Help for package recurrentpseudo

Title:

Creates Pseudo-Observations and Analysis for Recurrent Event Data

Version:

1.0.0

Description:

Computation of one-, two- and three-dimensional pseudo-observations based on recurrent events and terminal events. Generalised linear models are fitted using generalised estimating equations. Technical details on the bivariate procedure can be found in "Bivariate pseudo-observations for recurrent event analysis with terminal events" (Furberg et al., 2021) <doi:10.1007/s10985-021-09533-5>.

License:

GPL-2 | GPL-3 [expanded from: GPL (≥ 2)]

URL:

https://github.com/JulieKFurberg/recurrentpseudo

Encoding:

UTF-8

RoxygenNote:

7.2.1

Imports:

survival, geepack, stringr, prodlim

Depends:

magrittr, dplyr

Suggests:

knitr, rmarkdown, testthat (≥ 3.0.0)

Config/testthat/edition:

VignetteBuilder:

knitr

NeedsCompilation:

Packaged:

2022-09-19 12:23:25 UTC; JUKF

Author:

Julie Kjærulff Furberg

[aut, cre]

Maintainer:

Julie Kjærulff Furberg <julie_furberg@hotmail.com>

Repository:

CRAN

Date/Publication:

2022-09-19 14:16:16 UTC

Function that makes GEE model fit for recurrent pseudo-observations

Description

This function fits a GEE model based on pseudo-observations of the marginal mean function, and the survival probability or cumulative incidences of two death causes as returned by pseudo.onedim() (marginal mean function), or pseudo.twodim() (marginal mean function and survival probability), or pseudo.threedim() (marginal mean function and cumulative incidences of death causes 1 and 2)

Usage

pseudo.geefit(pseudodata, covar_names)

Arguments

pseudodata

Data set containing pseudo-observations. Expecting output from pseudo.twodim()

covar_names

Vector with covariate names to be found in "pseudodata". E.g. covar_names = c("Z", "Z1")

Value

An object of class pseudo.geefit.

xi contains the estimated model parameters
sigma contains the estimated variance matrix corresponding to xi

References

Furberg, J.K., Andersen, P.K., Korn, S. et al. Bivariate pseudo-observations for recurrent event analysis with terminal events. Lifetime Data Anal (2021). https://doi.org/10.1007/s10985-021-09533-5

Examples

# Bladder cancer data from survival package
require(survival)

# Make a three level status variable
bladder1$status3 <- ifelse(bladder1$status %in% c(2, 3), 2, bladder1$status)

# Add one extra day for the two patients with start=stop=0
# subset(bladder1, stop <= start)
bladder1[bladder1$id == 1, "stop"] <- 1
bladder1[bladder1$id == 49, "stop"] <- 1

# Restrict the data to placebo and thiotepa
bladdersub <- subset(bladder1, treatment %in% c("placebo", "thiotepa"))

# Make treatment variable two-level factor
bladdersub$Z <- as.factor(ifelse(bladdersub$treatment == "placebo", 0, 1))
levels(bladdersub$Z) <- c("placebo", "thiotepa")

head(bladdersub)

# Two-dimensional (bivariate pseudo-obs) model fit

# Computation of pseudo-observations
pseudo_bladder_2d <- pseudo.twodim(tstart = bladdersub$start,
                                   tstop = bladdersub$stop,
                                   status = bladdersub$status3,
                                   id = bladdersub$id,
                                   covar_names = "Z",
                                   tk = c(30),
                                   data = bladdersub)

# Data in wide format
head(pseudo_bladder_2d$outdata)

# Data in long format
head(pseudo_bladder_2d$outdata_long)

# GEE fit
fit_bladder_2d <- pseudo.geefit(pseudodata = pseudo_bladder_2d,
                                covar_names = c("Z"))
fit_bladder_2d

Function that computes univariate pseudo-observations

Description

This function computes univariate pseudo-observations of the marginal mean function (in the presence of terminal events)

Usage

pseudo.onedim(tstart, tstop, status, covar_names, id, tk, data)

Arguments

tstart

Start time - expecting counting process notation

tstop

Stop time - expecting counting process notation

status

Status variable (0 = censoring, 1 = recurrent event, 2 = death)

covar_names

Vector containing names of covariates intended for further analysis

id

ID variable for subject

tk

Vector of time points to calculate pseudo-observations at

data

Data set which contains variables of interest

Value

An object of class pseudo.onedim.

outdata contains the semi-wide version of the computed pseudo-observations (one row per time, tk, per id).
outdata_long contains the long version of the computed pseudo-observations (one row per observation, several per id).
indata contains the input data which the pseudo-observations are based on.
ts vector with time points used for computation of pseudo-observations.
k number of time points used for computation of pseudo-observations (length(ts)).

References

Furberg, J.K., Andersen, P.K., Korn, S. et al. Bivariate pseudo-observations for recurrent event analysis with terminal events. Lifetime Data Anal (2021). https://doi.org/10.1007/s10985-021-09533-5

Examples

# Example: Bladder cancer data from survival package
require(survival)

# Make a three level status variable
bladder1$status3 <- ifelse(bladder1$status %in% c(2, 3), 2, bladder1$status)

# Add one extra day for the two patients with start=stop=0
# subset(bladder1, stop <= start)
bladder1[bladder1$id == 1, "stop"] <- 1
bladder1[bladder1$id == 49, "stop"] <- 1

# Restrict the data to placebo and thiotepa
bladdersub <- subset(bladder1, treatment %in% c("placebo", "thiotepa"))

# Make treatment variable two-level factor
bladdersub$Z <- as.factor(ifelse(bladdersub$treatment == "placebo", 0, 1))
levels(bladdersub$Z) <- c("placebo", "thiotepa")
head(bladdersub)

# Pseudo observations
pseudo_bladder_1d <- pseudo.onedim(tstart = bladdersub$start,
                                   tstop = bladdersub$stop,
                                   status = bladdersub$status3,
                                   id = bladdersub$id,
                                   covar_names = "Z",
                                   tk = c(30),
                                   data = bladdersub)
head(pseudo_bladder_1d$outdata)

# GEE fit
fit_bladder_1d <- pseudo.geefit(pseudodata = pseudo_bladder_1d,
                                covar_names = c("Z"))
fit_bladder_1d

Function that computes 3-dim pseudo-observations

Description

This function computes 3-dimensional pseudo-observations of the marginal mean function (in the presence of terminal events) and cumulative incidences of death causes 1 and 2

Usage

pseudo.threedim(tstart, tstop, status, covar_names, id, tk, data, deathtype)

Arguments

tstart

Start time - expecting counting process notation

tstop

Stop time - expecting counting process notation

status

Status variable (0 = censoring, 1 = recurrent event, 2 = death)

covar_names

Vector containing names of covariates intended for further analysis

id

ID variable for subject

tk

Vector of time points to calculate pseudo-observations at

data

Data set which contains variables of interest

deathtype

Type of death (cause 1 or cause 2)

Value

An object of class pseudo.threedim.

outdata contains the semi-wide version of the computed pseudo-observations (one row per time, tk, per id).
outdata_long contains the long version of the computed pseudo-observations (one row per observation, several per id).
indata contains the input data which the pseudo-observations are based on.
ts vector with time points used for computation of pseudo-observations.
k number of time points used for computation of pseudo-observations (length(ts)).

References

Furberg, J.K., Andersen, P.K., Korn, S. et al. Bivariate pseudo-observations for recurrent event analysis with terminal events. Lifetime Data Anal (2021). https://doi.org/10.1007/s10985-021-09533-5

Examples

# Example: Bladder cancer data from survival package
require(survival)

# Make a three level status variable
bladder1$status3 <- ifelse(bladder1$status %in% c(2, 3), 2, bladder1$status)

# Add one extra day for the two patients with start=stop=0
# subset(bladder1, stop <= start)
bladder1[bladder1$id == 1, "stop"] <- 1
bladder1[bladder1$id == 49, "stop"] <- 1

# Restrict the data to placebo and thiotepa
bladdersub <- subset(bladder1, treatment %in% c("placebo", "thiotepa"))

# Make treatment variable two-level factor
bladdersub$Z <- as.factor(ifelse(bladdersub$treatment == "placebo", 0, 1))
levels(bladdersub$Z) <- c("placebo", "thiotepa")
head(bladdersub)

# Add deathtype variable to bladder data
# Deathtype = 1 (bladder disease death), deathtype = 2 (other death reason)
bladdersub$deathtype <- with(bladdersub, ifelse(status == 2, 1, ifelse(status == 3, 2, 0)))
table(bladdersub$deathtype, bladdersub$status)

# Pseudo-observations
pseudo_bladder_3d <- pseudo.threedim(tstart = bladdersub$start,
                                     tstop = bladdersub$stop,
                                     status = bladdersub$status3,
                                     id = bladdersub$id,
                                     deathtype = bladdersub$deathtype,
                                     covar_names = "Z",
                                     tk = c(30),
                                     data = bladdersub)
pseudo_bladder_3d

# GEE fit
fit_bladder_3d <- pseudo.geefit(pseudodata = pseudo_bladder_3d,
                                covar_names = c("Z"))
fit_bladder_3d

Function that computes bivariate pseudo-observations

Description

This function computes bivariate pseudo-observations of the marginal mean function (in the presence of terminal events) and the survival probability

Usage

pseudo.twodim(tstart, tstop, status, covar_names, id, tk, data)

Arguments

tstart

Start time - expecting counting process notation

tstop

Stop time - expecting counting process notation

status

Status variable (0 = censoring, 1 = recurrent event, 2 = death)

covar_names

Vector containing names of covariates intended for further analysis

id

ID variable for subject

tk

Vector of time points to calculate pseudo-observations at

data

Data set which contains variables of interest

Value

An object of class pseudo.twodim.

outdata contains the semi-wide version of the computed pseudo-observations (one row per time, tk, per id).
outdata_long contains the long version of the computed pseudo-observations (one row per observation, several per id).
indata contains the input data which the pseudo-observations are based on.
ts vector with time points used for computation of pseudo-observations.
k number of time points used for computation of pseudo-observations (length(ts)).

References

Furberg, J.K., Andersen, P.K., Korn, S. et al. Bivariate pseudo-observations for recurrent event analysis with terminal events. Lifetime Data Anal (2021). https://doi.org/10.1007/s10985-021-09533-5

Examples

# Example: Bladder cancer data from survival package
require(survival)

# Make a three level status variable
bladder1$status3 <- ifelse(bladder1$status %in% c(2, 3), 2, bladder1$status)

# Add one extra day for the two patients with start=stop=0
# subset(bladder1, stop <= start)
bladder1[bladder1$id == 1, "stop"] <- 1
bladder1[bladder1$id == 49, "stop"] <- 1

# Restrict the data to placebo and thiotepa
bladdersub <- subset(bladder1, treatment %in% c("placebo", "thiotepa"))

# Make treatment variable two-level factor
bladdersub$Z <- as.factor(ifelse(bladdersub$treatment == "placebo", 0, 1))
levels(bladdersub$Z) <- c("placebo", "thiotepa")
head(bladdersub)

# Pseudo observations
pseudo_bladder_2d <- pseudo.twodim(tstart = bladdersub$start,
                                   tstop = bladdersub$stop,
                                   status = bladdersub$status3,
                                   id = bladdersub$id,
                                   covar_names = "Z",
                                   tk = c(30),
                                   data = bladdersub)
head(pseudo_bladder_2d$outdata)

# GEE fit
fit_bladder_2d <- pseudo.geefit(pseudodata = pseudo_bladder_2d,
                                covar_names = c("Z"))
fit_bladder_2d