as the transperitoneal approach. However, the results of these 2 methods could not be compared in our series, because their indications were different. Lumboscopic drainage can be repeated easily under brief general anesthesia and avoids repeated opening of the peritoneal cavity and its possible contamination. This surgical approach is not exclusive and may be combined to a secondary laparotomy when needed. The preferred indications of this method are heterogeneous collections of necrosis with bacterial contamination. Even in an evident context of sepsis, FNA should be performed to decide about surgery and to adapt antibiotherapy.

Reprints: Luc P. Gambiez, MD, Clinique Chirurgicale Ouest, Hôpital Claude Huriez, CHRU, F 59037 Lille Cedex, France.