Countdown to Immunity: Testing Vaccine Onset of Protection

When we discuss the performance of a vaccine, properties such as its immunogenicity, adverse event frequency, and any correlates of protection are usually among the first to be mentioned. A property which is considered less often is when a patient can expect to be protected after they have received the vaccine. This is known as the onset of protection of the vaccine. A detailed understanding of the onset of protection is vital to ensure that an accurate picture of the efficacy of the vaccine is established and that the deployment strategy for the vaccine can be properly designed. Here, we’ll examine why onset of protection warrants inclusion in a vaccine testing programme, describe how a trial might be designed to perform that testing, and discuss statistical methods which can be used to analyse the data collected.

One might be forgiven for forgetting that the immunity provided by a vaccine is not present immediately after the dose is administered. Rather, it merely begins a physiological process which can take days or even weeks to reach its conclusion.

Vaccines are designed to stimulate the adaptive immune system, which acts as the immune system’s “memory”, allowing the body to recognise and swiftly target pathogens which it has already encountered. This system relies on B memory cells and T effector cells, which often require several days to proliferate and mature after exposure to a novel antigen. This means that the protection provided by the adaptive immune system also takes time to build.

This delay between administration of a vaccine and full immunity can be crucial when the vaccine is deployed in the field, particularly in outbreak scenarios. During the COVID-19 pandemic, for example, public health officials were quick to emphasise the two-week period between the final dose of a COVID-19 vaccine and a recipient acquiring full immunity for fear that those recipients might take undue risks while still lacking protection. Similarly, one might choose to prioritise the vaccination of healthcare workers in an outbreak to ensure a healthcare system continues to function. In this case, it is vital to know how long to wait before one can safely redeploy those vaccinated back onto the front line.

It is also important to account for onset of protection when conducting clinical trials which test the efficacy of the vaccine. These often compare outcomes, which might include contracting the disease, hospitalisation, and/or death, among vaccinated populations to those in unvaccinated populations. If one were to naïvely count these outcomes from the date of administration of the vaccine, there is a chance that some of those observed will have occurred before the onset of protection – that is, before the immunity provided by the vaccine has been fully realised. This can dilute the measured efficacy of the vaccine in the trial, making the vaccine appear less effective than it actually is.

There are several approaches which can be used to determine the onset of protection of a vaccine. As with many other components of the vaccine development and testing process, the choice of which to use is influenced by several factors, not least the severity and lethality of the disease in question. Let’s take two diseases as examples, and examine how studies finding the onset of protection of vaccines for those diseases might be performed.

Marburg Virus Disease (often known as simply Marburg) is a very severe and highly deadly disease caused by a virus in the same family as Ebola. The disease was first reported in outbreaks caused by laboratory work using African primates in Germany, but the disease is more commonly observed in sub-Saharan Africa, where its case fatality rate has been observed to be as high as 88% in some outbreaks. While the disease is highly contagious, outbreaks of Marburg are thankfully rare.

A population study for Marburg would be difficult to conduct due to the low prevalence of the virus, and its high severity means that human trials involving exposure to the virus would be unethical. A challenge trial in an animal model, therefore, would be an appropriate choice for a trial involving Marburg virus. Specifically, non-human primates (NHPs) are often used as their immune systems are similar to those in humans, meaning it is likely that results in NHPs will translate to humans.

To determine the onset of protection in such a study, groups of subjects are vaccinated on different days pre-challenge. For example, one group of subjects might be vaccinated four weeks before the challenge day, the next three weeks, and so on. One would expect that subjects for whom the onset of protection fall after the challenge day will have different outcomes to those for whom then onset of protection falls before the challenge day. In some studies,For severe diseases, however, it can be assumed that the challenge would be fatal for all subjects, meaning the control group may be omitted.

After challenge with the Marburg virus, the subjects are observed, and the progression of the disease is assessed. A simple method to infer the onset of protection of the vaccine is to examine the number of subjects who survive to a certain day post-challenge. This post-challenge period ought to be sufficiently long that one would expect that few of a control group will survive to that day. The survival rate in each vaccinated group can then be compared to the survival rate in a control group. The vaccination date of the latest-vaccinated treatment group which had a statistically significantly greater survival rate than a control group (whether included or hypothetical) informs us of the onset of protection of the vaccine.

For example, imagine we performed a study which returned the following results. Note that study day 0 is the challenge day.

As expected, none of the control group survived to the end of the test period. Similarly, the treatment group vaccinated one day before the challenge did not see any benefit from the vaccine, indicating that there had not been enough time for protection to develop. There is potentially evidence of weak protection among group 3 vaccinated eight days before challenge, but we see that the earliest-vaccinated group with significant protection is group 2. We can conclude, therefore, that the time to onset of protection of the vaccine is at most 15 days based on this evidence.

Imagine that a group had developed a new formulation of an existing chickenpox vaccine with the intention of decreasing its time to onset of protection without causing an increase of side effects. Assuming that the new formulation has been proven safe and effective, how might we demonstrate that it’s time to onset of protection is decreased?

We could perform a similar experiment as for the Marburg virus – vaccinate groups of subjects at different times, expose them to the virus, and see how outcomes differ. Unlike Marburg, however, chickenpox is a widespread disease which rarely has severe consequences for those who contract the disease. This means that population studies of human subjects of the performance of a chickenpox vaccine are both feasible and ethical.

So, we might instead offer the parents of children who are receiving a routine chickenpox vaccination the opportunity to participate in testing the new formulation. The children of those who agree would be randomised into either the treatment group and receive the new formulation, or into the control group and receive the old formulation. These children would then go back into the world to be exposed to chickenpox, which some would then contract. The parents would then report when they observed onset of chickenpox symptoms.

One can determine the onset of protection in this trial design using survival analysis. This utilises continuous time-to-event data with the potential to give a more precise measurement of the onset of protection than using a binary alive/dead or symptomatic/healthy measurement as used in the Marburg challenge trial discussed previously. Here, the event is initial detection of chickenpox symptoms, and the time to that event is the time since vaccination.

The additional precision comes with the trade-off of requiring a larger sample size in the study. The data from a population study is typically noisier than that from a challenge study due to a range of real-world factors, such as reporting errors or differences in the extent to which subjects are exposed to the chickenpox virus.

A common visualisation used when performing a survival analysis is the Kaplan-Meier plot. This plots the proportion of subjects in each group which remain chickenpox-free against time. Over time, this proportion decreases as subjects contract chickenpox, meaning we see a decreasing curve on a plot.

If a placebo group had been included, then results from this group would be included too. However, it is unlikely that a true placebo would be included in a trial which includes children, even for a generally harmless disease such as chickenpox. Instead, one might use data simulated from historical data on chickenpox spread in unvaccinated children.

Early in the trial period, the curves for all groups are likely to be very similar as neither formulation will have reached its onset of protection. For longer times, however, one would expect that the curves for each formulation would flatten off as the vaccine began to take effect. Specifically, the proportion of each treatment group who have contracted chickenpox will decrease at a lower rate than that for a group who had received a placebo.

The onset of protection can be determined from a Kaplan-Meier plot by finding where the divergence between the curves for the treatment groups and the placebo group occurs. In our case, we’d be looking for that divergence to occur earlier for the new formulation of the vaccine than for the old formulation. This would indicate that the new vaccine formulation is having an effect earlier than the old formulation, meaning it’s onset of protection occurs sooner after delivery of the vaccine.

Vaccine Development: Correlates of Protection Explained

Choosing a Path: Natural History Studies for Vaccine Development

Study Populations: Clinical Trial Design

Alongside measures of survival of trial subjects, it can be important to determine the onset of protection of a vaccine in terms of correlates of protection. These are biomarkers which are present at certain levels when a subject is protected by a vaccine. For example, the presence of antibodies for a target pathogen might indicate that a subject has gained protection against that pathogen.

Examination of correlates of protection are particularly important for vaccines targeting severe diseases – such as Marburg – for which human efficacy trials are unlikely to be possible. In these cases, an understanding of how can by approximated with measurable biomarkers is crucial as it means that the onset of protection in humans can be detected without exposing volunteers to potentially deadly diseases.

To measure the biomarkers associated with protection in a challenge trial, a baseline measurement of one or more biomarker is taken form a serum sample collected pre-vaccination. Then, a further sample is collected before challenge and the same measurement made.

This is repeated for all or a subset of the subjects in the trial. These are then challenged with the target pathogen and progress through the trial period, with their outcomes recorded. A regression model is then fit which associates the probability of a subject’s survival at the end of the trial period with the biomarker measurements from immediately pre-challenge (but post-vaccination). From this model, the concentration of a biomarker which correlates with a certain survival percentage can be found.  For example, we might want to know what concentration of a certain antibody correlates with a 50% survival rate at the end of the trial – . When we plot the survival rate against the biomarker concentration and fit the model, we can read off the IC₅₀ from the fitted model. This can be done for any given percentage, the concentration which correlates with an 80% survival rate, or IC₈₀, is another commonly measured quantity.

These measurements tell us the concentration of key biomarkers associated with a subject being protected. That means that we can determine the onset of protection from the time after vaccine delivery when the concentration of those biomarkers reach the critical levels. If the onset of protection cannot be directly measured by exposing human subjects to a disease, then this can be a useful method of finding how long a vaccine takes to be effective.

Understanding the onset of protection is far more than an academic curiosity. It’s a pivotal factor shaping real-world vaccine deployment strategies, influencing everything from outbreak responses to vaccination schedules. By accurately determining how soon after vaccination protection arises, we help ensure that individuals and communities are not unknowingly left vulnerable in critical periods. Survival analyses and carefully structured clinical or challenge trials are powerful tools for identifying precisely when a vaccine starts working, while correlates of protection offer valuable biomarkers, especially when direct human trials aren’t feasible. Ultimately, these methods allow vaccine developers, public health officials, and medical professionals to design safer and more effective immunisation programmes.