Research blog: Improving our answers to key questions in clinical trials

Professor Sabine Landau is Professor of Biostatistics at the Institute of Psychiatry, Psychology & Neuroscience at King’s College London.  She is a member of the UK Mental Health Research Network Methodology Research Group, and with Professor Andrew Pickles, is co-author of a major National Institute of Health Research (NIHR) Health Technology Assessment (HTA) methodological report, Evaluation and validation of social and psychological markers in randomised trials of complex interventions in mental health: a methodological research programme.  Here, Professor Landau - who is part of the NIHR Maudsley Biomedical Research Centre (BRC) - explains the report’s findings, and reports from a workshop presenting the work to clinicians and researchers.

Researchers have a long tradition of designing randomised clinical trials to answer the question “Does this treatment work?” or “Is there a beneficial effect of the treatment compared with some other treatment or treatment as usual?”.  However, they often have little knowledge or experience to answer questions like: “How does it work?”, “What are the underlying mechanisms or targets of the treatment?” or “What factors involved in the therapy make it work better?”. 

Our work looks at the methods which researchers use to answer these questions, and seeks to improve them through better statistical techniques and a new approach to methodology.

Many treatments in modern clinical practice, especially psychological therapies, are what we call “complex interventions” – they have multiple components, or they are delivered by multiple practitioners, and can have beneficial effects on patient outcomes when delivered in combination. 

The development of a complex intervention is usually based on theories about how it works.  For example, one of the ways by which cognitive behavioural therapy (CBT) for psychosis is thought to bring about improvements in symptoms is by reducing reasoning biases, or increasing someone’s flexibility about their beliefs. 

Researchers developing complex intervention often theorise that patients have to have certain characteristics to be able to benefit from a particular treatment. For example, CBT might only be beneficial for those patients who are able to form a strong working relationship with their therapist.   

I have been working with researchers from the University of Manchester, the MRC Biostatistics Unit and King’s College London, to develop a methodological research programme which reviews existing ways of evaluating complex interventions, outlines their limitations, and proposes some new statistical methods that can answer these questions. The major extension of our work is that we are explicit about the underlying assumptions of these methods, and when these might not hold.

Amongst other things, we have proposed a new trial design and associated strategy for analysing trial results, which will allow researchers to use biomarkers – biological characteristics or “signatures” which are used as indicators of a person’s health – to potentially test the notion that the right treatment can be given to the right patient at the right point in time.

In January, we organised a one day workshop in Manchester. The workshop launched our report to around 70 colleagues from across the UK and Ireland, reaching clinicians, triallists, applied statisticians and researchers from other disciplines who are engaged in the development and evaluation of complex interventions, from academia and industry.

Our work provides a framework for posing relevant questions in clinical trials – whether and how an intervention works, what makes it work better, and who it works for – and our workshop introduced attendees to the most appropriate (and readily available) statistical methods for addressing these questions.    

A key finding of our work is that patients who have a strong, engaging relationship with their therapist (“therapeutic alliance”) receive positive benefits from receiving therapy for the symptoms of psychosis, whereas therapy can actually be detrimental to patients with a poor therapeutic alliance.  This tallied with the experience of many of the clinicians attending the workshop, and seeing this factor so clearly displayed in our analysis provoked an interesting discussion.

There was also lots of interest in our new clinical trial design, accompanied by some scepticism about how trials in personalised medicine – which aim to determine which subsets of a population might most benefit from a treatment – can ethically recruit, when not every participant is thought likely to benefit from the treatment.

We will be holding dissemination events and workshops at several major conferences over the coming year, and look forward to sharing our work with more researchers and clinicians and gaining their feedback.   I am currently leading a research project, funded by the MRC Methodology Research Programme, which builds on this work to develop methods that can exploit modern trial designs to understand mechanisms in complex interventions, and we will also be conducting future clinical trials of complex interventions using these new methods. 

To find out more about this research and forthcoming workshops, please contact Professor Sabine Landau at

Tags: Bioinformatics & statistics - Informatics -

By NIHR Maudsley BRC at 24 Feb 2016, 16:39 PM

Back to Blog List