Developing statistical methods to test mental health treatments

Randomised trials are the ‘gold standard’ for testing if a treatment works. Since 2008, over three million volunteers have taken part in UK trials, and National Institute for Health Research (NIHR) spends over £100million on them annually. However, half the trials fail to show that a new treatment is better than an existing or ‘dummy’ treatment.

The key questions that patients and researchers want to answer in randomised trials are: 

  1. Are treatments effective?
  2. How do they work?
  3. What factors make them work more effectively?
  4. For whom does it work?

Statistical methods in clinical trials

Trialists know how to design trials to answer whether the treatment is effective but have limited knowledge of how to answer the others. Our National Institute for Health Research (NIHR) Maudsley Biomedical Research Centre theme has led the development of statistical methods to answer all these questions in collaborative work funded by NIHR/MRC and summarised in an open access NIHR monograph and tutorial papers.

These new methods are applied in NIHR-funded mental health trials including: examining the role of belief flexibility in paranoia (SlowMo); parent-mediated interventions for children with autism (PACT, PACT-G); voice-hearing in psychosis (AVATAR); illness perception in irritable bowel syndrome (ACTIB) and chronic fatigue syndrome (PACE).

External research groups also apply our methods in their studies (Oxford – paranoia, FeelingSafe trial; Manchester – suicide prevention, CARMS trial), and in clinical disciplines including surgery, stroke, inflammatory diseases, cancer and diabetes. We have international collaborations (Harvard, Arizona State) and contributed to guidelines for reporting the results of studies; first to understand how treatments work (the AGReMA statement) and second for studies investigating what factors help treatments work in health services (PCORI complex interventions).

Ensuring real-world effects

These methods and analyses are fundamental to what is called translational research – research that moves successful laboratory interventions to real-world settings.

To ensure that research proposals will have real world effects, we offer workshops for NIHR panel members, who evaluate proposals, and academics, patient and public involvement groups and clinical trial statisticians. Publicly available software code illustrates our methods, so they can be implemented by the community. Using novel designs to test treatments should increase the number that will benefit patients, and also help us understand how they work so we can boost all outcomes.



Developing Resources for Research | Data and Analytics to Drive Healthcare