Pancreatic fluid collections (PFCs) may be caused by acute or chronic pancreatitis (CP), trauma, surgery, or malignancy. Depending on the time of formation and the presence or absence of solid debris, the 2012 revised Atlanta classification defines PFCs as acute peripancreatic fluid collections, acute necrotic collections, pseudocysts, or walled-off necroses [1]. Indications for the drainage of PFCs include pain, gastric outlet obstruction, infection, and biliary obstruction [2-4].

The endoscopic management of symptomatic PFCs, particularly pseudocysts, is safe and effective [5-7]. Options for endoscopic management include transmural drainage (TM) alone, involving either a cyst-gastrostomy or cyst-duodenostomy, transpapillary drainage (TP) alone, involving an endoscopic retrograde pancreatography (ERP) procedure, or a combination of both drainage method (CD). As CD involves an ERP procedure in addition to endoscopic ultrasound (EUS), it may carry risks of additional complications such as ERP-associated pancreatitis, bleeding, and perforation beyond those associated with TM alone. There have been conflicting reports about the optimal method. Whereas in 2006, Hookey et al. found no significant difference in the treatment success rate of 116 patients who had undergone CD versus TM alone (91% vs. 95%), in 2010, a subsequent study of 110 patients by Trevino et al. did show an advantage from CD (97.5% vs. 80%; adjusted relative risk (RR)=1.14; 95% confidence interval [CI], 1.01–1.29; p=0.04) [4,8].

The aim of this study was to perform a meta-analysis of the existing literature to determine the optimal method for the endoscopic drainage of symptomatic PFCs in terms of technical success, clinical success, recurrence, and complications.

Our study used the meta-analysis of observational studies in epidemiology (MOOSE) guidelines [9], which provide a checklist for the reporting of meta-analyses of observational studies in epidemiology.

A total of 1,388 records were retrieved based on our search strategy (Fig. 1). After first excluding 208 review articles, the titles and abstracts of the remaining articles were reviewed to eventually select 11 articles for a full text review. After the full text review, one article was excluded as it contained a series of PFCs that overlapped with those in a previously published study. In total, 9 studies were included in the meta-analysis: 2 prospective observational studies, and 7 retrospective observational studies [4,8,10,13-18]. There have been no case control studies or randomized control trials published on PFC drainage.

The study characteristics of the 9 included observational studies are given in Table 1. The included studies contained populations from Germany, The Netherlands, Belgium, Brazil, and the United States. The mean age of the patients ranged from 38 to 57 years. The mean PFC diameter ranged from 6 to 13 cm. The majority of the studies received a NOS score of 5, with 2 studies scoring 7 or higher (Tables 1, 2). Two of the studies included patients with PFCs other than pseudocysts; however, the majority of the PFCs included in these studies were pseudocysts (81% and 71%) [4,8]. There were no data available in these studies to select only the pseudocyst cases for inclusion in the analyses.

A total of 604 drainage procedures were included in the analysis: 373 (62%) TMs and 231 (38%) CDs. All of the studies reported the rates of clinical success, technical success, recurrence, and complications; however, not all studies stratified these data by drainage route. Therefore, only certain studies could be included for each outcome of interest.

After applying a random-effects model to the data, the CD of PFCs was not shown to be superior to TM alone for any of the four outcomes (Figs. 2-5). With regards to the technical success of the procedures, four studies were included in the analysis, with a pooled OR of 1.12 (95% CI, 0.37–3.37; p=0.85) (Fig. 2). For the primary outcome (clinical success), all 9 studies were included in the model. The resultant pooled OR of 1.11 (95% CI, 0.65–1.89) was not statistically significant (p=0.70) (Fig. 3). An I² value of 26.94% suggested minimal heterogeneity (Fig. 3). For the recurrence rate, 6 studies were included, with a pooled OR of 1.49 (95% CI, 0.24–1.90; p=0.45) (Fig. 4). Finally, no difference was noted between the complications from CD and those from TM (OR, 1.15; 95% CI, 0.61–2.18; p=0.67); however, relevant data were only available from four studies (Fig. 5).

The results of this meta-analysis suggested that transpapillary stenting provides no added clinical benefit for the TM of PFCs. The TM and CD methods were statistically equivalent in terms of technical success, clinical success, recurrence, and complication rates.

Our findings have several explanations and implications. First, the TM drainage of PFCs is associated with a high technical success rate, particularly when performed under EUS guidance. In the studies included in our meta-analysis, the technical success of TM drainage alone varied from 82% to 97%. By contrast, the technical success rate of TP drainage, if defined as the placement of a stent across a pancreatic duct (PD) disruption/leak, is much lower. In two recent studies by the same group, bridging PD stents were successfully placed in only 42%–49% of patients with pseudocysts, and in 5%–17% of patients with walled-off necrosis [8,19]. Common reasons for failure include complete ductal disruption, gastric or duodenal obstruction, failed PD cannulation, or an altered surgical anatomy [8,10]. As such, the majority of the benefits from CD may come from the TM component of the procedure. Even among patients in whom bridging PD stents had been successfully placed, a sub-group analysis of the above series using a multivariate logistic regression showed no additional clinical benefit from PD stenting (OR, 1.699; 95% CI, 0.561–5144; p=0.34) [19]. The mechanistic explanation proposed by Hookey et al. is that pure TM may allow the cyst-enterostomy fistula to remain open longer and therefore to mature, whereas the addition of a transpapillary stent may hinder this process [4].

This study had several strengths. We used a rigorous search strategy to summarize and integrate the results from 8 observational studies using a validated methodology (NOS and MOOSE). This allowed us to compensate for the small sample sizes of the individual studies and to reconcile the conflicting published results. Furthermore, minimal heterogeneity was encountered in our analysis.

The study also presented limitations. Most importantly, we were unable to control for patients with a PD leak on the pancreatogram. In the majority of the included studies, an ERP was performed prior to the attempted TM. If a PD leak was observed on the pancreatogram, the patient underwent transpapillary stent placement, with a stent placed either across the defect or into the PFC itself. Conversely, if no leak was seen on the pancreatogram, the endoscopist would proceed to TM. The decision to perform CD or TM was highly personalized, and was not a random decision. Second, due to the varying follow-up lengths, there was considerable heterogeneity in the definitions of clinical success and recurrence in the studies. As such, studies that adopted less stringent criteria may have reported higher rates of success for all procedures. Nevertheless, as these definitions were consistent within the studies, the relative ratios of success between CD and TM were preserved, and they were accurately reflected in the ORs. However, conclusions about specific time points cannot be inferred from our study. Third, the data were primarily compiled from retrospective observational studies, as there have been no randomized, prospective trials to evaluate the utility of combined modality drainage. Fourth, two of the included studies contained PFCs other than pseudocysts [4,8]. Although the majority of the PFCs included in these studies were indeed pseudocysts (81% and 71%, respectively), data were not available to stratify the outcomes by PFC type and drainage route. We chose not to exclude these studies, as they are large studies that are frequently cited as evidence to support divergent answers to the question at hand. Finally, some of the earlier studies included in the meta-analysis did not use EUS guidance for TM. However, several prospective studies and one meta-analysis have now reported that EUS guidance achieves higher rates of technical success without greater clinical success or more complications than non-EUS guided drainage [20-23].

In conclusion, this meta-analysis supports the notion that TP provides no additional clinical benefit over TM of PFCs, particularly pseudocysts. As a consequence of this finding, patients undergoing TM of symptomatic pseudocysts may not require an ERP procedure, and may therefore be spared the risks associated with it, notably that of post-ERP pancreatitis. Nevertheless, our study is not without limitations, and it is impossible to draw causal links based on this analysis. Future prospective studies should place an emphasis on randomization to ensure that the patients undergoing TM and CD have similar baseline characteristics, particularly in terms of the presence or absence of pancreatic ductal disruption.