Association between Candida esophagitis and esophagojejunal leakage following total gastrectomy: a retrospective cohort study

Abstract

Background and objectives

Esophagojejunal (EJ) leakage is a serious complication following total gastrectomy for gastric cancer. While several nutritional and treatment-related risk factors have been described, the role of Candida esophagitis (CE) in anastomotic failure has not been previously investigated.

Methods

This retrospective cohort study included 268 patients with gastric adenocarcinoma who underwent total gastrectomy with EJ anastomosis. The study was conducted between March 2021 and March 2025 at a tertiary referral center. CE was diagnosed by histopathologic examination of proximal esophageal margins submitted during surgery. Univariable and multivariable logistic regression analyses were performed to determine predictors of EJ leakage. Best subsets variable selection using Akaike’s Information Criterion (AIC) and the Bayesian Information Criterion (BIC) guided final model development.

Results

Among the 268 patients, 48 (17.9%) developed EJ leakage. Multivariable analysis identified CE (OR: 2.19, p = 0.043), hypoalbuminemia (< 3.5 g/dL) (OR: 3.08, p = 0.007), BMI ≥ 25 kg/m² (OR: 3.68, p = 0.004), and administration of neoadjuvant therapy (OR: 2.34, p = 0.024) as independent predictors of EJ leakage. Additional analyses of diagnostic timing indicated that CE detected only on permanent histology (delayed treatment) was associated with higher odds of leak (adj OR 3.62; 95%CI: 1.42–9.23; p = 0.007), whereas CE detected on intraoperative frozen section was not.

Conclusion

CE was associated with increased odds of EJ leakage after adjustment, but causality cannot be inferred from this retrospective study. The finding that delayed CE diagnosis was linked to higher leak risk suggests diagnostic timing may matter. Prospective validation of targeted esophageal assessment and timed antifungal strategies is warranted. Elevated BMI, hypoalbuminemia, and neoadjuvant therapy also contributed to higher odds of EJ leakage.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1007/s00423-025-03942-y.

Introduction

Gastric cancer is the fifth most common cause of cancer death worldwide [1]. Esophagojejunal (EJ) anastomotic leakage remains one of the most serious and life-threatening complications following total gastrectomy for gastric cancer [2]. Reported incidence rates vary between 1% and 20%, depending on institutional experience, patient comorbidities, and surgical technique [3]. EJ leakage continues to be a major source of postoperative morbidity, prolonged hospitalization, and mortality among gastric cancer patients [4]. Despite advancements in surgical practice, including high-volume centers and improved perioperative care protocols, the multifactorial nature of anastomotic leakage still presents challenges for prediction and prevention [5].

Several clinical and nutritional factors have been proposed to influence the risk of EJ leakage, such as poor nutritional status, chemotherapy, comorbidities, and body mass index (BMI) [4, 6]. In recent years, microbial dysbiosis has also been considered as an underexplored contributor to impaired anastomotic healing [7]. Candida esophagitis (CE) has received attention for a possible association with upper gastrointestinal epithelial dysplasia and malignancy, although causal evidence remains limited [8, 9]. Proposed mechanisms include chronic mucosal inflammation, production of carcinogenic compounds, triggering the release of reactive oxygen species, reshaping of the immune microenvironment, and stimulation of oncogenic signals [8]. Moreover, fungal infections may contribute to local tissue inflammation, compromise mucosal integrity, and alter microvascular perfusion [10]. These factors could, in turn, impair the healing of gastrointestinal anastomoses. It should be noted that no prior study has comprehensively evaluated the association between histologically confirmed CE and the development of EJ leakage following total gastrectomy.

Accordingly, we aimed to investigate the relationship between CE and the risk of EJ leakage in patients undergoing total gastrectomy for gastric adenocarcinoma at a high-volume tertiary referral center. We hypothesized that the sole presence of CE, regardless of the timing of diagnosis or antifungal treatment, could be associated with a higher likelihood of postoperative EJ leakage. we also sought to identify other independent predictors of leakage using multivariable modeling. This approach helps develop a clearer understanding of risk factors in the high-risk surgical population.

Materials and methods

Study design and population

This retrospective cohort study was conducted at Sina Hospital, a high-volume tertiary referral center affiliated with Tehran University of Medical Sciences (TUMS), Tehran, Iran. The study population consisted of patients with histologically confirmed gastric adenocarcinoma who underwent total gastrectomy with EJ anastomosis between March 2021 and March 2025. All surgeries were performed by board-certified attending surgeons at Sina hospital. Patients were identified through the hospital’s database.

The study protocol was approved by the Research Ethics Committee of TUMS (ethical code: IR.TUMS.SINAHOSPITAL.REC.1402.056). The study was conducted in accordance with the principles of the Declaration of Helsinki [11]. All participants provided written informed consent for the use of their medical data in research. This study was designed and conducted in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines for cohort studies (see Supplementary Materials).

Patients aged ≥ 18 years with histologically confirmed gastric adenocarcinoma that underwent total gastrectomy with EJ anastomosis were included in the study. Moreover, as part of our institutional protocol, proximal esophageal margins were routinely submitted for histopathological evaluation during the study period, regardless of gross abnormality. This standardized approach allowed for retrospective identification of CE in all patients. Patients were excluded if they had undergone subtotal or partial gastrectomy, had incomplete pathological or operative records (including missing CE diagnosis or anastomotic data), or died within 72 h postoperatively before EJ leak assessment could be completed.

From a total of 372 patients initially assessed, 268 met the eligibility criteria and were included in the final analysis.

Diagnosis and staging of gastric cancer

The diagnosis of gastric cancer was made through histopathological analysis of endoscopic specimens. Patients typically presented with nonspecific gastrointestinal symptoms, such as dyspepsia, early satiety, or unintentional weight loss, prompting upper gastrointestinal endoscopy. Multiple mucosal biopsies were obtained from the suspected lesion and submitted for histological analysis.

Staging was performed in accordance with the 8th edition of the American Joint Committee on Cancer (AJCC) TNM classification [12]. All patients underwent a comprehensive staging workup, including contrast-enhanced computed tomography (CT) scans of the chest, abdomen, and pelvis to assess for local tumor invasion, lymphadenopathy, and distant metastases. Endoscopic ultrasound (EUS) was used selectively to better evaluate tumor depth and regional lymph node involvement when imaging results were inconclusive. Accordingly, patients were eligible for total gastrectomy with curative intent when clinically indicated, in line with international treatment guidelines.

Neoadjuvant therapy protocol

Neoadjuvant chemotherapy was administered to patients with locally advanced gastric adenocarcinoma when clinically appropriate, based on tumor stage, performance status, and multidisciplinary tumor board recommendations.

All patients who received neoadjuvant therapy were treated with the FLOT regimen, consisting of: 5-Fluorouracil (2,600 mg/m²) as a 24-hour continuous intravenous infusion, Leucovorin (200 mg/m²) intravenously, Oxaliplatin (85 mg/m²) intravenously, and Docetaxel (50 mg/m²) intravenously.

Surgical procedure

All patients underwent elective open total gastrectomy with curative intent, performed by board-certified attending general surgeons at Sina Hospital. No procedures were performed by resident physicians, and no laparoscopic or minimally invasive approaches were utilized during the study period, ensuring uniformity in operative technique.

The standard procedure involved en bloc resection of the entire stomach along with the proximal duodenum and distal esophagus, with regional D2 lymphadenectomy in accordance with international oncologic guidelines. Splenectomy and/or distal pancreatectomy were performed only when directly indicated by tumor invasion.

Reconstruction was achieved by creating a Roux-en-Y esophagojejunostomy (EJ anastomosis). In all cases, the esophagojejunal anastomosis was performed in an end-to-side fashion using hand-sewn suture technique. No stapled anastomoses were performed. The anastomosis was constructed with double-layered interrupted absorbable sutures to optimize strength and reduce the risk of leakage.

Diagnosis of CE and EJ leakage

CE was diagnosed based on histopathologic examination of proximal esophageal tissue obtained during gastrectomy. As part of our institutional surgical pathology protocol, proximal esophageal margins were systematically submitted for histological evaluation, regardless of gross appearance. Routine preoperative upper endoscopy in our institution was performed primarily for tumor diagnosis and staging; systematic preoperative screening for CE (with targeted esophageal biopsies or fungal testing) was not performed and preoperative endoscopy reports did not include standardized documentation of CE for all patients.

Two diagnostic modalities were employed for the detection of CE: [1] Intraoperative frozen section analysis, which allowed for immediate pathological assessment in patients. Frozen sections were processed intraoperatively by the pathology laboratory and the results were communicated to the operating team during the procedure. When fungal elements were identified on frozen section, antifungal therapy (fluconazole) was started in the immediate postoperative period. Precise timestamps documenting the minute of frozen-section reporting and the exact hour of the first antifungal dose were not uniformly recorded in the retrospective charts; [2] Permanent paraffin-embedded histopathologic analysis, conducted postoperatively. CE was defined by the presence of fungal hyphae infiltrating the squamous epithelium, typically accompanied by inflammatory cell infiltration, as observed on hematoxylin and eosin (H&E) staining, and periodic acid Schiff (PAS) staining (Fig. 1).

Fig. 1.

Histopathological examination of the proximal margin of gastrectomy; arrows indicate fungal yeasts and hyphae in H&E staining (left) and PAS staining (right). (X400)

EJ leakage was diagnosed based on a combination of clinical, laboratory, and radiologic findings. Patients were suspected to have leakage when they exhibited symptoms such as unexplained fever, tachycardia, and peritoneal irritation, accompanied by elevated levels of C-reactive protein (CRP) and leukocytosis (white blood cells > 11000/µl) with neutrophilia [13]. Neutrophilia is defined as an increase in the absolute neutrophil count of approximately more than 7700 neutrophils/µl or 70% of leukocytes [14]. The diagnosis was confirmed by an abdominal computed tomography (CT) scan after the surgery, which revealed varying amounts of fluid collection or well-formed abscesses in the appropriate subhepatic space. Contrast swallow studies and upper endoscopy (gastroscopy) were used selectively at the discretion of the attending surgical team when CT findings were equivocal, minor leaks were suspected, or when direct visualization was required for planning endoscopic therapy, or when contrast studies were preferred for clinical reasons.

Management of CE and EJ leakage

Patients diagnosed with CE based on intraoperative frozen section received empiric antifungal therapy in the immediate postoperative period. Treatment consisted of Fluconazole 400 mg intravenously once daily, continued for a minimum of 14 days. The treatment was adjusted based on clinical response and microbiological consultation. Patients in whom CE was identified only in the permanent pathology report received delayed antifungal therapy initiated after histopathological confirmation.

EJ leakage was managed according to institutional protocols. Management decisions (conservative therapy, percutaneous drainage, or reoperation) were based on the combination of clinical, laboratory, and imaging findings. Patients with contained leaks and stable clinical status were managed conservatively with broad-spectrum intravenous antibiotics, nil per os (NPO) status, and enteral or parenteral nutritional support. In cases with significant fluid collections or abscesses, image-guided percutaneous drainage was performed. Patients with generalized peritonitis, hemodynamic instability, or failed non-operative management underwent surgical re-exploration and either primary repair, diversion, or creation of a feeding jejunostomy as appropriate. It should be noted that endoscopic vacuum (Endo-VAC) therapy was not available as part of the institutional treatment repertoire during the study period; consequently, no patients in this cohort received EVT.

Statistical analysis

All statistical analyses were conducted using Stata version 17.0 (StataCorp, College Station, TX, USA). Descriptive statistics were used to summarize the baseline characteristics of the study population. For categorical variables, data were presented as frequencies and percentages. Continuous variables were expressed as mean ± standard deviation (SD) or median with interquartile range (IQR), depending on their distribution.

Candidate predictors considered for multivariable analysis were chosen based on a combination of clinical relevance (variables previously reported as risk factors for esophagojejunal anastomotic leakage) and an initial data screen (univariable tests, p < 0.10), as is recommended in recent methodological. Clinically important variables available in our dataset included body mass index (BMI), serum albumin (nutritional status), receipt of neoadjuvant therapy, age, comorbidities (for example COPD, diabetes), smoking and steroid use, use of proton-pump inhibitors, ASA score, and tumor characteristics.

To avoid relying exclusively on univariable screening and to reduce model-building bias we then applied a best-subsets selection strategy using the Furnival-Wilson “leaps-and-bounds” algorithm as implemented in Stata’s “gvselect”. This procedure evaluates candidate predictor subsets by their log-likelihoods and reports the best models at each subset size; models were ranked by Akaike’s Information Criterion (AIC) and the Bayesian Information Criterion (BIC) and inspected for clinical plausibility [15].

Final model choice balanced model fit (AIC/BIC), parsimony and events-per-variable considerations. Hence, we selected a parsimonious primary multivariable model that included the exposure of interest (CE) plus the strongest predictors identified by model selection while keeping the number of predictors consistent with our event count to avoid overfitting. The variable-selection strategy follows current recommendations to combine subject-matter knowledge with information-criteria-based model comparison rather than rely on automatic stepwise rules alone [16].

Model performance was evaluated by calculating the area under the receiver operating characteristic (ROC) curve, and goodness-of-fit was assessed using the Hosmer–Lemeshow test. A two-sided p-value < 0.05 was deemed statistically significant.

Results

Study population and baseline characteristics

From an initial cohort of 372 patients, 268 met the eligibility criteria and were included in the final analysis. A detailed flowchart outlining patient selection is provided in Fig. 2.

Fig. 2.

Flow diagram illustrating patient selection criteria for the current study

Among the included patients, 48 (17.9%) experienced postoperative EJ leakage, while 220 (82.1%) did not. The mean age of the cohort was 63.9 ± 6.3 years and, 66.0% were male. The overall characteristics of the study population stratified by EJ leakage status are summarized in Table 1.

Table 1.

Characteristics of the study population based on esophagojejunal (EJ) leakage status

	(N = 268) Total	(n = 220) EJ Leakage -	(n = 48) EJ Leakage +	p-value*
Age, years Mean ± SD	63.9 ± 6.3	63.7 ± 6.1	64.5 ± 6.8	0.44
Sex (%)				0.57
Male	177 (66.0)	147 (66.8)	30 (62.5)
Female	91 (34.0)	73 (33.2)	18 (37.5)
BMI (%)				0.097
< 18.5 kg/m2	34 (13.3)	26 (12.4)	8 (17.0)
18.5 to 24.9 kg/m2	184 (71.9)	156 (74.6)	28 (59.6)
≥ 25 kg/m2	38 (14.8)	27 (12.9)	11 (23.4)
HTN (%)	73 (28.2)	56 (26.2)	17 (37.8)	0.12
Diabetes Mellitus (%)	35 (13.6)	26 (12.3)	9 (19.6)	0.19
COPD (%)	26 (10.1)	20 (9.4)	6 (13.6)	0.40
Neoadjuvant Therapy (%)	103 (41.2)	80 (38.8)	23 (52.3)	0.10
Alcohol Use (%)	54 (21.1)	45 (21.2)	9 (20.5)	0.91
Cigarette Smoking (%)	67 (26.3)	52 (24.9)	15 (32.6)	0.28
Steroid Therapy (%)	16 (6.4)	12 (5.8)	4 (8.9)	0.45
PPI Therapy (%)	90 (35.0)	74 (35.1)	16 (34.8)	0.97
Serum Albumin (%)				0.014
< 3.5 g/dl	208 (80.0)	178 (82.8)	30 (66.7)
≥ 3.5 g/dl	52 (20.0)	37 (17.2)	15 (33.3)
ASA Grade (%)				0.88
I	2 (0.8)	2 (0.9)	0 (0.0)
II	125 (48.1)	102 (47.7)	23 (50.0)
III	129 (49.6)	107 (50.0)	22 (47.8)
IV	4 (1.5)	3 (1.4)	1 (2.2)
Tumor Type (%)				0.87
Intestinal	220 (82.1)	181 (82.3)	39 (81.2)
Diffuse	48 (17.9)	39 (17.7)	9 (18.8)
Tumor Site (%)				0.57
Antrum	127 (47.4)	101 (45.9)	26 (54.2)
Cardia	69 (25.7)	60 (27.3)	9 (18.8)
Body	39 (14.6)	31 (14.1)	8 (16.7)
Fundus	33 (12.3)	28 (12.7)	5 (10.4)
Tumor Stage (%)				0.11
II	121 (45.1)	94 (42.7)	27 (56.2)
III	120 (44.8)	105 (47.7)	15 (31.2)
IV	27 (10.1)	21 (9.5)	6 (12.5)
CE (%)	75 (28.0)	59 (26.8)	16 (33.3)	0.36
CE Diagnostic Method (%)				0.55
Frozen Section	33 (44.0)	27 (45.8)	6 (37.5)
Permanent Pathology	42 (56.0)	32 (54.2)	10 (62.5)

EJ Esophagojejunal, BMI Body mass index, HTN Hypertension, COPD Chronic obstructive pulmonary disease, PPI Proton pump inhibitor, ASA American Society of Anesthesiologists, CE Candidal esophagitis

* P < 0.05 was considered statistically significant

Patients with EJ leakage had a higher proportion of hypoalbuminemia (< 3.5 g/dL) compared to those without leakage (33.3% vs. 17.2%, p = 0.014). Similarly, neoadjuvant therapy was more frequent among patients with EJ leakage (52.3% vs. 38.8%). Nonetheless, it did not reach statistical significance (p = 0.10). No significant differences were found for other clinical characteristics, including comorbidities, ASA class, tumor site, or tumor type.

The prevalence of CE did not differ significantly between patients with and without EJ leakage (33.3% vs. 26.8%, p = 0.36). Among the 75 patients with CE, 33 were identified by intraoperative frozen Sect. (6/33 with EJ leakage, 18.2%) and 42 by permanent pathology (10/42 with EJ leakage, 23.8%) (Supplementary Table S1). Further, as demonstrated in Table 1, the difference in leak rates between the two diagnostic/treatment-timing groups was not statistically significant (p = 0.55). The paired 2 × 2 table is presented in Supplementary Table S2. Using permanent histology as the reference standard, frozen section showed sensitivity 44.0% (95%CI: 32.5–55.9), specificity 100% (95%CI: 98.1–100), PPV 100% (95%CI: 89.4–100) and NPV 82.1% (95%CI: 76.6–86.8). Observed agreement was 84.3% and Cohen’s κ 0.53, indicating moderate agreement.

Univariable analyses

In the univariable logistic regression analysis, BMI ≥ 25 kg/m² (OR = 2.27, 95%CI: 1.01 to 5.09, p = 0.047) was significantly associated with increased odds of EJ leakage compared to normal BMI (18.5 to 24.9 kg/m²). Serum albumin < 3.5 g/dL (OR = 2.41, 95% CI: 1.18 to 4.91, p = 0.016) contributed to greater odds of EJ leakage as well. Tumor stage III (vs. stage II) was associated with lower odds of leakage (p = 0.047). However, CE, neoadjuvant therapy, and other variables did not demonstrate significant associations in univariable analysis. ASA Grade IV was excluded from the regression due to quasi-complete separation. Univariable associations between variables and EJ leakage after total gastrectomy is demonstrated in Table 2.

Table 2.

Univariable associations between patient and clinical characteristics and EJ leakage after total gastrectomy

Variables	Odds Ratio (95% CI)	p-value*
Age	1.02 (0.97 to 1.07)	0.44
Male Sex	0.82 (0.43 to 1.58)	0.57
BMI
18.5 to 24.9 kg/m2	Reference
≥ 25 kg/m2	2.27 (1.01 to 5.09)	0.047
<18.5 kg/m2	1.71 (0.70 to 4.17)	0.24
HTN	1.71 (0.87 to 3.37)	0.12
Diabetes Mellitus	1.74 (0.75 to 4.02)	0.19
COPD	1.52 (0.57 to 4.05)	0.40
Neoadjuvant Therapy	1.73 (0.90 to 3.32)	0.10
Alcohol Use	0.95 (0.43 to 2.13)	0.91
Cigarette Smoking	1.46 (0.73 to 2.92)	0.28
Steroid Therapy	1.58 (0.48 to 5.14)	0.45
PPI Therapy	0.99 (0.51 to 1.93)	0.97
Serum Albumin < 3.5 g/dl	2.41 (1.18 to 4.91)	0.016
ASA Grade
I	Reference
II	0.68 (0.67 to 6.80)	0.74
III	0.62 (0.61 to 6.21)	0.68
IV	Omitted†
Diffuse-type Tumor	1.07 (0.48 to 2.39)	0.17
Tumor Site
Antrum	Reference
Cardia	0.58 (0.26 to 1.33)	0.20
Body	1.00 (0.41 to 2.44)	0.99
Fundus	0.69 (0.24 to 1.97)	0.49
Tumor Stage
II	Reference
III	0.50 (0.25 to 0.99)	0.047
IV	0.99 (0.36 to 2.71)	0.99
CE	1.36 (0.70 to 2.67)	0.36
CE Diagnostic Method
Frozen Section	Reference
Permanent Pathology	1.41 (0.45 to 4.37)	0.56

CI Confidence Interval, BMI Body mass index, HTN Hypertension, COPD Chronic obstructive pulmonary disease, PPI Proton pump inhibitor, ASA American Society of Anesthesiologists, CE Candidal esophagitis

^* P < 0.05 was considered statistically significant

^†ASA IV was omitted from logistic regression due to quasi-complete separation related to sparse data

Associations between CE and potential correlates (hypoalbuminemia, neoadjuvant therapy, smoking, BMI, and others) were also examined. The results showed that CE was less frequent among patients who received neoadjuvant therapy (OR 0.51; 95% CI 0.28–0.92; p = 0.026). No significant associations were observed with hypoalbuminemia, smoking, or other variables (Supplementary Table S3).

Final model and multivariable analyses

Multivariable logistic regression identified several independent predictors of EJ leakage. After variable selection using best subsets modeling, based on AIC and BIC, the final model included CE, neoadjuvant therapy, serum albumin, and BMI. Accordingly, as shown in Table 3, BMI ≥ 25 kg/m², hypoalbuminemia, neoadjuvant therapy, and the presence of CE remained statistically significant in the adjusted model, each demonstrating a positive association with the likelihood of postoperative leakage. Among them, elevated BMI and hypoalbuminemia showed the strongest associations. Model performance assessment demonstrated acceptable discrimination and calibration. The receiver operating characteristic (ROC) curve and Hosmer-Lemeshow goodness-of-fit test results for the final multivariable model are provided in the Supplementary Material (Supplementary Figure S1).

Table 3.

Multivariable logistic regression model identifying independent predictors of EJ leakage following total gastrectomy

Variables	Odds Ratio (95% CI)	p-value*
BMI
18.5 to 24.9 kg/m2	Reference
≥ 25 kg/m2	3.68 (1.51 to 8.98)	0.004
<18.5 kg/m2	2.02 (0.74 to 5.47)	0.17
Neoadjuvant Therapy	2.34 (1.12 to 4.90)	0.024
Serum Albumin < 3.5 g/dl	3.08 (1.36 to 6.95)	0.007
CE	2.19 (1.03 to 4.69)	0.043

CI Confidence Interval, BMI Body mass index, CE Candidal esophagitis

^* P < 0.05 was considered statistically significant

To evaluate the influence of diagnostic and treatment timing, we performed additional exploratory analysis by creating a three-level variable (“CE Timing”: no CE/Early treatment (CE detected on intraoperative frozen section)/Late treatment (CE detected only on permanent pathology)) and fitted a multivariable model that included CE Timing, BMI category, serum albumin status, and neoadjuvant therapy (Supplementary Table S4). In this model, CE detected only on permanent pathology (delayed treatment) was associated with an increased odds of EJ leakage (adjusted OR 3.62; 95% CI 1.42–9.23; p = 0.007), whereas CE detected on frozen section (with immediate postoperative antifungal initiation) did not lead to significantly higher odds of EJ leakage. Other covariates, including high BMI, neoadjuvant therapy, and hypoalbuminemia, remained independently associated with EJ leakage in this model. The “CE Timing” model demonstrated good calibration and acceptable discrimination. Model comparison with our primary multivariable model with simpler binary-CE variable showed a marginal improvement in AIC for the timing model (AIC 206.32 vs. 207.18) but a higher BIC (230.42 vs. 227.83) and a non-significant likelihood-ratio test (p = 0.09). Because the timing model did not provide a clear statistical improvement over the simpler binary model, we present the “binary-CE” model as the primary multivariable analysis and the “CE Timing” model as an exploratory supplementary analysis. The ROC curve and Hosmer-Lemeshow goodness-of-fit test results for the additional exploratory multivariable model are also provided in the Supplementary Material (Supplementary Figure S2).

Discussion

In this retrospective cohort study, we identified CE as an independent predictor of EJ leakage following total gastrectomy in gastric cancer patients. Elevated BMI, hypoalbuminemia, and receipt of neoadjuvant therapy were also identified as independent determinants of EJ leakage. Our findings add to the growing body of evidence that link microbiological and immunologic factors to anastomotic failure.

The incidence of EJ leakage in our cohort was 17.9%. This rate is at the higher end of the reported range (typically 1–20%) in high-volume surgical centers [3]. The presence of CE, confirmed via histopathology either on frozen or permanent section, was associated with a more than twofold increase in the odds of leakage after adjusting for other variables. This association aligns with emerging evidence suggesting that fungal colonization and infection can negatively affect anastomotic healing by disrupting mucosal integrity, triggering local inflammatory responses, and impairing tissue perfusion [7, 10, 17].

In this study, EJ leakage was diagnosed based on clinical suspicion and confirmed primarily by contrast-enhanced abdominal CT, which is widely regarded as the principal imaging modality for detecting postoperative anastomotic leaks, as it reliably demonstrates extraluminal collections, abscesses, or contrast extravasation [18, 19]. In selected cases with diagnostic uncertainty, minor leak suspicion, or when direct mucosal evaluation was considered helpful for therapeutic planning, contrast swallow studies or endoscopic assessment were performed at the discretion of the treating team. Additional confirmation was obtained intraoperatively during re-exploration or through image-guided drainage when necessary. Routine diagnostic gastroscopy was not performed for every patient, which aligns with standard practice in high-volume centers and reflects the targeted use of adjunctive tests in clinically indicated situations [20].

The observation that CE was not significant in univariable analysis but became significant after multivariable adjustment highlights the complexity of interrelationships among perioperative factors. This phenomenon may reflect confounding (where covariates are associated with both CE and leakage) or suppression (where a variable masks another’s effect in crude analyses) [21, 22]. Importantly, other variables such as BMI and hypoalbuminemia demonstrated stronger and more consistent associations across models, suggesting they are major drivers of anastomotic risk. Given the retrospective design, limited events per subgroup, and potential for residual confounding, we therefore interpret CE as a potentially important associated factor rather than definitive proof of causality.

Our findings regarding hypoalbuminemia and elevated BMI are consistent with prior reports. Hypoalbuminemia has long been recognized as a marker of impaired nutritional and immunologic reserve [23, 24]. Accordingly, it has been identified as a significant contributor to EJ leakage following gastrectomy [25]. Likewise, elevated BMI is linked to technical challenges during surgery, and adversely affects perfusion across anastomoses, making it another strong predictor of leakage [26].

Interestingly, neoadjuvant therapy was also independently associated with EJ leakage. This observation adds to the growing but mixed body of evidence on the impact of neoadjuvant chemotherapy on anastomotic outcomes. The FLOT regimen has been shown to improve long-term survival outcomes in patients with locally advanced gastric cancer [27]. However, the independent impact of neoadjuvant therapy on the risk of EJ leakage remains an area of active investigation [28]. Moreover, in this study, CE was significantly less frequent among patients who received neoadjuvant therapy. This may be partly explained by selection bias and supportive-care differences rather than a protective biological effect of chemotherapy. However, residual confounding and center-specific practices may also contribute, so this remains a hypothesis-generating observation.

We observed that CE identified only on permanent pathology (and therefore treated later) was associated with increased odds of leakage, while CE identified on frozen section was not. This point is clinically intriguing and hypothesis-generating. It suggests that earlier recognition and initiation of antifungal therapy could potentially attenuate the deleterious effect of fungal colonization on anastomotic healing. However, as precise timestamps for antifungal administration are lacking, these findings cannot establish causality. They do, however, support the rationale for prospective evaluation of standardized preoperative esophageal assessment. For example, endoscopy with targeted esophageal sampling and timed antifungal protocols could prove beneficial in determining whether true preemptive therapy reduces EJ leak rates. Besides, all 33 intraoperative frozen-positive cases were confirmed on permanent histology in our series. This supports the clinical practice of initiating antifungal therapy when frozen section is positive. However, the limited sensitivity indicates that many CE cases are missed intraoperatively, which plausibly explains the higher leak odds observed in the delayed-diagnosis group (CE detected only on permanent histology).

Overall, considering our available data, two non-exclusive possibilities could be suggested. Firstly, CE may be a marker of a generalized impaired mucosal/host defense (for example, related to malnutrition, immunosuppression, or prior chemotherapy) that predisposes to leakage. Moreover, CE could act as a mediator, whereby fungal colonization directly impairs mucosal integrity and healing. However, it should be noted that our dataset does not definitively separate these possibilities and focused prospective evaluation is needed.

To our knowledge, this is the first study to systematically assess the association between CE and EJ leakage in a uniform cohort of patients undergoing total gastrectomy. A key strength of this study lies in its standardized institutional protocol. In this context, all proximal esophageal margins were routinely submitted for histopathological evaluation. This enabled us to reliably capture the presence or absence of CE in all included patients.

There are, however, several limitations to this investigation. Several studies have demonstrated that covered stents in combination with adequate drainage and nutritional support can achieve high rates of anastomotic leak closure after total gastrectomy [20, 29]. However, Endo-VAC therapy has also emerged as a promising endoscopic option for the management of upper gastrointestinal anastomotic leaks and has been associated with favorable outcomes [30]. It should be noted that Endo-VAC method was not part of our institutional practice during the study. Accordingly, the absence of Endo-VAC therapy may limit direct generalizability to centers where this approach is routinely employed. Comparative prospective studies are needed to establish whether incorporation of Endo-VAC therapy changes leak-related outcomes in patients with and without CE.

The retrospective design of the current investigation may introduce selection bias and limits control over unmeasured confounders. The single-center nature of the study may affect generalizability. We did not perform systematic preoperative endoscopic screening for CE; routine endoscopy was directed at tumor diagnosis and staging, and preoperative documentation of esophageal candidiasis was not uniformly available. Therefore, we could not evaluate the potential impact of preoperative detection and preemptive antifungal therapy on the risk of EJ leakage in this retrospective investigation. In addition, we did not uniformly capture detailed timestamps for frozen-section reporting or the exact timing of the first antifungal dose. This limits inferences about the effectiveness of true intraoperative treatment and practical applicability of frozen-section-guided therapy. A further limitation of our retrospective dataset is that the exact anastomotic extent (for example distance of the anastomosis above the esophagogastric junction or from the incisors) was not recorded in a standardized manner and thus could not be evaluated. We have attempted to mitigate this by restricting the cohort to consecutive total gastrectomy patients operated in a single high-volume center by attending surgeons using a consistent anastomosis technique. Finally, CE’s significance only after multivariable adjustment may reflect confounding or suppression effects and model instability related to subgroup sample sizes; therefore, the association should be considered exploratory and subject to prospective validation.

This retrospective, nonrandomized comparison is underpowered to exclude a clinically meaningful benefit from earlier identification or treatment. We recommend prospective studies that incorporate standardized preoperative esophageal assessment to determine whether preemptive therapy alters anastomotic outcomes.

Our findings raise practical questions about whether and how assessment of esophageal mucosal status should be incorporated into perioperative pathways for patients undergoing total gastrectomy. Practical implications for future evaluation include performing preoperative upper GI endoscopy (gastroscopy) in patients at higher baseline risk of EJ leakage (for example, BMI ≥ 25 kg/m², serum albumin < 3.5 g/dL, and/or receipt of neoadjuvant therapy) with systematic inspection of the esophagus and targeted biopsies of suspect mucosa. if CE is identified preoperatively, confirmatory testing and multidisciplinary review (surgery, oncology, infectious diseases) should determine treatment timing. Anti-fungal prophylaxis could be considered in high-risk patients. However, it should be noted that any antifungal strategy should carefully consider potential harms. Azole antifungals can cause hepatic toxicity and have important drug-drug interactions (notably with agents metabolized via CYP pathways) [31]. Hence, therapeutic protocols must include liver-function monitoring, interaction checks with neoadjuvant regimens, and microbiologic surveillance. It should be emphasized that that these points constitute a research pathway and implementation should follow prospective validation.

Conclusions

This study identified CE to be associated with EJ leakage following total gastrectomy for gastric cancer. However, prospective studies are needed to confirm causality and to assess whether earlier detection and treatment modify risk. in particular, targeted preoperative esophageal assessment and carefully designed trials of timed antifungal interventions in high-risk patients should be considered to determine whether earlier detection or treatment can reduce anastomotic complications. Additionally, hypoalbuminemia, elevated BMI, and receipt of neoadjuvant therapy were associated with increased odds of leakage. These findings underscore the importance of considering not only surgical and nutritional factors but also mucosal and microbiological conditions when evaluating anastomotic integrity. According to our findings, heightened vigilance and tailored postoperative care may be warranted in patients with CE.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

A.A and E.A drafted the primary version of the manuscript. S.A.M.Y conceptualized and designed the study. E.N and A.A gathered the data. A.A analyzed the data. S.A.M.Y and E.N revised the manuscript. All authors approved the final version of the manuscript.

Funding

No external funding was received for this study.

Data availability

The datasets generated and/or analyzed during the current study are available from the corresponding author at [amiratashi@tums.ac.ir](mailto:amiratashi@tums.ac.ir) upon reasonable request.

Declarations

Ethics approval

All items of the project were approved by the Ethics Committee of TUMS with the ethical code of IR.TUMS.SINAHOSPITAL.REC.1402.056.

Informed consent

Informed consent was obtained from all participants prior to enrollment.

Competing interests

The authors declare no competing interests.