Chris Cameron1, Emmanuel Ewara2, Florence R Wilson1, Abhishek Varu1, Peter Dyrda2, Brian Hutton3,4, Michael Ingham5. 1. Cornerstone Research Group Inc., Burlington, ON, Canada (CC, FRW, AV). 2. Janssen Canada Inc., Toronto, ON, Canada (EE, PD). 3. Ottawa Hospital Research Institute, Ottawa, ON, Canada (BH). 4. University of Ottawa School of Epidemiology, Public Health and Preventative Medicine, Ottawa, ON, Canada (BH). 5. Janssen Scientific Affairs, LLC, Horsham, PA, USA (MI).
Abstract
BACKGROUND AND AIMS: Adaptive trial designs present a methodological challenge when performing network meta-analysis (NMA), as data from such adaptive trial designs differ from conventional parallel design randomized controlled trials (RCTs). We aim to illustrate the importance of considering study design when conducting an NMA. METHODS: Three NMAs comparing anti-tumor necrosis factor drugs for ulcerative colitis were compared and the analyses replicated using Bayesian NMA. The NMA comprised 3 RCTs comparing 4 treatments (adalimumab 40 mg, golimumab 50 mg, golimumab 100 mg, infliximab 5 mg/kg) and placebo. We investigated the impact of incorporating differences in the study design among the 3 RCTs and presented 3 alternative methods on how to convert outcome data derived from one form of adaptive design to more conventional parallel RCTs. RESULTS: Combining RCT results without considering variations in study design resulted in effect estimates that were biased against golimumab. In contrast, using the 3 alternative methods to convert outcome data from one form of adaptive design to a format more consistent with conventional parallel RCTs facilitated more transparent consideration of differences in study design. This approach is more likely to yield appropriate estimates of comparative efficacy when conducting an NMA, which includes treatments that use an alternative study design. CONCLUSIONS: RCTs based on adaptive study designs should not be combined with traditional parallel RCT designs in NMA. We have presented potential approaches to convert data from one form of adaptive design to more conventional parallel RCTs to facilitate transparent and less-biased comparisons.
BACKGROUND AND AIMS: Adaptive trial designs present a methodological challenge when performing network meta-analysis (NMA), as data from such adaptive trial designs differ from conventional parallel design randomized controlled trials (RCTs). We aim to illustrate the importance of considering study design when conducting an NMA. METHODS: Three NMAs comparing anti-tumor necrosis factor drugs for ulcerative colitis were compared and the analyses replicated using Bayesian NMA. The NMA comprised 3 RCTs comparing 4 treatments (adalimumab 40 mg, golimumab 50 mg, golimumab 100 mg, infliximab 5 mg/kg) and placebo. We investigated the impact of incorporating differences in the study design among the 3 RCTs and presented 3 alternative methods on how to convert outcome data derived from one form of adaptive design to more conventional parallel RCTs. RESULTS: Combining RCT results without considering variations in study design resulted in effect estimates that were biased against golimumab. In contrast, using the 3 alternative methods to convert outcome data from one form of adaptive design to a format more consistent with conventional parallel RCTs facilitated more transparent consideration of differences in study design. This approach is more likely to yield appropriate estimates of comparative efficacy when conducting an NMA, which includes treatments that use an alternative study design. CONCLUSIONS: RCTs based on adaptive study designs should not be combined with traditional parallel RCT designs in NMA. We have presented potential approaches to convert data from one form of adaptive design to more conventional parallel RCTs to facilitate transparent and less-biased comparisons.
Authors: Munyaradzi Dimairo; Philip Pallmann; James Wason; Susan Todd; Thomas Jaki; Steven A Julious; Adrian P Mander; Christopher J Weir; Franz Koenig; Marc K Walton; Jon P Nicholl; Elizabeth Coates; Katie Biggs; Toshimitsu Hamasaki; Michael A Proschan; John A Scott; Yuki Ando; Daniel Hind; Douglas G Altman Journal: BMJ Date: 2020-06-17
Authors: Bruce E Sands; Adam S Cheifetz; Chudy I Nduaka; Daniel Quirk; Wenjin Wang; Eric Maller; Gary S Friedman; Chinyu Su; Peter D R Higgins Journal: J Crohns Colitis Date: 2019-09-19 Impact factor: 9.071
Authors: Munyaradzi Dimairo; Philip Pallmann; James Wason; Susan Todd; Thomas Jaki; Steven A Julious; Adrian P Mander; Christopher J Weir; Franz Koenig; Marc K Walton; Jon P Nicholl; Elizabeth Coates; Katie Biggs; Toshimitsu Hamasaki; Michael A Proschan; John A Scott; Yuki Ando; Daniel Hind; Douglas G Altman Journal: Trials Date: 2020-06-17 Impact factor: 2.279