Risk stratification system to predict recurrence of intrahepatic cholangiocarcinoma after hepatic resection

Intrahepatic cholangiocarcinoma (ICC) is the second most common primary liver cancer accounting for 15 to 20% of all primary liver malignancies [ 1]. It is a relatively rare disease with dismal prognosis due to frequent recurrence after surgical operation, lymph node metastasis, and rapid progression of the tumor, and 5-year mortality of patients who received surgical treatments remains higher than 50% [ 2]. Therefore, the interest for ICC has been rising since the incidence and mortality have been increasing [ 3].

To date, hepatic resection remains the most frequent option for patients with ICC to prolong their survival. However, numerous patients present with unresectable ICC at the time of diagnosis due to lack of typical clinical manifestation and rapid progression of the tumor [ 4]. Moreover, patients with underlying liver diseases, such as liver cirrhosis, are very common where hepatitis B virus (HBV) infection is endemic, so that a majority of patients have no chance to receive hepatic resection. To date, several evaluation procedures for resection range in cirrhotic patients have already been developed and proved to be effective. However, another important factor that needs to be focused is recurrence of the tumor which significantly reduce survival outcomes. Therefore, in order to identify who should be treated by hepatic resection, it is vital to comprehensively confirm factors that are associated with recurrence of the tumor.

Precise prediction of prognosis is important for patients as well as physicians to consider postoperative management. According to the seventh American Joint Committee on Cancer (AJCC) cancer staging manual, the staging system for ICC was separated from hepatocellular carcinoma [ 5]. In spite of the modification, the predictive accuracy of conventional AJCC staging system was found to be lower in Eastern population which was validated by Chinese nomogram (concordance index [C-index]: 0.65 vs. 0.74) [ 6]. However, the nomogram was established on the overall survival of patients, indicating that the nomogram could be influenced by a number of confounding factors rather than the recurrence of the tumor, rendering us to interpret that currently available nomograms are insufficient to accurately predict relapse of ICC after hepatic resection on the basis of the Eastern population. Therefore, we conducted our nomogram to speculate risk index to precisely predict the recurrence of ICC based on our single center experience.

Most patients with ICC present with advanced disease at the time of diagnosis due to lack of typical clinical manifestations in early stages. Despite surgical approaches, recurrence is more frequent than other malignancies, such as hepatocellular carcinoma, resulting in much more less than a half of patients achieving favorable long-term prognosis [ 10]. Current perspectives on stratifying patients into stages mainly focus on TNM factors which were proven to significantly influence prognosis of patients with ICC [ 11]. However, some recent studies indicated that additional risk factors, such as HBV infection, were also found to be significantly associated with prognosis of patients with ICC after hepatic resection [ 12, 13]. In the present study, we performed univariate analyses on all potential factors that might have affected recurrence of the tumor and included all the variables that showed significant association with the recurrence of the tumor in developing the nomogram, and demonstrated excellent accuracy in predicting recurrence of ICC after hepatic resection (C-index, 0.71; 95% CI, 0.65 to 0.77). Furthermore, unlike previous nomograms that predicted overall survival outcomes, the purpose of our nomogram was to precisely stratify patients according to risk index for recurrence of ICC and exclude irrelevant factors, such as surgical complications, which would lead to unfavorable survival outcomes [ 6, 14].

The earliest nomogram for ICC was developed in 2013 by Wang et al. [ 6] on 367 patients with ICC who received partial hepatectomy from 2002 to 2007. Their nomogram was based on the independent factors for overall survival, including serum level of CA19-9, tumor size, number of tumor, vascular and direct invasion, and lymph node and local extrahepatic metastasis. The C-index for the prediction of survival was 0.74, which was found to be higher than AJCC, Nathan, Liver Cancer Study Group of Japan, and Okabayashi staging systems, thereby proving prognostic accuracy of nomogram. The second nomogram was built on basis of 514 patients from international 13 major hepatobiliary centers [ 14]. The nomogram included age at the time of diagnosis, number and size of tumor, presence of cirrhosis, nodal status, and vascular invasion, which is an initial nomogram that included additional factors rather than TNM factors. However, they put the presence of underlying cirrhosis into the nomogram as an important factor, regardless of its no significant association with survival outcomes in the univariate analysis ( P = 0.19) and the highest HR of 2.10 (95% CI, 1.46 to 3.01) was found in macrovascular invasion. More recently, Doussot et al. [ 15] validated the above two nomograms in 188 patients with a median follow-up period of 41 months and reported that both nomograms provided accurate estimation of prognosis (C-index, 0.72 and 0.66), indicating that current nomograms are sufficient to predict overall survival of patients with ICC. In the present study, our data revealed that seropositivity of HBsAg is associated with decreased incidence of recurrence. In addition, tumor size larger than 5 cm, Child-Pugh score of B, and the presence of lymph node metastasis at the time of surgery were associated with more frequent relapse of the tumor. Our nomogram comprehensively included all the above mentioned independent prognostic factors related to the recurrence of ICC, which have not been included in any other staging systems or nomograms.

There has been a controversy on survival outcomes between patients with HBV-associated ICC and patients without HBV infection [ 16– 20]. Several studies demonstrated that HBV infection is a favorable prognostic factor in patients with ICC who underwent hepatic resection [ 17]. Unfortunately, the presence of HBV infection was excluded in Wang’s nomogram. On the contrary, Tao et al. [ 21] indicated that HBV infection negatively influenced prognosis of patients with ICC. Another publication from Korea showed that there was no impact of HBV infection on survival outcomes of patients with ICC, and they suggested that early diagnosis of unexpected ICC in HBV carriers might result in relatively favorable outcomes [ 22]. However, most recently, Wang et al. [ 23] found that HBV-associated ICC patients achieved significantly better survival outcomes than the patients with hepatolithiasis-associated ICC by analyzing 731 patients with ICC undergoing R0 hepatic resection, and suggested that HBV infection might influence biological malignant invasiveness of the tumor rather than stages of ICC. Comprehensively, a recent meta-analysis demonstrated that HBV infection is a favorable prognostic factor that significantly improves survival outcomes in patients with ICC [ 24]. The meta-analysis also revealed that HBV infection was associated with underlying cirrhosis (odds ratio [OR], 6.44; 95% CI, 4.33 to 9.56), capsule formation (OR, 6.04; 95% CI, 3.56 to 10.26), and infrequent lymph node metastasis (OR, 0.39; 95% CI, 0.25 to 0.58). Most recently, Hiroya et al. [ 25] indicated that the absence of viral hepatitis was associated with more frequent incidence of lymph node metastasis in ICC, which led to poor prognosis. However, their study did not separate HBV and HCV, due to limited number of patients ( n = 32). As shown in our previous study, we preliminarily hypothesized that decreased lymph node metastasis of HBV-associated ICC might be derived from inhibited mesenchymal-to-lymphatic endothelial transition of cancer-associated fibroblasts in stromal areas, followed by reduced number of microlymphatic vessels participating in cancer-associated lymph node metastasis [ 26]. However, current biological evidences are insufficient to confirm the existence of mesenchymal-to-lymphatic endothelial transition in ICC, which awaits our future research.

The nomogram also contained Child-Pugh score to be an important factor in the estimation of recurrence. This scoring system is composed of ascites, bilirubin, albumin, prothrombin time, and encephalopathy that were considered to be crucial in evaluating liver reserving functions of patients with chronic liver diseases as well as HCC [ 27]. In previous years, the application of Child-Pugh score to ICC was not established in our center due to the following reasons: (1) liver dysfunction of patients with ICC is partially attributed to the tumor lesions, (2) cirrhosis is uncommon in typical ICC, and (3) subjective evaluation of ascites and encephalopathy. In the present study, merely 6 patients were found to be Child-Pugh score of B and no patient manifested encephalopathy, thereafter actually only 4 variables were included. On the other hand, a previous case-control study of 160 patients with ICC and 2498 patients with HCC by Lee et al. [ 28] indicated that HBV-associated ICC shared common disease process with HCC in terms of carcinogenesis. In addition, our previous publication indicated that HBV-associated ICC is a different entity with distinctive clinicopathological characteristics, including preponderance of male and younger patients, frequent elevation of serum aspartate transaminase and AFP, and a relatively lower level of CA19-9 [ 26]. Therefore, as our results indicated, the Child-Pugh scoring system needs to be applied to more precisely predict recurrence of ICC.

However, there are a few underlying limitations that remain to be further confirmed. Firstly, limited number of patients might need an additional modification by future large-scaled studies. Secondly, all the involved patients were from Eastern Asia, so the nomogram needs to be further validated by patients from other regions. Thirdly, specific region of lymph node metastasis was not collected due to retrospective collection of data. Despite the above limitations, our nomogram represents the first stratification system to predict recurrence of ICC after hepatic resection on the basis of Eastern population.

The nomogram could precisely stratify the patients according to the risk index for the recurrence of ICC by applying seropositivity of HBsAg, tumor size, Child-Pugh score, and lymph node metastasis. The accuracy of the nomogram revealed to be higher than the AJCC staging system in both the study set (C-index, 0.71 vs. 0.66) and the validation set (C-index, 0.65 vs. 0.57). According to the retrospective nature and limited number of patients, our risk stratification system calls for modification by large-sized studies.