Received: 6 July 2016 Revised: 20 September 2017 Accepted: 6 November 2017
DOI: 10.1002/pst.1843
MAIN PAPER
Use of a historical control group in a noninferiority trial
assessing a new antibacterial treatment: A case study and
discussion of practical implementation aspects
David Dejardin
1
Paul Delmar
1
Charles Warne
2
Katie Patel
2
Joost van Rosmalen
3
Emmanuel Lesaffre
4
1
Department of Biostatistics, F.
Hoffmann-La Roche, Basel, Switzerland
2
Department of Biostatistics, Roche
Products Ltd, Welwyn Garden City, UK
3
Department of Biostatistics, Erasmus
University Medical Center, Rotterdam, the
Netherlands
4
Interuniversity Institute for Biostatistics
and Statistical Bioinformatics, KU Leuven,
Leuven, Belgium
Correspondence
David Dejardin, Department of
Biostatistics, F. Hoffmann-La Roche,
Basel, Switzerland.
Email: david.dejardin@roche.com
When recruitment into a clinical trial is limited due to rarity of the disease of
interest, or when recruitment to the control arm is limited due to ethical reasons
(eg, pediatric studies or important unmet medical need), exploiting historical
controls to augment the prospectively collected database can be an attractive
option. Statistical methods for combining historical data with randomized data,
while accounting for the incompatibility between the two, have been recently
proposed and remain an active field of research. The current literature is lacking
a rigorous comparison between methods but also guidelines about their use in
practice. In this paper, we compare the existing methods based on a confirma-
tory phase III study design exercise done for a new antibacterial therapy with a
binary endpoint and a single historical dataset. A procedure to assess the relative
performance of the different methods for borrowing information from histori-
cal control data is proposed, and practical questions related to the selection and
implementation of methods are discussed. Based on our examination, we found
that the methods have a comparable performance, but we recommend the robust
mixture prior for its ease of implementation.
KEYWORDS
commensurate prior, historical control, Bayesian analysis, power prior, robust mixture prior
1 INTRODUCTION
Statistical methods for combining historical controls with randomized clinical trial data constitutes an active field of
research. Methods have been recently introduced to account for incompatibility between the historical control data and
the randomized control data. As discussed in the seminal article by Pocock,
1
even if historical and randomized control
data are strictly matched, a difference in central tendency or “drift” between historical and randomized control can never
be totally ruled out and should be accounted for in the analysis. We will use here the term “compatibility” for “absence of
drift.” In principle, the risk of bias due to the drift between randomized and historical controls could be mitigated by the
use of a “dynamic borrowing” approach, where the degree of borrowing depends on the compatibility of the randomized
and historical control data. This paper takes the perspective of a clinical trial practitioner in comparing these methods.
P. Delmar and D. Dejardin have contributed equally to the manuscript.
Pharmaceutical Statistics. 2017;1–13. wileyonlinelibrary.com/journal/pst Copyright © 2017 John Wiley & Sons, Ltd. 1