Cancer mortality in Europe, 1970-2009: an age, period, and cohort analysis.

The aim of this study was to interpret long-term trends in cancer mortality. We analyzed age, period of death, and cohort of birth effects for 18 major cancer sites and all neoplasms combined in the European Union and 16 European countries over the period 1970-2009 using data from the WHO mortality database. We used a log-linear Poisson model with a likelihood penalizing function to solve the identifiability problem. The present comprehensive analysis confirms the appreciable declines of cancer mortality since the late 1980s in most European countries. For several major sites, the declines were appreciably greater for cohort effects than for period ones. This reflects - besides the inherent characteristics of the model used - a major impact of tobacco smoking on total cancer mortality, particularly in men, as the role of tobacco in mortality of subsequent generations is largely - although not totally - a cohort effect. For neoplasms largely affected by improvements in management and therapy (e.g. testis, Hodgkin lymphoma, leukemia, but also breast and colorectal cancer), a favorable period effect is evident. However, there is also a cohort effect as the advancements have been generally greater in younger generations. For several cancer sites, there were considerable differences in cohort and period effects across Europe, and particularly high age-specific estimates were observed in eastern countries.