International Journal of Molecular Sciences (Mar 2023)

Prediction Model with <i>HLA-A*33:03</i> Reveals Number of Days to Develop Liver Cancer from Blood Test

  • Nao Nishida,
  • Jun Ohashi,
  • Goki Suda,
  • Takehiro Chiyoda,
  • Nobuharu Tamaki,
  • Takahiro Tomiyama,
  • Sachiko Ogasawara,
  • Masaya Sugiyama,
  • Yosuke Kawai,
  • Seik-Soon Khor,
  • Masao Nagasaki,
  • Akihiro Fujimoto,
  • Takayo Tsuchiura,
  • Miyuki Ishikawa,
  • Koichi Matsuda,
  • Hirohisa Yano,
  • Tomoharu Yoshizumi,
  • Namiki Izumi,
  • Kiyoshi Hasegawa,
  • Naoya Sakamoto,
  • Masashi Mizokami,
  • Katsushi Tokunaga

DOI
https://doi.org/10.3390/ijms24054761
Journal volume & issue
Vol. 24, no. 5
p. 4761

Abstract

Read online

The development of liver cancer in patients with hepatitis B is a major problem, and several models have been reported to predict the development of liver cancer. However, no predictive model involving human genetic factors has been reported to date. For the items incorporated in the prediction model reported so far, we selected items that were significant in predicting liver carcinogenesis in Japanese patients with hepatitis B and constructed a prediction model of liver carcinogenesis by the Cox proportional hazard model with the addition of Human Leukocyte Antigen (HLA) genotypes. The model, which included four items—sex, age at the time of examination, alpha-fetoprotein level (log10AFP) and presence or absence of HLA-A*33:03—revealed an area under the receiver operating characteristic curve (AUROC) of 0.862 for HCC prediction within 1 year and an AUROC of 0.863 within 3 years. A 1000 repeated validation test resulted in a C-index of 0.75 or higher, or sensitivity of 0.70 or higher, indicating that this predictive model can distinguish those at high risk of developing liver cancer within a few years with high accuracy. The prediction model constructed in this study, which can distinguish between chronic hepatitis B patients who develop hepatocellular carcinoma (HCC) early and those who develop HCC late or not, is clinically meaningful.

Keywords