pgx.endpoints: Derive pharmacogenetic (PGx) safety endpoints from...

In tobyjohnson/gtx: Genetics ToolboX

Description

Function to derive safety endpoints for continuous, ordinal, binary, and time-to-event analyses, from longitudinal data typical of a clinical trial.

Usage

pgx.endpoints(d, measure, exposure,
           mtime, mvalue, threshval,
           msign = c("greater", "less"), mstrict = c(TRUE, FALSE), 
           estart = "exstdtR", eend = "exendtR",
           dby = NULL, subjid = "USUBJID")

Arguments

d	A data frame containing subject identifiers, and an optional stratifying variable
measure	A data frame containing time series trait measurements
exposure	A data frame containing therapy exposure records
mtime	Name of the variable in measure that describes time of measurement
mvalue	Name of the variable in measure that describes value of measurement
threshval	Critical value, for defining events for “sufficient exposure” to treatment and for survival analysis
msign	Indicator how mvalue is to be compared to threshval
mstrict	Indicator how mvalue is to be compared to threshval
estart	Name of the variable in exposure that describes exposure start time
eend	Name of the variable in exposure that describes exposure end time
dby	Name of the variable in d to stratify on for computing median exposure times
subjid	Name of the variable in d, measure, and exposure, that identifies individual subjects

Details

This function derives safety endpoint variables that can be used for continuous, ordinal, binary, and time-to-event analyses, from clinical trial data consisting of: (i) time series trait measurements, typically indicating the presence or severity of symptoms for the safety event; and (ii) therapy exposure records.

TO DO: Allow elapsed time, cumulative dose

Value

A data frame of derived variables.

Author(s)

Toby Johnson [e-mail]

Examples

library(survival)
data(clinex) # like plain text export of SAS files from CDER/CDISC
             # compliant data
## format dates for R
clinex$LAB <- within(clinex$LAB, {
  lbdtR <- as.Date(LBDT, "%d%b%Y")                
  grade <- grade(LBSTRESN, c(1, 5, 10), strict = TRUE)
})
clinex$EXPOSURE <- within(clinex$EXPOSURE, {
  exstdtR <- as.Date(EXSTDT, "%d%b%Y")                
  exendtR <- as.Date(EXENDT, "%d%b%Y")                
})

pgx <- clinex$DEMO
pgx <- merge(pgx,
             pgx.endpoints(clinex$DEMO, clinex$LAB, clinex$EXPOSURE,
                        mtime = "lbdtR", mvalue = "LBSTRESN", threshval = 5,
                        msign = "greater", mstrict = TRUE))
pgx <- merge(pgx,
             pgx.endpoints(clinex$DEMO, clinex$LAB, clinex$EXPOSURE,
                        mtime = "lbdtR", mvalue = "grade", threshval = 2,
                        msign = "greater", mstrict = FALSE))
pgx <- within(pgx, {
  LBSTRESN.binary <- ifelse(!LBSTRESN.exposed, NA, 
                            ifelse(LBSTRESN.event == 1, 1,
                                   ifelse(LBSTRESN.max <= 1, 0, NA)))
  ATRTGRP <- as.factor(ATRTGRP)
})

summary(lm(LBSTRESN.max ~ ATRTGRP, data = pgx, subset = LBSTRESN.exposed))
summary(glm(LBSTRESN.binary ~ ATRTGRP, family = binomial, data = pgx))
summary(coxph(Surv(LBSTRESN.time, LBSTRESN.event) ~ ATRTGRP, data = pgx))

tobyjohnson/gtx documentation built on Aug. 30, 2019, 8:07 p.m.