Dose information from 1951 to 1988 was used in calcu-
lating excess relative risk for this cohort. The data were
grouped into categories, defined in terms of the following
covariates: age (5-year intervals; table 2), calendar year (5-
year intervals), occupational group (table 3), time since first
exposure (5-year intervals), and cumulative whole body
dose (table 1). Person-years at risk and cases were allocated
on the basis of these categories and were computed accord-
ing to the same methods as in the mortality study (3).
Standardized incidence ratios were calculated using
Canadian cancer incidence rates stratified by age, sex, and
calendar year. Two-sided confidence intervals on the stan-
dardized incidence ratios were calculated under the assump-
tion that the observed incidence follows a Poisson distribution
(29).
Relative risk regression models were fitted to the data
with the AMFIT program (30), which calculates excess rel-
ative risk from internal comparisons of cumulative dose and
disease status. Specifically, the expected number of incident
cases was described by a linear excess relative risk model
Njj(1 j), where Nj is the number of person-years at risk
in the j-th stratum, j is the baseline incidence rate in the
absence of radiation exposure, dj is the cumulative dose
weighted by the number of person-years at risk, and is the
excess relative risk per unit of dose. Two-sided 90 percent
confidence limits on were also calculated. The strata were
constructed from the same age, sex, and calendar year
groupings as in the standardized incidence ratio calcula-
tions. The weighting in the calculation of dj addresses the
fact that a person’s cumulative dose generally increases over
the years that contribute to a given stratum. Confidence
intervals were based on the profile likelihood, calculated
with the AMFIT PROFILE command (30).
The cumulative dose was lagged 2 years for leukemia and
10 years for solid tumors to allow for a latent period of cancer
induction. Consequently, exposures occurring within either 2
or 10 years of incidence due to leukemia or solid tumors,
respectively, were excluded from the cumulative dose.
Dose information from 1951 to 1988 was used in calcu-lating excess relative risk for this cohort. The data weregrouped into categories, defined in terms of the followingcovariates: age (5-year intervals; table 2), calendar year (5-year intervals), occupational group (table 3), time since firstexposure (5-year intervals), and cumulative whole bodydose (table 1). Person-years at risk and cases were allocatedon the basis of these categories and were computed accord-ing to the same methods as in the mortality study (3). Standardized incidence ratios were calculated usingCanadian cancer incidence rates stratified by age, sex, andcalendar year. Two-sided confidence intervals on the stan-dardized incidence ratios were calculated under the assump-tion that the observed incidence follows a Poisson distribution(29). Relative risk regression models were fitted to the datawith the AMFIT program (30), which calculates excess rel-ative risk from internal comparisons of cumulative dose anddisease status. Specifically, the expected number of incidentcases was described by a linear excess relative risk modelNjj(1 j), where Nj is the number of person-years at riskin the j-th stratum, j is the baseline incidence rate in theabsence of radiation exposure, dj is the cumulative doseweighted by the number of person-years at risk, and is theexcess relative risk per unit of dose. Two-sided 90 percentconfidence limits on were also calculated. The strata wereconstructed from the same age, sex, and calendar yeargroupings as in the standardized incidence ratio calcula-tions. The weighting in the calculation of dj addresses thefact that a person’s cumulative dose generally increases overthe years that contribute to a given stratum. Confidenceintervals were based on the profile likelihood, calculatedwith the AMFIT PROFILE command (30). The cumulative dose was lagged 2 years for leukemia and10 years for solid tumors to allow for a latent period of cancerinduction. Consequently, exposures occurring within either 2or 10 years of incidence due to leukemia or solid tumors,respectively, were excluded from the cumulative dose.
การแปล กรุณารอสักครู่..