Toxicity Testing in the 21st Century: A View from the Pharmaceutical Industry
James S. MacDonald, Richard T. Robertson · Toxicological Sciences · 2009
The report by the U.S. National Research Council entitled Toxicity Testing in the 21st Century: A Vision and a Strategy (National Research Council, 2007) lays out a bold vision for the future of toxicity testing of chemicals based on the explosive changes that have been and are occurring in the basic biological sciences. Our understanding of basic cellular biology has grown remarkably in the last years to the point where many cellular processes are well characterized at the molecular level. In contrast to this dramatic change, the approaches that have been taken to assess human risk of adverse effect from chemical exposure have changed little over the last several decades. This is reflected in the global regulatory requirements for registration of new agricultural, veterinary, and human pharmaceutical chemicals; the data requirements for these classes of chemicals have changed little since their establishment three and sometimes four decades ago despite the dramatic advances in the sciences that are used for this activity. This fact was a central point raised in the Food and Drug Administration's assessment of the processes used specifically for drug development in their Critical Path document (FDA, 2004). Improvements in the ability to predict important human adverse effects were seen by this Agency to be central to the ability to accelerate the new drug discovery and development process and bring more effective medicines forward safely and quickly. Although it is clear that there is a need for improvement in the processes used for human hazard identification and risk assessment, it is important to view this goal from the perspective of the reality of what it is we are intending to do. Protection of human safety is a primary objective of toxicology and it is imperative that the tools used to accomplish this task are fully adequate to enable effective risk management decisions. Toxic effects generally result from numerous complex and interrelated events in intact organisms. Although our basic understanding of molecular and cell biology is increasing at a remarkable rate, there is still much we do not understand. As individuals who have actively practiced toxicology for over three decades in the pharmaceutical industry, we heartily endorse the need for change. Although we feel it is imperative that we move forward aggressively to understand how best to use the new information in the basic sciences, we think it is necessary to do this from a realistic perspective. As drug hunters we are, unfortunately, often reminded that when dealing with novel clinically unprecedented targets, the extensive in vitro and in vivo preclinical work does not predict clinical outcomes. In Part I of this Forum series, Andersen and Krewski describe a very desirable ultimate approach to the identification of human chemical hazards and a significantly improved risk assessment process from these data. The toxicity testing strategy outlined involves four main components: chemical characterization, toxicity pathways and targeted testing, dose-response and extrapolation modeling, and human exposure data (Andersen and Krewski, 2009). It is useful to illustrate the magnitude of the challenge in achieving this vision from the perspective of the pharmaceutical industry by raising several points to consider for these components. A central element of the vision is an understanding of what are described as “toxic pathways.” Years of research and experience in the pharmaceutical industry have taught us that adverse events result from a diverse array of stimuli and not infrequently from “exaggerated pharmacology.” At the high end of the dose-response continuum, the desired pharmacologic effect is deleterious to the organism and expressed as an adverse reaction. It is useful to evaluate the broad spectrum of adverse effects that can be elicited by pharmaceutically active agents. Although this spectrum should not be a surprising observation given the fact that these agents are designed to produce some specific perturbation in a cellular process (in order to achieve a beneficial therapeutic outcome), it is illustrative to consider the breadth of adverse effects drugs can elicit to better understand the magnitude of the challenge we are facing when we consider how best to predict—and prevent—these events in patients. A survey of the reasons for drug withdrawals after introduction into the marketplace from 1998 to 2008 from the US, European, or Asian markets is provided in Table 1. Table 2 shows a summary of the primary reasons for these withdrawals. Smith and Schmid provide a similar survey and group the toxicity into four primary categories ranging from exaggerated pharmacology, secondary (off-target) pharmacology, or direct organ toxicity (Smith and Schmid, 2006). It is important to emphasize that the toxicity leading to withdrawal is complex even within a specific tissue or target organ system. For example, the hepatoxicities that led to withdrawals in this period resulted from a wide range of specific molecular events including mitochondrial toxicity, cholestasis, direct cytotoxicity from a classical reactive metabolite-mediated pathway, to unknown idiosyncratic events which culminated in severe hepatotoxicity. Similarly, the cardiac toxicity cited here resulted from alterations of valvular function, fatal arrhythmias, alterations in ion channel function (leading to QTc prolongation), and other effects of unknown mechanism. It is encouraging to recognize that, because of a greater understanding of many of these mechanisms and the employment of screening systems in the development process (in vitro and in vivo), many of these serious toxicities can be avoided today (e.g., cardiac arrhythmias resulting from human ether-a-go-go-related gene channel blockage). Survey of Drug Withdrawals over the Period of 1998–2008 Note. ADHD, attention deficit hyperactivity disorder; CNS, central nervous system; COMT, catechol-O-methyl transferase; CV, cardiovascular; GI, gastrointestinal; NMDA, N-methyl-D-aspartic acid; NSAID, non-steriodal anti-inflammatory drug; PNS, peripheral nervous system; QTc, corrected QT interval. Survey of Drug Withdrawals over the Period of 1998–2008 Note. ADHD, attention deficit hyperactivity disorder; CNS, central nervous system; COMT, catechol-O-methyl transferase; CV, cardiovascular; GI, gastrointestinal; NMDA, N-methyl-D-aspartic acid; NSAID, non-steriodal anti-inflammatory drug; PNS, peripheral nervous system; QTc, corrected QT interval. Summary of Primary Reasons for Drug Withdrawals from United States, European, or Asian Markets over the Period of 1998–2008. Other includes the following: renal dysfunction, accelerated carcinogenicity or death, mutagenesis, severe drug-drug interactions with ethanol, hypersensitivity, pulmonary hypertension. N = 51 instances of drug withdrawals. Summary of Primary Reasons for Drug Withdrawals from United States, European, or Asian Markets over the Period of 1998–2008. Other includes the following: renal dysfunction, accelerated carcinogenicity or death, mutagenesis, severe drug-drug interactions with ethanol, hypersensitivity, pulmonary hypertension. N = 51 instances of drug withdrawals. Perhaps more instructive for the present purpose is an understanding of those toxicities that have caused either significant warnings to be added to product labeling worldwide after registration and product launch or those effects that have been observed in animal studies that cause significant disruption to or termination of new drug development programs. A listing of such effects from a survey of global regulatory action letters and from an analysis of drug candidate liabilities in our laboratories over three decades is shown in Table 3. (Similar diverse adverse responses are confirmed by reports in the literature such as Giezen et al., 2008; Schuster et al., 2005; Wysowski and Swartz, 2005.) Although this list is not meant to be a comprehensive survey, it is intended to illustrate the diverse effects and mechanisms experienced routinely in pharmaceutical development (and by inference, that may be expected to play in role in chemically mediated toxicity generally). To effectively protect public health, any new approach to identification of potential human health hazards will need to encompass issues of this breadth to be of real value. Examples of Drug-Induced Toxicities in Humans that Led to Regulatory Action Letters or Toxicities in Animal Studies that Led to Alteration or Termination of Drug Development Programs Examples of Drug-Induced Toxicities in Humans that Led to Regulatory Action Letters or Toxicities in Animal Studies that Led to Alteration or Termination of Drug Development Programs If data generated from a collection of cell-based systems can be used to categorize xenobiotics, then we must have human cell-based systems which maintain both genotype and phenotype. Examples of the challenges include attempts in the pharmaceutical industry to understand the interactions of drug candidates with the CAR (constitutive androstane receptor) and PPAR (peroxisome proliferator–activated receptor) nuclear receptor families. CAR is a nuclear hormone receptor with a very promiscuous, but in some cases species-specific, collection of ligands capable of activating all or part of its cellular regulation systems. CAR has been implicated in both natural and xenobiotic biotransformation, lipid regulation, thyroid homeostasis, hepatocellular proliferation, and carcinogenesis, to name a few. In spite of a great deal of effort, there have been no successful attempts at maintaining the activity of this multi-functional system in cell constructs. As mentioned earlier, there are major differences in the ligand selectivity between species including ligands that are potent activators in mice with no activity in rats, and the relevance of either response to humans cannot be assessed with present technology. The same challenges exist in attempts to maintain the complex functioning of the PPAR nuclear receptor system with its attendant comodulators, stimulants, and repressors in cell-based systems. Although we understand much about the molecular pathways upstream and downstream of these receptors, it is not currently possible to assess the human relevance of these interactions using in vitro cell-based systems. Evaluating the potential adverse effect of immunomodulating agents further illustrates the challenges presented by the proposed systems-based modeling effort. These agents interact with another complex system of multicellular functions with built in costimulatory and inhibitory pathways, and genotypic variabilities which have proven understandably difficult to model accurately. The same is true of the central nervous system, where both in vitro and in vivo models have proven very poor surrogates for neuropharmacologic agents. In fact, most of the predictive in vivo models for neuropharmacologic activity were “validated” after activity had been documented in humans. Reproducing in multiple in vitro test systems the complex intracellular dynamics in a manner that permits the test systems to be reliable predictors of apical responses in exposed individuals is a significant challenge. Even seemingly homogeneous cell populations respond differently to chemical stimuli. Newly available technologies permit retrospective understanding of why these responses differ but are still very difficult to use in a prospective, predictive manner (Cohen et al., 2008). The National Research Council (NRC) report acknowledges the difficulty in predicting in vivo biotransformation and kinetics from in vitro systems. The pharmaceutical industry has had a great deal of experience in this area—and a reasonable level of success particularly in recent years. Using in vitro cell constructs and single dose multispecies screening, our experience is that we can predict the bioavailability and clearance of potential drug candidates in humans within a factor of 2 about 60% of the time and within a factor of four greater than 80% of the time if the principle clearance is via Phase I oxidative pathways. However, if we restricted our in vivo testing to rodents as suggested in the NRC report, the predictive power and success rate drops significantly in our experience. The NRC report proposes the use of in vitro approaches to determine outcome of chemical biotransformation as a means of providing further information on the potential toxic effect of test agents. A particular challenge faced in the pharmaceutical industry can help to illustrate the complexity of even such a seemingly simple process. We have reasonable success using relatively high throughput screening systems to identify agents that can covalently react with cellular nucleophiles. Although it has been proposed that this is a viable means of determining which compounds have the potential to cause direct reactive metabolite-mediated toxicity and thus can be eliminated from early screening tests (Evans et al., 2004), there is enough uncertainty around the outcome of such studies to make it difficult to utilize data from these screens alone for decision making even in early stages of drug development (Obach et al., 2008). This difficulty becomes even more apparent when considering those compounds that can form acyl glucuronides. Although there is ample evidence for the reactivity of this metabolic moiety (and associated significant human toxicity), the implication for potential toxicity is not simply a matter of formation or even reactivity of the chemical species (Bolze et al., 2002; Dong et al., 2005; Spahn-Langguth et al., 1996). It is true that with application of newer analytical technologies (e.g., proteomics, systems modeling) we are making important progress on integrating our understanding of protein modification by chemicals and the consequent clinically relevant health effects (Liebler, 2008). At the moment, however, these exciting and powerful tools are able to permit hypothesis testing rather than definitive human hazard identification or risk assessment. Biotransformation and generation of reactive intermediate metabolites have been linked to many modes of cellular toxicity including idiosyncratic reactions (Park et al., 1998). These represent one of the most difficult issues to address in drug development as our predictive tools are seldom capable of detecting even serious, life-threatening adverse reactions that are seen in only a very small percentage of the exposed population. It is inadequate to simply be able to identify the possibility of a chemical modification; understanding the specific of specific metabolites and metabolic pathways in humans in to understand the of the chemical is evidence that, as in the of all of dose is It has been that very idiosyncratic drug reactions have been seen with drugs given at a dose of or In approaches to chemicals and potential specific adverse effects is increasing Programs such as and are used routinely in the stages of drug discovery to identify those molecular that be associated with specific adverse effects to direct in a The primary today is These are effective at identification of associated with but are designed to be at hazard If we used to we have that all for These tools are used in potential from toxicity that be expected from and in active pharmaceutical and For this these in tools on what tests may be necessary or further is et al., these tools are however, to predict more complex events system toxicity, for the predictive As these are and the predictive with the used for this in predictive is not as the available for more complex toxicities is of these will the of these screening tools at the moment, their is restricted as definitive predictors of adverse A array of high throughput screens for of is in These include both cell and cell-based systems. The years has seen dramatic in the application of cell-based systems which the same approach to the effects of drug candidates on cellular pathways as in the NRC This is in the discovery to target are of effects and that the desired biology is as specific and potent as necessary to achieve the desired therapeutic compounds are actively that produce significant toxicity in either the preclinical toxicology animal testing in human clinical In the experience of the pharmaceutical industry, it can be that greater than of the adverse effects drug are based effects at the end of the dose-response The NRC report that in vitro test systems such as those proposed for a toxicology screening approach have been in use for many years in the pharmaceutical industry Although this is as described these approaches are a specific that is to have potential therapeutic This is very from how these approaches are proposed to be used in the NRC In contrast to the use in the pharmaceutical industry, the proposed approach involves screening with an It is not clear at this point how data from the proposed screens in the NRC vision enable a of an that be deleterious to the exposed It is possible to changes in cellular pathways at very of potent active agents but a of these changes represent a toxic hazard to humans will be difficult at best with these data from cellular test systems. This points out a challenge faced by our attempts to move our approach to that by the NRC is a need to the new approaches in a that potential adverse effects in humans. We can responses to chemical exposure in some in animal studies using approaches (e.g., of in and small of of organ with the (and tools of clinical and clinical and such as and in This us to the between these responses and those seen with some of the newer approaches that on the “toxic in the NRC report in an to the As mentioned however, we often that the “toxic and cellular regulatory are one and the same with the differences in of response only by the magnitude of effect (and to the exposure at the target The classical approach to generally taken and by in such as on the of Animal and for the of is to how well new systems those seen in using Although the outcome of such studies our understanding of the role of in cellular pathways on apical responses in it does little to our ability to predict with any greater the potential in humans which of the ultimate goal of all of these of this is seen in attempts to to assessment of human risk of from chemical The currently is to on data from several in vitro and in vivo toxicity with the of 2 and studies at high to determine the potential of chemicals Although a of have been proposed that a more global approach all available data et al., 2004), all of the attempts to new approaches use data from the for Although this the new approach for ability to predict the outcome of the it does little to the predictive power of the for true human A further of the challenges faced by any new approach to of human risk is provided by the work several years ago to to understand how well the currently approaches predict human risk of toxicity from pharmaceutical agents. et describe a of work the and group to survey pharmaceutical and those compounds for which human toxicity had been and for which preclinical data were from compounds were with over human adverse Although this the of animal toxicity screening, there were still many important human toxicities that were not in either or tests between human toxicity and animal toxicity was if both and data were and 60% for alone and for only At level of understanding of the cellular responses that to human toxicity, it is that a predictive approach that on in vitro even the level of predictive seen with animal studies It is clear that there is a need to the process of human risk assessment from chemical of the pharmaceutical industry is the need to address the very of new and chemicals for which there are very data is a to move forward with new the pharmaceutical industry, this need is by many by the move to and more diverse toxicity Although the is to still on animal tests in a wide of in vitro are used early in the drug discovery to to the of the more toxicology studies designed to human Our from these new can be important to how one can newer technologies such as those proposed by the NRC report in a manner and the of more approaches to assessment of potential human we have experienced is that the use of in approaches can be very powerful in understanding the of action of a pharmaceutical chemical the tissue response is in This information can thus be used to evaluate the of the for human safety with data to the for either the forward or development et al., Using the same data however, is much more today is how are used in this several years at data the response in of at a single time point to be a very powerful to out potential toxic In however, the last several years have seen many move from using this approach in a manner on such technologies to help better the process of understanding the of observed in animal The that most is to a Using of toxic potential with a application of all of the new technologies in a manner will to the best possible human risk assessment using a global of evidence approach that all the available data as that proposed by the human relevance et al., this in the manner proposed by the NRC report will enable the most of the newer technologies and enable true to our task rather than simply another We feel that it is imperative that data from the technologies not be to the process of human risk assessment the means to assess the relevance of the data that are from these are fully To do will these very We endorse this and comprehensive A presented in the pharmaceutical is the extensive human clinical experience that is available on many A retrospective of the of the responses seen in animal studies with a assessment of the effects on pathways accelerate our understanding of the of the new approaches and help to and technologies that can risk assessment in the desired target for this the development of new and improved predictive It is important in providing on a this to the of in any is difficult and The heartily endorse to and the process of human risk assessment using the best available and the NRC report a bold vision in this This new process however, in the of the complexity of the challenge presented to on the of the vision and provide new approaches of real in this important We to their to for in the available literature on toxicities and the information in the