Skip to contents

CER: Calculating Cumulative Excess Risks

CER.Rd

Compute the cumulative excess risk (CER) attributable to radiation exposure for an individual scenario or cohort, integrating excess risks over age up to a specified maximum while accounting for competing all-cause mortality and model uncertainty. This is the core engine for lifetime risk estimation in CanEpiRisk.

Usage

CER(exposure, reference, riskmodel, option)

Arguments

exposure

list. Exposure scenario with components:

  • agex: age(s) at exposure (scalar or vector).

  • doseGy: dose(s) in gray (Gy); same length as agex if vectorized.

  • sex: sex indicator (1 = male, 2 = female).

reference

list. Baseline reference data for the same population/region with:

  • baseline: data frame of site-specific baseline rates (incidence or mortality), with columns age, male, female on ages 1:100.

  • mortality: data frame of all-cause mortality (same columns/age grid).

riskmodel

list. Radiation risk model definition with sublists for excess relative risk (ERR) and excess absolute risk (EAR), e.g.:

  • err/ear: each a list containing para (numeric parameter vector), var (variance–covariance matrix) or ci (confidence interval for 1-parameter models), and f (function of the form f(beta, data, lag) returning age-specific excess risk).

option

list. Optional settings:

  • maxage: maximum attained age for accumulation (e.g., 100).

  • err_wgt: weight to blend ERR vs EAR (1 = pure ERR; 0 = pure EAR; intermediate values allowed).

  • n_mcsamp: Monte Carlo sample size for uncertainty propagation (e.g., 10000).

  • alpha: significance level for interval estimation (default 0.05).

Value

A named numeric vector with point and interval summaries of cumulative excess risk, typically including:

  • mle: point estimate,

  • mean, median: Monte Carlo summaries,

  • ci_lo, ci_up: confidence intervals.

Values are per person; multiply by 1e4 or 1e5 to report per 10,000 or 100,000.

Details

Calculating Cumulative Excess Risks (CER)

The function integrates age-specific excess risks implied by the supplied ERR/EAR model(s) across the attained-age grid (typically ages 1–100). When n_mcsamp > 1, parameter uncertainty is propagated via Monte Carlo sampling using either the model variance–covariance matrix (var) or a 95% CI (ci) for 1-parameter models. For protracted exposures (vectorized agex/doseGy), contributions are summed across exposure segments. Baseline site-specific rates and all-cause mortality must correspond to the same population/region and share the same age grid and sex coding.

Units & Alignment

Doses must be in Gy. Ensure baseline and mortality tables are from the same population/region and aligned on age (1–100) and sex coding.

See also

population_LAR, population_YLL, YLL

Examples

set.seed(100)
# Example 1: All-solid cancer mortality (Region 1), female, 0.1 Gy at age 15
exp1 <- list(agex = 15, doseGy = 0.1, sex = 2)
ref1 <- list(
  baseline  = Mortality[[1]]$allsolid,
  mortality = Mortality[[1]]$allcause
)
mod1 <- LSS_mortality$allsolid$L
opt1 <- list(maxage = 100, err_wgt = 1, n_mcsamp = 10000)

CER(exposure = exp1, reference = ref1, riskmodel = mod1, option = opt1) * 10000
#>         mle        mean      median  ci_lo.2.5% ci_up.97.5% 
#>    156.0667    158.3529    156.4740    114.7944    214.0962 

# Example 2: Leukaemia incidence (Region 4), male, 100 mGy evenly across ages 30–45
exp2 <- list(agex = 30:44 + 0.5, doseGy = rep(0.1/15, 15), sex = 1)
ref2 <- list(
  baseline  = Incidence[[4]]$leukaemia,
  mortality = Mortality[[4]]$allcause
)
mod2 <- LSS_incidence$leukaemia$LQ
opt2 <- list(maxage = 60, err_wgt = 0, n_mcsamp = 10000)

CER(exposure = exp2, reference = ref2, riskmodel = mod2, option = opt2) * 10000
#>         mle        mean      median  ci_lo.2.5% ci_up.97.5% 
#>  3.68374227  3.62558094  3.63197695 -0.06062042  7.27448084