Evaluation of the New American Joint Committee on Cancer Staging Manual 8th Edition for Perihilar Cholangiocarcinoma

Background The aim was to compare the prognostic accuracy of cross-sectional imaging of the 7th and 8th editions of the American Joint Committee on Cancer(AJCC) staging system for perihilar cholangiocarcinoma(PHC). Methods All patients with PHC between 2002 and 2014 were included. Imaging at the time of presentation was reassessed and clinical tumor–node–metastasis (cTNM) stage was determined according to the 7th and 8th editions of the AJCC staging system. Comparison of the prognostic accuracy was performed using the concordance index (c-index). Results A total of 248 PHC patients were included;45 patients(18.1%) underwent a curative-intent resection, whereas 203 patients(81.9%) did not because they were unfit for surgery or were diagnosed with locally advanced or metastatic disease during workup. Prognostic accuracy was comparable between the 7th and 8th editions (c-index 0.57 vs 0.58). For patients who underwent a curative-intent resection, the prognostic accuracy of the 8th edition (0.67) was higher than the 7th (0.65). For patients who did not undergo a curative-intent resection, the prognostic accuracy was poor in both the 7th as the 8th editions (0.54 vs 0.57). Conclusion The 7th and 8th editions of the AJCC staging system for PHC have comparable prognostic accuracy. Prognostic accuracy was particularly poor in unresectable patients.


Introduction
Perihilar cholangiocarcinoma (PHC) is the most common malignancy of the bile ducts. 1 Overall survival differs strongly between PHC patients, ranging from 12 months in palliative treatment to 40 months after curative-intent resection. [2][3][4] Prognostic studies typically focus on patients undergoing curative-intent resection. However, the majority of patients with PHC have metastatic or locally advanced disease at the time of presentation. 2,[5][6][7] One of the most commonly used staging systems is the American Joint Committee on Cancer (AJCC) staging system. Recently, the AJCC released the AJCC 8th edition cancer staging manual, which came into effect on January 1, 2018. The 7th edition of the AJCC staging system was the first to stage PHC and distal cholangiocarcinoma separately. The new 8th edition for PHC contains four significant changes (Tables 1 and 2). Bilateral second-order bile duct involvement (i.e., Bismuth classification IV) is no longer classified as T4 in the 8th edition. Other reasons for T4 (e.g., main portal vein Bas Groot Koerkamp and François E. J. A. Willemssen contributed equally to this work. involvement) are reclassified as stage IIIb rather than stage IVa. Positive lymph nodes beyond the hepatoduodenal ligament (e.g., aortocaval or celiac nodes) have become M1 disease (stage IVb) rather than N2 disease in the 7th edition. Instead, in the 8th edition, N2 disease (stage IVa) is classified as 4 or more positive regional lymph nodes.
AJCC staging systems are intended to be applicable to all cancer patients, regardless whether they undergo curativeintent resection, palliative treatment, or best supportive care. As the majority of patients with PHC is not eligible for curative-intent resection, the AJCC staging involves assessment of cross-sectional imaging in most patients, rather than pathological evaluation of resected tumor specimens. Therefore, the aim of this retrospective study was to evaluate the 8th edition of the AJCC staging system for all patients with PHC and compare the prognostic value of the 7th and 8th editions of the AJCC staging system for PHC.

Study Population and Data Acquisition
All patients with PHC between 2002 and 2014 in Erasmus MC University Medical Center, Rotterdam, the Netherlands, were included. PHC was defined as a mass or malignantappearing stricture at or near the biliary confluence, arising between the origin of the cystic duct and the segmental bile ducts. 8 A multidisciplinary team diagnosed PHC based on clinical characteristics, radiological characteristics, endoscopic findings, and follow-up, if histopathological evidence was not available. 9 Patient and tumor characteristics, clinical parameters, and laboratory results were retrospectively collected from electronic patient records.
Experienced abdominal radiologists revised all imaging from the time of first presentation. Tumor diameter, presence and location of suspicious lymph nodes, presence of distant metastases, and vascular involvement was reassessed. Suspicious lymph nodes were defined as nodes larger than 1.0 cm in short-axis diameter, with central necrosis, an irregular border, or hyper-attenuation compared to liver parenchyma in the portal-venous contrast-enhancement phase. 9,10 Vascular involvement was defined as tumor contact of at least 180 degrees to the unilateral or main portal vein or hepatic artery. 9 Tumor-node-metastasis (TNM) stage was determined Tumor invades main PV or its branches bilaterally, or the common hepatic artery, or unilateral second-order biliary radicals with contralateral portal vein or hepatic artery involvement.
Node (N) stage N0 No regional lymph node metastasis No regional lymph node metastasis N1 Regional lymph node metastasis: hilar (along CBD, cystic duct, HA, or PV) One to three positive lymph nodes typically involving the hilar, cystic duct, common bile duct, hepatic artery, posterior pancreatoduodenal, and portal vein lymph nodes N2 Metastasis to periaortic, pericaval, SMA, or coeliac lymph nodes Four or more positive lymph nodes from the sites described for N1 Metastasis (  according to both the 7th and 8th editions of the AJCC staging system (Table 1). TNM stages I and II were combined, since cT1 (stage I) and cT2 (stage 2) cannot be reliably distinguished on imaging. 11 The Institutional Review Boards of Erasmus MC University Medical Center approved the study, and the need for informed consent was waived.

Statistical Analyses
Statistical analyses were performed using IBM SPSS Statistics for Windows version 21.0 (IBM Corp., Armonk, NY, USA) and R (a language and environment for statistical computing) version 3.3.3 for Windows (R Foundation for Statistical Computing, Vienna, Austria). Continuous data are reported as mean with standard deviation (SD) or median with interquartile range (IQR). Categorical parameters are reported as counts and percentages. Survival was measured from the date of first presentation. Survival probabilities were estimated using the Kaplan-Meier method and compared with the log-rank test. Survival status was updated using the municipal records database on December 21, 2017.
Comparison of the staging systems was performed using the concordance index (c-index) and Brier score. The concordance index is used to evaluate whether a staging system can correctly discriminate between two patients at different stages of disease. It is calculated as the probability that for two random patients with different stages, the patient at the lower stage has a longer survival. A c-index of 0.5 means that the predictive ability is no better than random chance. A c-index of 0.7 indicates a good model and an c-index of 1 means perfect prediction. The Brier score is used to measure the difference between observed and predicted survival per stage. As opposed to c-indices, a lower Brier score is better and a score of 0 means total accuracy, while a score of 0.250 indicates no prognostic value.

Patient Characteristics
A total of 248 patients were included; 45 patients (18.1%) underwent a curative-intent resection and 203 patients (81.9%) did not undergo a curative intent resection because they were unfit for surgery or were diagnosed with locally advanced or metastatic disease during workup (Fig. 1). Patient characteristics are summarized in Table 3. Patient characteristics are summarized in Table 3. The median age was 65 years (IQR 55-73) and 150 patients (60.5%) were male. Most patients (n = 106, 44.0%) had an ECOG performance status of 0 and 87 patients (35.1%) had a tumor larger than 3 cm on imaging. Unilateral involvement of the portal vein

Survival across Stages
The median OS for patients staged according to the 7th or 8th edition per TNM stage were as follows: stage I/II (17.0 vs. 17.0 months), stage III (10.5 vs. 10.9 months), and stage IV (7.03 vs. 5.6 months), respectively (p value between stages in Prognostic Accuracy Table 6 shows the concordance indices and Brier scores for the two editions of the AJCC staging system. Prognostic accuracy of the 8th editions of the main stages of the AJCC staging systems was slightly higher than the 7th edition (cindex 0.59 vs 0.61). Expanding the 7th edition to include substages (e.g., IIIa and IIIb) slightly diminished its prognostic accuracy (c-statistic from 0.59 to 0.57). Expansions of the 8th edition also diminished its prognostic accuracy (c-statistic from 0.61 to 0.58). Prognostic accuracy was comparable between the expanded 7th and 8th AJCC staging systems (cindex 0.57 vs 0.58). Subgroup analysis was performed to determine the prognostic accuracy of the AJCC staging system editions across treatment groups (Table 6). In both the 7th as the 8th editions, the AJCC staging system performed better in the subgroup of patients who underwent a curative-intent resection compared to the entire cohort (0.65 vs 0.57 in the 7th edition, 0.67 vs 0.58 in the 8th edition). The 8th edition did have a slightly better prognostic value compared to the 7th edition in this subgroup (c-index of 0.65 vs 0.67).
Although the prognostic accuracy of the 8th edition of the AJCC staging system in patients who did not undergo a resection was slightly better when compared to the 7th edition (0.54), the prognostic accuracy was still very poor with a cindex of 0.57 in both the main as expanded staging system.

Discussion
In our all-comer cohort, we found that the prognostic accuracy of cross-sectional imaging for patients presenting with PHC was comparable across the 7th and 8th AJCC staging systems (c-index 0.57 vs 0.58). The prognostic accuracy of the 8th edition was higher in patients who underwent a curative- Table 5 Cross-tabulation of the substages of the 7th and 8th editions of the American Joint Committee on Cancer (AJCC) staging system 8th edition IIIa IIIb IIIc IVa IVb Total   7th edition I/II  33  0  0  0  0  0  33  IIIa  0  78  0  0  0  0  78  IIIb  0  0  0  18  7  0  25  IVa  0  0  11  17  13  0  41  IVb  0  0  0  0  0  71  71  Total 33  78  11  35  20  71  248 Each row shows how many patients at a specific 7th edition stage transitioned to other stages according to the 8th edition. Numbers in red refer to patients who moved to a different stage from the 7th to the 8th edition  I/II  III  IV  Total   7th edition  I/II  33  0  0  33   III  0  96  7  103  IV  0  28  84  112  Total  33  124  91  248 Each row shows how many patients at a specific 7th edition stage transitioned to other stages according to the 8th edition. Numbers in red refer to patients who moved to a different stage from the 7th to the 8th edition  intent resection compared with those who did not (0.67 and 0.57). Although prognostic accuracy of the 8th edition in patients who did not undergo a curative-intent resection was slightly better than the 7th edition, the prognostic accuracy of the AJCC staging system in these patients was still poor with a c-index of 0.57. The 8th edition AJCC staging system included four major modifications (Table 1). These modifications resulted in reclassification of 53 (21.4%) patients with consideration of substages (e.g., stages IIIa and IIIb) and 35 (14.1%) patients considering only the major stages. However, these modifications and concomitant reclassifications failed to significantly improve its prognostic accuracy.
Other studies evaluated the prognostic accuracy of the 7th edition of the AJCC staging system. [12][13][14] However, TNM stages were based on pathological evaluation (pTNM) of the resected specimen, rather than evaluating cross-sectional imaging (cTNM) as was performed in the present study. These studies excluded most PHC patients, because only a minority of PHC patients is eligible for a curative-intent resection. A large study comparing the 6th and 7th editions of the AJCC staging system in a cohort of 306 patients who underwent a resection found similar prognostic accuracy for the 7th edition with a c-index of 0.59 using only the main stages and 0.54 using substages. 12 . A Japanese study evaluated the 7th edition of the AJCC staging system and proposed a modified system. 13 This modification was the basis for the modification in T stage implemented in the 8th edition of the AJCC staging system: Bismuth type IV tumors were no longer considered as T4 and T4 tumors were downstaged from stage IVA to IIIb. However, external validation showed that the modified model did not improve prognostic accuracy compared to the 6th and 7th editions of the AJCC staging system. 12 This is the first study to evaluate the 8th edition of the AJCC staging system for all patients with PHC, regardless of subsequent treatment. AJCC stages were assigned based on cross-sectional imaging (cTNM). Stage assignment based on pathological evaluation (i.e., pTNM) was not possible, because most patients with PHC have locally advanced or metastatic disease or are unfit to undergo major surgery and therefore do not undergo a resection. Nevertheless, this study has some limitations that should be mentioned. The TNM stage was determined on cross-sectional imaging in all patients with PHC, rather than using pathological examination of resected specimens. Vascular involvement and the biliary extent of the tumor are often difficult to determine on cross-sectional imaging. However, the AJCC staging system is specifically developed to apply on both cross-sectional imaging and pathological examination of all PHC patients. In future studies, we would like to compare clinical and pathological staging, which would require detailed pathological reporting.
Because most patients with PHC have locally advanced or metastatic disease at presentation (or are unfit for major surgery), the prognostic accuracy of AJCC staging system editions should be based on cross-sectional imaging rather than pathological evaluation. In addition, staging has the most potential clinical implications in the preoperative period, where it can still influence the decision whether to try and perform a resection or not. Accuracy on imaging is therefore arguably the most important parameter. Future editions of the AJCC staging system should aim to improve the prognostic accuracy of AJCC staging system on cross-sectional imaging.

Conclusions
The prognostic accuracy of the 8th edition of the AJCC staging system was similar to the 7th edition. Prognostic accuracy was particularly poor in the majority of PHC patients who did not undergo a resection. Future editions of the AJCC staging system should aim to improve the prognostic accuracy of AJCC staging system on cross-sectional imaging.
Author Contributions MPG: study concept and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, statistical analysis, critical revision of the manuscript for important intellectual content, and approval of final draft submitted.
SB: acquisition of data, analysis and interpretation of data, statistical analysis, critical revision of the manuscript for important intellectual content, and approval of final draft submitted.
JLAvV: acquisition of data, analysis and interpretation of data, critical revision of the manuscript for important intellectual content, and approval of final draft submitted. JdJ: acquisition of data, analysis and interpretation of data, critical revision of the manuscript for important intellectual content, and approval of final draft submitted.
WGP: acquisition of data, analysis and interpretation of data, critical revision of the manuscript for important intellectual content, and approval of final draft submitted.
MD: acquisition of data, analysis and interpretation of data, critical revision of the manuscript for important intellectual content, and approval of final draft submitted.
JNMIJ: study concept and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, critical revision of the manuscript for important intellectual content, and approval of final draft submitted.
BGK: study concept and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, critical revision of the manuscript for important intellectual content, and approval of final draft submitted.
FEJAW: study concept and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, critical revision of the manuscript for important intellectual content, and approval of final draft submitted.
Open Access This article is distributed under the terms of the Creative Comm ons Attribution 4.0 International License (http:// creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.