The aim of this function is to demonstrate a clear and reproducible ability of the vaccine to protect animals from progressing to TB disease or to prevent Mtb infection. Different animal species, modes of infection and disease endpoints are used. Although there is no single standardised animal model of TB, there are several recognised animal models/study designs that can be used to compare different vaccine candidates. The most appropriate model or study design will vary according to the type of vaccine, but it is generally recommended to perform studies in mouse challenge models initially and progress to more complex models such as guinea pigs or nonhuman primates (NHPs) with promising candidates. It is recommended, particularly from Stage B onwards, to use independent laboratories for protection experiments which use standardised methodology, including blinded analyses to ensure un-biased interpretation of the results. As mentioned under pre-clinical immunogenicity, there is no validated animal model that is accepted as being predictive for TB vaccine efficacy in humans.
The ability of a candidate to reduce or limit the progression of experimental infection with Mtb must be demonstrated in an appropriate animal model. Different animal species, modes of infection and read-outs of disease are used. These are described in reviews such as Cardona and Williams 2017, Singh and Gupta 2018, Williams and Orme 2016. Typically, efficacy is first demonstrated in mice or guinea pigs. There are many examples of these species being used for testing the efficacy of TB vaccines which are illustrated in the following references; Stylianou et al 2018, Doherty et al., 2004, Reed and Lobet 2005, Williams et al., 2005, Brandt et al., 2004, Clark et al., 2016
There are two, commonly used laboratory-adapted strains of Mtb used for challenge, preferably via the respiratory route, these are the H37Rv and Erdman strains. There is no specific requirement to use these strains and it may be considered an advantage to use clinical strains. The challenge strain should be well-characterised and from a source that allows reproducibility between experiments. The design of the vaccination schedule should be informed by previously generated immunogenicity data. In the case of protection studies in mice, the strain of mice used should also be consistent with the strain used to demonstrate immunogenicity. In the absence of a correlate of protection, evaluating multiple vaccine dose levels in early-stage challenge studies should be considered. The outcome measures used to demonstrate protection typically include a reduction in bacterial burden in lungs and other relevant organs, survival, or quantifiable changes in pathological features of TB disease. A statistically significant change compared to a relevant control group must be demonstrated. Statistical power calculations based upon a pre-defined target level of protection should be used to design experiments with appropriate group sizes. The control group against which efficacy is compared must be justifiable and appropriate to the nature and intended TPP of the candidate. For example, a novel live-attenuated whole cell vaccine should show an improvement over BCG unless other significant benefits would justify an equivalent level of protection to BCG. If a candidate is a part of a heterologous prime-boost regimen, the combined regimen must be significantly more protective than the individual components. If the prime-boost regimen has BCG as a critical component e.g. boosting BCG in infants, there is an expectation that protection greater than BCG must be demonstrated. At stage A, in order to show an efficacy signal, it is advisable to use a model e.g. mice where there is greater statistical power to show the BCG-boost effect, with tractable group sizes. Appropriate comparators to judge the relative effect of the vaccine might include empty vectors or, if available, best-in-class vaccines that are already in development.
There is no published guidance about the most suitable model to use for specific vaccine candidates but organisations such as TBVI (TBVI services) or the Collaboration for Tuberculosis Vaccine Discovery (CTVD) will provide advice via, for example, product development teams (PDT – TBVI) or the research communities established by CTVD, in particular the NHP community.
At Stage B, the emphasis is on confirming the protective efficacy of the candidate. Confirmation should be shown in a second animal model to demonstrate that the effect is biologically robust. This second model might be a guinea pig model or a mouse model which is more stringent e.g. involves a more virulent M. tuberculosis challenge, or uses a mouse strain which is more susceptible to disease. However, if this is not feasible for the candidate (for example, if a virus is known to be permissive in only one of the standardised animal models available), then independent verification of protection in the same animal model in another laboratory is acceptable. Regardless, it is imperative that the data are robust and that the protective effect can be reproduced. Studies aimed at elucidating the mechanism of action should be considered (e.g., T-cell depletion or passive antibody transfer). Head-to-head evaluation of vaccine candidates in independent laboratories is available via TBVI (TBVI services - standard mouse and guinea pig primary infection models, post-exposure vaccination models in mice and guinea pigs, mouse models involving challenge with clinical strains and NIH (NIAID/NIH pre clinical models - standard mouse and guinea pig models, therapeutic vaccination mouse models), where standardised and well-characterised models and study designs are employed to ensure that comparison of data is feasible and head-to-head testing is encouraged. Immunogenicity and efficacy studies conducted in Stages A and B should be used to support the design and execution of experiments in the advanced model(s) in Stage C, including the identification of a primary endpoint upon which Stage C studies will be powered.
Given the close similarity of non-human primates (NHP) to humans, data generated in NHP models are the most likely to predict clinical immunogenicity and efficacy and therefore efficacy data generated in an NHP model will provide the greatest confidence for funders / stakeholders to invest in further development. Studies conducted in NHP allow parallels to be drawn directly with data generated in humans which (i) inform the design of early clinical studies for example by providing information on vaccine doses/ routes/ regimen/ type of immunological read–out and (ii) provide added significance to the clinical immune responses because they can be correlated to a relevant signal of protection. Ethical and financial considerations restrict the testing of candidates in NHP to those which have the greatest potential, hence the system of screening in more tractable animal models in stages A and B is designed to identify the most promising of vaccines. The global capacity for candidate evaluation in NHP has been increased and there are now standardised methodologies for measuring vaccine efficacy across different NHP sites (Laddy et al., 2018). Thus, there is an expectation and encouragement that protection in NHPs should be demonstrated in order to pass Gate C. However, there may be situations where a different, advanced animal model would be equally appropriate. Examples where alternative models may be pursued include using cattle field trials to demonstrate prevention of naturally acquired infection or using prevention of relapse models. Animal models for TB vaccines are still being developed and refined and important developments (such as the establishment of a controlled human challenge system) may have an impact upon the selection of the model used but at stage C, it is imperative that the model used must be justifiable and highly relevant to the nature and TPP of the vaccine and must be robust and reproducible so studies in non-standard models which are under development is discouraged. A primary endpoint should be selected and documented before initiation of the study, and the study adequately powered to produce clear, statistically significant results. Without setting specific levels of protection against myriad endpoints that would be required to support advancement of a candidate into the clinic, there is an expectation that protection be statistically significant. Investigators are encouraged to confer with the CTVD’s Nonhuman Primate Research Community for advice on the design and execution of NHP challenge studies.