Three-dimensional length from the center of the liver is a prognostic factor of colorectal cancer with liver metastasis: a retrospective analysis

Background Resectability of liver metastasis is important to establish a treatment strategy for colorectal cancer patients. We aimed to evaluate the effect of distance from metastasis to the center of the liver on the resectability and patient outcomes after hepatectomy. Methods Clinical data of a total of 124 patients who underwent hepatectomy for colorectal cancer with liver metastasis were retrospectively reviewed. We measured the minimal length from metastasis to the bifurcation of the portal vein at the primary branch of the Glissonean tree and dened it as “Centrality”. Predictive effects on positive resection margin and overall survival of centrality were statistically analyzed. Results The value as a predictive factor for the positive resection margin of centrality was analyzed by the receiver operating characteristic curve (area under the curve = 0.72, P<0.001) and centrality ≤ 1.5 cm was an independent risk factor the positive resection margin in multivariate analysis. Total number of metastases ≥ 3 and centrality ≤ 1.5 cm were signicant risk factors of overall survival after Cox regression analysis. Patients with these two risk factors (n=21) had worse 5-year overall survival (10.7%) than patients with one (n=35, 58.3%) or no risk factor (n=68, 69.2%). Conclusion Centrality was related with the positive resection margin of deeply located liver metastasis. Centrality should be considered to establish the surgical strategy for patients with advanced colorectal cancer with liver metastasis.


Introduction
Although curative resection of isolated hepatic metastasis improves long-term survival of patients with colorectal cancer, only 20-30% of colorectal liver metastasis (CRLM) is resectable at the time of diagnosis [1,2]. Several trials assessing the expansion of liver resectability by staged resection or systemic chemotherapy to convert unresectable metastases into resectable metastases have been conducted [3,4]. Generally, the tolerable liver remnant volumes are 20% in a normal liver, 30% in a liver with chemotherapy-induced injury, and 40% in cirrhotic liver [5]. Additionally, adequate resection means the removal of a negative margin in all viable tumors with adequate vascular in ow and out ow and biliary drainage [6]. Thus, a multidisciplinary setting is required to establish treatment strategies for CRLM to achieve favorable oncologic outcomes and preserve remnant liver function after curative resection.
Vascular proximity to the central vessels of the liver is an important factor in selecting a proper treatment strategy. With metastasis occasionally observed near the central vessels, a large-volume resection is required, and subsequently, complete resection is considered di cult. Thus, this study aimed to evaluate the effect of distance from metastasis to the center of the liver on the resectability of metastases and to demonstrate its association with patients' survival.

Data collection
We retrospectively reviewed the medical records of patients who underwent hepatectomy for CRLM in our institution from January 2012 to December 2017. Patients who were diagnosed with colorectal adenocarcinoma with liver-only metastasis and underwent resections with curative intent were included. Patients with other liver diseases, including liver cirrhosis or hepatocellular cell carcinoma, were excluded.
The clinical characteristics of patients, such as age at diagnosis of the primary cancer, sex, primary cancer location, and TN categories, were investigated. Characteristics associated with liver metastasis, such as carcinoembryonic antigen (CEA) level at the diagnosis of liver metastasis, neoadjuvant chemotherapy (NAC), and the longest diameter, total number, and lobar involvement of liver metastases, were also investigated. Operative factors associated with hepatectomy, such as positive resection margin, length of the closest resection margin, intraoperative radiofrequency ablation (IORFA), and synchronous resection with primary tumor, were investigated. Positive resection margin was de ned as the presence of malignant tumor cells on the border of the specimen's resection margin after hepatectomy.
Decision-making in a multidisciplinary team All the treatment strategies for these patients were established by a multidisciplinary team (MDT) in our institution. The MDT for CRLM comprised a group of experienced specialists from medical oncology, colorectal and hepatobiliary surgery, radiation oncology, pathology, and diagnostic and interventional radiology. The resectability of CRLM was assessed by two experienced hepatobiliary surgeons in MDT.
IORFA was performed by radiologic interventionists in MDT only for the tumor located at the periphery of the liver and with no abutting to any vascular structure and less than 2cm in size.
Centrality "Centrality" was de ned as the minimal length from metastasis to the center of the liver, the bifurcation of the portal vein at the primary branch of the Glissonean tree. To measure the centrality, a process of threedimensional (3D) interpolation from two-dimensional (2D) computed tomography (CT) or magnetic resonance (MR) images was performed using a 3-or 5-mm slice thickness (a gap between two serial images) of CT or MR images ( Figure 1). For 3D interpolation, three vertices of A, B, C were required (A = bifurcation of the portal vein, B = point where the liver lesion began to appear, C = projection of A on the axial image where B appears). First, the rst axial image of A and the second axial image of B were identi ed. The length on the Y-axis was measured as the distance between the two images, a multiple (n) of slice thickness and the length between B and C (ω) can be measured on the same axial image. Thus, centrality was calculated using the Pythagorean theorem as follows: In case of multiple metastatic lesions, centralities were calculated for all lesions close to the hilum, and the minimum value was selected and de ned as a nal centrality for that case.

Statistical analyses
Data were analyzed using the Statistical Package for the Social Sciences (SPSS) software version 20.0 (International Business Machines Corporation SPSS, Chicago, IL, USA). Discrete values, such as sex, TN categories, and primary tumor locations, were compared using the Pearson's χ 2 test. Student's t-test was used to compare continuous values such as age, CEA level, and total number and longest diameter of liver metastases. To compare the predictive value of centrality for positive resection margin after hepatectomy, the receiver operating characteristic (ROC) curve analysis and multivariate analysis using the logistic regression model were used. Overall survival (OS), from diagnosis of liver metastasis to death or the last follow-up, was analyzed using the Kaplan-Meier method and log-rank test. Cox proportional hazards regression analysis was used to evaluate the risk factors for OS. A two-sided P value < 0.05 was considered statistically signi cant.

Results
A total of 124 patients were included in this study ( Table 1). The median age at diagnosis of primary tumor was 60 years (interquartile range [IQR], 54-70), and the present study predominantly comprised men (n=77, 62.1%). Left colon and rectal cancer were more frequently observed than right colon cancer (42.7%, 37.9%, and 19.4%, respectively). Eighty-one (65.3%) patients were simultaneously diagnosed with primary cancer and liver metastasis. The median length of liver metastasis was 2 cm (IQR, 1.2-4), and the median number of total liver metastasis was 2 (IQR, 1-4). The median centrality was 4.4 cm (IQR, 2.3-6.0), and 56 (45.2%) patients showed metastasis on both the right and left lobes. Neoadjuvant chemotherapy was performed in 42 (33.9%) patients, and among them, 83.3% received a combination of neoadjuvant chemotherapy with a target agent bevacizumab or cetuximab. Synchronous resections with primary tumor were performed in 73 (58.9%) patients, and 16 (12.9%) patients received IORFA. The median length of the closest resection margin of patients with negative resection margin was 0.4 cm (IQR, 0.1-0.8), and the proportion of patients with negative resection margin was 71.0% (n=88).
The distribution of centrality is shown in Figure 2(a), and a total of 25 (20.1%) patients had centrality ≤ 1.5 cm. The ability of centrality to predict positive resection margin was analyzed using the ROC curve (area under the curve=0.72, P<0.001, Figure 2(b)). Selecting cut-off value from ROC curve was not possible due to the optimal value with high sensitivity and speci city could not be obtained from ROC curve. We set the cut-off value of centrality as 1.5cm according to opinions of hepatobiliary surgeons in our institution. More major hepatectomies were performed in the group of patient with centrality ≤ 1.5cm than the group of patient with centrality > 1.5cm (9 (36.0%) vs 20 (20.2%)), but there was no statistical difference (P=0.095, Supplement Table 1). In the multivariate analysis, centrality ≤ 1.5 cm and bilobar involvement were a signi cant risk factor for positive resection margin (Supplement Table 2). The characteristics of the two groups, patients with centrality > 1.5 cm (n=99) and patients with centrality ≤ 1.5 cm (n=25), were compared, and the results are shown in Table 2. The length of the longest diameter, total number of metastasis, proportion of bilobar involvement, and NAC were higher in patients with centrality ≤ 1.5 cm than those in patients with centrality > 1.5 cm. The proportion of positive resection margin after hepatectomy was also higher in patients with centrality ≤ 1.5 cm than that in patients with centrality > 1.5 cm. Moreover, the resection margin in patients with centrality ≤ 1.5 cm was closer than that in patients with centrality > 1.5 cm.
The median follow-up period of all patients was 39.2 months (min: 5.9, max: 145.8). According to a multivariate analysis, risk factors related with OS were the number of liver metastasis > 3 and centrality ≤ 1.5 cm ( Table 3). All patients were classi ed based on the number of the risk factors observed, and the distribution of patients is shown in Figure 3. Z0 (zone 0) represents patients with no risk factor (centrality > 1.5 cm and total number of metastasis < 3 [n=68]). Z1 represents patients with single risk factor (centrality ≤ 1.5 cm or total number of metastasis ≥ 3 [n=35]), and Z2 represents patients with both risk factors (centrality ≤ 1.5 cm and total number of metastasis ≥ 3 [n=21]). Figure 4 presents the Kaplan-Meier curves of OS of Z0, Z1, and Z2 (5-year OS: Z0, 69.2%; Z1, 58.3%; Z2, 10.7%). Z2 showed worst OS among the three groups, and statistically signi cant differences in OS among the three groups were observed (Z0 and Z1, P=0.026; Z1 and Z2, P=0.016; Z0 and Z2, P<0.001). Additionally, patients in Z0-Z2 were strati ed based on NAC, and their differences in OS between subgroups were analyzed ( Figure 5). After performing a subgroup analysis, patients receiving NAC in Z2 showed better OS than patients not receiving NAC in Z2 (3-

Discussion
This study is considered signi cant considering that centrality, which is de ned as the proximity with the liver hilum at the portal bifurcation of liver metastasis, was proven to be another morphologic factor in deciding the resectability of CRLM. During the resection of liver metastasis, how to optimally solve two con icting issues of preserving maximum liver function and securing su cient resection margin after hepatectomy should be signi cantly considered. Traditionally, volumetric parameters such as tumor size and number of CRLM have been shown to be signi cant factors associated with resectability and prognosis after resection [7,8]. However, the concept of resectability of CRLM has been changed over the last decade with the introduction of several treatments. Surgical techniques, such as portal vein embolism (PVE) or two-stage hepatectomy (TSH) are proven to be safe and curative in selective cases of advanced CRLM with inadequate future liver remnant (FLR) or underlying liver diseases [9,10]. Combined resection with IORFA is bene cial in preserving FLR with favorable oncologic outcomes [11,12].
Additionally, NAC with a target agent and rescue surgery for unresectable CRLM results in tumor downsizing with survival bene ts [13]. Considering that the bene ts of these treatments have been proven, active resections have been widely performed, and the criteria for resectability have focused on how to perform R0 resection with su cient liver function [7]. Thus, factors associated with positive resection margin should be comprehensively studied to de ne the resectability of CRLM.
Although the standard width of the resection margin of CRLM remains unknown, traditionally, resection margin of 1 cm has been accepted as a minimal margin to obtain favorable oncologic outcomes [14]. Certainly, R1 resection, i.e., microscopic identi cation of malignant cells on the resection margin, is a negative factor for survival after resection based on previous reports [15]. However, other studies reported that the width of surgical margin, even < 1 cm, does not affect survival in patients with R0 resection [16]. Thus, hepatic resection tends to be actively performed even when the length of the resection margin is < 1 cm on imaging modality. However, all hepatectomy for CRLM should be performed with efforts to achieve su cient resection margin length even for the centrally-located tumors.
Troisi et al. suggested several factors in affecting the di culty of hepatectomy, such as previous hepatectomy, disease type (hepatocellular carcinoma, benign or metastatic lesion), liver function, and tumor size and number [17]. They also reported that performing hepatectomy is highly di cult when the tumor is located in deeper layers or proximal to the branch of the Glissonean tree. As the tumor is closer to the primary branch of the Glissonean tree, the hilar dissection or anatomical resection is required to completely resect the tumor. The centrality in this study was at the bifurcation of the primary Glissonean branch; thus, we could assume that the centrality was a factor indicating the di culty of hepatectomy to secure su cient resection margin near the center of the liver.
Subsequently, how should we approach CRLM with high centrality? Per et al. demonstrated that associating liver partition and portal vein ligation (ALPPS) made major resections possible with improved resection rates compared with conventional TSH [18]. They also reported the rate of severe complications and negative resection margins were comparable with TSH. Enhanced one-stage hepatectomy (E-OSH) is another technique that makes major hepatectomy possible especially for deep-located nodules [19,20]. E-OSH utilizes the intraoperative ultrasound and vascular manipulations to achieve complete vascular control and detachment of liver metastasis from intrahepatic vascular structures. This approach was reported to be helpful to overcome the disease progression between the two hepatectomies which was a major drawback of TSH and also showed a comparable survival with ALPPS. These approaches, ALPPS and E-OSH, should be considered for advanced liver metastasis with high centrality. Additionally, recent studies reported a greater conversion rate of initially unresectable CRLM after NAC with regimens of oxaliplatin, irinotecan, and other target agents [21]. Thus, NAC should be considered in patients with high centrality, so that hepatectomy can be performed after securing su cient resection margin at the central side. However, prolonged chemotherapy may also have a detrimental effect on the hepatic parenchyma due to oxaliplatin-induced sinusoidal obstruction or irinotecan-induced steatosis [22,23]. Hence, liver function and remnant liver volume should be carefully monitored during chemotherapy. Taken together, various treatment tools should be considered when establishing appropriate strategies for CRLM with high centrality, and treatments should be customized based on patient's condition.
This study has several limitations. First, the de nition of the center of the liver, bifurcation of the portal vein at the rst branch of the Glissonean tree, was subjective and did not include other important structures. For example, proximity to the inferior vena cava is an important factor in deciding adequate vascular out ow, but it was not considered in this study. Second, the factor of centrality ≤ 1.5 cm was selected empirically according to the opinion of hepatobiliary surgeons in our MDT team. According to the ROC curve analysis, the speci city for positive resection margin of centrality ≤ 1.5 cm was 91.9%. However, the sensitivity was only 30.6%, suggesting that factors other than centrality should be considered when deciding the positive resection margin of CRLM. Third, this study was retrospectively conducted based on previous pathologic reports, which contained only the shortest length of resection margin of specimens and did not indicate whether the resection margin was of the central side. Fourth, although the difference was not statistically signi cant, the difference of the long-term overall survival between of -NAC group in Z1 and that of + NAC group in Z1 was relatively signi cant. This might be caused by the small volume of subgroups of Z1 (number of NAC-in Z1 = 13 and number of NAC+ in Z1 = 22). Fifth, the reason why the centrality is related with a worse survival was not de nite. More major hepatectomy were performed in the high centrality group and major hepatectomy is related with posthepatic liver failure [24]. We carefully assumed that the liver function of the high centrality group might be deteriorated due to complications after major hepatectomy such as liver cirrhosis or biliary stula in the mid or long term period of follow-up and treatments such as repeat resections for recurrent hepatic metastasis or compliance to the chemotherapy might be affected to survival of patients. However, the tumor burden of the high centrality group seemed to be higher than that of the low centrality group, thus the heterogeneity in tumor burden might affect the results of survival of patients with high centrality.

Conclusion
Centrality, proximity to the central vessels of liver metastasis, had a potency in predicting the positive resection margin and a worse overall survival after hepatectomy for CRLM. Centrality might be bene cial in establishing treatment strategies especially for patients with advanced colorectal cancer liver metastasis.

Consent for publication
Not applicable

Con ict of interest
There is no con ict of interest in this study.

Data Availability Statement
The data that support the ndings of this study are available from the corresponding author upon reasonable request.  Figure 1 Measurement of centrality using three-dimensional interpolation from two-dimensional images of liver metastasis. Centrality was de ned as the minimal length from liver metastasis to the center of the liver to the bifurcation of the portal vein at the primary branch of the Glissonian tree. 〖Centrality〗^2 〖= y〗^2+ω2, y = slice thickness × n, ω = length between B and C, A = bifurcation of the portal vein, B = point where the liver lesion began to appear, C = projection of A on the axial image where B appears