Survival model methods for analyses of cancer incidence trends in young adults.

Recent studies have reported increases in cancer incidence in adults under 50 years. However, there remains uncertainty about whether these are true increases or a result of incidental findings from increased medical imaging. To evaluate these trends, we propose an alternative method to age-period-cohort analyses based on survival modeling. Simulations show that our method is capable of quantifying cohort effects within various backgrounds including increasing medical imaging. We applied the method to analyze the changes in cancer incidence rates for 44 anatomic sites, stratified by sex, by birth cohort for individuals born from 1945 to 1969 in the US based on incidence data from the Surveillance, Epidemiology, and End Results (SEER) program, and tested the validity of our models using later birth cohorts (1970-1974 and 1975-1979). We found that cancer risks have increased significantly in 15 sites (9 in men and 11 in women) for 25-49 year-olds. These results were consistent with previous findings from age-period-cohort analyses. Furthermore, based on our simulations, these increases were independent of increased medical imaging and support substantial, increased extrinsic risks in the identified cancers. Although our approach has several limitations including the restriction to the younger age range and requirement of complete data for all ages of interest, we demonstrate many advantages of our approach including the ease in implementation and interpretation of cohort effects, robustness to various period backgrounds, and ability to make predictions. Our approach should help epidemiologists evaluate cohort effects using incidence data for cancer or other diseases.