br Table br Cumulative ambient exposure by demographics and
Cumulative ambient exposure by demographics and exposure activities.
f/cc, fibers per cubic centimeter. DK/M, do not know or missing.
a Asbestos exposure based on calculation of modeled residential ambient exposure and months of residence. No calculated cumulative ambient asbestos exposure for 687 cohort participants.
To analyze outcomes by exposure concentration, we took the nat-ural-log transformed cumulative ambient exposure and stratified by percentiles < 50th, 50th-75th, and > 75th.
2.3. Cancer incidence and mortality outcomes
The study cohort (n = 5848) was linked to the Minnesota Cancer Surveillance System (MCSS) to identify incident cases of mesothelioma (ICD-O-3 histology codes: 9050–9055), lung cancer (ICD-O-3 site codes: C340-C349), and all-cancers combined that were diagnosed during 1988–2010, among Minnesota residents. Follow-up began in 1988, the year MCSS was initiated and one year before the close of the vermi-culite processing facility in 1989. A descriptive case review of work history, asbestos exposure, disease onset, and cause of death was con-ducted for incident cases of mesothelioma.
The study cohort was also linked to Minnesota vital records for the PD98059 1988–2011 (matches = 793 deaths). The remaining cohort members presumed alive (n = 5054) were submitted to National Death Index (NDI-PLUS) for the years 2002–2011 to obtain vital status and cause of death information. Those known to be deceased, but missing a cause of death code, were also submitted to NDI based on their avail-able date of death. Altogether, from January 1, 1988 to December 31, 2011 state and national death records searches identified 847 deaths in the cohort. Our mortality analysis was restricted to 526 deaths that occurred from 2001 to 2011, or persons who were alive at the time of the initial interview.
2.4. Statistical analysis
Mesothelioma and lung cancer proportional incidence ratios (PIRs), along with 95% confidence intervals (CIs), were calculated for the years 1988–2010 and adjusted by age (20 year groupings: 20–39, 40–59, 60–79, 80+) and gender. Exact confidence intervals (95%) were cal-culated assuming a Poisson distribution for the observed frequency. PIRs compare the ratio of a specific cancer (e.g., mesothelioma) to all cancers in the study population and in a reference (or standard) po-pulation (Boyle and Parkin, 1991; Mills and Shah, 2014; Ross et al., 2003).
R = observed cases for the cancer of interest in the group under study.
E = expected cases for the cancer of interest in the group under study.
ri* = number of cases of the cancer of interest in the age group i in the standard population.
ti* = number of cases of cancer (all sites) in the age group i in the standard population.
ti = number of cases of cancer (all sites) in the age group i in the study cohort.
In selecting the reference population, we found little difference between the seven-county Minneapolis-St. Paul metropolitan area me-sothelioma rates and the Minnesota statewide mesothelioma rates; we used the statewide rate for greater precision. The PIR can be biased if the overall cancer rate in the study popu-lation differs from the reference population. To assess this possible bias, cancer registry data from two adjoining zip codes that contain the study area were used to estimate the all-cancer SIRs for the cohort. This analysis indicated about a 4% all-cancer deficit in males and a 6% all-cancer deficit in females (5% overall) in the population of the two zip codes compared to statewide cancer rates. The cohort PIRs were then adjusted for this lower cancer incidence using the simple adjustment
method developed for proportional mortality ratios (Wong and Decoufle, 1982).
We compared the mortality rates of the cohort with that of the Minnesota population to calculate standardized mortality ratios (SMRs) and 95% CIs adjusted for gender, five year age groups, and calendar period. Person-time at risk was accrued from the date the participant was enrolled until the date of death or the end of the follow-up period (December 31, 2011); there were 51,006 person-years at risk for the cohort. The expected number of deaths were calculated by applying age, calendar time, and cause-specific mortality rates of the Minnesota population to the person-year observations of the study population. SMRs were obtained by computing the ratio of the observed-to-ex-pected number of deaths for the overall mortality and specific causes of death. The mortality datasets were created in SAS 9.3 and SMRs were calculated using the Life Table Analysis System (LTAS) V.3.1 software (National Institute for Occupational Safety and Health (NIOSH), 2013).