The EMBaRC (Enabling Methods for Bayesian Randomised Clinical trials) lab, led by Dr. Anna Heath, aims to improve the design of randomised clinical trials (RCTs) through innovative Bayesian statistical methodology. With an applied focus, we combine methodological research and expertise with collaboration, guidance, education and meaningful patient engagement.

What is a Randomised Clinical Trial?

Clinical trials study the safety, efficacy and effectiveness of healthcare interventions in humans. There are many different types of clinical trial, with each different type aiming to answer a different research question. RCTs are used to compare the impact of different interventions on an outcome of interest. Randomisation is used to assign which intervention each trial participant receives, which aims to ensure that the only differences between the groups is the intervention that they receive. This means that any observed differences in the outcomes can be attributed to the intervention, making RCTs an important tool to understand the differences between healthcare interventions.

Picture of a randomised clinical trial, demonstrating how trial participants are selected from the population before being randomised to either an intervention or comparator. Outcomes are measured and then compared.

Are there issues with Randomised Clinical Trials?

RCTs are costly to run, take a long time to complete, and require specialist involvement from a range of different perspectives. From a statistical perspective, RCTs require rigorous statistical design to avoid or minimise potential biases, draw generalised conclusions that help clinicians decide the best treatment for their patients, and to ensure that resources are used effectively to maximise impact. Current statstical methods for RCT design often fail when the available sample size is limited. This is particularly a problem in rare diseases, where only a small number of patients are available to enroll in the RCTs, and in paediatrics, where ethical concerns can limit recruitment further. There is also a growing understanding that patients often respond differently to interventions, which is challenging to evaluate in standard RCTs.

What is Bayesian Statistics?

Historically, RCTs have been analysed using statistical methods that draw conclusions evaluating the frequency of events “in the long run” across an infinite number of hypothetical similar trials, known as frequentist methods. Bayesian inference provides an alternative framework for drawing insights from data, compared to this standard “frequentist” approach. The Bayesian approach combines the data from the current study (i.e., the RCT that has just been completed) with information available before the trial, which is specified using a prior distribution. From these two sources of evidence, we obtain a summary that provides information on the possible values of the differences in outcomes between the interventions (see Figure). Recently, Bayesian methods have been gaining popularity in clinical trials because:

  1. They can require smaller sample sizes
  2. They can include external evidence in the prior distribution to be included in the RCT analysis
  3. They can be more flexible, allowing trials to respond quickly, such as during the Covid-19 pandemic.
  4. They closely align with clinical decision making, as the probability of a clinically important benefit can be computed
  5. The computational power required to use Bayesian methods is now available.

A graphic displaying a Bayesian analysis

RCTs using Bayesian Decision theory

The EMBaRC lab has substantial expertise in using Bayesian decision theory in the design of RCTs and clinical research more broadly. These methods, known as Value of Information, have been most commonly explored alongside health economic modelling to calculate the monetary or health benefit associated with research. The goal of using Value of Information methods in RCTs is to maximize the impact of clinical research by prioritizing research with high value. The EMBaRC lab works on both theoretical and applied projects to improve the use of Value of Information.

A depiction of the evidence cycle for RCT design in Value of Information.

Overall, advancing Bayesian statistical methodology aims to develop efficient, feasible and cost-effective RCTs across a range of disease areas. Ultimately, improved design for RCTs ensures that research funding is used appropriately and that effective interventions efficiently translate into practice and improve outcomes.