Simultaneous synthesis of treatment effects and mapping to a common scale: an alternative to standardisation. (23rd January 2015)

• Record Type: Journal Article
• Authors: Ades, A. E.; Lu, Guobing; Dias, Sofia; Mayo‐Wilson, Evan; Kounali, Daphne
• Is Part Of: Research synthesis methods. Volume 6:Number 1(2015:Jan./Mar.)
• Page Start: 96
• Page End: 107
• Publication Date: 2015-01-23
• Subjects: Research -- Methodology -- Periodicals; Research -- Statistical methods -- Periodicals; 507.2
• DOI: 10.1002/jrsm.1130
• Languages: English
• ISSNs: 1759-2879

Abstract

<abstract abstract-type="main"> <title> <x xml:space="preserve">Abstract</x> </title> <sec id="jrsm1130-sec-0001" sec-type="section"> <title>Objective</title> <p>Trials often may report several similar outcomes measured on different test instruments. We explored a method for synthesising treatment effect information both within and between trials and for reporting treatment effects on a common scale as an alternative to standardisation</p> </sec> <sec id="jrsm1130-sec-0002" sec-type="section"> <title>Study design</title> <p>We applied a procedure that simultaneously estimates a pooled treatment effect and the "mapping" ratios between the treatment effects on test instruments in a connected network. Standardised and non‐standardised treatment effects were compared. The methods were illustrated in a dataset of 22 trials of selective serotonin reuptake inhibitors against placebo for social anxiety disorder, each reporting treatment effects on between one and six of a total nine test instruments.</p> </sec> <sec id="jrsm1130-sec-0003" sec-type="section"> <title>Results</title> <p>Ratios of treatment effects on different test instruments varied from trial to trial, with a coefficient of variation of 18% (95% credible interval 11–29%). Standardised effect models fitted the data less well, and standardised treatment effects were estimated with less relative precision than non‐standardised effects and with greater relative heterogeneity.</p> </sec> <sec id="jrsm1130-sec-0004" sec-type="section"> <title>Conclusion</title> <p>Simultaneous synthesis of treatment effects and mapping to a common scale make fewer assumptions than standardising by dividing effects by the sample standard deviation, allow results to be reported on a common scale, and deliver estimates with superior relative precision. © 2015 The Authors. <italic>Research Synthesis Methods</italic> published by John Wiley & Sons, Ltd.</p> </sec> </abstract>