Nonparametric comparison of two survival functions with dependent censoring via nonparametric multiple imputation

Chiu-Hsieh Hsu, Jeremy M G Taylor

Research output: Contribution to journalArticle

8 Citations (Scopus)

Abstract

When the event time of interest depends on the censoring time, conventional two-sample test methods, such as the log-rank and Wilcoxon tests, can produce an invalid test result. We extend our previous work on estimation using auxiliary variables to adjust for dependent censoring via multiple imputation, to the comparison of two survival distributions. To conduct the imputation, we use two working models to define a set of similar observations called the imputing risk set. One model is for the event times and the other for the censoring times. Based on the imputing risk set, a nonparametric multiple imputation method, Kaplan-Meier imputation, is used to impute a future event or censoring time for each censored observation. After imputation, the conventional nonparametric two-sample tests can be easily implemented on the augmented data sets. Simulation studies show that the sizes of the log-rank and Wilcoxon tests constructed on the imputed data sets are comparable to the nominal level and the powers are much higher compared with the tests based on the unimputed data in the presence of dependent censoring if either one of the two working models is correctly specified. The method is illustrated using AIDS clinical trial data comparing ZDV and placebo, in which CD4 count is the time-dependent auxiliary variable.

Original languageEnglish (US)
Pages (from-to)462-475
Number of pages14
JournalStatistics in Medicine
Volume28
Issue number3
DOIs
StatePublished - Feb 10 2009

Fingerprint

Dependent Censoring
Multiple Imputation
Survival Function
Imputation
Censoring
Wilcoxon Test
Log-rank Test
Two-sample Test
Auxiliary Variables
Nonparametric Statistics
Kaplan-Meier
Censored Observations
Survival Distribution
Non-parametric test
Clinical Trials
Categorical or nominal
CD4 Lymphocyte Count
Count
Simulation Study
Model

Keywords

  • Auxiliary variables
  • Dependent censoring
  • Log-rank test
  • Multiple imputation
  • Wilcoxon test

ASJC Scopus subject areas

  • Epidemiology
  • Statistics and Probability

Cite this

Nonparametric comparison of two survival functions with dependent censoring via nonparametric multiple imputation. / Hsu, Chiu-Hsieh; Taylor, Jeremy M G.

In: Statistics in Medicine, Vol. 28, No. 3, 10.02.2009, p. 462-475.

Research output: Contribution to journalArticle

@article{5c9016b6918c421792745782181c17a6,
title = "Nonparametric comparison of two survival functions with dependent censoring via nonparametric multiple imputation",
abstract = "When the event time of interest depends on the censoring time, conventional two-sample test methods, such as the log-rank and Wilcoxon tests, can produce an invalid test result. We extend our previous work on estimation using auxiliary variables to adjust for dependent censoring via multiple imputation, to the comparison of two survival distributions. To conduct the imputation, we use two working models to define a set of similar observations called the imputing risk set. One model is for the event times and the other for the censoring times. Based on the imputing risk set, a nonparametric multiple imputation method, Kaplan-Meier imputation, is used to impute a future event or censoring time for each censored observation. After imputation, the conventional nonparametric two-sample tests can be easily implemented on the augmented data sets. Simulation studies show that the sizes of the log-rank and Wilcoxon tests constructed on the imputed data sets are comparable to the nominal level and the powers are much higher compared with the tests based on the unimputed data in the presence of dependent censoring if either one of the two working models is correctly specified. The method is illustrated using AIDS clinical trial data comparing ZDV and placebo, in which CD4 count is the time-dependent auxiliary variable.",
keywords = "Auxiliary variables, Dependent censoring, Log-rank test, Multiple imputation, Wilcoxon test",
author = "Chiu-Hsieh Hsu and Taylor, {Jeremy M G}",
year = "2009",
month = "2",
day = "10",
doi = "10.1002/sim.3480",
language = "English (US)",
volume = "28",
pages = "462--475",
journal = "Statistics in Medicine",
issn = "0277-6715",
publisher = "John Wiley and Sons Ltd",
number = "3",

}

TY - JOUR

T1 - Nonparametric comparison of two survival functions with dependent censoring via nonparametric multiple imputation

AU - Hsu, Chiu-Hsieh

AU - Taylor, Jeremy M G

PY - 2009/2/10

Y1 - 2009/2/10

N2 - When the event time of interest depends on the censoring time, conventional two-sample test methods, such as the log-rank and Wilcoxon tests, can produce an invalid test result. We extend our previous work on estimation using auxiliary variables to adjust for dependent censoring via multiple imputation, to the comparison of two survival distributions. To conduct the imputation, we use two working models to define a set of similar observations called the imputing risk set. One model is for the event times and the other for the censoring times. Based on the imputing risk set, a nonparametric multiple imputation method, Kaplan-Meier imputation, is used to impute a future event or censoring time for each censored observation. After imputation, the conventional nonparametric two-sample tests can be easily implemented on the augmented data sets. Simulation studies show that the sizes of the log-rank and Wilcoxon tests constructed on the imputed data sets are comparable to the nominal level and the powers are much higher compared with the tests based on the unimputed data in the presence of dependent censoring if either one of the two working models is correctly specified. The method is illustrated using AIDS clinical trial data comparing ZDV and placebo, in which CD4 count is the time-dependent auxiliary variable.

AB - When the event time of interest depends on the censoring time, conventional two-sample test methods, such as the log-rank and Wilcoxon tests, can produce an invalid test result. We extend our previous work on estimation using auxiliary variables to adjust for dependent censoring via multiple imputation, to the comparison of two survival distributions. To conduct the imputation, we use two working models to define a set of similar observations called the imputing risk set. One model is for the event times and the other for the censoring times. Based on the imputing risk set, a nonparametric multiple imputation method, Kaplan-Meier imputation, is used to impute a future event or censoring time for each censored observation. After imputation, the conventional nonparametric two-sample tests can be easily implemented on the augmented data sets. Simulation studies show that the sizes of the log-rank and Wilcoxon tests constructed on the imputed data sets are comparable to the nominal level and the powers are much higher compared with the tests based on the unimputed data in the presence of dependent censoring if either one of the two working models is correctly specified. The method is illustrated using AIDS clinical trial data comparing ZDV and placebo, in which CD4 count is the time-dependent auxiliary variable.

KW - Auxiliary variables

KW - Dependent censoring

KW - Log-rank test

KW - Multiple imputation

KW - Wilcoxon test

UR - http://www.scopus.com/inward/record.url?scp=63049126380&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=63049126380&partnerID=8YFLogxK

U2 - 10.1002/sim.3480

DO - 10.1002/sim.3480

M3 - Article

C2 - 18991250

AN - SCOPUS:63049126380

VL - 28

SP - 462

EP - 475

JO - Statistics in Medicine

JF - Statistics in Medicine

SN - 0277-6715

IS - 3

ER -