Using desirability indices for decision making in drug development
D. Renard, K. Wu, R. Wada, G. Flesch
Novartis Pharma AG
Objectives: The clinical utility index (CUI) has been proposed as an integrated measure of clinical benefit/risk [1] [2]. Its usage has focused on characterizing optimal dose ranges and selecting compounds when decisions are based on multiple attributes (e.g., safety and efficacy outcomes, quality of life benefits, drugability properties, etc.). By definition the CUI is a weighted sum and requires defining utility functions to represent expected clinical value of possible outcomes. We propose using the desirability index (DI) [3] as an alternative measure to the CUI.
Methods: Desirability concepts have been developed for the optimization of complex industrial processes which involve competing properties. The weighted geometric mean is a popular choice as a summary measure in this context and one of its features naturally translates to drug development applications, namely that if one of the process’ attributes is unacceptable, the process as a whole becomes unacceptable. In our approach DI is derived while accounting for two sources of uncertainty: variability in estimated dose-response relationships and random variation in desirability functions, which are inherently subjective, in order to achieve a more robust assessment. These naturally lend themselves to a Bayesian decision-analytic framework, where the desirability index acts as a gain function. The distribution of DI can easily be obtained through simulation-based methods, either in a Bayesian or frequentist setting.
Results: We highlight some limitations of CUI and show how it can be framed within a broader desirability context. We illustrate the dose optimization and compound comparison problems using a (blinded) case study where one wants to optimize on a single efficacy and single safety outcome. We also illustrate how weights can be employed to generate a desirability surface that better reflects the risk-benefit assessment through equi-desirable contours.
Conclusions: Desirability indices provide a general and flexible framework to extend CUI as an integrated measure of clinical benefit/risk and support dose and compound decisions in drug development.
References:
[1] Korsan B., Dykstra K. and Pullman W. (2005). Transparent trade-offs – A clinical utility index (CUI) openly evaluates a product’s attributes and chance of success. Pharmaceutical Executive (March 2005).
[2] Roy A. and Pfister M. (2006). Optimizing Dose Selection with Respect to Multiple Safety/Efficacy Endpoints Using Clinical Utility Concepts. PAGE2006.
[3] Harrington E.C. (1965). The desirability function. Industrial Quality Control, 21, 494-98.
