COVID-19: Does a 4th Vaccine Dose Prevent Infection?
by Raywat Deonandan, PhD
Epidemiologist & Associate Professor
University of Ottawa
(I add my credentials to these COVID-19 blog posts in case they get shared. I want readers to know that my opinion is supposedly an educated and informed one)
Earlier this year, I wrote a post explaining how prior to the rise of the Omicron variant, COVID-19 mRNA vaccines were actually not bad at preventing both infection and transmission of the disease. Hence the strong evidence base for policy measures like vaccine mandates at work and vaccine passports at places like restaurants.
Omicron changed all that to a high degree by greatly reducing the vaccines’ ability to curtail infection and transmission, but not their ability to provide strong protection against the worst measurable COVID outcomes, like hospitalization and death.
That analysis was based on the assumption that most people had received only two doses of the vaccine. But I ended that post by citing some evidence showing that a third dose restored much of the jab’s effectiveness at preventing infection.
As I write this, that original Omicron strain has been supplanted by more transmissible subvariants. Currently, the BA.4 and BA.5 subvariants account for the majority of cases in North America. Both subvariants show an ability to overcome immunity garnered from both vaccination and prior infection from Omicron, suggesting that even three doses are no longer effective in preventing infection and transmission.
We are now in the era of 4th doses. So I thought it would be helpful to revisit the evidence around whether four doses can help prevent infection.
Spoiler alert: the evidence, as it is, suggests that four doses definitely increases our ability to avoid infection altogether. But there is debate about exactly how much that increase is, and how long it will last.
Infection vs Symptomatic Disease
First, a reminder. All clinical trials of COVID vaccine effectiveness have a hierarchy of outcomes:
infection -> symptomatic disease -> hospitalization -> death
Vaccine efficacy tends to increase along that spectrum from left to right. And measuring efficacy is easier, going left to right. Keep in mind that efficacy is essentially the comparison between the outcome as seen among vaccinated people vs that same outcome as seen among unvaccinated people.
Measuring death and hospitalization among trial participants is straightforward. Measuring symptomatic disease is also manageable, since it entails waiting for someone to report known COVID symptoms, then giving them a PCR test to confirm infection status.
But measuring actual infection among trial participants is practically impossible. Think about it. There are so many instances of asymptomatic infection, or instances of people not recognizing symptoms when they arise, that relying on participants to develop and acknowledge symptoms and thence to seek testing will necessarily miss a great many cases of infection. That’s why trials stress that VE (vaccine efficacy) is measured with respect to symptomatic disease, and not to infection.
To properly measure VE against infection in such a scenario, participants would have to be PCR tested pretty much daily for weeks and months. When thousands of people are enrolled, that’s much too impractical and expensive. It does happen on occasion, though.
More commonly, researchers use a “test-negative design” (TND) to estimate VE against infection. A TND is like a case-control study (if you know what that is), in which the odds of being vaccinated among people known to be infected is compared with the same odds among people who are not infected, matching for important confounders like age. It’s a lot more complicated than I make it sound and usually requires a robust surveillance system and testing infrastructure to make it meaningful.
TNDs suffer from a variety of biases (as do all designs), including the possibility that health-seeking behaviours might differ between infected (i.e., cases) and non-infected people (i.e., controls), and whether the vaccine is “leaky“, as some have described the COVID vaccines. In other words, all of these data are subject to criticism, but are nevertheless useful to some degree.
As a result, most of the good trials looking at vaccine effectiveness have as their first measurable outcome of interest, symptomatic disease, which, as I will repeat, is not the same as infection.
To date, one of the best studies of the effectiveness of 4th doses was done by Magen et al, who looked at 182 thousand matched pairs of people over 60 in Israel, comparing those who received 4 doses with those who only received 3 doses (the controls). They found a 55% 4-dose VE (vaccine efficacy) against symptomatic disease, as compared to 3 doses.
In other words, a 4th dose reduced the incidence of symptomatic disease by 55%, or a 55% reduction in the number of cases we would have expected had they not been given a 4th dose.
Note that though the VE against symptomatic disease (again, comparing 4th doses to 3 doses) was 55%, their computed VE against hospitalization was 68%, and that against death was 74%. As noted, the effectiveness improves as we go up the chain of severity. So please don’t just focus on infection or symptomatic disease when discussing vaccine efficacy.
Infection vs Transmission
Second, it’s important to draw a distinction between VE against infection and VE against transmission, as the two commonly get conflated.
VE against infection, as noted above, can be computed by comparing the incidence rate of infection amongst vaccinated people with the rate of infection amongst unvaccinated people.
But transmission is another beast entirely. To best get a handle on transmission likelihood, we look at something called a secondary attack rate (SAR), which is the incidence of infection among the contacts of vaccinated infected people. To that end, it’s often not that useful to compare the SAR for infected vaccinated people against the SAR for infected unvaccinated people. What is often more useful is to simply look at the rate of SARs among the vaccinated.
Let me rephrase that. It’s not that it’s not useful to compare SARs among vaccinated and unvaccinated. It’s just that SARs are more useful when comparing circumstances for transmission, and not necessarily who is transmitting. For example, SARs can vary with the size and behaviour of the household, the amount of virus circulating in the community, as well as the vaccination status of people in the household. So it’s a messy measure.
As result, there have been precious few good studies on VE against transmission for COVID vaccines at any point throughout the pandemic, let alone this time of new doses and multiple new variants.
All of this is why in this post, I will not be touching the VE against transmission, but will instead focus solely on VE against infection. However, if you don’t get infected in the first place, then you can’t transmit it to anyone else. So taking care of infection pretty much takes care of transmission. (But it’s also true that avoiding symptomatic disease also decreases your chances of passing it on to others, since symptomatic people are more likely to project their goodies onto your face.)
The Studies That I Found
Magen et al
I cited Magen et al above. Their design was elegant, as it matched subjects according to a series of variables, including age, sex, and chronic conditions. This reduced the influence of confounding factors. However, their study was restricted to just people 60 and above, was done when the first Omicron variant (B.1.1.529) –not BA.5– was dominant, and enrolled only people who had had their third dose at least 4 months earlier.
Despite my suggestion above that frequent testing would be too onerous, that appears to be what they did with their subjects. So in fact they were able to compute a reliable VE against infection, not just against symptomatic disease.
They found that, when assessed 7-10 days after the 4th dose, effectiveness against PCR-confirmed infection was 45%, meaning that those who got a 4th dose were 45% less likely to get infected than were those who only had 3 doses.
That might sound small, but that ain’t nothin’ to sneeze at. Remember, this is relative to 3 doses, not relative to being unvaccinated at all. The VE against infection versus complete non-vaccination will be at least as large, probably larger.
Chariyalertsak et al
Another (non peer-reviewed) study, that of Chariyalertsak et al, looked primarily at the AstraZeneca vaccine given as a 4th dose, but in a heterologous configuration, meaning that at least one of the first three doses was not AstraZeneca. They looked a large surveillance database of adults over 18 in Thailand. Now, because this study used surveillance data, it was not done in a clinical trial environment, meaning that subjects weren’t followed and tested until an infection arose. Rather, the nation’s COVID reporting system maintained records of people coming forward for testing, some of whom were asymptomatic. As a result, I am certain many asymptomatic cases were not captured.
Therefore their reported VE is likely an overestimate.
Nevertheless, they found a VE of 31% for three doses against Omicron, and a VE of 71-73% for four doses.
Grewal, et al
This study of Ontario Long Term Care (LTC) residents of course only looked at people over 60. They looked at about 14,000 people who tested positive and compared them to over 200,000 controls who did not.
A VE score of 4th dose protection was computed to be 19%, compared against those with only three doses, at at least 7 days post jab.
Again, do remember that this is relative to 3rd doses. VE relative to no vaccination at all will be higher.
There are a smattering of other studies. But the ones I’ve found are either very small or rely mostly on antibody titers and not actual real-world patient data, so are excluded from this summary. Your mileage may vary.
The evidence, what there is of it, shows that a fourth dose of a COVID vaccine increases the ability of the vaccine to prevent actual infection, not just symptomatic disease. The extent of that increase is variable. And –critically– it is completely unknown how long that increased protection will last. The Israeli data shows that the the two incidence curves (4th dose vs 3rd dose) continue to diverge at least a month after that 4th dose. So it’s likely that that increased protection will last at least a few months.
Is that a sufficient advantage to make vaccination a requirement for your institution or social gathering? That’s a question that must be answered by each such institution, factoring in its values, vulnerability of its population, and presence of other mitigating factors, such as mask mandates and ventilation improvements.
The European Medicines Agency has a good summary:
“Evidence on the effects of a fourth dose comes largely from Israel, where data indicate that a second booster given at least 4 months after first booster restores antibody levels without raising any new safety concerns. Data also suggest that a second booster provides additional protection against severe disease, although the duration of the benefits is not yet known and the evidence is still limited.”
So there you go. I got my 4th dose because my reading of the evidence suggests to me that it’s the obvious choice, not only to maximize my protection against the worst outcomes, but also to increase –however marginally– my ability to avoid infection altogether.