Predictive model and prognostic insights into duodenal stump fistula following radical gastrectomy with Roux-en-Y reconstruction for gastric cancer

Abstract

BACKGROUND

Duodenal stump fistula (DSF) is a rare yet serious complication following gastric cancer surgery. The risk factors associated with DSF, as well as the predictive models, remain insufficiently elucidated.

AIM

To identify DSF risk factors following radical gastrectomy with Roux-en-Y anastomosis, develop a predictive model, and evaluate impact on prognosis.

METHODS

This retrospective cohort study was conducted on patients undergoing radical gastrectomy with Roux-en-Y anastomosis for gastric cancer at Juntendo University from 2015 to 2021 (n = 325). Univariate and multivariate analyses were performed to identify the risk factors associated with DSF. Based on the independent risk factors, a predictive nomogram was developed and subsequently evaluated using receiver operating characteristic curve analysis. Kaplan-Meier survival curves were utilized to assess the impact of DSF on overall survival (OS), cancer-specific survival (CSS), and disease-free survival (DFS).

RESULTS

Among the 325 patients analyzed, DSF was observed in 7 (2.2%) cases. No DSF was observed in 110 patients where the duodenal stump suturing fixation technique to the jejunal wall was used. Multivariate analysis confirmed that age [odds ratio (OR) = 1.17, P = 0.015] and obstructive ventilatory failure (OVF) (OR = 14.03, P = 0.001) were independent risk factors for DSF. The predictive nomogram was constructed based on age and OVF, which exhibited strong performance (area under the curve = 0.90, 95% confidence interval: 0.82-0.99). Kaplan-Meier analysis revealed a statistically significant reduction in CSS for patients with DSF, whereas no significant differences were observed in OS or DFS.

CONCLUSION

Age and OVF are independent risk factors for DSF, which worsens CSS. A nomogram predicts DSF accurately, and innovative surgical techniques may reduce its occurrence.

Key Words: Gastric cancer; Duodenal stump fistula; Roux-en-Y reconstruction; Predictive nomogram; Cancer survival

Core Tip: This retrospective cohort study included 325 patients who underwent radical gastrectomy with Roux-en-Y reconstruction to identify risk factors for duodenal stump fistula (DSF). Age and obstructive ventilatory failure were identified as independent risk factors for DSF. A nomogram incorporating these risk factors demonstrated robust predictive performance for DSF. DSF significantly affected cancer-specific survival. Suturing fixation of the duodenal stump to the jejunal wall has shown potential as an effective strategy for preventing DSF.

INTRODUCTION

Gastric cancer ranks fifth in incidence and fourth in mortality globally[1]. Radical resection accompanied by lymphadenectomy is the standard and well-established surgical methodology[2]. However, postoperative complications, particularly duodenal stump fistula (DSF), pose a significant challenge. DSF predominantly arises following non-Billroth I gastrojejunostomy and presents a low incidence rate, ranging from 0.7% to 7.7%[3-6]. Despite its relative rarity, DSF is associated with high lethality, exhibiting a mortality rate between 7.8% and 28.0%[3-7]. This high mortality can be primarily ascribed to severe complications linked to DSF, such as sepsis, intra-abdominal abscesses, and pneumonia[8]. These complications consequently result in significantly extended hospital stays[3].

As a result, the grave consequences and lethal potential of DSF have thrust it into the forefront of clinical concern, underscoring the critical necessity for prompt and decisive action. Especially refining prevention strategies to avoid DSF is of paramount importance and priority. Several studies have pinpointed predictive factors for DSF. These factors encompass a range of elements. These include baseline patient characteristics, such as age and comorbidities[7,9]; preoperative nutritional status, including hypoalbuminemia and anemia[5,10]; tumor-related factors like size and stage[9,11]; and operative parameters such as contamination and blood loss[5,7,12]. In addition, preventive practices related to the actual surgical techniques include modified purse-string, patches, and duodenal decompression probes[13-15].

Nonetheless, existing studies for DSF following gastrectomy exhibit several limitations. These include the absence of a reliable DSF prediction model, inadequate reporting of intraoperative stump management, and an overreliance on single surgical approaches. Such challenges hinder data heterogeneity and comparability. To address these shortcomings, this retrospective cohort study investigated 325 cases of radical gastrectomy with Roux-en-Y reconstruction for gastric cancer. All cases were collected over 7 years at a single institution. With the focus on comprehensively identifying DSF risk factors, establishing a predictive nomogram based on multimodal analyses and its impact on prognostic outcomes.

MATERIALS AND METHODS

Study design and patients

Between January 2015 and December 2021, 842 patients who underwent gastrectomy for gastric cancer at the Department of Esophageal and Gastroenterological Surgery at Juntendo University were eligible for this study. Patients who underwent non-Roux-en-Y reconstruction or were clinically diagnosed with synchronous distant metastases were excluded. We also excluded patients with incomplete or missing comprehensive medical records, surgical videos, pathological results, or follow-up data, as well as those who received preoperative chemotherapy or radiotherapy. Ultimately, 325 patients who underwent gastrectomy with Roux-en-Y reconstruction were included in this analysis (Figure 1). This study was approved by the Institutional Ethics Committee at our institute (No. H20-0396).

Figure 1 Flowchart of patient selection.

Surgical procedure

Surgical treatments were performed in accordance with the sixth edition of the Japanese Gastric Cancer Treatment Guidelines[2]. All patients underwent radical surgery, comprising gastrectomy with lymphadenectomy. Radical gastrectomy includes distal gastrectomy, defined as the removal of at least two-thirds of the stomach when an adequate proximal margin is achievable. This margin is > 3 cm for Borrmann types I and II, and > 5 cm for types III and IV. Total gastrectomy is performed when such a margin cannot be secured. The extent of lymphadenectomy was determined by tumor location and stage, in accordance with treatment guidelines. After complete lymph node dissection, the esophagus, stomach, and duodenum were transected with an endoscopic linear stapler. All patients underwent Roux-en-Y reconstruction, involving gastrojejunostomy or esophagojejunostomy. This was followed by a jejunojejunostomy approximately 40 cm distal to the primary anastomosis. All duodenal stumps were embedded using purse-string sutures with absorbable thread, extending to the depth of the full thickness or the seromuscular layer. In recent surgical cases (n = 110), the surgeon has judged and selected this approach after free mobilization and complete embedding of the duodenal stump. The residual stump and the jejunal sidewall (Roux limb) are suture-fixed to ensure that the reconstructed intestinal tract is not twisted or excessively strained (Figure 2). One or two drainage tubes were routinely positioned intraoperatively in the abdominal cavity.

Figure 2 Intraoperative images of the duodenal stump and suturing fixation of the stump to the jejunal sidewall. A: Initiation of the suturing process to embed the remnant stump with barbed suture; B: Complete embedding of the remnant stump following continuous suturing; C: Suturing fixation of the remnant stump with the jejunal lateral wall (Roux limb); D: Tightening of the sutures to finalize the procedure. Confirm that the entire Roux leg fixed to the duodenum from the esophagus or stomach and jejunal anastomosis is not under excessive tension and that no abnormalities are present in the subsequent placement of the intestinal tract in the Y leg.

Data collection

Comprehensive clinical and pathological data were systematically collected for eligible patients. Clinicopathological characteristics were collected for age, gender, body mass index (BMI), American Society of Anesthesiologists (ASA) score, comorbidities, and preoperative tumor markers. Other variables included tumor location, tumor diameter, pathological findings, non-anastomotic complications (≥ grade 3a), and anastomotic complications (≥ grade 3a). ASA score was assessed according to the ASA classification system[16]. The diagnosis of obstructive ventilatory failure (OVF) in this study was defined as a forced expiratory volume in one second (FEV1)/forced vital capacity ratio < 0.70 on preoperative spirometry testing, referring to the Global Initiative for Obstructive Lung Disease criteria[17]. Tumor locations followed the Japanese Gastric Cancer Association (JGCA) classification system[2]. Pathological types were classified according to the World Health Organization classification of tumors[18]. Pathological tumor node metastasis staging followed the guidelines of the eighth edition of the American Joint Committee on Cancer (AJCC) staging manual[19]. Intraoperative stump management was confirmed by thoroughly reviewing surgical records and videos. Postoperative complications were evaluated according to the Clavien-Dindo classification system[20]. Perioperative outcomes included surgical approach, conversion to open surgery, gastrectomy type, lymphadenectomy extent, combined resection, surgical time, blood loss, and retrieved lymph nodes. Additional factors included postoperative hospital stay, reinforcement of the duodenal stump, embedding degree, muscularis and serosa injury, bleeding during reinforcement, reinforcement duration, and whether the stump was sutured to the intestinal wall. Reinforcement of the duodenal stump involves continuous suturing and embedding using absorbable sutures. The embedding degree is defined as the proportion of the original stump closure line covered after completing the embedding procedure. Muscularis and serosa injury refers to the count of mucosal injuries caused by improper handling during the suturing and embedding process. Bleeding in reinforcement refers to the estimated blood loss observed during the stump embedding phase based on surgical video recordings. Reinforcement duration is measured as the time, in minutes, required for duodenal stump suturing and embedding, starting from the initial needle insertion to the suture cutting. The diagnostic criteria for DSF included an amylase level in the drainage fluid exceeding three times the standard value postoperatively. Alternatively, DSF was defined by the presence of peritoneal fluid leakage through the abdominal wall accompanied by severe upper abdominal pain. A definitive diagnosis required correlation with findings from upper gastrointestinal angiography or abdominal exploration[21]. All patients underwent standardized postoperative follow-up. Assessments were conducted every 3-6 months during the first 2 years after surgery, and every 6-12 months thereafter. Follow-up was continued until death or at least 3 years after surgery, whichever occurred first.

Outcomes and definitions

The primary outcome of this study was the incidence of DSF. The secondary outcome was to identify risk factors for DSF. We also aimed to investigate the survival parameters outcomes: Overall survival (OS), cancer-specific survival (CSS), and progression-free survival (PFS), all defined according to AJCC guidelines[19]. OS was defined as the time from surgery to death from any cause or the last follow-up. CSS was defined as the time from surgery to death, specifically due to gastric cancer, with deaths from other causes censored at the time of occurrence. PFS was defined as the time from surgery to the first event of disease progression, recurrence, or death from any cause. Censored data arose from patients lost to follow-up or alive at the end of the study period. These were handled using standard survival analysis methods, including Kaplan-Meier estimation for survival curves and Cox proportional hazards regression for assessing risk factors.

Statistical analyses

The baseline characteristics and treatment details were compared between patients with and without DSF. Normally distributed continuous variables were presented as mean ± SD and analyzed using an independent t-test or one-way analysis of variance. Non-normally distributed continuous variables were reported as median [interquartile range (IQR)] and analyzed using the Mann-Whitney U test. Categorical and ordinal variables were summarized as frequencies and percentages [n (%)] and analyzed using the χ² test or Fisher’s exact test when expected cell frequencies were less than 5. All analyses reported the corresponding P values. Statistical significance was set at a two-sided P value of < 0.05. Statistically significant variables in univariate analysis (P < 0.05) were included in the multivariate analysis. To reduce confounding from treatment-related factors, the variable “stump sutured to the intestinal wall” was also included, based on its theoretical and practical relevance. Collinearity diagnostics were performed before multivariate analysis to ensure no significant collinearity existed among the variables. Multivariate logistic regression analysis was performed to identify independent risk factors, with results presented as odds ratios (ORs) alongside 95% confidence intervals (CIs) and P values. To address statistical randomness and bias due to the small sample size in the analysis of DSF risk factors, four additional statistical methods were applied. These methods were specifically designed for rare events: Firth penalized logistic regression[22], Bayesian logistic regression[23], rare events logistic regression[24], and Poisson regression[25]. Significant variables identified in the multivariate and rare events analyses were incorporated into a nomogram for DSF risk prediction. The predictive performance of the nomogram was assessed using receiver operating characteristic curves and calibration curves. The latter were used to evaluate the agreement between predicted and observed outcomes. Decision curve analysis (DCA) was performed to quantify net clinical benefit across a range of risk thresholds. To minimize the potential overestimation of covariate effects due to unaccounted competing events, survival outcomes were analyzed using the Fine-Gray competing risks model. This model is a modified Kaplan-Meier method that accounts for competing risks. Outcomes were compared between the DSF and none-DSF (nDSF) groups using the log-rank test. All statistical analyses were conducted using R software (version 4.3.3; R Foundation for Statistical Computing) and SPSS software (version 27.0; IBM Corp., Armonk, NY, United States).

RESULTS

Clinicopathological characteristics

Table 1 shows the results of the clinicopathological characteristics of the patients. The incidence of postoperative DSF among the 325 patients was 2.2% (7 cases), and nDSF was 97.8% (318 cases). The DSF group had a significantly higher median age compared to the nDSF group [82.0 years (IQR: 79.5-85.0) vs 72.0 years (IQR: 63.2-78.0), P = 0.001]. No significant differences were observed in sex, BMI, and ASA score. In comorbidities, the prevalence of OVF in the DSF group was significantly higher than that in the nDSF group (57.1% vs 5.7%, P < 0.001). There was no difference in preoperative tumor markers between the two groups. The middle stomach was the most frequent tumor location in both groups, with no significant difference between the DSF and nDSF groups (57.1% vs 42.1%, P = 0.839). The DSF group had a significantly higher proportion of patients with pN2-3 stage disease than the nDSF group (28.6% vs 12.6%, P = 0.022), but there was no significant difference in pstage (P = 0.612). The incidence of grade 3a or higher non-anastomotic complications showed a significant difference between the DSF and nDSF groups (71.4% vs 16.4%, P < 0.001). In contrast, there was no significant difference in grade 3a or higher anastomotic complications other than DSF (0% vs 3.5%, P = 0.993).

Table 1 Clinicopathological characteristics, mean ± SD/n (%).

Variables	Categories	nDSF group (n = 318)	DSF group (n = 7)	P value
Age		72.0 (63.2, 78.0)	82.0 (79.5, 85.0)	0.001a
Gender	Female	102 (32.1)	2 (28.6)	1
	Male	216 (67.9)	5 (71.4)
BMI		22.3 ± 3.7	22.3 ± 3.7	0.620
ASA score	1	71 (22.3)	1 (14.3)	0.757
	2	222 (69.8)	5 (71.4)
	3	25 (7.9)	1 (14.3)
Comorbidity	OVF	18 (5.7)	4 (57.1)	< 0.001b
	Heart failure or angina	7 (2.2)	0 (0.0)	1
	Bowel obstruction	3 (0.9)	0 (0.0)	1
	Digestive ulcer history	5 (1.6)	0 (0.0)	1
	Abdominal surgery history	44 (13.8)	0 (0.0)	0.617
Tumor marker	CEA	5.7 ± 21.2	5.3 ± 20.1	0.170
	CA-199	44.5 ± 228.9	31.0 ± 85.0	0.737
	AFP	15.4 ± 102.3	9.8 ± 77.3	0.758
Tumor location	Upper stomach	97 (30.5)	1 (14.3)	0.839
	Lower stomach	71 (22.3)	2 (28.6)
	Middle stomach	134 (42.1)	4 (57.1)
	Gastroesophageal junction	13 (4.1)	0 (0.0)
	All or multiple	3 (0.9)	0 (0.0)
Tumor diameter (mm)		63.6 ± 43.0	60.8 ± 41.6	0.291
Pathological type	Adenocarcinoma	197 (61.3)	5 (71.4)	0.993
	Undifferentiated carcinoma	114 (35.8)	2 (28.6)
	Neuroendocrine tumor	4 (1.3)	0 (0.0)
	Lymphoma and others	4 (1.3)	0 (0.0)
	Special types	1 (0.3)	0 (0.0)
pT	pT1-2	134 (42.1)	3 (42.9)	0.602
	pT3-4	184 (57.8)	4 (57.1)
pN	pN0-1	278 (87.4)	5 (71.4)	0.022a
	pN2-3	40 (12.6)	2 (28.6)
Distant metastasis		43 (13.5)	1 (14.3)	1
pStage	I-II	188 (59.1)	3 (42.9)	0.612
	III-IV	130 (40.9)	4 (47.2)
Non-anastomotic complication (≥ Grade 3a)		52 (16.4)	5 (71.4)	< 0.001b
Anastomotic complications except DSF (≥ Grade 3a)		11 (3.5)	0 (0.0)	0.993

^aP < 0.05.

^bP < 0.001.

AFP: Alpha-fetoprotein; ASA: American Society of Anesthesiologists; BMI: Body mass index; CA-199: Carbohydrate antigen 199; CEA: Carcinoembryonic antigen; OVF: Obstructive ventilatory failure; pN: Pathological nodal stage; pT: Pathological tumor stage; DSF: Duodenal stump fistula; nDSF: Non-duodenal stump fistula.

Perioperative outcome

Table 2 shows the perioperative outcomes. There was no difference between the two groups in surgical approach (P = 0.210), conversion to open surgery, type of gastrectomy (P = 0.621), lymphadenectomy extent (P = 0.800), and combined resection. There were no significant differences in surgical time (P = 0.369) and blood loss (P = 0.659). The DSF group experienced significantly longer postoperative hospital stays compared to the nDSF group (38.7 ± 11.7 days vs 17.5 ± 20.2 days, P < 0.001). The proportion of patients undergoing reinforcement of the duodenal stump showed no significant difference between the DSF and nDSF groups (100% vs 91.5%, P = 0.910). Similarly, associated parameters-including embedding degree, muscularis and serosa injury, bleeding during reinforcement, and reinforcement duration did not differ significantly. Among 110 patients who underwent duodenal stump suturing fixation to the jejunal side wall after embedding, none developed DSF. The information on the seven patients who had DSF is presented in Supplementary Table 1.

Table 2 Perioperative outcomes, mean ± SD/n (%).

Variables	Categories	nDSF group (n = 318)	DSF group (n = 7)	P value
Surgical approach	Laparotomy	129 (40.6)	1 (14.3)	0.210
	Laparoscopy	146 (45.9)	6 (85.7)
	Robot-assisted	43 (13.5)	0 (0.0)
Conversion to open surgery		11 (3.5)	0 (0.0)	1.000
Gastrectomy	DG	129 (40.6)	4 (57.1)	0.621
	TG	189 (59.4)	3 (42.9)
Lymphadenectomy extent	D1, D1 +	146 (45.9)	3 (42.9)	0.800
	D2, D2 +	170 (54.0)	4 (57.1)
Combined resection		48 (15.1)	1 (14.3)	1.000
Surgical time (minute)		324.8 ± 100.6	348.3 ± 62.8	0.369
Blood loss (mL)		72.7 ± 102.9	77.1 ± 90.5	0.659
Retrieved lymph nodes		39.1 ± 17.6	44.9 ± 17.6	0.288
Postoperative hospitalization (days)		17.5 ± 20.2	38.7 ± 11.7	< 0.001b
Reinforcement of duodenal stump		291 (91.5)	7 (100.0)	0.910
Embedding degree (%)		94.2 ± 13.9	90.7 ± 8.3	0.820
Muscularis and serosa injury		36 (11.32)	2 (28.6)	0.418
Bleeding in reinforcement (mL)		1.5 ± 3.5	1.7 ± 1.9	0.259
Reinforcement duration (minute)		9.1 ± 5.2	9.3 ± 3.2	0.659
Stump suture fixed to the intestinal wall1		110 (34.6)	0 (0.0)	0.131

^bP < 0.001.

¹In select cases (n = 110), the stump was fully mobilized after embedding and then sutured to the jejunal side wall.

DG: Distal gastrectomy; TG: Total gastrectomy; DSF: Duodenal stump fistula; nDSF: Non-duodenal stump fistula.

Results of univariate and multivariate logistic regression analyses for DSF risk factors

The results of univariate and multivariate analyses of risk factors for DSF are shown in Tables 3 and 4, respectively. Univariate logistic regression analysis demonstrated significant associations between DSF and age (P < 0.001), OVF (P < 0.001), carbohydrate antigen 19-9 (P = 0.033), alpha-fetoprotein (P = 0.042), and blood loss (P = 0.020). Variance inflation factors (VIFs) were calculated for the significant variables identified in the univariate analysis. All VIF values ranged from 1.000 to 1.413, indicating no significant multicollinearity (VIF < 5), ensuring the variables’ suitability for multivariate analysis. Multivariate logistic regression analysis confirmed that age (OR = 1.17, 95%CI: 1.05-1.34, P = 0.015) and OVF (OR = 14.03, 95%CI: 3.74-27.62, P = 0.001) were identified as independent risk factors for DSF. Several rare events regression analyses further confirmed the independent risk factors identified in the multivariate analysis (Tables 5, 6, 7 and 8). All these analyses validated age and OVF as independent risk factors for DSF.

Table 3 Univariate logistic regression analysis of the duodenal stump fistula risk factors.

Variables (clinicopathological factors)	P value	Variables (perioperative factors)	P value
Age	< 0.001b	Non-anastomotic complications1	0.073
BMI	0.945	Anastomotic complications2	1
Gender	0.310	Surgical approach	0.190
ASA score	0.757	Laparoscopic surgery	0.084
OVF	< 0.001b	Conversion to open surgery	1
Heart failure or angina	1	Gastrectomy	0.622
History of bowel obstruction	1	Lymphadenectomy extent	1
History of digestive ulcer	1	Combined resection	1
History of abdominal surgery	0.777	Surgical time	0.350
CEA	0.494	Blood loss	0.020a
CA-199	0.033a	Retrieved lymph nodes	0.423
AFP	0.042a	Reinforcement of duodenal stump	1
Tumor location	0.735	Embedding degree	0.142
Tumor diameter	0.394	Muscularis and serosa injury	0.395
Pathology type	0.912	Bleeding in reinforcement	0.790
pT	0.276	Reinforcement duration of duodenal stump	1
pN	0.582	Stump suture-fixed to the intestinal wall	0.100
Distant metastasis	1
pStage	0.505

^aP < 0.05.

^bP < 0.001.

¹≥ Grade 3a.

²≥ Grade 3a and except duodenal stump fistula.

AFP: Alpha-fetoprotein; ASA: American Society of Anesthesiologists; BMI: Body mass index; CA-199: Carbohydrate antigen 199; CEA: Carcinoembryonic antigen; OVF: Obstructive ventilatory failure; pN: Pathological nodal stage; pT: Pathological tumor stage.

Table 4 Multivariate logistic regression analysis of the duodenal stump fistula risk factors.

Variables	OR	95%CI	P value
Age	1.201	1.039-1.387	0.013a
CA-199	0.971	0.916-1.03	0.332
AFP	0.96	0.675-1.365	0.819
OVF	22.431	3.263-154.217	0.002a
Blood loss	0.998	0.992-1.004	0.523
Stump sutured to the intestinal wall	0.032	0.002-1.028	0.695

^aP < 0.05.

AFP: Alpha-fetoprotein; CA-199: Carbohydrate antigen 199; OVF: Obstructive ventilatory failure; OR: Odds ratio; CI: Confidence interval.

Table 5 Firth penalized logistic regression analysis of duodenal stump fistula risk factors.

Variables	OR	95%CI	P value
Age	1.157	1.038-1.325	0.006a
CA-199	1.001	0.994-1.002	0.217
AFP	1.005	0.997-1.015	0.139
OVF	13.075	2.643-67.87	0.002a
Blood loss	0.999	0.995-1.001	0.280
Stump sutured to the intestinal wall	0.112	0.004-1.346	0.094

^aP < 0.05.

AFP: Alpha-fetoprotein; CA-199: Carbohydrate antigen 199; OVF: Obstructive ventilatory failure; OR: Odds ratio; CI: Confidence interval.

Table 6 Bayesian logistic regression analysis of duodenal stump fistula risk factors.

Variables	OR	95%CI	P value
Age	1.187	1.06-1.355	0.003a
CA-199	0.971	0.909-1.099	0.530
AFP	0.916	0.693-1.003	0.079
OVF	15.623	2.764-86.577	0.004a
Blood loss	0.996	0.986-1.001	0.158
Stump sutured to the intestinal wall	0.084	0.004-1.185	0.136

^aP < 0.05.

AFP: Alpha-fetoprotein; CA-199: Carbohydrate antigen 199; OVF: Obstructive ventilatory failure; OR: Odds ratio; CI: Confidence interval.

Table 7 Rare events logistic regression analysis of duodenal stump fistula risk factors.

Variables	OR	95%CI	P value
Age	1.157	1.038-1.325	0.006a
CA-199	1.001	0.994-1.002	0.217
AFP	1.005	0.997-1.015	0.139
OVF	13.075	2.643-67.87	0.002a
Blood loss	0.999	0.995-1.001	0.280
Stump sutured to the intestinal wall	0.110	0.003-1.346	0.094

^aP < 0.05.

AFP: Alpha-fetoprotein; CA-199: Carbohydrate antigen 199; OVF: Obstructive ventilatory failure; OR: Odds ratio; CI: Confidence interval.

Table 8 Poisson regression analysis of duodenal stump fistula risk factors.

Variables	OR	95%CI	P value
Age	1.146	1.026-1.281	0.016a
CA-199	0.982	0.941-1.024	0.394
AFP	0.991	0.759-1.294	0.947
OVF	10.789	2.181-53.37	0.004a
Blood loss	0.998	0.992-1.004	0.585
Stump sutured to the intestinal wall	0.075	0.016-1.105	0.994

^aP < 0.05.

AFP: Alpha-fetoprotein; CA-199: Carbohydrate antigen 199; OVF: Obstructive ventilatory failure; OR: Odds ratio; CI: Confidence interval.

Results of the predictive nomogram of DSF and its diagnostic performance

The results of the DSF predictive nomogram and its diagnostic performance are shown in Figures 3 and 4. A DSF prediction nomogram was constructed based on age and OVF, the significant variables identified in the multivariate analysis. The nomogram exhibited high predictive performance, achieving an area under the curve of 0.90 (95%CI: 0.82-0.99). The calibration curve aligns well with the ideal line, indicating good overall predictive performance. The decision curve shows higher net benefit at risk thresholds between 0.1 and 0.6, supporting the nomogram’s clinical value. The model demonstrated the highest standardized net benefit when the risk threshold ranged from 20% to 40% in DCA. Therefore, we selected 30% as the optimal cutoff for high-risk stratification. Based on this threshold, patients older than 84 years without preoperative OVF or older than 66 years with preoperative OVF were identified as high-risk individuals.

Figure 3 The nomogram for predicting duodenal stump fistula. OVF: Obstructive ventilatory failure; DSF: Duodenal stump fistula.

Figure 4 Evaluation of the nomogram’s predictive performance. A: Receiver operating characteristic curve; B: Calibration curve; C: Decision curve analysis curve. AUC: Area under the curve.

Survival outcomes

Kaplan-Meier survival curves based on the competitive risk model were constructed for 325 patients who underwent gastrectomy with Roux-en-Y reconstruction for gastric cancer stratified by the presence of DSF (nDSF group, n = 318; DSF group, n = 7) (Figure 5). In the CSS analysis, the DSF group exhibited a significantly lower 5-year survival rate than the nDSF group [57.1% vs 83.3%, P = 0.04; hazard ratio (HR) = 3.18, 95%CI: 0.61-1.95]. In contrast, the analyses of the OS (57.1% vs 76.7%, P = 0.17; HR = 2.19, 95%CI: 0.58-1.35) and PFS (57.1% vs 72.3%, P = 0.31; HR = 1.80, 95%CI: 0.59-1.00) revealed no statistically significant differences between the DSF and nDSF groups.

Figure 5 Survival analysis based on competitive risk model. A: Overall survival; B: Cancer-specific survival; C: Disease-free survival.

DISCUSSION

DSF is an intractable and life-threatening complication after gastrectomy that cannot be dismissed. However, the low incidence of DSF limits large-scale prospective studies and hinders validation of its risk factors. In addition, its complex etiology involving oncologic characteristics, surgical techniques, and patient comorbidities further complicates risk identification.

The KSCC DELICATE study reported that patients with DSF had a higher proportion of ≥ pN1 disease[6]. The clinicopathological analyses of our study, also showed a higher proportion of pN2-N3 cases in DSF patients. Yu et al[12] and Paik et al[9] reported a significantly increased DSF incidence in advanced-stage gastric cancer. These results may suggest that DSF may be more likely to occur in more advanced gastric cancer patients. This may be partly explained by the need for more extensive surgical procedures and prolonged operative time in advanced-stage disease. These procedures are often associated with increased tissue fragility and heightened perioperative physiological stress. All of these factors can impair anastomotic healing and thereby elevate the risk of DSF.

In addition to tumor-related factors, several surgical and intraoperative conditions have been implicated in the development of DSF. Although one study reported a 5.6-fold increased risk associated with laparoscopic surgery[3], subsequent investigations found no significant association[9,12,26]. This includes our own findings. This inconsistency may reflect improvements in laparoscopic techniques and increased surgical proficiency over time. Massive intraoperative bleeding and contamination have been consistently identified as independent risk factors[5,10,12]. These findings underscore the importance of careful hemostasis and adherence to aseptic protocols to reduce DSF risk.

The impact of duodenal stump reinforcement on DSF prevention has been widely explored, but findings remain inconsistent. Several studies support its protective effect, particularly with seromuscular or laparoscopic techniques[6,27,28]. In contrast, other studies found no consistent benefit. Sun et al[13] and Yan et al[29] reported no overall difference in DSF incidence between reinforced and non-reinforced groups, although subgroup analyses suggested potential advantages of purse-string sutures. In our study, 91.7% of patients received absorbable purse-string sutures, but reinforcement was not associated with a reduced DSF rate. All DSF cases occurred despite reinforcement, and no significant differences were found in technical parameters between groups, suggesting that surgical detail and standardization may play a more critical role than the use of reinforcement alone.

Our center implemented a novel technique for stump fixation involving suturing the stump to the jejunal side wall after satisfactory embedding and mobilization of the duodenal stump. The concept of this technique goes beyond preventing twisting of the reconstructed intestine and internal hernia. It also aims to cover the duodenal stump with a healthy jejunal wall to further reinforce the stump and avoid even the risk of a minor leak. This concept includes the possibility that if DSF does occur, drainage through the sutured jejunal wall into the intestinal lumen could still be expected. In this study, 110 patients (33.8% of cases analyzed) underwent this technique, and none experienced DSF, although this difference did not reach statistical significance in the group comparison (P = 0.131).

Preoperative comorbidities are increasingly recognized as significant contributors to DSF risk. Multiple studies have identified cardiovascular, hepatic, and pulmonary conditions as key factors. Notably, the presence of multiple comorbidities has been consistently associated with increased DSF incidence[9,10]. Among these, obstructive ventilatory dysfunction (OVF)-encompassing chronic obstructive pulmonary disease (COPD), asthma, and bronchiectasis has drawn particular attention due to its link with postoperative complications. For instance, COPD has been shown to predict both overall and major complications following gastrectomy[30]. Additionally, reduced pulmonary function (FEV1% < 70%) is also associated with worse postoperative outcomes[31]. Asthma has even been identified as an independent risk factor for DSF-related mortality[6], further reinforcing the relevance of OVF-related conditions. Collectively, these findings underscore the importance of integrating pulmonary function assessment into perioperative risk stratification for gastric cancer surgery.

However, the reason patients with OVF are more likely to develop postoperative complications including DSF remains unclear. Several mechanisms may be involved. OVF patients are more prone to respiratory tract infections[32] and often have immune dysfunction[33]. Postoperative hypoxemia is also common in this group and may contribute to abnormal inflammatory responses while promoting ischemia-reperfusion injury[34]. In particular, hypoxia may impair tissue healing, which is critical for the recovery of anastomoses[35]. These conditions are potential risk factors for postoperative complications.

Age was also confirmed as a significant risk factor and has been consistently associated with DSF in previous studies as a reliable predictor of postoperative complications[31,36]. Based on these findings, a nomogram-based risk stratification model was developed facilitate the identification of high-risk patients.

Guided by DSF risk stratification, we propose a perioperative management strategy focused on patients with high-risk features such as OVF. Preoperative measures include enteral nutrition and optimization of comorbidities, such as diabetes and hypoalbuminemia. Pulmonary rehabilitation through smoking cessation and respiratory training is also recommended. Intraoperative strategies involve reducing anastomotic tension with reinforced sutures, preserving blood supply by minimizing electrocautery injury and using indocyanine green fluorescence imaging. Fixation of the duodenal stump to the jejunum may enhance healing and facilitate drainage of minor leaks. Postoperative care includes early computed tomography surveillance on postoperative days 3-5 and duodenography if needed. It also involves early oral intake and the use of proton pump inhibitors to minimize anastomotic irritation.

Although our study found no significant differences in OS and PFS between patients with and without DSF, a notably lower CSS was observed in the DSF group, suggesting a potential impact of DSF on cancer-specific outcomes. This aligns with previous findings, such as those by Yoo et al[11], which also identified DSF as an independent risk factor for CSS in advanced gastric cancer. While the adverse prognostic role of postoperative leakage remains debated, inflammatory responses induced by complications may promote tumor progression, especially in the presence of exfoliated cancer cells. Unlike more common anastomotic leaks, DSF involves a different digestive fluid composition. It remains a rare but clinically significant event, warranting larger, well-designed studies to elucidate its oncologic implications. Ultimately, improving preoperative risk stratification, optimizing stump management, and enhancing perioperative care may reduce the incidence of DSF and its negative prognostic impact.

This study has several limitations. First, given the retrospective nature of this study, the potential for information bias arising from incomplete or imprecise documentation cannot be definitively excluded. Although multivariate analysis was conducted, residual confounding factors may still be present. Second, this single-center study included an older population with a high prevalence of comorbidities, which could restrict the generalizability of its findings. Additionally, in accordance with JGCA guidelines, neoadjuvant therapy is infrequently administered to Japanese gastric cancer patients. To minimize potential confounding, patients who received neoadjuvant treatment were excluded. However, this may limit the generalizability of our findings to Western populations, where neoadjuvant therapy is more widely adopted. Finally, the limited sample size, particularly the small number of DSF cases, hindered the ability to further conduct subgroup analyses to explore DSF-related factors and outcomes.

CONCLUSION

Advanced age and OVF have been identified as significant risk factors for DSF incidence following radical gastrectomy for gastric cancer. The predictive nomogram we have developed for DSF risk can be implemented with a high degree of accuracy. DSF profoundly impacts CSS, highlighting the critical need for prevention, early diagnosis and intervention, and aggressive management. Suturing fixation of the duodenal stump to the jejunal wall may be one option to avoid DSF effectively.

ACKNOWLEDGEMENTS

We thank all the members for their insightful discussions and thank you to Ms. Nojiri S for her professional support in statistics and Ms. Ecoff K for proofreading the manuscript in English.