A Comparison of Endoscopic Versus Surgical Creation of a Cystogastrostomy to Drain Pancreatic Pseudocysts and Walled-Off Pancreatic Necrosis in 5500 Patients

Abstract

Objective:

To determine the success, morbidity, and mortality rates of endoscopic and surgical creation of pancreatic cystenterostomies for the drainage of peripancreatic fluid collections, pseudocysts with necrotic debris, and walled-off pancreatic necrosis.

Summary Background Data:

Endoscopic methods of cystenterostomy creation to drain pancreatic pseudocysts (with and without necrotic debris) and infected peripancreatic fluid collections are perceived to be less morbid than surgery. Contemporary reports document a very high complication rate with endoscopic methods.

Methods:

A meta-analysis of 5500 patients.

Results:

Open and laparoscopic surgical techniques to drain chronic pancreatic pseudocysts, infected pancreatic fluid collections, and walled-off pancreatic necrosis are more successful with less morbidity and mortality than endoscopic methods.

Conclusions:

In regards to a surgical step-up approach to treat chronic infected pancreatic fluid collections or walled-off pancreatic necrosis, surgical creation of a cystenterostomy is more successful with fewer complications than endoscopic methods and should be given priority if less invasive or conservative methods fail.

Mini-abstract: A meta-analysis is presented that compares the success, morbidity, and mortality rates of endoscopic and surgical creation of pancreatic cystenterostomies for pancreatic pseudocysts.

INTRODUCTION

Complications of pancreatitis are life threatening and frequently require complex treatment algorithms. Historically, surgical management carried a high morbidity with prolonged length of stay and risk of repeat procedures. Therefore, in 2013, the first prospective randomized trial was performed and reported that endoscopic and surgical cystogastrostomy had equal efficacy for drainage of pancreatic pseudocysts (PP).¹ A subsequent study documented similar findings.² While these initial reports dealt exclusively with the drainage of pancreatic pseudocysts, the endoscopic procedure has now been clinically extended, beyond its original intent, for drainage of infected peripancreatic fluid collections (IPFC, includes abscesses, infected pseudocysts, or infected liquefactive necrosis) and walled-off pancreatic necrosis (WON, sterile).^3,4 This has marked a change in philosophy with many acute complications now being treated with internal, rather than external, drainage. While the initial results using endoscopic techniques to drain pseudocysts had been good, these expanded attempts to drain IPFCs and WON have been marred by bleeding complications, perforations, failure to adequately drain infected pseudocysts leading to sepsis, multiple reinterventions, and death.^3,4 Despite these reports, endoscopic drainage of IPFC and WON has increased in volume and fewer surgical cyst-enteric drainage procedures are being performed, although a laparoscopic surgical approach has been shown to have a very short length of stay, a high safety profile, and a low cost.⁵

Referral patterns are evolving away from surgery due to high-perceived morbidity and mortality associated with early operation on necrotizing pancreatitis, before a walled-off or mature pseudocystic collection has formed. Previous reports documented that surgery on necrotizing pancreatitis within 72 hours resulted in mortality rates up to 65%.^6,7 The surgical techniques used in these cases were distal pancreatectomies, transperitoneal approaches with debridement and packing, or transperitoneal resection with lavage.^8–10 Contemporary surgeries are now carried out through videoscopic-assisted retroperitoneal debridement (VARDS), or a transgastric laparoscopic approach, with much less morbidity. Since a pseudocapsule has not yet formed early in the disease process, any attempt at a transperitoneal approach would open the lesser sac and create a vast communication between the retroperitoneum and the abdominal cavity, leading to a high rate of pancreatic fistulas and intraabdominal abscesses in these patients.^11,12 It is now evident that patients who can avoid operative intervention for 12 or more days enjoy a greatly decreased mortality.^13,14 This delay allows enough time for an early pseudocapsule to form within the retroperitoneum and contain the inflammatory process. If surgery can be delayed even further to 4–6 weeks, a true pseudocyst will have formed and a surgical cystogastrostomy can be created with a simultaneous debridement or necrosectomy. This procedure has a dramatically lower mortality risk than an early pancreatic necrosectomy.⁵

Some authors have reported that the highest success rates with endoscopic cystogastrostomy occur when there is radiographic documentation of a well-loculated, purely fluid-filled pseudocyst without any necrotic debris or infection.^15,16 Infected pseudocysts often harbor thick pus and granular material that do not drain well via a small stent. Other reports have also documented that infected pseudocysts, or sterile pseudocysts that become infected after endoscopic stenting, often plug the stent with pus and debris, which allows the infection to persist, and can perpetuate ongoing sepsis and led to death.^17,18 Thus, it would appear that endoscopic cystogastrostomy has proved its effectiveness for uncomplicated pseudocysts, while surgical cystogastrostomy would be more appropriate for infected pseudocysts or those with a lot of necrotic debris. Academic medical centers have standardized their approach and refer to “step up” algorithms, which match techniques to individual patient conditions.¹⁹ However, this approach is not what is always occurring in the clinical setting, and the belief that endoscopic therapy is less morbid than surgical therapy has allowed the expansive use of endoscopy for mature cystic collections of every type. Several recent reports aggregating patients from the last 10 years have documented a substantial morbidity and mortality with endoscopic cystogastrostomy for IPFCs and WON, far above what has been historically reported for surgical cystogastrostomy.^15–20

Patients with pancreatitis can develop a wide-spectrum of IPFCs or WON. Each collection will vary in the amount of debris, fluid, infection, loculation, anatomical position, necrosis, and size. These patients may be hemodynamically stable or in septic shock, with varying degrees of organ failure. Research reporting on an isolated minority of homogenous fluid collections, for example, WON or PP alone, will have meaningful results that would apply to only that population, which does not reflect the variety of mature peripancreatic fluid collections that are seen in actual practice. Complicating the analysis is the fact that most published reports on endoscopic stenting of pancreatic pseudocysts include patients with infection and WON as well.^{17–19,21–26} Selection bias enters most comparisons as many endoscopic reports document results that are performed on stable patients in both the inpatient and outpatient settings, while surgical results contain both stable and septic patients who are hospitalized.^22–26 This would, in theory, favor the endoscopic results. An appropriate way to determine the true morbidity, reintervention risk, and success of this heterogeneous group is to analyze all of the patient data together, using volume to create uniformity in the patient populations. Once the appropriate population is defined, one needs to compare analogous procedures, those being endoscopic cystogastrostomy with surgical cystogastrostomy, and to not include early pancreatic debridement or trans-peritoneal debridement.

There have been no recent reports in the last 5 years comparing the results of endoscopic therapy with surgical therapy for the drainage of all mature PP, IPFCs, and WON taken as a whole. It was the authors’ hypothesis that, based upon the myriad of case reports documenting complications from endoscopic therapy, surgical therapy, using either open or laparoscopic techniques, may actually have less morbidity with a higher success rate. With this in mind, the authors sought to review all of the reported series of patients undergoing endoscopic or surgical drainage of mature PP, IPFCs, or WON over the last 20 years, and determine the optimal approach based upon successful drainage, complications, reinterventions, and mortality.

METHODS

Data Collection

An online PubMed and internet review of all articles dealing with mature PP, IPFCs, or WON over the last 20 years were indexed. All English language articles in adult or pediatric populations were reviewed, and German, Russian, French, and Spanish articles if an English translation of the article could be procured. All series that had 5 patients or more who underwent endoscopic, laparoscopic, or open surgical therapy for mature PP, IPFCs, or WON were included. Authors who published many series containing the same patient population had only their largest series included to avoid duplications. Series that reported external catheter drainage were excluded. Case reports or series with less than 5 patients were excluded. Previously published meta-analysis or review articles were excluded. Series dealing with early pancreatic debridements, and not mature cystic collections, were excluded. Surgical reports of transperitoneal debridements were excluded since this is an outdated method and not technically comparable to endoscopic cystogastrostomy. All etiologies of pancreatic disease were included. Endoscopic and surgical reports that described any type of cystenterostomy creation, including cystogastrostomy, cystoduodenostomy, or cystojejunostomy, were included.

Since the majority of initial reports of endoscopic cystenterostomy began to be published 10 years ago, almost all of the endoscopic articles are from the last 10 years. To have a large enough comparison group and because laparoscopic and open surgical procedures have been performed for a longer time period, the surgical literature search was extended back 20 years.

Abstracts were examined for patient series of endoscopic or surgical drainage of mature PP, IPFCs, and WON since all of these disease processes are addressed by both open surgical, laparoscopic, and endoscopic methods in practice, and the authors were seeking “real world” results. After all of the abstracts were reviewed, those matching the topics of interest were isolated, and the full-length manuscript was obtained. Each article was read in its entirety, and the following information was extracted: number of patients, age, gender, etiology of pancreatitis, morbidity, and bleeding complications, resolution of the fluid collection, recurrence of fluid, reintervention, length of stay, follow-up, and mortality. Any kind of reintervention, endoscopic or surgical, to address a persistent fluid collection after the initial procedure, including postprocedure abscess or other complications such as bleeding, was counted as a reintervention. The only exception was endoscopic stent removal after successful drainage, which was not counted as a reintervention since this is a planned necessity poststenting. Resolution was defined as successful drainage and clinical improvement after a single intervention. The term “clinical resolution,” often used in the literature but having a variety of meanings, was not assumed to be true resolution unless there was associated radiographic documentation that the cyst had been drained to less than a 2-cm diameter (as suggested by other authors), and no further interventions were required because the patient was asymptomatic and without complications.^27–29 If such documentation was not described in the manuscript or if no radiological follow-up was obtained in the study group, the resolution was labeled as “NR,” not reported. The number of patients in each study who had unsuccessful attempts at endoscopic drainage did not equal the number of reinterventions because many patients had reinterventions for complications, and many had more than 1 reintervention.

End Points

The primary endpoint of this study was to evaluate the efficacy (cyst resolution) and safety (complications, adverse events) of endoscopic cystenterostomy and surgical cystenterostomy. Secondary endpoints were reintervention rates, fluid or symptom recurrence, and mortality.

Definitions

According to the revised Atlanta criteria, PP were defined as collections of fluid that were observed after a period of more than 4 weeks, were encapsulated within a well-defined inflammatory wall, and contained no solid components or necrosis.³⁰ Walled-off pancreatic necrosis (WON) was defined as a mature, encapsulated collection of pancreatic and peripancreatic necrosis that presented with a well-defined inflammatory wall and was present for more than 4 weeks, and clinically presumed to be sterile. Infected peripancreatic fluid collections (IPFCs) were PP or WON or other chronic, encapsulated, noncategorized fluid accumulations with documented infection. Many of the articles described “pseudocysts” with various amounts of necrotic debris, that is, 30% or 50% solid component, but since all of these collections were being included in this review, the exact naming of PP versus WON did not make any difference. Recurrence was defined as the presence of a new pseudocyst or reexpansion of the original cyst on imaging after successful treatment and shrinkage of the initial PP or WON.

Statistical Analysis

A statistical analysis was performed using a random effects model to account for between study heterogeneity. Specifically, we used a random intercept logistic regression model for each binary outcome of interest (bleeding, resolution, reintervention, recurrence, morbidly, and mortality).³¹ The model was fit using the LME-4 package in R, which uses maximum likelihood techniques.^32,33 For each outcome, we obtained estimates, standard errors, and 95% confidence intervals (CI) for the mean proportion of individuals with that outcome after receiving surgery or endoscopy. We also obtained estimates, standard errors, 95% CI, and P values (testing for a difference from 0) for the log odds ratio between surgery and endoscopy for that outcome. For each continuous outcome of interest (ie, length of stay and follow-up), we estimated the overall mean using a random intercept model with inverse variance weighting for pooling each study’s mean.³⁴ As we were unable to obtain standard deviations for all the individual studies, we used conservative estimates of the studies’ standard deviations in our analysis. The model was fit using the meta package in R.^33,35 For each outcome, we obtained estimates, standard errors, and 95% confidence intervals for the mean after receiving surgery and endoscopy. We also obtained estimates, standard errors, 95% confidence intervals, and P values (testing for a difference from 0) for the difference in means between surgery and endoscopy for that outcome.

To evaluate if the endoscopic and surgery patient populations were similar, we compared the age, gender, and etiology of pancreatitis between the groups. Additional comorbidity data were very limited or absent in most of the endoscopic manuscripts, which limited further comparison of other patient characteristics. Institutional review board approval for this study was obtained from the St Joseph Mercy Health System.

RESULTS

We identified 1346 articles that covered the topics of interest, and from those, there were 27 surgical articles^20,36–61 and 58 endoscopic articles^{15–18,20–29,59,62–104} that met criteria. From these 85 articles, we reviewed the results for 5588 patients. See Tables 1 and 2 and PRISMA flow diagram.

TABLE 1.

Results of Surgical Cystenterostomy

Author	n	Bleeding	Resolution*	Reintervention	Recurrence	LOS	Follow-up	Morbidity	Mortality
Boutros	19	0	19 (100%)	0	0	7 d	31 mos	NR	0
Palanivelu	100	2 (2%)	100 (100%)	2 (2%)	1 (1%)	6 d	54 mos	NR	0
Yin	12	1 (8%)	12 (100%)	1 (8%)	1 (8%)	NR	108 mos	3 (25%)	0
Hauters	12	0	12 (100%)	0	1 (8%)	6 d	12 mos	1 (8%)	0
Parks	28	0	28 (100%)	0	0	7 d	NR	10 (36%)	0
Hindmarsh	15	0	15 (100%)	0	2 (13%)	NR	37 mos	2 (13%)	0
Bansal	134	3 (2%)	128 (96%)	0	0	4 d	66 mos	14 (10%)	0
Khaled	54	0	53 (98%)	0	1 (2%)	8 d	18 mos	12 (23%)	1 (2%)
Gibson	44	1 (2%)	44 (100%)	1 (2%)	1 (2%)	6 d	74 mos	1 (2%)	0
Hamza	29	0	29 (100%)	0	2 (7%)	3 d	15 mos	1 (3%)	0
Melman	38	1 (3%)	35 (92%)	1 (3%)	1 (3%)	9 d	10 mos	10 (25%)	0
Barragan	8	0	8 (100%)	0	0	6 d	NR	0	0
Obermeyer	6	0	6 (100%)	0	0	5 d	44 mos	1 (17%)	0
Mori	14	0	14 (100%)	1 (7%)	1 (7%)	NR	NR	NR	0
Saluja	20	0	20 (100%)	0	0	5 d	24 mos	2 (10%)	0
Robin	119	2 (1.7%)	111 (93%)	4 (3.4%)	9 (7.6%)	10 d	12 mos	36 (30%)	2 (1.7%)
Martinez	111	6 (5%)	111 (100%)	0	7 (6%)	NR	NR	25 (23%)	2 (1%)
Marino	48	0	44 (92%)	5 (10%)	0	10 d	NR	5 (10%)	0
Gerin	11	1 (9%)	11 (100%)	0	1 (9%)	16 d	10 mos	2 (18%)	0
Crisanto	20	0	19 (94%)	0	0	7 d	40 mos	1 (6%)	0
Simo	22	0	22 (100%)	0	0	14 d	2 mos	1 (4.5%)	0
Ramachandran	5	0	5 (100%)	0	0	4 d	12 mos	0	0
Oida	7	0	7 (100%)	0	0	8 d	65 mos	0	0
Kulkarni	22	0	20 (90%)	1 (4.5%)	1 (4.5%)	7 d	12 mos	7 (32%)	0
Malik	12	0	12 (100%)	0	0	4 d	NR	2 (17%)	0
Yoder	13	1 (8%)	12 (92%)	1 (8%)	0	4.5 d	NR	1 (8%)	0
Driedger	178	6 (3%)	NR	23 (13%)	12 (7%)	8 d	21 mos	68 (38%)	4 (2%)
Total	1101

*Resolution was defined as success after a single procedure.

NR indicates not reported.

TABLE 2.

Results of Endoscopic Cystenterostomy

Author	n	Bleeding	Resolution*	Reintervention	Recurrence	LOS	Follow-up	Morbidity	Mortality
Teoh	59	3 (5%)	NR	20 (34%)	2 (3.4%)	11 d	11 mos	5 (8.4%)	2 (3%)
Shekhar	100	4 (4%)	89 (89%)	11 (11%)	1 (1%)	NR	12 mos	9 (9%)	0
Melman	45	2 (4%)	23 (51%)	16 (36%)	NR	NR	NR	7 (16%)	0
Rasmussen	22	NR	13 (59%)	9 (41%)	4 (18%)	NR	NR	NR	0
Yamamoto	9	1 (11%)	7 (78%)	2 (22%)	NR	NR	NR	2 (22%)	0
Attam	10	1 (10%)	9 (90%)	10 (100%)	2 (20%)	NR	NR	3 (30%)	1 (10%)
Nelsen	5	0	2 (40%)	0	3 (60%)	NR	10 mos	0	0
Sharaiha	124	2 (1.6%)	NR	30 (24%)	6 (5%)	NR	3 mos	23 (19%)	0
Will	27	1 (4%)	NR	27 (100%)	NR	NR	NR	3 (11%)	1 (3.7%)
Ahlawat	12	0	NR	3 (25%)	2 (17%)	NR	4 mos	3 (25%)	0
Xu	108	6 (5.5%)	NR	8 (7%)	9 (8.3%)	NR	6 mos	8 (7%)	4 (3.7%)
Gambitta	91	1 (1%)	73 (80%)	24 (26%)	17 (19%)	NR	NR	5 (5%)	2 (2%)
Laique	88	2 (2%)	70 (80%)	10 (11%)	6 (7%)	6 d	14 mos	19 (22%)	2 (2%)
Garg	30	3 (10%)	23 (77%)	27 (90%)	5 (17%)	NR	22 mos	7 (23%)	0
Shin	27	2 (7%)	NR	NR	9 (30%)	NR	8 mos	6 (22%)	0
Saluja	35	0	27 (78%)	6 (17%)	NR	6 d	24 mos	10 (29%)	0
Yamauchi	36	NR	NR	NR	NR	NR	56 mos	20 (56%)	2 (6%)
Yao	125	5 (4%)	113(90%)	17 (14%)	3 (2.4%)	NR	26 mos	24 (19%)	2 (1.6%)
Yang	205	NR	186 (91%)	19 (9%)	29 (14%)	5 d	6 mos	28 (13%)	0
Puga	41	6 (15%)	NR	NR	NR	NR	29 mos	11 (27%)	0
Venkata	116	1 (0.8%)	110 (95%)	8 (7%)	1 (0.8%)	3 d	6 mos	13 (11%)	4 (3.4%)
Brimhall	249	20 (8%)	134 (54%)	11 (4.4%)	13 (5.2%)	NR	NR	51 (20%)	2 (0.8%)
Ge	52	2 (4%)	NR	8 (15%)	NR	8 d	NR	4 (8%)	0
Keane	109	2 (2%)	76 (70%)	31 (28%)	8 (7%)	4 d	11 mos	13 (12%)	0
Siddiqui	80	6 (7.5%)	72 (90%)	60 (75%)	1 (1%)	NR	8 mos	13 (16%)	NR
Yuan	47	NR	NR	NR	3 (6%)	11 d	48 mos	18 (38%)	0
Bapaye	21	1 (5%)	21 (100%)	1 (5%)	0	NR	1.5 mos	2 (10%)	0
Walter	61	4 (7%)	51 (83%)	23 (38%)	NR	NR	NR	21 (34%)	0
Chandran	47	3 (5.5%)	36 (77%)	10 (21%)	4 (8.5%)	NR	8 mos	26 (55%)	4 (8.5%)
Lin	90	4 (4%)	85 (94%)	7 (8%)	5 (5.5%)	10 d	48 mos	13 (14%)	0
Ng	61	1 (2%)	46 (75%)	13 (21%)	6 (10%)	NR	11 mos	16 (26%)	0
Due	10	NR	6 (60%)	4 (40%)	4 (40%)	NR	4 mos	4 (40%)	1 (10%)
Bang	122	3 (2%)	102 (84%)	21 (17%)	NR	NR	NR	7 (6%)	1 (1%)
Mukai	89	2 (2%)	70 (79%)	37 (42%)	NR	NR	24 mos	11 (12%)	3 (3%)
Smoczynski	97	19 (20%)	78 (80%)	35 (36%)	19 (20%)	29 d	31 mos	29 (30%)	2 (2%)
Kato	67	NR	NR	11 (16%)	16 (24%)	NR	34 mos	1 (1.5%)	0
Binmoeller	14	1 (7%)	11 (79%)	5 (36%)	NR	NR	3 mos	3 (21%)	0
Rische	40	NR	NR	106 (265%)†	4 (10%)	30 d	31 mos	14 (35%)	3 (7.5%)
Weilert	18	0	14 (78%)	8 (44%)	NR	NR	NR	6 (33%)	3 (17%)
Fabbri	20	NR	17 (85%)	2 (10%)	1 (5%)	NR	20 mos	3 (15%)	0
Puri	40	1 (2.5%)	39 (98%)	1 (2.5%)	1 (2.5%)	NR	48 mos	3 (8%)	0
Seewald	80	12 (15%)	58 (73%)	22 (28%)	9 (11%)	NR	31 mos	21 (26%)	0
Varadarajulu	211	3 (1%)	180 (85%)	34 (16%)	NR	3 d	12 mos	17 (8%)	5 (2%)
Ahn	47	1 (2%)	42 (89%)	7 (15%)	5/41 (12%)	NR	17 mos	5 (11%)	0
Gornals	19	0	NR	13 (68%)	3 (16%)	NR	12 mos	5 26%)	0
Cavallini	55	2 (4%)	43 (78%)	12 (22%)	8 (14%)	NR	34 mos	9 (16%)	1 (1.8%)
Sharma	144	8 (5.5%)	140 (97%)	24 (17%)	13 (9%)	NR	18 mos	26 (18%)	1 (0.7%)
Petrone	67	5 (7%)	NR	14 (21%)	2 (3%)	NR	NR	16 (24%)	NR
Aburajab	47	0	44 (94%)	4 (8.5%)	4 (9%)	NR	NR	12 (26%)	0
Ruckert	51	2 (4%)	22 (43%)	25 (49%)	7/30 (23%)	NR	42 mos	16 (31%)	3 (6%)
Wantanabe	103	2 (2%)	80 (78%)	37 (36%)	10/75(13%)	NR	38 mos	15 (15%)	3 (3%)
Nabi	30	4 (13%)	28 (93%)	7 (23%)	2 (7%)	NR	28 mos	10 (33%)	0
Ang	49	1 (2%)	36 (73%)	21 (43%)	NR	NR	NR	5 (10%)	0
Vazquez	211	15 (7%)	178 (84%)	33 (16%)	NR	NR	NR	44 (21%)	NR
Thompson	60	8 (13%)	52 (86%)	21 (35%)	8 (13%)	NR	17 mos	NR	0
Dhir	47	NR	42 (89%)	12 (26%)	2 (4%)	NR	10 mos	4 (9%)	0
Fugazza	311	22 (7%)	272 (87%)	56 (18%)	27 (9%)	NR	5 mos	74 (24%)	2 (0.6%)
Parsa	306	12 (4%)	281 (92%)	45 (15%)	NR	13 d	6 mos	53 (17%)	0
Total	4487

*Resolution was defined as success after a single procedure

^†Every patient had at least 2 repeat interventions.

NR indicates not reported

The patients within the endoscopy and surgery groups were comparable, with similar age (surgery mean 48 years, CI, 44–52, endoscopy mean 51 years, CI, 49–53), gender (surgery males 69%, CI, 62–75%, endoscopy males 68%, CI, 65–71%), and percentage of the group having alcohol as the etiology of their pancreatitis (surgery 29%, CI, 23–35%, endoscopy 37%, CI, 32–42%). See Table 3 for results from significance tests comparing these patient populations, and see Tables 4 and 5 for raw data.

TABLE 3.

Demographics Comparison

	Surgery	Endoscopy	P
Mean age	48, CI, 44–52	51, CI, 49–53	0.12
Male gender	69%, CI, 62–75	68%, CI, 65–71	0.87
Alcohol etiology	29%, CI, 23–35	37%, CI, 32–42	0.06

TABLE 4.

Demographics of Surgical Cystenterostomy

	n	Mean Age	Gender % Male	Alcohol Etiology(%)
Boutros	19	56	63	42
Palanivelu	100	44	70	19
Yin	12	38	64	46
Hauters	12	46	42
Parks	28	50	55	33
Hindmarsh	15	59	33	20
Bansal	134	36	86	25
Khaled	54	53	70	40
Gibson	44	54	66	32
Hamza	29	53	64	39
Melman	38	49	55
Barragan	8
Obermeyer	6	39	67	33
Mori	14	55	79	29
Saluja	20	37
Robin	119	52	90	45
Martinez	111		74
Marino	48	48
Gerin	11	61	67	23
Crisanto	20	40	54	18
Simo	22	51	65	29
Ramachandran	5	56	55	14
Oida	7	61	60	0
Kulkarni	22	51	100	71
Malik	12	52	59	14
Yoder	13	10	75	0
Driedger	178	51	64	31
Total	1101

TABLE 5.

Demographics of Endoscopic Cystenterostomy

Author	n	Mean age	Gender % Male	Alcohol Etiology (%)
Teoh	59	45	81	34
Shekhar	100	56	59	38
Melman	45	52	71
Rasmussen	22	51	59	41
Yamamoto	9
Attam	10	53	80	10
Nelsen	5	57	100
Sharaiha	124
Will	27
Ahlawat	12
Xu	108	52	56	18
Gambitta	91
Laique	88	53	60	48
Garg	30	38	73	27
Shin	27	56	81
Saluja	35	37	80	34
Yamauchi	36	54	83	47
Yao	125	48	66
Yang	205	55	63	32
Puga	41
Venkata	116	53	67	36
Brimhall	249	48	65	31
Ge	52	50	56
Keane	109	55	55	19
Siddiqui	80
Yuan	47	48	62	43
Bapaye	21	47	86	63
Walter	61
Chandran	47	51	68	38
Lin	90	49	67	16
Ng	61	49	70	66
Due	10	53	80
Bang	122	49	52	38
Mukai	89	56	81	56
Smoczynski	97
Kato	67	56	75
Binmoeller	14	50	64	50
Rische	40	56	75
Weilert	18	50	67	44
Fabbri	20	57	85	35
Puri	40	39	78	60
Seewald	80	56	61	10
Varadarajulu	211	48	58	33
Ahn	47	46	85	68
Gornals	19	57	68	32
Cavallini	55	54	55	4
Sharma	144	36	82	47
Petrone	67	59	69	12
Aburajab	47
Ruckert	51
Wantanabe	103	55	71	50
Nabi	30		73
Ang	49	53	51
Vazquez	211
Thompson	60	54	60	28
Dhir	47
Fugazza	311
Parsa	306	55	70
Total	4487

For the outcome of postprocedure bleeding complications, the percentage of surgical patients with a bleeding complication was 1.8% (CI, 0.0097–0.0342) and for endoscopy was 4.3% (CI, 0.0337–0.0558). There was a significantly decreased risk of postoperative bleeding after surgery, with a log odds ratio of –0.906 (SE 0.362, P = 0.012; see Figs. 1 and 2).

FIGURE 1.

Estimated odds ratio of each parameter listed for surgery compared to endoscopy.

FIGURE 2.

Estimated proportion of each parameter listed for surgery compared to endoscopy.

When comparing the chance of complete pseudocyst resolution after having surgery as compared to endoscopic intervention, surgical patients had complete resolution of their cyst 98.5% (CI, 0.959–0.995) of the time, compared to 83.5% (CI, 0.796–0.868) after endoscopic interventions. There was a much higher chance of having complete pseudocyst resolution after surgery, with a log odds ratio of 2.64 (SE 0.582, P < 0.0001).

In terms of repeat interventions postprocedure, including repeat interventions to drain persistent fluid collections, repeat interventions for bleeding, occluded stents, abscesses, or other complications, the risk of repeat interventions after surgical drainage was 3.9% (CI, 0.0266–0.0572) and after endoscopic procedures was 25.8% (CI, 0.1934–0.3354). This does not include repeat endoscopic intervention for planned stent removal after pseudocyst resolution, which is an expected event. Thus, surgical patients were at a greatly reduced chance of requiring any type of reintervention with a log odds ratio of –3.33 (SE 0.508; P < 0.0001).

The risk of pseudocyst recurrence, when given as a statistic, must be indexed for the length of time that a patient is followed, as the risk increases over time. Without this adjustment, the results would be skewed in favor of the group with shorter follow-up because pancreatitis tends to be a chronic and recurrent disease. Therefore, the results of pseudocyst recurrence are examined in conjunction with length of follow-up. The surgical patients had much longer follow-up, with a mean of 34.0 months (CI, 20.5–47.3). Mean endoscopic follow up was 19.4 months (CI, 15.0–23.8). The difference was small, but statistically significant (P = 0.04). However, despite the longer follow-up in the surgical group, the recurrence risk was much less at 2.3% (CI, 0.0124–0.0439) compared with the recurrence risk with endoscopic intervention, which was 8.8% (CI, 0.0679–0.1130). Thus, even taking into account the differences in follow-up, the risk of pseudocyst recurrence is much less with surgical cystogastrostomy, with a log odds ratio of –1.405 (SE 0.366, P < 0.001).

Although surgical cystogastrostomy is often cited as “more risky” than endoscopic therapy, the overall morbidity with surgery was actually less than with endoscopic stenting (13.5%, CI, 0.0934–0.1913 versus 17.8%, CI, 0.1509–0.2089). The log odds ratio slightly favored surgery at –0.326, although not statistically significant (SE 0.232, P = 0.16). Both procedures had a very low mortality, with surgery slightly less at 0.5% (CI, 0.00121–0.0201) compared with endoscopic interventions at 0.9% (CI, 0.0051–0.0157). Again the log odds ratio slightly favored surgery at –0.673 but was not significant (SE 0.943, P = 0.48).

There was no significant differences in postprocedure length of hospital stay, with a mean length of stay after surgery of 7.1 days (CI, 5.9–8.4) and a mean length of stay after endoscopic interventions of 9.4 days (CI, 4.2–14.5, P value for the difference = 0.61).

DISCUSSION

Despite the common assumption, the endoscopic therapy is safer and less morbid than surgical intervention on the pancreas; the results of this study overwhelmingly favor surgery for several important endpoints. The data could be used to challenge the assumption that surgery can always be performed if endoscopic therapy fails, as this logic is clearly flawed. Failed endoscopic therapy increases the chances of bleeding, prolongs the duration of sepsis for IPFCs, and has a greater morbidity than surgical intervention alone. Other authors have also shown that a patient’s risk of complications goes up when surgery is performed after failed endoscopic therapy, as opposed to those who went directly to surgery.³⁶

From a pathophysiological perspective, a large surgical cystogastrostomy will allow ongoing pancreatic fluid from ductal disruption to drain into the stomach until the pancreas heals. In those patients who develop a pancreatic pseudocyst due to a significant disruption in the duct of Santorini or the duct of Wirsung, an ongoing pancreatic leak and fistula can be expected, and are slow to heal even if the duct is stented by ERCP.⁹³ For these patients who have had a surgical cystogastrostomy or cystoduodenostomy, the ongoing fistula will continue to drain into the foregut and the output will be controlled internally. However, if these patients were treated by an endoscopic stent, the stent will need to remain in place indefinitely until the leak resolves, which can take up to 13 months or longer^93,105 Some authors even recommend leaving the stents in place indefinitely.¹⁰⁶ This is problematic since the longer a stent stays in place, the higher the complication rate from erosion into the surrounding vessels or viscera.^{65,67,71,75,84,88}

Historically, all pseudocyst walls are biopsied at the time of cystogastrostomy creation to ensure that the pseudocyst is not really a cystadenocarcinoma of the pancreas. In all of the endoscopic series reviewed in this article, very few of the endoscopic reports described performing a biopsy at the time of cystogastrostomy creation. Other authors have recommended sampling and testing the cyst fluid for carcinoembryonic antigen or CA 19-9 to ensure no malignancy is present.⁹⁴ While the risk of missing a cancer is considered low, it is not zero, and cases of pancreatic cancer are still being reported from presumed pseudocysts.^95,107,108 A recent report found a 1.8% incidence of pancreatic cancer by brush cytology of WOPN, while another found a frequency of 1.25% in peripancreatic fluid collections.^96,109

Many endoscopic cystogastrostomy studies document “clinical success,” which they define as symptomatic improvement only, although some authors consider “success” to be any decrease in the size of the pseudocyst by more than 1 cm.⁷⁵ Other authors have defined “clinical success” to be resolution of symptoms and a decrease in the size of the pseudocyst by >40%.⁸⁰ While other authors define success as the pseudocyst shrinking to less than 3 cm.⁶⁷ Others think that clinical success is a size reduction of >50% or to a size <5 cm.^83,110 While still other authors feel that clinical success is defined by “a distinct decrease in the size of the cyst after a drainage period of six months.”⁹⁷ This variable and misleading idea of success leads to the propagation of the belief that endoscopic techniques are more successful than they actually are. We only found 5 endoscopic reports that documented follow-up CT scans or ultrasound imaging in all of their patients.^{22,24,91,97,98} Not surprisingly, all of these studies reported very high rates of reintervention for the persistent pseudocysts. Not only is the idea of “clinical success” misleading, but it also can be dangerous since we know from the natural history of pseudocysts that their persistence often leads to vascular erosion and hemorrhage, as one of the authors has previously reported.¹¹¹ The presence of metal stents within the pseudocyst also appears to significantly increase the risk of hemorrhage.^{24,99,100,109}

Multiple studies from the literature show that endoscopic stenting nearly always fails when there is greater than 50% parenchymal necrosis.^101,112,113 Three multicenter trials in patients with primary pancreatic necrosis found that attempts at endoscopic necrosectomy had very high morbidity (26%–33%) and mortality rates (6%–11%).^112–114 Another recent multicenter study involving 333 endoscopic procedures for peripancreatic fluid collections, many with a large necrotic component, found an adverse event rate of 24%.¹⁰² Thus, while endoscopic cystogastrostomy with stenting is useful for PP that are purely cystic with minimal necrotic debris, any IPFC or WON with a large solid or granular component or >50% parenchymal necrosis will be better served going directly to surgery. Surgery as primary therapy has less morbidity, mortality, recurrence, reintervention, and greater resolution than endoscopic treatments.

This study is limited by the fact that most of the reviewed articles included in the analysis were retrospective studies, and only a few were prospective trials. Thus, the analysis carries over the retrospective biases present in the original articles, including potential confounding variables and selection bias due to nonrandomization. All of the results can be interpreted as associational. However, the causal interpretation should be further assessed. Also, grouping together the disease processes into 1 large group may lead to potential bias. For example, if a majority of papers focus on a specific type of peripancreatic fluid collection, then the results will also be weighted toward results from that process. We did find, fortunately, that the results are not strongly influenced by any single paper within the study.

This article has ramifications for policy makers, administrators, and those institutions that have begun to follow a step-up approach to IPFCs, since endoscopic therapy is not a benign procedure and carries greater morbidity and mortality risk than surgery for the creation of a cystogastrostomy.