These are used in cases where you can measure something, such as the how dark some liquid gets after some reaction, but cannot, or don't bother, to turn that into some "real" number such as the number/concentration of particles in the liquid causing the effect.

So these units are comparable across several experiments or measurements using the same protocol. Sometimes they are tied to the equipment used and can be compared to other data from the same equipment. But other than that, only the relative values or, as in this case, the finding of proportionality is meaningful.

But, to get a general idea: just look at the values they got across their population, remove any extreme outliers, and put it next to some chart using units you want to convert to. Make sure they are on the same scale (i. e. linear or log). The mean and other quantiles should correspond.

Or, depending on what you're looking for, just eyeballing fig. 1 (https://www.medrxiv.org/content/medrxiv/early/2021/06/24/202...) itself might be enough. It shows (little surprise) that the bulk of measurements cluster in "safe" territory. That asignificant number of people show stronger reactions than that bulk, by sometimes orders of magnitude, but don't derive (much/yet) additional protection from it. And that a number of individuals show a weak response and are at risk, without that group being distinct from the rest in any specific way. Rather, it's a continuum