Managing Toxicology for the Future

Drug toxicity remains a major hurdle to reducing risk and improving productivity in pharmaceutical R&D. About one third of all drugs fail preclinical or clinical testing because of apparent or suspected drug toxicities. In addition, over the past few years, several lucrative FDA-approved drugs have been withdrawn from the market due to serious side-effects, resulting not only in huge financial losses for the companies, but also in diminished public trust in the FDA's and the industry's ability to bring effective and safe drugs to the market. According to industry estimates, companies spend about $2 billion annually on toxicity-related drug failures. The inaccuracy of safety screens in place today can also lead to enormous lost opportunity costs. Termination of a single potential blockbuster drug in development or on the market can lead to lost revenue far exceeding the total direct cost of industry failures.

Despite the staggering financial losses associated with toxicity-related drug failures, companies typically spend only about 5% of their total R&D budget on drug safety evaluations. Moreover, toxicity testing has not changed fundamentally over the past several decades and animal models remain the gold standard, firmly ensconced by international regulations requiring extensive animal testing before first use of a drug in humans. Nevertheless, animal testing has serious limitations, both in terms of predictive power and throughput capacity. Because of interspecies differences in physiology, animal data are not always predictive of the human situation, and because of their low-throughput, animal models cannot be used to assess drug toxicity early in discovery nor to detect rare adverse drug events prior to wide-spread clinical use.

Recognizing the toxicity bottleneck, companies recently began to pay much attention to the area of ADME/Tox testing (adsorption, distribution, metabolism, elimination and toxicity), with the goal of developing alternative in vitro and in silico models to evaluate drug toxicity. The development of alternative models depends on a better understanding of the underlying mechanisms that lead to the clinical manifestation of drug toxicity. New insights can be gained through the generation of experimental data using traditional (e.g. animal) and novel tools (e.g. microarrays, proteomics, metabonomics), and through the analysis of existing public and proprietary data of known toxic compounds. A mechanistic understanding of drug-induced toxicities is the first step towards developing more predictive human-based in vitro and in silico models. The second step is the adaptation of alternative predictive models to a high-throughput format in order to assess drug toxicity early in discovery, rather than in late stage clinical trials when much of a drug's R&D budget has been spent.

Developing predictive models in toxicology requires a fundamentally different and highly interdisciplinary approach that involves heretofore separate disciplines, ranging from life sciences and engineering to computer science. No single technology or approach will be the magic bullet that will provide the answer to the problem of drug safety. Instead, advances in toxicology will come from a gradual paradigm shift based on an integrated approach linking experimentation and computation and using both existing and radically new tools to generate and analyze toxicity data (e.g., established tools in genomics and bioinformatics and next-generation "omics" tools and computational approaches, such as systems analysis, borrowed from other fields).

Novel methods and technologies to assess drug toxicity will continue to be developed both internally (in pharmaceutical companies) and by technology platform companies, with expertise ranging from informatics to device fabrication and assay development. Since integration of different approaches and access to large data sets will be key to effectively address the problem of drug toxicity, the industry increasingly recognizes the need for consortium business models that draw on the expertise of multiple organizations.

In the short-term, the industry will use alternative toxicity models during lead optimization to prioritize drug candidates, and during preclinical/clinical testing to complement and guide the interpretation of animal and human data. In the long-term, computational models will be used extensively to assist with the design of chemical structures that are less likely to be toxic in humans - much in the same way computer simulations are used to design airplanes capable of safe flight from the moment they leave the manufacturing plant.

The development and implementation of alternative toxicity models needs to occur within the current regulatory framework and new approaches need to be carefully validated against human in vivo data and standard animal models. Clearly, it will take a long time before the FDA will agree to replace animal models by human-based in vitro and in silico models, and any small step in this direction will require close collaboration between industry and the FDA. Any missteps (or inaction) by regulatory authorities can also have the effect of hindering innovation in the field.

©Drug and Market Development 2003

To view and purchase D&MD reports click here!