The two authorizations issued by the FDA for COVID-19 vaccines come because of clear data that they limit infections by the SARS-CoV-2 virus, and ensure that any ensuing cases are mild. Studies have also indicated that the vaccine triggers the development of antibodies specific to the virus. Oddly, however, we don't have good data on an obvious question: is there a causal relationship between the two? In other words, we haven't determined whether production of anti-SARS-CoV-2 antibodies is a necessary step towards providing protection, or how long that protection lasts.

The have been a few small studies that hint at answers to these key questions, but significant uncertainties have remained. Now, a massive study out of Oxford University Hospitals provides a clear indication that high levels of antibodies are protective. But, even with 12,500 participants, the study doesn't eliminate the uncertainties.

The good news

To get some good numbers, Oxford University Hospitals tested its entire staff of healthcare workers, both for the presence of viral RNA, and for antibodies that indicated a past exposure to the virus. Following the initial tests, all the staff had the option of being retested for the presence of virus every two weeks, and antibodies every two months. Testing started back in April, when the first wave of infections was still happening, and continued through the end of November, when the second wave was still building. While many of the hospital staff were busy enough that they took longer than two weeks for follow-up testing, the hospital was able to track over 12,500 people.

Already at the start of the study, 1,265 people had been infected with the virus. Many of them had been exposed or experienced symptoms before testing was widespread in the UK, so we can only infer they'd been infected based on the presence of antibodies.

Advertisement

Over the course of the study, 225 ended up having a positive test, with a bit under half of those positive results coming from asymptomatic cases. Most of those new cases came towards the end of the study. A grand total of two of these people were among those who had antibodies to the virus during the original testing, suggesting that they had been reinfected. Put differently, the rate among the health care workers in general was 1.1 cases for every 10,000 days of risk during the study period. Among those who had a positive antibody test, the rate was 0.13 cases per 10,000 days of risk. These repeat infections were both asymptomatic.

We'll come back to those two cases in a bit, but let's take a moment to focus on the good news. The antibody tests used here don't produce binary yes-no answers; instead, they're quantitative, measuring the levels of antibodies against a specific target. Or, in this case, two targets, as the researchers measured antibodies against both the spike protein on the virus' surface, and a protein that is embedded in the membrane that surrounds the virus' genetic material.

In both cases, there was a strong inverse relationship. The higher the levels of antibody present, the less likely someone is to be infected. That was true for antibodies against both of the target proteins. This suggests that antibodies are either directly involved with lowering the risk of infection or are clearly correlated with something that is. Since the highest risk came roughly six months after most people in the study were initially exposed, this also provides evidence that the immunity endures at least that long.

The caveats, of which there are many

If you were paying careful attention there, you might detect a potential issue: the fact that the levels of antibody correlated with infection risk implies that there are intermediate states. One such state happens when you don't have high levels of antibody, but still get some protection. This is definitely the case with this data. To get to the "only two reinfections" figure, the authors had to choose a threshold on the levels of antibody that indicated having a previous infection.

Advertisement

Below that threshold, a person wouldn't be considered to have been infected previously, but they may still have some antibodies that react with SARS-CoV-2 proteins. This may be from an earlier infection that triggered a weak immune response, it may be from an infection by a related virus (like the ones that cause cold symptoms), or it may just be random chance. So, it's possible that a larger fraction of the study population actually had experienced an earlier infection.

The flipside of this is that, even using the best available equipment, false positive tests are nearly inevitable. With 12,500 people participating, there's a realistic chance that both of the "reinfections" seen here were simply the product of false positive antibody tests. There are also similar issues with the RNA-based tests, which also produce false positives and negatives. The researchers note that one potential reinfection cases had seen a positive test, followed by two negative tests over the next several days, suggesting that the first was a false positive.

Finally, there was the issue that, on average, the participants were screened for viral RNA every 10 weeks. There's certainly time during those gaps for an asymptomatic infection to start and finish without ever coming near a test kit.

So, if you look to long at the details, there are enough questions left that it's easy to convince yourself that we don't really know anything. But that would be missing the forest by focusing on a few stray trees. Overall, it appears that the more antibodies you produce, the more likely you are to be immune to reinfection (although again, we can't say whether the antibodies themselves produce this immunity). And that protection lasts at least six months after the initial infection.

Even if there are some exceptions to this, it's a finding that bodes well for the approved vaccines, which also induce an immune response that includes significant levels of antibodies. And, long term, these results should help us piece together a clearer picture of what SARS-CoV-2 immunity looks like.

New England Journal of Medicine, 2020. DOI: 10.1056/NEJMoa2034545 (About DOIs).