About 25% of compounds fail due to toxicity after they reach the clinic, making the early elimination of toxic compounds important for companies looking to cut the cost of drug development. But there is still a relative dearth of scientific understanding of toxicological mechanisms, and the lack of reliable high throughput toxicological screening leaves most companies with no alternative to traditional, low throughput in vivo testing.

Dale Johnson, founder and chief scientific officer of ddPlatform LLC (Emeryville, Calif.), an in silico drug discovery consulting firm, estimates that 75% of R&D costs are related to compound failures, of which a third are related to toxicity problems.

At the same time, everything going on in drug development today - from the growing number of targets to the hits generated by high throughput screening - points to the need to be able to prioritize and kill compounds earlier in the process.

While it is impossible to predict how much toxicology screening will change the efficiency of drug discovery, James Moe, head of global toxicology for Pharmacia Corp. (PHA, Peapack, N.J.), believes a reasonable goal of predictive toxicology is to increase the success rate in early clinical development by two-fold.

But before the financial benefits of toxicity profiling can be recognized, the hard work of gaining a better understanding of biological mechanisms of toxicity still needs to be done.

Defining toxicity

Indeed, the manifestations of toxicity must be characterized before toxic properties can be predicted. And toxicology encompasses many biological systems and processes, creating a need to develop high throughput screening technologies and to identify surrogate markers of toxicities.

For example, one of the most advanced areas of study is cardiac toxicity. Only a few years ago, cardiotoxicity studies were almost all done in low throughput whole animal systems, measuring ECG changes, effects on heart rate and blood pressure. In fact, until recently, there were no good animal models or surrogate markers available for reliable cardiotox screening, according to Tom Colatsky, executive vice president and chief scientific officer at Physiome Sciences Inc. (Princeton, N.J).

However, the identification of associations between marketed drugs and long QT syndrome, which can lead to fatal arrythmias, provided impetus to develop better screening for cardiotoxicity. Associations with long QT syndrome are the reason the nonsedating antihistamine Seldane was pulled off the market in 1997 and the gastrointestinal prokinetic agent Cisapride was pulled last year.

In 1998, Rockefeller University professor Roderick MacKinnon described in Cell the structure of the HERG potassium channel that is responsible for many arrythmias, including long QT arrythmias. Last October, John Mitcheson, lecturer at the University of Leicester in the U.K., and Michael Sanguinetti, professor of medicine at the University of Utah in Salt Lake City, published in the Proceedings of the National Academy of Sciences the structural basis for drug-induced long QT syndrome, providing a target to screen against for cardiotoxicities.

Those studies provided the basis for current screening assays to detect interactions between lead compounds and the HERG potassium channel. The latest cardiotoxicity screens are done in tissue and the next step will be to screen for interactions with ion channels in cell cultures, Moe said.

Also relatively well advanced is an understanding of gene toxicity - the ability of a compound to disrupt gene structure or function with physiological consequences such as carcinogenicity. Michael Taylor, executive director of drug safety and evaluation for Durect Corp. (DRRX, Cupertino, Calif.), noted that gene toxicity is well understood because it has been studied for decades in relation to environmental and chemical toxins.

On the other hand, the knowledge base is less developed for hepatoxicity, which has been responsible for the largest number of drug failures due to toxicity. In vitro screens for interactions with cytochrome P450 enzymes, which metabolize drugs and can create hepatotoxic metabolites, are widely available and specific enzyme induction can be quantified. But the fact remains that hepatotoxicity cannot be well predicted by animal models due to specific differences in metabolism (see "Hepatotoxicity: From Rodents to Man").

Likewise, the understanding of nephrotoxicity, immunotoxicity and neurotoxicity is less developed than for cardiotoxicity and carcinogenicity.