iCARE (version 1.0.0)

computeAbsoluteRiskSplitInterval: Building and Applying an Absolute Risk Model: Compute Risk over Interval Split in Two Parts

Description

This function is used to build an absolute risk model that incorporates different input parameters before and after a given time point. The model is then applied to estimate absolute risks.

Usage

computeAbsoluteRiskSplitInterval(apply.age.start, apply.age.interval.length, 
      apply.cov.profile, model.formula, model.disease.incidence.rates, model.log.RR, 
      model.ref.dataset, model.ref.dataset.weights=NULL, model.cov.info, use.c.code=1, 
      model.competing.incidence.rates=NULL, return.lp=FALSE, apply.snp.profile=NULL, 
      model.snp.info=NULL, model.bin.fh.name=NULL, cut.time=NULL, 
      apply.cov.profile.2=NULL, model.formula.2=NULL, model.log.RR.2=NULL, 
      model.ref.dataset.2=NULL, model.ref.dataset.weights.2=NULL, model.cov.info.2=NULL, 
      model.bin.fh.name.2=NULL, n.imp=5, return.refs.risk=FALSE)

Arguments

apply.age.start

single integer or vector of integer ages for the start of the interval over which to compute absolute risk.

apply.age.interval.length

single integer or vector of integer years over which absolute risk should be computed.

apply.cov.profile

dataframe containing the covariate profiles for which absolute risk will be computed. Covariates must be in same order with same names as in model.formula.

model.formula

an object of class formula: a symbolic description of the model to be fitted, e.g. Y~Parity+FamilyHistory.

model.disease.incidence.rates

two column matrix [ integer ages, incidence rates] or three column matrix [start age, end age, rate] with incidence rate of disease. Must fully cover age interval for estimation.

model.log.RR

vector with log odds ratios corresponding to the model params; no intercept; names must match design matrix arising from model.formula and model.cov.info; check names using function check_design_matrix().

model.ref.dataset

dataframe of risk factors for a sample of subjects representative of underlying population, no missing values. Variables must be in same order with same names as in model.formula.

model.ref.dataset.weights

optional vector of sampling weights for model.ref.dataset.

model.cov.info

contains information about the risk factors in the model ; a main list containing a list for each covariate, which must have the fields:

• "name": a string with the covariate name, matching name in model.formula
• "type": a string that is either "continuous" or "factor".

If factor variable, then:

• "levels": vector with strings of level names
• "ref": optional field, string with name of referent level

use.c.code

binary indicator of whether to run the c program for fast computation.

model.competing.incidence.rates

two column matrix [ integer ages, incidence rates] or three column matrix [start age, end age, rate] with incidence rate of competing events. Must fully cover age interval for estimation.

return.lp

binary indicator of whether to return the linear predictor for each subject in apply.cov.profile.

apply.snp.profile

data frame with observed SNP data (coded 0,1, 2, or NA). May have missing values.

model.snp.info

dataframe with three columns [ rs number, odds ratio, allele frequency ]

model.bin.fh.name

string name of family history variable, if in model. This must refer to a variable that only takes values 0,1, NA.

cut.time

integer age for which to split computation into before and after

apply.cov.profile.2

see apply.cov.profile, to be used for estimation in ages after the cutpoint

model.formula.2

see model.formula, to be used for estimation in ages after the cutpoint

model.log.RR.2

see model.log.RR, to be used for estimation in ages after the cutpoint

model.ref.dataset.2

see model.ref.dataset, to be used for estimation in ages after the cutpoint

model.ref.dataset.weights.2

see model.ref.dataset.weights, to be used for estimation in ages after the cutpoint

model.cov.info.2

see model.cov.info, to be used for estimation in ages after the cutpoint

model.bin.fh.name.2

see model.bin.fh.name, to be used for estimation in ages after the cutpoint

n.imp

integer value for number of imputations for handling missing SNPs.

return.refs.risk

binary indicator of whether to return the absolute risk prediction for each subject in model.ref.dataset.

Value

• This function returns a list of results objects, including:
  • risk: absolute risk estimates over the specified interval for subjects given byapply.cov.profile
  • details: dataframe with the start of the interval, the end of the interval, the covariate profile, and the risk estimates for each individual
  • beta.used: the log odds ratios used in the model
  • lps.1: linear predictors based on first set of parameters for subjects inmodel.cov.profile, if requested byreturn.lp
  • lps.2: linear predictors based on second set of parameters for subjects inmodel.cov.profile, if requested byreturn.lp
  • refs.risk: absolute risk estimates for subjects inmodel.ref.dataset, if requested byreturn.refs.risk; computes for first age interval provided

Details

Individualized Coherent Absolute Risk Estimators (iCARE) is a tool that allows researchers to quickly build models for absolute risk and apply them to estimate individuals' risk based on a set of user defined input parameters. The software gives users the flexibility to change or update models rapidly based on new risk factors or tailor models to different populations based on the specification of simply three input arguments:

• (1) a model for relative risk assumed to be externally derived
• (2) an age-specific disease incidence rate and
• (3) the distribution of risk factors for the population of interest.

The tool can handle missing information on risk factors for risk estimation using an approach where all estimates are derived from a single model through appropriate model averaging.

Examples

data(bc_data, package="iCARE")

form <- caco ~ famhist + as.factor(parity)
results <- computeAbsoluteRiskSplitInterval(model.formula=form, 
                                         cut.time = 50,
                                         model.cov.info       = bc_model_cov_info,
                                         model.snp.info       = bc_15_snps,
                                         model.log.RR         = bc_model_log_or,
                                         model.log.RR.2       = bc_model_log_or_post_50,
                                         model.ref.dataset    = ref_cov_dat,
                                         model.ref.dataset.2  = ref_cov_dat_post_50,
                                         model.disease.incidence.rates   = bc_inc,
                                         model.competing.incidence.rates = mort_inc, 
                                         model.bin.fh.name = "famhist",
                                         apply.age.start    = 30, 
                                         apply.age.interval.length = 40,
                                         apply.cov.profile  = new_cov_prof,
                                         apply.snp.profile  = new_snp_prof, 
                                         return.refs.risk   = TRUE)
summary(results$risk)
plot(density(results$risk, na.rm=TRUE))
boxplot(results$risk ~ new_cov_prof$famhist, na.rm=TRUE)