Obtaining valid and reliable estimates of treatment effectiveness from evidence synthesis of psychiatric trials.

Establishing the most effective treatment(s) for depression is a priority for population health. By 2020 depression will be the second highest cause of disease worldwide with the cost to the NHS currently around #9 billion. Health Technology Assessment (HTA) supports healthcare decision-making through the evaluation of the clinical and cost-effectiveness of medical treatments and the publication of treatment recommendations and guidelines for the NHS. Gold standard evidence for HTA comes from meta-analysis of randomised controlled trials (RCTs). A meta-analysis combines the data from multiple RCTs for the same condition in one analysis to produce more powerful evidence of whether a treatment is effective or not. However, empirical research suggests that RCTs in psychiatry are particularly susceptible to bias, which can lead to under- or over-estimation of treatment effectiveness. New methods for controlling for bias in meta-analysis have been proposed which generate un-biased estimates of treatment effectiveness. This project evaluates the potential benefits and reliability of these new methods compared to standard meta-analysis. The project uses these methods to evaluate the effectiveness of new-generation antidepressants, psychotherapies and older treatments, such as tricyclic antidepressants, to answer the question which is the most effective treatment for depression?

Background: Establishing the most effective treatment(s) for depression is a population health priority. By 2020 depression will be the second highest cause of disease worldwide, with the cost to the NHS around #9 billion. Evidence synthesis using randomised controlled trials (RCTs) underpins Health Technology Assessment (HTA) and clinical guidelines. However, empirical evidence suggests psychiatric RCTs are susceptible to bias, raising questions about reliability and validity of effect estimates from psychiatric HTA. Two recent methodological developments make it possible to control for bias in psychiatric HTA; Bias-adjusted meta-analysis and Mixed Treatment Comparisons.

Aims & Objectives: to examine the suitability of MTC for psychiatric RCTs and assess how biases in evidence synthesis of psychiatric RCTs can be controlled and mitigated to generate more precise, valid estimates of treatment effect. I will;
1. explore the benefits of bias-adjusted MTC compared to standard meta-analysis in the presence of quality-related biases and effect modifiers,
2. assess the effect of extending an MTC network of evidence for depression to include all relevant treatments and assess the ‘risk of bias‘ for each included RCT;
3. explore and compare the effects of bias adjustment on treatment effects estimates in the extended depression network of evidence, using approaches proposed by Turner, Welton and Dias.

Design & Methodology: A review of NICE guidelines and Cochrane reviews will be conducted to identify and assemble psychiatric MTC networks of evidence. The networks inform a simulation study to examine the reliability of bias-adjustment, comparing effect estimates from meta-analysis, unadjusted MTC and bias-adjusted MTC in the presence of bias. The substantive part of the project extends an existing network of new-generation antidepressants for depression, including additional treatments and assessing each RCT for bias using the Cochrane risk of bias tool. The extended network forms the basis for an empirical study to explore and compare the effects of bias-adjusted MTC on depression effect estimates, using three recently proposed bias-adjustment methods. Analyses use a Bayesian approach implemented in WinBUGS.

Scientific Opportunities: Project outputs will
generate validated, reliable treatment effect estimates and rankings for depression;
inform a future grant proposal of bias-adjusted MTC for all depression treatments to identify the most effective treatment overall;
evaluate the benefits of bias-adjusted MTC compared to meta-analysis in psychiatric HTA
inform NICE‘s research agenda regarding the validity and reliability of MTC;
identify whether the three MTC bias-adjustment methods can be combined into a single approach.