Correlates of Protection: Vaccine Development under FDA Animal Rule

One key challenge when developing vaccines for human use is that it is often infeasible or unethical to assess the efficacy of the candidate vaccine directly in humans. As we’ve discussed previously, diseases which are severe, rare, or both are difficult to accommodate in human testing. This poses a quandary: these diseases are often the ones against which we most want to vaccinate against, but also the ones for which vaccine development is the most obstructed. Thankfully, pathways exist for developing and licensing vaccines without the need for efficacy testing in humans, such as the FDA Animal Rule. Here, we’re going to examine the FDA Animal Rule, how it applies to vaccine development, and discuss a key component for demonstrating efficacy under the rule: correlates of protection.

Recognising that there are some limited scenarios in which it is necessary to license a novel therapeutic without human efficacy data, the FDA outline an approval process which uses efficacy data from animal studies alone.

This is known as the Animal Rule: the specific regulatory requirements for vaccines are detailed in 21 CFR 601, and guidance outlined in the document Product Development Under the Animal Rule: Guidance for Industry.

The Animal Rule does not only apply to vaccines, but in many scenarios where testing a therapeutic on human subjects is not possible. This includes prophylactics for nerve agent poisoning, mitigation of cyanide poisoning and radiation sickness, as well as treatments for Botulism, Anthrax, Plague, and Smallpox. To test for efficacy on human subjects in any of these scenarios would expose volunteers to life-threatening outcomes, meaning conducting such trials would be impossible on ethical grounds.

The FDA guidance outlines four criteria all of which must be met for a product to be considered eligible to be licensed under the Animal Rule:

1. There is a reasonably well-understood pathophysiological mechanism of the toxicity of the substance and its prevention or substantial reduction by the product
2. The effect is demonstrated in more than one animal species expected to react with a response predictive for humans, unless the effect is demonstrated in a single animal species that represents a sufficiently well-characterized animal model for predicting the response in humans
3. The animal study endpoint is clearly related to the desired benefit in humans, generally the enhancement of survival or prevention of major morbidity
4. The data or information on the kinetics and pharmacodynamics of the product or other relevant data or information, in animals and humans, allows selection of an effective dose in humans.

The guidance states “If all of these criteria are met, it is reasonable to expect the effectiveness of the drug in animals to be a reliable indicator of its effectiveness in humans”.

The guidance also outlines requirements specifically for vaccines. This includes the establishment of a relationship between the vaccine dose and the desired immune response. The key component, however, is developing an approach for bridging between responses in animals and in humans. This often involves careful selection of one or more biomarkers known as correlates of protection.

A Correlate of Protection (CoP) is an immune marker whose concentration in an organism shows statistical correlation with protection against a particular disease. While other types of drug development also use biomarkers or surrogate endpoints as substitutes for clinical endpoints, such as survival or disease progression, CoPs are specific to vaccines. That’s because, unlike most medical interventions, vaccines are a prophylactic treatment. In other settings, we are interested in measuring the effect of a treatment using biomarkers. For vaccines, we are using CoPs to predict the effectiveness of the vaccine at preventing the disease in question.

There are a couple of ways CoPs can be classified:

Mechanistic: A mechanistic CoP (mCoP) is a biomolecule which is directly responsible for the protection with which it is correlated. An example might be a neutralising antibody for a target antigen – the antibody directly attacks the pathogen, so a higher concentration of the antibody in a serum sample indicates a greater protection against that pathogen

Non-mechanistic: A non-mechanistic CoP (nCoP) is a biomolecule which, while correlated with the protection provided by a vaccine, is not itself part of the protective mechanism. A classic example is serum antirotavirus IgA, an antibody which is associated with protection against the childhood gastrointestinal disease rotavirus. Studies have found that antirotavirus IgA in serum does not participate in protection against the rotavirus, but it’s presence can be indicative of an immune response in the intestine which is protective.

Absolute: An absolute CoP is one for which a concentration above a certain threshold is deemed to be indicative of protection. A subject for whom the CoP is below the threshold is unlikely to be protected, while one for whom the CoP is above the threshold is likely to be protected against the target pathogen.

Relative: A relative CoP is one for which a continuous relationship between its concentration and protection is observed. The greater the concentration of the CoP, the more a subject is likely to be protected.

It is important to note that CoPs indeed indicate correlation with protection, and are not a guarantee. The immune system is highly complex, and involves several interacting components which act together to protect against pathogens. Further, individuals respond in different ways to vaccines. All that means is that no CoP is an all or nothing indicator of protection, but they can provide a useful and often necessary tool to help determine a subject’s state of protection.

Countdown to Immunity: Testing Vaccine Onset of Protection

In Vitro vs In Vivo Potency Assays: Modern Vaccine Batch Release

Understanding Preclinical Vaccine Safety: Regulatory Guidelines Explained

Identification of CoPs is a key goal in many preliminary vaccine trials as they can enable significant acceleration and resource savings in the vaccine testing and licensing process.

An example of this is immunobridging. This is where the efficacy of a vaccine in a target population can be established by comparison with another population where its efficacy is already accepted. One common use of this is to accelerate licensure for an updated formulation or delivery method for an existing vaccine. For example, if an updated formulation of a vaccine generates at least as strong an immune response as the previous formulation – as demonstrated by measuring serum antibody levels, say – then it can typically be taken to have demonstrated acceptable efficacy without the need for a full clinical trial.

Similarly, the immunobridging process can be used to extend the populations in which a vaccine is approved for use beyond those in which the vaccine was originally trialled. Many vaccine trials are conducted in adults for ethical and feasibility reasons. When looking to extend approval into elderly or paediatric groups, it can often be sufficient to demonstrate that a CoP generated by the vaccine is present at levels at least as high as in the populations originally tested. This avoids the need for further efficacy trials, and allows the benefit of the vaccine to be extended to vulnerable groups when those trials are not possible.

Of course, the logical extension of this is to demonstrate efficacy in humans by comparing CoP data with that from non-human trials, such as those which utilise the FDA Animal Rule. As we’ve discussed, this allows human vaccines to be developed for diseases where traditional efficacy trials would be impossible due to low prevalence or the ethical impossibility of exposing human volunteers to potentially deadly diseases.

How does one establish the efficacy of a vaccine using CoPs? In their 2014 paper Estimating vaccine efficacy using animal efficacy data, Yellowlees and Perry explore statistical techniques which allow vaccine efficacy to be estimated from CoP data in animals. In the paper, they define vaccine efficacy as . This is specifically tackling efficacy of preventing the target disease outright – one could also examine efficacy against preventing severe disease or death.

The process outlined in the paper goes as follows:

1. Establish the relationship between the chosen CoP and vaccine efficacy in an animal study. This requires careful study design to ensure that the curve outlining this relationship is well characterised.

Typically, the animal study will follow a challenge design. This involves several study groups which are given differing doses of the vaccine candidate before being challenged with the target disease. CoP concentration is measured for all subjects at a series of time points through the trial period and compared to baseline measurements taken pre-challenge. The probability of protection (i.e. no disease at the end of the trial period) is then plotted against the CoP concentration and a statistical model – typically a logistic model – is fit to the data.

The resulting curve defines the relationship between the CoP concentration and the probability of protection, which is equivalent to the vaccine efficacy. If we assume that all unvaccinated subjects become sick – – the vaccine efficacy simplifies to . Since an animal can only be either diseased or undiseased, this means that the vaccine efficacy is the probability of protection.

2. Once the protection curve for the chosen CoP has been established, the process of determining the efficacy in humans can begin. To do this, a group of human subjects is dosed according to the anticipated dosing regimen for the vaccine. Serum samples are taken from the subjects following a specified period after vaccination and the concentration of the CoP found.

The immune response will be stronger for some subjects and weaker for others, meaning there will be a range of CoP concentrations measured – indeed, this is the very reason why a range of different doses are used. These measurements are expected to follow a log-normal distribution.

3. Once the CoP concentrations have been collected, these can then be translated into protection probabilities using the protection curve. Specifically, one can read up to the curve from the CoP concentration on the x-axis, and then across to the associated protection probability on the y-axis.

This is done for all the collected CoP concentrations, and the mean protection probability is found. Given the assumptions outlined previously, this is the estimated efficacy of the vaccines in humans.

A wide variety of vaccines have CoPs acknowledged by regulators. We’ve listed some examples below:

The search for a CoP against the COVID-19 demonstrated how rapidly such information can be established in modern vaccine development. By 2021, evidence from phase III clinical trials, alongside both human and non-human challenge trials, had identified that neutralising antibody titre was a key CoP for COVID-19 vaccines. This was acknowledged by regulators as an acceptable CoP for vaccine updates and delivery expansions without further efficacy trials, and used in support of authorisation for boosters against the Omicron variant. Antibody levels provide a relative CoP for COVID-19: there is not definitive antibody titre above which strong protection is observed. Instead, a protection curve exists which outlines how protection against the disease increases with measured antibody titre.

Influenza is an example of a disease where an absolute CoP is well established. Specifically, a 1:40 serum titre in a haemagglutinin-inhibition (HAI) assay is indicative of 50% protection against influenza illness in young adults with prior exposure to influenza. While this does not extend to all groups who receive the vaccine – protection is indicated by higher titres in older adults or young children without prior exposure – this threshold is used as a benchmark for developing and updating ‘flu vaccines. This is particularly useful for ‘flu vaccines as the require annual updates to target the strains expected to circulate each winter. Annual efficacy trials would be prohibitively expensive, meaning approval based on CoP data is vital for ensuring the availability of the seasonal ‘flu vaccine.

As mentioned previously, the rotavirus vaccine generates an example of a non-mechanistic CoP. The vaccine is thought to work by generating mucosal immunity in the gut, where intestinal IgA prevents viral replication. As mucosal IgA levels are difficult to measure, serum IgA levels have been identified as an indirect measurement of antibody levels in the intestines. Research has found that a serum anti-rotavirus antibody concentration of greater than 20 units per millilitre is correlated with reduced rates of rotavirus disease. This finding, however, has not been consistent in all groups studied: studies in children in low-income countries showed little correlation between anti-rotavirus antibody seropositivity and disease prevention. While serum IgA can prove a useful CoP for monitoring batch consistency and testing new formulations, therefore, it is not regarded as a full surrogate for efficacy across all populations, meaning full phase III efficacy trials are required for licensing rotavirus vaccines.

Vaccines for Human Papillomavirus (HPV) are an interesting case of an effectively qualitative CoP. The vaccines generate very high levels of neutralising antibodies to the extent that almost everybody vaccinated in efficacy trials showed almost complete protection and remained seropositive. This meant that no minimum antibody level for protection could be determined. What was apparent, however, was that the presence of any vaccine-induced anti-HPV antibodies is protective against the disease, while the zero-antibody state is not. This data has been used to support extending HPV vaccination drives to adolescents, and is particularly useful in the development of novel vaccination approaches due to the long timescales for clinical outcomes (i.e. cancers prevented) to become apparent.

CoPs function as an immunological dictionary that enables translation of vaccine efficacy between different populations. The FDA animal rule is built on this foundational concept: once a reliable relationship between a CoP and protection is established in animal models, human vaccine responses can be assessed against this benchmark to estimate efficacy. By replacing clinical outcome data with inference based on immunological data, decisions about vaccine efficacy can be made when trials in humans are unethical or infeasible. This enables vaccine investigations for rare or severe diseases for which such treatments may not have otherwise been developed.

As we have seen, however, the utility of CoPs extends well beyond FDA Animal Rule contexts. The same principles are used to accelerate vaccine development and extend approvals across populations. For example, leveraging CoPs can enable a vaccine shown to be effective in one age group to be licensed in another – such as extending a vaccine from adults to children – by showing similar or superior immune responses. It’s also used to compare new formulations or delivery methods against existing ones, significantly reducing the need for large-scale efficacy trials.

Together, CoPs and immunobridging form a strategic framework that allows developers to infer protection across species, formulations, and populations, allowing faster and more efficient vaccine development. Within the FDA Animal Rule, they’re not just useful tools: they’re regulatory lynchpins allowing animal data to be used for human benefit.