Revisional surgery for delayed gastric conduit emptying after esophagectomy: the RESTORANT project

Summary

Delayed gastric conduit emptying (DGCE) is a debilitating functional complication after esophagectomy. Because of a lack of large-scale outcome data, this project evaluated surgical revisional strategies for refractory DGCE within a recently published mechanism-informed classification system. REvisional Surgery for delayed gasTRic conduit emptying (RESTORANT) was a multicentric, retrospective cohort study including adult patients undergoing revisional surgery for DGCE from January 1, 2015, to December 31, 2024, across 26 high-volume centers that performed 11,807 esophagectomies during this period. Patients were eligible irrespective of the date of their index esophagectomy. Interventions were stratified into Class 1 (pylorus-related, C1), Class 2 (conduit-related, C2), and Class 3 (hiatus-related/paraconduit hernia, C3), with class assigned retrospectively according to the revisional procedure performed. The primary endpoint was reported symptom resolution. A total of 196 patients underwent 263 surgical interventions for DGCE. Sustained symptom relief was comparable across all classes: C1 (47.2%), C2 (58.2%), and C3 (63.9%; P = 0.165). For C1, pyloroplasty (47.6%) and Roux-en-Y diversion (46.7%) showed similar efficacy. In C2, conduit replacement (75.0%) and remodeling (61.3%) were significantly more effective than simple mobilization (50.0%; P = 0.035). For C3, paraconduit hernia repair achieved higher resolution than crural division (65.8% vs. 47.1%). The overall complication rate was 32.7%, with significantly higher major morbidity and mortality in C2 (20.0% and 3.6%) compared with C3 (13.4% and 1.7%) and C1 (16.7% and 2.8%, P = 0.047). Multivariable analysis identified no independent predictors of surgical success but was limited by heterogeneity and sample size. RESTORANT provides the first multicentric analysis of surgical DGCE repair. While symptom resolution is achievable across all classes, C2 interventions carry increased risks. These findings may help structure future prospective audits.

INTRODUCTION

Esophagectomy with gastric conduit reconstruction is the curative cornerstone for esophageal cancer. As survival rates improve, focus has shifted toward long-term physiological consequences, since up to 80% of patients report lasting digestive discomfort.^1–6 Delayed gastric conduit emptying (DGCE) is perhaps the most challenging sequela, frequently resulting in debilitating symptoms such as regurgitation, early satiety, and significant weight loss.^1,7–9

The pathophysiology of DGCE involves complex, often overlapping anatomical and functional abnormalities. While international consensus has recently standardized diagnostic criteria,^10,11 our group has further refined the management framework by publishing a classification system based on the underlying mechanism: Class 1 (pylorus-related), Class 2 (conduit-related, such as redundancy/dilation), and Class 3 (hiatus-related, specifically paraconduit hiatal hernia [PCH] and hiatal narrowing).¹²

Standard therapy for DGCE relies primarily on conservative and endoscopic interventional measures. However, a minority of patients suffer from refractory symptoms or mechanical obstacles, which necessitate surgical revision.^9,13–15 While these procedures aim to address the respective mechanical or functional roots, current literature is limited to small, single-center case series.¹⁴ This lack of robust, comparative data hinders the development of standardized treatment pathways.

The REvisional Surgery for delayed gasTRic conduit emptying (RESTORANT) project was established to address this gap. This multicentric, international cohort study systematically evaluated the clinical presentation, surgical repair techniques, and outcomes of revisional surgery across the diverse DGCE spectrum. By analyzing a large-scale, real-world dataset, we aimed to provide a descriptive account of contemporary practice and outcomes, and thereby a basis for the design of future prospective studies in this complex functional complication.

METHODS

Study design, cohort, and data collection

This multicentric retrospective analysis included patients (≥18 years) undergoing revisional surgery (elective or emergent) for DGCE and PCH from January 1, 2015, to December 31, 2024, across 26 high-volume centers (≥25 esophagectomies/year). Expert centers were identified and invited through personal invitations and network (International Society for Diseases of the Esophagus) advertisement (invitation sent to 42 centers). Ethical approval was obtained from the lead center in Zurich, Switzerland (BASEC-No. 2024-00015) and participating sites.

De-identified data included demographics, index esophagectomy details, and complications (ECCG standardized data¹⁶), with major morbidity defined as Clavien-Dindo ≥IIIa.¹⁷ DGCE symptoms were assessed via Konradsson criteria¹⁰; patients with PCH were included only if they reported at least two moderate-to-severe DGCE symptoms. Interventions were categorized by a recently published classification: Class 1 (antro-pyloric, C1), Class 2 (conduit-related, C2), and Class 3 (hiatus-related, C3).¹² DGCE class was assigned retrospectively by the central study team on the basis of the dominant pathology addressed by the revisional procedure, as documented in the operative report. Class assignment was not based on preoperative phenotyping, as no standardized and validated tests for DGCE itself or its respective classes exist. This design therefore carries an inherent risk of circularity, in that the presumed mechanism is inferred from the treatment chosen by the operating surgeon rather than established independently. Preoperative diagnostic findings were captured separately and are reported alongside the assigned class. Eligibility required at least one revisional surgical intervention for DGCE performed within this 10-year window. Patients were eligible irrespective of the date of their index esophagectomy. For every eligible patient, the complete DGCE-related interventional history was recorded, including surgical and endoscopic procedures performed before 2015 or after 2024; recorded interventions therefore span a longer time frame than the 10-year study period. The final follow-up cutoff was October 1, 2025. The lead center in Zurich audited all securely transmitted data for integrity and completeness.

Endpoints and symptom assessment

Baseline demographics, index esophagectomy characteristics, the incidence of revisional DGCE surgery, and follow-up duration are reported at patient level (n = 196). Symptom resolution, procedural characteristics, complications, mortality, hospitalization, and repeat interventions are inherently procedure-specific and are reported at intervention level (n = 263); the applicable denominator is stated in the legend of each table and figure. Forty-six patients contributed more than one intervention.

The primary endpoint was symptom resolution following surgical intervention, measured on a 5-point Likert scale (ranging from ‘absence of symptoms/a lot better’ to ‘worsening’), and separated by DGCE class and surgical intervention. Preoperative symptoms and symptom resolution were collected retrospectively from contemporaneous clinical records; no validated patient-reported outcome instrument was applied, and baseline symptom severity was reconstructed from documentation rather than prospectively graded. Secondary endpoints included overall and major morbidity, the efficacy of pre- and postoperative endoscopic interventions, and the identification of predictors of successful surgical outcomes, which were defined as ‘absence of symptoms/a lot better.’

Statistical analysis

Baseline characteristics were summarized at patient level, while procedural characteristics were analyzed at intervention level (specified for each analysis in the manuscript). Comparisons in intervention-level descriptive tables were adjusted for clustering of repeated interventions within patients using survey-based clustered tests. Categorical variables are presented as number (n)/percentage (%) and compared with Fisher’s exact test. Continuous variables are expressed as median and interquartile range (IQR) and compared by the Wilcoxon rank-sum test. Associations between variables and binary outcomes were analyzed using generalized multivariable linear mixed-effects models, with study center and patient identifier included as random effects, thereby accounting for both between-center heterogeneity and within-patient clustering of repeated interventions. Results are reported as odds ratios (ORs) with corresponding 95% confidence intervals (CIs). Covariates were selected based on clinical rationale; variables for multivariable analysis were chosen based on a P-value threshold of <0.2 in univariate analysis. Missing covariate data were handled by complete-case analysis. The number of interventions and outcome events entering the model, the number of covariates and the resulting number of events per variable are reported in the respective text passages and table legends. Statistical significance was defined as P < 0.05. R V4.0.2 (Vienna, Austria) was used for statistical analyses and figure generation.

RESULTS

Study cohort and participating centers

Across 26 international centers (24 in Europe, 1 in Asia, and 1 in North America), which performed a total of 11,807 esophagectomies (center-specific range: 169–1673 index operations) during the study period (8418 minimally invasive, 71.3%), data from 366 patients undergoing 888 interventions between May 2008 and August 2025 were recorded. Following the exclusion of 170 patients with interventions unrelated to DGCE or PCH not fulfilling DGCE criteria, the final study population comprised 196 patients undergoing 263 surgical interventions (Fig. 1). The overall median of contributed cases per center was 9 (range 1–48) and the median of revisional surgeries for DGCE after exclusion of the above-mentioned cases was 4 (range 0–33). The overall rate of revisional DGCE surgery to primary esophagectomies was 1.66% and ranged from 0% to 8.36% across centers.

Fig. 1.

Study flow chart. (A) Patient selection (patient level): of 366 patients screened, 38 were excluded because all recorded procedures were unrelated to delayed gastric conduit emptying (DGCE) and 132 because they underwent paraconduit hernia repair or adhesiolysis without fulfilling DGCE criteria, leaving a final study population of 196 patients. (B) Distribution of the 263 revisional surgical interventions performed in these 196 patients (intervention level), each retrospectively assigned to DGCE Class 1 (n = 36), Class 2 (n = 55), or Class 3 (n = 172) according to the dominant pathology addressed by the procedure; because 46 patients contributed more than one intervention, the class groups are not mutually exclusive at patient level. The 208 endoscopic procedures recorded in this cohort are also intervention-level data.

Revisional interventions and DGCE class distribution

Interventions were retrospectively classified by the primary pathophysiological mechanism: C1 targeting pyloric dysfunction (13.6%, pyloroplasty, n = 21; Roux-en-Y diversion, n = 15), C2 targeting conduit dilatation/redundancy (20.9%, conduit repositioning, n = 20; remodeling (revision of the existing gastric conduit intended to correct geometry and/or outflow, n = 31); replacement (removal of the gastric conduit with reconstruction using an alternative substitute, e.g., colon or small intestine, n = 4), and C3 targeting hiatus-related pathology (65.3%, PCH repair, n = 155; crural division, n = 17). Mesh use for PCH repair was reported in 19.1% (7.9% biological resorbable, 5.3% synthetic resorbable, 5.2% synthetic non-resorbable, and 0.7% unknown mesh) and had no influence on the need for repeat surgical intervention. Timing of revisional surgery varied by class; C1 and C3 procedures were performed at a median of 455 and 375 days post-esophagectomy, respectively, whereas C2 re-operations occurred significantly later (median 879 days).

Patient and index esophagectomy characteristics

The study cohort was predominantly male (78.1%) with a median age of 61 years (IQR: 54–69) and a BMI of 25 (IQR: 22–28). Index esophagectomies were performed between June 1972 and November 2024, primarily for adenocarcinoma (77%) and squamous cell carcinoma (17%) and did not differ significantly between classes. Most tumors were located at the gastroesophageal junction (39.8%) or lower esophagus (42.9%), with 94.4% of patients receiving an intrathoracic anastomosis. Index esophagectomies were largely minimally invasive (abdominal: 82.7%; thoracic: 60.7%). Routine pyloric drainage at the time of surgery had been performed in 25% of patients, including dilatation (n = 30), pyloroplasty/−myotomy (n = 17), G-POEM (Gastric Peroral Endoscopic Myotomy, n = 1), or diversion (n = 1). Indication, era and surgical parameters of the index esophagectomy, stratified per DGCE class, are displayed in Table 1 (patient level, n = 196).

Table 1.

Indication, era and technical parameters of the index esophagectomy, stratified by DGCE class (patient level data, n = 196, classified by first revisional surgery)

	Class 1	Class 2	Class 3	P-value	Total
	(N = 29)	(N = 37)	(N = 130)		(N = 196)
Indication for index esophagectomy
Adenocarcinoma	18 (62.1%)	26 (70.3%)	105 (80.8%)	0.126	149 (76.0%)
Squamous cell carcinoma	7 (24.1%)	6 (16.2%)	19 (14.6%)		32 (16.3%)
Other/benign	4 (13.8%)	5 (13.5%)	6 (4.6%)		15 (7.7%)
Era of index esophagectomy
Before 2010	3 (10.3%)	4 (10.8%)	3 (2.3%)	0.034	10 (5.1%)
2010–2020	11 (37.9%)	23 (62.2%)	70 (53.8%)		104 (53.1%)
2021–2025	15 (51.7%)	10 (27.0%)	57 (43.8%)		82 (41.8%)
Level of anastomosis
Cervical	3 (10.3%)	4 (10.8%)	4 (3.1%)	0.28	11 (5.6%)
Upper mediastinum	24 (82.8%)	29 (78.4%)	115 (88.5%)		168 (85.7%)
Lower mediastinum	2 (6.9%)	4 (10.8%)	11 (8.5%)		17 (8.7%)
Interponate
Tubulized stomach (<5 cm)	18 (62.1%)	24 (64.9%)	103 (79.2%)	0.047	145 (74.0%)
Tubulized stomach (≥ 5 cm)	2 (6.9%)	5 (13.5%)	17 (13.1%)		24 (12.2%)
Whole stomach	3 (10.3%)	5 (13.5%)	2 (1.5%)		10 (5.1%)
Missing/Other	6 (20.7%)	3 (8.1%)	5 (6.1%)		19 (7.2%)
Abdominal access
Open	5 (17.2%)	6 (16.2%)	17 (13.1%)	0.005	28 (14.3%)
Laparoscopic	21 (72.4%)	27 (73.0%)	88 (67.7%)		136 (69.4%)
Robotic	2 (6.9%)	0 (0%)	24 (18.5%)		26 (13.3%)
Missing	1 (3.4%)	4 (10.8%)	1 (0.8%)		6 (3.1%)
Thoracic access
Open	11 (37.9%)	9 (24.3%)	42 (32.3%)	<0.001	62 (31.6%)
Thoracoscopic	14 (48.3%)	20 (54.1%)	43 (33.1%)		77 (39.3%)
Robotic	1 (3.4%)	2 (5.4%)	39 (30.0%)		42 (21.4%)
None/Missing	3 (10.3%)	6 (16.2%)	6 (4.6%)		15 (7.7%)
Preexisting hiatal hernia	1 (3.4%)	5 (13.5%)	35 (26.9%)	0.01	41 (20.9%)
Intraoperative transsection crura	7 (24.1%)	16 (43.2%)	22 (16.9%)	0.004	45 (23.0%)
Prophylactic hiatal closure	4 (13.8%)	4 (10.8%)	15 (11.5%)	0.926	23 (11.7%)
Prophylactic pyloric drainage procedure
None	21 (72.4%)	31 (83.8%)	95 (73.1%)	0.005	147 (75.0%)
Pylorus dilation	1 (3.4%)	1 (2.7%)	28 (21.5%)		30 (15.3%)
Pyloromyotomy	2 (6.9%)	2 (5.4%)	3 (2.3%)		7 (3.6%)
Pyloroplasty	4 (13.8%)	3 (8.1%)	3 (2.3%)		10 (5.1%)
G-POEM	1 (3.4%)	0 (0%)	0 (0%)		1 (0.5%)
Diversion (gastro-jejunostomy)	0 (0%)	0 (0%)	1 (0.8%)		1 (0.5%)

Abbreviation: DGCE: delayed gastric conduit emptying.

Postoperative complications following index esophagectomy included anastomotic leakage (13.8%), pulmonary complications (33.7%), infections (18.4%) and a 19.5% 90-day readmission rate. The median initial hospital stay was 13.0 days (IQR: 10.0–21.8), with a median ICU (intensive care unit) stay of 2 days and a 13.8% escalation-of-care rate.

Preoperative symptoms, diagnostics, and non-interventional therapies

Prior to revisional surgery, the most prevalent symptoms were early satiety/fullness (67.5% moderate/severe), reduced oral intake (64.1%), dysphagia (65.3%), nausea (58.8%), vomiting (53.7%), and regurgitation (54.8%). Conversely, reflux (38.9%), pain (31.3%), aspiration (17.9%), and dumping (3.7%) were less frequent (Fig. 2). Pain was the only symptom significantly associated with a specific class, occurring more often in C3/PCH (38.1% vs. C1: 23.5% and C2: 15.2%; P = 0.017).

Fig. 2.

Bar charts depicting the preoperative distribution of satiety, intake inability, regurgitation, and pain across delayed gastric conduit emptying classes (n = 263, intervention level).

Preoperative workup included CT imaging (87.5%), endoscopy (C1: 94.4%, C2: 87.3%, C3: 62.2%), and contrast radiography (C1: 86.1%, C2: 80.0%, C3: 25.0%), though the latter two were used significantly less in PCH. Functional testing was reported in only 4.6% (gastric emptying scintigraphy n = 4, functional lumen imaging probe (EndoFLIP) n = 3). The most common reported diagnostic finding was paraconduit herniation, especially in C3 (76.2%), followed by conduit redundancy, dilatation and pyloric stenosis (Table 2; intervention level, n = 263). Agreement between preoperative diagnostic findings and the retrospectively assigned class was moderate and incomplete. Class-defining findings were documented in 63.9% of C1 interventions (pyloric stenosis), in 60.0% (conduit dilatation) and 47.3% (conduit redundancy) of C2 interventions, and in 76.2% of C3 interventions (paraconduit herniation). Conversely, class-defining findings were simultaneously recorded in a relevant proportion of interventions assigned to a different class, for example pyloric stenosis in 40.0% of C2 and 32.0% of C3 interventions, and conduit redundancy in 39.0% of C3 interventions. A formal concordance statistic could not be calculated, as preoperative findings were neither mutually exclusive nor uniformly documented across all interventions.

Table 2.

Preoperative diagnostic modalities and findings in relation to the assigned DGCE class (intervention level data, n = 263 interventions in 196 patients)

	Class 1	Class 2	Class 3	P-value	Total
	(N = 36)	(N = 55)	(N = 172)		(N = 263)
Pre-revisional diagnostic modalities
Endoscopy	34 (94.4%)	48 (87.3%)	107 (62.2%)	<0.001	189 (71.9%)
Contrast radiography	31 (86.1%)	44 (80.0%)	43 (25.0%)	<0.001	118 (44.9%)
Computed tomography	31 (86.1%)	50 (90.9%)	149 (86.6%)	0.576	230 (87.5%)
Functional testing	3 (8.3%)	2 (3.6%)	7 (4.1%)	0.305	12 (4.6%)
Pre-revisional diagnostic findings
Pyloric stenosis	23 (63.9%)	22 (40.0%)	55 (32.0%)	0.007	100 (38.0%)
Conduit dilatation	21 (58.3%)	33 (60.0%)	54 (31.4%)	<0.001	108 (41.1%)
Conduit redundancy	16 (44.4%)	26 (47.3%)	67 (39.0%)	0.555	109 (41.4%)
Conduit torsion	9 (25.0%)	22 (40.0%)	15 (8.7%)	<0.001	46 (17.5%)
Conduit stenosis	3 (8.3%)	14 (25.5%)	23 (13.4%)	0.069	40 (15.2%)
Hiatal obstruction	10 (27.8%)	15 (27.3%)	46 (26.7%)	0.993	71 (27.0%)
Paraconduit herniation	5 (13.9%)	10 (18.2%)	131 (76.2%)	<0.001	146 (55.5%)
Intestinal obstruction	0 (0%)	2 (3.6%)	36 (20.9%)	<0.001	38 (14.4%)
Anastomotic stenosis	2 (5.6%)	3 (5.5%)	9 (5.2%)	0.996	14 (5.3%)
Pre-revisional conservative management
Diet counseling	33 (91.7%)	35 (63.6%)	106 (61.6%)	0.005	174 (66.2%)
Prokinetics	31 (86.1%)	45 (81.8%)	66 (38.4%)	<0.001	142 (54.0%)
Proton pump inhibitors	34 (94.4%)	46 (83.6%)	151 (87.8%)	0.755	231 (87.8%)

^* n = 263 interventions in 196 patients* delayed gastric conduit emptying (DGCE) class was assigned retrospectively according to the revisional procedure performed. This table is provided to allow appraisal of the agreement between preoperative diagnostic findings and the assigned class. Findings were not mutually exclusive and were not uniformly documented across all interventions.

Conservative management included dietary counseling (66.2%; highest in C1, P = 0.004), proton pump inhibitors in the majority of patients (87.8%), and prokinetics (54.0%). Prokinetic use was significantly lower in C3 (38.4%) compared with C1 (86.1%) and C2 (81.8%; P < 0.001, Table 2).

Postoperative symptom relief

Sustained symptom relief, assessed per surgical intervention (n = 263), was comparable across all three DGCE classes (C1: 47.2%, C2: 58.2%, C3: 63.9%; P = 0.165). Rates of ‘no change/worsening’ were low, ranging from 2.8% to 9.1% (Fig. 3A). Within C1, pyloroplasty (47.6%) and Roux-en-Y diversion (46.7%) achieved nearly identical resolution rates (P = 0.67). In C2, conduit replacement (75.0%) and remodeling (61.3%) were more effective than simple mobilization or reduction (50.0%; P = 0.035). For C3, hiatal hernia repair yielded numerically higher symptom resolution than division of the right crus (65.8% vs. 47.1%), without reaching statistical significance (P = 0.189). Forty-six patients underwent more than one surgical intervention, most frequently repeat PCH repair (n = 20), while seven patients underwent conduit remodeling (C2) after initial PCH repair/division of the right crus. Median follow-up duration was 525 days after revisional surgery (range: 20–6320 days), with no difference between classes (C1: 553 days, C2: 731 days, C3: 468 days).

Fig. 3.

Symptom relief following surgical interventions (A), preoperative (B) and postoperative (C) endoscopic interventions. Data at intervention level, n = 263 surgical interventions and n = 208 endoscopic procedures. Abbreviation: DGCE = delayed gastric conduit emptying.

Endoscopic procedures before and after revisional surgery

A total of 208 endoscopic procedures for treatment of DGCE were recorded in the database. Of these, 131 interventions in 62 patients (median 1, maximum 8) were performed before surgical revision and uniformly targeted C1/pylorospasm, including 108 pyloric dilatations, 13 botulinum toxin injections, 6 pyloric stent placements, and 4 G-POEMs. As failure of endoscopic management is a natural prerequisite for revisional surgery in patients with a condition potentially addressable by endoscopy (e.g. pylorspasm), endoscopic therapy in this treatment-refractory and selected patient population had by definition not provided durable relief. Overall symptom resolution after endoscopic treatment was reported in 14.3% (highest for balloon dilatation), with most procedures yielding transient (63.8%) or slight (6.7%) improvement only. The distribution among preoperative endoscopic intervention types and corresponding symptom response is displayed in Fig. 3B.

77 endoscopic interventions after surgical revision were reported in 38 patients (median 1, maximum 8). All but 2 interventions (1 pyloric stenting, 1 pyloric dilatation targeting C2) also targeted C1/pylorospasm, including 57 pyloric dilatations, 3 botulinum toxin injections, 13 pyloric stent placements, and 2 G-POEMs. Overall symptom resolution was 21.7% with 14.5% slight improvements and 45% temporary improvements. The distribution among postoperative endoscopic intervention types and corresponding symptom response is displayed in Fig. 3C.

Complications and postoperative hospitalization

The overall complication rate of the 263 interventions was 32.7%, with 17.5% major complications (Clavien-Dindo ≥IIIa) and a 2.3% mortality rate. Major morbidity and mortality were significantly higher in C2 (20.0% and 3.6%) compared with C1 (16.7% and 2.8%) and C3 (13.4% and 1.7%; P = 0.047, Table 3). Primary complications included pulmonary (12.9%), infectious (9.9%), and gastrointestinal (9.5%) events. Gastrointestinal leakage occurred in 2.7%; no conduit necrosis was reported. Deaths (n = 6) were attributed to cardiac (n = 2), infectious/pulmonary (n = 2), leakage (n = 1), and unspecified causes (n = 1).

Table 3.

Complications after revisional surgery, stratified by DGCE class (data are reported at the level of the surgical intervention, n = 263 interventions in 196 patients). Major complications correspond to Clavien-Dindo grade ≥ IIIa

	Class 1	Class 2	Class 3	P-value	Total
	(N = 36)	(N = 55)	(N = 172)		(N = 263)
Time between index esophagectomy and revisional surgery (days)
Median [Q1, Q3]	456 [165, 877]	879 [251, 2000]	375 [155, 794]	<0.001	462 [163, 941]
Overall complication rate
Letal	1 (2.8%)	2 (3.6%)	3 (1.7%)	0.047	6 (2.3%)
Major (Clavien-Dindo grade ≥ IIIa)	6 (16.7%)	11 (20.0%)	23 (13.4%)		40 (15.2%)
Minor (Clavien-Dindo grade < IIIa)	9 (25.0%)	12 (21.8%)	19 (11.0%)		40 (15.2%)
None	20 (55.6%)	30 (54.5%)	127 (73.8%)		177 (67.3%)
Gastrointestinal leakage	2 (5.6%)	2 (3.6%)	3 (1.7%)	0.432	7 (2.7%)
Pulmonary complications	6 (16.7%)	13 (23.6%)	15 (8.7%)	0.011	34 (12.9%)
Chyle leak	2 (5.6%)	1 (1.8%)	0 (0%)	0.023	3 (1.1%)
Gastrointestinal complications	8 (22.2%)	4 (7.3%)	13 (7.6%)	0.026	25 (9.5%)
Cardiac complications	0 (0%)	5 (9.1%)	3 (1.7%)	0.022	8 (3.0%)
Infectious complications	4 (11.1%)	8 (14.5%)	14 (8.1%)	0.373	26 (9.9%)
Wound complications	0 (0%)	0 (0%)	7 (4.1%)	0.148	7 (2.7%)
Other complications	5 (13.9%)	4 (7.3%)	8 (4.7%)	0.111	17 (6.5%)
ICU stay (days)
Median [Q1, Q3]	0 [0, 1]	0 [0, 1]	0 [0, 0]	0.225	0 [0, 1.00]
Escalation in care	1 (2.8%)	4 (7.3%)	12 (7.0%)	0.723	17 (6.5%)
Hospital stay (days)
Median [Q1, Q3]	12 [7.50, 22.0]	11 [6.00, 16.0]	6 [4.00, 13.0]	0.003	8.00 [5.00, 15.0]
Surgery related readmission	2 (5.6%)	5 (9.1%)	14 (8.1%)	0.901	21 (8.0%)

Abbreviation: DGCE, delayed gastric conduit emptying

The median hospital stay was 8 days, ranging from 6 days for PCH repair to 20.5 days for conduit replacement. While ICU admission was not required for most patients, 6.5% required escalation of care (unplanned readmission to intensive or intermediate care unit). The surgery-related readmission rate was 8.0%.

Multivariable analysis

A multivariable model was constructed to identify predictors of successful outcomes following revisional surgery based on results of univariable testing of potential factors influencing postoperative outcomes. The multivariable model comprised 255 interventions with complete covariate data (n = 8 excluded due to missing data in one covariate) and 157 events (successful outcome, defined as ‘Absence of symptoms/a lot better’ on the 5-point Likert scale), with 13 variables included in the multivariable model because they fell below the predefined P-value threshold, corresponding to 12 events per variable. None of the evaluated variables—including patient demographics, index esophagectomy technical aspects, preoperative diagnostic findings, or DGCE class—were independently associated with surgical success (Table 4).

Table 4.

Uni- and multivariate analysis of factors predicting successful surgical treatment of DGCE (defined as ‘absence of symptoms/a lot better’ on 5-point Likert scale)

Variable		Unsuccessful n (%)	Successful n (%)	OR (univariable)	OR (multivariable)
Age (years)	Mean (SD)	58.9 (12.0)	58.7 (12.8)	1.00 (0.98–1.02, P = 0.887)	-
Gender	Female	29 (49.2)	30 (50.8)	-	-
	Male	77 (37.7)	127 (62.3)	1.59 (0.89–2.86, P = 0.117)	1.40 (0.71–2.77, P = 0.329)
BMI	Mean (SD)	24.6 (4.3)	25.6 (4.6)	1.05 (0.99–1.12, P = 0.107)	1.03 (0.97–1.11, P = 0.315)
Indication for esophagectomy	Adenocarcinoma	82 (41.2)	117 (58.8)	-	-
	Squamous cell carcinoma	14 (31.8)	30 (68.2)	1.50 (0.76–3.08, P = 0.251)	-
	Other/benign	10 (50.0)	10 (50.0)	0.70 (0.28–1.78, P = 0.449)	-
Abdominal accesss	Open	21 (50.0)	21 (50.0)	-	-
	Laparoscopic	77 (41.8)	107 (58.2)	1.39 (0.71–2.73, P = 0.337)	1.58 (0.59–4.27, P = 0.365)
	Robotic	7 (22.6)	24 (77.4)	3.43 (1.26–10.21, P = 0.020)	2.45 (0.51–11.95, P = 0.261)
	Missing	1 (16.7)	5 (83.3)	5.00 (0.72–100.19, P = 0.157)	4.20 (0.30–107.07, P = 0.295)
Thoracic access	Open	34 (42.5)	46 (57.5)	-	-
	Thoracoscopic	51 (44.3)	64 (55.7)	0.93 (0.52–1.65, P = 0.798)	0.72 (0.35–1.47, P = 0.374)
	Robotic	13 (26.0)	37 (74.0)	2.10 (0.99–4.66, P = 0.059)	1.01 (0.32–3.34, P = 0.981)
	None/Missing	8 (44.4)	10 (55.6)	0.92 (0.33–2.65, P = 0.880)	0.69 (0.16–2.78, P = 0.601)
Year of esophagectomy	Before 2020	63 (48.1)	68 (51.9)	-	-
	2020 and after	43 (32.6)	89 (67.4)	1.92 (1.17–3.18, P = 0.011)	1.81 (0.88–3.73, P = 0.106)
Year of revisional surgery	Before 2022	55 (46.2)	64 (53.8)	-	-
	2022 and after	51 (35.4)	93 (64.6)	1.57 (0.95–2.58, P = 0.076)	0.99 (0.50–1.94, P = 0.977)
Pre- or intraoperativ pylorus drainage procedure	Yes	28 (40.6)	41 (59.4)	0.98 (0.56–1.73, P = 0.957)	-
Preexisting hiatus hernia	Yes	26 (41.3)	37 (58.7)	0.95 (0.54–1.70, P = 0.858)	-
Intraoperative transection of crura	Yes	31 (50.0)	31 (50.0)	0.60 (0.33–1.06, P = 0.077)	1.03 (0.51–2.10, P = 0.937)
Prophylactic hiatal closure	Yes	7 (20.6)	27 (79.4)	2.94 (1.29–7.57, P = 0.015)	2.44 (0.97–6.80, P = 0.069)
Antro-pyloric dysmotility/pyloric stenosis	Yes	45 (45.0)	55 (55.0)	0.73 (0.44–1.21, P = 0.225)	-
Conduit dilatation	Yes	51 (47.2)	57 (52.8)	0.61 (0.37–1.01, P = 0.057)	0.82 (0.39–1.70, P = 0.587)
Conduit redundancy	Yes	50 (45.9)	59 (54.1)	0.67 (0.41–1.11, P = 0.122)	0.94 (0.48–1.85, P = 0.858)
Conduit torsion	Yes	24 (52.2)	22 (47.8)	0.56 (0.29–1.06, P = 0.073)	1.00 (0.44–2.29, P = 0.998)
Conduit stenosis	Yes	21 (52.5)	19 (47.5)	0.56 (0.28–1.10, P = 0.090)	0.58 (0.26–1.26, P = 0.167)
Hiatus obstruction	Yes	29 (40.8)	42 (59.2)	0.97 (0.56–1.70, P = 0.913)	-
Paraconduit herniation	Yes	56 (38.4)	90 (61.6)	1.20 (0.73–1.97, P = 0.472)	-
Endoscopic pretreatment	Yes	37 (41.6)	52 (58.4)	0.92 (0.55–1.56, P = 0.764)	-
DGCE classification	Class 1	19 (52.8)	17 (47.2)	-	-
	Class 2	23 (41.8)	32 (58.2)	1.55 (0.67–3.66, P = 0.306)	1.86 (0.73–4.82, P = 0.192)
	Class 3	64 (37.2)	108 (62.8)	1.89 (0.91–3.92, P = 0.086)	1.38 (0.60–3.18, P = 0.447)

Analyis on intervention level, n = 263. Center and patient ID included as random factors to account for both between-center heterogeneity and within-patient clustering of repeated interventions.

Abbreviation: DGCE, delayed gastric conduit emptying.

Number in data frame = 263, Number in model = 255, Missing = 8, AIC = 356.4, C-statistic = 0.687, H&L = Chi-sq(8) 10.15 (P = 0.255).

DISCUSSION

The RESTORANT project represents the largest multicentric, international analysis to date focusing specifically on outcomes after surgical revision for DGCE refractory to medical or interventional treatment. Our principal finding is that surgical treatment results in clinically meaningful symptom improvement in a majority of patients. Approximately half of patients achieved complete or near-complete symptom resolution, while only a small minority (2–9%) experienced no benefit. The observed outcomes are consistent with earlier small monocentric series^14,18–20 and underscore the effectiveness of surgical approaches for this challenging functional syndrome in carefully selected patients.

Symptom relief was broadly comparable across all three DGCE classes, which is compatible with, but does not confirm, the utility of a mechanism-informed approach. DGCE is not a uniform entity but a constellation of distinct pathophysiological processes including antro-pyloric dysmotility, conduit-related abnormalities, and hiatal-related mechanical obstruction. However, it is essential to recognize that these DGCE classes frequently overlap. A patient may present with both conduit redundancy (C2) and hiatal narrowing (C3), making it difficult to isolate the primary driver of stasis. This overlap likely explains why outcomes are not perfect and a substantial subset of patients required multiple surgical interventions; if surgical therapy addresses only one specific pathologic mechanism while another remains untreated, symptom resolution may be incomplete. For the current analysis, classification was based on the dominant mechanism inferred from the surgical procedure.

However, the clinical trajectories and anatomical findings at revision differed substantially between classes, suggesting that the classification may provide an exploratory framework for structuring diagnostic workup. In this study, however, patients were grouped retrospectively based on the surgical procedure performed. This design carries an inherent risk of circularity, as the pathophysiological mechanism was often inferred from the chosen surgical procedure rather than independently established through preoperative diagnostics. Consistent with this, the agreement between preoperative findings and the assigned class was moderate at best, and a formal concordance statistic was not computable because findings were neither mutually exclusive nor uniformly documented. This highlights a central challenge in DGCE management: the difficulty of precise preoperative phenotyping in the absence of standardized functional data. Consequently, while our ‘mechanism-informed,’ so far unvalidated, classification may serve as a conceptual framework for identifying surgical targets, the specific technical approach remains a critical determinant of the clinical outcome.

Class-specific outcomes offered further nuance, particularly regarding the timing of intervention. For C1, success rates for pyloroplasty and Roux-en-Y diversion were consistent with historical data and suggest that once the pylorus is addressed, the long-term prognosis is favorable. In contrast, C2 interventions occurred significantly later. Besides technical and geometric issues of the index esophagectomy (width, rotation, and angulation of the conduit) as an underlying cause, this temporal pattern might reflect a ‘material fatigue’ mechanism, where the denervated and amotile gastric conduit undergoes degenerative remodeling and redundancy over years of exposure to chronic, raised intraluminal pressure, although this remains speculative. Our data show that corrective procedures for conduit geometry—specifically remodeling or replacement—were significantly more effective than simple mobilization. This underscores that established structural pathology in C2 DGCE requires anatomical correction rather than mere positional adjustment.

Our analysis of C3 DGCE highlights the importance of distinguishing between different hiatal pathologies. While mechanical obstruction at the diaphragm was the most frequent indication for surgery, the underlying mechanism dictated the choice of procedure. Hiatal repair (hiatoplasty/crurorrhaphy) was used for PCH, whereas crural division was reserved for cases of a narrow hiatus causing conduit outlet stenosis. The observation that crural division was associated with poorer sustained resolution (47.1%) than hiatoplasty for PCH (65.8%) suggests that hiatal stenosis may be a more difficult surgical target or that its contribution to the obstructive phenotype is less easily resolved by simple release compared with the definitive anatomical restoration required for PCH.

Another key observation is the low rate of durable symptom resolution following endoscopic therapies in this cohort, with complete relief reported in fewer than 20% of cases. Because the RESTORANT population consists exclusively of patients who required surgical escalation, however, these results are effectively a self-fulfilling prophecy; by definition, endoscopic management had already failed in these individuals. These endoscopic outcomes are therefore reported descriptively within a surgically selected, refractory cohort and permit no inference of superiority of surgical over endoscopic management. The results do not suggest that endoscopic therapy is ineffective in the broader post-esophagectomy DGCE population—where pneumatic dilatation is highly effective with reported success rates of >90% and remains the gold standard for first-line treatment^21–23—but rather that persistent symptoms despite repeated endoscopic intervention identify a subgroup for whom definitive surgical correction is required. These proportions should also be viewed against the population-level burden of DGCE. A recent systematic review and meta-analysis of 125 studies reported a pooled incidence of 15.9% (95% CI 11–21%) for early and 9.4% (95% CI 7.1–11.9%) for late DGCE after esophagectomy, with heterogeneity driven largely by inconsistent diagnostic criteria,²⁴ while a companion systematic review of procedural treatments identified endoscopic balloon dilatation and botulinum toxin injection as preferred first-line interventions with reported success rates of 50–100%.²⁵ Against this background, the observed rate of 1.66% of all esophagectomies proceeding to revisional surgery indicates that operative correction is ultimately required in only a small minority of affected patients, and that the present cohort represents the refractory tail of the DGCE spectrum rather than the condition as a whole. A complete denominator of all patients developing DGCE within the 11,807 index esophagectomies could not be reconstructed retrospectively, as non-operatively managed patients are frequently treated in peripheral hospitals or by gastroenterologists, and diagnostic thresholds were not standardized across centers and eras.

Despite encouraging functional outcomes, revisional surgery for DGCE remains technically demanding and carries substantial morbidity (32.7%) and mortality (2.3%) rates. These figures are comparable to other complex revisional esophageal procedures, such as surgery for failed anti-reflux repairs or reconstructions following caustic injury. The observed complication profile mandates that revisional surgery be considered only after structured diagnostic workup and after non-operative/endoscopic strategies are exhausted, with shared decision-making that explicitly addresses expected benefit and perioperative risk. It furthermore strongly supports the recommendation that surgical treatment of refractory DGCE should be centralized to high-volume expert centers.

While the multicenter design, involving 26 centers, helps mitigate center-specific selection bias, several limitations must be acknowledged. First, the infrequent use of objective functional testing, such as EndoFLIP or scintigraphy (<5%), before surgery precluded a standardized diagnostic workup and limited our ability to correlate anatomical repair with physiological improvement. Second, the retrospective assessment of symptoms via medical records introduces inherent subjectivity; baseline severity was reconstructed rather than prospectively graded, and no validated patient-reported outcome measure was applied. Third, DGCE class was derived from the procedure performed rather than from standardized preoperative phenotyping, so that the classification must be regarded as hypothesis-generating until tested prospectively. Fourth, 46 patients contributed more than one intervention; although within-patient clustering was addressed by clustered descriptive tests and patient-level random effects, residual dependence between observations cannot be excluded. Furthermore, our multivariable analysis identified no independent predictors of success. Given the limited number of events per variable, the cohort’s high heterogeneity and within-patient multiplicity, the model had restricted statistical power for multivariable discovery, and the absence of statistically significant associations should not be interpreted as evidence that clinically relevant predictors do not exist. However, it may also underscore that both the indication for revisional surgery in DGCE and the expected treatment response remain highly individualized. Future prospective audits using standardized diagnostic criteria such as emptying times on contrast imaging²⁶ and isotope emptying studies,²⁷ as well as validated patient-reported outcome measures and symptom response questionnaires,²⁸ are essential to move beyond subjective Likert scales. In conclusion, RESTORANT demonstrates that revisional surgery is a high-risk but frequently effective therapeutic option in carefully selected patients with refractory DGCE, and provides a descriptive, hypothesis-generating framework that may help to structure future prospective evaluation of mechanism-based treatment allocation.

Specific author contributions: Marcel André Schneider (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Validation, Visualization, Writing—original draft, Writing—review & editing), Andrew Richard Davies (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing—review & editing), Catherine O’ Brien (Data curation, Writing—review & editing), Cara Baker (Data curation, Writing—review & editing), James Gossage (Data curation, Writing—review & editing), Jessie Elliott (Conceptualization, Data curation, Formal analysis, Methodology, Writing—review & editing), Nicola Raftery (Data curation, Writing—review & editing), Rakesh Ahmed (Data curation, Writing—review & editing), Fredrik Klevebro (Conceptualization, Data curation, Methodology, Writing—review & editing), Magnus Nilsson (Data curation, Writing—review & editing), Nora Norén (Data curation, Writing—review & editing), Olga Meier (Adamenko) (Data curation, Writing—review & editing), Ewen Griffiths (Data curation, Writing—review & editing), William Butterworth (Data curation, Writing—review & editing), Hans A Schlößer (Data curation, Writing—review & editing), Christiane Bruns (Data curation, Writing—review & editing), Eleni Tzitzili (Data curation, Writing—review & editing), Felix Nickel (Conceptualization, Data curation, Methodology, Writing—review & editing), Marius Kemper (Conceptualization, Data curation, Writing—review & editing), Petro Zgurskyi (Data curation, Writing—review & editing), Misha DP Luyer (Data curation, Writing—review & editing), Tessa Geraedts (Data curation, Writing—review & editing), Luigi Bonavina (Data curation, Writing—review & editing), Alberto Aiolfi (Data curation, Writing—review & editing), Ines Gockel (Data curation, Writing—review & editing), René Thieme (Data curation, Writing—review & editing), Stefan Niebisch (Data curation, Writing—review & editing), Marko Kraljevic (Data curation, Writing—review & editing), Beat Müller-Stich (Data curation, Writing—review & editing), Lucien Cron (Data curation, Writing—review & editing), Felix Berlth (Data curation, Writing—review & editing), Alfio Milazzo (Data curation, Writing—review & editing), Bruno Sgromo (Data curation, Writing—review & editing), Holly Harris (Data curation, Writing—review & editing), Lucia Moletta (Data curation, Writing—review & editing), Michele Valmasoni (Data curation, Writing—review & editing), Michal Hubka (Data curation, Writing—review & editing), Simone Giacopuzzi (Conceptualization, Data curation, Writing—review & editing), Alice Collizzolli (Data curation, Writing—review & editing), Riccardo Rosati (Data curation, Writing—review & editing), Agnese Carresi (Data curation, Writing—review & editing), Silvia Battaglia (Data curation, Writing—review & editing), Stefano Santi (Data curation, Writing—review & editing), Monica Gualtierotti (Data curation, Writing—review & editing), Fabrizio Rebecchi (Data curation, Writing—review & editing), Lorenzo Viggiani D’Avalos (Data curation, Writing—review & editing), Manuel Pera (Data curation, Writing—review & editing), Enrique Biel (Data curation, Writing—review & editing), Arnulf Hoelscher (Data curation, Writing—review & editing), Martin Hemmerich (Data curation, Writing—review & editing), Edward Cheong (Conceptualization, Data curation, Writing—review & editing), Elke Van Daele (Data curation, Writing—review & editing), Hanne vanommeslaeghe (Data curation, Writing—review & editing), Giulia Nezi (Data curation, Writing—review & editing), Bastiaan R Klarenbeek (Data curation, Writing—review & editing), Peter P Grimminger (Data curation, Writing—review & editing), Christian Gutschow (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Supervision, Writing—original draft, Writing—review & editing).