Detecting a cohort effect for cancer mortality data using varying coefficient model

INTRODUCTION:  Cancer mortality is increasing with the aging of the population in Japan. Cancer information obtained by a suitable method becomes a basis for planning an effective cancer control program. Our purpose in this study is to visualize cancer mortality risk as the curved surface on the age-period plane and find a cohort-effect. 

METHODS:  To visualize cancer mortality risk as the curved surface on the age-period plane, we introduce a varying coefficient, which is estimated by model non-parametric and parametric. The non-parametric model is based on a geographically weighted regression by regarding age and period as a geographical coordinate. The parametric model is based on an interaction model which can detect a cohort effect and assess its significance. The proposed method is applied to data of liver cancer mortality in Vital Statistics Japan. Liver cancer is said to have a cohort effect such that the birth cohort around 1935 suffers a high risk. We tried to detect such cohort effect.

RESULTS:  We visualized the age-period trend for data of liver cancer mortality in Japan by non-parametric and parametric model. The estimated trend of parametric model was similar to that of non-parametric model. The contour map of risk suggested a cohort effect such that the birth cohort around 1935 suffers a high risk. This cohort effect was highly significant.

CONCLUSIONS:  Non-parametric model is useful to grasp the age-period trend of mortality, but no statistical assessment on cohort effect. On the other hand, parametric model is useful for both visualization and statistical test. For data of liver cancer mortality in Japan, the parametric model was well fitted and detected the significant cohort effect.