Pancreaticogastrostomy Decreased Relaparotomy Caused by Pancreatic Fistula After Pancreaticoduodenectomy Compared With Pancreaticojejunostomy

Hypothesis  Pancreaticogastrostomy (PG) is associated with a lower relaparotomy rate following pancreaticoduodenectomy (PD) than pancreaticojejunostomy (PJ).

Design  Retrospective clinical trial.

Setting  Department of digestive surgery and transplantation.

Patients  Between 1987 and 2001, 250 consecutive patients underwent PD in our institution. Among them, 83 patients underwent PJ and 167, PG.

Main Outcome Measures  Preoperative clinicopathological features, intraoperative parameters, in-hospital mortality, postoperative morbidity, pancreatic fistula (PF), relaparotomy rates, and length of hospital stay were analyzed and compared between 2 reconstructive methods, PJ and PG, after PD.

Results  The morbidity rate, including PF, was lower in the PG group (38.3%) than in the PJ group (53.0%; P = .02). The mortality rate did not differ between the PG group (2.9%) and PJ group (2.4%). Conversely, the incidence of PF and the mean ± SD length of hospital stay were significantly lower in the PG group (2.3% and 17.2 ± 7.7 days) than in the PJ group (20.4% and 23.3 ± 11.7 days; P<.001 for both variables). Moreover, the overall relaparotomy rate was significantly lower in the PG group (4.7%) than in the PJ group (18.0%; P = .001). Nine (52.9%) of 17 patients with PF in the PJ group underwent relaparotomy. These 9 patients underwent subsequent completion pancreatectomy (n = 7) or removal of peripancreatic necrotized tissue (n = 2) with a postoperative mortality rate of 22.2%. However, no patient required relaparotomy for PF in the PG group because medical therapy succeeded in all 4 patients with PF. Moreover, no mortality related to PF occurred in the PG group.

Conclusion  The PG procedure is a safe method of reconstruction after PD, with a significantly lower rate of PF and relaparotomy.

Pancreaticoduodenectomy (PD) has been used increasingly to resect a variety of malignant diseases of the pancreatic head and the periampullary area. As a consequence of recent advances in surgical techniques and perioperative management, the mortality rate has decreased to less than 4%.^1-6 However, the morbidity rate after PD remains high, up to 50%,^6-10 and has yet to be remarkably reduced. Pancreatic fistula (PF) still remains the most serious postoperative complication after PD. The incidence of PF has been reported to be about 10% to 20%.^8,9,11-13 Intra-abdominal abscess, sepsis, and hemorrhage are the most common complications of PF associated with a high mortality rate of 20% to 40%.^6,14-16 Pancreatic fistula requires relaparotomy in about one third of patients¹³ and can be directly implicated in postoperative mortality.^1,6,11,13-16

Pancreaticogastrostomy (PG) has been investigated in the last decade as an alternative reconstruction method compared with pancreaticojejunostomy (PJ) and recommended by several authors.^17-26 In addition, several comparative studies between PJ and PG investigated which reconstruction method could more efficiently prevent PF after PD.^8,24,27-34 Among them, only 1 study was a prospective trial,⁸ which could not show a significant difference between both groups concerning morbidity and mortality rates. However, only a few comparative studies have reported series of more than 100 patients who underwent PG.^8,24,27-33

Delcore et al³⁵ reported a method of PG in which 2 to 3 cm of pancreatic remnant was telescoped into the gastric lumen. They reported a 0% incidence of PF after this method. The efficacy of this method was later investigated in several reports.^{24,27,28,32-34,36,37}

Almost all studies in which PG was performed show the incidence of PF and the rates of morbidity and mortality. However, few comparative studies have reported how many patients with postoperative complications, especially PF, required relaparotomy after PD when the reconstruction method was PG.^17,21,30

Our study retrospectively reviewed the surgical outcome in a consecutive group of 250 patients undergoing PD in a single institution. The aim of the present study was to analyze the early surgical outcome of 2 reconstruction methods and investigate whether PG with the telescoped pancreatic remnant could prevent PF or reduce the incidence of relaparotomy caused by PF.

From January 1987 to December 2001, 250 patients with pancreatic ductal adenocarcinoma and periampullary diseases underwent PD in our institution. Four surgeons experienced in hepatobiliary and pancreatic surgery performed the PDs. Patient characteristics and perioperative parameters are summarized in Table 1.

Among the 250 consecutive patients, 83 patients underwent reconstruction with PJ (PJ group) between 1987 and 1999. Until 1993, PJ was preferentially performed after PD. In 1994, PG was introduced in our institution and, after preliminary evaluation, was adopted as the only reconstruction method in February 1999 (Figure 1). Therefore, 167 patients underwent PG from 1994 to 2001 (PG group). During the transitional study period when both PJ and PG were being performed, selection of the reconstruction method was based on the personal preference of the surgeon.

The surgical procedure included either resection of the distal stomach (232 patients [92.8%]) or pylorus preservation (18 patients [7.2%], 11 undergoing PG and 7 PJ). Portal with or without superior mesenteric vein resections were associated in 39 patients (15.6%). Among these 39 patients, 29 underwent reconstruction with PG and 10 with PJ. The procedures of vascular reconstruction and lymphadenectomy were previously described.³⁸

After pancreatic transection, the hemostasis of individual bleeding points on the cut edge of the pancreatic remnant were suture-ligated with 5-0 and 6-0 polypropylene sutures (Prolene; Ethicon, Brussels, Belgium).

We performed PG according to the method described by Delcore et al³⁵ in which 3 cm of the pancreatic remnant was telescoped into the gastric lumen without any stenting of the main pancreatic duct. In brief, the proximal 3 cm of the pancreatic remnant was freed from the surrounding retroperitoneal tissue (Figure 2A) after PD. Interrupted 3-0 adsorbable sutures allowed leaning of the posterior seromuscular wall of the stomach against the anterior surface of the pancreatic parenchyma approximately 3 cm from the cut edge of the pancreas (Figure 2B). A small gastrotomy was made in the posterior gastric wall. Then, 2 cm of the pancreatic remnant was telescoped into the gastric lumen. The posterior surface of the pancreatic remnant and the gastric seromuscular layer were interruptedly sutured with 3-0 adsorbable sutures (Figure 2C). Finally, this anastomosis was secured through the gastric lumen by a continuous second layer using 4-0 adsorbable running sutures, which leaned the gastric mucosa against the serosal surface of the telescoped pancreatic remnant (Figure 2D). Suturing was done through either the gastric stump or through an anterior-wall short gastrotomy in the case of a pylorus-preserving procedure. After completion of the second layer, the permeability of the main pancreatic duct was checked with a catheter.

We performed PJ as a 1-layered, end-to-side anastomosis in which the pancreatic remnant was telescoped into the jejunal lumen with (n = 7) or without stenting (n = 76) of the main pancreatic duct. We performed PJ as the first anastomosis with the jejunal loop followed by the hepaticojejunostomy and then by the gastrojejunostomy in cases of gastric resection or by duodenojejunostomy in cases of pylorus-preserving PD.

Preoperative biliary drainage was performed routinely in patients with altered general health who on initial examination had either high serum levels of total bilirubin (>300 µmol/L [17.5 mg/dL]) or acute renal failure.

During the postoperative period, patients received antibiotic, antithrombotic, and peptic-ulcer prophylaxes. Octreotide (300 µg/d) was administered prophylactically in both groups until postoperative day 8. An upper gastrointestinal tract x-ray film controlled experiment with a water-soluble contrast substance was performed between the fifth and seventh postoperative days to assess anastomotic leakage and gastric emptying.

Mortality was recorded until the 30th postoperative day. Postoperative complications, particularly delayed gastric emptying, and PF were diagnosed according to the criteria defined by Yeo et al.⁸ We diagnosed PF when more than 50 mL of drainage fluid with an amylase concentration of more than 3-fold the normal upper limit in serum was obtained on or after postoperative day 10 or when pancreatic anastomotic disruption was demonstrated radiologically. Delayed gastric emptying was diagnosed when a nasogastric tube was placed on postoperative day 10 or later in addition to the patient's failure to progress to oral feeding.

Surgical outcome and relaparotomy rate caused by PF or upper gastrointestinal bleeding were compared between the 2 groups.

Data are expressed as mean ± SD. All statistical analyses, including the χ² test, unpaired t test, Mann-Whitney U test, and multifactorial regression analysis, were performed with the computer software package Statview (Abacus Concepts Inc, Berkeley, Calif). A difference was considered significant when P<.05.

Age, preoperative clinical history, preoperative total bilirubin and creatinine serum levels, length of preoperative hospital stay, type of PD, number of concomitant mesentericoportal venous resections, operative time, blood transfusion requirement, and rate of chronic fibrosis in the pancreatic remnant did not differ significantly between the PG and PJ groups (Table 1). Conversely, the proportion of men, rate of preoperative biliary drainage, and postoperative octreotide administration in the PG group were significantly higher than in the PJ group (Table 1). Indications for PD are listed in Table 2 and did not differ between the 2 groups.

Details of postoperative complications are shown in Table 3. The overall morbidity rate, including PF, was significantly lower in the PG group (38.3%) than in the PJ group (53.0%; P = .02). The rate of PF and the mean ± SD length of hospital stay were significantly lower in the PG group (2.3% and 17.2 ± 7.7 days, respectively) than in the PJ group (20.4% and 23.3 ± 11.7 days, respectively; P<.001 for both variables). Moreover, complications directly related to PF, such as postoperative erosive hemorrhage and relaparotomy rates, were significantly lower in the PG group than in the PJ group (Table 3). However, the rate of postoperative bleeding from the operative field and the rate of upper gastrointestinal bleeding did not differ between the PG and PJ groups. Preoperative biliary drainage did not influence the morbidity rate (46% for those with biliary drainage vs 43% for those without biliary drainage). The improvement of operative and perioperative techniques during the first 2 study periods did not influence the incidence of PF after PJ. Indeed, 9 (16.9%) of 53 patients and 8 (26.6%)of 30 patients developed PF after PJ from 1987 to 1993 and from 1994 to 1998, respectively. However, hospital volume significantly decreased the incidence of PF after PG (P = .01). In fact, 4 (5.8%) of the 68 patients operated on between 1994 and 1998 developed PF, whereas no PF has occurred since 1999.

Univariate analysis of morbidity showed that the older the patients (P = .01), reconstruction method with PJ (P = .02), and normal texture (without fibrosis) of the pancreatic remnant (P = .03) were significantly associated with an increased overall morbidity. However, in a multivariate analysis of morbidity, only PJ and the age of patients were independent factors significantly associated with postoperative complications (P = .02 and P = .01, respectively).

Seven patients died during the postoperative period within 2 months after surgery (2.8%) (Table 3). The mortality rate did not differ between the PG group (2.9%) and PJ group (2.4%). Mortality was related directly to PF in 2 (11.7%) of 17 patients in the PJ group. No death was related to PF occurrence in the PG group.

Table 4 shows indication for relaparotomy. The overall relaparotomy rate was significantly lower in the PG group (4.7%) than in the PJ group (18%; P = .001). The rate of relaparotomy required to treat complications of PF after PD was 52.9% (9/17) in the PJ group. In the PG group, none of the 4 patients who developed PF required relaparotomy.

One patient with PF in the PJ group underwent a percutaneous drainage of intra-abdominal fluid. Relaparotomy for PF was only performed in cases of severe clinical symptoms, bleeding, or hemodynamic instability. In 10 of the 15 patients who underwent relaparotomy in the PJ group, pancreatic surgery was performed. The pancreatic surgery included completion pancreatectomy for PF in 6 patients, removal of peripancreatic necrotized tissue with peritoneal drainage for PF in 2 patients, and completion pancreatectomy for uncontrollable hemorrhage from the cut surface of the pancreatic remnant in 1 patient. The 10th patient underwent a reconstruction with a salvage PG at postoperative day 4. After a hospital stay of 37 days, this patient recovered and lived for 41 months. The 2 postoperative deaths in the PJ group occurred in 2 patients who underwent completion pancreatectomy, making the relaparotomy mortality rate for PF 22.2% (2/9). The patients died in the postoperative period at days 9 and 43, respectively, from multiple organ failure. In the other 5 of 15 patients who underwent relaparotomy in the PJ group, a hemostasis of the operative field was performed in 1 patient and a repair of the gastrojejunal anastomosis was performed in 4 patients.

The causes for the 8 relaparotomies in the PG group are summarized in Table 4. None of these 8 patients underwent a relaparotomy to treat a complication related to PF. Three patients in the PG group died within 1 month after relaparotomy: the first from a mycotic aneurysm of the hepatic artery (complication of a biliary leakage), the second from a liver necrosis (complication of a celiac-axis compression), and the third from a multiple organ failure (complication of bleeding from the cut surface of the pancreatic remnant). Two patients in the PG group died from medical complications: fungal septicemia in one patient and liver failure in the other, who had hepatitis C and cirrhosis of the liver.

The 4 patients in the PG group who developed PF did not require relaparotomy for drainage or completion pancreatectomy or revision of the pancreatic anastomosis. All PF healed following a conservative medical treatment that included a total parenteral nutrition, continuous intravenous perfusion of somatostatin (6 mg/d; UCB SA Secteur Pharma, Braine-l'Alleud, Belgium), and maintenance of nasogastric tube suction and intraoperatively placed drains for 10 to 14 days. One patient needed percutaneous drainage of an intra-abdominal collection. All 4 patients recovered within 2 months (19-59 days).

Several parameters, including postoperative prophylactic octreotide administration, did not differ between those with PF (n = 21) and without PF (n = 229) (Table 5). However, normal texture of the pancreatic remnant, periampullary disease, and reconstruction with PJ were significantly associated with an increased risk of PF (Table 5). In addition, descriptions of 2 contingency table analyses for PF and morbidity are added for each group (Table 6 and Table 7). The mean length of hospital stay (P<.001), sepsis (P<.001), postoperative erosive hemorrhage due to PF (P<.001), intra-abdominal collection (P = .050), and wound infection (P = .051) were significantly higher in patients with PF compared with those without PF.

In a multivariate analysis of PF, the reconstruction method and the presence of a periampullary disease were independently implicated in the occurrence of postoperative PF (P<.001 and P = .04, respectively).

The present study showed that PG, in which 3 cm of pancreatic remnant was telescoped into the gastric lumen, significantly reduced the incidence of PF and the overall morbidity rate compared with PJ. Moreover, this series showed that PF after PG could be managed successfully without relaparotomy and related mortality.

The present single institutional study is a retrospective, nonrandomized trial. However, once again, this study shows, as the other retrospective and prospective nonrandomized series, that PG is associated with a PF rate of less than 5% provided that a telescoped technique has been used (Table 8). Moreover, the mortality rate after relaparotomy due to PF was particularly high in the PJ group (22.2%) compared with the PG group (0%). Furthermore, relaparotomy was not necessary to successfully treat PF occurring after PG.

There was no difference between the 2 groups concerning patient characteristics, preoperative status, type of PD, texture of the pancreatic remnant, associated vascular resection, and indications for PD. However, there were some minor differences between the 2 groups: preoperative biliary drainage, sex, and postoperative octreotide administration. These differences between the 2 groups could influence the interpretation of the presented data but do not constitute an important bias. First, significantly more patients (13.1% vs 4.8%) in the PG group underwent, when indicated, preoperative biliary drainage in the later part of the study. Sohn et al³⁹ and Heslin et al⁴⁰ both observed that preoperative biliary drainage increased the rate of PF and the rate of postoperative infectious complications after PD. According to these reports, an increased postoperative complication rate could be expected, including PF, particularly in patients undergoing PG. In contrast, our results showed significantly lower overall morbidity and PF rates in the PG group compared with the PJ group, despite a significantly higher proportion of patients undergoing preoperative biliary drainage. Second, in the present series, significantly more male patients (65.8% vs 53.0%) were present in the PG group. Marcus et al⁴¹ reported that male sex constitutes a significant factor for PF predisposition. As quoted previously concerning preoperative biliary drainage, increased incidence of PF could be expected in the PG group. This was not the case. Therefore, these observations suggest that PG is safer. Third, in the present study, significantly more patients (89.9% vs 77.2%) in the PG group received prophylactic administration of octreotide postoperatively and developed lesser PF. Prophylactic administration of octreotide could constitute an additional therapy to prevent PF. However, results available in the literature are controversial.^42-47 Four European multi-institutional studies showed that prophylactic administration of octreotide decreased the incidence of PF after PD.^42-45 Conversely, 2 unicentric, prospective American studies failed to demonstrate any advantage.^46,47 Our results also did not show a benefit related to prophylactic postoperative octreotide administration in terms of decreased incidence of PF as demonstrated by univariate and multivariate analysis of risk factors for PF.

A recent report already clearly showed the influence of hospital volume on morbidity, mortality, and length of hospital stay, as well as a reduction of complications during the more recent study period.⁴⁸ In the current study, the hospital volume increased during the 3 periods (Figure 1) from 7.5 PD per year in the first period (only PJ) to 19.6 PD per year in the second period (both PG and PJ) and 33 PD per year during the last period (only PG). Indeed, analysis of the PF rate during the intervals showed that there was no significant difference in the different periods of study for PF after PJ (16.9% from 1987-1993 vs 26.6% from 1994-1998) excluding any effect of hospital volume on surgical outcome after PJ. However, the same analysis showed a benefit of hospital volume on PF rate after PG (5.8% from 1994-1998 vs 0% from 1999-2001). Therefore, hospital volume, period in which PJ and PG were performed, and selection bias of PJ or PG may have some effect on the results of the present study, although these effects were not statistically demonstrated after PJ. Numerous reconstruction techniques of the pancreatic-enteric anastomosis have been developed, but none of these techniques was unanimously accepted to be safer. Postoperative complications include delayed gastric emptying, PF, and hemorrhage. In a large prospective randomized study, Yeo et al⁸ reported similar morbidity after PG compared with PJ. Moreover, they stated that an obvious relationship exists between PF and the occurrence of other complications such as delayed gastric emptying, wound infection, cholangitis, pneumonia, and development of intra-abdominal abscess. In concordance with these results, our data support the relationship between PF and other postoperative complications.

Several comparative studies between PJ and PG investigated which reconstruction method can better prevent PF after PD.^24,27-34 However, few reports discussed the modality treatment of PF and the relationship between the occurrence of PF and the necessity for relaparotomy. Indeed, Cullen et al¹³ were interested in targeting their report on the management of PF after PJ. They showed that 10 (15%) of their 66 patients required surgical intervention to treat anastomotic leaks, 7 (10.6%) of whom underwent completion pancreatectomy. In their series, 5 patients (8%) died following relaparotomy because of complication directly related to PF. However, Van Berge Henegouwen et al¹⁶ did not observe mortality after surgical treatment of PF occurring after PJ, provided that early relaparotomy is performed. In our series, we noticed a similar mortality rate of 11.7% (2 of 17 with PF) in the PJ group, whereas relaparotomy (9/17; 52.9%) and completion pancreatectomy (7/17; 41%) rates were higher. The severity of PF explained our radical strategy in the management of this complication after PD. In the absence of active bleeding, peritoneal signs, sepsis, or hemodynamic instability associated with PF, a conservative medical therapy and minimal invasive procedure, such as percutaneous radiography, to guide drainage of the intra-abdominal collections should be attempted first to avoid the need for relaparotomy.

Previous reports on the surgical management of PF after PD included completion pancreatectomy and surgical drainage.^13,16 The surgeon during relaparotomy chose the most accurate modality. Recently Schlitt et al³⁴ reported a conversion of PJ to PG in 4 patients. In the present study, a salvage PG was used to successfully treat a PF occurring after PD in 1 patient who previously had a PJ. This advantageous effect of PG was first demonstrated in experimental acute necrotizing pancreatitis in dogs.⁴⁹

Many reports in this decade have focused on PG as an alternative reconstruction method after PD.^8,17-37 Fabre et al¹⁷ reported a large multi-institutional study of PG, and Kapur et al¹⁸ reported a large single institution study of PG. Table 8 summarizes comparative studies between PG and PJ. There have been 10 studies in this decade in which PG and PJ were compared.^8,24,27-34 Yeo et al⁸ performed the first prospective randomized study of PG vs PJ. They reported that the incidence of PF was 12.3% in the PG group and 11.1% in the PJ group and that PG did not reduce the incidence of PF compared with PJ.⁸ However, they did not use a telescoped or invaginated procedure. Andivot et al³⁰ also reported identical results concerning the incidence of PF after PG (14.0%) and PJ (12.5%) using the same technique of anastomosis. In contrast, a recent controlled study supports the hypothesis that PG is safer than PJ, particularly with regards to the incidence of PF.³² Eight of the 11 studies, including ours, reported a reduced incidence of PF after PG compared with PJ.^{24,27,28,31-34} They used either a telescoped or invaginated technique.

Several potential advantages of PG have been previously advocated. They include the prevention of pancreatic enzyme activation by gastric acidity and also technical aspects related to the proximity of the pancreas to the gastric wall and reduced anastomosis in a single jejunal loop. The excellent blood supply, the thickness of the gastric wall, and emptying of the stomach through nasogastric tube suction represent additional advantages in favor of PG. However, in the present study, safety of PG was mainly based on technical aspects of the pancreaticogastric anastomosis. The so-called binding technique in which a small posterior-wall gastrotomy associated with a running second layer, which leans the gastric mucosa to the pancreatic serosa, constituted the main technical measure to prevent pancreatic leakage. This binding technique has been successfully used to prevent PF after PJ.⁵⁰ Other technical procedures have been recommended to reduce the PF rate after PJ. They include a 2-layer technique using a duct-to-mucosa anastomosis and a stenting of the main pancreatic duct.^2,51 Recent data suggested that a duct-to-mucosa/2-layer anastomosis of PJ was associated with a lower PF rate compared with an end-to-end/invagination method of PJ.^52-54 Therefore, the method of PJ performed in the present study may have influenced the rate of PF. Indeed, in the present study, a single-layer anastomosis of PJ was used, whereas a 2-layer technique according to Delcore et al³⁵ was performed for PG.

Early postoperative hemorrhage after PG can occur either from the anastomotic site or from the cut edge of the pancreas.^22,25,35 This complication almost always requires relaparotomy to control the bleeding vessels.^22,25,35 To avoid such complication, care should be taken not to cauterize vessels, even oozing points, located in the cut edge of the pancreas. Suture ligations are the only guarantee to prevent bleeding. In addition, the second layer of the PG, which leans the gastric mucosa to the pancreatic serosa, allows hemostasis of the well-vascularized gastric submucosa and contributes to decrease the bleeding rate from the anastomotic site. In the present study, the low rate of upper gastrointestinal bleeding (1.7%) after PG suggests that these technical aspects are essential. In this study, the incidence of the upper gastrointestinal bleeding was not different between the 2 groups. Therefore, these results suggest that PG cannot be incriminated in predisposing upper gastrointestinal bleeding.

Not many reports describe incidence of relaparotomy. Andivot et al³⁰ reported a reduced incidence of relaparotomy after PG compared with that after PJ. Recently, in their large series Schlitt et al³⁴ reported 29 patients who needed relaparotomy for PF. However, they did not specify the ratio of relaparotomy in each group treated with either PJ or PG. In the present study, relaparotomy because of PF was performed in 9 (52.9%) of 17 patients in the PJ group with a high postoperative mortality rate of 22.2% (2/9). The overall number of patients undergoing relaparotomy was significantly reduced in the PG group (4.7%) compared with the PJ group (18.0%). Moreover, the 4 patients with PF in the PG group were successfully treated and recovered within 2 months without relaparotomy and related mortality.

In conclusion, these results suggest that PG, in which pancreatic remnant is telescoped into the gastric lumen, can be recommended based on the reduced incidences of PF and relaparotomy compared with PJ. However, additional prospective randomized trials should be conducted to confirm this issue.

Corresponding author and reprints: Daniel Jaeck, MD, PhD, Centre de Chirurgie Viscérale et de Transplantation, Hôpital de Hautepierre, Strasbourg, Avenue Molière, 67098 Strasbourg Cedex, France (e-mail: Daniel.Jaeck@chru-strasbourg.fr).

Accepted for publication November 3, 2003.