Immediate Versus On‐Demand Direct Endoscopic Necrosectomy for Necrotizing Pancreatitis: A Systematic Review and Meta‐Analysis

ABSTRACT

Background

The optimal timing of direct endoscopic necrosectomy (DEN) for Necrotizing Pancreatitis (NP) remains unclear. This meta‐analysis aimed to evaluate the efficacy and safety of immediate DEN compared with on‐demand DEN.

Methods

PubMed, Embase, Web of Science, and Cochrane Library databases were systematically searched to identify studies. The primary outcome measure was clinical success.

Results

Seven studies (4 RCTs and 3 observational studies) comprising 616 patients were included. Immediate DEN did not demonstrate a statistically significant advantage in clinical success (OR: 1.92, 95% CI 1.01–3.63). No significant differences were found in technical success (OR: 1.62, 95% CI: 0.43–6.09), total number of DEN (standardized mean difference: 0.31, 95% CI: −0.21–0.84), need for additional percutaneous drainage (OR: 0.53, 95% CI: 0.27–1.03), need for salvage surgery (OR: 0.60, 95% CI: 0.28–1.27), procedure‐related adverse events were comparable between groups (OR: 0.81, 95% CI: 0.51–1.28), disease‐related adverse events (OR: 0.84, 95% CI: 0.34–2.13), mortality (OR: 1.11, 95% CI: 0.47–2.59). In RCT‐only analyses, these findings remained consistent. Immediate DEN lowered stent migration risk (OR: 0.31, 95% CI: 0.11–0.92), but lost significance in RCT‐only analysis.

Conclusion

This meta‐analysis found no overall superiority of immediate DEN over on‐demand DEN for patients with NP, highlighting the need to identify specific subgroups that benefit from an upfront necrosectomy strategy.

Key Summary

• Summarize the established knowledge on this subject

  • ◦The optimal timing of direct endoscopic necrosectomy (DEN) for necrotizing pancreatitis (NP) remains uncertain, and current practice is largely guided by scarce evidence and endoscopists' preferences, highlighting a limited evidence base in this patient population.
  • ◦Immediate DEN has been proposed as a strategy to directly remove the source of infection and necrotic tissue, with recent randomized trials indicating the benefits including shorter time to clinical success, fewer reinterventions, and reduced major complications.

• What are the significant and/or new findings of this study?

  • ◦Our pooled analysis found that immediate DEN was associated with a significantly lower risk of stent migration, suggesting that timely removal of pus, debris, and necrotic tissue may promote more effective stent drainage and enhance stent stability.
  • ◦Our pooled analysis revealed no overall superiority of immediate DEN over on‐demand DEN, highlighting the need to identify specific subgroups (stratified by collection characteristics such as size, extent of solid necrosis, and encapsulation stage) that may truly benefit from an upfront necrosectomy strategy.

Abbreviations

ANC

acute necrotic collections

CI

confidence interval

DEN

direct endoscopic necrosectomy

EUS

endoscopic ultrasound

IQR

interquartile range

LAMS

lumen‐apposing metal stents

NP

necrotizing pancreatitis

OR

odds ratio

RCT

randomized controlled trial

SMD

standardized mean difference

WON

walled‐off necrosis

1. Introduction

Acute pancreatitis is one of the most common gastroenterological diagnoses worldwide, bringing substantial medical and social burden [1, 2]. Approximately 20%–30% of patients progress to necrotizing pancreatitis (NP), with local complications classified as acute necrotic collections (ANC) or walled‐off necrosis (WON) based on disease duration and encapsulation [3].

Endoscopic drainage is widely accepted for symptomatic NP because it avoids the risks of pancreatic fistula and exo/endocrine insufficiency, improves quality of life, reduces complications, and shortens hospital stays [4, 5]. Necrosectomy is a treatment option that can effectively remove solid debris from necrotic collections [6]. Over half of WON patients required reinterventions after endoscopic ultrasound (EUS)‐guided drainage, most commonly direct endoscopic necrosectomy (DEN) [7]. Timely removal of necrotic tissue is essential for preventing the progression of infected necrosis [8].

In the past decade, the “step‐up approach” has become the standard. The operators initially place a percutaneous drainage catheter and then perform a minimally invasive retroperitoneal necrosectomy for patients not improving after drainage [9]. In endoscopic drainage, to avoid unnecessary interventions and potentially lethal adverse events, the current recommendations also advocate sparing necrosectomy for a few days after initial stent placement [10]. However, with improved technical safety of DEN, another viewpoint holds that immediate DEN can directly remove the source of infection and necrotic tissues promptly, potentially preventing systemic inflammatory response or other serious complications caused by persistent infection. Currently, there is conflicting evidence regarding the efficacy of immediate versus on‐demand necrosectomy. Some studies support immediate necrosectomy [11, 12], while others do not show clear benefits [13].

Given emerging evidence, this meta‐analysis aimed to systematically review the current literature to provide an updated estimate of the efficacy and safety of immediate DEN and to compare the priority of immediate versus on‐demand DEN after EUS‐guided drainage.

2. Materials and Methods

2.1. Protocol and Registration

This systematic review and meta‐analysis followed the reporting guidelines of the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) (Supporting Information S1: Table S1) [14], and was prospectively registered in the Prospective International Register of Systematic Reviews (PROSPERO) (CRD420261341447). For the included observational studies, the MOOSE reporting guidelines were followed (Supporting Information S1: Table S2).

2.2. Search Strategy and Selection Criteria

PubMed, Embase, Web of Science, and the Cochrane Library were systematically searched from inception to March 18, 2026, without any restrictions on language or article types. The search strategy is described in Supporting Information S1: Appendix 1. Two authors (Wang, R. and Hu, P) independently reviewed the titles and abstracts of the potentially relevant articles and excluded duplicate publications and irrelevant studies during the preliminary screening. Subsequently, a full‐text review of the selected articles was conducted to determine their eligibility. Any disagreements were resolved by a third author (Kong, QZ).

2.3. Inclusion/Exclusion Criteria

We included the study according to the following criteria (1): patients with NP undergoing EUS‐guided drainage using plastic or metal stents (2); available data comparing the safety and efficacy of immediate DEN (perform DEN at the time of initial stent placement for EUS‐guided drainage) and on‐demand DEN (place the stent for EUS‐guided drainage and undergo DEN if the condition does not improve) (3); randomized controlled trials (RCTs), non‐RCTs and observational studies were all considered for inclusion. Duplicate articles, reviews, case reports, single‐arm studies, and non‐peer‐reviewed articles were excluded.

2.4. Data Extraction

Two authors (Wang, R. and Hu, P) independently extracted the data using a standardized extraction form. Variables included author, publication year, country, study type, size of collection, stent types, clinical success rate, technical success rate, times of DEN, need for percutaneous drainage or salvage surgery, adverse events, recurrence rate, mortality, cost, and hospitalization length. Another author (Kong, QZ) verified the data. If the results were presented as median [interquartile range] in the original articles, we estimated the values using the reported median and range, as described by Wan et al. [15] and Luo et al. [16] methods.

2.5. Definitions

Clinical success was defined as the resolution of symptoms and a decrease in collection size during the follow‐up period [11, 17]. Technical success was defined as the completion of planned transmural stent placement and drainage via endoscopy [13]. The total number of DEN procedures performed per patient to achieve clinical success was defined as the total number of DEN procedures performed [18]. Need for salvage surgery was defined as any surgical intervention required due to unresolved WON or ANC [19]. Adverse events were classified according to the American Society for Gastrointestinal Endoscopy lexicon [20], including procedure‐related adverse events (such as bleeding, perforation, stent occlusion, stent migration, etc.) and disease‐related adverse events (such as intestinal fistulas, choledocholithiasis, organ failure, etc.) [17].

2.6. Risk of Bias and Quality Assessment

Two authors (Wang, R. and Hu, P) estimated the quality of RCTs using the revised Cochrane Risk of Bias Tool (RoB 2.0) and of observational studies using the Newcastle–Ottawa Quality Assessment Scale (NOS) [21, 22]. The Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach was used to evaluate the quality of evidence for outcomes [23]. Any disagreements or discrepancies were resolved through consensus with a third author (Kong, QZ).

2.7. Statistical Analysis

The effectiveness and risk of intervention bias were assessed using RevMan software (Review Manager Version5.4, The Cochrane Collaboration, 2020). Dichotomous outcomes were presented as odds ratio (OR) with 95% confidence intervals (CIs) using the Mantel‐Haenszel random‐effects model, and continuous outcomes as standardized mean difference (SMD) with 95% CI using the inverse‐variance method. Standard deviation (SD) was calculated from standard errors, p values, confidence intervals, or plots when SDs were lacking [24]. This applied to the overall pooled analyses combining RCTs and observational studies, as well as analyses stratified by study design. The index of heterogeneity (I²) was calculated using the chi‐square test. Heterogeneity was reported as low (I² = 0%–25%), moderate (I² = 26%–50%), and high (I² > 50%). Publication bias was assessed using Egger's test if more than 10 studies were finally included. Statistical significance was determined by a two‐sided p value < 0.05.

3. Results

3.1. Study Selection and Characteristics

The study selection process is illustrated in Figure 1. Initially, 5116 studies were identified through database searches. 3099 studies remained after removing duplicate articles. Of these, 3068 studies were excluded based on title and abstract assessment. 24 articles were excluded after reviewing the full text, including 14 single‐arm studies, four studies evaluating percutaneous necrosectomy without endoscopic drainage, five clinical trial registries without reported outcome data, and one duplicate publication from the same cohort. Seven studies were ultimately included in the meta‐analysis. There were four RCTs [11, 12, 17, 25] and three retrospective cohort studies [13, 18, 26]. Four studies were conducted in the United States, one in Europe (Denmark), one in East Asia (Japan), and one in West Asia (Iran).

FIGURE 1.

Flow diagram of studies included in the present meta‐analysis.

3.2. Characteristics of the Study

Of the 616 patients, 228 (37.01%) received immediate DEN therapy, and 388 (62.99%) received on‐demand DEN. The RCTs included 107 patients in the immediate DEN group and 108 in the on‐demand DEN group, while the observational studies enrolled 121 patients treated with immediate DEN, and 280 receiving on‐demand DEN. Two RCTs included a total of 23 patients with ANC(11, 17), and the remaining patients had WON. Age ranges were similar between the groups. Alcohol consumption and cholelithiasis were the main etiologies in the studies. All studies used 15–20 mm diameter lumen‐apposing metal stents (LAMS), except for the study by Saito et al., which used plastic stents for two patients in the on‐demand DEN group. Additional study and population characteristics are summarized in Table 1. The technical characteristics of endoscopic necrosectomy procedures are summarized in Supporting Information S1: Table S3.

TABLE 1.

Baseline characteristics of patients and intervention details of the included studies in the meta‐analysis.

Study	Study design	Country	Center type	Interventions	Patient, n	Age, years	Male sex, n (%)	Diameter of collection, cm a	Infection as DEN indication, n (%)	Stent type (diameter)
Mohamadnejad, M. 2026 [25]	RCT	Iran	Single‐centre	Immediate DEN	25	53.2.0 (15.2)	19 (76)	10.4 (3.27)	25 (100%)	LAMS (10 or 20 mm)
				On‐demand DEN	25	50.9 (16.4)	14 (56)	10.1 (3.9)	25 (100%)	LAMS/(10 or 20 mm)
Saito, T. 2026 [11]	RCT	Japan	Multicentre	Immediate DEN	33	59.0 (15.5)	22 (67)	13.0 (11.0–18.5)	25 (76%)	LAMS (NR)
				On‐demand DEN	37	59.8 (17.8)	28 (76)	13.5 (10.0–16.1)	25 (68%)	LAMS/Plastic stent (NR)
Olsen, G. A. 2025 [12]	RCT	Denmark	Single‐centre	Immediate DEN	12	56.5 (34.5–70.0)	7 (58)	23.0 (20.4–26.8)	11 (92%)	LAMS (20 mm)
				On‐demand DEN	13	53.0 (48.0–74.0)	8 (62)	21.7 (20.1–25.0)	10 (77%)	LAMS (20 mm)
Bang, J. Y. 2024 [17]	RCT	USA	Multicentre	Immediate DEN	37	45.7 (14.4)	26 (70%)	11.6 (8.9–14.1)	37 (100%)	LAMS (20 mm)
				On‐demand DEN	33	53.2 (14.9)	20 (61%)	10 (8.5–15.0)	33 (100%)	LAMS (20 mm)
Abdullah, A. 2024 [26]	Retrospective cohort study	USA	Single‐centre	Immediate DEN	9	58 (14.2)	8 (88.9%)	8.8 (3.3)	NR	LAMS (15 mm)
				On‐demand DEN	41	51.3 (12.2)	32 (78.1%)	10.1 (4.3)	NR	LAMS (15.1 mm)
Pawa, R. 2023 [13]	Retrospective cohort study	USA	Single‐centre	Immediate DEN	43	52.2 (15.4)	24 (56%)	128.4 (107)	20 (47%)	LAMS (15 mm)
				On‐demand DEN	37	57.5 (13.8)	25 (68%)	114.7 (79.4)	14 (38%)	LAMS (15 mm)
Yan, L. 2019 [18]	Retrospective cohort study	USA	Multicentre	Immediate DEN	69	54.8 (8.8)	51 (74%)	9.6 (4.8)	NR	LAMS (14.9 mm)
				On‐demand DEN	202	54 (8.8)	73 (36%)	11.1 (4.8)	NR	LAMS (14.9 mm)

Note: Data are presented as mean (SD) for normally distributed continuous variables or as median (IQR) for non‐normally distributed variables.

Abbreviations: DEN, direct endoscopic necrosectomy; IQR, interquartile range; LAMS, lumen‐apposing metal stent; NR, not reported; RCT, randomized controlled trial; SD, standard deviation.

^a The study of Pawa, R. 2023 reported collection size as area (cm²).

3.3. Clinical Success

Six of seven studies compared the clinical success. The immediate DEN group showed a higher success rate than that of the on‐demand DEN group, but no statistical difference was observed (OR: 1.92, 95% CI: 1.01–3.63, p = 0.05), and no heterogeneity was detected (I² = 0%, p = 0.70) (Figure 2A). The analysis limited to RCTs also showed no significant statistical difference (OR: 3.09, 95% CI: 0.98–9.70, p = 0.05).

FIGURE 2.

Forest plots for clinical success (A) and technical success (B).

3.4. Technical Success

Three studies reported the technical success of all patients (100%). For these studies, a forest plot was not appropriate because there was no effect heterogeneity. Technical success did not significantly differ between the groups. In the remaining RCTs, the pooled OR was 2.75 (95% CI: 0.11–69.94, p = 0.54); in the remaining observational studies, the pooled OR was 1.45 (95% CI: 0.34–6.22, p = 0.61). No significant differences were observed between the subgroups (p = 0.72) (Figure 2B).

3.5. Total Number of DEN

All seven studies were included to analyze the mean number of DENs. There was no significant difference between the two groups, with SMD of 0.31 (95% CI: −0.21–0.84, p = 0.24), though high heterogeneity was detected (I² = 87%, p < 0.00001). There was no statistical difference between the two groups in the RCTs (SMD = 0.52, 95% CI: −0.02–1.06, p = 0.06, I² = 72%) or the observational studies (SMD = 0.06, 95% CI: −0.81–0.93, p = 0.89, I² = 91%) (Figure 3A). Because of significant heterogeneity across studies, subgroup analysis was conducted by LAMS diameter. High heterogeneity still exists in the 15 mm diameter LAMS subgroup (SMD = 0.06, 95% CI: −0.81–0.93, P = 0.89, I² = 91%), while no heterogeneity was detected in the 20 mm diameter LAMS subgroup (SMD = 0.13, 95% CI: −0.27–0.54, P = 0.52, I² = 0%) (Supporting Information S1: Figure S1). Three studies were included in the analysis of patients who received necrosectomy once, twice, or three times or more. The pooled OR for one, two, and three or more procedures were 0.97 (95% CI: 0.09–10.52, p = 0.98, I² = 96%), 2.38 (95% CI: 0.46–12.37, p = 0.30, I² = 88%), and 0.62 (95% CI: 0.15–2.63, p = 0.52, I² = 85%), respectively (Supporting Information S1: Figure S2).

FIGURE 3.

Forest plots for total number of DEN (A), need for additional percutaneous drainage (B), need for salvage surgery (C).

3.6. Need for Additional Percutaneous Drainage

The analysis of six studies resulted in an OR of 0.53 (95% CI: 0.27–1.03, p = 0.06), with no heterogeneity (I² = 0%, p = 0.77) regarding the need for percutaneous drainage. In RCTs only, there was no significant difference in the OR (OR: 0.48, 95% CI: 0.21–1.09, p = 0.08) for requiring additional percutaneous drainage between the immediate DEN and on‐demand DEN groups (Figure 3B).

3.7. Need for Salvage Surgery

There was no significant difference in the need for salvage surgery between the two groups, with an overall pooled OR (0.60, 95% CI: 0.28–1.27, p = 0.18). and an RCT‐only pooled OR (0.35, 95% CI: 0.05–2.31, p = 0.27). No significant heterogeneity was observed between the studies (I² = 0%, p = 0.91) (Figure 3C).

3.8. Procedure‐Related Adverse Events

All seven studies compared the procedure‐related adverse events. The pooled rate of procedure‐related adverse events was comparable between immediate DEN and on‐demand DEN (15.85% vs. 18.25%, OR: 0.81, 95% CI: 0.51–1.28, p = 0.37), and no heterogeneity was observed (I² = 0%, p = 0.53) (Figure 4A). Subgroup analyses of RCTs (OR: 0.84, 95% CI: 0.36–1.98, p = 0.69, I² = 39%) and observational studies (OR: 0.74, 95% CI: 0.39–1.40, p = 0.36, I² = 0%) revealed no statistically significant difference between the two groups. The reported subtypes of adverse events included bleeding, stent occlusion, stent migration, and perforation. No significant differences between treatment strategies were observed for bleeding, stent occlusion, and perforation (Figure 5A, 5B and Supporting Information S1: Figure S3). Stent migration was reported in three patients in the immediate DEN group and 24 patients in the on‐demand DEN group. In the overall pooled analysis, there was a higher incidence of stent migration in the on‐demand DEN group (1.39% vs. 6.4%, OR: 0.31, 95% CI: 0.11–0.92, p = 0.03) (Figure 5C). Subgroup analyses by study design showed no statistically significant difference in RCTs (OR: 0.35, 95% CI: 0.08–1.61, p = 0.18) or observational studies (OR: 0.28, 95% CI: 0.06–1.27, p = 0.10), with no significant differences between subgroups (p = 0.85) and no within‐subgroup heterogeneity (I² = 0%). Specific adverse events in individual studies are summarized in Table 2.

FIGURE 4.

Forest plots for procedure‐related adverse events (A), disease‐related adverse events (B), and mortality (C).

FIGURE 5.

Forest plots for bleeding (A), stent occlusion (B), and stent migration (C).

TABLE 2.

Adverse events between the immediate and on‐demand DEN groups were reported in the included studies.

Study	Bleeding: n (%)		Perforation: n (%)		Stent occlusion: n (%)		Stent migration: n (%)		Infection: n (%)
	Immediate DEN	On‐demand DEN	Immediate DEN	On‐demand DEN	Immediate DEN	On‐demand DEN	Immediate DEN	On‐demand DEN	Immediate DEN	On‐demand DEN
Mohamadnejad, M. 2026 [25]	4 (16.0)	1 (4.0)	NR	NR	NR	NR	1 (4.0)	1 (4.0)	NR	NR
Saito, T. 2025 [11]	7 (21.0)	4 (11.0)	2 (6.1)	2 (5.4)	1 (3.0)	0 (0)	1 (3.0)	5 (14.0)	NR	NR
Olsen, G. A. 2025 [12]	1 (8.33)	2 (15.38)	1 (8.33)	1 (7.69)	NR	NR	NR	NR	NR	NR
Bang, J. Y. 2024 [17]	4 (10.81)	4 (12.12)	NR	NR	NR	NR	0 (0)	1 (3.03)	2 (5.41)	1 (3.03)
Abdullah, A. 2024 [26]	2 (22.22)	6 (14.63)	0 (0)	1 (2.44)	2 (22.22)	23 (56.09)	0 (0)	6 (14.63)	1 (11.11)	8 (19.51)
Pawa, R. 2023 [13]	1 (2.32)	1 (2.70)	NR	NR	2 (4.65)	3 (8.11)	0 (0)	1 (2.70)	NR	NR
Yan, L. 2019 [18]	0 (0)	8 (3.96)	0 (0)	3 (1.49)	1 (1.45)	9 (4.46)	1 (1.45)	10 (4.95)	5 (7.25)	20 (9.90)

Abbreviations: DEN, direct endoscopic necrosectomy; NR, not reported.

3.9. Disease‐Related Adverse Events

No significant difference was observed between immediate and on‐demand DEN groups (21.85% vs. 26.21%, OR: 0.84, 95% CI: 0.34–2.13, p = 0.72), and moderate heterogeneity was observed (I² = 49%, p = 0.10) (Figure 4B). Among RCTs, disease‐related adverse events showed no significant difference (OR: 0.72, 95% CI: 0.29–1.81, p = 0.48, I² = 50%).

3.10. Mortality

A total of 13 patients in the immediate DEN group and 12 patients in the on‐demand DEN group died. There was no significant difference between the two therapies (OR: 1.11, 95% CI: 0.47–2.59, p = 0.82), and no heterogeneity was observed (I² = 0%, p = 0.50) (Figure 4C). In RCTs only, mortality showed no significant difference (OR: 0.84, 95% CI: 0.28–2.58, p = 0.77, I² = 0%).

3.11. Recurrence

Three studies comprising 376 patients showed no significant difference in the rate of recurrence with a pooled OR of 1.64 (95% CI: 0.16–17.27.0, p = 0.68), and high heterogeneity was observed (I² = 72%, p = 0.03). Only one RCT reported recurrence data, which showed no statistically significant between‐group difference (OR: 2.40, 95% CI: 0.19–30.52, p = 0.50) (Supporting Information S1: Figure S4).

3.12. Duration of LAMS Placement

Four studies reported the duration of LAMS placement and found no significant difference between the groups. The pooled SMD was −0.28 (95% CI: −0.56–‐0.00, p = 0.05), and there was no heterogeneity (I² = 0%, p = 0.85). In RCTs, there was no statistically significant between‐group difference (SMD: −0.41, 95% CI: −0.82–‐0.00, p = 0.05). (Supporting Information S1: Figure S5).

3.13. Length of Hospitalization

Hospitalization length was reported in six studies. Overall, there was no significant difference between immediate DEN and on‐demand DEN groups, with SMD of −0.20 (95% CI: −0.66–0.26, p = 0.40). However, a high degree of heterogeneity was observed between studies (I² = 75%, p = 0.001). Among RCTs, the result remained non‐significant (SMD: −0.30, 95% CI: −0.87–0.27, P = 0.31) (Supporting Information S1: Figure S6).

3.14. Costs

Three RCTs showed no significant difference in the total costs, with SMD of −0.34 (95% CI: −0.81–0.12, p = 0.14), and there was moderate heterogeneity between studies (I² = 50%, p = 0.13) (Supporting Information S1: Figure S7).

3.15. Risk of Bias

Four of the seven studies were rated as low risk, and the remainder as moderate risk (Supporting Information S1: Figure S8, Table S4). We evaluated the quality of evidence for primary and secondary outcomes. Evidence quality was reduced owing to the inclusion of non‐randomized studies (Supporting Information S1: Figure S9). Publication bias was not assessed because fewer than 10 studies were included in this meta‐analysis.

4. Discussion

Although recent RCTs have individually shown that immediate DEN reduces the number of reinterventions, decreases major complications, or shortens the time to clinical success, our meta‐analysis failed to demonstrate improved outcomes with immediate DEN compared with on‐demand DEN, except for stent migration, which was significantly lower in the immediate DEN cohort. This finding revealed study heterogeneity and suggested that the benefits of immediate DEN may depend on specific factors, such as collection characteristics and healthcare institutions.

Immediate DEN is considered a more proactive intervention. Performing immediate DEN at the time of stent placement may allow adequate drainage and infection control through early mobilization, and contribute to a shortened time to clinical success, as was confirmed in the WONDER‐01 trial [11]. Although the pooled OR for clinical success favored immediate DEN, this meta‐analysis did not demonstrate a statistically significant difference in clinical success between immediate and on‐demand DEN. Given the limited number of studies and patients, the precision of the effect estimate is limited. The current evidence is insufficient to establish the superiority of immediate DEN. In the study by Pawa et al., clinical success was defined as the resolution of the WON cavity at the time of LAMS removal, rather than at the 6‐month follow‐up period [13]. This may result in the omission of patients who could have been included in the clinical success rate during the 6‐month follow‐up period, thereby affecting the final determination. Future RCTs with larger samples are needed to verify whether the observed effect estimate reflects a true clinical benefit.

Earlier studies suggested that 16%–50% of WON could resolve with drainage alone, supporting a step‐up approach that may avoid unnecessary necrosectomy procedures [27, 28]. However, our analysis did not reveal statistically significant differences between the two therapeutic approaches regarding the total number of DENs or the need for percutaneous drainage or surgery. Notably, the included studies exhibited heterogeneity in outcomes related to the number of DENs. High heterogeneity still persisted after stratification by study design. Previous studies revealed that larger diameter stents, dual‐modality drainage and multi‐gateway techniques are associated with fewer necrosectomy processes [29]. Our subgroup analysis did not identify stent diameter as a significant contributor to the observed heterogeneity. The differences in procedural protocols may be the source of heterogeneity. In Yan et al.’s study, all patients underwent DEN routinely, which may have inflated the mean value in the on‐demand group. In subsequent RCTs, the WONDER‐01 trial's protocol required 72–96 hours of observation and failure of at least two drainage‐based interventions before DEN initiation [11]. In contrast, the ACCELERATE trial [12], the DESTIN trial [17], and the trial by Mohamadnejad et al. [25] initiated DEN or additional drainage without mandated delays for repeated drainage in the on‐demand arms. Meanwhile, the reduction of necrotic collections was assessed by cross‐sectional imaging at 72 hours after the initial intervention in the DESTIN trial [17] and the trial by Mohamadnejad et al. [25]. In contrast, other studies used later or less strictly protocolized imaging assessments, including reassessment approximately 1 week after drainage, or imaging caused by persistent or recurrent symptoms. Later, or symptom‐triggered imaging may delay the identification of repeating DEN. Therefore, heterogeneity in DEN frequency may reflect variability in protocols across centers. A protocolized strategy specifying early repeat cross‐sectional imaging and planned necrotic tissue removal achieved faster necrotic reduction, fewer adverse events, and shorter intensive care unit stay [30]. More future trials with standardized protocols are expected to validate the superiority of different strategies.

Previously, most endoscopists delayed DEN to allow for mature tract formation and to reduce the risk of adverse events [31]. However, no significant differences in bleeding, stent occlusion, or perforation were observed between immediate and on‐demand DEN in our analysis. Pathophysiologically, when DEN is performed immediately after drainage, the sinus tract may not have sufficiently matured. Aggressive device manipulation may increase tissue damage and the risk of perforation or severe bleeding [29]. The absence of significant differences in the comparative data does not necessarily exclude the potential risk of immediate DEN. Therefore, the timing of DEN should also consider the stability of transmural channels and the clinical condition of patients. Notably, stent migration was significantly lower with immediate DEN. Mechanistically, timely removal of pus, debris, and necrotic tissue promotes more effective drainage through the stent, thereby reducing the effects on stent stabilization. The consistency of trends across the included studies suggests that this effect may reflect a genuine clinical difference. However, the low incidence in original studies limits statistical power. Therefore, future RCTs with larger sample sizes are needed to verify this protective effect. In the included studies, LAMS was used in all the patients receiving immediate DEN and in all but two patients receiving on‐demand DEN. Therefore, the pooled estimates, including the one for stent migration, reflect the effect of DEN timing in LAMS‐based drainage. The average LAMS duration for both treatment groups exceeded 4 weeks in three of the four reported studies. Given that recent evidence shows no clear increase in bleeding or overall adverse events with LAMS placement time beyond 4 weeks [32], LAMS duration is unlikely to confound adverse event comparisons between the two DEN strategies.

LAMS maintains apposition between the gastrointestinal lumen and the necrotic cavity, providing a stable passage for the endoscope. This technical characteristic may contribute to the preference for LAMS among patients requiring DEN [33]. Unlike LAMS, plastic stents provide a smaller access tract and require transmural tract dilation before DEN [34]. Current evidence does not demonstrate a clear superiority of LAMS over plastic stents in overall treatment outcomes [33]. However, evidence on plastic stents has largely been derived from step‐up strategies. Whether immediate DEN can achieve comparable safety in patients with plastic stent drainage remains uncertain.

The potential heterogeneity of collection characteristics may also have diminished the differences between the two methods. The ACCELERATE trial focused specifically on large WONs (> 15 cm) containing predominantly solid necrosis, and every patient in the on‐demand arm ultimately required necrosectomy. In this population, immediate DEN improved the composite outcome of mortality, major complications, and hospital stay [12]. These findings suggest that patients with large volumes and substantial necrotic burden benefit more from immediate DEN than from waiting for drainage failure alone. In contrast, the remaining RCTs included moderate‐sized collections, and a certain proportion of patients in the on‐demand DEN group achieved clinical success with drainage alone. In particular, Mohamadnejad et al. found no benefit from immediate DEN. This observation suggests that the benefit of immediate DEN may not be uniform across all patients and that an on‐demand strategy may be sufficient for selected patients with a lower necrotic burden or a lower anticipated need for necrosectomy. Infected necrosis is an important indication for intervention. The proportion of patients with infected status also varied across studies. Notably, the DESTIN trial and the trial by Mohamadnejad et al. enrolled only patients with infected necrotizing pancreatitis but reported different comparative outcomes, suggesting that infection status alone may not identify patients who benefit from immediate DEN. However, these interpretations remain hypothesis‐generating. The quadrant, necrosis, and infection (QNI) classification system demonstrated that the WON cavity with multi‐quadrant distribution, elevated necrosis percentage, and infected status was significantly associated with a higher number of necrosectomies [35]. The QNI classification has been validated externally [36] and prospectively [37] as effective in stratifying patients by their risk of a complex clinical course. This stratification approach provides a new perspective for explaining the heterogeneity among studies on the timing of DEN. It lays a theoretical foundation for the design of future individualized treatment trials. An additional factor is the timing of initiating drainage. Endoscopic intervention performed early or delayed may achieve comparable technical and clinical outcomes in selected patients [38]. In this meta‐analysis, most patients underwent drainage after the WON formation, whereas 24% in the WONDER‐01 trial and 8.5% in the DESTIN trial were ANC patients with incompletely encapsulated lesions. Because of the lack of separate outcome data, we could not perform subgroup analyses by necrosis extent, size, or collection type. Future research should stratify patients by size, solid extent, and encapsulation stage to identify the subgroup most likely to benefit from immediate DEN.

There are several limitations to this meta‐analysis. First, a substantial portion of the pooled evidence was derived from observational studies, which are susceptible to residual confounding, selection bias, and clinical practice patterns. Therefore, we performed overall and RCT‐only analyses to evaluate the robustness and interpretability. RCTs with larger samples in the future are needed to provide research data support. Second, the number of eligible studies was relatively small, and most studies included limited sample sizes, resulting in reduced statistical power and potential type II errors. The small number of studies also restricted the ability to perform more comprehensive subgroup analyses. Third, we did not analyze procedural duration, disease recurrence, or readmission, as these parameters have not been considered in all studies.

5. Conclusion

This meta‐analysis demonstrated that immediate DEN can achieve efficacy comparable to on‐demand DEN in patients with NP and causes fewer stent migration events. Future research should focus on more detailed stratification of necrotic collections by morphology to identify phenotypes most likely to benefit from an immediate DEN strategy.

Author Contributions

Rui Wang: conceptualization, data curation, formal analysis, writing – original draft. Qingzhou Kong: conceptualization, data curation, investigation, writing – original draft. Ping Hu: data curation, formal analysis, investigation, methodology. Yu Ma: methodology, resources, software, supervision. Xiao Liang: methodology (equal), resources (equal), software (equal), supervision (equal). Ning Zhong: supervision, validation, visualization. Zhen Li: funding acquisition, writing – review and editing (lead).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting Information S1

Data Availability Statement

All data generated or analyzed during this study are included in this published article.