Efficacy of CT Value Along Portal Vein for Preoperative Prediction of Portal Vein Resection in Pancreatic Head Cancer

Background It is important for surgeons to determine whether combined portal vein (PV) resection (PVR) is necessary before surgery. The present study aimed to determine the ability of computed tomography (CT) value along the PV in predicting the necessity for concomitant PVR. Methods A total of 107 consecutive patients who underwent pancreaticoduodenectomy (PD) for invasive ductal carcinoma of the pancreatic head at our institute between September 2007 and September 2020 were reviewed retrospectively. Univariate analysis to predict PVR was performed with preoperative radiological valuables acquired by Synapse Vincent. The resected specimen near the PV or the PV notch was analyzed by histopathological findings. Results Only the CT value of the PV was independently associated with PVR (Mann-Whitney U test; P = .045, logistic regression test; P = .039). The outer boundary of the PV was unclear in the cases without pathological PV invasion and PVR due to the development of smooth muscle in the outer membrane of the PV and the proliferation of collagen fibers. The elastic fibers were arranged regularly in the notch portion of the PV in cases wherein PVR was not performed. Discussion The CT value along the PV was independently associated with PVR and is the only predictor of PVR. These results were very useful in predicting PVR preoperatively and were histopathologically supportive.


Introduction
The prognosis of pancreatic cancer is extremely poor.2][3] A reason for the poor survival rate is that above 50% of the cases are diagnosed at an advanced stage.Peripancreatic vascular invasion, which has a major impact on prognosis, is an important criterion to consider while assessing resectability. 4 For patients with peripancreatic venous involvement, pancreaticoduodenectomy (PD) with portal vein (PV) resection and reconstruction substantially increases the probability of achieving R0 resection. 5herefore, accurate preoperative evaluation of the presence of vein invasion is essential. 6,7everal models have been proposed for predicting preoperative peripancreatic venous invasion.However, in actual surgery, combined resection of the PV is required in many cases because it cannot be safely detached otherwise.This is true for not only the invasion of cancer but also the effects of concomitant pancreatitis and fibrotic changes around the pancreas.According to recent reports, concomitant PV resection (PVR) during PD in high-volume centers is safe, with lower morbidity and mortality rates than earlier. 8Therefore, it is important for surgeons to diagnose PV invasion and determine whether combined resection of the PV is necessary before surgery.
The present study aimed to determine the ability of computed tomography (CT) value along the PV in predicting the necessity for concomitant PVR in pancreatic head carcinoma.

Methods
This research has been approved by Tokyo Medical University Hachioji Medical Center Ethics Committee.This research has obtained informed consent from all participants.

Patients
In total, the data of 107 consecutive patients who underwent PD for invasive ductal carcinoma of the pancreatic head at our institute from September 2007 to September 2020 were collected retrospectively.Five patients were excluded from the study, and data from the remaining 102 patients were evaluated.Patients with the following histological tumor types were excluded from the study: mucinous cystadenocarcinoma, intraductal papillary adenocarcinoma, and neuroendocrine carcinoma. 7

Radiological Analysis
All patients underwent preoperative abdominal CT during the arterial-phase, portal-phase, and equilibrium phases using Aquilion ONE (Toshiba Medical Systems, Tokyo, Japan), a LightSpeed Volumetric CT (GE, Waukesha, WI, USA), and Revolution Frontier (GE, Waukesha, WI, USA) reconstructed at 1.25 mm intervals. 7Unenhanced and three-phase contrast-enhanced helical CT images were obtained.The beginning of image acquisition for each phase after the injection of the contrast material was timed using the automated Dual Shot GX7 (Nemoto Kyorindo, Tokyo, Japan).Arterial phase and portal venous phase image acquisition started at 10 and 40 s after contrast injection, respectively. 7Delayed-phase image acquisition started 110 s after the contrast injection.
Contrast material (mean, 520 mg iodine/kg body weight) was delivered over a period of 30 s. 7

Calculation of CT Value Along PV by Synapse Vincent
All CT data were analyzed using the Synapse Vincent system (Fuji Film).Images were developed using multiplanar reformation (Figure 1A).Next, we confirmed the most accessible part of the cancer centering on the PV and confirmed the axial image (Figure 1A, left upper panel).We then made fine adjustments in the sagittal image (Figure 1A, right upper panel) and coronal image (Figure 1A, left lower panel) to follow the PV course.Finally, after reconstructing the image of the most approaching part adjusted parallel to the PV (Figure 1A, right lower panel), the CT value was measured for individual patients so that the part along the PV had an area of approximately 80 mm 2 , in which the PV was not included (Figure 1B).All evaluations were performed by two investigators (S.O. and N.C., with 8 and 20 years of post-training experience in surgery and CT imaging, respectively), and the average values were adopted.

Surgical Procedure
The PV around the pancreas was dissected as much as possible, regardless of whether it was due to cancer invasion or surrounding inflammation.Concomitant resection of the PV was performed when dissection became difficult.The surgery proceeded with monitoring of the pathological findings for the frozen section collected from around the PV.At the location where cancer invasion was observed, concomitant resection was performed immediately.All authors performed the surgery, and the procedures were unified.

Histopathological Analysis
The resected specimens were fixed in formalin and embedded in paraffin.For each case, a representative tissue block showing pancreatic ductal carcinoma and the PV was selected.In the cases without PVR, the part of the pancreatic tissue the PV was probably in contact with was selected.Serial 3 μm-thick sections were prepared from the paraffin-embedded tissue blocks and stained with hematoxylin-eosin, Elastica van Gieson, and Azan.

Statistical Analysis
All statistical analyses were performed using SPSS (version 24.0;IBM Corp., Armonk, NY, USA).Categorical variables were evaluated using the chisquare test or Fisher's exact test.Continuous variables were compared using Student's t-test or the Mann-Whitney U test (for variables with non-normal distribution).All variables were incorporated into a univariate analysis, and only those variables showing statistical significance (P < .05)were evaluated by univariate logistic analyses to identify the independent predictive factors for concomitant PVR in patients with pancreatic head cancer.Survival analysis was performed using the Kaplan-Meier method, and the log-rank test was performed for comparison between groups.Receiver operating characteristic curves and the corresponding area under the curve (AUC) were used to assess the performance of the predictive model with the test data.1A).Next, we confirmed the most accessible part of the cancer centering on the PV and confirmed the axial image (Figure 1A, left upper panel).We then made fine adjustments in the sagittal image (Figure 1A, right upper panel) and coronal image (Figure 1A, left lower panel) to follow the PV course.Finally, after reconstructing the image of the most approaching part adjusted parallel to the PV (Figure 1A, right lower panel), the CT value was measured for individual patients so that the part along the PV had an area of approximately 80 mm 2 , in which the PV was not included (Figure 1B).
Table 2 shows the two groups stratified by the presence or absence of concomitant PVR.The CT value along the PV was significantly higher in the PVR (+) group than in the PVR (À) group (P = .039).Other variables that were significantly different between the two groups were neoadjuvant chemotherapy (NAC), operative time, blood loss, and Union for International Cancer Control (UICC) stage.These were understandable results because of the many advanced cases in the PVR (+) group.
The overall 5-year survival rate in our series was 16.2%, with a median survival of 20.0 months.As previously reported, our series also showed significantly better survival in patients with positive pathological PV invasion than in those with negative results (data not shown).In addition, no significant difference in overall survival in cases without pathological PV invasion was noted between the PVR (+) and PVR (À) groups (Figure 2).

Predictive Factors for PVR
In this series, univariate analysis showed that sex, age, CEA, CA19-9, tumor size on CT, PV length of tumor contact on CT, and CT value along the PV were analyzed to predict PVR (Table 3).Only the CT value along the PV was independently associated with PVR.The CT value along the PV in the PVR (+) group was significantly higher than that in the PVR (À) group (Mann-Whitney U test; P = .045,logistic regression test; P = .039)(Figure 3A).The AUC of the CT value along the PV for the prediction of PVR was .616(Figure 3B).

Histopathological analysis
The representative histopathological findings along the PV of cases with positive pathological PV invasion,

Discussion
Preoperative diagnosis of pathological PV invasion is of great significance.However, it is difficult to predict PV invasion based on morphological changes alone. 5Although many reports have proposed various nomograms to predict PV invasion, they are unclear because of subjective evaluation of the CT images. 6Only morphological measurements with low subjectivity were examined.Significant factors that can predict pathological PV invasion include tumor diameter, length of tumor-PV contact, PV circumference, and PV wall irregularities. 9,10However, in this study, neither tumor diameter nor the length of the tumor-PV contact were significant factors in predicting concomitant PVR.This is because the objective variable was not pathological invasion but the presence or absence of concomitant PVR.In some cases with concomitant PVR, the tumor was not in contact with the PV.However, in many cases, concomitant PVR was not required even if the tumor was  in contact with the PV.However, in cases with pathological PV invasion, the tumor was in contact with the PV.Preoperative diagnosis of clinical noninvasion of the PV does not coincide with the need for concomitant PVR.From the data of this study, 35% (15 of 43 cases) had pathological PV invasion among the cases with concomitant PVR (40%-60% in other reports). 8,11Pathologically unnecessary PVR was performed in the remaining 28 cases.Reasons other than pathological infiltration include inflammation around the portal vein, pancreatitis, and our surgical concept.Our surgical concept of PVR is to dissect the area around the portal vein and perform PVR when it is judged to be difficult to dissect.Various biases in this research have arisen from this surgical concept.In other words, in addition to predicting PV invasion preoperatively, it is also important to diagnose whether preoperative concomitant PVR is necessary.There are various concepts regarding surgical procedures.If PV invasion, that is, PV stenosis or obstruction, is apparent on the CT image, the detachment of that part of the PV should be minimized, and combined resection should be performed.However, if the PV invasion is not apparent on the CT image but still suspected, detachment of the area around the PV may be needed or combined resection if detachment is deemed difficult.On the other hand, if clinical PV invasion is suspected, some surgeons perform concomitant PVR in every case.Even if PV invasion is not clinically suspected, difficulties in detachment around the PV are encountered due to inflammatory and fibrotic changes in the surrounding areas.Therefore, it is also important to determine whether concomitant PVR is necessary before surgery.
It is clear that positive pathological PV invasion is a clear prognostic factor.In cases without pathological PV invasion, there was no significant difference in the incidence of postoperative complications due to concomitant PVR, and the prognosis did not change.Therefore, our surgical policy was considered acceptable.In addition, the determination of the need for concomitant PVR before surgery could enable appropriate surgery preparation.The rate of unexpected PVR may also decrease.Pathological PV invasion could not be predicted using the CT value of the PV used in this study.In other words, there was no change in the CT value of the PV values in the two groups with and without pathological PV invasion  The CT value along the PV in the PVR (+) group was significantly higher than that in the PVR (À) group (Mann-Whitney U test; P = .045,logistic regression test; P = .039)(Figure 3A).The AUC of the CT value along the PV for the prediction of PVR was .616(Figure 3B).
(data not shown).However, it is necessary to pay attention to differences between facilities, in which particularly those with less experience reconstructing the portal vein.
If a black band on the CT image is visible at the border of the pancreas in the closest part of the tumor along the PV, it is presumed that it is a fat layer and the CT value is lowered accordingly.In such cases, it is not necessary to measure the CT value, and it is determined that PVR may not be necessary.On the other hand, PVR can be considered necessary even when the PV is narrowed or obstructed due to obvious cancer invasion in the absence of such a fat layer.Difficult cases showed no further fat layer or any obvious deformation of the PV.In this case, it is speculated that measuring the CT value of the closest part would enable a meaningful determination of the need for concomitant PVR.
In the histopathological findings of this study, the border outside the PV was unclear in cases without pathological PV invasion and with PVR, and hyperplasia of the collagen fibers was observed.These facts provide evidence to support the previous hypothesis.In addition, the notch of the PV in cases without PVR showed hyperplasia of the elastic fibers; however, the arrangement was well-organized and the presence of any outer fat layer was not observed.The possible reason is that the fat layer disappeared during the process of resection and specimen preparation.
The limitations of this study are that the number of cases was small and it was retrospective in nature.Although it is preliminary data (prospective study), it is suggested that the CT value along PV is a useful predictive factor for PVR before resection (data not shown).In addition, the subjective element in the measurement of the CT value along the PV has not been completely eliminated.The measurement area along the PV and CT values in contrast-enhanced PV imaging has a small range, which makes it more objective but not perfect.Furthermore, the effects of morphological change on the CT value due to the effect of NAC were not examined.

Figure 1 .
Figure1.Assessment of CT value along PV by using the Synapse Vincent software.All images acquired by CT were developed using multiplane reformation (Figure1A).Next, we confirmed the most accessible part of the cancer centering on the PV and confirmed the axial image (Figure1A, left upper panel).We then made fine adjustments in the sagittal image (Figure1A, right upper panel) and coronal image (Figure1A, left lower panel) to follow the PV course.Finally, after reconstructing the image of the most approaching part adjusted parallel to the PV (Figure1A, right lower panel), the CT value was measured for individual patients so that the part along the PV had an area of approximately 80 mm 2 , in which the PV was not included (Figure1B).
negative pathological PV invasion, with PVR performed, and without PVR performed are shown in Figure4.The outer boundary of the PV without cancer invasion was clear in cases with pathological PV invasion.On the other hand, the outer boundary of the PV was unclear in the cases without pathological PV invasion and PVR due to the development of smooth muscle in the outer membrane of the PV and the proliferation of collagen fibers.The elastic fibers were arranged regularly in the notch portion of the PV in cases wherein PVR was not performed.The fat layer on the outside was not found pathologically.

Figure 2 .
Figure 2. Cumulative survival rates in the cases without pathological PV invasion.No significant difference in overall survival in cases without pathological PV invasion was noted between the PVR (+) and PVR (À) groups.

Figure 3 .
Figure3.CT value along PV compared with PV resection and receiver operating characteristics curve.Only the CT value along the PV was independently associated with PVR.The CT value along the PV in the PVR (+) group was significantly higher than that in the PVR (À) group (Mann-Whitney U test; P = .045,logistic regression test; P = .039)(Figure3A).The AUC of the CT value along the PV for the prediction of PVR was .616(Figure3B).

Figure 4 .
Figure 4. Histopathological findings along portal vein (PV) of representative cases.The representative histopathological findings along the PV of cases with positive pathological PV invasion, negative pathological PV invasion, with PVR performed, and without PVR performed are shown in Figure 4.The outer boundary of the PV without cancer invasion was clear in cases with pathological PV invasion.On the other hand, the outer boundary of the PV was unclear in the cases without pathological PV invasion and PVR due to the development of smooth muscle in the outer membrane of the PV and the proliferation of collagen fibers.The elastic fibers were arranged regularly in the notch portion of the PV in cases wherein PVR was not performed.The dotted line represents the outer boundary of the PV, and the solid circle represents the cancer invasion to the PV.

Table 2 .
Patient Characteristics Stratified by Portal Vein Reconstruction.

Table 3 .
Univariate Analysis of Risk Factors for Portal Vein Reconstruction.