Prognostic Accuracy of the Eighth Edition of TNM Classification Compared with the Seventh and Sixth Edition for Perihilar Cholangiocarcinoma

Background: This study was conducted to evaluate and compare the prognostic value and performance of the 6th, 7th, and 8th editions of the American Joint Committee on Cancer (AJCC) staging system when applied to patients undergoing surgery for perihilar cholangiocarcinoma (PHC). Methods: Patients undergoing liver resection with curative intention for PHC between 2002 and 2019 were identified from a prospective database. Histopathological parameters and stage of the PHC were assessed according to the 6th, 7th, and 8th editions of the tumor node metastasis (TNM) classification. The prognostic accuracy between staging systems was compared using the area under the receiver operating characteristic curve (AUC) model. Results: with revealed male gender, positive resection margins, presence of distant metastases, poorly tumor differentiation, lymph no as independent predictors of poor survival (p <


Introduction
Perihilar cholangiocarcinoma (PHC) is a relatively uncommon disease, but the treatment of this cancer remains a medical challenge. Chemotherapy and radiotherapy, either alone or in combination, have failed to provide marked improvements in the long-term patient outcomes, and operative resection still represents the only possible curative treatment for these patients [1,2].
However, surgery for PHC is complex and continues to be a most difficult task for hepato-pancreatobiliary surgeons. Hepatectomy, with en bloc resection of the caudate lobe and extrahepatic duct and lymph node dissection, together with resection of the portal vein and/or hepatic artery when indicated, is an aggressive approach for PHC that can result in disease-free and overall survival [3][4][5]. Recent studies have reported 5-year survival rates following curative-intent liver resection varying from 25% to 40% [1,3,[6][7][8]. Unfortunately, the majority of patients at presentation have metastatic or locally advanced disease; consequently, they do not benefit from resection [9][10][11][12][13].
PHC has a large number of factors associated with its prognosis, and yet prediction of long-term survival can be difficult to determine and accurate stratification of patient prognosis remains challenging [11,14,15]. Several prognostic schemes have attempted to aggregate the most relevant clinicopathological factors to provide estimates of long-term survival [6,7,14,[16][17][18]. Commonly used prognostic tools include the Bismuth-Corlette classification and the Blumgart Tstage system, which focus on determining resectability using preoperative imaging [7,19,20]. However, the American Joint Committee on Cancer (AJCC) staging system is still the most widely used for determining prognosis and appropriate treatment and for comparing outcomes between other centers [21].
In the 6th edition (2003) of the AJCC staging system, all patients with extrahepatic bile duct tumors were covered by a single classification [22]. By contrast, the 7th edition (2010) was the first system to stage perihilar (proximal) and distal cholangiocarcinoma separately [23,24]. The recently published 8th edition of the AJCC staging manual now has introduced notable changes to the T-, N-, and overall stage category classification schemes [16,20,21,23,25] (Table   1). In particular, the T4 disease category excludes bilateral second-order bile duct extension and is now defined as a tumor invading the main portal vein or its branches bilaterally, the common hepatic artery, or the unilateral, second-order biliary radicals with contralateral portal vein or hepatic artery involvement [26].
This new definition now means that T4 tumors are downstaged from stage IVA to IIIB and a newly introduced stage IIIC (Any T, N1, M0). The 8th edition has also introduced substantial changes in the nodal staging of patients with PHC. With respect to N-stage, the N1-and N2-stage categories have changes in the counts of lymph nodes (N1: involvement of 1-3 regional lymph nodes; N2: involvement of 4 and more regional lymph nodes).
The current literature contains few studies that have utilized the newly published AJCC 8th edition staging system for patients, and especially those undergoing liver resection for PHC with curative intent at Western high-volume hepatopancreato-biliary (HPB) centers [25]. Against this background, the aim of the current study was to compare the prognostic accuracy and validation of the newly published tumor node metastasis (TNM) classification of the AJCC 8th edition with the 7th and 6th editions for patients with PHC following liver resection. Specifically, we wanted to evaluate whether the new PHC classification provides better differentiation between tumor stages and more accurate prediction of patient survival. We also wanted to evaluate survival rates with regard to biological tumor markers and pathological characteristics.

Study Population
This study included patients whose prospective data was contained in a database at the University Hospital of Leipzig. The study was approved by the local Ethics Committee (AZ EK: 243-14-14072014). Only patients with resected and histologically confirmed PHC were included in the study. Patients with final pathology indicative of a diagnosis other than PHC, those who had suffered inhospital mortality (such as patients who underwent an R2 resection), those who had undergone a palliative operation, and those who had undergone ablation were excluded from this study.
PHC was defined as cancer involving the hilar bile duct (the duct located topologically between the right side of the umbilical portion of the left portal vein and the left side of the origin of the right posterior portal vein) [27].

Patient management
Preoperatively, a diagnostic workup was performed for a multidisciplinary decision.
Before surgery, every patient was staged by anamnesis, physical examination, and cross-sectional imaging with computed tomography (CT) of the abdomen and chest, as well as magnetic resonance imaging (MRI) with magnetic resonance cholangiopancreatography (MRCP) of the liver at or before initial referral, ideally prior to biliary drainage. This imaging allowed evaluation of distant metastases, the biliary extent of the tumor, and involvement of the portal vein and hepatic artery.
In recent years, preoperative biliary drainage of the presumed remnant liver was carried out in some cases percutaneously, rather than endoscopically, to minimize bacterial contamination of the biliary tract. However, many of the included patients had undergone endoscopic drainage before referral. A brush was typically used at the time of preoperative biliary drainage, but pathologic confirmation of malignancy was not required to proceed with surgical resection.
Diagnostic laparoscopy to rule out distant metastatic disease was frequently performed, but not in all patients, given the decreasing yield as preoperative imaging improved over time.
Patients considered to have an insufficient size and function of the future liver remnant underwent a portal vein embolization (PVE). Neoadjuvant chemotherapy was administered in fewer than 10% of the study patients.
With some exceptions, patients with involvement of the main portal vein, the portal vein bilaterally, the common hepatic artery or the hepatic artery bilaterally were considered unresectable. Patients with involvement of the second-order bile ducts were considered resectable if clear margins were anticipated at the segmental bile ducts. The presence of extensive bilobular metastases, such as peritoneal dissemination, was also considered to indicate unresectable disease.
The operative technique consisted of extrahepatic bile duct resection, with complete dissection of the hilar structures, as well as the fatty and lymph node tissue surrounding the common hepatic artery, the main portal vein, and the bile duct. Lymph nodes of the hepatoduodenal ligament, the proper hepatic artery, and the posterior surface of the head of the pancreas were dissected routinely and retrieved. Interaortocaval lymph nodes were retrieved when they appeared macroscopically abnormal.
The type of hepatectomy was determined by the relative location and extent of the tumor, according to the bismuth classification [19,28].
Resection/reconstruction of the portal vein or hepatic artery was undertaken when macroscopic vascular invasion was encountered during surgery.
Prior to December 2007, caudate lobectomy (TCL) was performed at the discretion of the surgeon. If, during the operation, the surgeon felt the tumor had infiltrated the caudate bile duct, TCL was performed to obtain a radical resection.
From January 2008 onward, TCL was customarily performed with prophylactic portal vein resection and extended hepatectomy as a part of no-touch technique [3].

Data Collection
Standard patient demographic and clinicopathological characteristics, including age, sex, American Society of Anesthesiologists (ASA) classification, and presence of cirrhosis, were collected and analyzed. Serum levels of carcinoembryonic antigen (CEA), Cancer Antigen (CA) 19-9, and preoperative bilirubin levels were also analyzed.
We also collected data regarding the treatment characteristics, including administration of neoadjuvant chemotherapy, type of surgery, preoperative drainage/endoscopic stenting, portal venous embolization (PVE), and administration of adjuvant treatments.
Resection margin status was classified as microscopically negative (R0) or microscopically positive (R1). Macroscopically positive tumors (R2 tumors) were not considered for analysis of the TNM classification.
Tumor-specific characteristics, including tumor size and number of tumors, liver capsule involvement, histological grade, morphological type, the number and site of lymph nodes, and the number of collected metastatic lymph nodes, were evaluated. The presence of lymphovascular/perineural/biliary invasion and direct invasion of the portal vein, hepatic artery, and contiguous organs were also recorded. Table 1 presents differences in the AJJC staging classifications for the 6th, 7th, and 8th editions. With the help of pathological, surgical, and radiological reports, data on tumor stage were collected and each patient was classified separately according to the 6th, 7th, and 8th edition AJCC staging systems.

Statistical Analysis
Continuous variables were summarized as mean/median values with interquartile range (IQR)/ standard deviation depending on the normality of the distribution.
Categorical variables were expressed as whole numbers and percentages.
The outcome for survival analyses was overall survival (OS), defined as the time interval between the surgery and the date of death or last follow up for living patients.
Survival rates were measured using the Kaplan-Meier methodology, and a logrank test was used to compare survival curves.
Cox proportional hazards models were used to evaluate associations between tumor stage and OS. In the survival analysis, clinicopathological and demographic variables resulting in a statistically significant value (P < 0.05) in the univariate analysis were entered into the multivariate analysis to assess their independency.
The coefficients from the Cox models were reported as hazard ratios (HRs) with corresponding 95% confidence intervals (CIs).
The area under the receiver operating characteristic curve (AUC) was calculated to test the discriminative power of each TNM edition for whether a patient would survive or die. Calibration was assessed using the Hosmer-Lemeshow Chi-square test to determine the goodness of fit, and patients with ongoing follow-up were censored at the last time point of examination.

Baseline Characteristics
A total of 189 consecutive patients underwent surgery between 2002 and 2019.
In 95 patients, liver resection was performed with curative intention, whereas 94 (49%) patients underwent an explorative laparotomy due to peritoneal carcinosis, reduced liver function, or major tumor extension.

Stage transitions
The 6th edition of the AJCC staging categorized 5 (

Univariate and multivariate analysis of factors associated with OS after curative resection
The median follow-up time of patients after liver resection was 4.3 ± 2.9 years.
We tested several clinical and pathological factors on OS. As shown in Table 4, OS was significantly longer in female patients (p = 0.02), patients with no extrahepatic distant metastases (p < 0.001), those with negative resection margins (p = 0.003), those aged < 65 years (p = 0.03), those with good/moderate tumor differentiation (p = 0.01), and patients with no microvascular invasion (p < 0.01).
Lymph node status was strongly associated with OS. Patients who were lymph node negative had a 5-year overall survival of 33.3% compared with 24.3% and 0% for patients with 1-3 and >3 metastatic lymph nodes (p < 0.01).
Portal vein resection was not associated with significantly better overall survival (p > 0.05), but patients who underwent caudate resection had a 5-year OS of 28.7% compared with 17.1% for those who did not undergo caudate lobe resection (p = 0.004).
The multivariate analysis ( remained associated as independent predictors of OS.

Survival Across Stages
Overall, an unbalanced distribution was observed that favored stage IIa (40%) in the 6th edition and stage II (54.7%) in 7th and 8th editions.  Table 7 presents the ROC analysis and Hoshmer-Lemeshow test (HLT) for discrimination and calibration of all the evaluated staging systems. The prognostic accuracy was remarkably similar across the three analyzed AJCC staging systems. The prognostic accuracy of the main stages was higher for the 8th edition than for the 7th edition (AUC: 8th edition versus 7th edition: 0.66 versus 0.61) but the difference did not reach statistical significance. Expanding the 7th edition to include substages slightly diminished its prognostic accuracy (AUC: from 0.61 to 0.58), whereas, expansion of the 8th edition with substages increased the prognostic accuracy (AUC main stages versus substages: 0.66 to 0.69) and a statistical trend was observed (p = 0.09). With regard to calibration of the evaluated stagings systems, the HLT shows acceptable calibration and fitness of the analyzed 7th and 8th staging systems ( Table 8).

Discussion
PHC is a challenging disease to manage, and only selected patients are candidates for curative-intent surgery [29,30]. Therefore, accurate staging with a standardized staging system is essential not only for planning the frequency of post-operative surveillance, making therapeutic decisions, and/or selecting patients for adjuvant treatment, but also for informing both patients and physicians about long-term clinical outcomes and for allowing for a standardized exchange of information about the extent of the disease [1,2]. While the optimal method for risk stratification of patients with PHC remains unclear, the TNM system provided by the AJCC and the International Union Against Cancer (UICC) is the most commonly used method for stratifying outcomes among cancer patients [15,20,24,31].
In this context, the newly released 8th edition of the AJCC staging system has introduced four major modifications (Table 1) and has especially included the recommendations by Ebata et al. [15], who modified the T-categories of the previous 7th TNM edition in a large survival analysis of 1399 patients undergoing liver resection for PHC at 8 Japanese liver centers [15]. These authors reported that survival was slightly better for patients with advanced disease (T4 tumors, stage IVa) than for patients with regional lymph-node positive disease (stage IIIb), a finding confirmed in our study. However, as already recognized, and as previous data have shown, marked differences exist in peri-and operative strategies and outcomes between Eastern and Western centers; therefore, external validation of the prognostic accuracy of staging systems in Western patients is essential [32,33].
A recently published study from two Western hepatobiliary centers evaluated the prognostic accuracy of the 8th TNM classification of the AJCC staging system in a cohort of 214 patients undergoing liver resection for PHC. In that study population, about 40% of the patients changed their stages from the 7th to the 8th AJCC edition. The authors determined that the new 8th TNM edition was only slightly better than the previous 7th edition [25].
Our analysis confirmed findings that could validate the new 8th TNM classification. For example, we observed a reclassification in a total of 25 patients The extent of intrahepatic biliary invasion is a critical factor that determines surgical indications and the type of hepatectomy that should be performed in patients with PHC. To ensure a curative liver resection (R0 margins) in patients with Bismuth type II and III PHC, the resection should include a hemihepatectomy combined with en bloc resection of the first segment (caudate lobe), as well as an extrahepatic bile duct resection to ensure negative resection margins [25,28].
However, in patients with Bismuth type IV disease (characterized by having bilateral involvement of the second-order intrahepatic bile ducts), the standard operative approach remains unclear, and many centers still consider Bismuth type IV tumors unresectable [2,[34][35][36][37]. Conversely, as surgical techniques and perioperative management have improved, some surgeons have suggested that a type IV tumor does not always imply unresectability [6,15,34,38]. In this context, these surgeons have proposed a more aggressive operative approach for patients with a Bismuth type IV PH-CCA, including an extended hepatectomy combined with caudate lobectomy, extended lymph node dissection, and vascular resection with portal vein and/or hepatic artery reconstruction, to achieve negative resection margins [3,8,15,34,39,40]. The results of previous studies indicate that this aggressive strategy could further improve long-term outcomes and prolong patient survival; therefore, this procedure could be established as a state-of-the art operation for patients with locally advanced PHC, and especially for Bismuth-Corlette types III-IV hilar cholangiocarcinomas [3,34,41,42].
The Japanese group associated with Ebata et al. [35] was the first to report that patients with P0M0N0 disease who underwent an R0 liver resection showed no significant differences in survival when compared with patients with Bismuth type I/II, III and IV tumors, as the 5-year OSs were 63.1%, 65.6%, and 59.2%, respectively [35]. Similar findings were obtained in a recently published study by Ruzzenenete et al. [25], as well as in the present study, as the 5-year survival rate was 50.1% in our 30 patients with pM0N0 who underwent R0 resection for a Bismuth Type IV PHC. Although these patients represent a highly selected subset of patients undergoing aggressive surgery, these findings nevertheless suggest that an aggressive operative approach with or without vascular resection could offer a curative option even for patients with Bismuth type IV PH-CCA [25,35].
Furthermore, the T staging should separate the longitudinal extension of the disease from its vascular invasion.
Collectively, these data suggest that while PHC generally has a poor prognosis, large differences in survival can be observed among patients based on a subset of clinical and pathological factors. In our cohort of patients, several factors, including gender, age, surgical radicality, extent of liver resection (± caudate lobe resection), tumor differentiation and distant metastases (e.g., lymph node status), were independent prognostic factors for long-term survival, in agreement with previous observations [12,41,42].
In this context, the prognostic impact of the lymph node (LN) status is considered as one of the most important factors for long-term survival following liver resection for PHC [5,18,[43][44][45]. A certain benchmark number of retrieved lymph nodes is necessary to secure a representative and adequate staging. However, the extent of the lymphadenectomy during the resection of PHC, such as the minimum number of lymph nodes to be retrieved, remains under controversial debate, and marked differences in the operative approach between Western and Eastern centers have been reported [43,44,46,47]. Recent SEER register analyses showed that the retrieved LN counts are independent prognostic survival factors for node-negative and node-positive PHC patients [48,49].  [43,44,47,50,51]. The authors of a previously published systemic review of 20 studies suggested that a LN count greater than or equal to 7 is adequate for prognostic staging, while a LN count greater than or equal to 15 does not improve detection of patients with positive LNs [43]. By contrast, a recently published international, multi-center study that included 437 patients with PHC who underwent liver resection reported a required median number of retrieved LNs of 3 (IQR 1-7), and the incidence of LN metastasis was 36%, similar to other published values from Western centers [46]. The results of the current study should be interpreted while considering several limitations. One limitation is that, although our study is surely one of the larger Western series of patients with resected PHC to be reported in the literature, our sample size is nevertheless still insufficient to allow drawing of definitive conclusions about small modifications in this staging system. The retrospective nature of our study and the long study time (over 17 years) presents a risk of selection and confounding biases regarding the diagnosis and treatment options of the patients. Therefore, larger cohorts of patients are needed to test the TNM classification, although this is challenging given the rare incidence of PHC.
A second limitation is our inclusion only of patients who underwent a liver resection. The TNM classification offers a precise tumor-staging algorithm aside from the treatment set and is widely used in cancer staging. However, only surgery and pathological examination are able to provide an accurate confirmation of the TNM stages. The advantage of our study is that pathological confirmation of the TNM stage was available for all patients. The disadvantage is that the observed results cannot be extrapolated to patients who did not undergo resection because of extrahepatic metastasis or locally advanced tumor disease.
However, we included only resected patients because the assessment of both Tstage and N-stage is inaccurate without microscopic evaluation of a resected specimen. In this context, the accuracy of estimation of the nodal status based on an evaluation of the size criteria on cross-sectional imaging is very limited. For example, Ruys et al. [54] showed sensitivity and specificity estimates of 61% and 88% for lymph node involvement based on CT staging of patients with PHC.
Furthermore, only 37% of the lymph nodes >30mm in diameter were confirmed as positive after analyzing slides of 147 patients undergoing liver resection for PHC. These findings indicate a very low diagnostic accuracy of cross-sectional imaging for staging of perihilar cholangiocarcinoma, and especially for nodal involvement [52].
A third limitation is that our study included patients undergoing nonradical (R1) and suboptimal tumor resections (distal bile duct resections, no caudate lobe resections) and/or patients with involvement of main portal vein or common hepatic artery. The improvements in standard operative and perioperative approaches, in combination with new neoadjuvant or adjuvant chemotherapeutic agents, necessitates inclusion of sensitivity analyses of those patients who underwent R0 extended hepatectomy with caudate lobe resections to compare them with the whole cohort of patients.

Conclusions
An optimal staging system should provide information about the prognosis, guide the therapy, and allow comparison with different staging systems. However, none of the currently existing PHC staging systems fulfills these criteria. Historically, the AJCC staging is based mainly on the anatomic extent of the tumor. Although non-anatomic factors have been introduced into the staging of some cancers (e.g., mitotic rate in melanoma), with reference to PHC, the 6th, 7th, and 8th editions of the AJCC staging system have all adhered to the anatomic extent of the tumor, as in all other hepatobiliary and pancreatic cancers [11,23]. Despite

Declarations Acknowledgements
We acknowledge support from the German Research Foundation (DFG) and University of Leipzig within the program of Open Access Publishing.

Funding
This work was supported by the German Research Foundation (DFG) and Leipzig University within the program of Open Access Publishing. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Availability of data and materials
Our database contains highly sensible data which may provide insight in clinical and personnel information about our patients and lead to identification of these patients. Therefore, according to organizational restrictions and regulations these data cannot be made publically available. However, the datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Author contributions
HMH, FM and RS were responsible for the study conception and design; HMH, FM, SR, RS and DS were responsible for data acquisition; HMH, FM, SR and DS analyzed and interpreted the data; HMH, FM and RS drafted the manuscript; and HMH, FM, SR, RS and DS critically revised the manuscript. All authors read and approved the final manuscript.

Ethics approval and consent to participate
The study was approved by the local ethical commission board from the University of Leipzig (AZ-EK: 243-14-14072014).
Due to the retrospective design of the study and accordingly national guidelines, the local ethic committee confirmed, that informed consent was not necessary from participants.

Consent for publication
Not applicable Tumor invades the main portal vein or its branches bilaterally; or the common hepatic artery; or unilateral second order biliary radicals with contralateral portal vein or hepatic artery involvement NX Regional lymph nodes cannot be assessed Regional lymph nodes cannot be assessed Regional lymph nodes cannot be assessed N0 No regional lymph node metastasis No regional lymph node metastasis No regional lymph node metastasis N1 Regional lymph node metastasis are the cystic duct, pericholedochal, hilar, peripancreatic (head only), periduodenal, periportal, celiac, and superior mesenteric nodes Regional lymph node metastasis including nodes along the cystic duct, common bile duct, common hepatic artery, and portal vein Metastasis to 1-3 regional lymph nodes N2 Metastasis to 4 or more regional nodes      Table 5: Cross-tabulation of the 6th, 7th and 8th editions of the American Joint Committee on Cancer (AJCC) staging system