Abstract 1636: A model-based approach toward clinical pipeline optimization

Publisher Website
Google Scholar
Abstract
Introduction What is the optimal “shape” for a drug pipeline: A flute? A megaphone? A trumpet? In other words, what set of stringencies for passage of candidates to the next stage in development maximizes the value of the pipeline as a whole, and how do assays predicting drug quality influence this optimum? Methods A linear programming formulation was developed for four phases of a drug pipeline: Preclinical, Phase I, Phase II, and Phase III, with variables including: the total budget for the entire pipeline (B), and the number (Ni), cost (ci), and net present value (vi) of drugs in each phase (i). The model is subject to two constraints: (1) the pipeline must reduce in size with each subsequent phase (i.e. Npre≥ N1 ≥ N2 ≥ N3), and (2) the sum total of the cost times the number of drugs in each phase must be less than or equal to the total budget (i.e. cpreNpre + c1N1 + c2N2 + c3N3 ≤ B). The overall goal is to maximize the total value of drugs progressing through the pipeline (i.e. maximize vtot = vpreNpre + v1N1 + v2N2 + v3N3). Additionally, pipeline evolution was modeled to evaluate the implications of a “predictive” drug quality assay. Hypothetical “drug quality scores” were drawn from normal distributions which reflect industry probabilities of success between phases. Drugs that are above a cutoff stringency advance forward, with the process repeated for all phases. The cost of each phase and value for successful drugs are recorded for each simulation. The process is repeated many times, with the expected value calculated at each iteration. Results Beginning with a pool of drug candidates, the optimal pipeline structure should approach equal sized pipeline widths as rapidly as possible. Monte Carlo simulations demonstrated that increasing the stringency of a “quality assay” in Phase I results in a linear increase in expected value. Additionally, if the quality test is improved, e.g. by reducing the variability of the assay, the improvement in expected value will increase in an exponential manner. Conclusions The simulations predict that to maximize pipeline value, drug companies should: (a) pick the winners as early as possible (i.e. Preclinical or Phase I), (b) set stringent go/no-go criteria in Preclinical and Phase I development, and (c) perform prospective clinical trial simulations (Preclinical to Phase I, Phase I to Phase II, Phase II to Phase III) to improve the ability to predict winners. Citation Format: Jackson Burton, William Antebi, Christopher J. Zopf, Dean Bottino, Shu-Wen Teng, Ryan Nolan, Arijit Chakravarty. A model-based approach toward clinical pipeline optimization. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 1636. doi:10.1158/1538-7445.AM2015-1636

This publication has 0 references indexed in Scilit: