Pragmatic Evidence-Based Medicine: closing the implementation gap by promoting 'user-friendly' evidence

Lead Research Organisation: University of Oxford
Department Name: Population Health


When a patient visits his doctor with a sore back, he wants to know whether the experimental intervention is better than the best existing intervention. This information is best obtained in an ‘active controlled trial‘. Yet, current wisdom has it that the best way to measure effectiveness is in a placebo controlled trial. My thesis examined the arguments that placebo controls are methodologically superior to active controls, and found them to be wanting - the apparent tension between good methods and usefulness is largely illusory. I intend to expand my thesis results in two ways. First, I will conduct social research into what patients/clients, practitioners, and policy makers regard as ‘user-friendly‘ evidence, i.e. evidence that would be easy to put into practice. Second, I intend to investigate the methods for comparing the experimental intervention with all available interventions for a given ailment (rather than just one or two). The impact of my project, both in medical and social science research, could be to have an immediate effect on the delivery and uptake of the most effective interventions.

Technical Summary

Objective I. Disseminate of my thesis findings to methodological and medical , and social intervention communities.
My doctoral and proposed extension of my doctoral work are both of particular interest to both the MRC and the ESRC. In particular my work will be able to inform any revisions of the ESRC‘s Research Methods Programme, and will contribute to the MRC‘s Methodology Research Programme.
An ‘active‘ controlled trial is one that compares the experimental intervention with existing treatment. A placebo controlled trial is one that compares the experimental intervention with a placebo, or ‘sham‘ intervention (i.e. sugar pill). Placebo controlled trials are regarded as methodologically superior to active controlled trials (Temple 2000; ICH E10 2000), but active-controlled trials are more ‘user-friendly‘, since they help the used decide between two or more available alternatives. My investigation concluded that the arguments for the methodological superiority of placebo over active controls are on the whole wanting. The importance of my work for the methodology of social science and medicine, where placebo controls are often difficult to design, is clear.

Objective II: (Thesis Extension) How do the relevant groups in society define user-friendly evidence
The arguments for the superiority of placebo controlled trials are made mostly by statisticians and researchers. It would be interesting to know whether potential users (i.e. patients), practitioners, and policy makers regard the information gained from active controlled trials to be more practically useful. In this section of my research I will conduct a small pilot study consisting of semi-structured interviews of patients, practitioners, policy makers, as well as statisticians and clinical researchers. I will use the results of this study to guide a methodological analysis of indirect comparisons of competing interventions.

Objective III: (Thesis Extension): Maximally User-Friendly Evidence
Even more useful than simple active controlled trials, studies that compare all available interventions for a given ailment are more useful than studies that compare only two or three interventions. Such a study would providee way to accomplish this is to make ‘indirect comparisons‘ of the effectiveness of each intervention trials that have already been accomplished (Glenny, Altman, et al: 2005). Beyond a few notable examples (Caldwell et al: 2005) the potentially useful indirect comparisons have few advocates and are rarely done. Further, indirect comparisons are regarded as methodologically problematic, not least of all because the populations could be different across trials. I propose to analyze methodological and empirical problems with indirect comparisons of all available interventions.


10 25 50