Estimation of pancreatic histology and likelihood of postoperative pancreatic fistula using extracellular volume fraction from contrast-enhanced computed tomography

Abstract

BACKGROUND

Pancreatic fibrosis, which decreases risk of postoperative pancreatic fistula (POPF), can be estimated using extracellular volume fraction (ECVf).

AIM

To investigate the correlation between ECVf and pancreatic histology, as well as the usefulness of ECVf in predicting POPF.

METHODS

In 71 patients who underwent pancreatic resection, we caluculated pancreatic ECVf by comparing absolute enhancements of the pancreas and aorta between pre-contrast and equilibrium phases. Areas of fibrosis, fat, acini, and islets were calculated based on resection specimens.

RESULTS

ECVf correlated with fibrosis (r = 0.724; P < 0.001) and negatively correlated with acini (r = -0.510; P < 0.001). Among 48 patients who underwent pancreatoduodenectomy, 21 developed POPF. Main pancreatic duct diameter ≤ 2 mm and ECVf < 36% were selected as risk factors by multivariate analysis [respective odds ratios (OR) and P values, 4.26 and P = 0.048; OR = 11.07 and P = 0.036]. Using these factors as a risk score (0-2 points), POPF occurred in 0%, 50%, and 70% of patients with 0, 1, and 2 points, respectively.

CONCLUSION

ECVf is useful in predicting acinar loss and pancreatic fibrosis, and ECVf < 36% may be a risk factor for POPF.

Key Words: Pancreas; Extracellular volume fraction; Fibrosis; Acini; Pancreatic resection; Pancreatoduodenectomy; Postoperative pancreatic fistula

Core Tip: This study evaluated the utility of extracellular volume fraction (ECV), a computed tomography (CT)-based quantitative marker, in estimating pancreatic tissue composition and predicting postoperative pancreatic fistula (POPF). ECV was strongly correlated with histological fibrosis and inversely with acinar cell content. In patients undergoing pancreatoduodenectomy, low ECV (< 36%) and small main pancreatic duct diameter (≤ 2 mm) were identified as independent risk factors for POPF. These findings suggest that CT-derived ECV can noninvasively reflect pancreatic histology and may serve as a valuable preoperative imaging biomarker to stratify the risk of POPF.

INTRODUCTION

The mortality rate following pancreaticoduodenectomy (PD) has decreased, now reported to be approximately 2%. However, the complication rate remains high, at about 30%[1]. The most severe complication is postoperative pancreatic fistula (POPF), which occurs in 5%-26% of patients undergoing PD[2]. POPF after resection has been classified by the International Study Group on Pancreatic Fistula Definition[3]. Clinically significant POPF requiring intervention, classified as Grade B or Grade C, is termed as clinically relevant POPF, which is the usual context for POPF. Various risk factors for POPF have been investigated. Pancreatic texture has long been recognized as a key risk factor for POPF following PD, and various risk scores incorporating texture and main pancreatic duct (MPD) diameter have been proposed[4-7]. In 2023, Schuh et al[8] developed a simpler scoring system based on only 2 factors: Pancreatic texture and MPD diameter, with soft texture and MPD diameter ≤ 3 mm representing the highest-risk conditions. Moreover, pancreatic texture has been shown to correlate with histologic fibrosis, with harder pancreas associated with more advanced fibrosis[9-11].

Extracellular volume (ECV) comprises the extravascular-extracellular and intravascular compartments. The ECV fraction (ECVf) can be calculated using contrast-enhanced computed tomography (CE-CT) or magnetic resonance imaging (MRI) and reflects tissue characteristics such as fibrosis, edema, or structural remodeling, which are critical in diagnosing and assessing severity of various diseases[12,13]. The correlation between ECVf and fibrosis was first demonstrated by cardiologists as a prognostic indicator in heart disease[12,13], and later confirmed by hepatologists in studies of liver fibrosis[14-16]. In the pancreas, several studies have shown that ECVf correlates with both imaging findings and histologic fibrosis[17-22]. Among these, Sofue et al[18] and Zhu et al[21] specifically validated the association between ECVf and fibrosis, as well as its predictive value for POPF.

Texture, a subjectively assessed tissue property, is difficult to assess objectively before surgery. A preoperative imaging finding that could predict texture and act as a proxy risk factor for POPF would be valuable. Fibrosis has been shown to correlate with pancreatic texture, and ECVf has been reported to correlate with fibrosis. Based on this relationship, we hypothesized that ECVf might serve as a surrogate for pancreatic texture and a potential risk factor for POPF. However, studies investigating ECVf in the pancreas are limited, and most previous work has focused solely on its correlation with fibrosis. Little is known about how ECVf reflects other histologic components of the pancreas. To address this, we first examined how ECVf corresponds to various histologic features, including but not limited to fibrosis. We then investigated whether ECVf differs between soft and hard textures, and whether ECVf is associated with the risk of clinically relevant POPF following PD. Finally, we explored the possibility that ECVf could be used in place of intraoperative texture assessment to construct imageng-based indicator for POPF prediction.

MATERIALS AND METHODS

Study population

This retrospective study enrolled 178 consecutive patients for whom pancreatectomy was performed in our hospital between January 2018 and June 2024. Excluded patients were those who underwent total pancreatectomy; lacked precontrast images; had a different tube voltage or slice thickness between precontrast CT and CE-CT; were unsuited to measurement of CT values in regions of interest (ROI) because of severe pancreatic atrophy (thickness < 10 mm) on transverse CT images; and for whom no pancreatic stump sample was available. Ultimately, 71 patients were enrolled in this study (Figure 1). All were included in analysis of the relationship between ECVf, texture, and histology. For the analysis of POPF risk factors, we focused exclusively on the 48 patients who underwent PD, as most established POPF risk models and scoring systems have been developed and validated in PD cases[4-8]. Among these patients, 21 (44%) developed POPF and 27 (56%) did not. Cases with incomplete imaging or pathological data were excluded during the screening process. As a result, no missing data were present in the final analytic cohort. This study was approved by our Institutional Review Board (2023-290-A). Consent followed an opt-out method.

Figure 1 Flow chart of study population. PD: Pancreaticoduodenectomy; DP: Distal pancreatectomy; MP: Middle pancreatectomy; POPF: Postoperative pancreatic fistula; BL: Biochemical leak.

CT acquisition

CT scans of the pancreas were acquired using 64-row CT scanners (Aquilion 64; Toshiba Medical Systems, Otawara, Tochigi, Japan). After precontrast scans were acquired, contrast-enhanced scans were obtained at 180 s (equilibrium phase) after contrast-agent administration. The iodinated nonionic contrast agent (Iopromide, 370 mg I/mL; Bayer Healthcare, Leverkusen, North Rhine-Westphalia, Germany) was administered via an antecubital vein using a mechanical power injector at a dose of 550 mg I/kg of body weight given over a fixed interval of 30 second. All acquisitions were performed using the following parameters: Tube voltage 100 kVp, with the tube current automatically modulated (Auto Exposure Control, GE Healthcare, Chicago, Illinois, United States) during CT displaying the anatomic ROI. The detector configuration included 64-detector rows with 0.625 mm detector element width. Table increment was 20.62 mm/rotation; acquisition matrix, 512 × 512; gantry rotation speed, 0.4 second; and maximum allowable tube current, 700 mA. CT data were reconstructed to create 5 mm-thick axial sections with a standard soft-tissue kernel and a preset noise index of 8.00 HU for a 5-mm slice thickness.

Image analysis

Quantitative image analysis was performed by consensus between one radiologist (Yuki Tashiro) and one gastrointestinal surgeon (Yuki Takahashi), with both blinded to patients’ clinical data and pathologic findings. Preoperative CT images were used to measure CT values of the pancreatic parenchyma overlying the portal vein, reflecting the pancreatic resection line, according to ROI. Measurements were performed in both precontrast and equilibrium phases, carefully avoiding tumors, blood vessels, and pancreatic ducts. Additionally, CT values of the abdominal aorta (Ao) were measured using ROI in both precontrast and equilibrium phases. Each CT value was measured 3 times, and the average value was used. ROI sizes were not strictly standardized across patients, as individual variation in pancreatic thickness and the need to avoid adjacent vascular structures required flexible placement. In each case, the largest feasible ROI was used within the target region to ensure measurement reliability. The median ROI areas (interquartile ranges) were as follows: Pancreas (precontrast), 13.3 mm² (8.8-18.8 mm²); pancreas (equilibrium phase), 14.1 mm² (9.5-21.4 mm²); Ao (precontrast), 39.3 mm² (25.8-57.7 mm²); and Ao (equilibrium phase), 38.8 mm² (25.7-55.7 mm²). All image analyses were performed under the supervision of a board-certified radiologist (Yuki Tashiro), ensuring technical consistency. Although inter- and intraobserver variability was not formally assessed, we acknowledge this as a limitation of the present study. The pancreatic CT-ECV fraction was calculated using the following formula: ECV fraction (%) = [100-hematocrit (%)] × (ΔHUpancreas/ΔHUAo), where ΔHUpancreas and ΔHUAo represent absolute enhancement as the CT values in equilibrium-phase images minus CT values in precontrast images of the pancreatic parenchyma and abdominal Ao, respectively (Figure 2). The thickness of the pancreas and MPD diameter were also measured over the portal vein, near the pancreatic resection line.

Figure 2 Computed tomography from a 51-year-old male with pancreatic neuroendocrine neoplasm. A: Represents the precontrast phase; mean computed tomography (CT) value of the pancreas (ellipse), 45.2 HU and mean CT value of the aorta (circle), 43.6 HU; B: Represents the equilibrium phase, with the mean CT value of the pancreas (ellipse) being 78.3 HU and the mean CT value of the aorta (circle) being 125 HU. Extracellular volume fraction in this case was calculated to be 23.3%.

Histologic analysis

Slides were prepared from formalin-fixed, paraffin-embedded 5-μm-thick sections from resected pancreatic specimens. Serial sections of specimens within 1 cm from the pancreatic resection margin were stained with hematoxylin-eosin (HE) (Figure 3A) and Masson trichrome (Figure 3B), and were immunostained for BCL-10 (333.1) (Figure 3C) and Insulin (Figure 3D). BCL-10 was used to visualize acinar cells, as it is a known cytoplasmic marker of pancreatic acinar differentiation[23], while insulin staining was employed to identify islets. These sections were used to quantitively evaluate features of pancreatic parenchyma such as fat, fibrous tissue, acinar cells, and islets, respectively. The use of Masson trichrome and immunostaining facilitated accurate delineation of fibrous tissue, acinar cells, and islets, thereby enabling more precise area measurement. Fat was identified on HE sections as unstained or pale areas, allowing clear distinction from other tissues without additional stains. For quantitative evaluation, areas of target tissues were calculated using the ImageJ software program, version 1.54 g (National Institutes of Health, Bethesda, MA, United States). The area was measured in 3 randomly selected fields under a 10 × objective lens (1.466 mm² per field); the mean value was used. All histologic evaluations were carried out by 2 pathologists (Akihiro Nakamura and So Murai) acting by consensus; they were blinded to patients’ clinical data and imaging findings. So Murai is a board-certified pathologist.

Figure 3 Stained sections from specimens within 1 cm of the pancreatic resection margin. A: Hematoxylin-eosin staining, 10 × objective lens; B: Masson trichrome staining, 10 × objective lens; C: BCL-10 (333.1) staining, 10 × objective lens; D: Insulin staining, 10 × objective lens.

Surgical procedure and postoperative management

All procedures were performed by 2 hepatobiliary-pancreatic surgeons (Kenichi Matsuo and Kuniya Tanaka) or under their direct supervision. Surgical procedures included PD in 48 cases, distal pancreatectomy (DP) in 22, and middle pancreatectomy (MP) in 1. All PDs were subtotal stomach-preserving pancreaticoduodenectomies performed according to a previously reported standardized technique with a modified Child’s method. Pancreatojejunostomies used a double-layer duct-to-mucosa, end-to-side anastomosis with an external or internal stent tube[24]. Two or three multichannel silicone drains were placed near the pancreatic and biliary anastomoses[25]. For DP, pancreatic transection was performed using a stapling device, with splenectomy performed in all cases. A multichannel silicone drain was placed near the pancreatic stump and the left subphrenic space. The single MP case was performed for a NET G1 in the pancreatic body. The pancreas was transected over the superior mesenteric vein using a stapler, as in DP. Additionally, the pancreas was transected caudally to the tumor, partially resecting the pancreas. The pancreatic tail was reconstructed with a double-layer duct-to-mucosa end-to-side anastomosis using an external stent tube. A multichannel silicone drain was placed near the pancreatojejunostomy site and the foramen of Winslow. Pancreatic texture was assessed intraoperatively by the surgeons through visual inspection and manual palpation of the remnant pancreas. However, as there is no universally accepted definition of “soft” or “hard” pancreas, texture assessment remains subjective and relies entirely on the surgeon’s intraoperative judgment, as reported in previous studies[9,10,25].

Based on the definition of the International Study Group of Pancreatic Surgery[3], POPF grade B and grade C were classified as the POPF group, while the remaining cases were categorized as the non-POPF group. Amylase concentrations in drain fluid were measured daily until the third day after surgery. CE-CT was performed on postoperative day 4. If no POPF was detected and no peripancreatic fluid collection was observed on CE-CT, the drain was removed. Octreotide was not routinely administered.

Statistical analysis

All continuous variables are presented as medians and interquartile ranges; all categorical variables are presented as numbers and percentages. The Wilcoxon test was used to compare continuous variables as some variables were not normally distributed and we opted to apply a consistent analytical approach across all variables. Categorical variables were compared between groups using Fisher’s exact test, which is appropriate for small sample sizes. A P value less than 0.05 was considered statistically significant. Correlations between ECVf and histologic findings were analyzed by using Spearman correlation analysis. Multivariate logistic regression analysis was performed to identify risk factors POPF. Variables used in the multivariate analysis were selected based on a P value of < 0.05 in univariate analyses, combined with a stepwise selection method to refine the final model. Continuous variables were dichotomized based on cut-off values determined using receiver operating characteristic (ROC) curves before inclusion in the multivariate analysis. Statistical analyses were performed using JMP Pro 17.0 (SAS Institute, Cary, NC, United States). As this was a retrospective study, no formal sample size calculation was performed. The final sample size of 71 patients was determined based on the inclusion period and availability of appropriate imaging and histological data. We acknowledge that this may limit the statistical power, particularly for multivariate modeling, and this limitation is discussed accordingly.

RESULTS

Characteristics of all patients are shown in Table 1. Correlations between ECVf and histologic findings are shown in Figure 4. Acini showed a moderate negative correlation with ECVf (r = -0.510, P < 0.001; Figure 4A). Fibrosis exhibited a strong positive correlation with ECVf (r = 0.724, P < 0.001; Figure 4B). In contrast, islets (r = 0.228, P = 0.056; Figure 4C) and fat (r = 0.075, P = 0.534; Figure 4D) showed no significant correlation with ECVf.

Figure 4 Correlations between extracellular volume fraction and histologic findings. A: Acini showed a moderate negative correlation with extracellular volume fraction (ECVf) (r = -0.510; P < 0.001); B: Fibrosis showed a strong positive correlation with ECVf (r = 0.724; P < 0.001); C: Islets showed no significant correlation with ECVf (r = 0.228; P = 0.056); D: Fat showed no significant correlation with ECVf (r = 0.075; P = 0.534).

Table 1 Patient characteristics, n (%)/median (25^th-75^th percentiles).

Patient background	All (n = 71)
Age (years)	72 (62-78)
Sex
Male	38 (54)
Female	33 (46)
BMI	23.1 (20.4-25.1)
Indication
PDAC	29 (41)
Vater Ca	10 (14)
IPMN	10 (14)
Bile duct Ca	7 (10)
PanNEN	6 (8)
MCN	2 (3)
Other	7 (10)
Blood test findings
Hemoglobin (g/dL)	13.3 (12.2-14.0)
Hematocrit (%)	40.1 (37.2- 43.0)
Platelet (× 104/µL)	24.0 (19.1-28.4)
Creatinine (mg/dL)	0.68 (0.57-0.81)
Albmin (g/dL)	4.2 (4.0-4.4)
Prognostic Nutritional Index	49.0 (46.0-53.1)
Image findings
MPD diameter (mm)	3 (2-4)
Thickess of pancreas (mm)	13 (11-15)
CT value on precontrast (HU)	41.0 (38.1-44.4)
CT value on equilibrium phase (HU)	102.4 (92.5-113.4)
ECVf (%)	32.0 (27.1-36.7)
Operative and postoperative findings
Blood loss (mL)	370 (200-616)
Operative time (min)	590 (471-675)
Texture
Soft	51 (72)
Hard	20 (28)
Day 3 drain amylase (U/L)	304 (93-616)
Histologic findings
Area of acini (mm2)	0.879 (0.729-1.027)
Area of islets (mm2)	0.027 (0.019-0.037)
Area of fibrosis (mm2)	0.027 (0.013-0.092)
Area of fat (mm2)	0.030 (0.007-0.073)

BMI: Body mass index; PDAC: Pancreatic ductal adenocarcinoma; Vater Ca: Carcinoma of papilla of Vater; IPMN: Intraductal papillary mucinous neoplasm; Bile duct Ca: Bile duct cancer; Pan NEN: Pancreatic neuroendocrine neoplasm; MCN: Mucinous cystic neoplasm; MPD: Main pancreatic duct; ECVf: Extracellular volume fraction.

In comparing imaging and histologic findings between cases with hard and soft pancreatic textures (Table 2), several significant differences were observed. The MPD diameter was greater in the hard-texture group than in the soft-texture group (4 mm vs 2 mm; P = 0.019). The CT value on the equilibrium phase (CE-CT value) and ECVf were higher in the hard-texture group than in the soft-texture group (113.5 HU vs 99.5 HU; P = 0.001, 39.9% vs 30.3%; P < 0.001). The area of fibrosis was larger in the hard-texture group than in the soft-texture group (0.100 mm²vs 0.019 mm²; P < 0.001).

Table 2 Hard-texture vs soft-texture in imaging and histologic findings, median (25^th-75^th percentiles).

Patient background	Hard (n = 20)	Soft (n = 51)	P value
Imaging findings
MPD diameter (mm)	4 (2-5)	2 (2-3)	0.019
Thickess of pancreas (mm)	14 (12-15)	12 (10-15)	0.274
CT value in precontrast (HU)	40.0 (38.1-42.2)	41.7 (38.1-44.7)	0.140
CT value in equilibrium phase (HU)	113.5 (101.5-124.5)	99.5 (88.9-107.0)	0.001
ECVf (%)	39.9 (36.2-45.4)	30.3 (26.4-34.4)	< 0.001
Histological findings
Area of acini (mm2)	0.778 (0.514-1.027)	0.880 (0.805-1.027)	0.195
Area of islets (mm2)	0.028 (0.021-0.044)	0.027 (0.019-0.035)	0.222
Area of fibrosis (mm2)	0.100 (0.024-0.174)	0.019 (0.011-0.040)	< 0.001
Area of fat (mm2)	0.018 (0.005-0.070)	0.032 (0.008-0.073)	0.374

MPD: Main pancreatic duct; CT: Computed tomography; ECVf: Extracellular volume fraction.

Univariate analysis of POPF risk factors (Table 3) identified several parameters that were significantly different between non-POPF and POPF groups. MPD diameter was significantly smaller in POPF group than in non-POPF group (4 mm vs 2 mm; P < 0.001; Figure 5A). The CT value at equilibrium phase and ECVf were lower in POPF group (110.5 HU vs 94.3 HU, P = 0.001; Figure 5B, 37.2% vs 30.5%, P = 0.003; Figure 5C). In addition, amylase concentrations in drain fluid on postoperative day 3 were significantly higher in POPF group than in non-POPF group (154 U/L vs 612 U/L; P < 0.001). Histologic findings, including acini (Figure 5D) and fibrosis (Figure 5E), showed no significant differences.

Figure 5 Comparison between non-postoperative pancreatic fistula group and postoperative pancreatic fistula group in the imaging and histopathological findings. A: Main pancreatic duct diameter (4 mm vs 2 mm; P < 0.001); B: Computed tomography value in equilibrium phase (110.5 HU vs 94.3 HU; P = 0.001); C: Extracellular volume fraction (37.2% vs 30.5%; P = 0.003); D: Area of acini (0.802 mm²vs 0.880 mm²; P = 0.127); E: Area of fibrosis (0.041 mm² vs 0.026 mm²; P = 0.110). ^aP < 0.001; ^bP < 0.01. POPF: Postoperative pancreatic fistula; MPD: Main pancreatic duct; CT: Computed tomography; ECVf: Extracellular volume fraction.

Table 3 Patient characteristics of pancreaticoduodenectomy and univariate analysis of risk factors for postoperative pancreatic fistula, n (%)/median (25^th-75^th percentiles).

Patient background	All patients (n = 48)	non-POPF (n = 27)	POPF (n = 21)	P value
Age (years)	73 (64-78)	74 (66-79)	73 (63-77)	0.176
Sex				0.082
Male	27 (56)	12 (44)	15 (71)
Female	21 (44)	15 (56)	6 (29)
BMI	23.1 (20.3-24.8)	22.2(19.7-24.1)	23.7 (21.1-25.1)	0.154
Indication				0.382
PDAC	20 (42)	13 (48)	7 (33)
Non-PDAC	28 (58)	14 (52)	14 (67)
Blood test findings
Hemoglobin (g/dL)	13.3 (12.2-14.2)	13 (11.6-13.9)	13.9 (12.5-14.7)	0.058
Hematocrit (%)	40.1 (37.4-43.0)	39.5 (37.1-42.3)	41.0 (38.7-43.3)	0.216
Platelet (× 104/µL)	23.8 (19.5-28.6)	23.6 (17.9-28.5)	24.0 (20.8-31.3)	0.430
Creatinine (mg/dL)	0.66 (0.57-0.76)	0.64 (0.57-0.75)	0.68 (0.62-0.83)	0.266
Albmin (g/dL)	4.2 (3.8-4.3)	4 (3.7-4.2)	4.2 (4.1-4.4)	0.063
Prognostic nutritional index	48.1 (46.0-53.8)	47.1 (45.6-50.8)	51.0 (47.2-54.9)	0.099
Image findings
MPD diameter (mm)	3 (2-5)	4 (3-5)	2 (2-3)	0.001
Thickess of pancreas (mm)	13.5 (11-15)	13 (11-15)	14 (11.5-16)	0.530
CT value on precontrast (HU)	41.0 (37.8-44.2)	40.8 (37.7-42.7)	42.7 (37.2-46.0)	0.304
CT value on equilibrium phase (HU)	102.7 (92.7-116.2)	110.5 (100.9-125.2)	94.3 (88.3-103.9)	0.001
ECVf (%)	31.5 (27.3-39.1)	37.2 (28.6-46.1)	30.5 (26.2-32.5)	0.003
Operative and postoperative findings
Blood loss (mL)	462 (293-670)	440 (200-600)	540 (380-683)	0.053
Operative time (minute)	643 (574-690)	625 (566-680)	661 (597-734)	0.424
Texture				0.060
Soft	34 (71)	16 (59)	18 (86)
Hard	14 (29)	11 (41)	3 (14)
Day 3 drain amylase (U/L)	279 (82-638)	154 (24-373)	612 (261-2665)	< 0.001
Histologic findings
Area of acini (mm2)	0.847 (0.659-1.018)	0.802 (0.424-1.013)	0.880 (0.767-1.023)	0.127
Area of islets (mm2)	0.030 (0.019-0.037)	0.033 (0.019-0.041)	0.027 (0.018-0.036)	0.318
Area of fibrosis (mm2)	0.030 (0.014-0.122)	0.041 (0.017-0.191)	0.026 (0.011-0.081)	0.110
Area of fat (mm2)	0.026 (0.004-0.072)	0.014 (0.002-0.065)	0.035 (0.007-0.076)	0.232

POPF: Postoperative pancreatic fistula; BMI: Body mass index; PDAC: Pancreatic ductal adenocarcinoma; MPD: Main pancreatic duct; CT: Computed tomography; ECVf: Extracellular volume fraction.

Among variables that showed significant differences in the univariate analysis, MPD diameter, the CE-CT value, and ECVf, which can be measured preoperatively, were selected for further analysis. ROC curves were constructed to determine cut-off values for each parameter. These values were as follows: MPD diameter, 2 mm (AUC = 0.786), CE-CT value, 105 HU (AUC = 0.787), and ECVf, 36% (AUC = 0.750). These parameters were dichotomized based on respective cut-off values. Then, using a stepwise method, variables were narrowed down to MPD diameter and ECVf for inclusion in a multivariate analysis. Results of the multivariate analysis are shown in Table 4. MPD diameter ≤ 2 mm and ECVf < 36% were identified as independent risk factors for POPF [respectively, odds ratio (OR) 4.26, 95%CI: 1.01–17.94, P = 0.048, and OR 11.07, 95%CI 1.17-104.76, P = 0.036]. Among cases with MPD ≤ 2 mm, 71% (15/21) developed POPF, while 61% (20/33) of cases with ECVf < 36% had POPF. Numbers of these 2 factors (MPD ≤ 2 mm and ECVf < 36%) present in each case was categorized into 3 groups representing 0, 1, or 2 points to consider the prevalence of POPF (Table 5). Occurrence of POPF was 0% in the 0-point group, 50% in the 1-point group, and 70% in the 2-point group, representing a stepwise increase. Using a cutoff score of 2 points to predict POPF, the sensitivity was 66.7%, specificity 77.8%, positive predictive value 70.0%, and negative predictive value 75.0%.

Table 4 Multivariate analysis of risk factors for postoperative pancreatic fistula.

	Odds ratio (95%CI)	P value
MPD diameter ≤ 2 mm	4.26 (1.01-17.94)	0.048
ECVf < 36%	11.07 (1.17-104.76)	0.036

MPD: Main pancreatic duct; ECVf: Extracellular volume fraction.

Table 5 Computed tomography risk score for postoperative pancreatic fistula.

Score	Non-POPF	POPF	Rates	P value
CT risk score1				< 0.001
0	14	0	0%
1	7	7	50%
2	6	14	70%

¹Number of cases meeting two criteria. Main pancreatic duct diameter ≤ 2 mm and extracellular volume fraction < 36%. POPF: Postoperative pancreatic fistula; CT: Computed tomography.

DISCUSSION

In this study we examined the relationship between ECVf and histology, the correlation between pancreatic texture, imaging, and histologic findings, and risk factors for POPF. The results demonstrated a strong correlation between ECVf and fibrosis (r = 0.724; P < 0.001), consistent with previous reports[7-11]. Notably, our findings extend prior research by identifying a negative correlation between ECVf and acinar cells (r = -0.510; P < 0.001). Acinar cell loss and fibrosis progression are largely driven by pancreatic stellate cells, which are activated by factors such as inflammation, alcohol, and hypoxia. These cells then produce extracellular matrix, promoting fibrotic remodeling[26-28]. Since acinar cell loss and progression of fibrosis occur together, one could reasonably expect that ECVf, which correlates strongly with fibrosis, would show a negative correlation with acinar cells. On the other hand, ECVf does not correlate with either islets or fat. Even in conditions such as chronic pancreatitis, which includes progressive pancreatic fibrosis, islets are relatively preserved[26]. As for fat, Harrel et al[29] reported no clear relationship between fat and fibrosis. Very likely, then, islets and fat, which do not correlate strongly with fibrosis, do not correlate with ECVf. Given its strong correlation with fibrosis and negative correlation with acinar cells, ECVf may also serve as a useful imaging biomarker for chronic pancreatitis. Since evaluating fibrosis in chronic pancreatitis remains clinically difficult[27], ECVf may offer a promising non-invasive and objective alternative[28,29].

In our analysis of texture and histologic findings, the hard-texture group showed significant progression of fibrosis, while no significant differences were observed in acinar cells, islets, or fat. Nahm et al[30] found that hard pancreatic texture correlated with increased fibrosis and decreased acinar tissue in HE-stained sections from 61 pancreatectomy cases. In estimating acinar cells, Nahm et al[30] appear to allocate all non-fibrotic, non-fatty areas to acinar cells. In contrast, we employed BCL-10 immunostaining, a specific cytoplasmic marker of pancreatic acinar cells, allowing for precise identification. This enabled us to clearly differentiate acinar cells from adjacent islets and other non-acinar components, which were thereby excluded from quantification. This methodological distinction may explain the differences in acinar content reported between our approach and that of Nahm et al[30].

In our analysis of texture and imaging findings, the hard-pancreas group showed higher ECVf values, reflecting progression of fibrosis. Additionally, CE-CT values were also higher in the hard-pancreas group. In fibrotic pancreases, contrast enhancement has been reported to gradually increase and decrease more slowly, leading to higher CT values in the late contrast phase compared to cases with normal texture[31]. Furthermore, MPD was dilated in the hard-pancreas group, which is consistent with findings reported by Sugimoto et al[10]. These results suggest that texture, which is assessed subjectively during surgery, accurately reflects histologically demonstrable fibrosis, and that ECVf can serve as an objective preoperative predictor of pancreatic texture.

In our study, small MPD, low CE-CT values, and low ECVf were identified as potential risk factors for POPF by univariate analysis. Further, MPD diameter ≤ 2 mm and ECVf < 36% were found to be independent risk factors in multivariate analysis. Notably, ECVf < 36% had an OR of 11.07, supporting the findings of Sofue et al[18] who also demonstrated usefulness of ECVf in predicting POPF risk. Although stepwise selection in multivariate modeling may pose a risk of overfitting, we conducted a bootstrap analysis with 1000 replications to assess the stability of the selected predictors. The bootstrapped OR and 95%CI were consistent with the original estimates, suggesting that the final model is robust and generalizable. The association between small MPD and increased POPF risk has been highlighted in previous studies[21,32,33]. This could reflect technical challenges in pancreatic duct anastomosis, increased risk of pancreatic juice stasis leading to pancreatitis, and a greater likelihood of high intraductal pressure[25]. Regarding equilibrium-phase CT values, which showed a significant difference in univariate analysis, Sofue et al[18] also reported significantly lower values in POPF group (91.3 HU vs 82.4 HU; P = 0.003), consistent with our findings. This result likely reflects degree of fibrosis, resembling the comparison between hard and soft pancreas.

Several studies have explored histologic risk factors for POPF. Early reports in 2010 and preceding years identified pancreatic fat infiltration as a risk factor[34-36]. Later studies focused on the paucity of fibrous tissue as a POPF risk factor[22,25,37-39]. Recently attention has turned to acinar cells, with studies by Umezaki et al[40] and Partelli et al[41] showing high acinar cell content as a significant POPF risk factor; the latter study specifically stated that acinar content exceeding 60% increased risk. In the present study we did not find significant differences in acini, islets, fibrosis, or fatty tissue between POPF and non-POPF groups. Although the POPF group tended to show larger acinar areas (0.802 mm² vs 0.880 mm²; P = 0.127) and smaller fibrotic area (0.041 mm² vs 0.026 mm²; P = 0.110), these differences did not reach statistical significance. While these observations are inconclusive, future studies with larger cohorts may help clarify whether these patterns reflect meaningful histologic differences. In comparing ECVf with histologic findings, we found that an ECVf cutoff value of 36% approximately corresponded to an acinar area of 0.8 mm² (55%) and a fibrosis area of 0.1 mm² (7%) along a line of regression. Considering that Partelli et al[41] likely included islet content and vasculature in their acinar content measurements, our result likely seems consistent with theirs. These findings suggest that while individual histologic parameters did not reach statistical significance, likely due to limited sample size, ECVf may serve as a more sensitive and integrative marker of histologic changes relevant to POPF risk.

In our study, MPD ≤ 2 mm and ECVf < 36% were identified as risk factors for POPF. Incidence of POPF in cases with ECVf < 36% was 61%, but only 53% in those with soft texture. This suggests that ECVf < 36% is a reliable indicator of POPF risk, like texture. To evaluate a risk score based on 2 CT factors, MPD ≤ 2 mm and ECVf < 36%, we used a score with 0, 1, or 2 points. In the 0-point group, no cases of POPF were found; in the 1-point group, 50% developed POPF; and in the 2-point group, 70% developed POPF. These results suggest that a score of 2 points may indicate a high risk of POPF. Although the scoring system is simple, it achieved moderate-to-good predictive performance, with sensitivity and specificity values of 66.7% and 77.8%, respectively. These findings suggest that the scoring system may be useful for assessing POPF risk in clinical practice.

To our knowledge, this is the first study to explore the potential of a POPF risk assessment approach based solely on CT-derived imaging features, particularly ECVf and MPD diameter, without relying on intraoperative findings. Moreover, while previous studies have mainly focused on the correlation between ECVf and pancreatic fibrosis, our study is also novel in demonstrating the relationship between ECVf and other histologic components such as acinar cells, islets, and fat. While further validation is needed, this imaging-based method represents a novel step toward objective and preoperative risk stratification that could support surgical decision-making. In particular, the observed OR of 11.07 for ECVf < 36% indicates a substantially elevated risk of POPF. This value is notably higher than previously reported OR for well-established predictors such as soft pancreatic texture (OR = 4.24, 95%CI: 3.67–4.89) and MPD diameter ≤ 3 mm (OR = 3.66, 95%CI: 2.62–5.12)[8]. Such a strong association may have practical implications, such as prompting the use of modified anastomotic techniques, prophylactic administration of Octreotide, or enhanced postoperative monitoring in high-risk patients. As this information can be obtained preoperatively, it also has potential utility in the informed consent process, allowing for better communication with patients regarding individualized risk and perioperative management strategies. However, this study has several limitations. First, it is a single-center retrospective study with a small sample size; statistical accuracy could be improved with a larger sample size. In this study, analysis of POPF risk factors was conducted only for PD. If more DP cases were accumulated, further investigation of POPF risk factors in DP cases could be conducted. The small sample size resulted from differences in CT imaging conditions, making many cases ineligible. An ideal study would standardize CT imaging protocols through a prospective design. Second, the study used only CT for evaluation. ECVf also can be measured using MRI, and combining both modalities could allow for more accurate evaluation. Unfortunately, contrast-enhanced MRI is not typically included in routine preoperative assessments, which is why this study relied solely on CT. Third, in scoring of risk, both MPD diameter ≤ 2 mm and ECVf < 36% were assigned 1 point each, but the OR for ECVf < 36% was higher. Separately analyzing cases with only ECVf < 36% or only MPD ≤ 2 mm might be informative. However, because only 1 case had MPD diameter ≤ 2 mm without ECVf < 36%, statistical analysis was limited, so these values each were assigned 1 point to maintain simplicity and usability of the score in clinical settings. As more cases accumulate, it may become possible to further subdivide the 1-point group into MPD diameter ≤ 2 mm only and ECVf < 36% only.

CONCLUSION

The study suggests that pancreatic tissue characteristics such as fibrotic and acinar components as well as texture, may be reflected by ECVf obtained from preoperative CT. Unlike texture, which is subjective, CT evaluation enables an objective approach. In our analysis of POPF risk factors, both MPD diameter ≤ 2 mm and ECVf < 36% were identified as independent risk factors. Cases meeting both criteria are considered at high risk for POPF. As these factors are measurable preoperatively by CT, ECVf may serve as a useful marker to support POPF risk prediction and perioperative management. Although a relatively small sample size limits this study, our approach is novel in that the proposed POPF risk score is derived solely from preoperative CT imaging parameters. Further validation through prospective or multicenter studies is warranted to assess its generalizability and to advance the development of a simple and objective preoperative risk assessment tool using imaging-based parameters.