Pancreatic-fluid collections: a randomized controlled trial regarding stent removal after endoscopic transmural drainage

ORIGINAL ARTICLE: Clinical Endoscopy

Pancreatic-fluid collections: a randomized controlled trial regarding stent removal after endoscopic transmural drainage Marianna Arvanitakis, MD, Myriam Delhaye, MD, PhD, Maria Antonietta Bali, MD, Celso Matos, MD, `re, MD, PhD Viviane De Maertelaer, PhD, Olivier Le Moine, MD, PhD, Jacques Devie Brussels, Belgium

Background: Endoscopic transmural drainage is obtained by creating a communication between the intestinal tract and the pancreatic-fluid collection, and then inserting 1 or more stents. Collection recurrence after therapy is noted in 10% to 30% of cases. It is not known whether leaving the stents in position reduces recurrence rates. Objective: To test the hypothesis that patients who have undergone previous successful pancreatic-collection drainage and whose stents are retrieved have higher recurrence rates. Design: Randomized controlled trial. Setting: Tertiary referral center. Patients: During a period of 27 months, 46 of 77 patients who had undergone endoscopic transmural drainage for pancreatic collections met inclusion or exclusion criteria, and 28 of these patients were randomized. Interventions: Fifteen patients were assigned to group A, whose stents were left in place, and 13 were assigned to group B, whose stents were removed after collection resolution. The remaining 18 patients, who were not randomized, also had their stents left in place. All 46 patients were similarly followed. Main Outcome Measurement: Recurrence of the same pancreatic collection that required therapy. Results: All patients were followed for a median period of 14 months (interquartile range, 8.2-22 months) after treatment. The primary end point was reached in 5 patients in group B (stent retrieval), as opposed to none in group A (P Z .013). Moreover, no recurrence was observed in the remaining 18 nonrandomized patients. Limitations: Small sample size. Conclusions: In patients who underwent successful transmural drainage of pancreatic collections, stent retrieval was associated with higher recurrence rates. (Gastrointest Endosc 2007;65:609-19.)

Pancreatic-fluid collections (PFC) may occur as a result of acute pancreatitis (AP) or chronic pancreatitis (CP),1 pancreatic or abdominal surgery,2,3 and abdominal trauma.4 Pathogenesis is based on either a disruption of the pancreatic-ductal system, with subsequent fluid leakage,5 or progressive liquefaction of pancreatic necrosis.6 A widely accepted nomenclature for classifying PFCs that occur as a complication of AP or CP has been established and includes acute fluid collections, acute pseudocysts, pancreatic abscess, organized pancreatic necrosis, and chronic pseudocysts.7,8 See CME section; p. 672. Copyright ª 2007 by the American Society for Gastrointestinal Endoscopy 0016-5107/$32.00 doi:10.1016/j.gie.2006.06.083

Therapeutic options for PFC drainage include surgery, endoscopy, and imagery-guided percutaneous drainage. Surgical cyst-enteric anastomosis has traditionally been indicated, although morbidity rates ranged from 7% to 37%.9,10 Percutaneous drainage offers satisfying results but requires the presence of an indwelling catheter for several weeks and is occasionally associated with external fistula formation.11 During the last 10 years, endoscopic drainage has been increasingly used.8,12-28 Two endoscopic approaches have been described: through the GI wall by creating a cystogastrostomy or a cystoduodenostomy (transmural drainage), or through the papilla (transpapillary drainage), followed, in both cases, by the insertion of 1 or more plastic stents.29 Results are satisfactory, with a 70% to 87% clinical success rate, and a complication rate that ranges from 11% to 34%.21,22 However, the

www.giejournal.org

Volume 65, No. 4 : 2007 GASTROINTESTINAL ENDOSCOPY 609

Stent removal after endoscopic transmural drainage

progress made, thanks to increased experience and new technology, including the introduction of EUS and the development of large-channel echo endoscopes, has led to advances that include the ability to drain pancreatic abscesses,26 organized liquefied necrosis,6,8,23 and nonbulging distant PFCs,18 thus extending the indications for transmural drainage. In most series involving transmural drainage, endoscopic removal of the stent(s) was routinely performed within 2 weeks of documented PFC resolution.8,20,21,27 The rate of PFC recurrence that required further endoscopic, surgical, or percutaneous drainage varied from 10% to 30%, and recurrences usually occurred during the year after treatment.8,13,14,20,21 However, there are limited data concerning the exact time of recurrence and the optimal duration of stent placement. Transmural stents could also be left in place, even after PFC resolution, because the stents mainly serve as guides to keep the fistula patent between the PFC and, indirectly, the pancreatic duct and the digestive tract.18,22 Therefore, pancreatic secretions are effectively drained even if stent occlusion occurs. This is particularly the case for complete main pancreatic duct (MPD) rupture, when the MPD segment upstream of the rupture is disconnected and exclusively drained through the fistula, a situation in which our policy is always to leave the stent in place. Furthermore, a recent retrospective study revealed that drainage lasting longer than 6 weeks was associated with a more favorable outcome and a lower recurrence rate, thus underlining the need for potential long-term stent placement.21 The aim of this randomized, controlled trial was to test the hypothesis that patients with AP or CP who had undergone previous successful endoscopic transmural drainage of PFC and whose stents had been left in place had lower PFC recurrence rates than those whose stents had been retrieved within 2 weeks of PFC resolution.

Arvanitakis et al

Capsule Summary What is already known on this topic d

Previous studies suggest that, after endoscopic transmural pancreatic fluid collection drainage, stent retrieval should be performed because stent occlusion might lead to a recurrence of the collection and because the stent might act as a foreign body and lead to infectious complications.

What this study adds to our knowledge d

In a randomized controlled trial of 46 patients who had successful transmural drainage of PFC, stent retrieval was associated with higher recurrence rates than stent maintenance, especially in patients with MPD rupture.

bleeding, infection) that required multiple subsequent endoscopic or radiologic interventions or surgery; (5) combined PFC drainage modalities (percutaneous and transpapillary); and (6) follow-up impossible (eg, patients referred from other institutions or from abroad for endoscopic therapy only). PFC recurrence was defined as a recurrence of the same symptomatic PFC after initial successful resolution that required endoscopic, radiologic, or surgical therapy.

Treatment techniques

This was a single-center, prospective, randomized trial, conducted at the Erasme University Hospital. The study was approved by the institutional review board of our institution, and all patients gave their written informed consent before participating. Consecutive patients presenting with 1 or more PFCs with a diameter exceeding 30 mm, caused by AP or CP, who had undergone successful exclusive transmural endoscopic drainage with stent insertion and PFC resolution, were eligible for inclusion in the trial. Exclusion criteria were the following: (1) PFC resulting from pancreatic surgery, including necrosectomy; (2) early stent migration before randomization; (3) pregnancy; (4) major complications occurring during the drainage procedure (eg,

Before therapy, all patients had magnetic resonance imaging (MRI), with secretin-enhanced MRCP (S-MRCP), to define the location, the size, and the content of the PFC,30 as well as to explore the possible presence of a MPD rupture, defined as a focal interruption of ductal continuity on MRCP sections31 after intravenous secretin administration. Contrast-enhanced MRI and S-MRCP also provided useful information regarding the presence of a pseudoaneurysm,32 proximity with adjacent organs (GI tract or common bile duct) and ductal abnormalities suggestive of CP.33 All endoscopic procedures were performed with the patient under general anesthesia with endotracheal intubation. Prophylactic antibiotics (amoxicilline-clavulanate) were administered intravenously during the procedure and for the subsequent 48 hours. A therapeutic duodenoscope (TJF-160 or TJF-140; Olympus Optical Co, Ltd, Tokyo, Japan) and/or a linear EUS scope (GFUCT-140 [Olympus]; EG-3830UT [Pentax, Hamburg, Germany]) were used in all cases. Pancreatography and deep cannulation of the MPD were attempted in patients with CP and ductal abnormalities (strictures, stones), as well as in patients with a suspicion of a distal, corporealocephalic rupture based on S-MRCP. PFC communication was assessed based on the pancreatogram. Complete MPD rupture was confirmed as an extravasation of contrast medium from the ductal

610 GASTROINTESTINAL ENDOSCOPY Volume 65, No. 4 : 2007

www.giejournal.org

PATIENTS AND METHODS General methods, inclusion criteria, and definitions

Arvanitakis et al

system when the main duct proximal to the rupture was not opacified or as a complete stop at the pancreatogram (disconnected duct syndrome).34,35 Transpapillary MPD drainage started with pancreatic sphincterotomy of the major and/or the minor papilla as previously described.29 In the case of MPD strictures, dilation was performed by using either graduated Teflon bougies (6F-10F; Cook Endoscopy, Winston-Salem, NC) or 4- to 6-mm hydrostatic MaxForce balloon catheters (Microinvasive Endoscopy, Boston Scientific Corp, Natick, Mass).29 Nasopancreatic catheters and/or pancreatic polyethylene straight stents were placed if required. All patients underwent successful endoscopic transmural drainage, with or without EUS guidance. In most cases, 1-step EUS-guided transmural drainage was performed with a therapeutic EUS scope. Puncture was performed with the electrosurgical needle of a 10F cystenterostome (Cystotome; Cook Endoscopy). Once the needle with its catheter was in the PFC, a sample of its content was taken for bacteriologic analysis and carcinoembryonic antigen measurement. This was performed as a routine analysis to exclude the possibility of a cystic neoplasm, although results of previous clinical data and S-MRCP suggested a PFC in all patients. Afterward, a 0.035-inch guidewire was inserted. The tract was then enlarged by passing the cystenterostome into the PFC, followed by the positioning of a double-pigtail stent (7F-10F, 3-12 cm in length between pig tails) and/or a nasocystic catheter (6F-7F). For the placement of multiple stents, further dilation of the transmural path was performed with an 8- to 15-mm hydrostatic balloon catheter (MaxForce or CRE; Microvasive), or a second puncture was performed during a second procedure. Patients who had large or infected PFCs and/or those with PFCs that contained purulent or necrotic debris initially had a nasocystic catheter placed to allow flushing of the PFC. The catheter was subsequently exchanged for 1 or 2 stents. After the final stent(s) insertion and before discharge, the patient underwent MRI and S-MRCP to assess the efficacy of drainage.

Stent removal after endoscopic transmural drainage

Patients who were not eligible for randomization were also followed prospectively.

Outcome All patients, whether or not they were randomized, had the same follow-up, which consisted of laboratory investigations, MRI and MRCP, and plain abdomen film 1, 3, and 6 months after endoscopic drainage and every 6 months thereafter. The primary end point was recurrence of the same symptomatic PFC that required endoscopic, radiologic, or surgical therapy. In case of spontaneous stent migration during follow-up, the stent placement period was calculated as the interval between stent insertion and the last follow-up visit, whereas the stent was still in place. Data were analyzed by intention-to-treat; therefore, each patient was considered in the group initially allocated, independent of stent migration during follow-up. All patients were followed for at least 6 months or until the end point was reached.

Statistical analysis

All patients were seen 1 month after the last procedure, after measurement of their hydrolases and C-reactive protein levels, MRI, MRCP, and plain abdomen film. They were asked to participate in the study if they fulfilled all the inclusion criteria, the PFC had resolved, and the stent(s) was still in place. In case of persistent PFC, the patients were reevaluated after 1 month. Patients were assigned to 2 groups (A and B) through a computer-generated randomization list, with a block size of 4. For group A patients, the stents were left in situ and for group B patients, they were removed. The latter underwent upper-GI endoscopy with stent removal within 2 weeks of randomization. If a patient had 2 stents that drained either one or different PFCs, both were retrieved.

The sample size was estimated by assuming that the overall rate of PFC recurrence after PFC resolution and stent retrieval was 25% after a follow-up of 12 months. This rate was based on the results of previous studies, by assuming that their retrospective design may have led to an underestimation of the PFC recurrence rate. Our hypothesis was that by maintaining the stents in place, the rate could be reduced to less than 5%. To detect such a difference, by using a 2-tailed testing, an alpha of 0.05, and a beta of 0.20, 60 patients would have been required in each group. An interim analysis of the results was planned if PFC recurrence rates reached 35% in either group. A P value of !.03 was selected at the study onset for stopping early. After the enrollment of 28 patients, the PFC recurrence rate reached 38.4% in group B (stent retrieval), which prompted an interim analysis of the findings. After this analysis, which showed a significant difference between the 2 groups for PFC recurrence, our institutional review board expressed reservations about proceeding further, and we decided to close the study to further enrollment. Nevertheless, follow-up was continued for all patients, whether or not they were randomized. Data were analyzed by intention-to-treat. Descriptive statistics were expressed as median and interquartile range (IQR) for continuous variables. Differences between groups were calculated by using the Mann-Whitney test for continuous variables. The Pearson c2 test or the Fisher exact tests, if appropriate, were used for categorical variables. Also, 95% confidence intervals (CI) on probabilities were calculated by the Wilson method.36 The KaplanMeier method was used to determine the times to event (same PFC recurrence that required therapy) and the log-rank test, to compare the difference between the 2 groups.37 In addition, univariate analysis was performed to assess the effect of the factors analyzed for the entire

www.giejournal.org

Volume 65, No. 4 : 2007 GASTROINTESTINAL ENDOSCOPY 611

Randomization

Stent removal after endoscopic transmural drainage

study population in connection with the probability of PFC recurrence requiring therapy. Statistical significance was considered for P values %.05. Statistical analyses were performed by using SPSS software, version 11.0 (SPSS Inc, Chicago, Ill) and CIA (confidence interval analysis) software, version 2.0.0 (Trevor N Bryant 2000, Southampton, UK).

RESULTS Participants flow In our institution, between June 2003 and October 2005, endoscopic transmural drainage was attempted in 80 patients with PFC. Successful drainage was achieved in 77 patients. Eleven patients did not participate in the trial because they had PFCs resulting from pancreatic surgery (n Z 6) or previous surgical necrosectomy for necrotizing AP (n Z 5). Twenty patients were referred to our department from abroad for the endoscopic procedure only and could not be followed prospectively; therefore, they were not included in the trial. Of the remaining 46 patients who had PFCs because of AP or CP, 18 patients were not randomized for the following reasons: stents were not in position at the time of randomization because of early spontaneous stent migration (n Z 4) or early deliberate stent retrieval (a cystogastric stent near the gastroesophageal junction, which resulted in dysphagia) (n Z 1), refusal to be randomized (n Z 10), and major complications occurring during the procedure, which required further endoscopic, radiologic, or surgical therapy (n Z 3). Consequently, 28 of the 46 patients were randomly assigned to stent maintenance (group A, n Z 15) or stent retrieval (group B, n Z 13) (Fig. 1). All 46 patients were followed prospectively, as described above.

Baseline characteristics Twenty-eight patients (23 men, 5 women) were randomized, with a median age of 49 years (IQR, 41-54 years). Etiology for the PFCs included AP (n Z 11) and CP (n Z 17). PFCs included acute pseudocysts (n Z 6), chronic pseudocysts (n Z 15), pancreatic abscesses (n Z 5), and organized liquefied necrosis (n Z 2). MPD rupture based on S-MRCP was present in 14 cases. Carcinoembryonic antigen levels in the PFC fluid were low in all patients (!30 ng/mL), excluding the possibility of a cystic neoplasm. Groups A and B were comparable for their baseline characteristics, listed in Table 1.

Arvanitakis et al

1-2). In 8 patients, a nasocystic catheter for drainage and flushing was placed before final stent insertion for a median period of 4 days (IQR, 3-20.5 days). Of the 17 patients with CP, 9 also underwent MPD drainage with major (n Z 9) or minor sphincterotomy (n Z 1), and stent insertion if a MPD stricture was present (n Z 6). Six patients with calcifying CP underwent extracorporeal shock wave lithotripsy for obstructive MPD stones. Four patients with moderate CP and no ductal dilatation did not undergo MPD drainage. Four patients had endoscopic transmural drainage for 2 concomitant PFCs. The following treatment-related minor complications occurred in 5 patients: retroperitoneal leakage, with no clinical impact (n Z 1); sepsis, treated successfully with antibiotics (n Z 2); temporary bleeding that required transfusion but no further therapy (n Z 1); and pain (n Z 1). Posttherapeutic assessment by MRI and MRCP performed before discharge after a median interval of 6 days (IQR, 3-11.5 days) after stent insertion revealed a decrease of O50% in PFC size in 18 patients (64.2%). PFC resolution was observed 1 month after therapy in 25 patients (89.2%). There were no significant differences between the 2 groups with regard to endoscopic therapy (Tables 2 and 3). In group B, stents were maintained in place for a median period of 2 months (IQR, 2-3.2 months), and stent retrieval was not associated with any immediate complications.

Primary outcome Randomized patients were followed for a median period of 14 months (IQR, 8.2-22 months) after endoscopic drainage and 11.5 months (IQR, 5-20.2 months) after group assignment. Five of the 13 patients (38.4%) in group B (stent retrieval) experienced recurrence of the same PFC that required therapy, whereas no patient in group A (stent maintenance) had PFC recurrence during follow-up (P Z .013; relative risk 0.385, 95% CI 0.0940.645). Recurrence occurred a median period of 6 months (IQR, 3.50-8 months) after endoscopic treatment and 4 months (IQR, 1-5.50 months) after group assignment and stent retrieval. Results from the Kaplan-Meier analysis were also similar (log-rank test, P Z .012) (Fig. 2). All patients who had PFC recurrence underwent successful repeat EUS-guided endoscopic transmural drainage with stent insertion and subsequent PFC resolution. The characteristics of the 5 patients with PFC recurrence are given in Table 4.

Adverse effects Endoscopic treatment All patients had successful endoscopic transmural drainage of the PFC with a cystogastrostomy (n Z 18) or a cystoduodenostomy (n Z 10), followed by 1 or 2 double-pigtail stent insertions (7-10F, 3-12 cm). Only 4 patients had PFC drainage without EUS guidance. The median number of endoscopic procedures was 2 (IQR,

Eleven months after initial PFC drainage, 1 patient with CP from group B (stent retrieval) presented with a new PFC in a different location and required transmural drainage. In 2 patients from group A (stent maintenance), stents were removed because of problems connected with stent placement. One of the patients had epigastric pain, and,

612 GASTROINTESTINAL ENDOSCOPY Volume 65, No. 4 : 2007

www.giejournal.org

Arvanitakis et al

Stent removal after endoscopic transmural drainage

Figure 1. Participant flow chart.

on a plain film, the proximal end of the cystogastric stent appeared partially dissected. The stent was retrieved 28 months after initial endoscopic therapy, and the pain subsequently resolved. The other patient underwent ERCP for biliary drainage 24 months after initial endoscopic therapy, and the 2 cystoduodenal stents were retrieved at that time because they were hindering a biliary sphincterotomy procedure. Neither patient had PFC recurrence.

In 3 other patients from group A (stent maintenance) stents migrated during follow-up after a stent placement period of 3 months (n Z 2) and 12 months (n Z 1). No PFC recurrence occurred in these patients. Stent diameter ranged from 7F to 10F, and the length was 4 cm (between pigtails) for all migrated stents. A 54-year old patient with alcoholic liver disease in group B (stent retrieval) died during follow-up, 24 months

www.giejournal.org

Volume 65, No. 4 : 2007 GASTROINTESTINAL ENDOSCOPY 613

Stent removal after endoscopic transmural drainage

Arvanitakis et al

TABLE 1. Baseline patient characteristics of the study population (n Z 46)

Median age, y (IQR)

Group A (n Z 15)

Group B (n Z 13)

P*

50 (44-58)

45 (37-52)

.170

Sex, no. (%) Male Female

Nonrandomized patients (n Z 18) 53 (42-62)

.195

.211

.330

11 (73.3)

12 (92.3)

13 (72.2)

4 (26.7)

1 (7.7)

5 (27.8)

Underlying disease, no. (%)

Py

.620

.154

AP

6 (40)

5 (38.5)

10 (55.6)

CP

9 (60)

8 (61.5)

8 (44.4)

Median PFC drainage time since onset of AP, d (IQR)z

90 (30-135)

60 (30-135)

.792

60 (30-97.5)

.654

Alcohol related AP or CP, no. (%)

12 (80)

12 (92.3)

.356

10 (55.5)

.028

PFC type, no. (%)

.545

.155

PC-AP

3 (20)

3 (23.1)

7 (38.9)

PC-CP

7 (46.7)

8 (61.5)

5 (27.8)

PA

3 (20)

2 (15.4)

6 (33.3)

OLN

2 (13.3)

d

PFC location, no. (%)

d .393

.157

Head

8 (53.3)

4 (30.8)

3 (16.7)

Body

5 (33.3)

5 (38.5)

8 (44.4)

Tail

2 (13.4)

4 (30.7)

7 (38.9)

Median PFC diameter, mm (IQR)

47 (40-70)

76 (48-110)

.186

70 (45-101)

.382

MPD rupture based on MRCP, no. (%)

8 (53.3)

6 (46.2)

.500

9 (50)

.618

MPD rupture based on ERCP, no. (%)x

6 (60)

2 (33.3)

.378

3 (33.3)

.788

Indication for therapy, no. (%) Pain Sepsis

.085 11 (73.3)

12 (92.3)

4 (26.7)

.083 9 (50)

d

2 (11.1) 6 (33.3)

Organ compression

d

1 (7.7)

Increase in size

d

d

1 (5.6)

OLN, Organized liquefied necrosis; PA, pancreatic abscess; PC, pseudocyst. *Difference between groups A and B. yDifference between randomized and nonrandomized patients. zFor the 21 of 46 patients of the study population who have AP (group A, 6; group B, 5; nonrandomized, 10). xFor the 26 of 46 patients of the study population in whom a pancreatogram was obtained (group A, 10; group B, 6; nonrandomized, 10).

Baseline characteristics were similar to those of randomized patients (Table 1). Major complications that required further endoscopic, radiologic, or surgical treatment

occurred in 3 patients. They consisted of sepsis, with subsequent abscess formation that required further endoscopic drainage (n Z 2), and late bleeding from the gastric wall that required angiography and surgery (n Z 1). When regarding the results of the initial therapy, there were no significant differences compared with randomized patients, except for higher rates of nasocystic catheter insertion (11/18 vs 8/28, P Z .030) and

614 GASTROINTESTINAL ENDOSCOPY Volume 65, No. 4 : 2007

www.giejournal.org

after initial drainage, because of biliary sepsis caused by CP-related biliary stricture and biliary-stent obstruction. No PFC recurrence occurred in this patient.

Results for nonrandomized patients

Arvanitakis et al

Stent removal after endoscopic transmural drainage

TABLE 2. Baseline characteristics regarding initial endoscopic therapy in study population (n Z 46) Group A (n Z 15)

Group B (n Z 13)

Type of drainage, no. (%)

P*

Nonrandomized patients (n Z 18)

.055

Py .197

CG

7 (46.7)

11 (84.6)

15 (83.3)

CD

8 (53.3)

2 (15.4)

3 (16.7)

13 (86.7)

11 (84.6)

.644

18 (100)

.125

2 (13.3)

6 (46.2)

.067

11 (61.1)

.030

EUS guidance, no. (%) NCC insertion before stent, no. (%) Cumulative stent diameter, no. (%)

.136

.132

7F

6 (40)

2 (15.4)

1 (5.6)

8.5F

3 (20)

1 (7.7)

3 (16.7)

10F

5 (33.3)

10 (76.9)

12 (66.7)

d

d

2 (11.1)

1 (6.7)

d

d

MPD drainage

9 (60)

6 (46.2)

.362

9 (50)

.526

Major EPS (current or previous)

9 (60)

6 (46.2)

.362

8 (44.4)

.763

Minor EPS (current or previous)

1 (6.7)

2 (15.4)

.199

1 (5.6)

.492

Pancreatic stent, no. (%)

4 (26.7)

2 (15.4)

.400

6 (33.3)

.288

ESWL, no. (%)

3 (20)

3 (20)

.600

2 (11.1)

.314

Minor complications, no. (%)

3 (20)

2 (15.4)

.257

2 (11.1)

.683

Major complications, no. (%)

d

d

d

3 (16.7)

.054

Median no. procedures (IQR)

2 (1-2)

2 (1-2)

.892

2 (1-3)

.903

Length of hospital stay, d (IQR)

7 (3-12)

10 (5-12)

.525

8 (6-22)

.701

18.5F (10F stent and 8.5F stent) 20F (two 10F stents)

CD, Cystoduodenostomy; CG: cystogastrostomy; NCC, nasocystic catheter; EPS, endoscopic pancreatic sphincterotomy; ESWL, extracorporeal shock wave lithotripsy. *Difference between groups A and B. yDifference between randomized and nonrandomized patients.

major complications that required further treatment (3/18 vs 0/28, P Z .054) in nonrandomized patients (Table 2). In this group of 18 patients, the stents were left in place, except for the 4 patients who experienced early spontaneous stent migration and 1 patient whose stent was deliberately retrieved because of symptoms before randomization. All patients were followed prospectively for a median period of 12 months (IQR, 5.25-20.25 months). No patient experienced any PFC recurrence that required therapy (Table 3). In 5 patients, the stents migrated spontaneously after a minimum stent placement period of 4 months (IQR, 2-12 months). The stent diameter ranged from 7F to 10F, and the length was 4 cm (between pigtails) for all migrated stents. In 1 patient, the cystogastric stent was deliberately retrieved after a stent placement period of 24 months. He developed epigastric pain and an ulcer from stent compression, seen in the gastric wall opposite

to the cystogastrostomy during upper-GI endoscopy. In another patient, the cystogastric stent was retrieved after 16 months, during ERCP performed for pancreatic stent exchange, because it had almost completely migrated into the gastric lumen. There was no PFC recurrence in these patients.

www.giejournal.org

Volume 65, No. 4 : 2007 GASTROINTESTINAL ENDOSCOPY 615

Univariate analysis Univariate analysis was applied to the entire study population (n Z 46). It revealed that PFC recurrence was significantly associated with stent retrieval (P Z .002) and a short stent-placement period (P Z .050). There were no differences between patients with and without PFC recurrence with regard to underlying disease, alcoholic etiology, initial PFC size and location, MPD rupture, type of transmural drainage (cystogastrostomy vs cystoduodenostomy), or associated MPD drainage (Table 5). In the subgroup of patients with stent retrieval (n Z 14: 13 patients from group B and 1 nonrandomized patient),

Stent removal after endoscopic transmural drainage

Arvanitakis et al

TABLE 3. Characteristics regarding outcome in study population (n Z 46)

O50% reduction in PFC size on first posttherapeutic assessment by MRI, no. (%) Median interval between PFC drainage and first posttherapeutic assessment by MRI, d (IQR)

Group A (n Z 15)

Group B (n Z 13)

12 (80)

6 (46.1)

.071

11 (61.1)

.535

6 (3-8.5)

.650

5.5 (2.7-11.2)

.991

6 (3-14)

P*

Nonrandomized patients (n Z 18)

Py

PFC resolution, no. (%) At 1 mo

14 (93.3)

11 (84.6)

.444

11 (61.1)

.030

At 3 mo

15 (100)

13 (100)

d

15 (83.3)

.054

Median follow-up since treatment, mo, no. (IQR)

16 (8-20)

11 (7-21)z

.316

12 (6.7-20.2)

.743

Median mo follow-up since randomization, no. (IQR)

12 (7-18)

9 (4.5-18)z

.201

d

d

Median mo stent placement duration, no. (IQR)

12 (3-20)

2 (2-3.5)

.004

4 (2-12)

.820

*Difference between groups A and B. yDifference between randomized and nonrandomized patients. zPFC recurrence occurring 6 mo (IQR, 3.5-8 mo) after treatment and 4 mo (IQR, 1-5.50 mo) after randomization in 5 patients from group B.

Many aspects of endoscopic transmural drainage have yet to be determined, such as optimal stent size and number, as well as stent placement duration. Previous data suggest that stent retrieval should be performed after PFC resolution, based on the rationale that (1) stent occlusion might lead to PFC recurrence, and (2) the stent might act as a foreign body and consequently lead to infectious complications. However, neither the appropriate attitude nor

the optimal stent-placement duration has yet been established. A previous study already showed that a drainage period of more than 6 weeks was associated with a more favorable outcome.21 The present study showed that, contrary to current assumptions based on retrospective studies, systematic stent removal after successful transmural drainage and PFC resolution resulted in higher recurrence rates. The explanation for our findings might be that by maintaining the stents in place, the cystenterostomy tract may remain patent. This patency might not be because of the patency of the stent itself but because the presence of the stent, which acts as a wick, maintains the cystenterostomy open. This is particularly important in patients with complete MPD rupture (disconnected duct syndrome). In these patients, optimal MPD drainage with bridging of the rupture site is very difficult and often impossible, compared with patients with partial MPD rupture.34,35 Therefore, a patent cystenterostomy may help to reduce the pressure in the disconnected MPD segment upstream of the rupture site, by improving drainage of the pancreatic secretions. Thus, pancreatic secretion drainage through a patent cystenterostomy can be successfully demonstrated by S-MRCP after secretin stimulation of the exocrine pancreas.18 In our series, MRCP showed that 4 of the 5 patients with PFC recurrence after stent retrieval had MPD rupture (confirmed as a complete MPD rupture by ERCP in 2 patients in whom a pancreatogram was obtained), thus underlining the need for long-term stent maintenance. According to univariate analysis of the whole study population, only stent retrieval and shorter stent-placement duration seemed to be associated with PFC recurrence. Although, logically, complete MPD rupture should have been associated with a higher risk of

616 GASTROINTESTINAL ENDOSCOPY Volume 65, No. 4 : 2007

www.giejournal.org

Figure 2. Kaplan-Meier analysis of the probability that PFC recurrence would occur according to group allocation for the 28 randomized patients.

PFC recurrence tended to be associated with the presence of MPD rupture (4/5 vs 2/9, P Z .063).

DISCUSSION

Arvanitakis et al

Stent removal after endoscopic transmural drainage

TABLE 4. Characteristics of the 5 patients with PFC recurrence Patient 1

Patient 2

Patient 3

Patient 4

Patient 5

26

35

45

41

42

Female

Male

Male

Male

Male

Underlying disease

AP

CP

AP

CP

CP

Alcohol-related disease

no

yes

yes

yes

yes

PC AP

PC, CP*

PA

PC, CP

PC CP

PFC size, mm

76

80

100

130

35

PFC location

Body

Tail

Body

Body

Body

MPD rupturey

Yes

Yes

Yes

No

Yes

MPD drainage

No

No

No

Yes

Yes

PFC drainage route

CG

CG

CG

CG

CG

O50% reduction in PFC sizez

No

No

No

No

Yes

Stent placement duration, mo

2

2

4

2

2

Pain

Pain, sepsis

Pain

Pain

Pain

6

1

4

5

1

MPD rupture

MPD rupture

MPD rupture

MPD obstructionx

MPD rupture

Age, y Sex

PFC type

Symptoms related to PFC recurrence Interval between stent retrieval and PFC recurrence, mo Presumed cause of PFC recurrence

CG, Cystogastrostomy; PA, pancreatic abscess; PC, pseudocyst. *Moderate CP with no MPD dilation. yBased on MRCP. zBased on the first posttherapeutic assessment by MRI after a median interval of 5 d (IQR, 2-8 d) after PFC drainage. xPatient with calcifying CP and MPD obstruction by stones.

recurrence, this was not suggested by the overall analysis but rather by the results for the subgroup of patients who underwent stent retrieval. The small number of patients who reached the end point might imply that there were limitations to the identification of predictive factors, and, understandably, our findings would have been more firmly established had we been allowed to include a larger number of patients. Nevertheless, if the present results are not convincing enough for indefinite stent maintenance to be accepted as a standard strategy in all patients who undergo PFC transmural drainage, they should at least underline the importance of leaving the stent(s) in position in the case of complete MPD rupture or of a communicating PFC in patients with CP who are undergoing endoscopic therapy. In this connection, this strategy probably affords a unique opportunity to provide a second route for drainage of pancreatic secretions, independent of stent patency. In addition to stent occlusion, another apprehension associated with leaving the stents in situ indefinitely is foreign-body–related infection. However, in our trial, no patient had severe adverse effects, because of stent maintenance throughout a follow-up period exceeding 2 years

in some cases. Nevertheless, the patient cohort is not large enough, nor is follow-up long enough to generalize these results. Almost a third of the study population in whom the stents were maintained presented spontaneous stent migration during follow-up, in agreement with the results of a previous prospective study concerning transmural endoscopic drainage of PFC.27 Stent size, number, and type of drainage (cystogastrostomy vs cystoduodenostomy) were not identified as factors predicting stent migration. We noted with interest that there was no PFC recurrence in these patients, which suggested that stent retrieval and spontaneous stent migration have different pathophysiologic implications. Thus, PFC resolution might entail a progressive joining of the PFC walls, which might, in turn, lead to a gradual migration of the stent toward the GI lumen. On the contrary, stent retrieval occurring before the virtual PFC cavity has completely disappeared might lead to PFC recurrence, especially if communication with the pancreatic duct is still active.8 This suggests that, in some patients, it is the duration of the stent placement that counts and not that the stents are still present or have been retrieved after an adequate stent-placement period.

www.giejournal.org

Volume 65, No. 4 : 2007 GASTROINTESTINAL ENDOSCOPY 617

Stent removal after endoscopic transmural drainage

Arvanitakis et al

TABLE 5. Univariate analysis: factors associated with PFC recurrence in the whole study population (n Z 46) Patients with PFC recurrence (n Z 5)

Patients without PFC recurrence (n Z 41)

Underlying disease, no. (%)

P .585

AP

2 (40)

19 (46.3)

CP

3 (60)

22 (53.7)

4 (80)

30 (73.2)

.609

60 (41-100)

.389

.597

Alcohol-related disease, no. (%) Median PFC size, mm, (IQR)

80 (55-115)

PFC location, no. (%) Head

1 (20)

14 (34.1)

Body

3 (60)

15 (36.3)

Tail

1 (20)

12 (29.3)

MPD rupture based on MRCP, no. (%)

4 (80)

19 (46.3)

.173

MPD drainage, no. (%)

2 (40)

22 (53.7)

.457

5 (100)

28 (68.3)

.173

Type of PFC drainage, no. (%) CG CD

d

13 (21.7)

Median mo stent placement duration, no. (IQR)

2 (2-3)

4 (2-12.5)

.050

Stent retrieval, no. (%)*

5 (100)

9 (21.9)

.002

CD, Cystoduodenostomy; CG, cystogastrostomy. *Stents were retrieved in 14 patients (13 allocated to group B, and 1 nonrandomized, in whom the stent was retrieved because of symptoms from stent placement).

Another interesting aspect of our study is that it constitutes the first prospective evaluation of endoscopic transmural PFC drainage by sequential MRI and MRCP and plain abdomen film assessments at predetermined intervals. MRI and MRCP estimate the presence of residual PFC and its contents, as well as the patency of the transmural tract, more accurately than abdominal US or CT.30 It is noteworthy that 10 of the 46 patients (21.7%) in our prospective series had persistent PFCs 1 month after drainage, and 3 (6.5%) had persistent PFCs 3 months thereafter. These findings suggest that stent retrieval 6 weeks after endoscopic therapy, as generally practiced, might be premature, even if abdominal US is reassuring at that time. 618 GASTROINTESTINAL ENDOSCOPY Volume 65, No. 4 : 2007

There are some limitations to our study. First, as mentioned above, the small sample size might be of concern, despite the significant difference in outcome between groups A and B and the associated factors. Nevertheless, the size of the whole study population (n Z 46), including the nonrandomized patients, is comparable with that of the populations in the 2 previous prospective studies concerning PFC transmural drainage.19,27 Moreover, the absence of PFC recurrence in the nonrandomized patients, to whom the same strategy was applied, as in group A (stent maintenance), corroborates the intention-to-treat comparison between groups A and B. Secondly, with regard to the possibility that the follow-up period was not long enough, most PFC recurrences in previous published retrospective studies occurred during the year after endoscopic drainage, so that our median follow-up of 14 months was probably adequate. Furthermore, group B patients tended to have larger PFCs, with slower resolution rates, than group A patients, even if this difference was not significant. This fact might imply that a larger PFC and a slower resolution predispose to PFC recurrence. These differences might have impacted the results without being detected by statistical analysis because of the small size groups. However, final resolution was observed in all our patients before stent retrieval, and PFC size was not identified as a predictive factor for PFC recurrence. Last, there were fewer cases of organized liquefied necrosis in our series (2/28 [7.14%]) than in the large retrospective series studied by Baron et al.8 This may be explained by the fact that most of these patients had previous pancreatic surgery with necrosectomy and, therefore, were not eligible for our trial. Nevertheless, recurrence rates reach 30% for patients with organized liquefied necrosis, according to published data8; therefore, it would have been interesting to include a higher percentage of patients with this condition. In summary, our findings suggest that stent retrieval within 2 weeks of PFC resolution was associated with a higher rate of PFC recurrence, occurring mostly in patients with MPD rupture. There were no severe adverse effects because of the maintenance of the stents in situ. Furthermore, a third of the stents migrated spontaneously during follow-up, without entailing any complications, thus suggesting that long-term stent placement could be applied for most patients and would be clearly indicated for those having complete MPD rupture or a communicating PFC in the setting of CP. Further prospective cohort studies with longer follow-up could provide information regarding the safety of maintaining the stents in place indefinitely.

DISCLOSURE The authors have no commercial associations (eg, equity ownership or interest, consultancy, patent and licensing agreement, or institutional and corporate www.giejournal.org

Arvanitakis et al

associations) that might give rise to a conflict of interest in relation to this article. Marianna Arvanitakis is the recipient of a Wilson Cook grant for clinical research in GI endoscopy. REFERENCES 1. Baillie J. Pancreatic pseudocysts (part I). Gastrointest Endosc 2004;59: 873-9. 2. Yeo CL, Cameron JL, Sohn TA, et al. Six hundred fifty consecutive pancreaticoduodenectomies in the 1990s: pathology, complications, and outcomes. Ann Surg 1997;226:248-60. 3. Arvanitakis M, Delhaye M, Chamlou R, et al. Endoscopic therapy for main pancreatic duct rupture after silastic-ring vertical gastroplasty. Gastrointest Endosc 2005;62:143-51. 4. Lin BC, Chen RJ, Faung JF, et al. Management of blunt major pancreatic injury. J Trauma 2004;56:774-8. 5. Kloppel G. Pseudocysts and other non-neoplastic cysts of the pancreas. Semin Diagn Pathol 2000;17:7-15. 6. Baron TH, Thaggard WG, Morgan DE, et al. Endoscopic therapy for organized pancreatic necrosis. Gastroenterology 1996;111:755-64. 7. Bradley EL III. A clinically based classification system for acute pancreatitis. Summary of the International Symposium on Acute Pancreatitis, Atlanta, GA, September 11 through 13, 1992. Arch Surg 1993;128:586-90. 8. Baron TH, Harewood GC, Morgan DE, et al. Outcome differences after endoscopic drainage of pancreatic necrosis, acute pancreatic pseudocysts, and chronic pancreatic pseudocysts. Gastrointest Endosc 2002; 56:7-17. 9. Lehman GA. Pseudocysts. Gastrointest Endosc 1999;49:S81-4. 10. Boerma D, van Gulik TM, Obertop H, et al. Internal drainage of infected pancreatic pseudocysts: safe or sorry? Dig Surg 1999;16:501-6. 11. Neff R. Pancreatic pseudocyst and fluid collections: percutaneous approaches. Surg Clin North Am 2001;81:399-403. 12. Cremer M, Devie`re J, Engelholm L. Endoscopic management of cysts and pseudocysts in chronic pancreatitis: long-term follow-up after 7 years of experience. Gastrointest Endosc 1989;35:1-9. 13. Binmoeller KF, Seifart H, Walter A, et al. Transpapillary and transmural drainage of pancreatic pseudocysts. Gastrointest Endosc 1995;42: 219-24. 14. Smits ME, Rauws EAJ, Tytgat GNJ, et al. The efficacy of endoscopic treatment of pancreatic pseudocysts. Gastrointest Endosc 1995;42: 202-7. 15. Catalano MF, Geenen JE, Schmalz MJ, et al. Treatment of pancreatic pseudocysts with ductal communication by transpapillary pancreatic duct endoprosthesis. Gastrointest Endosc 1995;42:214-8. 16. Barthet M, Sahel J, Bodiou-Bertei C, et al. Endoscopic transpapillary drainage of pancreatic pseudocysts. Gastrointest Endosc 1995;42: 208-13. 17. Venu RP, Brown RD, Marrero JA, et al. Endoscopic transpapillary drainage of pancreatic abscess: technique and results. Gastrointest Endosc 2000;51:391-5. 18. Sanchez Cortes E, Maalak A, Le Moine O, et al. Endoscopic cystenterostomy of nonbulging pancreatic fluid collections. Gastrointest Endosc 2002;56:380-6. 19. Beckingham IJ, Krige JEJ, Bornman PC, et al. Long term outcome of endoscopic drainage of pancreatic pseudocysts. Am J Gastroenterol 1999;94:71-4. 20. Harewood GC, Wright CA, Baron TH. Impact on patient outcomes of experience in the performance of endoscopic pancreatic fluid collection drainage. Gastrointest Endosc 2003;58:230-5.

www.giejournal.org

Stent removal after endoscopic transmural drainage 21. Cahen D, Rauws E, Fockens P, et al. Endoscopic drainage of pancreatic pseudocysts: long-term outcome and procedural factors associated with safe and successful treatment. Endoscopy 2005;37:977-83. 22. Hookey LC, Debroux S, Delhaye M, et al. Endoscopic drainage of pancreatic fluid collections in 116 patients: a comparison of etiologies, drainage techniques, and outcomes. Gastrointest Endosc 2006;63: 635-43. 23. Seewald S, Groth S, Omar S, et al. Aggressive endoscopic therapy for pancreatic necrosis and pancreatic abscess: a new safe and effective treatment algorithm. Gastrointest Endosc 2005;62:92-100. 24. Giovannini M, Bernardini D, Seitz JF. Cystogastrostomy entirely performed under endosonography guidance for pancreatic pseudocyst: results in six patients. Gastrointest Endosc 1998;48:200-3. 25. Vilmann P, Hancke S, Pless T, et al. One step endosonography-guided drainage of a pancreatic pseudocyst: a new technique of stent delivery through the echo endoscope. Endoscopy 1998;30:730-3. 26. Giovannini M, Pesenti C, Rolland AL, et al. Endoscopic ultrasoundguided drainage of pancreatic pseudocyst or pancreatic abscesses using a therapeutic echo endoscope. Endoscopy 2001;33:473-7. 27. Kru¨ger M, Scneider AS, Manns MR, et al. Endoscopic management of pancreatic pseudocysts or abscess after an EUS-guided 1-step procedure for initial access. Gastrointest Endosc 2006;63:409-16. 28. Yusuf TE, Baron TH. Endoscopic transmural drainage of pancreatic pseudocysts: results of a national and an international survey of ASGE members. Gastrointest Endosc 2006;63:223-7. 29. Delhaye M, Matos C, Devie`re J. Endoscopic management of chronic pancreatitis. Gastrointest Endosc Clin N Am 2003;13:717-42. 30. Morgan DE, Baron TH, Kevin Smith JK, et al. Pancreatic fluid collections prior to intervention: evaluation with MR imaging compared with CT and US. Radiology 1997;203:773-8. 31. Arvanitakis M, Delhaye M, De Maertelaer V, et al. Computed tomography and magnetic resonance imaging in the assessment of acute pancreatitis. Gastroenterology 2004;126:715-23. 32. Pilleul F, Beuf O. Diagnosis of splanchnic artery aneurysms and pseudoaneurysms, with special reference to contrast enhanced 3D magnetic resonance angiography: a review. Acta Radiol 2004;45:702-8. 33. Matos C, Metens T, Devie`re J, et al. Pancreatic duct: morphologic and functional evaluation with dynamic MR pancreatography after secretin stimulation. Radiology 1997;203:435-41. 34. Telford JJ, Farrell JJ, Saltzman JR, et al. Pancreatic duct placement for duct disruption. Gastrointest Endosc 2002;56:18-24. 35. Varadarajulu S, Noone TC, Tutuian R, et al. Predictors of outcome in pancreatic duct disruption managed by endoscopic transpapillary stent placement. Gastrointest Endosc 2005;61:568-75. 36. Wilson EB. Probable inference, the law of succession, and statistical inference. J Am Stat Assoc 1927;22:209-12. 37. Kaplan E, Meier P. Nonparametric estimation from incomplete observation. J Am Stat Assoc 1958;53:457-81.

Received April 13, 2006. Accepted June 20, 2006. Current affiliations: Department of Gastroenterology (M.A., M.D., O.L.M., J.D.), Department of Radiology (M.A.B., C.M.), Erasme University Hospital, Brussels, Belgium; IRIBHM, Statistical Unit (V.D.M.), Universite´ Libre de Bruxelles, Brussels, Belgium. M. Arvanitakis is the recipient of a fellowship from the ‘‘Fondation Erasme.’’ Reprint requests: Marianna Arvanitakis, MD, Department of Gastroenterology, Erasme University Hospital, Route de Lennik 808, 1070 Brussels, Belgium.

Volume 65, No. 4 : 2007 GASTROINTESTINAL ENDOSCOPY 619