Lecturer: Dr. Rajat Mukherjee, Founder and Chief Statistical Scientist at MuSigmas
Statistics and statisticians play a vital role not just in the analysis and interpretation of data generated in clinical trials but also in the design and conduct of clinical trials. For a fixed design, i.e. a design without interim looks, the statistician at a minimum calculates the sample size based on prior information or clinical assumptions, decides in discussion in the clinical team the primary endpoint(s) that will provide evidence towards the safety and efficacy of a drug, biologic, vaccine or a medical device and the statistical analysis to be conducted at the end of the trial. Since around 25 years or so, adaptive designs have become more and more popular not just in early phase discovery trials but also in late phase confirmatory trials. These adaptive designs gives clinical trialists the opportunity to rectify the design following interim looks in case some of the prior assumptions during trial design seem unrealistic in light of the interim data. Rectification of the design in turn minimizes the probability of failure of the trial to show the desired evidence even when the product works. In a nutshell, adaptive designs are designs where following one or more interim looks, pre-specified, data-driven changes are possible for the future course of the trial.
Some common adaptive designs can be constructed and optimized using asymptotic results, however, in most cases simulations are required for establishing the optimal design parameters and operating characteristics of the design. In any case, simulation help towards more rigorous planning. One special class of designs that are becoming very popular in oncology, rare diseases, pediatric indications and even in the vaccine space are Bayesian designs using Bayesian statistics. These are exclusively simulation-guided designs.
In this course we will learn the basics of such simulated-guided adaptive clinical trial designs both in the traditional frequentist as well as in the Bayesian framework via case studies (early and late phase clinical trials) and using some programs and software packages in R.