Abstract
Aim: We hypothesized that a nomogram can accurately predict overall survival (OS) for patients with intrahepatic cholangiocarcinoma (ICC) after hepatectomy. Materials and Methods: A nomogram to predict OS based was developed using data from 185 ICC patients who had undergone hepatectomy. The nomogram was evaluated by concordance index (C-index), as well as testing calibration of predicted OS with observed OS for both internal and external cohorts. Results: Ten clinicopathological independent factors for OS prediction were selected for use in the nomogram. For internal validation, the calibration curve for probability of OS showed good agreement between prediction by the nomogram and actual observation. In three external validation cohorts, the nomogram discrimination was also superior to two other staging systems. Conclusion: A nomogram integrating ten clinicopathological variables was developed that may assist in individual prognostic prediction of ICC after hepatectomy.
Abbreviations: ICC, Intrahepatic cholangiocarcinoma; LCSGJ, Liver Cancer Study Group of Japan; OS, overall survival; MF, mass forming; PI, periductal infiltrating; IP, intraductal papillary; HCC, hepatocellular carcinoma; AJCC, American Joint Committee on Cancer; ALT, alanine aminotransferase; AST, aspartate aminotransferase; CA 19-9, carbohydrate antigen 19-9; CEA, carcinoembryonic antigen; CT, computed tomography; C-index, concordance index.
Intrahepatic cholangiocarcinoma (ICC) originates from the epithelial cells of segmental or proximal branches of the bile duct (1). ICC is the second most common primary liver malignancy, accounting for 5% to 30% of all such malignancies (2, 3). The Liver Cancer Study Group of Japan (LCSGJ) proposed three gross pathological types of ICC, including the mass forming (MF), periductal infiltrating (PI) and intraductal papillary (IP) type (1). The incidence and mortality of ICC are steadily increasing worldwide (4). Hepatectomy has remained the curative treatment for ICC (5, 6). However, prognosis for hepatectomy is unsatisfactory due to a high incidence of locoregional recurrence and/or distant metastases (7-9). ICC differs from hepatocellular carcinoma (HCC) in carcinogenesis and biological behavior (10). ICC also differs from hilar and distal bile duct cholangio-carcinoma in clinical features, imaging manifestations and therapeutic strategies (11). Therefore, ICC needs a distinct prognostic predictive model for further formulation of treatment strategies.
The most commonly used staging system for ICC is the tumor-node metastasis (TNM) system. Okabayashi et al. (12) proposed a new staging system for the MF-ICC based on the TNM system, without including, however, the tumor's diameter. Yamasaki (13) reported that ICC larger than 2 cm had an inferior prognosis. Thus, 2 cm was regarded as a cut-off value for the T classification and adopted as an important parameter by the LCSGJ system (13). The 6th edition of the American Joint Committee on Cancer (AJCC) TNM staging system concludes that patients with ICC larger than 5 cm have a poorer prognosis; however, this is based on data for HCC. Thus, its applicability to ICC is uncertain (14, 15). Nathan et al. (16) proposed an improved TNM staging system to exclude the tumor diameter in the T classification. This opinion has been further adopted in the 7th edition of the AJCC TNM staging system (17). Lacking consensus on which staging system to use has led to considerable confusion.
Nomograms have been developed for most types of malignancy (18-20). Nomograms are comparable to traditional staging systems for many cancers and have been proposed as alternatives, or even as new standards (21-23). Although Wang et al. and Hyder et al. constructed prognostic nomograms for patients after partial hepatectomy for ICC (24, 25), several limitations of the study emerged. While Wang's study comprised mainly patients with diagnosis of MF-ICC, Hyder's study enrolled only minor portion of Asian-Pacific subjects (a pandemic area of hepatolithiasis and ICC) (26).
The present study aimed to establish a prognostic nomogram for resectable ICC and determine whether this model provides more accurate prediction of survival compared to the current commonly used staging systems. Furthermore, three external validation cohorts were used to increase the heterogeneity of patients and to validate their predictive power.
Materials and Methods
A retrospective study was conducted on a primary cohort of 224 patients who underwent hepatectomy for ICC between 1997 and February 2007 at Chang Gung Memorial Hospital, Linkou (L-CGMH), Taiwan. Inclusion criteria included the following: no history of previous anticancer therapy; no history of other malignancies; pathologically proven radical resection for tumor; and pathologically proven ICC. Exclusion criteria were as follows: hilar or extrahepatic ICC, including intrahepatic metastasis of extrahepatic ICC; adenocarcinoma of uncertain origin or probable metastatic tumor; histopathologically confirmed mixed-type primary liver cancer; and perioperative mortality.
From September 2007 to April 2010, an independent cohort of 54 patients (L-CGMH; n=54) who had undergone partial hepatectomy for ICC was prospectively studied to validate the nomogram. In addition to aforementioned group, the nomogram was also validated by two other external populations, including patients from the Taipei Veterans General Hospital (TVGH; n=67) and Chang Gung Memorial Hospital, Kaohsiung and Keelung branch (Kao/KL CGMH; n=38). The study was censored on June 31, 2013, and approved by the institutional ethics committee. (IBR No. 99-2886B for CN Yeh, YY Chen, KC Chiang and 2014-01-008BC for MH Chen).
Diagnosis and treatment. After a detailed history and complete physical examination, blood was taken for determining serum albumin, total bilirubin, alanine aminotransferase (ALT), aspartate aminotransferase (AST), γ-glutamyltransferase, α-fetoprotein, carbohydrate antigen 19-9 (CA 19-9) and carcinoembryonic antigen (CEA). Other routine investigations included chest X-ray, upper gastrointestinal (GI) endoscopy, abdominal ultrasound, contrast-enhanced computed tomography (CT) and/or magnetic resonance imaging, if needed.
A preoperative diagnosis of ICC was based on clinical and radiologic data, as well as elevated serum markers. Partial hepatectomy was carried out according to tumor diameter, location, presence or absence of cirrhosis and estimated volume of the future liver remnant after a consensus conference, with participation of a gastroenterologist and hepatobiliary surgeon. Liver resection was carried out based on Couinaud's segments, sectors and hemi-livers. Routine dissection of lymph nodes in the hepatoduodenal ligament and retropancreatic and/or para-aortic lymph nodes was performed. Direct invasions of adjacent structures, local extrahepatic metastases, as well as newly found intrahepatic nodules identified intraoperatively, were removed whenever possible. Hepaticojejunostomy was carried out in patients with tumors involving the primary and secondary biliary ducts.
Histopathologic study of the resected specimen was carried out independently by a pathologist of each Institute, respectively. All studies were reviewed by an experienced pathologist in our institute (TC Chen) (27). Pathologic features, such as tumor diameter, number, capsule, site, surgical margin, vascular invasion, lymph node metastasis and cirrhosis, were documented and the degree of cell differentiation was also determined.
Follow-up. Patients were followed up every 3 months in the first 2 years after surgery and every 3 to 6 months thereafter. At each of the follow-up visits, a detailed history and a complete physical examination were carried out. Blood was taken for tumor markers, including CA19-9, CEA and α-fetoprotein, for liver function tests; an abdominal ultrasound was also performed. Contrast-enhanced CT or magnetic resonance imaging was performed once every 6 months or earlier when tumor recurrence or metastasis was suspected. ICC recurrence or metastasis was defined as the appearance of a newly detected tumor, confirmed on two radiologic images, with or without elevation of serum tumor markers. Overall survival (OS) was defined as the interval between partial hepatectomy and death or the last date of follow-up.
Categorization of patients in different staging systems. Patients were categorized according to two staging systems (the 6th and 7th editions) of the AJCC TNM classification.
Statistical analysis. Statistical analyses were performed using SPSS 15.0 for Windows (SPSS, Chicago, IL, USA). Categorical variables were grouped and compared using the Chi-square test or Fisher's exact test. Continuous variables were compared using the t-test or Mann-Whitney U-test. Survival curves were depicted by the Kaplan-Meier curve and compared using the log-rank test. Cox regression analysis was used for multivariate analyses and to formulate a nomogram.
Algorithm for nomogram creation. ICC, intrahepatic cholangiocarcinoma; L-CGMH, Chang Gung Memorial Hospital, Linkou; TVGH, Taipei Veterans General Hospital; Kao/Kl-CGMH, Kaohsiung and Keelung Chang Gung Memorial Hospital.
A nomogram was depicted using R environment (version 2.14.1) with the rms package (http://www.r-project.org/). In order to make the nomogram practical, we used categorical variable based on the result of receiver operating characteristic (ROC) curve application.
The performance of the nomogram was measured by concordance index (C-index) and assessed by comparing nomogram predicted versus observed probability of survival (illustrated with a calibration curve). Bootstrapping with 1,000 resamples was used for validation. Comparisons between the nomogram and other staging systems were performed by computation of the C-index with the rcorr.cens package of R. The larger the C-index, the greater the accuracy of the power to differentiate between superior and inferior survival (28). All p-values<0.05 were considered statistically significant in this study.
Results
Clinicopathological characteristics of patients. There were 224 patients with ICC who underwent hepatectomy during the study period and 185 out of them met the inclusion criteria for nomogram creation. Patients who were excluded included: 11 who expired within one month, 16 with incomplete data and 12 lost to follow-up (Figure 1). As validation cohorts, we studied 54 consecutive patients treated in our Institute and 2 other groups (n=67 and 38, respectively). Table I summarizes the clinicopathological characteristics of both the training and validation group patients.
OS in the training cohort. The median follow-up time was 39.3 months (range=7.5-107.2). The median OS was 21.0 months (range=1.6-105.7) and the 1-, 3- and 5-year OS rates were 61.9%, 40.8% and 35.2%, respectively.
Independent prognostic factors in the training cohort. Table II summarizes the univariate analysis and Cox proportional analysis. Univariate analysis revealed 10 significant prognostic factors, including serum CEA, presence of symptoms, mucobilia, hepatolithiasis, tumor diameter, resection margin, gross pathology, tumor differentiation, vascular invasion and positive lymph node. Further Cox proportional analysis was performed to create the nomogram. Although some factors showed no significance in Cox analysis, all the input factors should be incorporated into the nomogram later as adjustment items (Table II and Figure 2).
Prognostic nomogram for OS. Figure 3 shows the prognostic nomogram that integrated significant factors for OS in the training cohort. The cut-off values for tumor size and CEA level were acquired based on the results of ROC curve evaluation. The C-index for OS prediction was 0.76. The calibration plot for the probability of survival at 3 and 5 years after surgery showed an optimal agreement between the prediction by the nomogram and actual observation (Figure 4A and B). The calibration plot of the external cohort for the probability of survival at 3 years after surgery also showed satisfactory compatibility (Figure 4C).
Clinicopathological features of intrahepatic cholangiocarcinoma patients in training and validation cohorts who underwent hepatic resection.
Comparison of predictive accuracy for OS between nomogram and conventional staging systems. As shown in Table III, our nomogram displayed better accuracy in predicting OS in the training cohort. The C-index of the nomogram was 0.76, which was significantly higher than both the AJCC 6th and 7th edition staging systems (0.63 and 0.68, respectively).
Validation of predictive accuracy of the nomogram for OS. In the external validation cohort from our institute (L-CGMH), our nomogram still displayed better accuracy in predicting OS than both the AJCC 6th and 7th edition staging systems (the C-index of the nomogram was 0.69 and 0.63, 0.66 for the AJCC 6th and 7th edition, respectively). In the other two external validation cohorts from other institutes (TVGH and Kao/KL CGMH), our nomogram still displayed better accuracy in predicting OS. For the TVGH cohort, the C-index of our nomogram was 0.64, which was superior to the AJCC 7th (0.59) and 6th edition staging systems (0.60) (p<0.001). For the Kao/KL CGMH cohort, the C-index of our nomogram was 0.79, which was significantly higher than the AJCC 7th (0.64) and 6th edition staging systems (0.67) (p<0.001).
Discussion
It may not be desirable to simply stratify risk and outcome in ICC according to current TNM systems. In addition to those in the AJCC staging systems, our prognostic nomogram comprises of several important factors related to prognosis, including tumor diameter. Additional risk factors, other than the items in the TNM system, such as tumor markers, hepatolithiasis, mucobilia and resection margin, and currently known gross ICC types are also considered.
Kaplan-Meier curves for significant factors derived from univariate analysis. CEA, Carcinoembryonic antigen.
Univariate analysis and Cox proportional analysis.
Nomogram based on a training cohort with 185 cholangiocarcinoma patients after hepatectomy.
Comparison of concordance index between variable systems.
In several staging systems, the relationship between tumor diameter and number, as well as survival of patients with ICC after hepatectomy, remains controversial. This controversy is clearly evident in the evolution from the 6th to 7th AJCC TNM systems. Neither tumor size nor number were considered in the 7th AJCC system. In our study, tumor size, but not tumor number, reflected the invasiveness of ICC and was significantly associated with prognosis (Table II).
Our nomogram also includes comprehensive laboratory indices, such as serum tumor markers, which have not been included as parameters in the other staging systems. Serum CA 19-9 and CEA levels have been suggested to be independent risk factors for prognosis of cholangiocarcinoma, including ICC (29, 30). Although tumor markers have not been included in ICC staging systems, their role in increasing predictive performance has been observed in HCC staging systems (31, 32). Furthermore, a nomogram always contains more prognostic variables than traditional staging systems (33). CA 19-9 and CEA levels have been reported to be useful for diagnosing ICC in primary sclerosing cholangitis (34-36). Serum levels of CA 19-9 are correlated to tumor burden and unresectability for ICC, indicating a dismal prognosis (10, 37, 38). In our study, CEA was suggested as a prognostic factor in ICC patients after hepatectomy; however, CA19-9 was not. According to evidence from literature and our study, it may be reasonable to include tumor markers in a nomogram.
Calibration curves with nomogram predicted survival and actually observed survival. (A) For 3-year survival and (B) 5-year survival of training cohort. (C) For 3-year survival of 3 external validation cohorts.
Additional important risk factors, including presenting symptoms, mucobilia and hepatolithiasis, have been incorporated in the prognostic nomogram. Patients diagnosed with asymptomatic ICC incidentally showed significantly better survival, although early detection of ICC is challenging. Mucobilia occurs in various conditions, including biliary papillomatosis, cholangiocarcinoma and biliary cystadenoma and cystadenocarcinoma of the liver (39-42), with the presence of mucobilia being an important clue to the early diagnosis of ICC (43). We previously demonstrated the association of hepatolithiasis with all common types of ICC (44). ICC with hepatolithiasis is most likely the cumulative result of many possible etiological influences (31, 45-47). Moreover, the presence of hepatolithiasis could be regarded as an important indication of unfavorable prognosis, possibly related to impaired hepatic function due to long-term inflammation in the liver. Hepatic resection is the preferred treatment for ICC (6, 30, 48). We further reported that an absence of curative hepatic resection in ICC patients reduced the long-term OS rate; thus, it is reasonable to include margin status in the nomogram.
Even Wang et al. (24) reported prognostic nomogram for ICC patients after partial hepatectomy, MF type ICC comprised of 94.0% of his study cohort, while the impact of other gross pathological types of ICC could not be determined. The performance of the nomogram for prognostic prediction in MF type ICC was separately assessed and demonstrated its predictive power with a higher C-index,; however, gross pathological classification, including only MF-ICC, is obviously inadequate for clinical use. Our nomogram may provide a solution for the drawback of their model.
Similar to Wang's study, we enrolled patients with a history of hepatitis B virus (HBV) infection, including some who were anti-hepatitis C virus (HCV)-positive. We previously reported that viral hepatitis B and C might be independent risk factors for ICC in Taiwan (49), confirmed by some studies in Japan (50). However, our study revealed that underlying HBV and HCV infection is not important to the prognosis in ICC after hepatectomy. Large-scale studies may be needed.
Although our nomogram showed higher predictive accuracy for survival than the current AJCC systems, several limitations still exist. First, the nomogram was based on patients from a single institution and validated by three external cohorts all of which were Asians. Primary sclerosing cholangitis (not considered in this study) is important in carcinogenesis of cholangiocarcinoma, especially in Western countries. Whether this staging system is applicable to Western patients is still unclear. Second, an assessment of no metastasis was based on the absence of lymphadenopathy on imaging and surgical exploration. Although enhanced CT or PET can have a high negative predictive value (nearly 99%) in patients without lymphadenopathy as our criteria have been adopted by other studies, this is still a limitation that might affect the results. Third, the tumor diameter of some cases was not easy to measure, especially for PI type. Although collaborative pathologists meticulously had evaluated the extension of tumor, it undoubtedly presents a limitation. Finally, whether this nomogram can be applied to patients not receiving partial hepatectomy remains to be determined.
In conclusion, the nomogram proposed in this study accurately predicted the prognosis of ICC patients after partial hepatectomy.
Footnotes
↵* These Authors contributed equally to this study.
- Received May 30, 2016.
- Revision received June 19, 2016.
- Accepted June 21, 2016.
- Copyright© 2016 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.