My expertise in this area lies in the use of prior knowledge and in modelling and simulation to help with the design of individual trials and sequences of trials. In addition to techniques for the appropriate use of historic data in building prior distributions, I have developed the Bayesian clinical trial simultation (BCTS) method, along with theory for trial design that recognises both the requirements of regulators and the internal company needs, expectations and values.
Even designing an individual trial is a complex business, particularly at phase 3, where likely outcomes are influenced not just by raw sample size but also by a host of other factors - the use of interim and futility analyses, the recruitment schedule, choice of comparators, adaptive design features and phase 2/3 combinations, etc. Clinical trial simulation attempts to compute power and other features for alternative complex designs. However, conventional clinical trial simulation fixes numerous parameters that, whilst not random, are still uncertain - the drug's efficacy, baseline incidence, compliance and drop-out, adverse event rates, etc. Interpreting the answers from alternative parameter specifications can be very confusing. My pioneering BCTS approach addresses these challenges in a way that delivers much clearer, more relevant information to decision-makers.
Company prior knowledge, in the form of expert judgement and/or historic data, can play a role in achieving efficiencies, but care is required to justify its use and to appropriately discount such data. I can offer expertise in expert knowlege elicitation and in determining effective sample sizes for priors based on analysis or meta-analysis of historic trial data.
I am happy to consult on these or other problems in drug development.
|
|
|
|
|
|
|