Many readers will have seen, for example, this new paper from The Lancet on a large study in Spain. Testing tens of thousands of people across the country continues to show that (on average) only about 5% of the population is seropositive (that is, has antibodies to the virus). There are a lot of interesting findings – such as rather large differences in those positive testing rates in different regions of the country, as well as the realization that at least one-third of the people who now test positive never showed any symptoms at all. But we are still not sure if this means that 95% of the Spanish population has never been exposed to the virus, because we don’t know how many people might have cleared it without raising enough of an antibody response to still be detectable. This paper does show that seroprevalence was about 90% in people 14 days after a positive PCR test, which indicates that most people do raise some sort of antibody response, but we don’t know how many of these people will still show such antibodies at later testing dates. Remember the paper discussed in that link in the first paragraph above, which found that 40% of asymptomatic patients went completely seronegative during their convalescence.

In other words, the Spanish survey may appear to show that 95% of the country has not yet been exposed to the coronavirus, but that’s almost certainly not true. The authors do mention that cellular immunity is important and not something that they were able to address, but the combination of that factor plus the apparent dropoff in antibody levels with time makes these large IgG surveys almost impossible to interpret. But note that if there are indeed many people who have been exposed but do not read out in such surveys, that we also have no idea how immune they are to further infection. At a minimum, you’d want to know antibody levels over time, T-cell response over time, and (importantly) what a protective profile looks like for both of those. We barely have insight into any of this: the large-scale data are just a snapshot of antibody levels, and that’s not enough.