Study characteristics
The systematic search of databases identified 5 723 articles and included a total of 30 studies in the quantitative systematic review and meta-analysis (Fig. 1). The 30 studies were published between 1995 and 2020 and enrolled 1 355 patients with a diagnosis of DPDS (Table 1). Twenty-seven studies were retrospective in nature [2-5,8,13,14,21-40], two studies were prospective in design [9,41], and one study included patients enrolled prospectively and patient data sourced retrospectively [42]. Studies were conducted in the United States (n=19) [2-5,8,9,14,22,25-29,32,35-38,41], India (n=4) [24,33,34,39], Poland (n=2) [30,31], Mexico (n=1) [42], Belgium (n=1) [40], China (n=1) [23], and Japan (n=1) [21]. One study enrolled patient from India and United States [13].
Quality assessment
Quality assessments of the studies are presented in Supplementary Table 1 and 2 in Appendix. Most studies scored between 4-6 points frequently (24/30) [2,4,5,8,9,13,22-24,26-31,33,39,41,42]. Four studies scored 7-8 points [3,14,21,25]. Two studies scored three points [32,40]. Two domains which consistently scored poorly were representativeness of the exposed cohort (13/30) [4,8,23,27-30,32-35,39,40] and assessment of outcome (19/30) [2,4,5,8,9,22,23,26,27,29-31,34-40]. All studies failed to score on the domain of a nonexposed cohort. Similarly, the overall risk of bias were moderate in 14 studies [2,4,9,13,21,22,24,26,28,29,34,37-39], serious in 8 studies [3,5,8,14,25,30,35,40], and critical in 8 studies [23,27,31-33,36,41,42], employing the ROBIN-I tool. The eight studies were at critical overall risk of bias due to significant deficiencies in the domain of confounding bias [23,27,31-33,39,40,42].
Presentation and Diagnosis
PFC was the most common presentation for DPDS (77.3%, 728/942) followed by EPF (18.6%, 175/942), recurrent pancreatitis (3.2%, 30/942), and ascites (1.0%, 9/942) [3-5,8,9,13,14,21-26,28-31,34,37-39,41,42]. Presentation of DPDS was not quantified or reported in 5 studies [2,27,32,35,36]. The total prevalence of walled-of necrosis (WON) and pseudocyst were 74.4% (486/653) and 25.6% (167/653) respectively in 16 studies reporting the type of PFC [3,4,8,13,21,25,26,29-31,34,37,38,40-42]. Acute pancreatitis (95.2%, 918/964) was the most common etiology for DPDS, followed by chronic pancreatitis (3.1%, 30/964) and trauma (1.7%, 16/964) [2-5,9,14,21-26,28,33-42]. The most common site of DPDS was the body of pancreas (47.0%, 117/249). DPDS at the neck (26.5%, 66/249) and head (21.3%, 53/249) of pancreas were also frequently observed [2,9,26,30,36,39,41]. The specific location was not reported in 21 studies (Supplementary Table 3) [3,5,9,13,14,21,23-25,27-29,31-38,42].
Twenty-four studies reported a definition for DPDS [2-5,9,13,14,22-25,27,28,31-33,35-42]. There were only four studies[14,35,37,38] that used the three criteria proposed by Sandrasegaran et al. [6]. Seven studies defined DPDS by two criteria, which were extravasation or cutoff appearance of MPD when injected with contrast material and the demonstration of a viable upstream pancreas on imaging [2,3,5,28,30,41,42]. Three studies required an additional criterion that was nonhealing EPF or PFC [9,22,25]. Eight studies (including five studies on complete duct disruption) used the aforementioned MPD appearance as the sole criterion to define DPDS [23,24,30-33,39,40]. Lastly, two studies used intraoperative findings to define DPDS, however the specific findings were not reported (Supplementary Table 3) [4,36].
Endoscopic Drainage
The average time between the onset of pancreatitis, EPF, or PFC and diagnosis of DPDS was between 56 days and 7.5 months in two studies [22,23]. 17 studies reported on endoscopic drainage of DPDS including transmural, transpapillary, and combined-modality drainage which included a total of 553 patients [2,3,13,14,22-24,27-33,40-42]. The weighted overall success rate for endoscopic drainage was 82.0% (95%-CI: 68.6-90.5%) (Fig.2a). Six studies reported treatment of a total of 62 patients with transpapillary drainage [2,23,27,31,32,40]. The weighted overall success rate in these studies was 58.5% (95%-CI: 36.7-77.4%) (Fig. 2b). Eleven studies reported treatment outcome of transmural drainage which included 381 patients [3,13,24,28,29,32,33,40-42]. Transmural drainage was associated with a weighted success rate of 90.6% (95%-CI: 81.0-95.6%) (Fig 2c). Six of the 11 studies solely performed endoscopic ultrasound (EUS) guided transmural drainage for a total of 84 patients [13,14,28,33,41,42]. The weighted success rate of the six studies were 91.7% (95%-CI: 83.5-96.0%) (Supplementary Fig. 1). Transmural drainage success rate on weighted analysis was significantly higher compared to transpapillary drainage. Nine of the 11 studies reported the duration of transmural stents that were left in-situ [3,13,14,24,28,29,33,41,42]. There were five studies that left stents in-situ indefinitely [28,29,33,41,42], three studies that routinely removed the stents [13,14,22], and one study that removed the stents routinely in initial years of practice but later left the stents in-situ indefinitely [3].
Five studies reported treatment of a total of 131 patients with combined-modality drainage [2,13,30,32,40]. Combined-modality drainage was associated with a weighted success rate of 64.6% (95%-CI: 42.7-81.6%) (Fig. 2d). Two of the five studies routinely removed transmural stents for their patients [2,13]. The other three studies did not report the duration of transmural stent [30,32,40].
Ten studies reported the type of transmural stents that were used [2,3,13,14,24,28,29,33,41,42]. Five studies used only double pig-tail stents [21,24,28,29,42]. Three studies used either solely double pig-tail stents, or metal stents that were later exchanged for double pig-tail stents for long-term drainage [3,33,41]. Lastly, two studies exclusively used metal stents for drainage which were routinely removed [13,14].
Six studies which left transmural stents in-situ indefinitely reported stent-related complications of 19.5% (23/118). All stents left in-situ were double pig-tail stents [21,24,26,28,34,42]. All complications were related to stent migrations except for one patient who experienced stent fragmentation and stent migration [28]. Eleven percent (13/118) of the stent-related complications were asymptomatic or incidental findings and 8.5% (10/118) of the complications were symptomatic. These included bowel obstruction (1.7%, 2/118), bowel perforation (1.7%, 2/118), recurrent PFC (1.7%, 2/118), and infection (3.4%, 4/118) [21,26,28,34,42]. Endoscopic and surgical treatment were needed in 2.5% (3/118) and 0.8% (1/118) of complications respectively [21,26,34,42]. All endoscopic procedures and related outcomes are shown in Supplementary Table 4.
Surgical treatment
The average time interval between onset of pancreatitis, fluid collection, or fistula and surgery was 3.9 to 6.1 months [4,5,9,36]. Surgery was used as the definitive treatment following failure with endoscopic drainage in 22.0% (84/382) of patients [8,36,37]. Ten studies reported on surgical treatment of DPDS including distal pancreatectomy and surgical drainage in 194 and 226 patients respectively [2,4,5,8,9,14,25,35-37]. The weighted overall success rate for surgical treatment was 87.4% (95%-CI: 81.2-91.8%) while the weighted overall success rate of surgical treatment published in the last 10 years was 84.7% (95%-CI: 78.7-89.2%) in five studies (Supplementary Fig. 2a and b) [5,14,25,36,37]. The weighted success rate for distal pancreatectomy and surgical drainage was 86.6% (95%-CI: 77.0-92.6%) and 85.8% (95%-CI: 80.7-89.8%) respectively (Supplementary Fig. 2c and d). No difference was observed on pairwise meta-analysis between distal pancreatectomy and surgical drainage (distal pancreatectomy, 86.8%, 168/194, vs. surgical drainage, 86.3%, 195/226, OR 0.99, 95%-CI: 0.30-3.21, P=0.981) (Fig. 3a). All surgical procedures and related outcomes are shown in Supplementary Table 5.
There was no difference was found between surgical treatment and endoscopic drainage on pairwise meta-analysis (surgery, 69.0%, 20/29 vs. endoscopic drainage, 50%, 16/32, OR 2.23, 95%-CI: 0.09-52.84, P =0.389) (Fig. 3b). Similarly, there was no difference found between weighted overall success rates of surgical treatment and endoscopic drainage (Fig 2a and Supplementary Fig. 2a).
Percutaneous drainage
Five studies reported on percutaneous drainage in 161 patients [2,8,32,38,39]. Percutaneous drainage was uniformly unsuccessful in three studies [2,8,32]. In the other two studies, percutaneous drainage resulted in successful treatment in 97.0% (32/33) and 22.2% (12/54) of patients (Supplementary Table 6).