Sensitivity of the hazard ratio to nonignorable treatment assignment in an observational study

Nandita Mitra, Daniel F. Heitjan

Research output: Contribution to journalArticle

43 Citations (Scopus)

Abstract

In non-randomized studies, estimation of treatment effects generally requires adjustment for imbalances in observed covariates. One such method, based on the propensity score, is useful in many applications but may be biased when the assumption of strongly ignorable treatment assignment is violated. Because it is not possible to evaluate this assumption from the data, it is advisable to assess the sensitivity of conclusions to violations of strong ignorability. Lin et al. (Biomet. 1998; 54:948-963) have implemented this idea by investigating how an unmeasured covariate may affect the conclusions of an observational study. We extend their method to assess sensitivity of the treatment hazard ratio to hidden bias under a range of covariate distributions. We derive simple formulas for approximating the true from the apparent treatment hazard ratio estimated under a specific survival model, and assess the validity of these formulas in simulation studies. We demonstrate the method in an analysis of SEER-Medicare data on the effects of chemotherapy in elderly colon cancer patients.

Original languageEnglish (US)
Pages (from-to)1398-1414
Number of pages17
JournalStatistics in Medicine
Volume26
Issue number6
DOIs
StatePublished - Mar 15 2007

Fingerprint

Observational Study
Hazard
Observational Studies
Covariates
Assignment
Ignorability
Propensity Score
Survival Model
Chemotherapy
Treatment Effects
Biased
Cimetidine
Cancer
Adjustment
Therapeutics
Medicare
Simulation Study
Colonic Neoplasms
Evaluate
Drug Therapy

Keywords

  • Accelerated failure time model
  • Propensity score
  • Sensitivity analysis

ASJC Scopus subject areas

  • Epidemiology

Cite this

Sensitivity of the hazard ratio to nonignorable treatment assignment in an observational study. / Mitra, Nandita; Heitjan, Daniel F.

In: Statistics in Medicine, Vol. 26, No. 6, 15.03.2007, p. 1398-1414.

Research output: Contribution to journalArticle

@article{ace8b8db40ef4676a78f1c8d8fad4b10,
title = "Sensitivity of the hazard ratio to nonignorable treatment assignment in an observational study",
abstract = "In non-randomized studies, estimation of treatment effects generally requires adjustment for imbalances in observed covariates. One such method, based on the propensity score, is useful in many applications but may be biased when the assumption of strongly ignorable treatment assignment is violated. Because it is not possible to evaluate this assumption from the data, it is advisable to assess the sensitivity of conclusions to violations of strong ignorability. Lin et al. (Biomet. 1998; 54:948-963) have implemented this idea by investigating how an unmeasured covariate may affect the conclusions of an observational study. We extend their method to assess sensitivity of the treatment hazard ratio to hidden bias under a range of covariate distributions. We derive simple formulas for approximating the true from the apparent treatment hazard ratio estimated under a specific survival model, and assess the validity of these formulas in simulation studies. We demonstrate the method in an analysis of SEER-Medicare data on the effects of chemotherapy in elderly colon cancer patients.",
keywords = "Accelerated failure time model, Propensity score, Sensitivity analysis",
author = "Nandita Mitra and Heitjan, {Daniel F.}",
year = "2007",
month = "3",
day = "15",
doi = "10.1002/sim.2606",
language = "English (US)",
volume = "26",
pages = "1398--1414",
journal = "Statistics in Medicine",
issn = "0277-6715",
publisher = "John Wiley and Sons Ltd",
number = "6",

}

TY - JOUR

T1 - Sensitivity of the hazard ratio to nonignorable treatment assignment in an observational study

AU - Mitra, Nandita

AU - Heitjan, Daniel F.

PY - 2007/3/15

Y1 - 2007/3/15

N2 - In non-randomized studies, estimation of treatment effects generally requires adjustment for imbalances in observed covariates. One such method, based on the propensity score, is useful in many applications but may be biased when the assumption of strongly ignorable treatment assignment is violated. Because it is not possible to evaluate this assumption from the data, it is advisable to assess the sensitivity of conclusions to violations of strong ignorability. Lin et al. (Biomet. 1998; 54:948-963) have implemented this idea by investigating how an unmeasured covariate may affect the conclusions of an observational study. We extend their method to assess sensitivity of the treatment hazard ratio to hidden bias under a range of covariate distributions. We derive simple formulas for approximating the true from the apparent treatment hazard ratio estimated under a specific survival model, and assess the validity of these formulas in simulation studies. We demonstrate the method in an analysis of SEER-Medicare data on the effects of chemotherapy in elderly colon cancer patients.

AB - In non-randomized studies, estimation of treatment effects generally requires adjustment for imbalances in observed covariates. One such method, based on the propensity score, is useful in many applications but may be biased when the assumption of strongly ignorable treatment assignment is violated. Because it is not possible to evaluate this assumption from the data, it is advisable to assess the sensitivity of conclusions to violations of strong ignorability. Lin et al. (Biomet. 1998; 54:948-963) have implemented this idea by investigating how an unmeasured covariate may affect the conclusions of an observational study. We extend their method to assess sensitivity of the treatment hazard ratio to hidden bias under a range of covariate distributions. We derive simple formulas for approximating the true from the apparent treatment hazard ratio estimated under a specific survival model, and assess the validity of these formulas in simulation studies. We demonstrate the method in an analysis of SEER-Medicare data on the effects of chemotherapy in elderly colon cancer patients.

KW - Accelerated failure time model

KW - Propensity score

KW - Sensitivity analysis

UR - http://www.scopus.com/inward/record.url?scp=33847694752&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33847694752&partnerID=8YFLogxK

U2 - 10.1002/sim.2606

DO - 10.1002/sim.2606

M3 - Article

C2 - 16779876

AN - SCOPUS:33847694752

VL - 26

SP - 1398

EP - 1414

JO - Statistics in Medicine

JF - Statistics in Medicine

SN - 0277-6715

IS - 6

ER -