Nomograms predict survival outcome of Klatskin tumors patients

Patient data

Stat version 8.2.1 of SEER was used to download data from all patients with a diagnosis of liver cancer and cholangiocarcinoma from 2005 to 2015, according to the ICD-0-3/WHO 2008 guidelines. The main inclusion criteria were as follows: age 0 to 104 years, liver cancer with cholangiocarcinoma as the main malignant cancer diagnosis, pathological types of Klatskin tumors (based on ICD-O-3 8162/3), ICCA (based on ICD-O-3 8160/3), and HCC (based on ICD-O-3 8170/3). We also examined the gender, ethnicity, lymph node metastases, histological grades I to IV, definite AJCC TNM stages, and type of therapeutic strategy. We excluded patients with an unclear T stage record, and unknown survival time, diagnostic confirmation, or surgery strategy. Patients were included based on a diagnosis of liver cancer and cholangiocarcinoma before 2014 to ensure sufficient time for follow-up (Fig. 1). The study was in compliance with the ethics statement of Zhongshan Hospital, Fudan University.

Statistical analysis

Normalized continuous variables with a homogeneity of variance were compared using the t test; the Mann–Whitney U test was performed for all other variable calculations. Multiple groups were compared using ANOVA and Tukey’s post hoc test. The chi-square test was applied to the categorical data. Patients were divided into three groups according to pathology: Klatskin tumor group, the ICCA group, and the HCC group. The demographic and clinical characteristics of the three groups were compared using the chi-square test. Continuous variables were presented as mean ± SE or median (minimum, maximum), and the categorical variables were calculated as a percentage. PSM analyses were performed based upon age, race, gender, grade, historic stage, AJCC 6th stage, tumor size, lymph node status, surgical method, alpha-fetoprotein (AFP) level, and fibrosis score at a 1:1 ratio to adjust for the differences among the Klatskin tumor, ICCA, and HCC groups. The Kaplan–Meier method was used to plot the survival curve. The unadjusted OS rate of the different histological subtypes was adjusted using the log-rank test. OS was defined from the date of diagnosis to the date of death or final follow-up. Cox proportional hazard regression models were used to evaluate prognostic factors and HR was used as the 95% confidence interval. Diagnostic age, summary stage, lymph node metastases, and the treatment of the primary tumor were included in the survival analysis. The above statistical data were analyzed using the SPSS version 22.0 software package (IBM SPSS Statistics, Chicago, IL, US) and p < 0.05 was considered statistically significant.

Demographic, tumor, and treatment characteristics of patients with Klatskin tumors, ICCA, or HCC

A total of 65,450 patients met the inclusion criteria for our study, including 317 Klatskin tumor patients, 7,316 ICCA patients, and 57,817 HCC patients. The demographic and clinical characteristics of all patients are summarized in Table 1. The composition of the three groups differed significantly by age, race, gender, degree of differentiation, AJCC stage, surgical method, alpha fetoprotein (AFP), and fibrosis score. Klatskin tumors occurred more frequently in older patients (72.78 ± 13.29, 66.76 ± 12.99 vs 63.62 ± 11.48, p < 0.001) and among white men, which is similar to the demographics of the ICCA and HCC groups (79.5%, 78.3% vs 68.3%, p < 0.001). The localized and regional historic stages for patients with Klatskin tumors was 31.2% and 31.9%, 26.8% and 30.2% for the ICCA group, and 49.7% and 27.9% for the HCC group, respectively (p < 0.001). We lacked sufficient data to generate statistically significant results when running comparison analyses of the treatment method. The overall incidence rate of Klatskin tumors decreased between 2004 and 2014 (r =  − 0.94, p < 0.001) while the rates of ICCA (r = 0.89, p < 0.001) and HCC (r = 0.73, p = 0.011) increased (Fig. 2).

Survival and prognostic factors for Klatskin tumor, ICCA, and HCC patients

Kaplan–Meier analysis was used to evaluate OS and cancer-specific survival (Fig. 2). Klatskin tumor patients had worse survival rates than ICCA and HCC patients (median OS: 5 months vs 9 months and 14 months, p < 0.001; median cancer-specific survival: 10 months vs 13 months and 23 months, p < 0.001). The 1-, 2- and 3-year OS rates of Klatskin tumor, ICCA, and HCC patients were 26.2% versus 62.2% and 72.4%, 10.7% versus 36.4% and 48.5%, and 3.4% versus 19.1% and 36.2%, respectively. The X-tile program was used to divide all patients into two groups according to age in order to more accurately determine the prognosis. Further stratification studies showed that age at diagnosis, race, gender, tumor differentiation, SEER historic stage, tumor size, lymph node status, surgical intervention (including local tumor destruction and surgery), AFP level, and fibrosis score had a significant impact on the OS of Klatskin tumor, ICCA, and HCC patients (p < 0.05) (Fig. 3).

The results of the Kaplan–Meier analysis of OS were further analyzed using univariate and multivariate Cox regression models. According to the univariate factor analysis, older age, worse pathological grade, larger tumor size, lymph node metastases, lower AFP level, and late TNM and SEER stages were significantly related to a worse prognosis (p < 0.05) (Table 2). In contrast, surgical intervention and histologic type were low risk factors (p < 0.001). Adjusting the variables did not impact the results of the multivariate analysis, except in regard to lymph node metastases (p = 0.089).

PSM analysis was performed to adjust for the unmatching cohort and a total of 317 Klatskin tumor patients were matched with 317 ICCA and 317 HCC patients (1:1:1) (Table 3). In the matching cohorts, the Klatskin tumor group had worse 1-, 2-, 3-year OS rates than the ICCA and HCC groups (1-year OS: 28.4% versus 71.9% and 85.6%; 2-year OS: 13.3% versus 52.5% and 73.0%; 3-year OS: 5.2% versus 38.1% and 66.1%). The median OS in the Klatskin tumor group was 5 months compared with 26 and 68 months in the ICCA and HCC groups, respectively. The median cancer-specific survival in the Klatskin tumor group was 10 months compared with 34 months in the ICCA group (p < 0.001) (Fig. 4). Kaplan–Meier analysis and univariate Cox regression was used for further stratification studies and the results were similar to those of the un-matched cohorts (Fig. 5). Based on multivariable analysis, the prognostic factors for OS were older age (HR 1.736, 95% CI [1.396–2.157], p < 0.001), lymph node metastasis (HR 1.406, 95% CI [1.093–1.808], p = 0.008), surgical intervention, lower AFP level (HR 0.757, 95% CI [0.587–0.976], p = 0.032) and histologic type (Table 4).

Klatskin tumor patient survival and prognostic factors

317 patients were identified with Klatskin tumors from 2004 to 2014 and a Kaplan–Meier analysis was performed based on patient characteristics (Table 1). Kaplan–Meier survival curves and log-rank analysis showed that a poorer prognosis was associated with older age (median OS: 3 months versus 7 months, p < 0.001), distant stage (median OS: localized 7 months, regional 6 months versus distant 2 months, p < 0.001), no surgery (median OS: 5 months vs 17 months, p < 0.001), and lymph node metastases (median OS: 3 months vs 7 months, p = 0.016) (Fig. 6).

Multivariate Cox analysis of the 317 Klatskin tumor patients showed that older age (HR 1.725, 95% CI [1.324–2.249], p < 0.001), late SEER historic stage (HR 3.594, 95% CI [1.251–10.326], p = 0.018), and lymph node metastases (HR=1.468, 95% CI [1.008–2.139], p = 0.046) were independent prognostic factors for worse survival. Conversely, cancer-directed surgery was an independent protective factor that decreased the risk of death by 44.5% for Klatskin tumor patients (HR=0.555, 95% CI [0.316–0.977], p = 0.041) (Table 5).

Prognostic nomogram for Klatskin tumor

The prognostic nomogram was used to integrate all of the significant independent factors for OS in the Klatskin tumor group (Fig. 7). The patients were divided into three groups according to age using the X-tile program. The optimal cut-off points were at 71 and 82 years of age. A prognostic nomogram was established to more accurately determine the survival of Klatskin tumor patients using all of the significant independent variables based on multivariate Cox analysis (Fig. 7A). The C-index for OS prediction was 0.651 (95% CI [0.607–0.695]). The calibration plot for the probability of survival at 1, 2, and 3 years showed an optimal agreement between the prediction by nomogram and actual observation (Figs. 7B-D). The ROC curve was used to predict the accuracy of the model and the analysis was in agreement with the nomogram with a value of 0.884 in the area under the curve (AUC) (p < 0.001, Fig. 7E).