Predictive value of portal venous phase computed tomography parameters for anastomotic stricture after D2 gastrectomy

Abstract

BACKGROUND

The most common complication in D2 gastrectomy for gastric cancer is the anastomotic stricture, with reported incidence between 3% and 15%. Appropriate tissue perfusion is essential at all time points during the inflammatory, proliferative, and remodeling phases of anastomotic healing. Portal venous phase computed tomography (CT), a study universally indicated for preoperative staging and postoperative re-evaluation, is an objective assessment of visceral hemodynamic status that may impact anastomotic healing. Yet the predictive value of preoperative CT portal venous parameters for risk of anastomotic stricture has not yet been investigated, and reliable preoperative stratification tools are urgently needed. Deep learning-based portal venous CT parameters have been particularly well-studied for liver cirrhosis and portal hypertension, but this is the first time they have been studied in predicting gastrointestinal anastomotic complications.

AIM

To prospectively explore the predictive power of preoperative enhanced CT portal venous phase quantitative parameters for anastomotic stricture after D2 gastrectomy for gastric cancer to establish a combined prediction model.

METHODS

Clinical data from 199 patients who underwent D2 gastrectomy for gastric cancer at our institution between January 2022 and June 2024 were retrospectively analyzed. Patients were categorized into stricture group (n = 23) and non-stricture group (n = 176) based on whether anastomotic stricture occurred within 12 months postoperatively. On preoperative enhanced CT portal venous phase images, main portal vein (PV) CT value (PV-HU), splenic vein CT value, hepatic parenchyma, and abdominal aorta were measured. Derived parameters including normalized PV-HU (nPV-HU) and PV-to-abdominal aorta CT value (aorta-HU) ratio were calculated. Multivariate logistic regression analysis was employed to identify independent predictive factors. Receiver operating characteristic curves were constructed to evaluate predictive performance, and a combined prediction model was developed.

RESULTS

The postoperative anastomotic stricture rate was 11.6% (23/199). The stricture group demonstrated significantly lower PV-HU, splenic vein CT value, PV-to-aorta-HU ratio, and nPV-HU compared to the non-stricture group (P < 0.05). Multivariate logistic regression analysis revealed that body mass index ≥ 25 kg/m², diabetes mellitus, total gastrectomy, 25 mm stapler diameter, decreased nPV-HU, and decreased PV-to-aorta-HU ratio were independent risk factors for anastomotic stricture (P < 0.05). For predicting anastomotic stricture, nPV-HU yielded an area under the curve (AUC) of 0.812, with an optimal cutoff value of 42.5%, sensitivity of 78.3%, and specificity of 72.7%; PV-to-aorta-HU ratio yielded an AUC of 0.768, with an optimal cutoff value of 1.45, sensitivity of 73.9%, and specificity of 71.0%. The combined prediction model incorporating clinical factors and CT parameters achieved an AUC of 0.893, with sensitivity of 87.0%, specificity of 80.1%, and negative predictive value of 97.9%.

CONCLUSION

Preoperative enhanced CT portal venous phase quantitative parameters nPV-HU and PV-to-aorta-HU ratio can effectively predict the risk of anastomotic stricture following D2 gastrectomy for gastric cancer. The prediction model combining clinical risk factors demonstrates high diagnostic performance and provides a strong imaging-based rationale for preoperative risk stratification, potentially guiding personalized perioperative management strategies.

Key Words: Gastric cancer; D2 gastrectomy; Anastomotic stricture; Enhanced computed tomography; Portal venous phase; Prediction model

Core Tip: Preoperative quantitative perfusion indicators of portal venous phase computed tomography (CT) relevant to the anastomotic healing. Increased risk of postoperative anastomotic stricture after D2 gastrectomy was significantly associated with decreased normalized portal vein (PV) CT value and reduced PV-to-aorta CT value ratio. Other independent risk factors included body mass index ≥ 25, diabetes, total gastrectomy and use of a 25-mm stapler. Among single markers, the normalized PV-HU value exhibited best predictive performance and a good clinical-imaging combined model showed an area under the curve of 0.893 and high negative predictive value (97.9%). These CT parameters provide preoperative risk stratification with the potential to assist in identifying patients who may benefit from increased surveillance or surgical management alteration. The CT-based tool is simple, non-invasive and tailored to each subject’s perioperative management in gastric cancer surgery.

INTRODUCTION

Gastric cancer is considered as one of the common malignant tumors of the digestive tract and has high incidence and mortality rates in China[1]. In this study, based on the overview of its contour distribution characteristics by histopathology and CT images in patients with advanced gastric cancer, D2 radical gastrectomy + standardized lymph node dissection has become the standard treatment for these patients[2]. However, postoperative anastomotic stricture after gastrointestinal reconstruction is a critical complication causing dramatic impairment of patients' dietary function and their life quality. According to the published literature, the incidence varies from 3% to 15%, with most of them requiring either endoscopic balloon dilation or reoperation[3]. Anastomotic stricture formation is currently believed to be linked to local tissue ischemia, inflammatory response, and fibrotic healing. In addition, adequate tissue perfusion is critical to anastomotic healing through the three canonical phases: (1) During the inflammatory phase (days 0-5), sufficient blood flow provides immune cells and clears debris; (2) During the proliferative phase (days 3-14), robust perfusion supports fibroblast migration, collagen synthesis, and angiogenesis; and (3) During the remodeling phase (weeks 2-12), sustained vascular supply is essential for collagen maturation and restoration of tissue tensile strength but methods able to predict this preoperatively have yet to be established[4].

Enhanced computed tomography (CT), which is a routine examination for preoperative staging of gastric cancer, provides tumor localization and lymph node metastasis information as well as an objective reflection of visceral hemodynamic status through portal venous phase images[5]. The perfusion characteristic of the portal venous system exhibits a close correlation with hepatic functional reserve and visceral microcirculatory status, which could directly affect anastomotic healing process[6]. Previous reports have explored portal vein (PV) CT value (PV-HU) application in the assessment of portal hypertension secondary to liver cirrhosis, but little is described regarding the relationship with gastrointestinal anastomotic healing[7]. Decreased portal venous perfusion parameters may reflect impaired visceral circulatory function, which could in turn result in decreased local blood supply at the site of the anastomosis, and increased risk of stricture.

Accordingly, the objective of this study was to investigate the relationship between portal venous phase quantitative parameters from preoperative enhanced CT and anastomotic stricture after D2 gastrectomy in gastric cancer. This study aimed to identify independent predictive factors and establish a combined prediction model by analyzing intergroup differences in main PV-HUs, SV-HUs, and their derived parameters in order to provide imaging-based evidence for early clinical identification of high-risk patients and optimization of perioperative management strategies[8].

MATERIALS AND METHODS

Study population

We conducted a retrospective review of clinical data from patients who underwent D2 gastrectomy for gastric cancer at the Affiliated Huaian No. 1 People’s Hospital of Nanjing Medical University between January 2022 and June 2024.

Inclusion criteria: (1) Pathologic confirmation of gastric adenocarcinoma at surgery; (2) D2 lymph node dissection according to standard procedure; and (3) Abdominal enhanced CT examination performed no longer than 14 days prior to operation with the imaging quality satisfying analytic standards and follow-up data available for at least 12 months postoperatively.

We excluded the following ineligible patients: (1) Preoperative neoadjuvant chemotherapy or radiotherapy; (2) Concomitant other malignant tumors; (3) Emergency surgery; (4) Severe hepatic or renal insufficiency or history of iodine contrast agent allergy; (5) CT images with severe artifacts that affected measurement; (6) Missing critical clinical or imaging data; and (7) Preoperative enhance CT examination was not performed with GE Revolution CT equipment designated for this study (this restriction on scanners was necessitated to ensure technical standardization and minimize additional inter-scanner variation in how attenuation is measured by CT, although it may limit immediate external generalizability of specific threshold values across different CT platforms).

In total, this led to the enrolment of 199 patients according to these criteria. This study was approved by the Ethics Committee of Huai’an First People’s Hospital (approval No. YX-Z-2025-063-01). Informed consent was waived because of the retrospective study design.

Diagnostic criteria for anastomotic stricture

Anastomotic stricture was defined as symptoms of feeding obstruction after anastomosis, with gastroscopy confirming the presence of a diameter < 10 mm and inability to pass a standard adult gastroscope (outer diameter approximately 9.0 mm) through the anastomosis. For anastomotic stricture to be diagnosed, both criteria need to be fulfilled at the same time. In this study, the endpoint of 12 months postoperatively was selected for outcome observation because events occurring within this time window were recorded as anastomotic stricture; however, cases developing stricture after the observation period but during follow-up are reported here separately and further discussed in the discussion section. Patients were categorized into stricture and non-stricture groups based on the presence of anastomotic stricture.

D2 gastrectomy for gastric cancer

The current standard surgical intervention for advanced gastric cancer is the D2 gastrectomy, which aims to achieve radical resection of the primary tumor and draining lymphatics. D2 lymph node dissection range according to the Japanese Gastric Cancer Association treatment guidelines. Surgical methods selected total or distal gastrectomy depending on tumor position, digestive tract reconstruction methods were: (1) Billroth I; (2) Billroth II; and (3) Roux-en-Y anastomosis. All the surgeries in this study were performed by senior attending physicians or higher level surgeons at our center who had performed over ≥ 100 gastric cancer radical resections. All anastomoses were performed with circular stapling devices (25 mm or 29 mm, depending on the conditions of patient), and type of stapler as well as technique of anastomosis were documented.

CT scanning protocol

This study uniformly employed the GE Revolution CT for abdominal enhanced scanning. During the study period, the equipment underwent daily quality control calibration and manufacturer maintenance calibration every 6 months to ensure scanning parameter stability. Scanning parameters were as follows: (1) Tube voltage 120 kV; (2) Automatic tube current modulation; (3) Slice thickness 5 mm; (4) Reconstruction slice thickness 1.25 mm; and (5) Pitch 0.984. Iohexol contrast agent (iodine concentration 350 mgI/mL) was administered at a dose of 1.5 mL/kg body weight with an injection rate of 3.5 mL/second. Threshold triggering technique was employed, with portal venous phase scanning performed 65 seconds after the abdominal aorta CT value (aorta-HU) reached 120 HU. Scanning delay time error was controlled within ± 2 seconds.

CT image analysis and quantitative parameter measurement

The following quantitative parameters were measured on portal venous phase axial images: Main PV-HU, measured by placing a circular region of interest (ROI) at the widest portion of the main PV with a standardized area of 70 mm² (allowable error ± 5 mm²); hepatic parenchyma CT value (liver-HU), measured by placing 3 ROIs in uniform areas of the right hepatic lobe (avoiding vessels, bile ducts, and lesions) and calculating the average; aorta-HU, measured by placing an ROI on the abdominal aorta at the celiac trunk level; SV-HU, measured by placing an ROI at the widest portion of the splenic vein lumen at the splenic hilum level, with ROI area adjusted according to splenic vein diameter (range 30-50 mm²), ensuring the ROI was completely within the vessel lumen without including the vessel wall.

Derived parameters were calculated as follows: (1) Hepatic portal perfusion index (PPI) = liver-HU/PV-HU; (2) PV-to-aorta-HU ratio = PV-HU/aorta-HU; and (3) Normalized PV-HU (nPV-HU) = (PV-HU - aorta-HU)/aorta-HU × 100 (expressed as percentage, reflecting the degree of PV enhancement relative to the aorta). All derived parameters were first calculated as individual patient values, then subjected to intergroup statistical analysis. Each parameter was measured 3 times and averaged for statistical analysis.

Clinical data collection

The following clinical data were collected: (1) Age; (2) Sex; (3) Body mass index (BMI); (4) Comorbidities (diabetes mellitus, hypertension); (5) Preoperative serum albumin level; (6) Tumor location (categorized as cardia/fundus, body, or antrum/pylorus); (7) Tumor tumor-node-metastasis stage (American Joint Committee on Cancer 8^th edition); (8) Surgical approach (total gastrectomy/distal gastrectomy); (9) Digestive tract reconstruction method (Billroth I/Billroth II/Roux-en-Y anastomosis); (10) Stapler model (diameter 25 mm/29 mm); (11) Postoperative complications; and (12) Postoperative adjuvant therapy.

Postoperative complications were assessed using the Clavien-Dindo classification system, recording the occurrence of grade ≥ II complications including anastomotic leakage, intra-abdominal infection, hemorrhage, pulmonary infection, and wound infection. It should be noted that anastomotic stricture, as the primary outcome measure of this study, was not included in the postoperative complications variable analysis to avoid confounding between independent and dependent variables.

Follow-up

Follow-up was conducted through outpatient visits, telephone calls, or hospital medical records. Follow-up was performed every 3 months postoperatively, with all patients followed for ≥ 12 months. The follow-up cutoff date was June 30, 2025, and the data lock date was July 15, 2025.

Statistical analysis

Data analysis was performed using SPSS 26.0 statistical software and MedCalc 20.0 software. Continuous variables were assessed for normality using the Shapiro-Wilk test. Normally distributed data were expressed as mean ± SD, with intergroup comparisons using independent samples t-test; non-normally distributed data were expressed as median (interquartile range), with intergroup comparisons using Mann-Whitney U test. Categorical variables were expressed as n (%), with intergroup comparisons using χ² test or Fisher’s exact test (when expected frequency < 5).

Variables with P < 0.10 in univariate analysis were included in multivariate binary logistic regression model, with backward stepwise selection used to identify independent predictive factors (variable elimination criterion P > 0.10). Receiver operating characteristic (ROC) curves were constructed for independent predictive factors, calculating area under the curve (AUC) with 95%CI. Optimal cutoff values were determined using the maximum Youden index method, with corresponding sensitivity, specificity, positive predictive value, and negative predictive value calculated. DeLong test was used to compare AUC differences between different CT quantitative parameters. A combined prediction model was constructed by integrating multiple independent predictive factors, with the Hosmer-Lemeshow test used to assess model goodness-of-fit (P > 0.05 indicating good model fit). All statistical tests were two-sided, with P < 0.05 considered statistically significant.

RESULTS

General characteristics

This study enrolled 199 patients following D2 gastrectomy for gastric cancer, including 128 males and 71 females, aged 38-79 years with a mean age of 61.4 ± 9.7 years. Within 12 months postoperatively, 23 patients (11.6%) developed anastomotic stricture and were assigned to the stricture group; the remaining 176 patients without anastomotic stricture were assigned to the non-stricture group. An additional 3 patients developed anastomotic stricture after 12 months postoperatively (at 14 months, 16 months, and 18 months, respectively) and were not included in the stricture group.

Comparison of clinical data between two groups

The comparison of clinical data between the two groups is presented in Table 1. Univariate analysis revealed statistically significant differences between the stricture and non-stricture groups in BMI, proportion with diabetes mellitus, preoperative serum albumin level, tumor location, surgical approach, digestive tract reconstruction method, stapler model, and postoperative complications (P < 0.05). No statistically significant difference was observed between groups in anastomotic technique (P > 0.05). Among postoperative complications, individual categories including anastomotic leakage, intra-abdominal infection, hemorrhage, pulmonary infection, and wound infection showed no statistically significant differences between groups (P > 0.05), but the overall complication rate was higher in the stricture group than in the non-stricture group (P < 0.05). No statistically significant differences were found between groups in age, sex, proportion with hypertension, tumor tumor-node-metastasis stage, or postoperative adjuvant therapy (P > 0.05).

Table 1 Comparison of clinical data between two groups, mean ± SD/n (%).

Variable	Stricture group (n = 23)	Non-stricture group (n = 176)	t/χ2/Z value	P value
Age (years)	63.2 ± 8.9	61.1 ± 9.8	1.024	0.307
Sex			0.313	0.576
Male	16 (69.6)	112 (63.6)
Female	7 (30.4)	64 (36.4)
Body mass index (kg/m2)	25.8 ± 3.2	23.4 ± 3.1	3.521	0.001
Diabetes mellitus	9 (39.1)	32 (18.2)	5.324	0.021
Hypertension	8 (34.8)	51 (29.0)	0.337	0.562
Preoperative serum albumin (g/L)	35.2 ± 4.8	38.6 ± 5.1	-3.092	0.002
Tumor location			8.762	0.013
Cardia/fundus	11 (47.8)	42 (23.9)
Body	6 (26.1)	52 (29.5)
Antrum/pylorus	6 (26.1)	82 (46.6)
Tumor-node-metastasis stage			2.156	0.541
Stage I	3 (13.0)	38 (21.6)
Stage II	8 (34.8)	62 (35.2)
Stage III	12 (52.2)	76 (43.2)
Surgical approach			6.891	0.009
Total gastrectomy	14 (60.9)	58 (33.0)
Distal gastrectomy	9 (39.1)	118 (67.0)
Digestive tract reconstruction			7.234	0.027
Billroth I	3 (13.0)	56 (31.8)
Billroth II	5 (21.7)	48 (27.3)
Roux-en-Y anastomosis	15 (65.2)	72 (40.9)
Stapler model			5.127	0.024
Diameter 25 mm	16 (69.6)	82 (46.6)
Diameter 29 mm	7 (30.4)	94 (53.4)
Anastomotic technique			1.738	0.187
End-to-end anastomosis	4 (17.4)	54 (30.7)
End-to-side anastomosis	19 (82.6)	122 (69.3)
Postoperative complications (≥ grade II)1	8 (34.8)	26 (14.8)	5.671	0.017
Anastomotic leakage	2 (8.7)	7 (4.0)		0.2682
Intra-abdominal infection	3 (13.0)	6 (3.4)		0.0542
Hemorrhage	1 (4.3)	4 (2.3)		0.4702
Pulmonary infection	2 (8.7)	8 (4.5)		0.3292
Wound infection	2 (8.7)	5 (2.8)		0.1812
Postoperative adjuvant therapy	15 (65.2)	98 (55.7)	0.758	0.384

¹The same patient may experience multiple complications simultaneously, so the sum of individual categories may exceed the total complication count.

²Fisher’s exact test was used.

Comparison of CT quantitative parameters between two groups

The comparison of portal venous phase CT quantitative parameters between the two groups is presented in Table 2. The stricture group demonstrated significantly lower PV-HU, SV-HU, PV-to-aorta-HU ratio, and nPV-HU compared to the non-stricture group (P < 0.05). PPI was higher in the stricture group than in the non-stricture group, but the difference was not statistically significant (P = 0.090). No statistically significant differences were found between groups in liver-HU or aorta-HU (P > 0.05).

Table 2 Comparison of portal venous phase computed tomography quantitative parameters between two groups, mean ± SD/median (interquartile rage).

Parameter	Stricture group (n = 23)	Non-stricture group (n = 176)	t/Z value	P value
PV-HU (HU)	128.6 ± 18.3	142.5 ± 21.6	-3.012	0.003
Liver-HU (HU)	89.4 ± 12.7	92.8 ± 14.2	-1.124	0.263
Aorta-HU (HU)	94.2 ± 15.8	92.6 ± 16.4	0.452	0.652
SV-HU (HU)	118.3 ± 16.9	132.7 ± 19.5	-3.452	0.001
Portal perfusion index	0.70 ± 0.09	0.66 ± 0.11	1.706	0.09
PV-HU/aorta-HU	1.38 ± 0.18	1.56 ± 0.22	-3.847	< 0.001
nPV-HU (%)	35.2 (28.6-42.8)	52.4 (41.6-64.3)	-4.126	< 0.001

PV-HU: Portal vein computed tomography value; Liver-HU: Hepatic parenchyma computed tomography value; Aorta-HU: Abdominal aorta computed tomography value; SV-HU: Splenic vein computed tomography value; nPV-HU: Normalized portal vein computed tomography value.

Multivariate logistic regression analysis of anastomotic stricture

Variables with P < 0.10 in univariate analysis (BMI, diabetes mellitus, preoperative serum albumin, tumor location, surgical approach, digestive tract reconstruction method, stapler model, postoperative complications, PV-HU, SV-HU, PPI, PV-to-aorta-HU ratio, nPV-HU) were included in the multivariate binary logistic regression model. Collinearity diagnostics performed before inclusion revealed severe collinearity between surgical approach and digestive tract reconstruction method (variance inflation factor > 10); therefore, only surgical approach was retained in the model. Following backward stepwise selection, preoperative serum albumin, tumor location, postoperative complications, PV-HU, SV-HU, and PPI were eliminated from the final model due to P > 0.10. Results demonstrated that BMI ≥ 25 kg/m², diabetes mellitus, total gastrectomy, 25 mm stapler diameter, decreased nPV-HU, and decreased PV-to-aorta-HU ratio were independent risk factors for anastomotic stricture (P < 0.05; Table 3).

Table 3 Multivariate logistic regression analysis of anastomotic stricture.

Variable	β	SE	Wald χ2	Odds ratio	95%CI	P value
Body mass index ≥ 25 kg/m2	1.186	0.512	5.367	3.274	1.201-8.928	0.021
Diabetes mellitus	1.082	0.498	4.721	2.951	1.112-7.832	0.03
Total gastrectomy	1.324	0.487	7.392	3.758	1.447-9.762	0.007
Stapler diameter 25 mm	1.156	0.476	5.901	3.177	1.250-8.074	0.015
nPV-HU (per 10% increase)	-0.892	0.318	7.869	0.41	0.220-0.765	0.005
PV-HU/aorta-HU (per 0.1 increase)	-0.756	0.287	6.938	0.469	0.268-0.824	0.008

Odds ratio < 1 for normalized portal vein computed tomography (CT) value and portal vein CT value/abdominal aorta CT value indicates that increased values are protective factors, meaning decreased values increase anastomotic stricture risk. aorta-HU: Abdominal aorta computed tomography value; nPV-HU: Normalized portal vein computed tomography value; PV-HU: Portal vein computed tomography value.

ROC curve analysis of CT quantitative parameters for predicting anastomotic stricture

ROC curves were constructed for CT quantitative parameters showing statistical significance in multivariate analysis (nPV-HU, PV-to-aorta-HU ratio). For predicting anastomotic stricture, nPV-HU yielded an AUC of 0.812 (95%CI: 0.726-0.898), with an optimal cutoff value of 42.5%, sensitivity of 78.3%, and specificity of 72.7%; PV-to-aorta-HU ratio yielded an AUC of 0.768 (95%CI: 0.672-0.864), with an optimal cutoff value of 1.45, sensitivity of 73.9%, and specificity of 71.0% (Figure 1, Table 4).

Figure 1 Receiver operating characteristic curves of computed tomography quantitative parameters for predicting anastomotic stricture after D2 gastrectomy for gastric cancer. It shows the receiver operating characteristic curve performance of four prediction models. The normalized portal vein (PV) computed tomography (CT) value (PV-HU) model demonstrated the best performance [area under the curve (AUC) of 0.812], followed by the splenic vein CT value model (AUC = 0.724), PV-to-abdominal aorta CT value ratio model (AUC = 0.768), and PV-HU model (AUC = 0.697). The points marked on the curves represent the optimal cutoff values for each model. The dashed line represents the reference line (AUC = 0.5), indicating no discriminative ability. All models showed AUC values significantly higher than the reference line, demonstrating good discriminative ability, with the normalized PV-HU model exhibiting the best diagnostic performance. aorta-HU: Abdominal aorta computed tomography value; AUC: Area under the curve; nPV-HU: Normalized portal vein computed tomography value; PV-HU: Portal vein computed tomography value; SV-HU: Splenic vein computed tomography value.

Table 4 Receiver operating characteristic curve analysis of computed tomography quantitative parameters for predicting anastomotic stricture.

Parameter	Area under the curve	95%CI	Optimal cutoff	Sensitivity (%)	Specificity (%)	Positive predictive value (%)	Negative predictive value (%)
nPV-HU	0.812	0.726-0.898	42.50%	78.3	72.7	27.3	96.2
PV-HU/aorta-HU	0.768	0.672-0.864	1.45	73.9	71	25	95.4
PV-HU	0.697	0.586-0.808	133.5 HU	69.6	64.8	20.5	94.2
SV-HU	0.724	0.618-0.830	124.0 HU	73.9	63.6	21	94.9

aorta-HU: Abdominal aorta computed tomography value; nPV-HU: Normalized portal vein computed tomography value; PV-HU: Portal vein computed tomography value; SV-HU: Splenic vein computed tomography value.

Comparison of AUC among different CT quantitative parameters

DeLong test was used to compare AUC values among different CT quantitative parameters. The AUC of nPV-HU was significantly higher than that of PV-HU (Z = 2.156, P = 0.031) and SV-HU (Z = 1.987, P = 0.047). No statistically significant difference was observed between the AUC of nPV-HU and PV-to-aorta-HU ratio (Z = 0.892, P = 0.372). The AUC differences between PV-to-aorta-HU ratio and PV-HU or SV-HU were not statistically significant (P > 0.05; Table 5).

Table 5 DeLong test comparison of area under the curve among different computed tomography quantitative parameters.

Comparison	ΔAUC	Z value	P value
nPV-HU vs PV-HU/aorta-HU	0.044	0.892	0.372
nPV-HU vs PV-HU	0.115	2.156	0.031
nPV-HU vs SV-HU	0.088	1.987	0.047
PV-HU/aorta-HU vs PV-HU	0.071	1.324	0.186
PV-HU/aorta-HU vs SV-HU	0.044	0.876	0.381
PV-HU vs SV-HU	-0.027	0.512	0.609

aorta-HU: Abdominal aorta computed tomography value; nPV-HU: Normalized portal vein computed tomography value; PV-HU: Portal vein computed tomography value; SV-HU: Splenic vein computed tomography value; ΔAUC: Difference in area under the curve.

Construction and evaluation of combined prediction model

A combined prediction model was constructed by integrating the 6 independent predictive factors identified in multivariate logistic regression analysis (BMI ≥ 25 kg/m², diabetes mellitus, total gastrectomy, 25 mm stapler diameter, nPV-HU, PV-to-aorta-HU ratio). The Hosmer-Lemeshow test indicated good model fit (χ² = 6.824, P = 0.556). The combined prediction model achieved an AUC of 0.893 (95%CI: 0.832-0.954), which was significantly higher than using nPV-HU alone (Z = 2.314, P = 0.021) or PV-to-aorta-HU ratio alone (Z = 2.876, P = 0.004). Using a predicted probability of 0.158 as the optimal cutoff, sensitivity was 87.0%, specificity was 80.1%, positive predictive value was 36.4%, and negative predictive value was 97.9% (Table 6).

Table 6 Comparison of diagnostic performance between combined prediction model and single computed tomography parameters.

Predictive indicator	Area under the curve	95%CI	Sensitivity (%)	Specificity (%)	Positive predictive value (%)	Negative predictive value (%)
Combined prediction model	0.893	0.832-0.954	87	80.1	36.4	97.9
nPV-HU	0.812	0.726-0.898	78.3	72.7	27.3	96.2
PV-HU/aorta-HU	0.768	0.672-0.864	73.9	71	25	95.4
Clinical factors model1	0.798	0.708-0.888	73.9	75.6	28.3	95.7

¹Clinical factors model includes four clinical factors: Body mass index ≥ 25 kg/m², diabetes mellitus, total gastrectomy, and 25 mm stapler diameter.

aorta-HU: Abdominal aorta computed tomography value; nPV-HU: Normalized portal vein computed tomography value; PV-HU: Portal vein computed tomography value.

Subgroup analysis

Subgroup analysis was performed according to surgical approach. Among patients who underwent total gastrectomy (n = 72), the stricture group (n = 14) showed significantly lower nPV-HU [33.8% (26.4%-40.2%) vs 50.6% (40.2%-62.8%), Z = -3.512,P < 0.001] and PV-to-aorta-HU ratio (1.35 ± 0.17 vs 1.54 ± 0.21, t = -3.124, P = 0.002) compared to the non-stricture group (n = 58). Among patients who underwent distal gastrectomy (n = 127), the stricture group (n = 9) also demonstrated lower nPV-HU [38.2% (30.4%-46.8%) vs 53.6% (42.8%-65.2%), Z = -2.687, P = 0.007] and PV-to-aorta-HU ratio (1.42 ± 0.19 vs 1.57 ± 0.23, t = -2.012, P = 0.046) compared to the non-stricture group (n = 118). These results suggest that the predictive value of nPV-HU and PV-to-aorta-HU ratio for anastomotic stricture is consistent across different surgical approaches (Table 7).

Table 7 Comparison of computed tomography quantitative parameters in subgroups by surgical approach, mean ± SD/median (interquartile rage).

Surgical approach	Stricture group	Non-stricture group	Statistic	P value
Total gastrectomy (n = 72)	n = 14	n = 58
nPV-HU (%)	33.8 (26.4-40.2)	50.6 (40.2-62.8)	Z = -3.512	< 0.001
PV-HU/aorta-HU	1.35 ± 0.17	1.54 ± 0.21	t = -3.124	0.002
Distal gastrectomy (n = 127)	n = 9	n = 118
nPV-HU (%)	38.2 (30.4-46.8)	53.6 (42.8-65.2)	Z =-2.687	0.007
PV-HU/aorta-HU	1.42 ± 0.19	1.57 ± 0.23	t = -2.012	0.046

aorta-HU: Abdominal aorta computed tomography value; nPV-HU: Normalized portal vein computed tomography value; PV-HU: Portal vein computed tomography value.

DISCUSSION

Anastomotic stricture after D2 gastrectomy for gastric cancer is a severe complication that deeply affects postoperative. The incidence rate of 11.6% noted in our cohort matches published literature[9]. The current clinical prevention of anastomotic stricture mostly relies on a standardized intraoperative approach and perioperative nutritional support, while preoperative risk assessment tools are still limited. This is the first systematic study that elucidates the association between rational CT portal venous phase quantitative parameters and gastrectomies-associated anastomotic stricture after surgery, revealing nPV-HU and PV-to-aorta-HU ratio as independent predictors, aiming to propose new imaging biomarkers for preoperative risk stratification.

We found that CT enhancement values of the portal venous system were always lower in the stricture group than those in non-strictured liver, which may correlate closely with visceral circulatory function status. Fundamentally, portal venous blood flow predominantly comes from the splenic vein and superior mesenteric vein; its perfusion status eminently depicts venous return from both intestine and abdominal organs[10]. Decreased presence of PV-HUs has been reported in cirrhotic portal hypertension patients, and was related to the establishment of collateral circulation and impaired hepatic synthetic function[11]. Despite not having overt liver cirrhosis, enrolled patients displayed a decrease in portal venous perfusion parameters suggesting subclinical visceral microcirculatory dysfunction. Certain investigators theorize that portal venous hemodynamic abnormalities can lead to congestion of the intestinal wall veins and interstitial edema which affect local nutrient delivery and oxygen supply at the anastomosis site, eventually ending in impaired healing[12].

In this study, nPV-HU was used as an observational marker, as it was a more effective predictor of clinical outcomes than absolute PV-HU. The better performance of nPV-HU needs some explanation. By normalizing PV enhancement to hepatic parenchymal enhancement, however, nPV-HU successfully controls for multiple sources of variability including patient-level differences in cardiac output and body composition, timing variations in contrast administration, and systemic factors that affect the regional distribution of systemic contrast. This normalization technique effectively removes the influence of the portal venous-to-hepatic perfusion gradient, which can be a more sensitive measure of splanchnic hemodynamic derangement as compared to the enhancement values in absolute terms. The ratio-based parameter has the advantage of better delineating the efficiency of portal venous inflow vs hepatic tissue perfusion and could reflect subtle changes in visceral circulatory dynamics that may be lost with absolute measurements, due to the inter-individual variability of baseline physiology and contrast pharmacokinetics. This finding supports results from hepatic imaging studies demonstrating that using the abdominal aorta as reference restores comparability of imaging-measured parameters, since it corrects interindividual variability in patterns of contrast agent metabolism and circulation time[13]. Herein nPV-HU was able to predict anastomotic stricture with AUC of 0.812 and optimal cutoff value of 42.5%, which indicated that the risk of developing anastomotic stricture is significantly greater in cases where PV relative enhancement is less than 42.5% of aortic enhancement level. The clinical utility of this cutoff is that it provides an objective and quantifiable standard of care during preoperative evaluation[14].

BMI ≥ 25 kg/m², diabetes mellitus, total gastrectomy and 25 mm stapler are independent risk factors for anastomotic stricture in patients receiving the same protocol of surgery. The adipose tissue of obese patients also shows a proinflammatory state that might impact normal-healing processes[15] and presents an increase in intra-abdominal fat accumulation, leading to increased surgical difficulty. Diabetes mellitus patients show microvascular disease and reduced ability for tissue repair, and their higher rate of post-operative anastomotic complications has been confirmed in several studies[16]. Esophagojejunal anastomosis is done at a site under more tension and with weaker local blood supply after total gastrectomy; therefore, strictures develop more frequently than distal gastrectomy[17]. In addition to that, it has also been argued that small-diameter staplers (25 mm) produce anastomoses with smaller initial luminal diameter, which are more likely to reach clinical stricture criteria with postoperative scar contracture[18].

This study built a combined prediction model by integrating CT quantitative parameters with clinical risk factors, which achieved an AUC of 0.893 and outperformed single indicators significantly. One model reached a negative predictive value of 97.9%, indicating that patients predicted as low-risk by the model have a very small probability of developing postoperative anastomotic stricture, which could be used to develop individualised follow-up strategies[19]. In high-risk patients, preoperative optimization may include increased nutritional support whereas intraoperatively, selection of larger diameter staplers or hand-sewn anastomosis to maximize anastomotic caliber may be indicated[20].

The subgroup analysis showed that regardless of total gastrectomy or distal gastrectomy, the predictive value of nPV-HU and PV-to-aorta-HU ratio on anastomotic stricture is not limited by surgical approach, suggesting that portal venous perfusion parameters can predict postoperative outcomes without being limited to surgical methods[21]. This is a clinically significant finding and indicates that CT quantitative parameters may serve as universally applicable preoperative assessment tools.

About the pathophysiological mechanisms by which defects in portal venous perfusion cause anastomotic stricture, we propose that obstruction of portal venous return may result submucosal venous congestion of gut wall with local tissue edema impairing early anastomotic healing[22]. Second, impaired function of the visceral circulatory is frequently a systemic inflammatory response and coagulation dysfunction that prevents regenerative processes in tissues[23]. Moreover, portal venous flow reduction may correspond to diminished hepatic clearance mechanisms for gut-derived endotoxins leading to a chronic inflammatory state that further enhances anastomotic fibrosis[24].

Limitations of this study include: First, being a single-centered retrospective study with small sample size (n = 23 in stricture group), statistical power might be affected[25]. Second, even prohibited CT scanning time and contrast media dose controls, but there are still significant interindividual differences in circulatory status of subjects measuring bias can affect the result[26]. Limiting analysis to one CT scanner improved technical consistency but may also limit the immediate generalizability of these specific threshold values to other CT platforms; however, the pathophysiological underpinnings should remain valid across different scanners. Third, this study did not take into account surgical-related factors, such as intraoperative anastomotic tension and stapler firing integrity; stricture is defined by the distance between them[27]. And finally, the stricture with onset after 12 months could be missed[28]. A prospective multicenter study is needed to validate the proposed cutoff values in varied patient populations and CT platforms. More direct mechanisms could be established by directly correlating preoperative CT parameters with intraoperative tissue perfusion measurements using indocyanine green fluorescence angiography or laser Doppler flowmetry in future translational research. Moreover, studies correlating these imaging biomarkers with serological markers of endothelial dysfunction (e.g., von Willebrand factor, soluble thrombomodulin), systemic inflammation (interleukin-6, C-reactive protein), and profibrotic activity (transforming growth factor-β, procollagen peptides) would inform comprehensive multidimensional risk prediction models. In a similar vein, subsequent investigations should assess whether preoperative optimization strategies aimed at modifying the identified hemodynamic derangements – including fluid management protocols, vasoactive agents, or nutritional interventions – can affect stricture risk in high-risk patients determined by CT parameters.

CONCLUSION

In summary, the new quantitative indexes of preoperative enhanced CT portal venous phase nPV-HU and PV-to-aorta-HU ratio can predict anastomotic stricture after postoperative D2 gastrectomy for gastric cancer. Combined prediction model with clinical risk factors has high diagnostic performance. This study has potentially important implications for clinical preoperative risk assessment by providing a straightforward, non-invasive imaging tool to identify patients at high-risk prior to surgery in order to optimize perioperative management.