Reoperation for Severe Pancreatitis:  A 10-Year Experience in a Tertiary Care Center FREE

THE REPORTED mortality rates of severe necrotizing pancreatitis in surgical series still vary today from 8% to 35% despite improvements in intensive care management and imaging techniques.^1- 4 Surgical treatment by necrosectomy followed by either local lavage or scheduled relaparotomy has reduced mortality compared with pancreatic resection.² The benefit of these complex surgical procedures is accepted in patients of necrotizing pancreatitis with infection but remains debatable in patients of sterile necrosis.^5- 7 Today, tertiary care centers frequently are asked to manage patients operated on elsewhere and referred because of deterioration of clinical status. Numerous series including only a few transferred and reoperated on patients reported higher mortality rates than those observed in "first-hand pancreatitis."^3,6,8 To our knowledge, there is no large study to date specifically devoted to this subgroup of patients. Therefore, we have reviewed our experience in the last decade in the management of those patients previously operated on elsewhere, transferred, and then reoperated on in our center, using an original procedure combining necrosectomy, open drainage, and prolonged lavage. The aims of this retrospective work were to analyze the specific problems encountered in treating these patients and to determine the benefit of such a complex and demanding procedure.

From 1986 to 1996, 44 patients with acute pancreatitis previously operated on elsewhere were secondarily referred for clinical deterioration and reoperated on in our unit, using a standardized technique of necrosectomy and drainage followed by continuous irrigation.

This study group consisted of 27 men and 17 women, with a mean (± SD) age of 47±13 years. Acute pancreatitis was due to gallstones in 20 patients, alcohol in 8 patients, abdominal trauma in 6 patients, hypertriglyceridemia in 3 patients, endoscopic sphincterotomy in 1 patient, and idiopathic in 6 patients.

At the initial center, the diagnosis of acute pancreatitis was suspected in only 20 (45%) of the 44 patients before surgery. Ranson score was never assessed properly in the first 48 hours of admission. Acute pancreatitis was confirmed or diagnosed intraoperatively in all patients with necrosis noted in 31 patients (70%). Necrosectomy was performed in 13 patients and 2 patients underwent distal pancreatectomy. Peripancreatic drainage was performed using rubber drains in 27 patients. No patient underwent continuous irrigation of the pancreatic bed. Bacteriological analysis of pancreatic necrosis was unavailable retrospectively in most cases.

Patients were referred to our center with a mean (± SD) delay of 21±14 days after the onset of the pancreatitis. Bacteriological analysis of drained fluids and blood cultures were systematically performed on admission to our intensive care unit.

Abdominal computed tomographic (CT) scan with bolus intravenous infusion of contrast material was performed if the patient's clinical status allowed this examination. Lesions were classified according to the Balthazar scoring system.^9- 10 There was no attempt at percutaneous puncture or drainage of the fluid collections found on CT scans.

Reoperations were decided on either because of a surgical complication of the first procedure (evisceration, undrained gastrointestinal (GI) tract fistula, or hemorrage), or because of clinical deterioration with a septic course and evidence of organ failure (hemodynamic shock, acute respiratory distress syndrome, or renal insufficiency) persisting after 48 hour of maximal intensive supporting therapy. The simplified acute physiologic score (SAPS)¹¹ was assessed in the 12 hour preceeding reoperation. Clinical findings and biological anomalies considered in our current practice in assessing patient's status¹² were collected. Reoperation was performed by a senior surgeon through the previous incision frequently extended (bilateral subcostal incisions extending to the flanks or xypho-pubic incisions). All potential sites of necrosis and collections were explored. The anterior aspect of the pancreas was exposed through large openings of the gastrocolic ligament and lesser omentum. The posterior aspect of the pancreatic head was exposed by a Kocher maneuver. Mobilization of the right and left colon exposed the right and left pararenal spaces. The peritoneum at the base of the mesentery was divided. Areas of fluid collections and necrosis were sampled for bacteriological analysis. A complete necrosectomy was performed. Mikulicz gauze packs (Figure 1) were placed in these necrotic sites and were brought out of the abdomen through the incision (Figure 2). The tail of the pancreas was drained by a Mikulicz gauze pack located behind the left colic angle and exteriorized in the left flank. An ileostomy was performed in case of dubious colic viability,¹³ while necrotic colon or colic fistula led to segmental colectomy and double stoma. Biliary drainage (cholecystostomy, transcystic drainage, or choledochal T tube) and a feeding jejunostomy were placed.

Continuous enteral nutrition was started as soon as postoperative ileus resolved. Mikulicz gauze packs were removed at the 14th day, leaving large channels giving access to the necrotic sites, into which large bore (36F catheter) 3-channeled soft silicon spiral drains¹⁴ ( Figure 1 and Figure 3) were inserted. A continuous irrigation with 2 L/d of dialysate added with 20 mL of povidone-iodine per liter of dialysate (Betadine; Asta Medica Laboratory, Merignac, France) and 50,000 UI/L of aprotinin (Trasylol; Bayer Pharma Laboratory, Puteaux, France) was immediately started through these drains. They were temporarily removed daily by the nursing staff for at least 3 weeks to perform local lavage with dialysate under pressure, to dislodge large solid necrotic debris. Antibiotics were adapted to the bacteriological analysis of the necrosis. Somatostatin analogues were not used in these patients.

Numeric results are expressed as mean±SD. Statistical analysis was performed using the χ² test with Yates correction and the Student t test where appropriate.

On admission to our intesive care unit, 41 patients (93%) received intravenous antibiotics. Peripheral bacteriological samples (drainages or blood cultures) were positive in 23 patients. Abdominal CT scan preceded reoperation in 27 patients. A Balthazar grade E CT scan was noted in 20 patients (74%) and a Balthazar grade D in 7 patients (16%). Lack of enhancement was noted in 13 patients (44%), affecting more than 50% of the gland in 10 patients, and less in 3 patients. Gas bubbles in peripancreatic collections were noted in 11 patients (6 patients had drains in place).

Reoperation was indicated because of continuing sepsis with evidence of organ failure after 48 hour of maximal therapy in 36 patients, sepsis with undrained GI tract fistula in 4 patients, sepsis with evisceration in 3 patients, and massive hemorrage exteriorized through a drain in 1 patient. The description of patient's status prior to reoperation is detailed in Figure 4. The mean value of the SAPS was 8.5±3.7.

Parenchymatous necrosis affected more than 50% of the pancreas in 18 patients, was less than 50% in 17 patients, and was absent in 9 patients. Extra pancreatic fat necrosis was always present. Necrosis was infected in 36 patients (82%) and sterile in 8 patients (18%) (Table 1). In 14 patients, a fistula of the upper GI tract was found affecting the stomach in 2 patients, the duodenum in 4 patients, the small bowel in 2 patients, and the biliary tract in 5 patients. They were due to traumatic injury by previous drainages, feeding tubes and endoscopic procedures, or the involvement of the viscera by extra pancreatic necrosis. These were treated by intubation or stoma. Nine segmental or subtotal colectomies were performed to treat 3 colic necrosis and 6 colic fistulae. Fifteen loop ileostomies were performed to divert a colon of dubious viability. One splenic injury necessitated a splenectomy. The abdominal wall was closed using an exclusive cutaneous closure in 21 patients with relaxation incisions in 6 patients.¹⁵

Ten patients died postoperatively after a median delay of 48 days (range, 6-250 days), accounting for a mortality rate of 23%. Causes of death were 5 multiple organ failures, 2 noscocomial pneumonias, 1 massive hematemesis, 1 cerebral stroke, and 1 pulmonary embolism. Age, sex ratio, delay of reoperation, bacteriological status of necrosis and associated GI tract lesions did not differ between survivors and nonsurvivors. The SAPS calculated prior to reoperation was significantly higher in patients who ultimately died (SAPS, 10.6±4.2 vs 7.8±3.4; P=.03).

Continuous enteral nutrition was started after a delay of 11±1 days and was used during 56±5 days. Continuous irrigation and drainage were maintained to ensure complete evacuation of necrotic debris over 40±6 days. Mechanical ventilation was used for 14±2 days. Forty-one surgical complications affected 31 patients (70%) during postoperative course and were treated either medically or with an iterative surgical procedure (18 patients) (Table 2). Pancreatic and GI tract fistulae were the most frequent complications affecting 10 and 7 patients, respectively. One gastric fistula occurred and was due to the reopening of a previous gastric fistula, present and sutured at first reoperation performed in our center. The remaining 6 GI tract fistulae occurred de novo. Iterative necrotic undrained collection justified a second surgical drainage in only 6 patients. No patient required more than 2 surgical drainages The mean stay in the intesive care unit was 66±8 days for survivors. Continuity of the GI tract was restored in all survivors. In 8 patients abdominal wall hernia causing esthetic or functional discomfort has necessitated delayed reparative surgery.

This case series demonstrates the value of a formal reoperation combining necrosectomy, prolonged irrigation of the pancreatic bed with active debridement of necrosis for patients previously operated on for severe acute pancreatitis and referred in a specialized center for clinical deterioration. Indeed, despite higher rates of superinfection of undrained necrosis and associated GI tract lesions than observed in de novo severe pancreatitis, this procedure yields a 23% in-hospital mortality rate comparable to that reported by specialized surgical teams for patients without operative intervention.^4,6,16

Prognostic scores commonly used for acute pancreatitis^17- 19 include data collected in the first 48 hour of first admission. These scores were therefore unsuitable for these reoperated on patients. The SAPS, which prognostic value has been established for de novo pancreatitis,¹¹ seems of comparable prognostic value for these reoperated on patients when it is calculated in the 12 hours prior to reoperation. Its high mean value reflects the severity of pancreatitis in this series.

Most patients were reoperated on because of severe clinical deterioration occurring despite appropriate intensive supportive treatment. We did not preoperatively assess the bacteriological status of necrosis by percutaneous CT-guided fine needle aspiration.²⁰ This strategy could be criticized by those operating exclusively infected necrosis,^{4- 5,21- 22} but the mortality of severe pancreatitits even with sterile necrosis can reach 38%.⁷ Moreover, the bacteriological status of the necrosis has no significant influence on the outcome for patients operated on after the third week of evolution of the pancreatitis²³ and the surgical debridment of symptomatic necrosis seems beneficial irrespective of its bacteriological status.⁶ Lastly, infection of necrosis was actually present in most patients, probably favored by previous inadequate prolonged surgical drainage.^24- 25

A poor correlation was found between CT scan and either extra pancreatic necrosis or outcome, consistent with previous findings.^10,26- 27 Gas bubbles within necrosis were not reliably attributable to infection in these operated on patients because of previously placed abdominal drains. We did not use the CT-guided percutaneous drainage of necrotic collections because of poor results in pancreatitis,²⁸ attributable to their inability to drain large solid necrotic debris.^6,24 Since our surgical procedure systematically explores all potential sites of necrotic collections, CT scan was not obtained in hemodynamically unstable patients before reoperation.

The transperitoneal route was preferred to the retroperitoneal approach recently promoted,^24,29 to perform simultaneously complete debridment of all necrotic sites, treatment of associated visceral lesions, and placement of a feeding jejunostomy. The procedure of initial packing followed by insertion of the silicon spiral drains that were removed daily for local lavage, combines the advantages of techniques of repeated necrosectomy, with prolonged evacuation of large solid necrotic debris^4,30- 31 with those of the closed irrigation, namely, dilution of toxic compounds, prolonged drainage, closure of the abdominal wall with protection GI tract fistulization, and early use of enteral nutrition.^1,32 The 14% incidence of undrained infected collections necessitating surgical drainage and the 4% incidence of late pseudocyst observed in this series compares favorably with previous reports,^1,6,8 and reflects the efficiency of this drainage procedure.

The 6 de novo GI tract fistulae (14%) observed during the postoperative course were attibutable either to the necrotizing process or more likely to GI tract traumatism by the removal of the Mikulicz gauze packs or by iterative mobilization of the spiral drains. These fistulae are, however, also observed with other techniques.³³ They could be well drained and irrigated by the spiral drains and were closed by delayed simple surgery. Pancreatic fistulae were the most frequent complication but closed spontaneously in 60% of cases. They constitute probably the unavoidable drawback of an efficient necrosectomy and drainage procedure and are observed with comparable frequency by several teams, no matter what surgical technique is used.⁴ The prophylactic use of somatostatin analogues may reduce the incidence of these pancreatic fistulae or favor a spontaneous closure.³⁴

Mortality of patients transferred from other units for severe acute pancreatitis has been reported to be greater than for direct admission in a specialized unit and to reach 35% to 75%.^3,6,8 Contrary to these previous reports including only a few patients operated on, transferred, and reoperated on in specialized centers,^3,6,8 the 23% mortality rate observed in this series is comparable to the mortality rates reported by us and others^2,6,12,29,35 for de novo necrotizing pancreatitis. The efficiency of our operative approach may be responsible for this difference. Its results argue for the early transfer of such patients in specialized centers and justify this complex and demanding treatment.³⁶

We thank John Serius, MD, for his assistance in reading the manuscript.

Corresponding author: Rolland Parc, MD, FACS, Service de Chirurgie Digestive, Hôpital Saint-Antoine, 184 rue du Faubourg Saint-Antoine, 75012 Paris, France.