Age, period, and cohort effects on gastric cancer mortality

The incidence of gastric cancer varies both within and among populations and can change rapidly without a corresponding decline in Helicobacter pylori prevalence. Here, we describe the changes in gastric cancer mortality in Japan in relation to age-period-cohort effects as well as the decline in H. pylori prevalence.

We used data from Japanese for men and women aged 30-94 for birth years 1875-1970 (calendar time 1950-2000) to observe the age, period and cohort effects on gastric cancer mortality rates. Additionally, we used Poisson regression to simultaneously adjust for concurrent age, period and cohort effects as well as for declining H. pylori prevalence in the Japanese population.

There was an approximate 60 % decline in gastric cancer mortality between 1965 and 1995. Detailed age, period and cohort analyses and Poisson regression analysis showed these factors interact in complex ways, analyses focused on one or two of these effects, such as birth cohort without considering concurrent age and period would obscure important interactions that affected different age groups at different times to produce this composite effect.

The underlying complexity in population-disease dynamics requires population-specific descriptions of trends using multiple methods to provide an in-depth analysis while simultaneously allowing for necessary statistical adjustments as well as identification of interactions. More thorough descriptions of the population-specific general trends in relation to changes in the population structure (age-period-cohort) enable better prevention and health care policy planning, and further, the descriptions enable hypothesis generation regarding causes of population-specific disease patterns.