Ever since the beginning of the covid pandemic, one of the big topics of discussion has been whether infection results in lasting immunity. Since the advent of the vaccines, that has expanded in to a discussion about whether prior infection or vaccination provides a higher degree of immunity.
Back in December, I wrote about a study that showed that 90% of people who get covid still have antibodies six months out from infection. This was encouraging news. However, all it really did was show that most people keep their antibodies for a decent period of time after infection. It didn’t actually tell us anything about the probability of being re-infected.
Antibodies are a “surrogate” marker. We think they might tell us something useful, but we can’t really be sure. It’s kind of like looking at the share of a population that have high blood pressure instead of looking at the proportion that are having strokes. We really don’t know whether the presence of antibodies after infection means that someone is immune, or whether the absence of antibodies means that someone has lost their immunity. In fact, we still don’t really know whether antibodies play a meaningful role in fighting covid or not. Correlation isn’t always causation. Antibodies appear to be a good marker for prior infection, but that doesn’t mean that they have a causal role in preventing a re-infection.
So, what we really need is a study that looks at the degree to which people actually get re-infected, not more studies that look at antibodies. Once we have that, we can do a comparison with the results of the vaccine trials, and then we will finally have a reasonably good estimate of whether prior infection or vaccination provides a higher level of immunity, or if they are equivalent. That is now exactly what we have, thanks to a study that was recently published in The Lancet.
This was a cohort study carried out in the UK that recruited 25,661 NHS hospital workers and then followed them for an average seven months. The study was funded by the UK government. Participants were divided in to two cohorts, a covid positive cohort and a covid negative cohort. The purpose of the study was to see what proportion of people in each cohort went on to develop covid-19. The data were collected during the second half of 2020.
Everyone who had or had previously had a positive antibody test or PCR test for covid-19 at the beginning of the study was placed in to the covid positive cohort, and everyone else was placed in to the covid negative cohort. The covid positive cohort contained 8,278 participants at the beginning of the study, while the covid negative cohort contained 17,383.
Since this was a study of healthcare workers, more than 80% were female, and since it was carried out in the UK, more than 80% were white. The median age was 46 years and 75% had no underlying health conditions. In other words, the results primarily apply to relatively young healthy white women. It’s actually a good thing that the participants were relatively young and healthy, because we want to compare the results we get here with the results from the vaccine trials, and the participants in those trials were also young and healthy. The fact that the study mainly consisted of white women shouldn’t be that much of a problem, since there is no evidence to suggest that non-whites or men are different in their ability to develop immunity after getting covid as compared to white women.
Questionnaires were sent out to participants every two weeks asking about whether they had recently had any possibly covid-related symptoms, and they were also tested at regular intervals with both PCR tests and antibody tests. The goal was to PCR test all participants every two weeks, and antibody test them once a month. In other words, the participants weren’t just tested if they had symptoms. They were continuously screened for covid.
This means that the risk of missing a case was very low. Rather the opposite, in fact. It means that they found a large number of asymptomatic cases, or as we normally call them in medicine, healthy people. This could have been a problem in terms of allowing us to compare the results of this study with the vaccine trials, since the vaccine trials only counted people as cases if they both had a positive PCR test and also had at least one symptom suggestive of covid-19. Luckily, it isn’t a problem, because this study has gathered and presented data on the proportion of those with a positive PCR test or antibody test that actually had symptoms, and the proportion that didn’t. So we have all the data we need to do an apples to apples comparison with the vaccine trials.
Ok, let’s get to the results.
Over the course of 2,047,113 days of follow-up in the covid positive group, there were 78 cases of symptomatic covid-19 (by which we mean a positive test + at least one symptom).
Over the course of 2,971,436 days of follow-up in the covid negative group, there were 1,369 cases of symptomatic covid-19.
This works out to a relative risk reduction 0f 92%. For comparison, the Pfizer vaccine trial reported a reduction of 95%, the Moderna trial reported an reduction of 94%, the Astra-Zeneca trial reported a reduction of 70%, and the Johnson&Johnson trial reported a reduction of 67%.
So, on the face of it, prior infection is equivalent to the Pfizer and Moderna vaccines in terms of the level of protection offered, and much better than the Astra-Zeneca vaccine and J&J vaccine. In light of this, it seems completely unnecessary for people who have had covid to get the vaccine. In fact, if the goal of governments is to get their populations to herd immunity as quickly as possible, it would make more sense to tell people who have had confirmed covid-19 that they don’t need to get vaccinated. Vaccinating people who have already had covid-19 means delaying vaccination of people who haven’t had it, which means delaying the onset of herd immunity.
There is one potential problem with taking the 92% number at face value, especially in relation to the results from the vaccine trials, and that is that this is an observational study, not a randomized trial, so there is significant scope for confounding. For example, it could easily be the case that the people who had already had covid at the beginning of the study were the people who were at highest risk of exposure. Maybe they were disproportionately front-line workers, caring for covid patients. In that case, they would be disproportionately more like to get exposed to covid again over the course of the study than the people in the covid negative group. If that was the case, it would make the risk reduction seem smaller than it really is.
Conversely, it might be the case that those who were covid positive at the start of the study were disproportionately working in areas that were hard hit during the first wave, and that therefore had already built up a high level of population immunity by the time the second wave came around. These areas would then be only mildly hit during the second wave. That would mean that those participants who were negative at the start of the study would disproportionately be working in areas that hadn’t been hit very hard in the first wave, and that would therefore likely be hit harder in the second wave. If that was the case, then the covid negative group would end up being more exposed to covid during the study than the covid positive group, which would make the risk reduction seem bigger than it is.
The researchers attempted to correct for confounders to the extent that they were able, and came up with a modified risk reduction of 93%. But correcting for confounding is really a kind of guessing game. It isn’t a very reliable technique. And for all the confounders that are known and that can be corrected for, there are plenty more that aren’t known and can’t be corrected for.
That being said, a 92% or 93% risk reduction is a huge reduction, not far off the difference in lung cancer rates seen between smokers and non-smokers, so even with unknown confounders pushing the results up or down, it is clear that prior infection provides a high degree of immunity.