Immortal time bias in observational studies of time-to-event outcomes

Mark Jones, Robert Fowler

Research output: Contribution to journalArticleResearchpeer-review

20 Citations (Scopus)
5 Downloads (Pure)

Abstract

PURPOSE: The purpose of the study is to show, through simulation and example, the magnitude and direction of immortal time bias when an inappropriate analysis is used.

MATERIALS AND METHODS: We compare 4 methods of analysis for observational studies of time-to-event outcomes: logistic regression, standard Cox model, landmark analysis, and time-dependent Cox model using an example data set of patients critically ill with influenza and a simulation study.

RESULTS: For the example data set, logistic regression, standard Cox model, and landmark analysis all showed some evidence that treatment with oseltamivir provides protection from mortality in patients critically ill with influenza. However, when the time-dependent nature of treatment exposure is taken account of using a time-dependent Cox model, there is no longer evidence of a protective effect of treatment. The simulation study showed that, under various scenarios, the time-dependent Cox model consistently provides unbiased treatment effect estimates, whereas standard Cox model leads to bias in favor of treatment. Logistic regression and landmark analysis may also lead to bias.

CONCLUSIONS: To minimize the risk of immortal time bias in observational studies of survival outcomes, we strongly suggest time-dependent exposures be included as time-dependent variables in hazard-based analyses.

Original languageEnglish
Pages (from-to)195-199
Number of pages5
JournalJournal of Critical Care
Volume36
DOIs
Publication statusPublished - Dec 2016
Externally publishedYes

Fingerprint

Observational Studies
Proportional Hazards Models
Logistic Models
Critical Illness
Human Influenza
Oseltamivir
Therapeutics
Regression Analysis
Survival
Mortality

Cite this

@article{a11f16eb21454975b83bf85d2b0cc4e8,
title = "Immortal time bias in observational studies of time-to-event outcomes",
abstract = "PURPOSE: The purpose of the study is to show, through simulation and example, the magnitude and direction of immortal time bias when an inappropriate analysis is used.MATERIALS AND METHODS: We compare 4 methods of analysis for observational studies of time-to-event outcomes: logistic regression, standard Cox model, landmark analysis, and time-dependent Cox model using an example data set of patients critically ill with influenza and a simulation study.RESULTS: For the example data set, logistic regression, standard Cox model, and landmark analysis all showed some evidence that treatment with oseltamivir provides protection from mortality in patients critically ill with influenza. However, when the time-dependent nature of treatment exposure is taken account of using a time-dependent Cox model, there is no longer evidence of a protective effect of treatment. The simulation study showed that, under various scenarios, the time-dependent Cox model consistently provides unbiased treatment effect estimates, whereas standard Cox model leads to bias in favor of treatment. Logistic regression and landmark analysis may also lead to bias.CONCLUSIONS: To minimize the risk of immortal time bias in observational studies of survival outcomes, we strongly suggest time-dependent exposures be included as time-dependent variables in hazard-based analyses.",
author = "Mark Jones and Robert Fowler",
note = "Copyright {\circledC} 2016 Elsevier Inc. All rights reserved.",
year = "2016",
month = "12",
doi = "10.1016/j.jcrc.2016.07.017",
language = "English",
volume = "36",
pages = "195--199",
journal = "Seminars in Anesthesia, Perioperative Medicine and Pain",
issn = "0883-9441",
publisher = "Elsevier BV",

}

Immortal time bias in observational studies of time-to-event outcomes. / Jones, Mark; Fowler, Robert.

In: Journal of Critical Care, Vol. 36, 12.2016, p. 195-199.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Immortal time bias in observational studies of time-to-event outcomes

AU - Jones, Mark

AU - Fowler, Robert

N1 - Copyright © 2016 Elsevier Inc. All rights reserved.

PY - 2016/12

Y1 - 2016/12

N2 - PURPOSE: The purpose of the study is to show, through simulation and example, the magnitude and direction of immortal time bias when an inappropriate analysis is used.MATERIALS AND METHODS: We compare 4 methods of analysis for observational studies of time-to-event outcomes: logistic regression, standard Cox model, landmark analysis, and time-dependent Cox model using an example data set of patients critically ill with influenza and a simulation study.RESULTS: For the example data set, logistic regression, standard Cox model, and landmark analysis all showed some evidence that treatment with oseltamivir provides protection from mortality in patients critically ill with influenza. However, when the time-dependent nature of treatment exposure is taken account of using a time-dependent Cox model, there is no longer evidence of a protective effect of treatment. The simulation study showed that, under various scenarios, the time-dependent Cox model consistently provides unbiased treatment effect estimates, whereas standard Cox model leads to bias in favor of treatment. Logistic regression and landmark analysis may also lead to bias.CONCLUSIONS: To minimize the risk of immortal time bias in observational studies of survival outcomes, we strongly suggest time-dependent exposures be included as time-dependent variables in hazard-based analyses.

AB - PURPOSE: The purpose of the study is to show, through simulation and example, the magnitude and direction of immortal time bias when an inappropriate analysis is used.MATERIALS AND METHODS: We compare 4 methods of analysis for observational studies of time-to-event outcomes: logistic regression, standard Cox model, landmark analysis, and time-dependent Cox model using an example data set of patients critically ill with influenza and a simulation study.RESULTS: For the example data set, logistic regression, standard Cox model, and landmark analysis all showed some evidence that treatment with oseltamivir provides protection from mortality in patients critically ill with influenza. However, when the time-dependent nature of treatment exposure is taken account of using a time-dependent Cox model, there is no longer evidence of a protective effect of treatment. The simulation study showed that, under various scenarios, the time-dependent Cox model consistently provides unbiased treatment effect estimates, whereas standard Cox model leads to bias in favor of treatment. Logistic regression and landmark analysis may also lead to bias.CONCLUSIONS: To minimize the risk of immortal time bias in observational studies of survival outcomes, we strongly suggest time-dependent exposures be included as time-dependent variables in hazard-based analyses.

UR - http://www.scopus.com/inward/record.url?scp=84995694810&partnerID=8YFLogxK

U2 - 10.1016/j.jcrc.2016.07.017

DO - 10.1016/j.jcrc.2016.07.017

M3 - Article

VL - 36

SP - 195

EP - 199

JO - Seminars in Anesthesia, Perioperative Medicine and Pain

JF - Seminars in Anesthesia, Perioperative Medicine and Pain

SN - 0883-9441

ER -