<img alt="" src="https://secure.lote1otto.com/219869.png" style="display:none;">
Skip to content

Frequentist? Time for an update!

Pantelis Vlachos, PhD, is a Director at Cytel Consulting. He works with a team of experts who regularly assist clinical study teams with the design and implementation of Bayesian methods. In this blog post, Pantelis describes key benefits of Bayesian analyses for clinical trials.


Pantelis Vlachos, PhD

Bayesian methods are being used more frequently for the design and analysis of clinical trials because their paradigm is well suited for updating the information that accrues during the course of a trial. This has the potential to allow  for smaller trials and for patients to receive better treatment. Bayesian analyses allow for the use of available patient-outcome information, historical data, and the synthesis of results of relevant trials. In medical device trials, for example, data are often scarce and expensive and the use of prior information is particularly valuable. Bayesian methods lend themselves naturally in this setting as medical devices are usually modifications of previous ones. 

Advantages of the Bayesian Approach include:

The direct nature of Bayesian probabilities: A direct probability statement allows a statistician to draw conclusions that are conditional upon the evidence in front of her. By contrast, frequentist probabilities assume that if we were to repeat the experiment an infinite number of times, a given probability would arise. Independently of this assumption, frequentist probabilities tell us nothing about what conclusions we may draw from the data that actually exist.

Priors are available when data are scarce: Prior distributions represent a summary of our preexisting understanding and beliefs regarding unknown model parameters. As a result, they are crucial when data are scarce (e.g. in medical device trials.)

Continual update of inferences given calculations of predictive probabilities: Predictive probabilities tell us what the probability will be that the next cohort of patients will respond to treatment A, given the information that we already have. As data accumulate, Bayesianism allows us to recalculate predictive probabilities, instead of relying on the probabilities with which we began the trial.

Flexibility in sequential monitoring of a trial: Bayesian methods are better able to adapt to circumstances when a trial needs to be stopped early, or when randomization is discovered to have been improper or inadequate.   

Bayesian  methods are relatively simple to implement with support from experienced statisticians. Cytel Consulting regularly helps clinical study teams adopt Bayesian methodology. Given the many benefits of Bayesian design, consider approaching Cytel Consulting as you prepare for your next study. 

Click here to learn about Cytel Consulting

Cytel is excited to bring Bayesian methods to the forefront of their software and consulting services. Currently Bayesian power calculations for normal and binomial endpoints (assurance) can be found in East BASE; Dose Escalation methods such as the Continual Reassessment Method (CRM), modified Toxicity Probability Interval (mTPI) and Bayesian Logistic Regression Model (BLRM) can be found in East ESCALATE; and multiple Bayesian dose finding designs can be found in Compass.

Related Items of Interest


Berry, D. A. (1993). A case for Bayesianism in clinical trials. Statistics in Medicine, 12, 1377–93. (Requires access to Wiley Online.) 

Food and Drug Administration (2010) Guidance for the Use of Bayesian Statistics in Medical Device Clinical Trials (Available to All) 

Kadane, J. B. (1995). Prime time for Bayes. Controlled Clinical Trials, 16, 313–18. (Requires access to Elsevier) 

Spiegelhalter, D. J., Abrams K. R. and  Myles J. P. (2004). Bayesian Approaches to Clinical Trials and Health-Care Evaluation, Wiley, Chirchester, England (Google Book)

5 Reasons to Invest in Bayesian Dose-Escalation (Cytel Blog) 



contact iconSubscribe back to top