Model based drug development: A clinical pharmacologist’s approach to quantitative decision-making
*Scott Marshall, Pfizer Ltd. 

Keywords: Model Based Drug Development; Biomarker; Quantitative Decision Criteria; Go/No-Go Decision; Positive Predictive Value; False Positive Outcomes; Maximum Net Benefit.

The term Model Based Drug Development (MBDD) has been coined, in the clinical pharmacology literature, to describe the process of capturing the totality of the data for a given drug and subsequent extrapolation to the optimized design of future trials (1). Within this framework, physiologically and pharmacologically principled models provide the foundation for prediction of exposure response across populations and endpoints. Translational elements are facilitated via biomarker-outcome linkage and an understanding of disease progression. While integration of emerging data with prior understanding extracted from established therapies is achieved by meta-analytic techniques. The central link between evidence synthesis and the assessment of future trial performance is provided by the development and evaluation of Quantitative Decision Criteria (QDC). While pre-specification of QDC to ensure unbiased go/no-go decision making in drug development is well recognized, application within the MBDD framework has recently been more thoroughly examined (2-3). For learning trials, the work highlights the value of unconditional trial operating characteristics such as the probability of a correct decision [P(Correct)] and the amount of risk discharged by a go decision given any particular combination of decision rule, prior expectation and study design (aka Positive Predictive Value, PPV). Equally important and less discussed, is a greater emphasis on establishing quantitatively robust product concepts as the foundation for effective decision rules. Incorporation of an accurate reflection of the challenging competitive landscape is required to ensure efficient early stage screening and prevention of late stage commercial failure. However, a balance needs to be sought between minimizing false positive outcomes in the learning phase and the need to progress compounds in order to assess their maximum net benefit after determining the optimal dose regimen and most appropriate patient population. This presentation will illustrate, by way of example, the use of MBDD approaches in establishing priors used in developing and evaluating robust QDC in the assessment of novel therapies.