Bayesian approach for assessing noninferiority in a three-arm trial with binary endpoint

Santu Ghosh, Ram C. Tiwari, Samiran Ghosh

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

With the recent advancement in many therapeutic areas, quest for better and enhanced treatment options is ever increasing. While the “efficacy” metric plays the most important role in this development, emphasis on other important clinical factors such as less intensive side effects, lower toxicity, ease of delivery, and other less debilitating factors may result in the selection of treatment options, which may not beat current established treatment option in terms efficacy, yet prove to be desirable for subgroups of patients. The resultant clinical trial by means of which one establishes such slightly less efficacious treatment is known as noninferiority (NI) trial. Noninferiority trials often involve an active established comparator arm, along with a placebo and an experimental treatment arm, resulting into a 3-arm trial. Most of the past developments in a 3-arm NI trial consider defining a prespecified fraction of unknown effect size of reference drug, i.e., without directly specifying a fixed NI margin. However, in some recent developments, more direct approach is being considered with prespecified fixed margin, albeit in the frequentist setup. In this article, we consider Bayesian implementation of such trial when primary outcome of interest is binary. Bayesian paradigm is important, as it provides a path to integrate historical trials and current trial information via sequential learning. We use several approximation-based and 2 exact fully Bayesian methods to evaluate the feasibility of the proposed approach. Finally, a clinical trial example is reanalyzed to demonstrate the benefit of the proposed approach.

Original languageEnglish (US)
Pages (from-to)342-357
Number of pages16
JournalPharmaceutical Statistics
Volume17
Issue number4
DOIs
StatePublished - Jul 1 2018

Fingerprint

Non-inferiority
Bayes Theorem
Bayesian Approach
Binary
Clinical Trials
Margin
Therapeutics
Efficacy
Effect Size
Beat
Bayesian Methods
Toxicity
Placebos
Learning
Drugs
Paradigm
Integrate
Subgroup
Metric
Unknown

Keywords

  • Bayesian method
  • Jeffreys prior
  • Markov chain Monte Carlo
  • assay sensitivity
  • noninferiorty margin

ASJC Scopus subject areas

  • Statistics and Probability
  • Pharmacology
  • Pharmacology (medical)

Cite this

Bayesian approach for assessing noninferiority in a three-arm trial with binary endpoint. / Ghosh, Santu; Tiwari, Ram C.; Ghosh, Samiran.

In: Pharmaceutical Statistics, Vol. 17, No. 4, 01.07.2018, p. 342-357.

Research output: Contribution to journalArticle

@article{c3882e891ab4400ebdb5413d827f727c,
title = "Bayesian approach for assessing noninferiority in a three-arm trial with binary endpoint",
abstract = "With the recent advancement in many therapeutic areas, quest for better and enhanced treatment options is ever increasing. While the “efficacy” metric plays the most important role in this development, emphasis on other important clinical factors such as less intensive side effects, lower toxicity, ease of delivery, and other less debilitating factors may result in the selection of treatment options, which may not beat current established treatment option in terms efficacy, yet prove to be desirable for subgroups of patients. The resultant clinical trial by means of which one establishes such slightly less efficacious treatment is known as noninferiority (NI) trial. Noninferiority trials often involve an active established comparator arm, along with a placebo and an experimental treatment arm, resulting into a 3-arm trial. Most of the past developments in a 3-arm NI trial consider defining a prespecified fraction of unknown effect size of reference drug, i.e., without directly specifying a fixed NI margin. However, in some recent developments, more direct approach is being considered with prespecified fixed margin, albeit in the frequentist setup. In this article, we consider Bayesian implementation of such trial when primary outcome of interest is binary. Bayesian paradigm is important, as it provides a path to integrate historical trials and current trial information via sequential learning. We use several approximation-based and 2 exact fully Bayesian methods to evaluate the feasibility of the proposed approach. Finally, a clinical trial example is reanalyzed to demonstrate the benefit of the proposed approach.",
keywords = "Bayesian method, Jeffreys prior, Markov chain Monte Carlo, assay sensitivity, noninferiorty margin",
author = "Santu Ghosh and Tiwari, {Ram C.} and Samiran Ghosh",
year = "2018",
month = "7",
day = "1",
doi = "10.1002/pst.1851",
language = "English (US)",
volume = "17",
pages = "342--357",
journal = "Pharmaceutical Statistics",
issn = "1539-1604",
publisher = "John Wiley and Sons Ltd",
number = "4",

}

TY - JOUR

T1 - Bayesian approach for assessing noninferiority in a three-arm trial with binary endpoint

AU - Ghosh, Santu

AU - Tiwari, Ram C.

AU - Ghosh, Samiran

PY - 2018/7/1

Y1 - 2018/7/1

N2 - With the recent advancement in many therapeutic areas, quest for better and enhanced treatment options is ever increasing. While the “efficacy” metric plays the most important role in this development, emphasis on other important clinical factors such as less intensive side effects, lower toxicity, ease of delivery, and other less debilitating factors may result in the selection of treatment options, which may not beat current established treatment option in terms efficacy, yet prove to be desirable for subgroups of patients. The resultant clinical trial by means of which one establishes such slightly less efficacious treatment is known as noninferiority (NI) trial. Noninferiority trials often involve an active established comparator arm, along with a placebo and an experimental treatment arm, resulting into a 3-arm trial. Most of the past developments in a 3-arm NI trial consider defining a prespecified fraction of unknown effect size of reference drug, i.e., without directly specifying a fixed NI margin. However, in some recent developments, more direct approach is being considered with prespecified fixed margin, albeit in the frequentist setup. In this article, we consider Bayesian implementation of such trial when primary outcome of interest is binary. Bayesian paradigm is important, as it provides a path to integrate historical trials and current trial information via sequential learning. We use several approximation-based and 2 exact fully Bayesian methods to evaluate the feasibility of the proposed approach. Finally, a clinical trial example is reanalyzed to demonstrate the benefit of the proposed approach.

AB - With the recent advancement in many therapeutic areas, quest for better and enhanced treatment options is ever increasing. While the “efficacy” metric plays the most important role in this development, emphasis on other important clinical factors such as less intensive side effects, lower toxicity, ease of delivery, and other less debilitating factors may result in the selection of treatment options, which may not beat current established treatment option in terms efficacy, yet prove to be desirable for subgroups of patients. The resultant clinical trial by means of which one establishes such slightly less efficacious treatment is known as noninferiority (NI) trial. Noninferiority trials often involve an active established comparator arm, along with a placebo and an experimental treatment arm, resulting into a 3-arm trial. Most of the past developments in a 3-arm NI trial consider defining a prespecified fraction of unknown effect size of reference drug, i.e., without directly specifying a fixed NI margin. However, in some recent developments, more direct approach is being considered with prespecified fixed margin, albeit in the frequentist setup. In this article, we consider Bayesian implementation of such trial when primary outcome of interest is binary. Bayesian paradigm is important, as it provides a path to integrate historical trials and current trial information via sequential learning. We use several approximation-based and 2 exact fully Bayesian methods to evaluate the feasibility of the proposed approach. Finally, a clinical trial example is reanalyzed to demonstrate the benefit of the proposed approach.

KW - Bayesian method

KW - Jeffreys prior

KW - Markov chain Monte Carlo

KW - assay sensitivity

KW - noninferiorty margin

UR - http://www.scopus.com/inward/record.url?scp=85049791472&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85049791472&partnerID=8YFLogxK

U2 - 10.1002/pst.1851

DO - 10.1002/pst.1851

M3 - Article

C2 - 29473291

AN - SCOPUS:85049791472

VL - 17

SP - 342

EP - 357

JO - Pharmaceutical Statistics

JF - Pharmaceutical Statistics

SN - 1539-1604

IS - 4

ER -