Everyone was wrong about Omicron reinfections

Infection with the newest variants of SARS-CoV-2 provide good immunity against reinfection

As it begins to dawn on even the most staunch advocates of “Zero covid”, that, for better or worse, the virus is here to stay for the foreseeable future, and as the majority of most national populations have now been infected at least once, much attention has been drawn to the nature of re-infection.

Being Covid-19, this attention has been far from dispassionate. There have been many spurious claims made about Covid-19 reinfection, but we will focus today on the one which is now most obviously wrong.

Following the emergence of the Omicron variant, the rate of reinfections rose significantly as existing immunity (through infection or vaccination) provided little protection against symptomatic infection, although good protection against severe illness remained.

However, there has been a prolific claim that infection with the original subvariants of Omicron (BA.1 or BA.2) did not provide protection against reinfection from the most recent subvariants, BA.4 and BA.5

Where did this come from?

Aside from the usual raft of anecdotal claims of someone’s aunties child who has been infected 8 times in the space of 3 hours, one particularly high profile research paper seemed to bring substance to the idea. Published in the high impact journal Science, the paper found that infection with Omicron provided little boost to antibody or T cell immunity to Omicron itself, and that Omicron infection following previous infection with the original, “Wuhan” strain of the virus also failed to boost immunity.

The paper garnered a lot of attention, not least because one of the senior authors of the study wrote an editorial in the Guardian newspaper, titled “Where’s the herd immunity?” stating that:

“Even having had Omicron, we’re not well protected from further infections.”

And furthermore:

“ …this is more like being trapped on a rollercoaster in a horror film.”

It’s easy to see how this theory gathered pace, and caused a huge amount of anxiety.

It’s completely wrong

Fortunately for us, this has turned out not to be true. A flurry of recent observational studies from multiple different countries has proven that infection with Omicron provides substantial protection against reinfection from other Omicron variants.

A paper from Qatar which has extremely high levels of testing, used a well established test-negative case-control design and found that previous Omicron infection provided 80% protection against any infection with BA.4/5.

A second study from Denmark, which has some of the best national registry systems in the world, found a prior Omicron provided 94% protection against reinfection with BA.5.

Most recently, a third study from Portugal (which has had a particularly big wave with BA.4/5) found that prior infection with Omicron BA.1/2 provided 80% protection against reinfection with BA.4/5.

Some important things to consider include that all these populations are highly vaccinated and so the findings might be different in a low vaccine setting, and that observational studies of this nature can have some test seeking bias (e.g. people who have already been infected recently may be less likely to seek testing subsequently), however it would be a very far stretch to suggest this could explain such large effect sizes as observed in the studies above.

What can we learn?

First and foremost, we can now be confident that previous infection with Omicron provides robust protection against reinfection with different Omicron subvariants - at least in the short term.

As a more general lesson, extrapolating from surrogate outcomes (in this example, levels of neutralising antibodies measured in a laboratory) to the hard outcomes which we care about (real life protection against infection) can be very difficult in biomedicine. Some surrogates certainly have better correlation than others, and over the past few years neutralising antibody levels have fared well as a correlate of protection against symptomatic infection with Covid-19 (we use them to compare booster vaccines in the Cov-Boost clinical trial). However, the numbers of participants in the different groups of the study which prompted much of this anxiety are quite small (not uncommon for lab studies of this nature), and it is possible that the findings might be different were a similar study to be performed again.

Ultimately, when we are not measuring the thing we actually care about, we would be wise to interpret scientific findings with a pinch of salt, and avoid hyperbole or over-interpretation. Mechanistic studies are incredibly useful when they are all we have, but it is much more useful to have direct measurements of the outcomes which matter to people.

Recently, Rob Hughes put it best;

Measurement eats mechanism for breakfast.