Materials and Methods
We applied maximum likelihood estimation (6) to SEER data (5) to estimate the parameters of a birth-cohort and age model (7). The SEER database contains information on cancer incidence in the United States by year of diagnosis and age from nine cancer registries for 1973 through 1991 and 11 registries for 1992 through 2000. The SEER registries represent 14 percent of the US population, are similar to the United States regarding measures of poverty and education, and tend to be more urban than the general US population (8–10). For males and females, we extracted counts of mesothelioma cases and corresponding population counts for each year and each 5-year age interval from 0–4 through 85 years or older. These data were realigned into 5-year birth-cohort categories (1885–1889 through 1970–1974).
Incidence estimates calculated from the model were combined with mortality rates for all causes in a survival analysis to calculate, for each birth cohort, the probability of mesothelioma occurring in each 5-year age interval, conditioned on cancer-free survival to that age interval (11). These probabilities were used to calculate the lifetime risk of mesothelioma for each birth cohort.
For all-cause mortality, we created a table of cohort mortality rates by 5-year age interval for each birth cohort. The cohort mortality rates were developed from two sources: 1) cross-sectional mortality rates for 5-year age intervals from the SEER system, years 1969–2000 (12); and 2) cross-sectional mortality rates by 10-year age intervals for individual years 1900 through 1968 (13). We factored the 10-year age intervals into two 5-year intervals and then realigned the results to form a table of cohort mortality rates for birth cohorts 1900–1904 through 1995–1999.
To project the annual number of future mesothelioma cases, we required natality data (14), mesothelioma probabilities for each age interval, and all-cause mortality rates for future birth cohorts. The number of births for all future birth cohorts was set equal to the number of births reported for the most recent birth cohort, 1995–1999.
Future birth-cohort mesothelioma probabilities for females, starting with the 1965–1969 cohort, were set to the average of past probabilities for females, which have been relatively constant since the 1900–1904 birth cohort. Mesothelioma probabilities for males have been declining toward those for females. Therefore, probabilities for males in future birth cohorts starting with 1965–1969 were set equal to the averages for females, which we interpreted as background rates for mesothelioma. All-cause mortality rates for the year 2000 were used for all future birth cohorts.