Validation and limitation of age–period–cohort model in simulating mortality due to hepatocellular carcinoma from 1940 to 2010 in Japan

Abstract

We aimed to simulate the mortality due to hepatocellular carcinoma (HCC) by the age-period-cohort (APC) model with use of sex- and age-specific mortality data, for the purpose of validating the utility and assessing the limitation of this model. Age-specific mortality due to HCC was gleaned from people aged 20-84 years during 1940 through 2010 in Japan. The APC model had a high performance in reproducing HCC mortality (modified determination coefficient R(2) COR ≥ 0.99). Risk of HCC increased with age in both sexes, while risk of period barely changed in both sexes. The birth cohort factor in the APC model in males highlighted the maximum point within birth years 1931-1935. The observed HCC mortality in 2010 in males (19 444) was lower than the predicted, and corresponded to 72.3% of the predicted 26 883.4, and in all age groups by 5-year increments (55.6-90.9%). In females, the observed mortality was lower than that predicted in those aged 64 years or less, but not in those aged 65 years or more. We applied the APC model to predict HCC mortality rate, and it reproduced the observed mortality rate faithfully. However, in the recent past, the observed morality rate in males was only 72.3% that of the predicted. Such differences would be attributed to combined effects of medical interventions, such as antiviral treatments and screening for hepatitis viruses implemented in the early 1990s in Japan.