A case for Variolation against COVID-19?

It seems plausible to me that being exposed to a small amount of covid results in much milder disease, while still providing significant immunity against further infection. (It also seems plausible it makes no difference, e.g. because a smaller initial dose just needs to multiply further before it's detectable by your immune system.)

What’s the current state of evidence on that question? I’ve seen periodic discussions (as a potential auxiliary benefit of masking, and in response to Robin Hanson advocating deliberate exposure) but they've been pretty unhelpful for understanding the actual evidence. Is this still an open question, or do we have a pretty strong guess that it wouldn't work?

If we just don't know, it's an interesting case because it seems relatively cheap to learn by exposing a moderate number of healthy participants to gradually increasing doses. You'd only need a few symptomatic infections before you learned quite a lot about whether it worked. So I think the main cost of the experiment is the time and hassle of running it rather than the risk to human participants. (I assume the main difficulty is controlling the amount of virus you're exposed to, which does sound difficult.) Unlike vaccine challenge trials, these experiments wouldn’t need to involve a vaccine manufacturer who would be scared of negative fallout.

In any case vaccination is probably better than low-dose exposure. But if vaccination provides imperfect immunity and the residual risk is high enough to matter, it seems like it could make sense to deliberately expose vaccinated individuals. If protection from the vaccine is effective this seems like it shouldn’t cause much harm, and in the best case it could eliminate much of the remaining risk.

One reason to think this wouldn’t work is that there are historical cases of deliberate low-dose exposure, but I’m not aware of any that have been successful at consistently inducing asymptomatic infections. That said, it seems like modern technology potentially gives us much more fine-grained control over exposure, and we can try more sophisticated policies like gradually ramping up exposure (or having repeated very-low-dose exposures) to ensure we really get the minimal dose necessary to lead to an infection.

People may avoid discussing this topic if there is concern that it could lead to people taking pointless and destructive risks in half-baked attempts to expose themselves (which likely wouldn’t actually provide much protection or help us figure out what’s going on). I’m obviously not advising anyone to deliberately expose themselves to covid. The main thing I’m wondering about is whether we are making or have made a mistake by not running these studies, which I find particularly interesting because they are so cheap. Or is this just another case where the world is more complicated and messy than it looks at first?

I’m also kind of curious about this as a response to regular cold and flu. Are there many diseases where deliberate low-dose exposure could plausibly eliminate symptomatic infections? I’ve mostly seen this practice discussed for smallpox (“variolation”)—are there other examples? Is it applicable to smallpox because it happens to work unusually where in that case (I guess because the active ingredient was the site of infection rather than the dose), or because it was an unusually costly and prevalent disease for which even rudimentary versions of exposure could add a lot of value?

Here is an example of people talking about it in a commentary in the NEJM: https://www.nejm.org/doi/full/10.1056/NEJMp2026913

A similar "personal view" in the lancet: https://pubmed.ncbi.nlm.nih.gov/33631099/

These are pretty light on evidence. But they have a lot of citations without me having seen any convincing dismissal, leading me to believe that the hypothesis probably can't be easily dismissed. But then that leaves open whether more serious implications aren't being explored because it's not the kind of thing people talk about in the medical establishment, or because the idea is in fact much more viable as an argument for non-pharmaceutical interventions than as a potential prevention strategy to be explored.

Here is Robin Hanson advocating for deliberate exposure last year: https://www.overcomingbias.com/2020/03/variolation-may-cut-covid19-deaths-3-30x.html

I think I feel intuitively even more optimistic about low-volunteer-count experiments. For example, suppose that I had even a single volunteer who was willing to be exposed to slowly-escalating covid doses. What's the probability that they ultimately have a symptomatic vs asymptomatic infection? I think the pessimistic story is that it's ~50%, whereas the optimistic story is <10%. So that's practically a factor of 2 of evidence about this hypothesis from n=1. (Though again, the main difficulty seems like organizing the exposure.)

(Less evidence if the person is vaccinated since we have <50% probability of a symptomatic infection, but also still informative about further facts like "what is the nature of protection from vaccines?" I'd be totally happy to participate in such an experiment, I think the bottleneck is very unlikely to be volunteers.)

My overall best guess is that this isn't super promising (and it would have to work quite well to overcome logistical hurdles, especially at this point), but that it is still a massive unforced error not to do the experiments and that it would be an extremely good idea to do them even now. I find it kind of amusing and perverse that "isn't it too late to matter now?" just keeps feeling like a good argument 18 months after Robin Hanson wrote "Some fear that it is now too late to consider variolation, as the pandemic peak may be only a few weeks away."

He followed it up with: "But lockdowns may succeed in substantially slowing Covid19 growth, and we may then be in for many months or years of alternating local waves of suppression and reappearance." which looks like a good prediction. In general my sense is that Robin's early writing about the pandemic and our response looks pretty good in retrospect.