The challenge of accurately predicting drug toxicities and efficacies is in part due to inherent species differences in drug metabolizing enzyme activities [4, 5] and cell type-specific sensitivities to toxicants . For example, a dramatic difference between rat and human drug metabolism is demonstrated by the metabolism of coumarin. Species differences in coumarin toxicity appear to be mediated through two major phase I metabolic pathways. The first involves the conversion of coumarin by cytochrome P450 CYP2A enzymes to the nontoxic metabolite 7-hydroxycoumarin. Whereas in humans this reaction is very efficient, in rats CYP2A enzymes preferentially catalyze 7α-hydroxylation of testosterone rather than coumarin 7-hydroxylation  and therefore the formation of 7-hydroxycoumarin is extremely low in rats and the lack of 7-hydroxycoumarin is thought to render rats more susceptible to hepatotoxicity . The second pathway involves detoxification of the epoxide intermediate coumarin 3,4-epoxide. In humans, coumarin 3,4-epoxide spontaneously rearranges to o-hydroxyphenylacetaldehyde, which is further detoxified by oxidation to o- hydroxyphenylacetic acid. In rats, this conversion to o- hydroxyphenylacetic acid is 50 times lower than in humans, and represents a second major species difference in coumarin-mediated heptotoxicity . Certainly, a clear understanding of the role of drug metabolism in toxicity and potential species differences in response to toxic metabolites will aid in the development of safer drugs.
Inherent differences in cell type-specific sensitivities between species may also be an important differentiating factor for the prediction of drug toxicities. An example of differential cell toxicities across species is observed with bizelesin, a potent synthetic derivative of the anticancer agent CC-1065 that preferentially alkylates and binds the minor groove of DNA. Results with myelopoietic cells in vitro, in the absence of liver metabolism, reproduced in vivo species differences in myelosuppression, hence showing that murine cells are 1,000-fold more sensitive than human or canine cells . In the case of the antidiabetic drug troglitazone, metabolism by CYP3A4 and CYP3a1 in human and rat, respectively, results in a reactive quinone intermediate that can bind to proteins and nucleic acids and can potentially produce oxidative stress through the generation of reactive oxygen species via redox cycling or depletion of the oxidative stress protective tripeptide, glutathione. Using rat and human hepatocyte monolayer cultures, Lauer and colleagues demonstrated strong species differences and marked alterations in expressions of genes involved in xenobiotic metabolism as well as oxidative stress after treatment with troglitazone . The higher sensitivity of human hepatocytes towards troglitazone treatment in contrast to the weaker effects observed in rat hepatocytes at least partially explains why the strong hepatotoxic effects of troglitazone in the human population could not be predicted from regulatory animal studies . The results of these in vitro studies suggest that in addition to differences in drug metabolism, target cell sensitivity differences may also account for species differences in toxicity.
A review of FDA-approved drugs released from 1975 to 1999 estimated that 2.9% of these marketed drugs were withdrawn from the market due to severe adverse drug effects . According to Temple and Himmel, hepatotoxicity was the most frequent single reason for removing drugs from the market during this timeframe, and probably will still be the most important single toxicity leading to withdrawal or significant modification of labeling for FDA-approved drugs going forward . Examples of drugs removed from the market due to hepatoxicity during this time period include ticrynafen, benoxaprofen, bromfenac, and troglitazone. In addition to hepatoxicity, six drugs were withdrawn from the market because of hemolytic anemia (nomifensine maleate), hemolytic anemia, renal and hepatic injury (temafloxacin), thromboembolism (azaribine), anaphylaxis (zomepirac sodium), acute but reversible renal failure (suprofen), and increased mortality (flosequinan). Based on estimates from Lasser and colleagues, the probability of a new drug acquiring black box warnings or being withdrawn from the market over 25 years is approximately 20% . Through development of human cell-based MPS, it is hoped that toxicities as mentioned above will be detected earlier, preferably during the preclinical phase of drug development.
Given the list of FDA drugs withdrawn due to unpredicted toxicities in the human population, a key goal of the NIH MPS Program is to accurately evaluate the predictive value of these systems in determining drug toxicity and efficacy. This goal can be achieved through either development of a single microphysiological system, such as a liver organoid system that can reliably predict human hepatoxicity, or through a more complex platform where multiple MPS predict multiorgan toxicities to a given challenge drug. The multisystem platform approach will also assist in understanding the time dependence and magnitude of drug-drug interactions; for instance, how one drug may increase the blood concentration of another, as seen in the effect of mibefradil on midazolam blood concentrations . Understanding the limitations of these model systems will also be important. For instance, in cases where recognition of toxicity in the human population has been delayed - such as seen in the marked hypo tension with clozapine, the pulmonary fibrosis and marrow toxicity with tocainide, the valvulopathy associated with fenfluramine hydrochloride and the subarachnoid hemorrhage associated with phenyl propanol-amine  - the application of relatively short-term exposures of drugs in the MPS may limit the predictive value of the system for delayed toxicities.