Prognostic model for hepatocellular carcinoma with time-dependent factors

• Published in: Acta medica Okayama Vol. 65; no. 1; p. 11
• Main Authors: Kuwaki, Kenji, Nouso, Kazuhiro, Kobayashi, Yoshiyuki, Nakamura, Shinichiro, Ito, Yoichi M, Iwadou, Shouta, Hagihara, Hiroaki, Yasunaka, Tetsuya, Toshimori, Junichi, Miyatake, Hirokazu, Miyoshi, Kenji, Onishi, Hideki, Miyake, Yasuhiro, Shoji, Bon, Takaki, Akinobu, Shiraha, Hidenori, Iwasaki, Yoshiaki, Kobashi, Haruhiko, Yamamoto, Kazuhide
• Format: Journal Article
• Language: English
• Published: Japan 01.02.2011
• Subjects: Adult; Aged; Aged, 80 and over; Carcinoma, Hepatocellular - mortality; Carcinoma, Hepatocellular - therapy; Female; Follow-Up Studies; Humans; Liver Neoplasms - mortality; Liver Neoplasms - therapy; Male; Middle Aged; Prognosis; Proportional Hazards Models; Survival Analysis; Time Factors; Young Adult
• ISSN: 0386-300X
• DOI: 10.18926/amo/43825

The purpose of this study was to build a prognostic model of hepatocellular carcinoma (HCC) using time-dependent covariates to re-evaluate the prognosis at any stage of the disease. The subjects were consecutive HCC patients who were treated at our institute between 1995 and 2007. We constructed time-fixed and time-dependent prognostic models with a training group (n=336) and compared the prognostic abilities between conventional Cancer of the Liver Italian Program (CLIP) scores, Japan Integrated Staging (JIS) scores, an Okuda classification, and our prognostic models in the testing group (n=227) with the c-index. The time-dependent prognostic model consisted of main tumor size, tumor number, portal vein invasion, distant metastasis, alpha-fetoprotein, des-gamma-carboxy prothrombin (DCP), bilirubin, and albumin and the weighted scores were set for each factor depending on the hazard ratio for the prognosis. The prognostic index was determined by summing the scores. The c-index values for the CLIP scores, JIS scores, Okuda classification, and our time-dependent model were 0.741, 0.727, 0.609, and 0.870, respectively. These results indicate that our time-dependent model can estimate the prognosis of HCC more precisely than traditional time-fixed models and can be used to re-predict the prognosis of HCC.