v
Search
Advanced Search

Publications > Journals > Journal of Clinical and Translational Hepatology > Article Full Text

  • OPEN ACCESS

A New Prognostic Model Covering All Stages of Intrahepatic Cholangiocarcinoma

  • Shuang-Nan Zhou1,# ,
  • Shan-Shan Lu2,3,#,
  • Da-Wei Ju4 ,
  • Ling-Xiang Yu2,
  • Xiao-Xiao Liang2,5,
  • Xiao Xiang6,
  • Suthat Liangpunsakul7,8 ,
  • Lewis R. Roberts9 ,
  • Yin-Ying Lu2,* and
  • Ning Zhang2,* 
 Author information
Journal of Clinical and Translational Hepatology   2022;10(2):254-262

doi: 10.14218/JCTH.2021.00099

Abstract

Background and Aims

Intrahepatic cholangiocarcinoma (ICC) is the second most common primary hepatic malignancy that causes a poor survival. We aimed to identify its prognostic factors and to develop a nomogram that will predict survival of ICC patients among all stages.

Methods

A total of 442 patients with pathology-proven ICC registered at the Fifth Medical Center of PLA General Hospital between July 2007 and December 2019 were enrolled. Subjects were followed for survival status until June 30, 2020. A prognostic model visualized as a nomogram was constructed in the training cohort using multivariate cox model, and was then validated in the validation cohort.

Results

The median age was 55 years. With a median follow-up of 50.4 months, 337 patients died. The median survival was 11.6 months, with 1-, 3- and 5-year survival rates of 48.3%, 22.7% and 16.2%, respectively. Factors associated with overall survival were multiple tumors, lymph node involvement, vascular invasion, distant metastasis, decreased albumin, elevated lactate dehydrogenase (LDH), decreased iron, elevated fibrinogen, elevated CA125 and elevated CA19-9. A nomogram predicting survival of ICC patients at the time of diagnosis achieved a Harrel’s c-statistic of 0.758, significantly higher than the 0.582 of the TNM stage alone. Predicted median survivals of those within the low, mid and high-risk subgroups were 35.6, 12.1 and 6.2 months, respectively.

Conclusions

A nomogram based on imaging data and serum biomarkers at diagnosis showed good ability to predict survival in patients with all stages of ICC. Further studies are needed to validate the prognostic capability of our new model.

Keywords

Intrahepatic cholangiocarcinoma, Prognostic model, Nomogram, Risk stratification

Introduction

Intrahepatic cholangiocarcinoma (ICC), a subgroup of cholangiocarcinoma, the second most common malignancy arising from the liver, originates from the peripheral bile ducts within the liver parenchyma, proximal to the secondary biliary radicals.1 The incidence of ICC has been increasing globally.2 Surgical treatment is the only potentially curative therapy in ICC, although many patients are not eligible for resection because of locally advanced or metastatic disease;1 even for the eligible ones, the prognosis is unsatisfactory, with a high recurrence rate.3 The 5-year overall survival rate is reportedly less than 5%4 and the 5-year survival after resection has been reported in the range of 22–44%.5 As such, an upfront comprehensive assessment of the prognosis is essential for the clinical decision of multidisciplinary treatment.

Unlike hepatocellular carcinoma, there is no internationally recommended staging-guided roadmap for the treatment of ICC. The American Joint Committee on Cancer (AJCC) TNM staging system is widely used, but it performs relatively poorly in differentiating between patients with various prognoses, with substantial inter-patient differences in the survival even among those within the same TNM stages.6,7 In addition to the AJCC system, several prognostic models or nomograms have been developed7–17 and additional prognostic factors have been reported to predict overall survival of ICC patients, such as levels of C-reactive protein,13 hepatoma-derived growth factor,18 DNA index,19 and Homer1.20

Most of the published prognostic nomograms for ICC were established based on the patients undergoing surgical resection.7–16 The resulting models rely on the parameters from surgical and pathological reports to predict the outcome post-hepatic resection, thus limiting the use of the models in the majority of patients who are non-surgical candidates.7–16 Given the lack of an ideal treatment method for ICC, the outcome of the disease depends on multidisciplinary collaboration. Therefore, a prognostic model based on pre-treatment parameters can help doctors and patients make the optimal choice, which needs to meet the following points: covering the population at all stages, instead of just the patients after surgery; involving common and easily accessible parameters in clinical practice; and long follow-up time, to obtain more endpoint events and predict a longer survival outcome. To address these key points, we explored novel prognostic factors and developed a nomogram to predict survival probability of ICC patients using variables available at the time of diagnosis and determining its risk stratification to guide optimal patient management.

Methods

Study cohort

During the study period (July 2007 and July 2019), a total of 1,240 in-hospital patients with suspected diagnosis of ICC were registered at the Fifth Medical Center of Chinese PLA General Hospital Beijing, China. We excluded patients who had imaging compatible with ICC but without pathological diagnosis (n=410), pathological diagnosis of ICC performed at an outside hospital (n=90), patients with peri-hilar (n=112) or distal (n=45) cholangiocarcinoma, patients with diagnosis of mixed or combined hepatocellular carcinoma-cholangiocarcinoma (n=68), and patients with gallbladder carcinoma (n=1). In addition, we also excluded patients who had died within 1 month after resection (n=5) and patients who were seen only at baseline without long-term follow up (n=67). Thus, a total of 442 patients with pathology-proven ICC constituted our study cohort, and the treatment modalities were decided by multidisciplinary consultation. The study was approved by the Ethics Review Committee at the Fifth Medical Center.

Data collection

All patients were evaluated with a baseline demographic history, physical examination, concurrent comorbidities, serum laboratory tests, and cross-sectional imaging either by contrast computed tomography or magnetic resonance imaging of the abdomen and pelvis. Tumor characteristics were abstracted from the imaging, including tumor size, tumor number, evidence of vascular invasion, and lymph node involvement. Information regarding treatment-related variables, such as type of therapy, and surgical and pathological features (for patients who underwent surgical resection; including nodal status, margin, and vascular invasion) were recorded. Tumor boundary type (distinct vs. obscure) was determined using the Fudan score.10 Treatment modalities were classified into four groups based on the primary therapy, namely surgery, surgery+regional therapy, regional therapy, systemic therapy and palliative care. All patients were followed up at the Fifth Medical Center for survival outcomes, with the last date of follow-up being June 30, 2020.

Statistical analysis

Baseline characteristics were described. Continuous variables were summarized using median and interquartile range, while categorical variables were summarized using n and percentage. Median overall survival (OS) was estimated using the Kaplan-Meier method, and differences were tested using the log-rank test. Median follow-up time for OS was estimated using the reverse Kaplan-Meier method. Predictors with greater than 20% missing values were dropped from analysis.

The patients were randomly assigned to a training cohort (n=342) and a validation cohort (n=100) using a random seed of 20210130. The modeling was constructed in the training cohort and different models were then compared in the validation cohort. The model with highest performance in the validation cohort was then chosen to build the prognostic score. The final model was then estimated for the whole dataset.

The prognostic model was constructed using a two-stage method in the training cohort. In the first stage, the predictors were tested in a univariate cox regression model. The univariate p values were adjusted by the Benjamini-Hochberg method to maintain the false discovery rate (FDR) under a prespecified threshold (i.e. 0.10 and 0.01). In the second stage, predictors that had passed the FDR threshold were included in a multivariate cox regression model using stepwise variable selection based on Akaike information criterion or Bayesian information criterion. A naïve model included TNM staging as the only classifier and was used as benchmark. Taken together, there were five candidate models in the training step.

With the models obtained in the training step, the validation cohort was used to compare the candidate models to select the one with the best performance. The Harrel’s c statistics as well as the area under the curve (AUC) by different landmark timepoints (6, 12, 18, 24, 30, 36 months) were calculated as the benchmark. The model with the highest Harrel’s c was then chosen and re-trained in the pooled dataset to obtain the final model.

The model was calibrated using a bootstrap resampling method to construct the 95% confidence interval (CI) of the predicted survival rate at 12 months and was then plotted against the observed 12-month survival rate.

The prognostic score was constructed from the final model in the pooled dataset. The patients were classified into low, mid or high-risk subgroups based on their prognostic score. Ultimately, a nomogram visualizing the prognostic score was provided. All the above analyses were performed by R, version 4.0.0.

Results

Baseline demographic and clinical characteristics of the study cohort

Detailed demographic, clinical and tumor characteristics are shown in Table 1. The median age of patients in our study cohort was 55 years-old and 69.5% were men. Due to the nature of the random split method, there was no notable difference nor any variation between the training and validation cohorts. The leading causes of underlying liver diseases were hepatitis B virus (HBV) infection alone (38.2%), HBV infection and alcohol abuse (12.4%), and alcohol abuse alone (9.5%). The clinical features of HBV-infected patients are shown in Supplementary Figure 1 based on the status of hepatitis B e antigen (HBeAg), HBV DNA, transaminase, etc. Most of the patients showed negativity for HBeAg (n=177, 79%). Only a small portion of patients (n=44, 19.6%) were treated by the nucleoside/nucleotide analogues and diagnosed as ICC during the routine follow-up, and most were admitted to the hospital because of the liver nodules instead of the HBV infection.

Table 1

Demographics and baseline characteristics

VariablesTraining cohort (n=342)Validation cohort (n=100)Total (n=442)
Age in years55 (48–61)54.4 (47–60)55 (48–61)
Sex-male, n (%)240 (70.2%)67 (67%)307 (69.5%)
Etiologies of liver diseases, n (%)
  HBV129 (37.7%)40 (40%)169 (38.2%)
  HBV + alcohol42 (12.3%)13 (13%)55 (12.4%)
  Alcohol33 (9.6%)9 (9%)42 (9.5%)
  Nonalcoholic fatty liver disease20 (5.8%)8 (8%)28 (6.3%)
  HCV9 (2.6%)1 (1%)10 (2.3%)
  HCV + alcohol1 (0.3%)1 (1%)2 (0.5%)
  Liver fluke1 (0.3%)0 (0%)1 (0.2%)
  Previous HBV infection (HBsAg-, anti-HBc+)66 (19.3%)19 (19%)85 (19.2%)
  Biliary disease18 (5.2%)10 (10%)28 (6.3%)
  ANA+3 (0.9%)0 (0%)3 (0.7%)
  AMA+0 (0.0%)1 (1%)1 (0.2%)
  Unknown (without any tendency)38 (11.1%)14 (14%)52 (11.8%)
Cirrhosis, n (%)177 (51.8%)50 (50%)227 (51.4%)
MELD score7 (6–8)6 (6–8)6 (6–8)
Child-Pugh score6 (5–7)5 (5–7)6 (5–7)
BMI in kg/m224.0 (21.8–26.0)25.1 (22.1–26.8)24.1 (21.9–26.2)
Comorbidities, n (%)
  Diabetes48 (14%)14 (14%)62 (14%)
  CAD9 (2.6%)2 (2%)11 (2.5%)
  Hypertension63 (18.4%)24 (24%)87 (19.7%)
Laboratory data
  ALB in g/L39 (36–42)39 (37–41)39 (36–42)
  Bilirubin in mg/dL0.8 (0.6–5.9)0.9 (0.6–7.5)0.8 (0.6–6.5)
  GGT in U/mL75 (35–165)67 (35.5–178.5)74 (37–164)
  Alkaline phosphatase in U/mL115 (50.8–180)109.5 (80.5–161.2)115 (89–174)
  Creatinine in mg/dL0.8 (0.7–0.9)0.8 (0.7–0.9)0.8 (0.7–0.9)
  LDH in U/L193 (162–228)197.5 (167.8–280.2)194.5 (163.0–174.0)
  Iron in µmol/L14.3 (9.2–18.6)14.1 (9.8–20.9)14.3 (9.3, 19.2)
  Cholesterol in mmol/L4 (3.4–4.8)4 (3.5–4.8)4.0 (3.4–4.8)
  CA125 in U/mL25.4 (12.8,72.4)21.9 (12.8–68.3)24.5 (12.8–72.2)
  CA19-9 in U/mL39.4 (15.5–24.7)37.2 (12–408.8)38.2 (14.7–260.8)
  CA724 in U/mL2.0 (1.2–4.4)2.2 (1.1–5.3)2.0 (1.2–4.7)
  CEA in U/mL2.3 (1.5–6.2)3.4 (1.5–6.3)2.4 (1.5–6.3)
  AFP in ng/mL4.3 (2.5–10.6)4.3 (2.5–13)4.3 (2.5–11.0)
Tumor characteristics, n (%)
Number of lesions
  Solitary209 (61.1%)62 (62%)271 (61.3%)
  Multiple, ≥2131 (38.3%)38 (38%)169 (38.2%)
Max-diameter in cm6.0 (4.5–8.5)5.4 (3.1–8.6)6.0 (4.0–8.5)
Max-diameter in cm
  <5 cm98 (28.7%)42 (42%)140 (31.7%)
  5–10 cm190 (55.6%)41 (41%)231 (52.3%)
  >10 cm50 (14.6%)17 (17%)67 (15.2%)
Vascular invasion139 (40.6%)50 (50%)189 (42.8%)
Lymph node involvement166 (48.5%)38 (38%)204 (46.2%)
Distal metastasis39 (11.4%)16 (16%)55 (12.4%)
Tumor differentiation
  Poor88 (25.7%)25 (25%)113 (25.6%)
  Moderate196 (57.3%)65 (65%)261 (59.0%)
  High7 (2%)1 (1%)8 (1.8%)
AJCC 8th TNM stage
  IA40 (11.7%)10 (10%)50 (11.3%)
  IB59 (17.3%)11 (11%)70 (15.8%)
  II158 (46.2%)54 (54%)212 (48.0%)
  IIIA5 (1.5%)2 (2%)7 (1.6%)
  IIIB38 (11.1%)7 (7%)45 (10.2%)
  IV18 (5.3%)9 (9%)27 (6.1%)
  IVB21 (6.1%)7 (7%)28 (6.3%)
Principal treatment modality, n (%)
  No treatment29 (8.5%)13 (13%)41 (9.3%)
  Surgery80 (23.4%)20 (20%)100 (22.6%)
  Surgery + regional therapy85 (24.9%)24 (24%)109 (24.7%)
  Regional therapy124 (36.3%)38 (38%)163 (36.9%)
  Systemic therapy24 (7.0%)5 (5%)29 (6.5%)
Nomogram for prediction of survival probability at 12, 24 and 36 months.
Fig. 1  Nomogram for prediction of survival probability at 12, 24 and 36 months.

To use the nomogram, find the position of each parameter on the corresponding axis, draw a line to the point axis at the top for the number of points, add the total points from all of the variables, and draw a line from the total point axis to determine the OS probability at different time points at the bottom of the nomogram.

It is noteworthy that 19.2% of patients were judged as previous HBV infection-positive anti-hepatitis B core protein (HBc) and -negative hepatitis B surface antigen (HBsAg) in serum, which may have contributed to the occurrence of ICC. Unlike the etiology of ICC in the West, no primary sclerosing cholangitis or primary biliary cholangitis were clearly diagnosed in our cohort. Twenty-eight of the patients had a history of biliary diseases, such as cholelithiasis and gallbladder polyps, and a history of cholecystectomy. Underlying cirrhosis was found in 51.4% of our study cohort, as assessed by the pathological report. The median serum levels of total bilirubin, albumin (ALB), carcinoembryonic antigen (CEA), carbohydrate antigen 19-9 (CA19-9), CA125, and LDH were 0.8 mg/dL, 39 g/L, 2.4 U/mL, 38.2 U/mL, 24.5 U/mL, and 194.5 U/L, respectively.

Regarding tumor characteristics, 61.3% of patients had a solitary tumor with a median diameter of 6.6 cm. Vascular invasion and lymph node involvement were found in 42.8% and 46.2% of patients, respectively. Twelve percent of patients had evidence of distant metastasis. The majority of patients had earlier stage (75% in stage I to II) tumors, based on the AJCC staging classification (8th edition). Most patients received regional therapy or surgery. The type of regional therapy and the proportions of patients who underwent such are listed in Supplementary Table 1. For the patients who received surgery and regional therapy, transcatheter arterial chemoembolization was the most frequently used regional therapy, accounting for 77.1%. For the patients who received regional therapy only, argon-helium cryoablation combined with intratumoral ethanol injection was the most frequently used modality, accounting for 45.4%, followed by microwave hyperthermia (28.3%) and transcatheter arterial chemoembolization (25.2%).

Model training and validation

Within the median follow-up of 50.4 months, 337 (76%) patients died. The median survival was estimated as 11.6 months (95% CI: 9.9–13.3 months). The 1-, 3- and 5-year survival rates were 48.3%, 22.7% and 16.2%, respectively.

Under the different FDR thresholds of 0.1 and 0.01, a total of 38 and 21 predictors were selected in the univariate stage. The number of selected variables in multivariate cox regression, as well as the model performance are listed in Supplementary Table 2. Among the four candidate models, model 2 with an FDR threshold of 0.01 and the Akaike information criterion showed the highest performance. The Harrel’s c statistic of that model was 0.758 and time-dependent AUC of 0.783, 0.732, 0.740, 0.758, 0.758 and 0.744 for 6, 12, 18, 24, 30 and 36-month OS rate (Supplementary Fig. 2B).

Table 2

Cox proportional hazards regression model for the independent predictors of survival in the training cohort (n=342)

VariablesUnivariate analyses
Multivariate analyses
Hazard ratio (95% CI)pFDRHazard ratio (95% CI)p
Multiple tumors2.01 (1.63, 2.68)<0.0001<0.00011.38 (0.98, 1.93)0.0624
Tumor diameter, per 5 cm1.99 (1.67, 2.34)<0.0001<0.0001
Lymph node involvement2.20 (1.72, 2.82)<0.0001<0.00011.82 (1.33, 2.50)0.0002
Pathology lymph node invasion1.89 (1.34, 2.59)0.00010.0003
Distant metastasis2.24 (1.57, 3.20)0.00120.00381.79 (1.10, 2.91)0.0190
White blood cell count1.09 (1.04, 1.14)0.00050.0015
Neutrophil-lymptocyte ratio1.04 (1.02,1.06)0.00050.0015
Red blood cell count0.69 (0.56, 0.85)0.00040.0013
Hemoglobin, per 10 g/L0.88 (0.83, 0.94)0.00010.0003
ALB, per 10 g/L0.55 (0.44, 0.70)<0.0001<0.00010.70 (0.48, 1.03)0.0714
GLO, per 10 U/mL1.46 (1.17, 1.81)0.00070.0019
Creatinine0.20 (0.09, 0.47)0.00020.0007
LDH, per 100 U/mL1.16 (1.12, 1.22)<0.0001<0.00011.45 (1.16, 1.80)0.0009
Creatine kinase, per 50 U/m0.70 (0.58, 0.84)<0.00010.0001
Iron, per 10 umol/L0.58 (0.48, 0.71)<0.0001<0.00010.76 (0.58, 1.00)0.0496
Apoprotein A10.35 (0.21, 0.60)0.00010.0003
FIB1.09 (1.05, 1.12)<0.0001<0.00011.05 (1.00, 1.11)0.0509
CA125, per 100 U/mL1.08 (1.05, 1.12)<0.0001<0.00011.04 (0.99, 1.09)0.1114
CA19-9, per 100 U/mL1.04 (1.04, 1.07)0.00010.00031.07 (1.02, 1.11)0.0029
CEA, per 100 U/mL1.36 (1.18, 1.56)<0.00010.0001
TNM-8: Stage IB-IIIA vs. Stage IA2.01 (1.32, 3.08)0.00120.0032
TNM-8: Stage IIIB-IV vs. Stage IA3.86 (2.42, 6.17)<0.0001<0.0001
Calibration plot for the new nomogram associated with the prediction of 12-month survival.
Fig. 2  Calibration plot for the new nomogram associated with the prediction of 12-month survival.

By applying model 2 to the pooled dataset, the variables selected by the final model were multiple tumors, lymph node involvement, distant metastasis, ALB, LDH, iron, fibrinogen (FIB), CA125 and CA19-9. Note that the estimate of final model is slightly different from model 2 in Table 2. This is because the final model, with the same parameter setting as model 2, was re-trained in the pooled dataset. The final model estimates as well as the risk score per unit change are listed in Table 3.

Table 3

Final model for the selected predictors of survival in the pooled cohorts (n=442)

VariablesCox regression model
Prognostic score
Hazard ratio (95% CI)pBaseline valuePrognostic score per unit changeMedian prognostic score
Multiple tumors1.65 (1.24, 2.22)0.0006No103.8
Lymph node involvement1.82 (1.38, 2.39)<0.0001No104.6
Distant metastasis2.00 (1.37, 2.93)<0.0001No141.7
ALB, per 10 g/L0.77 (0.55, 1.07)0.12445546.4
LDH, per 100 U/mL1.25 (1.10, 1.42)0.0005047.6
Iron, per 10 μmol/L0.80 (0.64, 1.01)0.060445412
FIB, per U1.05 (1.00, 1.11)0.067000.82.8
CA125, per 100 U/mL1.04 (1.00, 1.08)0.069200.80.16
CA19-9, per 100 U/mL1.09 (1.05, 1.13)<0.000101.40.56

The nomogram visualizing the above using the final model is presented in Figure 1. The predicted 12-, 24- and 36-month survival rates could be easily read from the graph. The calibration plot for the probability of survival at 12 months after diagnosis showed a good agreement between the predicted OS and the actual OS (Fig. 2).

Risk stratification of the ICC patients by the new model

The patients in the pooled cohorts were divided into three risk groups according to the predicted prognostic score cutoff by 31 and 47. The OS curve separated noticeably between the three risk groups (Fig. 3). The median OS for the low, mid and high-risk groups were 35.6 months (95% CI: 24.7-not estimable), 12.1 months (95% CI: 9.8–14.6) and 6.2 months (95% CI: 5.3–8.2).

Kaplan-Meier curves demonstrating the differences in OS among low, mid and high-risk patients stratified by the prognostic score for ICC.
Fig. 3  Kaplan-Meier curves demonstrating the differences in OS among low, mid and high-risk patients stratified by the prognostic score for ICC.

From Table 4, we can see the distribution of treatment methods across different risk stratifications. Patients with lower risk had been treated with surgery more frequently than patients with higher risk, which is consistent with the real situation that reflects the eligibility of surgery resection as a major prognostic factor for survival. Therefore, the new nomogram was able to stratify patients and give guidance for determining clinical strategy.

Table 4

Distribution of risk stratification across different treatment modality

Treatment modalityLow risk, n (%)Mid risk, n (%)High risk, n (%)
Palliative care1 (1.0)7 (6.8)24 (22.4)
Surgery45 (43.7)20 (19.4)15 (14.0)
Surgery + Regional therapy36 (35.0)33 (32.0)12 (11.2)
Regional therapy15 (14.6)38 (36.9)48 (44.9)
Systemic therapy6 (5.9)5 (4.9)8 (6.5)

Discussion

ICC, a major subclass of cholangiocarcinoma and the second most common primary hepatic malignancy, is a major public health problem, with significant morbidity and mortality. The OS of ICC patients is poor, as the majority of patients present at an advanced stage that prohibits curative surgical resection.21 Several prognostication systems have been developed to predict the OS of ICC patients after hepatic resection, but few were constructed based on a population including the patients who were not appropriate candidates.7–17

Nomograms are statistical tools that account for the ability of multiple variables to influence the probability of long-term survival for a particular patient. To our knowledge, a nomogram model to predict the OS of ICC patients in various stages from the time of the diagnosis which has utility to guide clinical decisions has not been reported. We have developed and validated a new nomogram in a large cohort of ICC patients with a fairy long follow-up that more accurately predicts the survival probability at the time of diagnosis.

Our model consists of nine variables based on the tumor features and clinical biomarkers. By integrating these variables into nomogram scores, we are able to stratify patients into different groups based on the prognosis and OS (Fig. 3). Several commonly used clinical blood parameters were independently associated with survival in our ICC patients, such as ALB, FIB, carbohydrate antigen 125 (CA125), LDH and iron. ALB has been used frequently as a prognostic indicator of ICC, for its relation with liver reserve function and nutritional status.22,23 FIB was also reported as a prognostic factor of CCA.24 Elevated FIB indicates a hypercoagulable state which may be related to tissue hypoxia caused by a growing tumor and procoagulant factors. Previous studies have shown the utility of both CA125 and LDH as prognostic biomarkers for various cancers, such as hepatocellular carcinoma,25–28 lung, ovarian, and cervical cancer.19–21 CA125 and LDH-A expression in ICC tissue samples are considered evidence-proven prognostic factors of poor survival,29,30 and we found a significant relationship between the serum levels of CA125 per 100 U/m: (hazard ratio=1.04, 95% CI: 1.00–1.08) and LDH per 100U/L (hazard ratio=1.25, 95% CI: 1.10–1.42) on the OS of patients with ICC. The level of serum iron was also introduced into the nomogram to jointly predict the prognosis of ICC, which is a novel biomarker, although iron homeostasis has been reported disordered in cholangiocarcinoma. 31,32

The strengths of our study are the well-characterized cohorts of patients with pathology-proven ICC, the follow-up data on the survival outcome, sample size, and the broader applicability. However, we acknowledge several limitations. Our nomogram was derived from a homogenous cohort who were seen at a single institution. Despite the fact that our model was internally validated, future studies are needed to determine the prognostic accuracy in another group of patients. In view of the distinguishment of etiology among different countries, the nomogram tends to be suitable for the ICC population mainly caused by HBV infection and admitted in the hospital for pathological analyses and further treatment. Additionally, almost half of the patients had liver cirrhosis. Although the degree of cirrhosis has weight in clinical decision-making, statistical analysis shows that, compared with factors associated with the tumor, cirrhosis has no significant effect on the prognosis. It can be seen from the model for end-stage liver disease score or Child-Pugh score in Table 1 that most of the patients involved in this study had good liver function reserve, and this selective bias may be caused by the diagnosis of ICC relying on pathological evidence so that patients with end-stage cirrhosis could not be involved due to the lack of pathology. Thus, the prognostic model will be updated with the improvement of diagnosis and treatment methods.

In conclusion, we proposed a new prognostic model suitable for the comprehensive ICC population at all stages, which can stratify the risk of patients before treatment to help doctors and patients make reasonable clinical decisions or seek new therapies. Further external studies are needed to validate the prognostic ability of this new model.

Supporting information

Supplementary Fig. 1

Clinical features of ICC patients caused by HBV infection.

The descriptive criteria refer to the European Association for the Study of the Liver 2017 Clinical Practice Guidelines on the management of HBV infection. NAs, nucleoside/nucleotide analogues.

(TIF)

Supplementary Fig. 2

Receiver operating characteristic curves and AUC predicting survival at different time points after diagnosis for Model 1 (S1A), Model 2 (S1B), Model 3 (S1C), and Model 4 (S1D).

(TIF)

Supplementary Table 1

Type and proportion of regional therapy.

(DOCX)

Supplementary Table 2

Candidate models and prediction performance in the validation cohort.

(DOCX)

Abbreviations

AJCC: 

American Joint Committee on Cancer

ALB: 

albumin

AUC: 

area under the curve

CA125: 

carbohydrate antigen 125

CA19-9: 

carbohydrate antigen 19-9

CEA: 

carcinoembryonic antigen

CI: 

confidence interval

FDR: 

false discovery rate

FIB: 

fibrinogen

HBc: 

anti-hepatitis B core protein

HBeAg: 

hepatitis B e antigen

HBsAg: 

hepatitis B surface antigen

HBV: 

hepatitis B virus

ICC: 

intrahepatic cholangiocarcinoma

OS: 

overall survival

Declarations

Data sharing statement

The data are available upon reasonable request.

Funding

This work was supported by the Capital’s Funds for Health Improvement and Research (No. Z181100001718 075) and Medical Big Data and AI R&D Project of General Hospital (2019MBD-025).

Conflict of interest

The authors have no conflict of interests related to this publication.

Authors’ contributions

Study concept and design (NZ, YYL), drafting of the manuscript (NZ, SNZ, DWJ), critical revision (LRR, SL), analysis and interpretation of data (SNZ, DWJ, XX), recruitment and follow-up (SSL, XXL, LXY), and collection of the data (SNZ, SSL).

References

  1. Buettner S, van Vugt JL, IJzermans JN, Groot Koerkamp B. Intrahepatic cholangiocarcinoma: current perspectives. Onco Targets Ther 2017;10:1131-1142 View Article
  2. Florio AA, Ferlay J, Znaor A, Ruggieri D, Alvarez CS, Laversanne M, et al. Global trends in intrahepatic and extrahepatic cholangiocarcinoma incidence from 1993 to 2012. Cancer 2020;126(11):2666-2678 View Article
  3. Tsilimigras DI, Sahara K, Wu L, Moris D, Bagante F, Guglielmi A, et al. Very early recurrence after liver resection for intrahepatic cholangiocarcinoma: Considering alternative treatment approaches. JAMA Surg 2020;155(9):823-831 View Article
  4. Banales JM, Cardinale V, Carpino G, Marzioni M, Andersen JB, Invernizzi P, et al. Expert consensus document: Cholangiocarcinoma: current knowledge and future perspectives consensus statement from the European Network for the Study of Cholangiocarcinoma (ENS-CCA). Nat Rev Gastroenterol Hepatol 2016;13(5):261-280 View Article
  5. Khan SA, Davidson BR, Goldin RD, Heaton N, Karani J, Pereira SP, et al. Guidelines for the diagnosis and treatment of cholangiocarcinoma: an update. Gut 2012;61(12):1657-1669 View Article
  6. Doussot A, Groot-Koerkamp B, Wiggers JK, Chou J, Gonen M, DeMatteo RP, et al. Outcomes after resection of intrahepatic cholangiocarcinoma: External validation and comparison of prognostic models. J Am Coll Surg 2015;221(2):452-461 View Article
  7. Wang Y, Li J, Xia Y, Gong R, Wang K, Yan Z, et al. Prognostic nomogram for intrahepatic cholangiocarcinoma after partial hepatectomy. J Clin Oncol 2013;31(9):1188-1195 View Article
  8. Okabayashi T, Yamamoto J, Kosuge T, Shimada K, Yamasaki S, Takayama T, et al. A new staging system for mass-forming intrahepatic cholangiocarcinoma: analysis of preoperative and postoperative variables. Cancer 2001;92(9):2374-2383
  9. Nathan H, Aloia TA, Vauthey JN, Abdalla EK, Zhu AX, Schulick RD, et al. A proposed staging system for intrahepatic cholangiocarcinoma. Ann Surg Oncol 2009;16(1):14-22 View Article
  10. Jiang W, Zeng ZC, Tang ZY, Fan J, Sun HC, Zhou J, et al. A prognostic scoring system based on clinical features of intrahepatic cholangiocarcinoma: the Fudan score. Ann Oncol 2011;22(7):1644-1652 View Article
  11. Hyder O, Marques H, Pulitano C, Marsh JW, Alexandrescu S, Bauer TW, et al. A nomogram to predict long-term survival after resection for intrahepatic cholangiocarcinoma: an Eastern and Western experience. JAMA Surg 2014;149(5):432-438 View Article
  12. Uenishi T, Ariizumi S, Aoki T, Ebata T, Ohtsuka M, Tanaka E, et al. Proposal of a new staging system for mass-forming intrahepatic cholangiocarcinoma: a multicenter analysis by the Study Group for Hepatic Surgery of the Japanese Society of Hepato-Biliary-Pancreatic Surgery. J Hepatobiliary Pancreat Sci 2014;21(7):499-508 View Article
  13. Lin ZY, Liang ZX, Zhuang PL, Chen JW, Cao Y, Yan LX, et al. Intrahepatic cholangiocarcinoma prognostic determination using pre-operative serum C-reactive protein levels. BMC Cancer 2016;16(1):792 View Article
  14. Sakamoto Y, Kokudo N, Matsuyama Y, Sakamoto M, Izumi N, Kadoya M, et al. Proposal of a new staging system for intrahepatic cholangiocarcinoma: Analysis of surgical patients from a nationwide survey of the Liver Cancer Study Group of Japan. Cancer 2016;122(1):61-70 View Article
  15. Yeh CN, Wang SY, Chen YY, Chen MH, Chiang KC, Cheng CT, et al. A prognostic nomogram for overall survival of patients after hepatectomy for intrahepatic cholangiocarcinoma. Anticancer Res 2016;36(8):4249-4258
  16. Zheng BH, Yang LX, Sun QM, Fan HK, Duan M, Shi JY, et al. A new preoperative prognostic system combining CRP and CA199 for patients with intrahepatic cholangiocarcinoma. Clin Transl Gastroenterol 2017;8(10):e118 View Article
  17. Ishimoto U, Kondo S, Ohba A, Sasaki M, Sakamoto Y, Morizane C, et al. Prognostic factors for survival in patients with advanced intrahepatic cholangiocarcinoma treated with gemcitabine plus cisplatin as first-line treatment. Oncology 2018;94(2):72-78 View Article
  18. Guo S, Liu HD, Liu YF, Liu L, Sun Q, Cui XJ. Hepatoma-derived growth factor: a novel prognostic biomarker in intrahepatic cholangiocarcinoma. Tumour Biol 2015;36(1):353-364 View Article
  19. Kamphues C, Al-Abadi N, Dürr A, Bova R, Klauschen F, Stenzinger A, et al. DNA index is a strong predictive marker in intrahepatic cholangiocarcinoma: the results of a five-year prospective study. Surg Today 2014;44(7):1336-1342 View Article
  20. Wu SY, Yu MX, Li XG, Xu SF, Shen J, Sun Z, et al. Identification of Homer1 as a potential prognostic marker for intrahepatic cholangiocarcinoma. Asian Pac J Cancer Prev 2014;15(7):3299-3304 View Article
  21. Olaizola P, Perugorria MJ, Banales JM. Toward personalized medicine for intrahepatic cholangiocarcinoma: Pharmacogenomic stratification of patients. Hepatology 2018;68(3):811-814 View Article
  22. Matsumoto T, Itoh S, Yoshizumi T, Kurihara T, Yoshiya S, Mano Y, et al. C-reactive protein : albumin ratio in patients with resectable intrahepatic cholangiocarcinoma. BJS Open 2020;4(6):1146-1152 View Article
  23. Ni JY, An C, Zhang TQ, Huang ZM, Jiang XY, Huang JH. Predictive value of the albumin-bilirubin grade on long-term outcomes of CT-guided percutaneous microwave ablation in intrahepatic cholangiocarcinoma. Int J Hyperthermia 2019;36(1):328-336 View Article
  24. Chen Q, Li J, Jin B, Wu X, Shi Y, Xu H, et al. Prognostic nomogram that predicts overall survival of patients with distal cholangiocarcinoma after pancreatoduodenectomy. Cancer Manag Res 2020;12:10303-10310 View Article
  25. Faloppi L, Bianconi M, Memeo R, Casadei Gardini A, Giampieri R, Bittoni A, et al. Lactate dehydrogenase in hepatocellular carcinoma: Something old, something new. Biomed Res Int 2016;2016:7196280 View Article
  26. Yada M, Miyazaki M, Motomura K, Masumoto A, Nakamuta M, Kohjima M, et al. The prognostic role of lactate dehydrogenase serum levels in patients with hepatocellular carcinoma who are treated with sorafenib: the influence of liver fibrosis. J Gastrointest Oncol 2016;7(4):615-623 View Article
  27. Faloppi L, Scartozzi M, Bianconi M, Svegliati Baroni G, Toniutto P, Giampieri R, et al. The role of LDH serum levels in predicting global outcome in HCC patients treated with sorafenib: implications for clinical management. BMC Cancer 2014;14:110 View Article
  28. Yang Z, Ye P, Xu Q, Lu Y, Tang B, Wang Q, et al. Elevation of serum GGT and LDH levels, together with higher BCLC staging are associated with poor overall survival from hepatocellular carcinoma: a retrospective analysis. Discov Med 2015;19(107):409-418
  29. Higashi M, Yamada N, Yokoyama S, Kitamoto S, Tabata K, Koriyama C, et al. Pathobiological implications of MUC16/CA125 expression in intrahepatic cholangiocarcinoma-mass forming type. Pathobiology 2012;79(2):101-106 View Article
  30. Yu Y, Liao M, Liu R, Chen J, Feng H, Fu Z. Overexpression of lactate dehydrogenase-A in human intrahepatic cholangiocarcinoma: its implication for treatment. World J Surg Oncol 2014;12:78 View Article
  31. Jamnongkan W, Thanan R, Techasen A, Namwat N, Loilome W, Intarawichian P, et al. Upregulation of transferrin receptor-1 induces cholangiocarcinoma progression via induction of labile iron pool. Tumour Biol 2017;39(7):1010428317717655 View Article
  32. Mancinelli R, Cutone A, Rosa L, Lepanto MS, Onori P, Pannarale L, et al. Different iron-handling in inflamed small and large cholangiocytes and in small and large-duct type intrahepatic cholangiocarcinoma. Eur J Histochem 2020;64(4):3156 View Article
  • Journal of Clinical and Translational Hepatology
  • pISSN 2225-0719
  • eISSN 2310-8819
Back to Top

A New Prognostic Model Covering All Stages of Intrahepatic Cholangiocarcinoma

Shuang-Nan Zhou, Shan-Shan Lu, Da-Wei Ju, Ling-Xiang Yu, Xiao-Xiao Liang, Xiao Xiang, Suthat Liangpunsakul, Lewis R. Roberts, Yin-Ying Lu, Ning Zhang
  • Reset Zoom
  • Download TIFF