No, coronavirus vaccines do not increase the risk of infection after three months
Shortly before Christmas, Statens Serum Institut (SSI), a Danish state-owned research institute, reported that the coronavirus vaccine protection decreases over time, but increases again after revaccination. This was established in a new study of the difference in the effectiveness of the two mRNA vaccines against the Omicron and Delta variants.
An important finding in the study, in addition to the booster effectiveness, was that both vaccines provide protection against the Omicron variant. However, it is significantly lower than against Delta, and the protection against Omicron decreases faster after full vaccination.
The study was published as a preprint, before it was peer reviewed, on MedRxiv.org, and it attracted attention in a different way than intended. At the end of the study, there is a table that shows that both mRNA vaccines have a negative effectiveness against Omicron 91 days after the full effectiveness of primary vaccination, i.e. 14 days after 1st dose.
“An estimated negative vaccine effectiveness simply does not make sense. Therefore, we have to say that it should be interpreted as no effectiveness,” says Palle Valentiner-Branth, acting head of the Department of Infectious Disease Epidemiology and Prevention at Statens Serum Institut, who has carried out the study together with his colleagues.
Anti-vaccine sentiment
In the scientific paper, the researchers themselves address the seemingly negative vaccine effectiveness and indicate that the effectiveness may be underestimated due to differences in behaviour and/or exposure patterns between the vaccinated and the unvaccinated group.
However, the negative figures for vaccine protection were quickly unreservedly circulated by websites such as the Canadian Rebelnews.com.
Under the title “Denmark study shows vaccinated more likely to catch Omicron”, the anonymous author quotes an episode of a YouTube show hosted by the Rebel News founder Ezra Levant, where he says, among other things:
“You are literally more likely to get infected if you are vaccinated with Pfizer or Moderna than if you're not. This is not my opinion, this is what this chart in this national Danish study, published online by a collaboration by Yale and the British Medical Journal show, and they're sponsored, like I said, by Mark Zuckerberg.”
There is a lot to unpack in the one sentence alone, but if one simply looks at the basic claim that the SSI study shows that vaccines make people more likely to get infected by the coronavirus, then there is no basis for that conclusion.
Bias or error in calculations?
However, refuting the claim somewhat unusually requires more than just looking up the original scientific article because the study actually shows a negative vaccine effectiveness (VE) of both mRNA vaccines against the Omicron variant. More specifically, a table in the article indicates that Pfizer (BNT162b2) and Moderna (mRNA-1273) vaccines have a VE of, respectively, -76.5 (from -95.3 to -59.5) and -39.3 (from -61.6 to -20.0) against Omicron after 91–150 days.
The negative figures after 91 days could formally be interpreted as meaning that three months after full vaccination, the risk of Omicron infection increases.
If you take the confidence intervals literally, the risk increases by at least 20 percent and may be up to twice as high as in the unvaccinated. And it should be pointed out that this is after a period in which the same table in the same study shows a protective effect of both vaccines against Omicron. And after 91 days, a protective effect of the same vaccines against Delta is still observed.
In a figure showing the result, the entire confidence interval of VE for both vaccines in the last time period is well below the baseline, which in terms of vaccine effectiveness is zero percent—where the risk of infection is the same for the vaccinated and the unvaccinated individuals.
VE is expressed as the percentage of protection in relation to the baseline—infection among the unvaccinated—and for Pfizer the protection in the Danish study is estimated at 55.2 against Omicron and 86.7 against Delta in the first month after primary vaccination for people aged 60 and over. In other words, the vaccine prevents well over half of the Omicron infections and almost 90 percent of the Delta infections in that age group.
And the key word here is “estimated”. There are two versions of VE, both of which can be useful: The golden standard for measuring the effect of drugs, including vaccines, are RCTs—randomized controlled trials—where half of the randomly selected participants receive the vaccine, while the other half serves as the control group and receives a placebo, an inactive imitation of the vaccine. Vaccines are approved on the basis of such trials.
Subsequent studies of VE for already approved vaccines will often be estimated from observations without intervention, which will always negatively affect the reliability of the study, no matter how many participants are involved and whether the data is processed completely by the book.
VE measured in an RCT is expressed as efficacy, which is the effect of a vaccine under ideal conditions.
VE based on observations is expressed as effectiveness.
Observations in the population
The SSI study is a cohort study in which well-defined groups are observed over time, and it is crucial that the exposed and the unexposed group are as similar and as comparable as possible, which in clinical trials is achieved by random selection.
The authors of the study themselves indicate several possibilities for bias or sources of error, which may have pushed the results systematically in a certain direction, even though the data and calculations are undeniably correct.
“It’s a report that covers the first 20 days of Omicron. The first cases were imported by people who had travelled to Southern Africa, and all other things being equal, they should be fully vaccinated. And then there were some superspreader events involving young people. Omicron had not yet reached the general public, so both the elderly and children are underrepresented, and we are probably looking at some groups that do not necessarily behave like the rest of the population,” Palle Valentiner-Branth says.
The figures in the study have been adjusted for age, sex, and geographical region, and calendar time was used as the underlying time scale, because the risk of Omicron infection has increased during the period.
SSI prepared an explanation of the negative estimates of vaccine effectiveness in the study: There are a number of reasons why the VE estimate might be negative. 1. In many places including Denmark, vaccinated individuals are tested more frequently than unvaccinated individuals. This causes the incidence rate to be higher in the vaccinated population and resultantly a negative VE estimate. 2. Denmark was very quick to conduct sequencing and to identify the first generations of Omicron cases in the country. Cases during this period occurred to an exaggerated extent in those who were travelling internationally, and those in the social and professional circles of travellers, and were largely vaccinated. We expect therefore that there was an overrepresentation of vaccinated people among the first generations of Omicron cases identified in Denmark, not because the vaccines weren’t protective, but because the variant hadn’t spread far enough into the general population, including into the unvaccinated population, to make for comparable infection rates. 3. VE estimation relies on vaccinated and unvaccinated individuals behaving in a similar fashion in their every-day lives with respect to COVID-19 precautions and exposure to infection risk. It is conceivable that the increasingly small cohort of unvaccinated individuals that remains in Denmark takes further precautions (precisely because they are not well protected), engage less in social activities, etc. Such discrepancies in risk behaviour between vaccinated and unvaccinated individuals will lead to an underestimated VE. On that basis it is reasonable to expect that the vaccine effectiveness estimates presented in our study are too low, not only for the fourth period (91-150 days after vaccination) but likely also for the earlier time intervals. To conclude, the vaccines’ protective effect may be low against infection with Omicron after 4 months, but it is most unlikely to be negative!Three possible explanations
Source: SSI
“But there are things we can’t adjust for,” Palle Valentiner-Branth says.
Before the summer holidays, there was a large proportion of unvaccinated people who voluntarily got tested, while today it is the vaccinated people who are most diligent in getting tested. Therefore, a larger proportion of coronavirus infection will be detected in the vaccinated population, while there is a greater hidden number in the unvaccinated population.
Studies have previously been done on the true prevalence of the coronavirus by taking blood samples from a representative sample of the population, but this was at the beginning of the pandemic. The stated “incidence” per 100,000 people in the daily reported cases must therefore be taken with a grain of salt, because it is based on a decision to get tested.
Problematic comparison with the unvaccinated population
The effectiveness for those who have been vaccinated without a booster has been determined by comparing them with those who are unvaccinated, which is the standard for VE in vaccine approval.
But in Denmark, vaccination status is not randomly distributed among the population—being unvaccinated is an expression of conscious choices or certain living conditions. And there are so few unvaccinated people that it can be problematic to compare the vaccinated with the unvaccinated. Both because one group is so small and because the groups are assumed to be heterogeneous.
“Maybe unvaccinated people also behave differently. We already know that they don’t get tested as much and perhaps they take more precautions to avoid infection. This may mean that we have to make reservations about the results,” Palle Valentiner-Branth says.
VE for the vaccinated has been determined by comparing the vaccinated with the unvaccinated, while the effectiveness of the booster has been determined by comparing the vaccinated with those who have already received the third dose. Those two groups should be more directly comparable.
However, at that time that still included very few people, and for Moderna there were not enough boosters to estimate the protection against Omicron. But for both variants, the Pfizer booster seems to bring the protection up close to what it was right after finishing the primary vaccination, and the same goes for Moderna’s protection against Delta.
The study is exclusively examining the risk of infection. There are other studies underway that examine the risk of hospitalization and death from Omicron compared to Delta, but they are not finished yet. And the first of this type of studies will, just as the estimate of vaccine protection, carry the inherent uncertainty that it is not a random sample of the population that is affected by the first wave of a new virus variant.
