Analysis of subgroup effects in randomized trials when subgroup membership is missing: Application to the second Multicenter Automatic Defibrillator Intervention Trial

Daniel O Scharfstein, Georgiana Onicescu, Steven Goodman, Rachel Whitaker

Research output: Contribution to journalArticle

Abstract

We develop and implement a methodology for drawing inference about subgroup effects in a two-arm randomized trial when subgroup status is known only for a non-random sample in each of the trial arms. Since the subgroup effects are not point identified from the distribution of the observed data, we show how to compute bounds on these effects by using scientifically plausible assumptions. We characterize the uncertainty of our procedure by using the Bayesian paradigm. The methodology is developed in the context of the second Multicenter Automatic Defibrillator Intervention Trial, which was a randomized trial designed to evaluate the effectiveness of implantable defibrillators on survival.

Original languageEnglish (US)
Pages (from-to)607-617
Number of pages11
JournalJournal of the Royal Statistical Society. Series C: Applied Statistics
Volume60
Issue number4
DOIs
StatePublished - Aug 2011

Fingerprint

Randomized Trial
Subgroup
Methodology
Paradigm
Uncertainty
Evaluate
Randomized trial

Keywords

  • Bounds
  • Expert opinion
  • Identifiability

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Cite this

@article{f0ea76ea376c429d9b1bbafcf612d1e1,
title = "Analysis of subgroup effects in randomized trials when subgroup membership is missing: Application to the second Multicenter Automatic Defibrillator Intervention Trial",
abstract = "We develop and implement a methodology for drawing inference about subgroup effects in a two-arm randomized trial when subgroup status is known only for a non-random sample in each of the trial arms. Since the subgroup effects are not point identified from the distribution of the observed data, we show how to compute bounds on these effects by using scientifically plausible assumptions. We characterize the uncertainty of our procedure by using the Bayesian paradigm. The methodology is developed in the context of the second Multicenter Automatic Defibrillator Intervention Trial, which was a randomized trial designed to evaluate the effectiveness of implantable defibrillators on survival.",
keywords = "Bounds, Expert opinion, Identifiability",
author = "Scharfstein, {Daniel O} and Georgiana Onicescu and Steven Goodman and Rachel Whitaker",
year = "2011",
month = "8",
doi = "10.1111/j.1467-9876.2010.00759.x",
language = "English (US)",
volume = "60",
pages = "607--617",
journal = "Journal of the Royal Statistical Society. Series C: Applied Statistics",
issn = "0035-9254",
publisher = "Wiley-Blackwell",
number = "4",

}

TY - JOUR

T1 - Analysis of subgroup effects in randomized trials when subgroup membership is missing

T2 - Application to the second Multicenter Automatic Defibrillator Intervention Trial

AU - Scharfstein, Daniel O

AU - Onicescu, Georgiana

AU - Goodman, Steven

AU - Whitaker, Rachel

PY - 2011/8

Y1 - 2011/8

N2 - We develop and implement a methodology for drawing inference about subgroup effects in a two-arm randomized trial when subgroup status is known only for a non-random sample in each of the trial arms. Since the subgroup effects are not point identified from the distribution of the observed data, we show how to compute bounds on these effects by using scientifically plausible assumptions. We characterize the uncertainty of our procedure by using the Bayesian paradigm. The methodology is developed in the context of the second Multicenter Automatic Defibrillator Intervention Trial, which was a randomized trial designed to evaluate the effectiveness of implantable defibrillators on survival.

AB - We develop and implement a methodology for drawing inference about subgroup effects in a two-arm randomized trial when subgroup status is known only for a non-random sample in each of the trial arms. Since the subgroup effects are not point identified from the distribution of the observed data, we show how to compute bounds on these effects by using scientifically plausible assumptions. We characterize the uncertainty of our procedure by using the Bayesian paradigm. The methodology is developed in the context of the second Multicenter Automatic Defibrillator Intervention Trial, which was a randomized trial designed to evaluate the effectiveness of implantable defibrillators on survival.

KW - Bounds

KW - Expert opinion

KW - Identifiability

UR - http://www.scopus.com/inward/record.url?scp=79960846650&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=79960846650&partnerID=8YFLogxK

U2 - 10.1111/j.1467-9876.2010.00759.x

DO - 10.1111/j.1467-9876.2010.00759.x

M3 - Article

AN - SCOPUS:79960846650

VL - 60

SP - 607

EP - 617

JO - Journal of the Royal Statistical Society. Series C: Applied Statistics

JF - Journal of the Royal Statistical Society. Series C: Applied Statistics

SN - 0035-9254

IS - 4

ER -