Dr Robert Mahar is a postdoctoral biostatistician who splits his time between the University of Melbourne, School of Population and Global Health and the Victorian Comprehensive Cancer Centre.
Along with medical oncologists based at St Vincent’s Hospital, Associate Professor Sue-Anne McLachlan and Dr Melissa Moore, Dr Mahar is supervising Dr Wei Hong, a medical oncologist who is undertaking a PhD to explore how to better use and understand Bayesian clinical trial designs in an oncological context. He spoke with us about the opportunity for clinical trials.
Bayesian methods are just like the classical statistical methods that most people are familiar with, in that they use data and probability theory to estimate things like means and variances. Bayesian methods are arguably more powerful because they can combine and analyse evidence from different sources (such as previous results, newly collected data, or even expert opinions), and the results can be continually updated as more evidence becomes available.
Although the underlying Bayesian theoretical concepts were first developed almost 250 years ago, they were not used much until the last two decades. Because... for most real-world problems, everyone had to wait until computers were powerful enough to apply them.
Bayesian trial designs, which are almost always adaptive, are exciting – they can allow trials to be modified in-progress as new information becomes available.
For example, a Bayesian adaptive trial may be modified in the following ways:
Stop early if a treatment is likely to be either efficacious or futile without running the trial to completion
Start with multiple treatment arms and drop futile treatments along the way
Adapt its randomisation schedule so that a larger proportion of patients are given treatments with higher probabilities of efficacy
Seamlessly transform from an early-phase to a confirmatory trial
By ‘learning as they go’, Bayesian adaptive trials are often more efficient in terms of participant numbers, duration, and cost than traditional trials.
Many adaptive trial design features can be implemented using classical statistical methods, but the mechanics of the Bayesian approach allow information to be combined much more naturally and with a more intuitive focus on probabilities. Compared to the classical approach, which focuses on more slippery concepts such as statistical power and type I error, the Bayesian approach better resembles how we think in the real-world.
Encouragingly, just last year the US Food and Drug Administration (FDA) finalised a guidance document on adaptive clinical trials that is boldly supportive of Bayesian adaptive trials in general, and it is really exciting to see this approach gaining more mainstream acceptance.
Because Bayesian adaptive trials can be designed to quickly recognise and abandon futile treatments, they are often of more immediate benefit to trial participants compared to traditionally designed trials.
Indeed, by increasing the chance that trial participants receive the most promising treatments, many Bayesian adaptive trials can be considered to be practice-changing by design. Also, Bayesian adaptive trials are often smaller, faster, and less expensive than traditional trials, so these designs are increasingly more attractive to researchers, funders, and society in general.
Continued improvement in outcomes for people dealing with the consequences of cancer requires innovative thinking from the lab to the bedside and everywhere in between. For the time being, most new cancer treatment discoveries rely on randomised controlled trials to prove their worth and Bayesian approaches are at the forefront of innovative thinking in this area.
Whether it is here in Victoria, or anywhere else in the world, if researchers are interested in designing more efficient and responsive trials, then they should be considering Bayesian approaches alongside the more familiar traditional trial designs.
By using the best tools for the job, scarce healthcare resources are less likely to be wasted on unnecessarily large and long-lasting trials, and patients are more likely to have timely access to promising anti-cancer drugs.
Read more about about adaptive trial designs. Dr Mahar and colleagues feature in the latest issue of Pursuit.
