doCoxph: Fit a Cox proportional hazards regression model

In TalhoukLab/biostatUtil: Utility Functions Used in Biostatistics Projects

Fit a Cox proportional hazards regression model

Description

Fits coxph on all survival endpoints. doCoxphGeneric fits univariable models for each variable in var.names, and doCoxphMultivariable fits a multivariable model for all variables together in var.names.

Usage

doCoxphGeneric(
  input.d,
  var.names,
  var.descriptions,
  show.var.detail = FALSE,
  show.group.name.for.bin.var = FALSE,
  var.ref.groups = NULL,
  var.names.surv.time = c("os.yrs", "dss.yrs", "rfs.yrs"),
  var.names.surv.time2 = NULL,
  var.names.surv.status = c("os.sts", "dss.sts", "rfs.sts"),
  event.codes.surv = c("os.event", "dss.event", "rfs.event"),
  surv.descriptions = c("OS", "DSS", "PFS"),
  var.strata = NULL,
  missing.codes = c("N/A", "", "Unk"),
  use.firth = 1,
  firth.caption = FIRTH.CAPTION,
  add_log_hr = FALSE,
  stat.test = "waldtest",
  bold_pval = FALSE,
  sig.level = 0.05,
  round.digits.hr = 2,
  round.digits.p.value = 4,
  round.small = FALSE,
  scientific = FALSE,
  caption = NA,
  html.table.border = 0,
  banded.rows = FALSE,
  css.class.name.odd = "odd",
  css.class.name.even = "even",
  split.table = 300,
  format = c("long", "wide"),
  ...
)

doCoxphMultivariable(
  input.d,
  var.names,
  var.descriptions,
  show.var.detail = FALSE,
  show.group.name.for.bin.var = FALSE,
  var.ref.groups = NULL,
  var.names.surv.time = c("os.yrs", "dss.yrs", "rfs.yrs"),
  var.names.surv.time2 = NULL,
  var.names.surv.status = c("os.sts", "dss.sts", "rfs.sts"),
  event.codes.surv = c("os.event", "dss.event", "rfs.event"),
  surv.descriptions = c("OS", "DSS", "PFS"),
  var.strata = NULL,
  missing.codes = c("N/A", "", "Unk"),
  use.firth = 1,
  firth.caption = FIRTH.CAPTION,
  add_log_hr = FALSE,
  stat.test = "waldtest",
  bold_pval = FALSE,
  sig.level = 0.05,
  round.digits.hr = 2,
  round.digits.p.value = 4,
  round.small = FALSE,
  scientific = FALSE,
  caption = NA,
  html.table.border = 0,
  banded.rows = FALSE,
  css.class.name.odd = "odd",
  css.class.name.even = "even",
  split.table = 300,
  format = c("long", "wide"),
  ...
)

Arguments

input.d	The data.frame containing the data
var.names	variables to include as predictors
var.descriptions	vector of strings to describe the variables as they are to appear in the table
show.var.detail	logical. If TRUE, details such as categories and the reference group for categorical variables are shown.
show.group.name.for.bin.var	logical. If TRUE, the non-reference group name is shown beside the hazard ratio for dichotomized variables.
var.ref.groups	a vector of reference groups. If NULL, assume all variables are binary/continuous. If an item in the vector is NA, assume that particular marker is binary or continuous (i.e., treat it as a numeric variable)
var.names.surv.time	variable names of survival time
var.names.surv.time2	optional variable names for interval ending times, used when survival is left-truncated. The interpretation of var.names.surv.time becomes interval starting times when this argument is not NULL.
var.names.surv.status	variable names of survival status
event.codes.surv	event coding of survival status variable
surv.descriptions	names abbreviated survival endpoints in returned output
var.strata	optional variable name to stratify by using survival::strata()
missing.codes	character strings of missing values used in input.d
use.firth	percentage of censored cases before using Firth's method for Cox regression. If use.firth = 1 (default), Firth is never used and if use.firth = -1 Firth is always used.
firth.caption	subscript in html table output indicating Firth was used
add_log_hr	if TRUE, show the log hazard ratio
stat.test	the overall model test to perform on the Cox regression model. Can be any of "waldtest", "logtest", or "sctest". If Firth is used, only "logtest" can be performed. Test p-values are never Firth corrected ( per instruction from Aline 2015-04-14,15).
bold_pval	logical; if TRUE, p-values are bolded if statistically significant at sig.level
sig.level	significance level; default 0.05
round.digits.hr	number of digits for hazard ratio
round.digits.p.value	number of digits for p-value
round.small	if TRUE, uses small number rounding via round_small() for p-values
scientific	if TRUE, uses scientific notation when rounding
caption	caption for returned object
html.table.border	the border type to use for html tables
banded.rows	logical. If TRUE, rows have alternating shading colour
css.class.name.odd	Used to set the row colour for odd rows
css.class.name.even	Used to set the row colour for even rows
split.table	number of characters per row before splitting the table. Applies to the pandoc table output.
format	Either a "long" or "wide" format for the result.table.bamboo result element
...	additional arguments to pander::pandoc.table.return()

Details

Please note the following assumptions:

• Marker can be binary, continuous or categorical.

• Missing survival time/status variables are coded as NA (i.e. will only be checked by is.na()).

• Survival time/status variable name specified in the following order: "os", "dss", "rfs".

• Coding of survival status is binary only (i.e. cannot take survival status of > 2 categories).

Value

A list with the following elements

• result.table: a data frame with the Cox model results.

• result.table.bamboo: results transformed into a pandoc-ready format. Display using cat().

• result.table.html: results transformed into an HTML-ready format. Display using htmlTable::htmlTable().

• cox.stats: present for doCoxphMultivariable() only. A list of additional Cox model output statistics.

Author(s)

Samuel Leung, Aline Talhouk, Derek Chiu

Examples

library(survival)
doCoxphGeneric(input.d = lung, var.names = "sex", var.descriptions = "Sex",
               var.names.surv.time = "time",
               var.names.surv.status = "status", event.codes.surv = "2",
               surv.descriptions = "OS", caption = "")

TalhoukLab/biostatUtil documentation built on April 14, 2025, 4:15 a.m.