Annual report 2006

In its first year, the Group has achieved the following:

(a) Initial development and evolution of research and teaching strategies.
To date, the Research Group has held three one-day meetings at the University of Manchester’s offices in Covent Garden. These have been used to explore members’ common research interests and likely methodological needs of the MHRN, and to plan the Group’s future research and teaching programmes.

(b) Planning of training programme and delivery of first short course(s).
We envisage a developing programme of courses expanding to 5 per year, the courses themselves being run on a self-financing basis. The intended audience is postgraduate research workers (from PhD students to professors) likely to be involved in MHRN sponsored and similar research, including applied statisticians, trialists and clinicians.

So far, the following two courses are available:

Structural equation and latent variable modelling with Mplus (Graham Dunn, Tim Croudace and Andrew Pickles). This three-day course was initially run in Manchester in January 2006 and will be repeated in Manchester in January 2007. A further course (to be held at the Institute of Psychiatry) is planned for the Spring of 2007.

Causal inference using Stata (Graham Dunn, Andrew Pickles and Ian White). This three-day course was first run in the University of Copenhagen in October 2006. We envisage running it again (in the UK) in 2007. A two-day version has twice been run by GD and AP at the University of Manchester.

Tim Croudace has developed a 10 week (10 2 hour sessions) lecture series Reading, understanding and writing about multivariate statistics that will be developed over the next year into an external programme. See http://www.bio.cam.ac.uk/gradschool/current/courses/multivar.html

(c) Initial development and maintenance of the Group’s web-site.
We intend to develop a national/international internet-based resource , specifically aimed at mental health cohorts and trials, which will provide methodological support for trialists and epidemiologists in the development of high quality and maximally-informative research studies. A research assistant with knowledge of statistics and associated quantitative methods, and with some expertise in programming web pages, will be recruited early in 2007.

(d) Methodological research and associated funding.
The following members of the Group - Graham Dunn (PI), Linda Davis, Andrew Pickles, Chris Roberts, Ian White and Frank Windmeijer – together with Jonathan Green (Child and Adolescent Psychiatrist, University of Manchester) were successful co-applicants for funding (about £365K) following MRC’s recent Methodology and Implementation call. Named collaborators on this project are Richard Bentall (clinical psychologist, Manchester), Christine Barrowclough (clinical psychologist, Manchester), Steve Birch (health economist, Manchester and McMaster) and Philippa Garety (clinical psychologist, Institute of Psychiatry). The three-year project (MRC reference G), which started on 1st October 2006, is entitled “Designs and methods of explanatory (causal) analysis for randomised trials of complex interventions in mental health”. Richard Emsley is the RA on the project.

This is a project for the development, evaluation and dissemination of statistical and econometric methods for the design and explanatory analysis of randomized controlled trials of psychological treatments. We are concerned with making valid causal inferences about the mechanisms of treatment-induced change in outcome. This is important to clinicians and scientists alike. The analyses will focus on the estimation of the joint effects of several postulated intermediate or process variables (mediators). Potential mediators include: programme participation (Was the offer of therapy taken up? How many of the planned number of sessions of therapy were actually attended?); the quality of the therapy – including fidelity to the underlying treatment model being evaluated and the strength of the working or therapeutic alliance (either assessed by the therapist or client, or independent observation), and other (random) therapist effects; and mediator variables arising from the treatment model (in the context of cognitive behaviour therapy for psychotic delusions, for example, these might include strength of the delusion, beliefs about the nature of the delusion and strategies for coping with the problem).

The three key research questions to be addressed in this project are:
(1) What is the best way of estimating the joint influence of participation (number of sessions attended) and therapeutic alliance (or fidelity to a specific treatment model) on the effects of treatment?
(2) How important are therapist effects and how much of the variability between therapists is attributable to fidelity to the given treatment model and to the strength of therapeutic alliance?
(3) How can we develop the models to analyse more complex longitudinal data and explore the potential mediating effects of intervening variables implied by particular theoretical treatment models?
Following this, we will investigate potential trial designs with the dual-purpose of facilitating both the evaluation of the effectiveness of a particular therapy (Does it work?) and the evaluation of the relevant causal mechanisms (How does it work?).

Chris Roberts, in collaboration with Professor Simon Thompson and others in the MRC Population Health Sciences Research Network (PHSRN) has obtained PHSRN funding to organise a one-day methodological workshop (for 60 participants) on clustering effects in individually randomised trials and other studies (this is primarily concerned with therapist and interviewer effects). The workshop is planned for November 2007. Chris Roberts and Graham Dunn also supervise Rebecca Walwyn (medical statistician at the Institute of Psychiatry) who has recently started an MRC Training Fellowship (held at the Institute) to carry out research on therapist effects in psychotherapy trials, leading to an external PhD of the University of Manchester.