Judith A. Hewitt, Office of Biodefense Research Affairs, Division of Microbiology and Infectious Diseases, National Institutes of Allergy and Infectious Diseases; Bethesda, MD
Animal models are critically important in the development of vaccines and therapeutics, not only for preliminary safety and efficacy testing to enable clinical studies in general, but in biodefense applications, they contribute the pivotal efficacy data. The gap between preliminary and pivotal animal models/studies is often judged as wide, but with the proper approach, animal models need not be the greatest hurdle to product licensure. Indeed, this perceived gap in animal models has often led to inordinate focus on this area, sometimes at the expense of other equally critical areas such as manufacturing, assays, and clinical development. Indeed, the potential for animal models to play a pivotal role is new, with the finalization of FDA’s Animal Efficacy Rule (21CFR314.610 and 21CFR601.91) in May 2002. Experience with this regulatory pathway is extremely limited, leading to intense focus on the uncharted aspect: the reliance on animal models for efficacy data. The reason for the intense focus on animal models for biological threats is twofold: first, no model has yet passed the ultimate test of supporting product licensure, and second, product developers generally don’t have the in-house capability to conduct animal model development in biocontainment, and researchers who do have the capability aren’t directly responsible for, and sometimes only peripherally involved in, product development. NIAID’s approach to animal model development is product-neutral, whereas manufacturing and clinical development are inherently product-specific activities. Product-neutral animal models are developed and assessed for their utility in testing product efficacy, and the development of the animal model is the primary outcome, even if a product is tested and information is
1 Disclaimer: This document was not prepared by the Committee on Animal Models for Assessing Countermeasures to Bioterrorism Agents or National Research Council (NRC) staff. It was provided as background information at the request of the Committee by Judith A. Hewitt, Office of Biodefense Research Affairs, Division of Microbiology and Infectious Diseases, National Institutes of Allergy and Infectious Diseases; Bethesda, MD.
gained about that product. Ideally, NIAID uses multiple products to ensure the robustness and neutrality of the animal models. Further product specific refinements may be required, but are left to the sponsors or advanced development contracts in the context of pivotal, product-specific studies. NIAID’s independent investment in the development of product-neutral animal models has been quite successful and this white paper will share examples of the most important product-neutral lessons learned and guiding principles to apply to other animal models moving forward. NIAID’s approach has resulted in the development of several animal models that are now waiting on product licensure decisions as the ultimate demonstration of their utility.
Safety testing of vaccines and therapeutics typically begins in animals, where initial assessments help product sponsors and FDA understand the potential risks in humans. Data are then collected over several phases of clinical trials, with adverse event reporting as well as post-licensure monitoring contributing to our overall understanding of the safety of drugs and vaccines. As new information comes to light, changes in a product’s label may be warranted, such as added warnings regarding special populations. Efficacy results are similar to safety results in that accrual of data supports continued development and use of the product.
Product licensure under the Animal Rule is only different from the usual pathway in that all the efficacy data comes from animal studies, by bridging data that can be obtained in both animals and humans, such as immunogenicity for vaccines and pharmacokinetics for drugs. Data that supports safety and contributes to the bridge to animal efficacy and therefore, presumably, human efficacy, must still be accrued in human subjects. FDA reviews all clinical trial protocols before execution, though there is no regulatory requirement for all animal efficacy protocols to be reviewed prior to execution. Even animal safety protocols need not be reviewed prior to execution, though executing protocols without prior review by FDA runs the risk that the data collected may not support the intended use in subsequent studies, whether human or animal.
The rigor required of any assay, animal study or clinical trial is directly determined by the decisions that will be based upon the resulting data. As a product progresses along its development pathway, increased rigor is demanded from the component assays, reagents, animal models, etc. Once each component is developed to a standard sufficient to support product licensure, further use in supporting other products of a similar nature should be straightforward.
Currently for biodefense product development against biological threats, no product is yet licensed under the Animal Rule for wide use in an event or even an emergency, therefore none of the animal models or assays has been determined to be sufficiently well developed. Indeed, the data in hand at the time of an emergency will be assessed as to their adequacy to support use of a product, therefore anticipating the nature of the emergency is the only way to potentially gauge the rigor required of the data prior to licensure; a hypothetical emergency is not the same as an actual emergency. There are two products licensed under the Animal Rule, both for chemical agents where the mechanism of action in animals and humans is extremely similar, and these products have limited licenses for use in military or first responder applications. Experience with Emergency Use Authorization for all biodefense medical countermeasures is equally limited.
Infectious disease animal models represent a dynamic system, with myriad possibilities for the relationship between host and pathogen. Both the host and the pathogen are biologic systems themselves, fraught with genetic and epigenetic variability, that when combined results in a highly dynamic situation with even more variability. An additional challenge peculiar to biodefense is that it may be difficult to relate the animal model to the human disease; we may have an incomplete picture of human disease, it may be outdated, or we may have no information at all in the case of emerging diseases. The challenge becomes even greater when using the animal model to assess the efficacy of a countermeasure—yet another player in the host-pathogen-countermeasure dynamic and yet another uncertainty in bridging efficacy in animals to humans. Will the countermeasure impact the host and
host-pathogen dynamic in a similar way in humans as in animals? This is a question scientists must wrestle.
2. DESCRIPTION OF NIAID’S ANIMAL MODEL DEVELOPMENT PROGRAM
NIAID began to address the challenge of developing biodefense medical countermeasures by first convening Blue Ribbon Panels, which developed a Strategic Plan for Biodefense Research and Research Agendas for CDC Category A Agents and Category B and C Priority Pathogens, in 2002-2003. All three documents highlighted the critical role for animal models and the needed investment. In response, NIAID’s Division of Microbiology and Infectious Diseases (DMID) funded an initiative to fill some of the identified gaps in animal models, using a flexible contracting mechanism with a substantial investment. This initiative was also aided by strategic agreements with other federal agencies, such as the United States Army Medical Research Institute for Infectious Diseases (USAMRIID). The purpose of this initiative was to ensure that animal models were developed and available for product testing, in contrast to pursuing animal model development within the context of advancing specific products through product development initiatives. This program has been very successful and has therefore been renewed and extended to encompass all pathogens in the DMID portfolio.
NIAID’s initial and primary focus was to develop the animal models needed for the highest priority and most advanced countermeasures: anthrax vaccines, antibiotics and antitoxins; smallpox vaccines; and plague antibiotics. Subsequent animal model development efforts expanded into other NIAID priority pathogens, as well as ensured that Project BioShield requirements would be supported by the fundamental research typically funded by NIH. NIAID participated in the development of the HHS Public Health Emergency Medical Countermeasure Enterprise (PHEMCE) Strategic and Implementation Plans, published in 2007. In 2004-05, NIAID funded animal models related to the emergence of SARS and increased emphasis on pandemic flu, especially highly pathogenic avian influenza animal models. In other words, NIAID has focused on developing infectious disease animal models in priority order and commensurate with the development of medical countermeasures.
In the immediate aftermath of 9/11 and the anthrax letters, and the 2002 finalization of the “FDA Animal Rule,” NIAID’s approach has been to ensure that countermeasure development goals are not hindered by lack of animal models and that those models meet the regulatory goals of the FDA. Generally, development and advancement of animal models is not a well-funded stand-alone area in NIH investigator-initiated research portfolio, as it is viewed as a rote activity requiring little or no innovation. Since gaps were highlighted by the Blue Ribbon Panel, NIAID sought to directly fill those gaps through contracts. NIAID’s emphasis in 2003-2004 was distinctly different than the current mission of the Department of Defense’s Transformational Medical Technologies Initiative, launched in 2006.
3. ISSUES AND CHALLENGES IN ANIMAL MODEL DEVELOPMENT
While the use of animal models is not new, what is new is their role in contributing critical path data, including pivotal studies, and the higher level of rigor required of those particular data. Given the dynamic nature of animal models, there are many potential challenges, and one is likely to face multiple challenges in any given program to develop a countermeasure. Anticipating and minimizing the impact of these potential challenges is important for timely progression in product development. Animal models of infectious diseases are a dynamic system involving initially just the host and the pathogen, and later including countermeasures. The actual conduct of an animal study also has an impact on its utility, not only in developing a product but informing future development efforts, referred to below as study-specific issues. This paper will first introduce these three challenges
generally (host, pathogen, study-specific issues), and then describe the practical issues experienced through NIAID’s programs in Section 4, along with how they were resolved or moved forward.
3.1 Host Issues
The choice of host species is the first critical decision to be made. In many cases, the susceptible species and the nature of their disease are known from the published literature and one can begin developing models and products in a very productive manner; if not, this must first be explored. Depending on the nature of the infection, there may be subtle differences between species and understanding those differences will ultimately lead to better choices in relating the animal disease to human disease. An example that can be drawn from studies with pyridostigmine bromide (PB), given prophylactically to prevent the effects of the nerve agent Soman, relates to the susceptibility of different animal species to Soman. Species that are typically used early, such as mice and rats, demonstrated small and inconsistent effects. Further studies in guinea pigs and rhesus macaques demonstrated efficacy, but it is very desirable to test compounds in lower species first. Later studies showed that giving rats a carboxylesterase inhibitor, which increased their susceptibility to Soman and resulted in carboxylesterase levels more similar to humans, allowed the demonstration of pyridostigmine bromide efficacy. Had the carboxylesterase levels been considered up front, the use of mice and rats would not necessarily have preceded the use of guinea pigs and non-human primates; however their use has contributed to a greater understanding of Soman poisoning, the efficacy of pyridostigmine bromide and greater confidence for PB use in humans.
The choice of host species based on the infectious disease will hopefully also be a good species for the countermeasure involved, though that is not always the case. Pharmacokinetic behavior of drugs in animals is frequently different than in humans. When a protective benefit is seen in animals with lower drug exposure, it is logical that humans with a higher drug exposure would derive the same benefit. But when the lower pharmacokinetic levels in animals are only partially protective, it is difficult to know whether a more favorable pharmacokinetic profile in humans would result in better efficacy. Such differences in efficacy unfortunately would require larger studies to be conducted to demonstrate statistical benefit, and therefore careful analysis of existing data and consideration of the path forward is crucial to limiting the use of animals needlessly.
Even within a host species, there may be strain or population differences that can have an impact. One needs to consider outbred or inbred for some species, and the choices for inbred mice in particular are numerous. Without an understanding of susceptibility, it is difficult to know if the best strain has been selected. Some species are predominantly available as outbred animals, in particular, non-human primates. In that case, country of origin of the animals or the breeding stock can have an impact, as well as prior exposure to pathogens, though these impacts may not be known a priori. The further a program develops using a particular choice, the harder it may be to change, even if a better choice exists.
3.2 Pathogen Issues
One of the next challenges encountered in developing animal models is the challenge material itself. Selecting a strain or isolate that is representative of the pathogen is a critical decision, as it is highly desirable to use one strain/isolate throughout a product or model development program. Once a body of data is obtained using a particular strain, continued use of that strain is more likely, unless there is a good rationale for changing. Understanding variability among strains/isolates is vital to making a wise selection. If one selects the most virulent strain/isolate, then it is logical that countermeasure efficacy would extrapolate to less virulent strains/isolates; the reverse may not be true. Once a strain/isolate is
selected, it is important to understand the variables contributing to its growth and virulence, in order to control them so that the challenge material is not a variable from study to study. Ideally, one would set down a master and working bank of the pathogen and any associated cells required for growth—best practices for propagation should not be an afterthought! Assays to assess the pathogen’s purity, identity and activity should also be understood and formalized as necessary to support their eventual use in quality driven studies. Genomic sequence data is particularly important for viruses, and stability programs may need to be considered for challenge material prepared in advance. While early phase studies may not require formal procedures to be established, it is advisable to adopt them as soon as possible to reduce variability. Once the strain/isolate and associated procedures are formalized, it is reasonable to also use them in discovery stage studies for future generation countermeasures. The standard challenge strain for animal studies should be included in panels for testing new candidates.
There are sometimes barriers to sharing challenge strains, beyond Select Agent regulations. Some laboratories treat a pathogen as something they own, though when it comes down to it, none outwardly claims ownership. Isolates are often considered unique, and having sole possession of the pathogen isolate makes an animal model using that isolate unique. In fact, pathogens can be propagated and perhaps that is the rationale for not sharing. NIAID firmly believes in and strives toward the sharing of strains and models and even in the replication of animal models at additional sites, as a vital component of good science. Concerns about wide distribution of strains are warranted, but distribution should not be so restricted that it prevents the development of animal models or restricts the development of countermeasures. It is important to recognize that multiple developers working on the same model serve to increase our understanding of models in a much more rapid manner than any one laboratory alone is likely to accomplish.
Having established procedures to make the challenge material, it is now time to select a challenge route and dose for testing countermeasure efficacy. There are a number of challenge routes of interest, and a number of relationships between animal model and human disease. For many diseases, the usual or expected route of transmission is known and therefore animal model develop can utilize the same route, where disease should progress similarly to humans. In some cases, the route for testing is not the natural route but rather one that has potential for biothreat use, and there may not be information on the human disease course with that route. In some cases the route of transmission is known, but replicating human disease progression in an animal model is the hurdle to be overcome. And some pathogens can be transmitted by several routes resulting in different diseases. Often the challenge route of interest is selected, either because it is known to be the natural route of infection or because it is the route anticipated in a deliberate release. If the challenge route is not known, it will need exploration.
Selecting a challenge dose requires additional information; the simplest scenario is a threshold dose required for disease, and a more complicated scenario may be encountered when a pathogen exhibits a dose response resulting in different disease patterns. Selection of a challenge dose and route may have an impact on the requirements for pathogen and therefore procedures to grow the challenge stock. Current technology for aerosol challenges delivers only a fraction of the aerosolized material into the animal(s), compared to parenteral challenge routes where the pathogen is delivered directly into the animal and one must only consider a slight overage to ensure quantitative transfer. The requirements for pathogens and concomitant changes in culture methods to accommodate challenge objectives may have unintended impacts if not carefully considered.
The impact of countermeasures on disease can similarly have a dependence on the challenge route. For newly emerging diseases, the inquiry is even greater and will rely on careful epidemiological investigations. Diseases with relatively few human cases, such as Ebola, may lead to an incomplete understanding of actual transmission, which can be further complicated by findings of
seroprevalence in the absence of disease. Of course, animal studies can also be used to help understand transmission.
Aerosol challenge technology has steadily evolved so that the challenge dose can be more precisely measured. One variable that is not as well understood is the influence of particle size on disease progression. Particle size will have an impact on deposition sites within the respiratory tract, but the impact of various deposition sites on disease progression in animal models is not fully understood. The particle size for aerosolized Bacillus anthracis spores has been well established through years of research, but our understanding is not as great when it comes to other pathogens, such as viruses and vegetative bacteria. The boundaries around particle size and deposition site for human to human transmission of smallpox are not understood, though NIAID has begun some preliminary studies to examine the effect of aerosol particle size on disease progression in rabbit/rabbitpox and cynomolgus/monkeypox models.
3.3 Study-Specific Issues
Finally, it cannot be overlooked that the conduct and reporting of specific studies can have an impact on a model and its further development. If one has a certain objective for a study, then one can only draw conclusions related to the observations made or endpoints measured within the boundaries of the quality systems applied and the intended use for the data. For example, the selection of a qualitative endpoint or assay only allows you to draw qualitative conclusions, even if an assay could be further developed to perform in a quantitative manner. The sampling or observation frequency limits the conclusions one can draw about timing or kinetics. An assay that provides pivotal information should be performed to the highest quality standard possible, and any assays performed below such a level should be acknowledged as such; in other words, it is important not to oversell assay results without building a good assay. Results reported in the scientific literature may lack sufficient detail for replication, such as methods for growth of the challenge material, critical parameters in the use of a particular assay, etc. Some of this information may be considered proprietary or have dual-use considerations, nevertheless, there is an impact on the ability of others to replicate the model if the reporting is not complete. Negative data may never be published, therefore not contributing to the scientific community’s understanding of the model, even if data are mentioned but not presented in sufficient detail and perhaps leading to others to repeat studies. Anecdotal evidence and casual observations can be critical to the advancement of animal models, yet they are difficult to handle in the scientific literature and to inform future studies.
4. 4. ANIMAL MODEL DEVELOPMENT AND TESTING IN PRACTICE
What NIAID has learned over the past seven years of our animal model development program not only advances specific models, it’s also translatable to other, future programs. This section will describe the NIAID’s most valuable lessons learned, organized by pathogen, concluding with some pathogen-independent observations on study-specific issues encountered. The examples presented here are germane to the development of animal models and do not represent all of the work performed for NIAID; product specific information such as correlates of protection, while applicable to any vaccine, do not necessarily inform animal model development.
One of the first and most successful models NIAID has developed is a rabbit model of inhalational anthrax that demonstrates added benefit of post-exposure vaccination in addition to a partially
protective antibiotic regimen. The antibiotic regimen was intentionally developed to be partially protective and not to model the licensed human regimen; the licensed regimen would have been far more protective in the context of an experimental animal study. This model was developed over a two- and-a-half year period, considerably longer than originally anticipated. It took even longer to get all the development reports and though they were formally submitted to FDA over two-and-a-half years ago, the model has yet to be rigorously tested in a regulatory context as supportive of Emergency Use Authorization (EUA), product licensure, or a label indication. However, NIAID is confident in this model and considers the product-neutral development to be completed. There may be refinements required along the regulatory pathway for specific products, but we now have a strong scientific basis for understanding how the model should behave.
Could this model have been developed in less than two-and-a-half years? Yes, under certain conditions that may or may not have been possible. The original antibiotic, ciprofloxacin, was not well tolerated by the rabbits; rabbits are not a species of choice for antibiotic testing and there were not enough data to support a final, best choice of antibiotic at the outset. The second antibiotic (levofloxacin) worked well; one could estimate that approximately 3-4 months could have been saved if this had been known through prior experience or the literature. The second obstacle encountered was related to the challenge material. The time course of disease was notably different in the first proof-of-concept study combining vaccination with antibiotics. While a trend toward added benefit was seen, it was not statistically significant as animals succumbed to disease more quickly than in previous similar studies, both with and without antibiotics. The anthrax spores used in the proof-of-concept study had passed virulence testing, but the timing of the disease was different from studies using other spore lots. Studies were repeated, using spores eliciting the typical disease course, along with additional, refined antibiotic regimens based on consultation with experts in the pharmacodynamics of the antibiotic. These repeat studies were successful and consistent, quickly leading to two more studies with two additional vaccines, demonstrating robustness of the model. The setback from this second obstacle amounted to a total loss of 9 months. The relationship between the qualities of the spores and the timing of disease is not fully understood; only that it is an important consideration in the context of this model and needs to be controlled. In hindsight, the aberrant spores had the same virulence when considering an absolute measure of virulence (LD50) but not when considering time to lethality; this was seen in multiple species, including the lot release test in guinea pigs. Lastly, NIAID originally only planned to test a second vaccine for robustness, but given the availability of a third vaccine, the issues encountered, and the high priority placed on anthrax preparedness, an additional vaccine study was added for greater robustness, adding 3 more months. Beyond the two-and-a-half year model development period, NIAID elected to perform an additional study to test an assumption made in earlier studies on the appropriate time to initiate antibiotics and vaccination. This study demonstrated that our assumption had indeed been correct and could not be further refined; had NIAID empirically determined the start time in the course of model development, or made an incorrect assumption, additional time would have been required. The development time for this model was impacted both positively and negatively based on the knowledge available for rational study design. In retrospect, the obstacles encountered led to a better understanding of some of the critical parameters around this animal model.
In contrast to the robust rabbit post-exposure vaccination model, development of a similar model in non-human primates has met many difficulties. In fact, six years after beginning this model, the body of evidence suggests that it may not be possible to develop such a model without using a very large number of animals. Ironically, it was anticipated that the non-human primate model would be developed ahead of the rabbit model, as there was already a publication combining antibiotics and vaccine (Friedlander et al, 1993) and used to license antibiotics; NIAID’s program began from this starting point. An obstacle was immediately encountered that set us back 2 years and 9 months before
finding an alternate path. Inhalational anthrax is considered to be highly lethal, yet a number of rhesus macaque controls survived challenge, which the literature and anecdotal evidence did not lead us to expect. Given the dynamic nature of animal models of infectious disease, there were multiple variables to consider. Initial efforts focused on the challenge spores and the aerosol delivery. It became clear that animals were being exposed to spores and resolving the infection, though doubts remained about the actual challenge dose received. Higher challenge doses were tested, to no avail. At that time, the cynomolgus macaque was beginning to gain favor as an alternate species, due to greater availability than rhesus macaques. When NIAID switched from using rhesus to cynomolgus macaques, the control survivor rate dropped from ~30% to <10%, low enough to rationally design reasonably-sized studies that could yield statistical significance. However, we quickly encountered the next obstacle: the response of non-human primates to antibiotics is strong yet variable. Further manipulation of the dose, duration and even the choice of antibiotic, still has not defined an antibiotic regimen that is consistently partially protective. NIAID now believes that the animals themselves are the source of this poorly understood variability, and that it may not be possible to control.
There is a study published on the combination of antibiotics and vaccines in a non-human primate model (rhesus) that achieved statistical significance for the added benefit of vaccination over antibiotics alone (Vietri et al, 2006). There are a couple of important differences between this study and the approach NIAID took. First, the antibiotics and vaccine were administered 1-2 hours after challenge. This most certainly does not reflect realistic capabilities and therefore may overestimate the efficacy of this combination regimen in a real-world scenario. Secondly, it uses a full human dose of vaccine, given three times. While data on the immune response were not presented, it is likely that the non-human primate immune response to this regimen exceeds that which can be achieved by humans, again potentially overestimating the value of vaccination in humans. It does, however, demonstrate that statistical benefit can be achieved under these experimental conditions. These data could be extrapolated to conclude that vaccines are capable of eliciting protective immune responses in animals that have been exposed and treated with antibiotics, under more realistic experimental conditions, even if the outcome (survival) does not allow us to differentiate the effect statistically. The 2007 Anthrax Vaccines: Bridging Correlates of Protection in Animals to Immunogenicity in Humans Workshop participants wrestled with this issue, though without the benefit of NIAID’s non-human primate data set. In other words: vaccine works under certain conditions, antibiotics work under certain conditions, neither interferes with the other and can be demonstrated to help the other under suboptimal conditions, therefore even under the best or real conditions, it is likely that co-administration may help and certainly won’t hurt. Indeed, physicians are likely to base their decisions not on whether an individual patient would individually experience added benefit, but whether it is reasonable and available. Basic/applied research data such as in the Vietri publication can be extremely valuable in providing context for animal models. It remains to be determined exactly what data will be required in a regulatory environment to support a post-exposure prophylaxis indication for anthrax vaccines.
In collaboration with USAMRIID, NIAID has been developing a treatment model of inhalational anthrax. USAMRIID compared the natural history of anthrax disease in rhesus macaques, cynomolgus macaques and African green monkeys, and have pursued the African green monkey model for further development. In the course of this work, some very important observations have been made, notably that animals with other infections are able to survive an aerosol challenge, presumably due to activation of innate immune function. Also, there were more rhesus macaque survivors than cynomolgus macaque survivors, in agreement with NIAID’s vaccine program. In most cases, USAMRIID identified the underlying infection, but not in all cases. It is an intriguing hypothesis that control survivors may have an undetected infection which alters immune function and therefore gives the animal advantage over the pathogen. Subtle differences are also seen in the rabbit model when using rabbits from different sources, which could be due to genetic differences and/or non-symptomatic underlying
infections. This is an important area for future research, as it may impact our ability to develop models for other diseases. A possible solution is to use pathogen-free animals on all studies in order to reduce variability due to immune activation, however, that does not reflect the diversity found naturally in the human population, and specific-pathogen free animals vary by source. Indeed the impact of the microbiome on health is a new area of investigation that holds much promise for better understanding the human diseases to be modeled in animals.
The development of a therapeutic model for inhalational anthrax has benefitted from the work pursuing vaccine models. There are many common aspects of vaccine and therapeutic models, such as understanding the natural progression of disease, using the same challenge material, and understanding adjunctive therapies/regimens that might be used for both or each as an adjunct to the other. In some cases, assays and SOPs developed for one model may serve the other. One unique feature of anthrax treatment models that required additional work was the development of an assay that could be used to diagnose a diseased state, akin to the symptoms that might cause a person to seek medical attention. Bacteremia has long been an integral assay in anthrax animal studies, but the classical culture assay requires overnight growth at a minimum. PCR approaches would yield more rapid results, but suffer the disadvantage that they don’t necessarily indicate the presence live organisms, rather they indicate the presence of genetic material. An electrochemiluminescent (ECL) assay was developed that detects the presence of toxin, notably the protective antigen component of toxin, as a reflection of bacteremia. This assay yields rapid results, in a matter of a few hours, thereby increasing the ability to initiate treatment in a timely fashion. The validation of this new assay for use as a treatment trigger requires a reasonable data set to demonstrate that a positive result is highly correlated with disease as assessed by other methods. Collection of such a data set is time consuming and will benefit from multiple sources of data being collected together as well as a good understanding of different assays and how they behave, or standardization of a single assay as a gold standard. The sooner these approaches are begun, the sooner the result will be achieved. It is noteworthy that the 2004 FDA Workshop on Strategies for Developing Therapeutics That Directly Target Anthrax and Its Toxins did not discuss the assay now being used. The US government had signaled a desire to purchase such products but had not yet communicated that a treatment indication was the most important aspect of the desired target product profile. USAMRIID had already developed an assay for detection purposes, yet the validation as a treatment trigger in animals had not begun and is still in progress.
Even after including the time to lethality as part of the LD50 spore release test, NIAID did have a treatment study that was an outlier relative to other treatment studies. We still do not fully understand this study, but there were several differences between it and other studies: the source of rabbits was different, including their specific pathogen free (SPF) status; the venous access ports malfunctioned resulting in greater handling of the animals; and the effect of spore lot cannot be ruled out.
Now that these models are nearly as far as they can go in a product-neutral fashion, the next step planned is replicating these models at additional sites, one of the final tests of a good model.
NIAID had very clear guidance from FDA that animal models to support licensure of a next generation smallpox vaccine would require a respiratory challenge route, not the intravenous challenge model that was the most advanced model at that time. Our approach was to compare three different respiratory challenge routes, namely intranasal, intratracheal and aerosol, by first determining the dose required to create disease most similar to human smallpox, and secondly to examine the course of disease with one dose by a pathogenesis study encompassing serial time points. It quickly became clear that this was a large amount of work and would be best accomplished across multiple sites. Therefore, in order to
reduce the variable of challenge strain, NIAID provided the actual challenge material. NIAID obtained the Monkeypox Virus Zaire 79 strain from USAMRIID because it had been used the most in intravenous studies. Another advantage to this approach was that the virus characterization could be carried out at the production site and then multiple sites could use the same material in animal studies. NIAID devised a testing scheme for identity, purity and activity testing of the virus stock. Upon testing for identity, it was discovered that the monkeypox isolate had an extremely low, but detectable, contamination with cowpox. NIAID recognized that it might not be appropriate to base critical decisions on the best challenge route using data based on a contaminated isolate, and while it was unlikely that the level of contamination (less than 10-6) had a major impact on the disease manifestation, it was perceived as risky. NIAID immediately sought to obtain an uncontaminated isolate and were eventually successful. In retrospect, there were no differences seen when using a contaminated and pure stock in two small studies, but nonetheless, the risk was too great to proceed with a known contamination.
The approach of using multiple sites from the outset was new for NIAID, so we chose to replicate intravenous data across all three sites, to understand the consistency of the model in the absence of prior knowledge. NIAID also sought to harmonize procedures across the sites to the extent possible, recognizing that data collection was most important. As smallpox is not uniformly fatal and lethality was not an endpoint, a definition of severe disease was adopted by all sites. An animal was recognized as having severe disease by exhibiting one or more of the following: death or euthanasia; a poor clinical assessment score of 7 or greater on a scale of 9; or a severe rash of 100 or more pox lesions. The most important variables have been controlled and others have been harmonized to the extent practicable. There are additional variables that cannot be controlled in this program, such as the source of animals. Whether the source of animals plays a role in disease progression is unknown at this point. NIAID has studied fewer animals in monkeypox within any one route, especially the multiple respiratory routes, than with other pathogens/models, to have the same understanding of variability of these models. On the positive side, there is little reason to expect variability in the course of disease (in contrast to anthrax) and indeed animals tend to present with similar signs at similar times. There is also a very visible and incontrovertible marker of disease, namely pox lesions. This disease does demonstrate dependence on the challenge dose, and so understanding that is very important. NIAID plans to test a vaccine and a therapeutic in a monkeypox aerosol challenge model for proof-of-concept of the model. As two smallpox countermeasures have been handed off from NIAID to BARDA for further development, it is unlikely that NIAID will contribute much more data to our understanding of these models. The disadvantage of further model development occurring in the context of specific product development pathways is that such data may not become publicly available for future countermeasures. There is not a groundswell of demand for a meta-analysis of data from across studies or study sites, though NIAID would certainly participate should a meta-analysis be performed.
In the course of understanding different respiratory challenge routes, NIAID also collected detailed information on the intravenous challenge route. This route is important for antiviral countermeasures, as product developers and others in the scientific community have argued that the intravenous route of challenge, while not natural, is a more stringent test of antiviral efficacy. NIAID’s studies have certainly informed the path forward for therapeutics, though the outcome remains to be seen. Earlier comments about study conduct certainly apply here. Ultimately, when all of NIAID’s data is submitted to FDA, our investment in understanding disease by different challenge routes will help all smallpox countermeasures.
NIAID’s biggest success in plague has been getting animal data to support a treatment indication for licensed antibiotics. This will not use the Animal Efficacy Rule, but rather 21CFR314.500, Subpart H, “Accelerated Approval of New Drugs for Serious or Life-Threatening Illnesses,” as was the case for licensure of ciprofloxacin for post-exposure prophylaxis against anthrax. Subpart H allows for accelerated approval based on a surrogate endpoint, which is the serum level of antibiotics. Since licensure is based on drug levels, the assays to measure drug level must be validated, and this was our biggest hurdle. Many of the antibiotics NIAID is testing have been licensed for many years and therefore the assay technology is old and may not have been validated to today’s standards. Performing pharmacokinetic studies under Good Laboratory Practices (GLP, 21 CFR 58) is a paradigm shift as well, but it’s critically important to choose an antibiotic dose for animal studies that is not more favorable than that anticipated for use in humans.
As a general rule, plague is fatal and animals in a challenge study present with disease within a few hours of each other. This led to rapid development of a treatment model based on temperature as the trigger for therapy. The natural history study leading to this decision happened to have two animals that did not get a fever or become bacteremic, therefore giving great confidence in the fever trigger. Those two animals received a very low challenge dose and provided an opportunity to improve SOPs to ensure that the challenge material was not compromised. There was not a positive control for these studies, as no antibiotics were known to work in a therapeutic setting. The first antibiotic tested worked reasonably well but was not completely effective; there was no way to know if this was a limitation of the antibiotic or the animal model. Further work demonstrated that two other antibiotics were completely effective, leading to great confidence in the model. In fact, NIAID has now transferred the animal model to two additional sites where it has behaved similarly, giving us the ultimate confidence in the model overall. Getting label indications for these antibiotics is now in a regulatory arena and is beyond the scope of this paper.
NIAID’s tularemia models are still under development; however, one observation thus far is that the methods for bacterial culture have an impact on the virulence and disease in small animal models. Vegetative bacteria are best cultured fresh, just before challenge, rather than prepared in advance and characterized by lot, as viruses and spores are typically handled. NIAID is currently working to harmonize the growth methods across various sites and models within our contracts, i.e., translating these small animal findings and methods into non-human primate studies.
4.5 General Observations on Study Conduct
This section will summarize some general lessons learned in the conduct of numerous animal studies. These are not unique to biodefense animal models, however they deserve mention regardless. NIAID aspires to develop animal models in accordance with the three R’s (replace, reduce, refine), and some specific examples follow. While searching for a partially protective antibiotic regimen and encountering difficulties, NIAID took a parallel approach of using in vitro hollow fiber studies to determine a regimen that would be expected to limit but not completely abolish bacterial growth, as a way to achieve a partially protective regimen in vivo. Unfortunately, the regimen modeled in the hollow fiber studies was still too protective in animals, most likely due to immune functions not represented in the hollow fiber system. In the course of a model development program, one begins to understand the behavior of the control groups over time, as well as treatment groups. NIAID has
refined our assumptions for appropriate power calculations, and thereby reduced the number of animals in particular arms such as the control group, and weighting group sizes appropriately based on anticipated effects. NIAID has also used smaller control groups and supplemented those controls with data from historical controls; this approach works in well characterized and uniform diseases. Attention to detail in the culturing of infectious agents is vital to successful studies and using fewer animals.
One can only address questions one sets out to ask. In hindsight, NIAID performed studies which would have benefitted from the collection of additional samples at different times; on a few occasions, there were samples available to perform additional assays. Thorough development of a protocol is very important, as is careful execution and analysis of data generated. NIAID staff are notorious for plotting multiple studies on one graph and even further, multiple studies from multiple sites, multiple countermeasures, etc. While combining data sets might be considered poor practice in a statistical or regulatory setting for a therapeutic or vaccine, it is a great tool to better understand animal models.
Finally, implementing GLP is often undertaken before it is necessary, perhaps due to optimistic expectations of individual studies (or products) rather than looking at a program as a whole, including the developmental status of the animal models as well as the product under consideration. NIAID has certainly learned over time how better to determine when to conduct a study under GLP, and this is difficult to relay without knowledge of specific studies under consideration and the data set to support that study design.
5. ANIMAL MODEL DEVELOPMENT FOR BIODEFENSE ACROSS THE US GOVERNMENT
There has been a high level of interagency discussion for years, though true coordination efforts are really just now maturing. Historically, the Department of Defense had been funding the vast majority of biodefense research and product development for years, and began sharing expertise in the immediate aftermath of 9/11 and the anthrax letters. NIAID really became a major player in biodefense beginning with the 2002 budget increase. The Department of Health and Human Services similarly became a major player with the passage of Project BioShield in 2004.
Early interagency coordination efforts focused on countermeasures and reported to the Weapons of Mass Destruction Medical Countermeasures Subcommittee, originally convened by the Office of Science and Technology Policy and later chartered under the Committee on Homeland and National Security, under the National Science and Technology Council. One of the working groups chartered under this committee structure was the Product Development Tools Working Group (PDT WG), which I was asked to co-chair, charged with ensuring the availability of biocontainment facilities, animal stocks, animal model development,” validated” experimental protocols and “validated” assays. This WG had representation from many agencies: DoD, DHS, HHS, FDA, NIH, and CDC. While the charge was expansive, the outputs of greatest interest focused on the status of animal models for various classes of countermeasures under consideration for acquisition. It quickly became apparent that the term “validated” meant different things to different people and that better terminology was required before moving forward. Indeed, one of the most widely used outputs of that group was the development of Technology Readiness Levels (TRLs) for product development tools such as animal models, assays and challenge material in 2006. These TRLs have been used within government for some time and are now available on the BARDA website (https://www.medicalcountermeasures.gov/TRLs_for_PDTs.aspx). Publicizing this assessment tool was delayed, in part by harmonization of DoD and HHS versions of the TRLs for countermeasures, resulting in slight adjustments in the PDT TRLs. The PDT WG performed an assessment of animal
models for the Public Health Emergency Medical Countermeasures Enterprise (PHEMCE) in early 2007, prior to finalization of the PHEMCE Implementation Plan. The PDT WG assessment confirmed that it was indeed reasonable to plan the acquisitions in the timeframes under consideration relative to the status of the product development tools. Indeed, many people lament the “lack” of animal models, but from the perspective of NIAID in 2010, the animal models have seldom been on the critical path for advancing product development. NIAID now requires animal model contractors to use this assessment tool in annual reports.
The PDT TRLs capture a few additional lessons learned by NIAID. A tool can only progress so far without concomitant investment in products, and both products and tools mature in parallel fashion. Once a tool is used to support licensure of a product, it can be used again for future products, and may or may not require refinement, depending upon the intended use of the tool and how similar new products are to products licensed using the tool. One can think of the PDT TRLs as capturing three successive phases of animal model development: product independent, product dependent and product specific. Product independent studies are the early studies of the disease itself, such as pathogenesis and natural history studies and do not require a product to be available. Product dependent studies include the proof-of-concept studies to demonstrate that an animal model can test the efficacy of a product, and therefore require a product, whether licensed or very early in development. Both product independent and product dependent studies are considered to be product neutral when considering animal model development. Product specific studies are those which are performed under GLP and in the context of testing a specific product, ideally produced under Good Manufacturing Practices (21 CFR Parts 210-211, 600-680) and using the final formulation and dose, along with assumptions about efficacy of a specific product (not a product class) for statistical power. NIAID has used products in our product-neutral animal model development program, through sources such as product-development contracts or informal partnerships, and NIAID places a strong emphasis on testing several products to ensure that the model is robust and product agnostic.
More recent efforts in interagency coordination have shifted from asking “is it possible” to “is there enough capacity” to conduct animal studies. One of these efforts is aimed toward defining the need for an expanded interagency role for testing countermeasures at USAMRIID. The other is part of the Integrated Portfolio for CBRN Medical Countermeasures, formally chartered in 2009, which is mandated to coordinate efforts between DoD and HHS. This effort makes use of many ad hoc Integrated Program Teams that contribute and consolidate information from various agencies to create one portfolio in specific countermeasure areas. Another group reporting within this structure is the Animal Studies Queue Evaluation (ASQE) Team. The ASQE Team, of which I am a co-chair, has been in existence for less than a year and has the heroic task of assessing availability of animal model capability to support all the products in a particular pipeline for a particular pathogen, and to recommend a path forward if there are constraints. Initiatives that fund countermeasure development sometimes do not take into consideration feasibility and capacity for animal studies when contracts are awarded, or perhaps only for that pathogen and not in the context of other initiatives competing for the same biocontainment space for animal studies. The US Government needs to consider this capacity and appropriate sequencing of contract awards and activities when publishing solicitations. The ASQE will strive to ensure that this is the case where animal models are concerned, but the ASQE is only a recommending body, not a decision making body.
6. FUTURE EFFORTS
NIAID has recently renewed and expanded our successful biodefense animal model contract program. The expansion represents additional models in other non-biodefense pathogens of interest to NIAID, though perhaps not of interest to this Committee. Certainly the experience and advanced level of our
biodefense animal models will inform future efforts in less mature or non-biodefense animal models. It is important that future models be developed and studies performed in the most appropriate facilities—early proof of concept models may not be the best utilization of resources if performed in the expensive and limited environment of a GLP facility. It would be highly appropriate to develop animal models in research facilities and successfully transfer them to a fully compliant GLP environment, and a history of smooth transition from research to GLP facilities will help establish a greater level of comfort in the appropriate placement of animal studies by the product development community. Research laboratories need to consider early choices that may impact the transferability of models, such as the creation of master and working banks of pathogens and standardization and definition of limits of procedures, in order to successfully transfer models to a GLP environment. The term “GLP-like” is often derided as meaningless, but it can be difficult to transfer a model from research laboratory practices to GLP without going through an intermediate of good laboratory habits and documentation, with an eye toward the ultimate goal. The current state of animal model development is progressive in nature, approached by incremental advances rather than cumulative advancements all in one study. Other approaches may be feasible, but are uncharted regarding the final goal of successful implementation in regulatory decisions.
Over the past seven years, NIAID has developed an extensive program devoted to animal models that are coordinated with, but not direct results of, product development efforts. The product-neutral nature of NIAID’s animal model program has focused on developing the best possible animal models, without the potentially competing interest of furthering a product. This approach has been very productive and has tremendously helped the regulatory framework for assessing product efficacy. NIAID’s models are assessed as models in their own right, without a concomitant assessment of a product. NIAID’s objective approach to animal model development, along with a high level of investment, has been very successful and should be viewed as the standard when approaching Animal Rule efficacy to support medical countermeasures for biological threats.
“Nothing happens quite by chance. It’s a question of accretion of information and experience.”
Friedlander AM, SL Welkos, MLM Pitt, JW Ezzell, PL Worsham, KJ Rose, BE Ivins, JR Lowe, GB Howe, P Mikesell and WB Lawrence, “Postexposure Prophylaxis Against Experimental Inhalational Anthrax,” J. Infectious Diseases, 167: 1239-42, 1993.
Vietri NJ, BK Purcell, JV Lawler, EK Leffel, P Rico, CS Gamble, NA Twenhafel, BE Ivins, HS Heine, R Sheeler, ME Wright and AM Friedlander, “Short-Course Postexposure Antibiotic Prophylaxis Combined with Vaccination Protects Against Experimental Inhalational Anthrax,” Proc. Natl. Acad. Sci., 103: 7813-7816, 2006.
The conclusions in this white paper are based upon projects managed by a number of people in the Office of Biodefense Research Affairs: Ken Cremer, Martin Crumrine, Kristin DeBord, Robert Johnson, Freyja Lynn, Tracy MacGill, Ed Nuzum, Blaire Osborn, and Thames Pickett. Thanks to Rose Aurigemma, Sue Garges, Irene Glowinski, and Mike Kurilla for critical reading of this manuscript.