Covariate imbalance between treatment groups makes it difficult to compare cumulative incidence curves in competing risk analyses. Different methods have been developed to estimate adjusted cumulative incidence curves, including inverse probability of treatment weighting and outcome regression modeling. For these methods to work, correct specification of the propensity score model or outcome regression model, respectively, is needed. We introduce a new doubly robust estimator, which requires correct specification of only one of the two models. We have conducted a simulation study to assess the performance of these three methods, including scenarios with model misspecification of the relationship between covariates and treatment and/or outcome. This repository provides an overview of these adjustments methods and their validation.
The repository contains the following functions related to covariate adjusted for competing risks:
- R/adjOR.R : A function to produce to perform covariate adjustment of cumulative incidence curves via outcome regression with standardization using cause-specific Cox proportional hazard models
- R/adjDR.R : A function to produce to perform covariate adjustment of cumulative incidence curves via doubly robust estimation using cause-specific Cox proportional hazard models
- R/adjIPW.R: A function which produces observation weights by taking the inverse of probability of allocated treatment estimated by standard logistic regression
The following scripts were used to validate these methods in a simulation study:
- adj_comprisk_simstudy.Rmd : The Rmd-script used to validate the correction methods. Produces the tables and figures of the manuscript related to the simulation study.
- R/longCSH.R : A helper function which translates output from survfit to a long format data frame
- R/simCSH.R: Simulation of competing risks data from a cause-specific hazard model as a function of random strata allocation, a discrete covariate, and a continuous covariate
- R/trueCSH.R: A function to analytically obtain the hypothetical true cause-specific hazard curve as a function of strata allocation and covariates
- R/confCSH.R: Simulation of competing risk data from cause-specific hazard models with biased strata allocation as a function of a discrete and continuous covariate
R users
remotes::install_github("survival-lumc/AdjCuminc")Git users
git clone https://github.com/survival-lumc/AdjCuminc.gitAlternatively, a ZIP file can be found at the top-right corner of this Github page, which can be stored and extracted a directory of choice.
The code can be accesed by opening the adjCuminc.Rproj file in Rstudio.
| Name | Affiliation |
|---|---|
| Patrick van Hage | Institute of Biology Leiden (IBL) |
| Saskia le Cessie | University Medical Centre Utrecht (NL) |
| Marissa van Maaren | University of Twente |
| Hein Putter | Leiden University Medical Center (NL) |
| Nan van Geloven | Leiden University Medical Center (NL) |