Bayesian analysis is a powerful method for model-based data analysis. In pharmacometrics, we use such models to predict drug effects and recommend targeted and cost-effective clinical trials. One of the main strengths of Bayesian analysis is that it allows statisticians to quantitatively integrate data and prior information such as data from other studies. This proves key when constructing sophisticated models for which data from a single trial may be sparse or incomplete.
In this Journal Club, we discuss advantages of using Bayesian analysis, as well as malpractices thereof, historical criticisms, and modern approaches to address these criticisms. For the latter part, we closely follow the argument by Gelman & Shalizi 1.
Bayesian analysis has been central to some of our work at Metrum. Examples include the construction of a drug-disease-trial model describing Alzheimer’s progression in patients 2, and the development of a method to compare the efficacy of different drugs 3. The Metrum Institute has also posted free online courses on the application to Bayesian analysis to pharmacometrics.
To view this journal club session, please visit our YouTube channel here.
1 Andrew Gelman, Cosma Rohilla Shalizi. Philosophy and the practice of Bayesian statistics. British Journal of Mathematical and Statistical Psychology, 2013.
2 James A Rogers, Daniel Polhamus, William R Gillespie, Kaori Ito, Klaus Romero, Ruolun Qiu, Diane Stephenson, Marc R Gastonguay, and Brian Corrigan. Combining patient-level and summary-level data for alzheimer’s disease modeling and simulation: a beta regression meta-analysis. J Pharmacokinet Pharmacodyn, Jul 2012.
3 Jorge Luiz Gross, James Rogers, Daniel Polhamus, William Gillespie, Christian Friedrich, Yan Gong, Brigitta Ursula Monz, Sanjay Patel, Alexander Staab, and Silke Retlich. A novel model-based meta-analysis to indirectly estimate the comparative efficacy of two medications: an example using DPP-4 inhibitors, sitagliptin and linagliptin, in treatment of type 2 diabetes mellitus. BMJ Open, 3:e001844 (http://bmjopen.bmj.com/content/3/3/e001844), 2013.