Complete clip closure with staged triple-lumen feeding tube reduces delayed adverse events after duodenal endoscopic submucosal dissection

Abstract

BACKGROUND

Given the thin duodenal wall and exposure of post-resection defects to bile and pancreatic juice, delayed adverse events (AEs) after duodenal endoscopic submucosal dissection (ESD) remain a major clinical challenge. While insufficient defect closure and suboptimal postoperative drainage have been implicated as potential risk factors, real-world evidence remains limited. We hypothesized that complete clip closure combined with structured postoperative management using a triple-lumen feeding tube would reduce delayed AEs by improving defect protection and reducing exposure to digestive secretions.

AIM

To explore the relationship between complete clip closure, a staged postoperative management protocol, and delayed AEs after duodenal ESD.

METHODS

We retrospectively analyzed 179 consecutive patients who underwent duodenal ESD. The patients’ baseline characteristics were compared according to delayed AE occurrence. Independent risk factors for postoperative complications were identified through univariate and multivariate logistic regression analyses. Clip closure quality (complete vs incomplete) and postoperative management strategy (triple-lumen feeding tube vs nasogastric tube) effects on clinical outcomes were evaluated through subgroup analyses. The relationship between drainage strategy and perioperative inflammatory markers and nutritional indicators was examined through covariance analyses.

RESULTS

Delayed AEs occurred in 12.8% of patients. Multivariable analysis demonstrated that complete clip closure [adjusted odds ratio (aOR) = 0.16, 95% confidence interval (CI): 0.12-0.22; P < 0.001] and postoperative triple-lumen feeding tube placement (aOR = 0.64, 95%CI: 0.46-0.87; P < 0.001) were independently associated with lower delayed AEs risk. Patients who received complete closure and triple-lumen feeding tube drainage had the lowest observed incidence there of (2.8%). On postoperative day 1, the triple-lumen feeding tube group had a significantly lower neutrophil percentage than the nasogastric tube group [model 2: Adjusted mean difference (AMD) = -1.82], whereas serum albumin (model 2: AMD = 0.41) and total protein levels (model 2: AMD = 0.85) were significantly higher (all P < 0.05).

CONCLUSION

Complete clip closure and postoperative triple-lumen feeding tube management were independently associated with fewer delayed AEs after duodenal ESD.

Key Words: Duodenum; Endoscopic submucosal dissection; Delayed adverse events; Mucosal defect closure; Triple-lumen feeding tube; Perioperative management

Core Tip: Delayed adverse events (AEs) remain a major challenge after duodenal endoscopic submucosal dissection (ESD). In this real-world cohort, complete clip closure and postoperative triple-lumen feeding tube management were independently associated with fewer delayed AEs, with patients receiving both measures exhibiting the lowest observed complication rate. Furthermore, triple-lumen feeding tube drainage was associated with attenuated early postoperative inflammation and improved nutritional status. These results provide practical evidence to inform perioperative management in high-risk duodenal ESD.

INTRODUCTION

Duodenal endoscopic submucosal dissection (ESD) is technically challenging due to the thin duodenal wall, rich vascular supply, and persistent exposure of the post-ESD ulcer to bile and pancreatic juice. Consequently, the risks of delayed bleeding and perforation after duodenal ESD are substantially higher than those observed in other gastrointestinal tract segments[1-4]. The incidence of delayed adverse events (AEs) after duodenal ESD is approximately 5%-17%[5], representing a major barrier to the broader clinical application of this technique.

To mitigate these risks, clinical practice has widely adopted prophylactic endoscopic closure of mucosal defects, and has been recommended in previous studies[6-8]. The available evidence has suggested that complete defect closure is associated with lower delayed AE incidence and attenuated postoperative inflammatory responses compared with partial or no closure[6]. However, anatomical constraints, lesion location and size, and technical complexity mean that achieving stable and durable complete clip closure remains challenging in a high-risk case subset.

Currently, systematic quantitative evaluation of the correlation between clip closure quality (whether complete closure is achieved) and the incidence of complications following duodenal ESD is lacking. Furthermore, current conventional postoperative management strategies are subject to limitations regarding the effectiveness of digestive secretion drainage and early enteral nutritional support. Notably, targeted and optimized perioperative management strategies for high-risk wounds with incomplete closure or wounds where complete closure cannot be achieved are lacking.

Against this background, the present single-center retrospective cohort study was conducted with two primary objectives. The first was to systematically evaluate the association between clip closure quality and the risk of delayed AEs after duodenal ESD. The second was to assess the effects of a staged postoperative management strategy using a triple-lumen feeding tube on delayed AEs and early postoperative inflammatory and nutritional indicators. By explicitly defining these two aims, we aimed to provide real-world evidence to guide the optimization of perioperative management in high-risk duodenal ESD.

MATERIALS AND METHODS

Study design and eligible patients

This single-center retrospective cohort study was approved by the Peking University Cancer Hospital Ethics Committee (approval No. 2017KT103). All patients granted written informed consent. A total of 200 consecutive patients who underwent duodenal ESD at our institution between February 2016 and April 2025 were initially screened. Lesions were selected for ESD based on clinical judgment, considering the factors of size, location, morphology, and suspected histology. Due to the retrospective design, detailed data on these factors were not consistently available for all cases and therefore were not included in the article. We excluded one patient who required conversion to open surgery due to intraoperative perforation, and 20 patients with incomplete records of postoperative drainage strategies. The final analysis included 179 patients.

ESD procedure

All procedures were performed under general anesthesia with endotracheal intubation by experienced endoscopists who had completed at least 50 duodenal ESD procedures independently. The procedural workflow followed the standardized technical steps recommended by the European Society of Gastrointestinal Endoscopy (ESGE)[5]. Patients with ampullary lesions underwent routine prophylactic pancreaticobiliary stenting. The clip closure quality was classified as complete or incomplete based on immediate post-procedural endoscopic assessment, as detailed in the section “criteria for assessment of clip closure quality”.

A therapeutic endoscope equipped with a water jet function (EG-760CT, Fujifilm, Tokyo, Japan) was used. Submucosal injection was performed using a mixed glycerol fructose (500 mL) and sodium hyaluronate (150 mL) solution. Mucosal incision and submucosal dissection were primarily performed using an endoknife (blade length, 1.5 mm; Micro-Tech, Nanjing, Jiangsu Province, China). A HookKnife (Olympus, Tokyo, Japan) was used when maneuverability was limited. Minor bleeding was controlled by coagulation using the sheath of the knife tip, whereas active spurting bleeding was treated with hemostatic forceps (Coagrasper; Olympus). All electrosurgical devices were connected to a high-frequency generator (ESG-400; Olympus, or VIO-300S; ERBE, Tübingen, Germany). The generator settings were as follows: Dry cut mode (effect 3, 30 W) for mucosal incision, swift coagulation mode (effect 4, 30 W) for submucosal dissection, and soft coagulation mode (effect 5, 50 W) for hemostasis.

Criteria for assessment of clip closure quality

Two experienced endoscopists who were blinded to the clinical outcomes evaluated the clip closure quality independently by reviewing immediate post-procedural endoscopic images (κ = 0.94). Disagreements were resolved by adjudication by a third senior endoscopist. In all cases, closure integrity was assessed through a standardized testing protocol. After clip closure with titanium clips, gas was first aspirated from the duodenal lumen, followed by air insufflation via the endoscope to check for leaks, which were identified by the presence of visible air bubbles (air insufflation test). Subsequently, normal saline was injected onto the ESD wound surface to assess for fluid leakage, defined as the escape of gas or fluid through the mucosal defect under water immersion (water immersion test). These tests were routinely performed on all patients, regardless of the presence of intra-procedural perforation. Complete clip closure was defined as satisfactory apposition of the mucosal edges without visible gaps and no evidence of air or fluid leakage for at least 30 seconds during these tests. Incomplete closure was defined by the presence of any of the following: (1) Visible gaps or poor approximation of the mucosal edges; or (2) Detectable air or fluid leakage during the above testing.

Post-ESD management

All patients were routinely managed with fasting, intravenous fluid supplementation, and proton pump inhibitor therapy (rabeprazole 20 mg/day or lansoprazole 30 mg/day) for 3 weeks. Ulcer healing was assessed using second-look endoscopy within 1-3 days after ESD. Given the long study duration (2016-2025), postoperative management strategies at our center had evolved from nasogastric tube placement to triple-lumen feeding tube placement. Therefore, the drainage strategy choice mainly reflected temporal changes in clinical practice rather than patient-specific allocation. Postoperative management was accordingly categorized into two groups: Triple-lumen feeding tube and nasogastric tube.

The triple-lumen feeding tube group followed a staged nutritional support protocol. Continuous negative-pressure drainage combined with total parenteral nutrition was provided during the first 72 hours after ESD. Intermittent negative-pressure drainage was applied from postoperative day (POD) 4 to POD 6, and enteral nutrition was initiated via the feeding tube. From POD 7 onward, suction was discontinued and oral intake was gradually resumed. The triple-lumen feeding tube was routinely removed 7-10 days after the procedure.

The nasogastric tube group received continuous negative-pressure drainage combined with total parenteral nutrition. The timing of nasogastric tube removal and resumption of oral intake was determined individually according to the drainage fluid characteristics and degree of abdominal distension. Typically, the nasogastric tube was removed after 5-6 days.

Definition of AEs and outcome measures

The primary outcome was delayed AEs occurring > 24 hours after ESD, excluding intraoperative complications. The main delayed AEs were delayed bleeding (postoperative bleeding requiring endoscopic or pharmacological intervention) and delayed perforation (confirmed by radiographic imaging, such as X-ray or computed tomography, or clinically manifested as intra-abdominal or retroperitoneal abscess). To ensure complete capture of delayed AEs, all patients were followed for at least 30 days post-ESD, including both in-hospital monitoring and outpatient or telephone follow-up after discharge. The secondary outcomes were perioperative inflammatory and nutritional status, assessed by dynamic changes in white blood cell (WBC) count, neutrophil percentage (Neu%), serum albumin (ALB), and total protein (TP) measured 1 day before ESD and on POD 1, POD 3, and POD 5.

Statistical analysis

All statistical analyses were performed using Python (version 3.10). Continuous variables are presented as the mean ± SD, and categorical variables are presented as n (%). Groups were compared according to data distribution: Continuous variables were analyzed using the independent-samples t-test or Welch’s t-test, as appropriate. Categorical variables were compared using the χ² test or Fisher’s exact test, as appropriate. All tests were 2-sided, and P < 0.05 was considered statistically significant. Independent factors associated with postoperative complications were identified using logistic regression analysis. The factors affecting postoperative complications after duodenal ESD were systematically evaluated using three models: Univariable (model 1), multivariable (model 2), and multivariable with interaction terms (model 3). The dependent variable was postoperative complications, and the primary independent variable was the degree of titanium clip closure. The covariates were prespecified clinical confounders [age, gender, body mass index (BMI), comorbidities, year of admission, lesion location, and drainage method]. The association between titanium clip closure and complications after adjustment for confounders was assessed using model 2, while model 3 included the interaction term “closure degree × drainage method” to evaluate potential joint effects. Adjusted odds ratios (aORs) with 95% confidence intervals (CIs) were reported. Due to the limited number of positive events (n = 23) and the inclusion of multiple covariates, multivariable analyses were performed using L2-regularized logistic regression to reduce the risk of overfitting while preserving the effects of all confounders.

Differences in incidence among subgroups were described using the χ² test. Logistic regression models were separately fitted in each subgroup to quantify the between-group incidence risk and explore differences in the protective effect of titanium clip closure among the subgroups. Furthermore, the effects of the drainage methods on early postoperative laboratory indicators (WBC, Neu%, ALB, and TP) were evaluated using analysis of covariance. Model 1 was adjusted for preoperative baseline values. Model 2 was based on model 1, with additional adjustment for demographic characteristics. The robustness of the results was assessed through sensitivity analyses that excluded patients with ampullary lesions (Supplementary Tables 1-5).

RESULTS

Baseline characteristics of the study population

The analysis included 179 patients who underwent duodenal ESD. The overall incidence of delayed AEs was 12.8% (23/179). The age, gender, BMI, comorbidities, year of admission, lesion location, and postoperative drainage strategy of the patients with and without delayed AEs were not significantly different (all P > 0.05) (Table 1). Contrastingly, the clip closure quality differed significantly between the two groups. The patients without delayed AEs had a substantially higher proportion of complete clip closure than those with delayed AEs [70.51% (110/156) vs 26.09% (6/23), P < 0.001].

Table 1 Baseline characteristics of patients grouped by complications, mean ± SD/n (%).

Variable	Complication-free group (n = 156)	Complication group (n = 23)	t/χ2 value	P value
Age (years)	58.63 ± 11.14	57.35 ± 12.14	0.477	0.637
Gender			1.300	0.254
Male	78 (50.00)	15 (62.22)
Female	78 (50.00)	8 (34.78)
BMI (kg/m2)	24.37 ± 3.42	24.06 ± 3.31	0.417	0.680
Hypertension			0.669	0.414
Yes	36 (23.08)	3 (13.04)
No	120 (76.92)	20 (86.96)
Diabetes			0.815	0.316
Yes	21 (13.46)	1 (4.35)
No	135 (86.54)	22 (95.65)
Coronary heart disease			0.062	1.000
Yes	13 (8.33)	1 (4.35)
No	143 (91.67)	22 (95.65)
Admission year			1.063	0.184
2016-2020	18 (11.54)	5 (21.74)
2021-2025	138 (88.46)	18 (78.26)
Lesion location			0.132	0.505
Ampullary area	19 (12.18)	4 (17.39)
Non-ampullary area	137 (87.82)	19 (82.61)
Titanium clip closure status			15.452	< 0.001
Complete	110 (70.51)	6 (26.09)
Incomplete	46 (29.49)	17 (73.91)
Drainage strategy			2.190	0.139
Triple-lumen feeding tube	97 (62.18)	10 (43.48)
Nasogastric tube	59 (37.82)	13 (56.52)

Comparisons between groups for continuous variables were analyzed using the t-test; categorical variables were analyzed using the χ² test. BMI: Body mass index.

Logistic regression analyses of factors associated with delayed AEs and interaction analyses

The univariate logistic regression analyses determined that complete clip closure, female gender, presence of comorbidities, year of admission (2021-2025), and postoperative placement of a triple-lumen feeding tube were associated with a lower risk of delayed AEs (all P < 0.001). After adjustment for prespecified confounders, complete clip closure (aOR = 0.16, 95%CI: 0.12-0.22; P < 0.001), postoperative placement of a triple-lumen feeding tube (aOR = 0.64, 95%CI: 0.46-0.87; P = 0.005), female gender (aOR = 0.69, 95%CI: 0.51-0.93; P = 0.016), and admission year between 2021 and 2025 (aOR = 0.57, 95%CI: 0.36-0.89; P = 0.015) remained independently associated with a reduced risk of delayed AEs, whereas ampullary location was independently associated with an increased risk (aOR = 1.91, 95%CI: 1.21-3.01; P = 0.005). In the model 3 interaction analysis, the directions of the main effects were generally consistent with those observed in model 2. The interaction between titanium clip closure status and drainage strategy had a significant negative effect (aOR = 0.49, 95%CI: 0.26-0.93, P = 0.030) (Figure 1 and Table 2).

Figure 1 Forest plot of univariate and multivariate logistic regression for delayed adverse events following duodenal endoscopic submucosal dissection. A: Forest plot of the univariate logistic regression analysis; B: Forest plot of the multivariable logistic regression analysis, in which all potential confounders were included for adjustment (age, gender, body mass index, comorbidity, titanium clip closure status, drainage strategy, year of admission, and lesion location); C: Forest plot of multivariate logistic regression including the interaction term “titanium clip closure status × drainage strategy” based on model 2. ^aP < 0.05. ^bP < 0.01. ^cP < 0.001. BMI: Body mass index; CI: Confidence interval.

Table 2 Univariate and multivariable logistic regression and interaction analyses of the factors associated with delayed adverse events after duodenal endoscopic submucosal dissection.

Variable		Model 1		Model 2		Model 3
		OR (95%CI)	P value	aOR (95%CI)	P value	aOR (95%CI)	P value
Age (per 1 SD increase)		0.89 (0.77-1.03)	0.127	0.88 (0.76-1.02)	0.091	0.86 (0.74-1.00)	0.050
BMI (per 1 SD increase)		0.88 (0.76-1.02)	0.093	0.88 (0.76-1.03)	0.107	0.87 (0.75-1.02)	0.080
Gender	Male (reference)	1	< 0.001	1	0.016	1	0.010
	Female	0.53 (0.40-0.72)		0.69 (0.51-0.93)		0.68 (0.50-0.92)
Comorbidity	Without (reference)	1	< 0.001	1	0.084	1	0.130
	With	0.56 (0.41-0.76)		0.75 (0.55-1.04)		0.78 (0.56-1.08)
Admission year	2016-2020	1	< 0.001	1	0.015	1	0.040
	2021-2025	0.47 (0.30-0.73)		0.57 (0.36-0.89)		0.62 (0.39-0.98)
Lesion location	Non-ampullary area (reference)	1	0.061	1	0.005	1	0.005
	Ampullary area	1.52 (0.98-2.35)		1.91 (1.21-3.01)		1.92 (1.22-3.03)
Titanium clip closure status	Incomplete (reference)	1	< 0.001	1	< 0.001	1	< 0.001
	Complete	0.15 (0.11-0.20)		0.16 (0.12-0.22)		0.22 (0.14-0.36)
Drainage strategy	Nasogastric tube (reference)	1	< 0.001	1	0.005	1	0.480
	Triple-lumen feeding tube	0.47 (0.35-0.63)		0.64 (0.46-0.87)		0.83 (0.50-1.39)
Interaction: TCCS × DS						0.49 (0.26-0.93)	0.030

Model 1 Univariable logistic regression. Model 2 Multivariable logistic regression adjusting for prespecified confounders (age, gender, body mass index, comorbidities, year of admission, lesion location, drainage strategy, and titanium clip closure status). Model 3 Multivariable logistic regression including the interaction term between titanium clip closure and drainage strategy in addition to the confounders in model 2. BMI: Body mass index; OR: Odds ratio; aOR: Adjusted odds ratio; CI: Confidence interval; TCCS: Titanium clip closure status; DS: Drainage strategy; “×” indicates interaction.

Subgroup analysis of the association between clip closure quality and risk of delayed AEs

The subgroup analyses consistently demonstrated an inverse association between complete clip closure and the risk of delayed AEs across the clinically relevant subgroups. Complete clip closure was associated with a greater relative risk reduction (RRR) in delayed AEs across the following subgroups: Age ≤ 59 years (RRR = 86.0%, P < 0.001), female gender (RRR = 92.3%, P = 0.002), non-ampullary lesions (RRR = 85.5%, P < 0.001), admission between 2021 and 2025 (RRR = 89.4%, P < 0.001), and triple-lumen feeding tube management (RRR = 87.3%, P = 0.002). When intervention combinations were compared, patients who with complete clip closure combined with triple-lumen feeding tube management had the lowest observed incidence of delayed AEs (2.8%, 2/71, P = 0.002), which was significantly lower than that in patients with incomplete clip closure managed with a nasogastric tube (33.3%, 9/27, P = 0.013) (Figure 2 and Table 3). Further stratified analyses demonstrated that complete clip closure was consistently associated with a lower risk of delayed AEs across the subgroups of age, gender, BMI, comorbidities, admission year between 2021 and 2025, non-ampullary lesion, and drainage strategy (all subgroups, P < 0.001, Table 4). The ampullary lesions and 2016-2020 admission subgroups demonstrated a similar trend toward risk reduction. However, the association did not reach statistical significance, possibly due to the small sample sizes in these subgroups (Figure 3 and Table 4).

Figure 2 Subgroup χ² analyses of the association between titanium clip closure quality and postoperative complications after duodenal endoscopic submucosal dissection. The patients were stratified by age, gender, body mass index, comorbidity, lesion location, drainage strategy, and year of admission. The χ² test was used within each subgroup to compare the incidence of complications between the two groups. BMI: Body mass index.

Figure 3 Subgroup logistic regression forest plot of titanium clip closure quality and postoperative complications following duodenal endoscopic submucosal dissection. ^cP < 0.001. BMI: Body mass index; CI: Confidence interval.

Table 3 Subgroup χ² analyses of the association between titanium clip closure quality and postoperative complications after duodenal endoscopic submucosal dissection.

Stratification variable	Subgroup	Sample Size (n)	Complete closure complication rate (%)	Incomplete closure complication rate (%)	RRR (%)	P value
Age (years)	> 59	85	5.6 (3/54)	19.4 (6/31)	71.3	0.067
	≤ 59	94	4.8 (3/62)	34.4 (11/32)	86.0	< 0.001
Gender	Male	93	8.3 (5/60)	30.3 (10/33)	72.5	0.014
	Female	86	1.8 (1/56)	23.3 (7/30)	92.3	0.002
BMI (kg/m2)	> 24.0	87	5.2 (3/58)	27.6 (8/29)	81.3	0.005
	≤ 24.0	92	5.2 (3/58)	26.5 (9/34)	80.5	0.008
Comorbidity	No	122	6.4 (5/78)	29.5 (13/44)	78.3	< 0.001
	Yes	57	2.6 (1/38)	21.1 (4/19)	87.5	0.038
Lesion location	Non-ampullary area	156	4.0 (4/101)	27.3 (15/55)	85.5	< 0.001
	Ampullary area	23	13.3 (2/15)	25.0 (2/8)	46.7	0.589
Drainage strategy	Nasogastric tube	72	8.9 (4/45)	33.3 (9/27)	73.3	0.013
	Triple-lumen feeding tube	107	2.8 (2/71)	22.2 (8/36)	87.3	0.002
Admission year	2016-2020	23	21.4 (3/14)	22.2 (2/9)	3.6	1.000
	2021-2025	156	2.9 (3/102)	27.8 (15/54)	89.4	< 0.001

RRR (%) = (complication rate in the incomplete closure group - complication rate in the complete closure group)/complication rate in the incomplete closure group. BMI: Body mass index; RRR: Relative risk reduction.

Table 4 Subgroup logistic regression analyses of the association between titanium clip closure quality and postoperative complications after duodenal endoscopic submucosal dissection.

Stratified variable	Subgroup	N1	n2	OR (95%CI)	P value
Age (years)	> 59	85	9	0.25 (0.16-0.38)	< 0.001
	≤ 59	94	14	0.10 (0.06-0.15)	< 0.001
Gender	Male	93	15	0.21 (0.14-0.32)	< 0.001
	Female	86	8	0.06 (0.04-0.09)	< 0.001
BMI (kg/m2)	> 24	68	7	0.16 (0.09-0.26)	< 0.001
	≤ 24	111	16	0.15 (0.10-0.22)	< 0.001
Admission year	2016-2020	23	4	0.95 (0.41-2.21)	0.913
	2021-2025	156	18	0.08 (0.06-0.11)	< 0.001
Comorbidity	Without	122	18	0.16 (0.11-0.24)	< 0.001
	With	57	5	0.10 (0.06-0.18)	< 0.001
Lesion location	Non-ampullary area	156	19	0.11 (0.08-0.15)	< 0.001
	Ampullary area	23	4	0.46 (0.20-1.09)	0.077
Drainage strategy	Triple-lumen feeding tube	107	10	0.10 (0.07-0.15)	< 0.001
	Nasogastric tube	72	13	0.20 (0.12-0.31)	< 0.001

¹Total number of patients in each subgroup.

²Number of patients with postoperative complications in each subgroup.

BMI: Body mass index; OR: Odds ratio; CI: Confidence interval.

Associations between drainage strategy and early postoperative inflammatory and nutritional indicators

After adjusting for baseline preoperative values (model 1), drainage strategy had a significant effect on the POD 1 WBC count [model 1: F = 4.303, adjusted mean difference (AMD) = -0.120, P = 0.040], Neu% (model 1: F = 19.084, AMD = -1.920, P < 0.001), ALB (model 1: F = 11.437, AMD = 0.400, P < 0.001), and TP (model 1: F = 16.221, AMD = 0.880, P < 0.001) (all P < 0.05). In model 2, after further adjustment for age, gender, BMI, comorbidities, lesion location, year of admission, and titanium clip closure status, the overall trend was consistent with that observed in model 1, and the between-group differences remained statistically significant. On POD 1, based on the AMDs, patients managed with a triple-lumen feeding tube had significantly lower WBC counts (model 2: F = 6.373, AMD = -0.140, P = 0.013) and Neu% (model 2: F = 16.507, AMD = -1.820, P < 0.001) than those managed with a nasogastric tube, whereas ALB (model 2: F = 11.120, AMD = 0.410, P < 0.001) and TP (model 2: F = 14.228, AMD = 0.850, P < 0.001) were higher. The two groups were not significantly different on POD 3 and 5 (all P > 0.05). The explanatory power of the models (R²) was relatively high in both model 1 and model 2, indicating a good fit for the outcome variables. Overall, the associations between drainage strategy and inflammatory and nutritional indicators were prominent only on POD 1 (Figure 4 and Table 5).

Figure 4 Associations between postoperative drainage methods and inflammatory and nutritional indicators following duodenal endoscopic submucosal dissection. A: Model 1. The crude associations between drainage methods and postoperative indicators were assessed using analysis of covariance. Model 1 was adjusted for preoperative baseline values; B: Model 2. Based on model 1, adjustments for age, gender, body mass index, comorbidity, year of admission, lesion location, and titanium clip closure status were made using analysis of covariance to comprehensively control for potential confounding by baseline patient characteristics and analyze the associations between drainage methods and postoperative indicators. ^aP < 0.05. ^bP < 0.01. ^cP < 0.001. ANCOVA: Analysis of covariance; CI: Confidence interval; WBC: White blood cell; TP: Total protein; Neu%: Neutrophil percentage; ALB: Albumin.

Table 5 Associations between postoperative drainage methods and inflammatory and nutritional indicators following duodenal endoscopic submucosal dissection.

	Model 1				Model 2
	F value	P value	AMD	R2 value	F value	P value	AMD	R2 value
WBC-POD 1	4.303	0.040	-0.120	0.919	6.373	0.013	-0.140	0.927
WBC-POD 3	1.534	0.217	0.460	0.071	1.669	0.198	0.490	0.073
WBC-POD 5	0.426	0.515	-0.240	0.002	0.357	0.551	-0.220	0.004
Neu%-POD 1	19.084	< 0.001	-1.920	0.804	16.507	< 0.001	-1.820	0.811
ALB-POD 1	11.437	< 0.001	0.400	0.959	11.120	< 0.001	0.410	0.960
ALB-POD 3	0.518	0.473	-0.350	0.045	0.151	0.698	-0.190	0.075
ALB-POD 5	0.552	0.459	0.360	0.040	0.260	0.610	0.260	0.059
TP-POD 1	16.221	< 0.001	0.880	0.945	14.228	< 0.001	0.850	0.946
TP-POD 3	0.674	0.413	0.690	0.017	0.141	0.708	0.320	0.044
TP-POD 5	0.350	0.555	0.440	0.002	0.830	0.364	0.700	0.017

Model 1 was adjusted for preoperative baseline values only. Model 2 was adjusted for age, gender, body mass index, comorbidity, year of admission, lesion location, and titanium clip closure status. Due to missing data, neutrophil percentage was available only on postoperative day 1. Adjusted mean difference (the difference in postoperative indicator means after adjustment for age, gender, body mass index, comorbidity, lesion location, and titanium clip closure status). F value calculated analysis of covariance test statistic. R² value coefficient of determination. WBC: White blood cell; TP: Total protein; Neu%: Neutrophil percentage; ALB: Albumin; POD: Postoperative day.

Sensitivity analysis excluding ampullary lesions

Among the 156 patients who remained after the exclusion of those with ampullary lesions, complete clip closure (aOR = 0.14, P < 0.001) and postoperative placement of a triple-lumen feeding tube (aOR = 0.52, P < 0.001) remained independently associated with a reduced risk of postoperative complications. The protective association of clip closure with postoperative complications was observed across all subgroups (each subgroup, P < 0.05). Patients receiving both strategies demonstrated the lowest rate of delayed AEs (3.6%, 2/55), consistent with earlier results. These results are presented in the Supplementary Figures 1-5.

DISCUSSION

The results demonstrated that complete titanium clip closure was significantly associated with a reduced risk of delayed AEs after duodenal ESD, and that the staged management strategy of postoperative indwelling of a triple-lumen feeding tube was associated with a lower incidence of complications. Notably, combining the two strategies resulted in the lowest incidence of postoperative complications (2.8%). The results suggested that complete titanium clip closure combined with triple-lumen feeding tube management may represent a perioperative management approach associated with better outcomes, and its potential beneficial mechanisms may be related to multiple factors, including early postoperative isolation from local digestive fluids, control of systemic inflammation, and improvement of nutritional status.

In this relatively large sample, the present study quantitatively demonstrated that complete titanium clip closure was an independent factor associated with a significantly reduced risk of postoperative complications (aOR = 0.16, P < 0.001), corresponding to an approximate 84% risk reduction. This conclusion was highly consistent with Kato et al[6], who demonstrated that complete defect closure significantly reduced the incidence of delayed AEs (OR = 0.055). Furthermore, our results quantified and supported the above conclusion and were highly consistent with the ESGE guideline recommendation[5] advocating that complete defect closure should be achieved whenever possible.

Mechanistically, the duodenal wall is thin, and post-ESD ulcers are persistently exposed to a highly digestive environment composed of bile and pancreatic juice. Residual gaps after incomplete closure may permit ongoing chemical injury to the ulcer base, predisposing patients to delayed bleeding or perforation[9]. Additionally, unstable approximation of the wound edges may expose the ulcer to repetitive mechanical shear stress during duodenal peristalsis and luminal content passage, which may further increase the risk of delayed AEs. Therefore, our results supported the idea that pursuing and achieving complete closure of the duodenal ESD wound using titanium clip technology is an important technical precaution for securing favorable outcomes.

In clinical practice, the titanium clip closure quality is substantially limited by anatomical particularities. Given its relatively low mucosal redundancy, the duodenal bulb is considered a technically challenging area for achieving complete titanium clip closure[10]. Our results were highly consistent with this consensus: As many as 87.3% (55/63) of cases with incomplete closure were concentrated in the duodenal bulb. Additionally, lesions distal to the ampulla had a higher rate of wound reopening at the second endoscopic examination among cases in which complete titanium clip closure was initially achieved. This may be related to limited operating space, difficulty in intraoperative assessment of closure quality, and early clip dislodgement[10,11]. Collectively, these mechanisms compromise the initial titanium clip closure durability and integrity, increasing the risk of delayed complications.

In summary, we believe that achieving and maintaining complete closure with titanium clips is an important technical goal for preventing delayed postoperative complications following duodenal ESD, particularly for lesions distal to the ampulla and in the bulb. Advanced closure techniques such as an over-the-scope clip[12], a string clip suturing method with an anchor (SCSM-A)[13], and threads combined with endoclips[14] have been applied to address these challenges. However, achieving stable and durable complete closure remains a major technical challenge in clinical practice, particularly for high-risk lesions, such as those with a longitudinal location in the duodenal flexure, size > 40 mm, and > 50% involvement of the duodenal circumference[15]. Furthermore, previous studies[6] have suggested that the proportion of delayed AEs was not significantly different between patients with partial closure of the mucosal defect and those in whom it was not closed at all. The present study introduced and evaluated a postoperative proactive management protocol incorporating a triple-lumen feeding tube, aiming to provide important healing support for high-risk wounds. Therefore, introducing proactive postoperative management strategies that compensate for the limitations of closure techniques while pursuing optimal closure methods is important for such high-risk wounds, where achieving and maintaining complete closure is unfeasible.

At the management level, the traditional dietary management protocol following endoscopic resection of ampullary lesions recommended by Espinel et al[16] is typically straightforward. However, this conventional approach may be insufficient for adequately supporting tissue healing and preventing complications following duodenal ESD procedures involving greater trauma and higher complication risks. Some researchers have suggested that delayed perforation following large-scale duodenal ESD procedures may be treated with endoscopic partial closure combined with adequate drainage[17]. The present study determined that leaving a triple-lumen feeding tube in place was independently associated with a lower risk of complications (aOR = 0.64, P < 0.001), reducing the risk of postoperative complications by approximately 36%. Furthermore, using both protective measures together minimized the risk of postoperative complications (2.8%). Based on these results, an interaction analysis between titanium clip closure status and drainage strategy was performed. The results demonstrated a negative correlation between the interaction term and postoperative delayed AEs, and the benefit of the triple-lumen feeding tube was more prominent in the patients with complete titanium clip closure.

Regarding the logistic regression model performance, both multivariable models demonstrated stability and good predictive value [model 2: Area under the curve (AUC) = 0.801, 5-fold cross-validation AUC = 0.786; accuracy = 72.6%, sensitivity = 78.3%, specificity = 71.8%; model 3: AUC = 0.796, 5-fold cross-validation AUC = 0.774; accuracy = 70.9%, sensitivity = 78.3%, specificity = 69.9%]. Overall, model 2 exhibited a slightly better predictive ability than model 3. Model 2 highlighted the independent protective effects of clip closure quality and postoperative drainage strategy in reducing the risk of delayed AEs. Incorporating interaction analysis in model 3 did not significantly improve the overall predictive performance, but revealed the important effect-modifying role of drainage strategy on clip closure status. Taken together, the three models consistently indicated that the strongest protective factor was complete clip closure, whereas ampullary lesions remained an important risk factor. The interaction analysis provided a stratified framework for the individualized selection of drainage strategies, suggesting that the protective effect of complete titanium clip closure was more pronounced in patients with a triple-lumen feeding tube. Overall, these results suggested that ensuring complete clip closure and prioritizing the use of a triple-lumen feeding tube were important strategies for reducing postoperative complications following duodenal ESD, providing essential evidence to guide clinical monitoring and patient management (Figure 1 and Table 2).

Furthermore, the triple-lumen feeding tube group demonstrated milder inflammatory responses and superior nutritional status as early as POD 1 compared to the nasogastric tube group. These early improvements suggested that this approach may reduce wound irritation by locally draining and isolating potent digestive fluids such as bile and pancreatic juice, while simultaneously improving systemic inflammation and nutritional status through early enteral nutritional support. However, the causal relationship with clinical outcomes remained uncertain, and these mechanisms are speculative. Therefore, these findings should be interpreted as hypothesis-generating instead of conclusive evidence of clinical benefit, and these mechanisms require validation in prospective studies.

Considering that ampullary lesions carry a higher risk of complications due to their anatomical specificity (aOR = 1.91, P = 0.005) and represented a smaller proportion in this cohort, we conducted a sensitivity analysis on 156 patients with non-ampullary lesions to validate the robustness of the primary results. The analysis demonstrated that the association between complete clip closure and lower complication risk remained significant in non-ampullary patients (aOR = 0.14, P < 0.001), indicating that this conclusion remained robust in the non-ampullary patient population constituting the main body of the study.

Second-look endoscopy after duodenal ESD remains a subject with limited evidence. The present study enrolled 179 patients, all of whom underwent second-look endoscopy 1-3 days postoperatively to evaluate ulcer healing and provide thorough postoperative monitoring. Although the ESGE guidelines[5] do not recommend the routine use of second-look endoscopy after duodenal ESD, our results provided valuable real-world data on its safety and feasibility, underscoring the need for prospective studies to assess its clinical benefit.

Based on these observations, we propose the following practical implications. Achieving complete clip closure for non-ampullary lesions should be prioritized as a core technical objective whenever feasible. Furthermore, advanced closure techniques may be considered when complete closure is anticipated to be difficult or deemed suboptimal intraoperatively[13,14]. From a perioperative management perspective, combining complete clip closure and staged management using a triple-lumen feeding tube may represent a potentially beneficial strategy, particularly for patients with high-risk features, such as large lesion size[15], distal ampullary location, duodenal bulb involvement, or uncertain closure quality.

For the high-risk subgroup with ampullary lesions, the potential benefits of this combined approach as an enhanced management strategy require validation through prospective studies. Regarding postoperative surveillance, delayed bleeding after duodenal ESD occurs most frequently within the first 3 weeks after the procedure[10]. In patients managed with the combined strategy, the emergence of abnormal or progressively elevated inflammatory markers in the early postoperative period may warrant increased clinical vigilance and timely imaging evaluation, such as contrast-enhanced computed tomography, to exclude delayed perforation or other serious complications.

The present study was subject to the following limitations. First, the single-center retrospective design was inherently susceptible to selection bias. Additionally, unmeasured confounding, including missing data on lesion size, morphology, and histological type, may have limited the comprehensiveness of the analysis and could have influenced the conclusions. Second, although sensitivity analyses supported the robustness of the main results, the relatively few ampullary lesions limited the generalizability of the conclusions for this high-risk subgroup. Third, the change in postoperative management from nasogastric tube to triple-lumen feeding tube reflected the evolution of clinical practice at our center over the study period. Although statistical adjustment was performed to account for potential time period effects, improvements in other aspects, including endoscopist experience, perioperative medication protocols, and nursing practices, may also have occurred over time and could have contributed to the better outcomes observed in the triple-lumen feeding tube group. Therefore, the observed benefits should be interpreted with caution. Finally, the lack of systematic long-term follow-up and serial monitoring of sensitive inflammatory biomarkers (C-reactive protein and procalcitonin) limited the evaluation of long-term outcomes and mechanistic pathways. These results should be validated and extended through further prospective, multicenter studies with predefined interaction analyses.

CONCLUSION

Complete clip closure and staged postoperative management using a triple-lumen feeding tube were independently associated with a lower risk of delayed AEs after duodenal ESD. In descriptive comparisons of the intervention combinations, patients with both complete clip closure and triple-lumen feeding tube management had the lowest incidence of delayed AEs. These results provided real-world evidence to inform the optimization of perioperative management in high-risk duodenal ESD.

ACKNOWLEDGEMENTS

We thank all members of the Peking University Cancer Hospital and Institute Endoscopy Center for the help rendered in this study.