Intrahepatic Cholangiocarcinoma Tumor Burden: Classification and Regression Tree Model to Define Prognostic Groups after Resection

Surgery
28 Nov, 2019 ,

The impact of intrahepatic cholangiocarcinoma tumor burden on overall survival was examined using a machine-learning tool. Intrahepatic cholangiocarcinoma tumor size and number demonstrated a strong nonlinear association with survival after resection of intrahepatic cholangiocarcinoma.

Source
Full content

Background

Tumor burden is an important factor in defining prognosis among patients with primary and secondary liver cancers. Although the eighth edition of the American Joint Committee on Cancer staging system has changed the criteria for staging patients with intrahepatic cholangiocarcinoma to better define the effect of tumor burden on prognosis, the impact of intrahepatic cholangiocarcinoma tumor burden on overall survival has not been examined using a machine-learning tool.

Methods

Patients who underwent resection of intrahepatic cholangiocarcinoma at 1 of 14 participating international hospitals between 1990 and 2015 were identified. Classical survival models and the Classification and Regression Tree model were used to identify groups of patients with a homogeneous risk of death and investigate the hierarchical association between variables and overall survival.

Results

Among 1,116 patients included in the analysis, tumor size was ≤5 cm in 447 (40.1%) patients and >5 cm in 669 (59.9%) patients. Although 82.9% (n = 926) of patients had a single intrahepatic cholangiocarcinoma, 9.9% (n = 110) and 7.2% (n = 80) of patients had 2 and ≥3 tumors, respectively. Patients with intrahepatic cholangiocarcinoma tumors ≤5 cm and >5 cm had a 5-year overall survival of 51.7% and 32.6%, respectively (P < 0.001). Five-year overall survival decreased from 44.6% among patients with a single intrahepatic cholangiocarcinoma to 28.1% and 14.2% among patients with 2 and ≥3 intrahepatic cholangiocarcinomas, respectively (P < 0.001). Among the combinations of tumor size and intrahepatic cholangiocarcinoma tumor number used to estimate tumor burden, logarithmic transformation of tumor size (log tumor size) and intrahepatic cholangiocarcinoma tumor number had the highest concordance index. The Classification and Regression Tree model identified 8 classes of patients with a homogeneous risk of death, illustrating the hierarchical relationship between tumor burden (log tumor size and number of intrahepatic cholangiocarcinomas) and other factors associated with prognosis.

Conclusion

Intrahepatic cholangiocarcinoma tumor size and number demonstrated a strong nonlinear association with survival after resection of intrahepatic cholangiocarcinoma. A log-model Classification and Regression Tree–derived tumor burden score may be a better tool to estimate prognosis of patients undergoing curative-intent resection of intrahepatic cholangiocarcinoma.