Peak and decline in cancer incidence, mortality, and prevalence at old ages.

BACKGROUND: Cancer incidence and mortality increase with age through much of adulthood, but earlier work has found that these rates decline among the very elderly. To compare incidence and mortality at the oldest ages, the authors investigated both in the same large population, which comprised 9.5% of the United States in 2000. The authors also report age-specific prevalence among the elderly, which has received little attention.
METHODS: Twenty-three cancer types were studied in men, and 24 cancer types were studied in women. Patient records were obtained from the SEER 9 cancer registries, and population figures were taken from the 2000 US Census. The authors explored the reliability of census data on the oldest old, which has been questioned.
RESULTS: Age-specific incidence, prevalence, and mortality results are presented for the years 1998 to 2002. Incidence and mortality usually decreased or plateaued at very old ages. Prevalence usually decreased swiftly at ages >90 years. When there was statistical power, incidence normally peaked between ages 75 years and 90 years, dropping abruptly afterward. With several large exceptions, peak incidence and mortality coincided within ±5 years. Both rates often trended toward zero among centenarians, who may be asymptomatic or insusceptible.
CONCLUSIONS: The current results were found to be consistent with autopsy and survival studies. Most age-specific models of carcinogenesis are based on cancer rate data for ages <85 years. The authors argue that these models could not fit the current results without fundamental modification and outline biologic mechanisms for such modification, mostly cellular and tissue senescence. They also recommend caution to researchers who use census data on the very elderly.