Analysis of Risk Factors for Postoperative Delayed Perforation Following Endoscopic Submucosal Dissection of Gastrointestinal Stromal Tumors

Abstract

Background:

Delayed perforation is a serious complication after endoscopic submucosal dissection (ESD) for gastrointestinal stromal tumors (GISTs). Understanding its incidence, associated risk factors, and clinical outcomes can guide preventative measures. This study aimed to determine the incidence of delayed perforation, identify independent risk factors, and describe the clinical course and management outcomes.

Methods:

We retrospectively identified 57 patients who developed delayed perforation (“cases”). Each case was matched 1:1 by key variables (eg, age, tumor location) with 57 controls who did not experience delayed perforation. Baseline characteristics and the timing of delayed perforation were collected. We compared clinical and procedural factors in univariate analysis and conducted multivariate conditional logistic regression. Clinical course and management outcomes for delayed perforation, and subgroup analyses were assessed robustness.

Results:

A total of 57 patients developed delayed perforation at a median of 2 days postprocedure (IQR 1 to 7). In univariate analysis, larger tumor size (P=0.02) and operator inexperience (<80 ESDs, P=0.03) were associated with an increased risk. In multivariate analysis, tumor size (adjusted OR 1.45 per cm, 95% CI: 1.04-2.02, P=0.03) and operator inexperience (adjusted OR 2.88, 95% CI: 1.21-6.81, P=0.02) remained significant risk factors. Most patients with delayed perforation presented with abdominal pain (61.4%), and diagnosis was primarily made via CT (66.7%). Management strategies included endoscopic closure (31.6%), surgical repair (35.1%), and conservative treatment (33.3%), with a mean hospital stay of 7.5±2.3 days. Sensitivity analyses confirmed the robustness of these findings.

Conclusions:

Larger tumor size and limited operator experience were independent risk factors. Endoscopic or surgical interventions resulted in satisfactory outcomes. These results highlight the need for standardized preventive measures and operator training to mitigate this complication.

Gastrointestinal stromal tumors (GISTs) are the most common mesenchymal tumors of the gastrointestinal tract, predominantly found in the stomach and small intestine.^1,2 The standard management for localized GISTs is surgical resection, which can be performed laparoscopically, while advanced or metastatic cases are treated with tyrosine kinase inhibitors like imatinib.^2–4 Endoscopic submucosal dissection (ESD) has emerged as a less invasive alternative for managing submucosal tumors like GISTs due to its precise dissection and high rates of en bloc resection.^1,5 However, issues including bleeding and perforation happen after ESD. Immediate perforation occurs during the procedure and can be managed promptly, whereas delayed perforation, occurring hours to days postprocedure, poses a greater clinical challenge due to its potential for severe complications and the need for surgical intervention.^1,5

Delayed perforation is a rare but serious complication following ESD for early gastric cancer, with incidence rates reported between 0.1% and 0.91% across various studies.^6–12 Some delayed perforation cases require emergency surgery, while others can be managed conservatively or with endoscopic closure.^7,11,12 Delayed perforation-related mortality rates are low, and many patients recover without significant long-term effects.^10,12 The pathophysiology of delayed perforation is often linked to excessive thermal injury and necrosis of the muscle layer during the ESD procedure.^7,9 Risk factors include the location of the lesion, prolonged cautery time, postoperative stomach conditions, histopathologic ulceration, and excessive electrocautery for hemostasis.^6,9,13 Delayed perforation in GIST ESD is poorly studied, and data on independent risk factors are inconsistent or scarce.^6–8

The primary objective of this study is to identify the independent risk factors following ESD for GISTs. In addition, the secondary objectives are to characterize the clinical presentations and diagnostic pathways of delayed perforation, and to evaluate the subsequent management strategies and outcomes. By comprehensively investigating these aspects, this research aims to enhance our understanding of delayed perforation after ESD, ultimately improving patient safety and informing best practices in endoscopic management of GISTs.

METHODS

Study Design and Setting

We conducted a retrospective, matched case-control study to investigate risk factors for postoperative delayed perforation following ESD of GISTs at The First Hospital of PuTian City. Consecutive patients who underwent ESD for GISTs between January 2015 and December 2022 were evaluated. The study protocol was approved by the Ethics Committees of The First Hospital of PuTian City. A waiver of informed consent was granted due to the retrospective nature of the study. All procedures followed the Declaration of Helsinki principles and local regulatory guidelines.

Inclusion Criteria

(1) Adult patients who underwent ESD for histopathologically confirmed GISTs; (2) complete medical records with sufficient follow-up data (at least 14 d) to assess postoperative complications; (3) procedure performed by board-certified endoscopists with documented experience in ESD.

Exclusion Criteria

(1) Immediate perforation identified and treated during the procedure; (2) immediate perforation identified intraprocedurally; (3) incomplete procedures or aborted ESD where submucosal dissection was not finished; (4) insufficient documentation (missing key variables or lost to follow-up before 30 d); (5) patients with recurrent GISTs who underwent repeat ESD or additional surgical resections within 30 days post-ESD.

Delayed Perforation

Delayed perforation was defined as a transmural defect identified after completion of the ESD procedure (postextubation), confirmed by symptoms, imaging (free air on CT), or endoscopic visualization. Our analytic window was 14 days. Case ascertainment extended to 30 days, and no events occurred beyond day 14 in this cohort. Because some groups define “delayed” as >24 hours, we also evaluated >24 hours and ≤7 days definitions in sensitivity analyses.

Immediate Perforation

Immediate perforation was defined as any intraprocedural perforation noted during the ESD, evidenced by direct visualization of the muscular layer defect or extraluminal structures. Immediate perforation was excluded from this study.

High Versus Low Operator Experience

A threshold of 80 prior ESD procedures was used to categorize operator experience, based on existing literature and consensus in endoscopic practice.

Case Identification

We searched the endoscopy database to identify patients who developed delayed perforation within 30 days post-ESD for GISTs. A total of 57 cases were confirmed.

Control Selection

From the same database and time period, we identified patients who did not develop perforation post-ESD for GISTs. For each case, one control was matched (1:1) on key confounders, including age (±5 y) and tumor location (eg, stomach vs. other GI sites), creating 57 matched pairs.

Endoscopic Procedure

All ESDs were performed with therapeutic‑channel upper GI endoscopes under CO₂ insufflation. Lesions were marked (perimeter coagulation), the submucosal layer was expanded with a tinted injectate to enhance plane identification, followed by mucosal incision and submucosal dissection. Knife selection (needle‑type, insulated‑tip, or hook‑type) and use of traction were at the operator’s discretion according to lesion characteristics. Electrosurgery used a pulsed‑cut mode for mucosal incision and controlled coagulation for submucosal dissection, using short, titrated activations to minimize thermal injury, and visible vessels were treated with hemostatic forceps. Specific brand/models, generator numerical settings, and exact injectate formulations varied by operator and case and were not uniformly recorded in the registry.

Defect Closure (Prophylactic) and Periprocedural Management

Prophylactic closure of the post‑ESD mucosal defect was at operator discretion, taking into account defect size, location, exposure of muscularis propria, and perceived risk of delayed adverse events. Closure methods comprised: (1) clip‑only approximation of the defect edges using through‑the‑scope clips; (2) clip-endoloop purse‑string closure for larger defects or locations requiring circumferential apposition. Prophylactic closure status (attempted vs. not) was abstracted from standardized operative reports. Closure adequacy/technique was not graded.

Postprocedural Management

At our center, advanced gastric endoscopic resections (including ESD) are performed with carbon‑dioxide insufflation and followed by a uniform postoperative protocol. For this ESD cohort, the standard regimen comprised:

Immediate Care and Decompression

Patients were observed in the recovery area with serial abdominal examinations and vitals monitoring. Gastrointestinal decompression (nasogastric tube to low intermittent suction) was routinely maintained in the early postoperative period, with removal based on clinical stability and return of bowel function.

Acid Suppression

Proton‑pump inhibitor therapy was initiated postprocedure (intravenous during the initial observation period, then oral on advancement of diet) to promote mucosal healing.

Antibiotics

Intravenous broad‑spectrum antibiotics were administered postoperatively and continued according to clinical assessment (eg, presence of fever, peritoneal signs, or imaging findings).

Diet and Activity

Patients were kept nil per os initially, with stepwise advancement from liquids to soft diet once pain, exam, and labs were reassuring. Early ambulation was encouraged.

Monitoring and Imaging Triggers

Routine labs and symptom assessment were performed during observation. Contrast‑enhanced CT (or targeted endoscopy) was obtained if patients developed abdominal pain, fever, leukocytosis, or peritoneal signs to evaluate for post‑ESD adverse events such as delayed perforation; this diagnostic pathway is reflected in our cohort (CT was the most frequent modality at diagnosis).

Escalation/Management Algorithm

When delayed perforation was diagnosed, management followed a predefined pathway: endoscopic closure for stable, contained defects; surgical repair (laparoscopic or open) for generalized peritonitis, large defects, or failed endoscopic therapy; and conservative treatment (decompression, antibiotics, PPIs, and close observation) for selected cases without diffuse peritoneal signs.

Statistical Analysis

Categorical variables were summarized as counts and percentages, while continuous variables were reported as mean±SD or median (interquartile range [IQR]) depending on distribution. We compared potential risk factors between cases and controls using appropriate tests for matched data. Categorical variables were assessed via McNemar test, and continuous variables were analyzed via paired t test or Wilcoxon signed-rank test if distributions were non-normal. Variables with P<0.20 in univariate analyses or deemed clinically relevant were entered into a conditional logistic regression model to identify independent risk factors. Adjusted odds ratios (OR) with 95% CI and P values were reported. To minimize overfitting with 57 events, we prespecified a parsimonious multivariable model (2 primary predictors, tumor size and operator experience, plus closure status). We used conditional logistic regression for inference in the matched case-control sample. Model discrimination was summarized by a C‑statistic (AUC) from an unconditional logistic model fitted on the matched sample with cluster‑robust standard errors, reported descriptively. Goodness‑of‑fit tests were not applied to the conditional model. We performed exploratory analyses based on tumor size categories, tumor location, and operator experience. To assess the robustness of findings, we repeated the multivariate models under varied definitions of “delayed” perforation and excluded borderline cases with uncertain timing. The primary threshold for operator experience was <80 lifetime ESDs. Given heterogeneity in the literature, we also evaluated alternative cut‑offs in sensitivity analyses. All statistical analyses were performed using R (version 4.4; The R Foundation, Vienna, Austria). A 2-sided P<0.05 was considered statistically significant.

RESULTS

A total of 57 patients who experienced delayed perforation underwent ESD for GISTs and 57 matched no-perforation controls who did not experience perforation were included in this study. The mean age was 62.2±9.9 years, and 63% were male (Table 1). The majority of tumors were located in the stomach (73.7%). Among the nongastric GISTs in the matched sample (30/114; 26.3%), the distribution was duodenum 12, colon/rectum 10, esophagus 6, and jejunum/ileum 2 (Table 1); by design, each group contributed 15 nongastric lesions (Table 1). Median tumor size was 2.8 cm (IQR 2.0 to 3.5) in delayed‑perforation cases and 2.5 cm (IQR 2.1 to 3.0) in matched controls (Table 1).

TABLE 1.

Baseline Characteristics of Patients With and Without Delayed Perforation

	Delayed perforation (n=57)	No delayed perforation (n=57)	P
Age (y), mean±SD	62.5±10.2	61.8±9.6	N/A
Sex, n (%)			N/A
Male	36 (63.2)	36 (63.2)
Female	21 (36.8)	21 (36.8)
BMI (kg/m2), mean±SD	24.5±3.2	24.2±3.5	0.72
Comorbidities, n (%)
Hypertension	20 (35.1)	16 (28.1)	0.38
Diabetes mellitus	14 (24.6)	10 (17.5)	0.36
Anticoagulant/antiplatelet Use	12 (21.1)	9 (15.8)	0.48
Tumor characteristics
Location (Stomach), n (%)	42 (73.7)	42 (73.7)	N/A
Location (Other GI sites), n (%)	15 (26.3)	15 (26.3)	N/A
Esophagus	3 (5.3)	3 (5.3)
Duodenum	6 (10.5)	6 (10.5)
Jejunum/Ileum	1 (1.8)	1 (1.8)
Colon/rectum	5 (8.8)	5 (8.8)
Tumor size (cm), median (IQR)	2.8 (2.0-3.5)	2.5 (2.1-3.0)	0.42
Depth (submucosal), n (%)	48 (84.2)	45 (78.9)	0.46
Procedure-related factors
Procedure duration (min), median (IQR)	65 (45-90)	60 (40-80)	0.39
Operator experience, n (%)
High (≥80 ESDs performed)	25 (43.9)	32 (56.1)	0.16
Low (<80 ESDs performed)	32 (56.1)	25 (43.9)
Prophylactic closure attempt, n (%)	10 (17.5)	16 (28.1)	0.19
Intraoperative bleeding, n (%)	5 (8.8)	3 (5.3)	0.72

For context, we provide a descriptive, unmatched comparison of immediate perforation cases from the study period alongside the delayed perforation cohort and their 1:1 matched no‑perforation controls (Supplementary Table S1, Supplemental Digital Content 1, http://links.lww.com/SLE/A512). Because the immediate‑perforation group was not part of the matched case-control design, no hypothesis testing is presented there.

The median time to clinical detection of perforation was 2 days (IQR 1 to 7) (Table 2). A majority of perforations (42.1%) were diagnosed within 24 hours, while only 8.8% were detected between Days 8 and 14 (Table 2). Although ascertainment extended to 30 days, all events occurred between days 1 and 14.

TABLE 2.

Incidence and Timing of Delayed Perforation Following ESD for GISTs

	Value
Total ESD procedures for GISTs during study period	4327
Incidence of delayed perforation, n (%)	57 (1.3)
Median time to detected perforation, days (IQR)	2 (1-7)
Range of time to perforation, days	1-14
Detailed time to event distribution, n (%)
0-24 h after ESD	24 (42.1)
Day 2-3	19 (33.3)
Day 4-7	9 (15.8)
Day 8-14	5 (8.8)

Larger tumor size (median 2.8 cm vs. 2.5 cm) and less operator experience (<80 ESDs) were more frequent in the delayed perforation group, both factors reached statistical significance in univariate testing (P=0.02 and P=0.03, respectively, Table 3). Other factors, including comorbidities (hypertension, diabetes), anticoagulant use, and prophylactic closure attempts, did not show significant associations with delayed perforation (P>0.05, Table 3). In the matched univariate model, the available binary depth proxy was not associated with delayed perforation (OR 1.45; 95% CI: 0.56-3.74; P=0.46), so it was not retained in the multivariable analysis. In matched univariate models, age showed no association with delayed perforation (OR 1.01; 95% CI: 0.98-1.05; P=0.65), and diabetes was not associated (OR 1.52; 95% CI: 0.61-3.77; P=0.36). Antiplatelet/anticoagulant use was also not associated (OR 1.40; 95% CI: 0.55-3.59; P=0.48). Procedure duration did not correlate with delayed perforation (P=0.39), and electrocoagulation time was not available.

TABLE 3.

Comparison of Clinical and Procedural Factors: Univariate Analysis

Variable	Univariate OR (95% CI)	P
Age (y), mean±SD	1.01 (0.98-1.05)	0.65
Male sex, n (%)	1.00 (0.50-2.01)	1.00
Hypertension, n (%)	1.38 (0.66-2.86)	0.38
Diabetes mellitus, n (%)	1.52 (0.61-3.77)	0.36
Anticoagulant/antiplatelet Use, n (%)	1.40 (0.55-3.59)	0.48
Location (stomach), n (%)	1.00 (0.46-2.17)	1.00
Tumor size (cm), median (IQR)	1.52 (1.07-2.16)	0.02
Depth (submucosal), n (%)	1.45 (0.56-3.74)	0.46
Procedure duration (min), median (IQR)	1.01 (0.99-1.03)	0.39
Operator experience <80 ESDs, n (%)	2.37 (1.07-5.26)	0.03
Prophylactic closure attempt, n (%)	0.55 (0.21-1.41)	0.09
Intraoperative bleeding, n (%)	1.73 (0.40-7.37)	0.72

Variables with P<0.20 in univariate analysis or deemed clinically important (eg, tumor size, operator experience, procedure duration) were entered into a conditional logistic regression model. Operator inexperience (<80 ESDs) was independently associated with a higher risk of delayed perforation [adjusted odds ratio (OR) 2.88, 95% CI: 1.21-6.81, P=0.02], whereas increasing tumor size was also a significant risk factor (adjusted OR 1.45 per 1 cm increase, 95% CI: 1.04-2.02, P=0.03, Table 4). Procedure duration showed a borderline association (P=0.05, Table 4). The Hosmer-Lemeshow test (P=0.38) and a C-statistic of 0.82 indicated good model fit (Table 4).

TABLE 4.

Multivariate Risk Factor Analysis for Delayed Perforation

Variable	Adjusted OR (95% CI)	P
Tumor size (per 1 cm increase)	1.45 (1.04-2.02)	0.03
Operator experience <80 ESDs	2.88 (1.21-6.81)	0.02
Prophylactic closure attempt	0.53 (0.19-1.47)	0.22

Because operator inexperience (<80 ESDs) and larger tumor size were the independent risk factors in the conditional model, we added a simple risk stratification using operator experience alone as a pragmatic clinical classifier. In the matched sample, 32/57 (56.1%) of delayed‑perforation cases occurred in the high‑risk group (<80 ESDs) versus 25/57 (43.9%) among matched controls, with the complementary pattern for the low‑risk (≥80 ESDs) group [25/57 (43.9%) vs. 32/57 (56.1%)]. These descriptive gradients mirror our inferential findings: operator inexperience was associated with delayed perforation in univariate conditional analysis (OR 2.37, 95% CI: 1.07-5.26) and remained significant in multivariable modeling (aOR 2.88, 95% CI: 1.21-6.81) (Supplementary Table S2, Supplemental Digital Content 2, http://links.lww.com/SLE/A513).

We evaluated prophylactic closure attempted (yes/no) as a proxy. Prophylactic closure showed a protective trend in univariate analysis (OR 0.55, 95% CI: 0.21-1.41; P=0.09) but was not independently associated with delayed perforation in the multivariable model (aOR 0.53, 95% CI: 0.19-1.47; P=0.22). These data do not support a statistically significant association between delayed perforation and closure status in our cohort, although the direction of effect suggests potential benefit.

Among the 57 patients with delayed perforation, the most common presenting symptom was abdominal pain (61.4%), followed by peritonitis (26.3%) and fever (17.5%; Table 5). Diagnosis was primarily made via CT scan (66.7%), often revealing pneumoperitoneum or fluid collection. Management strategies varied: 31.6% received endoscopic closure, 35.1% underwent surgical repair, and 33.3% were managed conservatively (eg, antibiotics and observation). The mean hospital stay was 7.5±2.3 days; overall outcomes were favorable, with an 8.8% complication rate (eg, abscess) and 1.8% mortality.

TABLE 5.

Clinical Course and Management of Delayed Perforation

Parameter	n (%)/value
Clinical presentation
Abdominal pain	35 (61.4)
Peritonitis/guarding	15 (26.3)
Fever	10 (17.5)
Mild or nonspecific symptoms	8 (14.0)
Diagnostic Method
CT scan (pneumoperitoneum/abscess)	38 (66.7)
Endoscopy (visualizing perforation site)	15 (26.3)
X-ray (pneumoperitoneum)	4 (7.0)
Combination of imaging/endoscopy	12 (21.1)
Management strategy
Endoscopic clipping or suturing	18 (31.6)
Surgical repair (laparoscopic or open)	20 (35.1)
Conservative treatment (eg, IV antibiotics, observation)	19 (33.3)
Outcome of Management
Mean hospital stay (d), mean±SD	7.5±2.3
Additional complications (eg, abscess)	5 (8.8)
30-day rehospitalization	4 (7.0)
Mortality	1 (1.8)

Using the adjusted effects from the final model, we constructed a simple points‑based index combining operator experience and tumor size. The rubric assigns 3 points for operator inexperience (<80 ESDs) and 0 to 3 points for tumor size (<2.0 cm=0; 2.0 to 2.9 cm=1; 3.0 to 3.9 cm=2; ≥4.0 cm=3). Model‑based relative odds of delayed perforation rose monotonically with the total score, from ∼1.0 at 0 points to ∼8.8 at 6 points, reflecting the multiplicative combination of a size effect (∼1.45 per cm) and operator inexperience (∼2.88) (Supplementary Table S3, Supplemental Digital Content 3, http://links.lww.com/SLE/A514).

To test the robustness of our findings, we performed several sensitivity analyses (Table 6). Excluding borderline cases (n=5) or redefining delayed perforation to ≤7 days did not materially change the significance of tumor size and operator inexperience in the multivariate model. Similarly, extending the follow-up window to 14 days increased the absolute number of delayed perforations to 57 but did not alter the primary conclusions.

TABLE 6.

Sensitivity Analyses for Delayed Perforation Risk Factors

Sensitivity analysis	Sample size (n)	Tumor size OR (95% CI)	Operator experience <200 ESDs OR (95% CI)	P
Excluding borderline cases	52 delayed cases	1.38 (1.00-1.91)	2.57 (1.10-6.20)	0.04
Narrower time window (≤7 d)	52 delayed cases	1.63 (1.11-2.40)	3.15 (1.16-8.57)	0.02
Extended follow-up (≤14 d)	57 delayed cases	1.30 (0.96-1.77)	2.92 (1.25-6.82)	0.06

DISCUSSION

Our study demonstrates that the incidence of delayed perforation following ESD for GISTs is a clinically significant concern. Tumor size and operator experience emerged as independent risk factors, highlighting that both intrinsic lesion characteristics and procedural expertise critically influence postoperative outcomes. Moreover, the frequent presentation of abdominal pain and peritonitis emphasizes the need for vigilant monitoring and timely use of imaging or endoscopic assessments to ensure early detection.

Our study identified a 1.3% incidence of delayed perforation following ESD for GISTs, which is higher than the 0.1% to 0.91% range reported in multiple retrospective and prospective analyses.^6–10,12 While some investigators have observed specific rates of 0.42% and 0.25% in their cohorts, others emphasize that delayed perforation remains clinically significant despite its rarity.^7,9,12 Our findings highlight the role of tumor size and operator experience—observations that partly align with studies citing older age, fibrosis, location in the upper stomach, and extended electrocautery time as additional contributors to delayed perforation.^9,13,14 Nonetheless, conflicting results exist in the literature, with certain research suggesting that tumor size and operator experience are less influential compared with other procedural or anatomic considerations.^13,15

Several factors may explain both the similarities and discrepancies between our findings and previous work. First, the retrospective nature of our study design was prone to selection biases and variability in follow-up.^16–18 Differences in sample size and inclusion criteria also shape reported incidences.^6,19 Moreover, technical variations in ESD procedures, such as the use of specific electrosurgical knives, prophylactic clipping methods, and each endoscopist’s learning curve, could account for varied perforation rates across centers.^6,13,19 Finally, biological features of GISTs, including tumor depth and location, may predispose certain lesions to higher perforation risk; tumors near the cardia or with prominent fibrosis pose unique challenges that may not be uniformly represented in all cohorts.^6,8,19

Delayed perforation after ESD for early gastric cancer and GIST is uncommon^6–8 and typically emerges 11–29 hours after resection.^12,20 In our matched case-control GIST cohort, the incidence was 1.3%, the median time to detection was ∼2 days (IQR 1 to 7) with 42.1% within 24 hours, and prophylactic mucosal defect closure again showed only a nonsignificant protective trend, underscoring that closure alone cannot offset deep muscle injury. Our analysis larger tumor size (aOR 1.45 per cm) and operator inexperience (<80 ESDs; aOR 2.88) were the independent predictors, prior study reported that patients with delayed perforation were significantly older than controls and had a median detection time of 28.8 hours (14 to 71 h), whereas Yamamoto et al⁹ found all cases within 24 hours and highlighted upper-stomach location and prolonged hemostatic cautery as key signals.¹² These differences likely reflect design and population: we matched on age and organ-level location, studied subepithelial GISTs rather than epithelial gastric neoplasia, and did not capture granular electrocautery/device metrics or gastric sublocation, limiting direct testing of age/location/cautery hypotheses. Nonetheless, all series indicate that most delayed perforations present early (≤24 to 48 h), underscoring the need for close early monitoring and prompt CT-based evaluation of symptoms after ESD.^9,12 While we cannot exclude occasional unrecognized micro‑perforations, our analyses did not show an adverse association with prophylactic closure (aOR 0.53), arguing against “inadequate suturing” as a primary driver. A limitation is that we did not capture closure adequacy or a standardized intraprocedural “suspected micro‑perforation” variable, precluding definitive attribution in early‑detected cases.

In clinical practice, recognizing that the incidence of delayed perforation in GISTs may be higher than previously reported underscores the importance of careful patient selection and thorough preoperative assessments. Given that both excessive electrocautery and technical inexperience have been identified as risk factors, endoscopists might benefit from more conservative dissection techniques, especially for large or fibrotic lesions, and from targeted training programs aimed at improving competency in ESD.^6,13,19 Endoscopists at earlier stages of their learning curve might consider collaborating with more experienced peers or adopting a stepwise mentoring process.^21–23 For patients deemed high risk, routine postoperative imaging within the first few days can facilitate early detection of perforation. In addition, standardized protocols incorporating prophylactic measures—such as preventive clipping or suturing—could help minimize perforation risk, particularly in high-risk patients (eg, older adults or those with a known fibrotic response).^14,15,24,25 Larger tumor size and operator inexperience were the only independent predictors in our study, supporting a pragmatic bedside cue that lesions ≥3 cm and experience <80 cases merit heightened vigilance and preventive strategies. To facilitate translation, our exploratory points-based index (size tiered with the <80-case threshold) provides a simple gradient for risk awareness, not clinical decision-making. On the basis of prior reports and guidance, additional factors are plausible contributors, including upper-stomach/lower-esophageal subsite, advanced age, diabetes, periprocedural NSAID exposure, and prolonged electrocautery/hemostasis time, and deserve prospective, standardized capture.^6,8,9,21 In our dataset, these could not be fully evaluated due to that gastric subsite and cautery time were not uniformly recorded, age was a matching variable, diabetes and antithrombotic use showed no significant association in matched analyses, and NSAID exposure was not systematically abstracted. Consequently, we recommend using size ≥3 cm and experience <80 as data-supported cues for case selection, closure strategy, and early post-ESD monitoring, while treating the other factors as testable hypotheses for future multicenter registries.^6,8,9,21

Delayed perforation likely reflects an interplay of patient, lesion, operator, device, and perioperative management domains. In our matched analysis, larger tumor size and operator inexperience were the only independent predictors, while comorbidities, anticoagulation, and prophylactic closure did not reach significance. We lacked device/accessory‑level data and a standardized closure completeness metric, so we could not evaluate those domains quantitatively. To illustrate clinical application, we provide an exploratory points‑based index, which yields a consistent gradient in model‑based relative odds. Prospective, multicenter datasets with device‑level annotation are needed to build and validate a comprehensive score that also incorporates accessory quality and perioperative protocols.

However, we must acknowledge the limitations inherent to a retrospective design, including potential selection bias and underreporting of complications, as well as a limited sample size in certain subgroups. In particular, granular depth stratification (eg, muscularis propria involvement), ulceration, and preoperative NSAID/COX‑2 exposure were not uniformly available and were therefore not included. We did not capture closure technique (eg, direct vs. purse‑string) or a standardized measure of closure adequacy, precluding a technique‑level comparison and limiting causal inference about suturing quality and delayed perforation. A standardized grading of “closure completeness” was not collected in this retrospective registry. Specifically, we did not capture a uniform score of tissue apposition, nor a protocolized intra‑procedural leak test. Therefore, “suturing incompleteness” could not be evaluated as a covariate. Future investigations could benefit from prospective validation of preventive strategies, potentially via randomized trials comparing different prophylactic closure techniques, and larger, multicenter studies or meta-analyses aimed at confirming our identified risk factors on a broader scale. In addition, innovations in endoscopic devices (such as advanced closure systems) and simulation-based training modules may help reduce the likelihood of delayed perforation, emphasizing the need for continued progress in endoscopic technology and operator education.

In this matched case-control cohort, larger tumor size and operator inexperience were independently associated with delayed perforation after GIST ESD. Given the retrospective design and limited sample size, these findings should be viewed as hypothesis‑generating and warrant prospective, multicenter confirmation that includes standardized depth and medication variables.

Supplementary Material

sle-36-e1438-s001.docx ^{(16.6KB, docx)}

sle-36-e1438-s002.docx ^{(14.5KB, docx)}

sle-36-e1438-s003.docx ^{(15.3KB, docx)}