The short version: Concerns regarding the long-term effects of COVID-19 vaccines are poorly founded and in the context of a pandemic should not be a deterrent to getting one. Clinical trials will catch all but the most rare adverse events, which will be caught by postmarketing surveillance, and given how rare these events are, they do not represent good reasons to avoid a vaccine. Alternatively, there is only one vaccine- the varicella vaccine- which can very rarely cause an adverse reaction years later (shingles). This is because the vaccine contains a weakened, but nonetheless active, virus that like the pathogenic kinds, can cause a persistent infection and reactivate. None of the vaccines for COVID-19 that have advanced to phase III of the clinical trials process (at the time of writing this) are live, and there is no evidence of persistent infection from SARS-CoV-2. These concerns are not well-founded.

Lately I’ve observed a lot of worry about long-term effects from COVID-19 vaccine candidates, which I fully understand. Almost. This seems to come from the perception that corners are being cut in the clinical trial process which is being dramatically accelerated compared with how long it usually takes for vaccines to be made.

The concern of long-term adverse outcomes relating to the receipt of a SARS-CoV-2/COVID-19 vaccine, however, are almost entirely baseless beyond the need for empirical verification that they do not happen.

Firstly, we need to be clear about our terminology here. When we say long-term effect from a vaccine, we can mean one of two things:

1. An effect of the vaccine that arises soon after its receipt and has long-term consequences on the vaccinee.

2. An effect of the vaccine that arises years after the initial vaccination (I would prefer to call this a latent adverse event following immunization (LAEFI) but this is what many are calling a long-term effect so I’ll be consistent with their language).

The former would be caught by the clinical trials process unless it were very rare (Phase III trials include tens of thousands of people at least and compared the incidence of the infection in the vaccinated vs. unvaccinated groups and their outcomes from infection). Examples of such phenomena include vaccine-associated paralytic polio (VAPP) which occurs very, very rarely from the oral polio vaccine (OPV) and not at all from inactivated polio vaccine (IPV) (nOPV2 is a polio vaccine that has additional mutations to prevent the reversion to neurovirulence that causes VAPP) or intussusception from the rotavirus vaccines (also very rare, and largely determined by when the vaccine is given which is why it is given on the schedule before the risk for intussusception peaks). Basically this sort of thing is exactly what clinical trials are examining right now, and as long as the event isn’t very, very rare like the previously mentioned, there’s no reason that these would be missed by clinical trials (in biostatistics there is a rule of three which when extrapolated to clinical trials means that for a trial of n people, if an event does not occur, the 95% confidence interval for the event in the population is (0, 3/n) e.g. for a trial of 30,000 people, if a given event does not occur then the 95% confidence interval for the population is (0, 3/30,000) which means we can be 95% confident that the true risk of the event occurring is between 0 and 1 in 10,000 times within the population; this is only an approximation and not rigorous but can help you determine appropriate sample sizes in a study).

The pertinent question to ask when considering the risks from these kinds of adverse consequences is how they compare to the risks of no vaccine i.e. the risks of getting the disease. VAPP is far rarer than paralytic poliomyelitis from poliovirus infection. The numbers for intussusception are less clear, but ACIP deemed that the increased risk was too low to warrant withdrawal of the vaccine (though history of intussusception is a contraindication for this vaccine). So regarding this kind of long term-effect- it’s possible that this will be missed if it is very, very rare and will then show up in post-marketing surveillance. But the question then is “How relevant is that to your decision-making calculus if it’s truly that rare?” Consider the ever-expanding litany of long-term consequences we are tracing to COVID-19 infection- consequences which are not uncommon. My personal view is if this is such a risk from a vaccine for COVID-19, it would likely still be worth getting because the risks posed by infection will still be far greater.

Then we run into the latter kind of long-term effects: the ones that stay dormant and then hit you months or years later. This is basically not a thing. There is one major exception though which is worth going into because it really helps to inform why this is.

Live attenuated varicella vaccines do contain actual varicella zoster virus (Oka strain), which can establish infection in hosts and undergo latency like other herpesviruses including the unattenuated virus. As a result they can very rarely cause shingles or meningitis years after the receipt of the vaccine. This is a consequence of 2 things: the use of a live vaccine, and the ability of herpesviruses like varicella to undergo latency:

I should take a moment however to remind readers that these events are FAR more common from varicella infection than from the vaccine, and the vaccine is absolutely worth getting for anyone not contraindicated from it. We know the vaccine can do this because there is vaccine strain varicella (Oka strain) which can be isolated from patients experiencing these diseases which is genetically distinct from wild type strains. Furthermore, the shingles that results from vaccine strains as a rule is much milder than that from wild type varicella.

Most of the vaccine candidates for COVID-19 gaining attention right now are either vectored vaccines which essentially give a nonpathogenic or minimally pathogenic viral vector a protein (typically the spike protein) of the SARS-CoV-2 virus or use RNA encoding the spike protein.

RNA vaccines have caused a wave of trepidation in some and I attempted to assuage those concerns in this post but the short of it is that owing to their incredibly brief lifetime within the cell, the possibility of any long-term effects of the former or latter kind are basically impossible. Of course, clinical trial data will be needed to confirm this and I am by no means stating that it should be skipped.

Vectored vaccines are live, similar to the varicella vaccine discussed above, but they have a few key differences. One concern that always exists with the use of live attenuated vaccines is that they will undergo a reversion to virulence- they will be introduced into a host and many generations of mutation later will regain their virulent abilities and cause disease. This is the basis for OPV causing VAPP and MIBE from measles vaccines. The use of a vector completely circumvents this by using a different virus which does not pose these risks. But we can actually take this a step further and the ChAdOx candidate did. The ChAdOx vaccine candidate uses a replication-deficient chimp adenovirus vector which expresses SARS-CoV-2’s spike protein. Replication-deficient refers to the fact that the vector, once introduced into the host, cannot replicate. The vector contains mutations in its E1 protein that make this impossible which is excellent because it means that the vaccine can even be given to immunocompromised people safely. Because the vector cannot replicate, there is no potential for it to return years later with a vengeance like the varicella vaccine, and with just the spike protein even if it could, there’s not all that much this could do unless there are heretofore undiscovered pathogenic roles of the spike protein beyond facilitating entry into the cell (which seems unlikely to me given that the virus has a suite of nonstructural proteins that wreak havoc on the immune system’s responses).

I should mention however that latency in the herpesvirus sense is not the only way a persistent infection can occur. It would be extremely unusual for latency to occur with a coronavirus, but persistent infections have been documented in the past with mouse hepatitis virus, a coronavirus of great experimental importance.

In general, persistence occurs because the immune system is not able to completely clear the virus from infected cells for any number of reasons. For example, some viruses can hide in the central nervous system, an immunoprivileged (sort of) site, and periodically cause infection. Alternatively, some viruses can become literally part of the host genome in the cells they infect, like HIV. Mouse (murine) hepatitis virus (MHV) is a coronavirus that has shown the ability to cause persistent infection by causing immunosuppression.

Because none of this would apply to a vaccine for COVID-19 (given the platforms being used; a live attenuated vaccine is very unlikely because of how well coronaviruses undergo recombination- a reversion to virulence is likely to occur with this strategy) the risk for latent long-term effects is so small as to be meaningless. There’s also the question of long-term effects from COVID-19 infection, and that is definitely not meaningless, with nearly 90% of patients who recovered from acute COVID-19 reporting at least 1 symptom months after the initial infection. The most common long-term effect of vaccines is obvious: immunity and protection against vaccine-preventable diseases.

Also perhaps the most important thing: FDA guidance recommends follow up for as long as possible but at least 1–2 years after vaccination:

which isn’t all that different from that of other vaccines in non-pandemic times:

The development process is going to be a lot faster because a lot of the regulatory fat is being trimmed and the entire world is working on the production of an effective COVID-19 vaccine. But that doesn’t have to mean that the end product will be less safe.

Addendum:

The amazing and indefatiguable Dr. Kizzmekia Corbett on reading this piece raised a very good point that I believe is worth addressing regarding the potential for antibody-dependent enhancement (ADE) of disease. ADE is as the name suggests, a process by which an infectious disease worsens because of an antibody-mediated process of some kind. The best-characterized example is with Dengue. Dengue viruses have several serotypes (strains) and antibodies that carry over from one strain might be recalled in infection by another strain. However, though these antibodies are able to bind Dengue virions, they are unable to actually neutralize them, which is to say, prevent entry into cells as one would hope (the ability to neutralize viruses or toxins is among the most important thing that antibodies do, though not the only thing). The virus can then spread to additional cells that contain receptors for antibodies called Fc receptors, which is thought to worsen disease. The exact mechanisms by which disease gets worse however, is not very well understood. Unfortunately, ADE is not a phenomenon limited to Dengue, and has been a major point of concern in the creation of SARS-CoV-2 based on prior history of attempting a vaccine against SARS-CoV (classic SARS).

There are a few points worth establishing however regarding the potential for ADE.

1. ADE is rare and limited to only a few infections.

2. ADE in vitro does not equate to ADE in vivo.

3. ADE in an animal model does not, unfortunately, predict ADE in humans.

4. ADE does not necessarily mean a dead end for vaccine prospects. Dengue, the most famous offender, will soon have a vaccine which does not cause ADE.

5. ADE does not occur at all levels of antibodies but rather within a specific window, typically as they start to wane.

That last point is the major concern, as if protection from a vaccine candidate or infection starts to wane, theoretically one could observe ADE occurring with reinfections, hence the need for robust immunity. We have observed some cases of true reinfection of SARS-CoV-2, with unfortunately no clear trend of how the second infection compares to the first, most probably reflecting not that the phenomenon is so rare but that we are not sequencing sufficiently many viral samples from people to notice it. With respect to a vaccine candidate, the only way to know how robust an immune response elicited by a vaccine is to wait long enough for it to actually wane. The good news is that there’s an easy fix to the matter- a booster dose. Further good news is that thus far, we have not seen good evidence right now of SARS-CoV-2/COVID-19 ADE, and given the fairly ubiquitous use of convalescent plasma, we might have expected to see something at this point. Vigilance is of course still warranted, but my personal view is that this is unlikely at this point.