Randomised trials of complex interventions are typically designed, conducted, and analysed as if they are drug trials. Although there are many parallels there are also a number of important distinctions, which are seldom considered when designing individual trials. One of these concerns the involvement of therapists in delivering psychotherapy. Systematic reviews and meta-analyses provide an opportunity for exploring the full range and complexity of issues encountered in realistically complex situations. The first objective of the thesis was therefore to develop a conceptual framework for understanding the role of the therapist in trial designs. It was addressed by a review of the psychotherapy and statistical literatures structured according to the broad concepts of precision, internal and external validity and refined on the basis of a systematic methodological review of Cochrane reviews meta-analysing trials involving psychotherapy. The second objective was then to review, adapt, illustrate and compare methods for meta-analysing psychotherapy trials with nested designs. Methods for meta-analysing ICC estimates, absolute and standardised mean differences were adapted to allow for heteroscedasticity between treatments at the therapist- and patient- levels. These were illustrated using the example of counselling in primary care, with comparisons being made between aggregate and one-step approaches to the meta-analysis of individual-patient-data.It was argued that the therapist has two roles in randomised trials. Firstly, they are one component of a multi-component intervention, and are thus a potential treatment variable. Second, the nesting of patients within therapists creates an additional level in the design, so the therapist is also an experimental unit. The inability to conceal or randomise allocations leads to observational components within the trial design and to heteroscedasticity which deserves more attention. Characterising complex interventions, like psychotherapy, with more than one treatment variable could facilitate greater understanding of their components, how they interact, which are important, to what extent, and for whom. It also brings what is currently referred to as process research into the remit of trials, enabling a more complete evaluation of the causal effects. The broad concept of multiple experimental units makes cluster-randomised, longitudinal, multi-centre, crossover, therapist- and group-based intervention trials special cases of a more general class of multilevel trial. All involve clustering effects; their nature and the appropriate statistical model varying according to the design. Methods were proposed for the meta-analysis of continuous outcome data for two-level nested designs. A general approach was adopted, where possible, to incorporate methods covering cluster-randomised trials and the Behrens-Fisher problem. It was clear that this is a relatively untouched methodological area in need of further exploration. For the same reasons as it became necessary to summarise clinical research, it is recommended that systematic methodological reviews be carried out on a larger scale in future.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:538418 |
| Date | January 2011 |
| Creators | Walwyn, Rebecca Elizabeth Ann |
| Contributors | Dunn, Graham ; Roberts, Christopher |
| Publisher | University of Manchester |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | https://www.research.manchester.ac.uk/portal/en/theses/therapist-variation-within-metaanalyses-of-psychotherapy-trials(be4f5031-fa4a-49d2-8772-5ea4a8fd7aef).html |
Page generated in 0.002 seconds