Let’s do an update on the antibody situation in this epidemic (a bit of antibody/immunology background here if you would like some). The word is today that the CDC is starting several efforts to do antibody testing. One of these is to look at people (from areas that showed widespread cases of coronavirus) who never showed any symptoms themselves. How many of these folks will turn out to have had asymptomatic cases anyway?
The second study is more wide-ranging across the US population – similar efforts are underway from others (here’s one being run out of Stanford, and there’s an interesting round of testing around Telluride, Colorado), and combining these will help to give us a picture that at the moment we just don’t have: knowing the actual prevalence rate of the disease is vital for figuring out our strategy in the coming months. The third CDC study is specifically looking at health care workers, a population that sadly has had every chance possible to be exposed to the virus, to see how many of them show exposure compared to the never-had-symptoms group and the general population.
A note about antibody testing is in order, and it’s the usual concern about diagnostic tests in general. The mathematics mean that the higher the prevalence of what you’re testing for in your population of subjects, the better the accuracy of the results. So if your disease has a low prevalence, your test has to have very low false negative and false positive rates to be meaningful or even useful at all. Here are a couple of blog posts that run through the numbers as illustrations of this, and here is a good Twitter thread on one of the tests (Cellex) that the FDA has recently given Emergency Use Authorization for. That one has a 93.8% sensitivity and a 95.6% specificity, which means that if (for example) 4.5% of the US population has had the virus, that a positive result with that test only means a 50% chance that you actually are positive. If 30% of the population has actually been infected, the test is right 90% of the time, and so on (see the link for details). I don’t know what tests the CDC and Stanford surveys are using and what their specificity/selectivity numbers are, although we can be sure that they know all about the statistics. But everyone else should, too.
One of the uses of antibodies from recovered patients, naturally, is “convalescent plasma” treatment. If someone has raised neutralizing antibodies to the coronavirus, their blood plasma can be given to someone else as an emergency measure, and those antibodies should work on them as well. This has already been used on a small scale (it’s an old technique), but unfortunately it’s not easy to scale up. Depending on the antibody titer of the recovered patient, it may be that one person might have enough antibodies to treat only one other person, and this is naturally a relatively labor-intensive process which will be reserved for the critically ill. Another thing to remember is that these antibodies are effective on infusion, but that you’re not giving the recipient their own lasting antibody response by doing so – their immune system will have to mount that.
But how many people have raised neutralizing antibodies? That’s another key piece of data that we don’t really know yet. This new report from Fudan is worth thinking about, where they’ve analyzed 175 patients discharged from hospitals in Shanghai after coronavirus infection. Neutralizing antibodies appear about 10 to 15 days after the onset of disease (which sounds about right) and target three different regions of the “spike” protein on the virus. (Interestingly, these do not cross-react with the earlier SARS coronavirus spike protein). The total amount in the blood (the titer) varied quite a bit between individuals – notably, younger patients had far lower levels than older ones did, which raises the question of how immune they really are. In fact, ten of those young patients had no detectable neutralizing antibodies at all (!) and overall, about 30% of the entire cohort failed to develop a high antibody titer (although they had similar disease progression before their recovery).
This presumably means that other parts of the immune system played a greater role in clearing the virus, which is fine – except that those long-lasting antibodies and memory B cells are the key to staying immune. A lot of people are (understandably) talking about having some sort of “immunological passport” system to clear people for work, etc. before we are able to vaccinate the population, and these results may be telling us that that will be a complicated process, one that might not clear as many people as one would hope.
We’ll see if the CDC numbers and others back up these results in our own populations. Stay tuned – the antibody story is a big part of all this, and it’s just going to get bigger.
Update: forgot to mention this paper. Thanks to a collaboration between Scripps and the University of Hong Kong, we now have an X-ray crystal structure of a neutralizing antibody bound to the receptor-binding domain of the coronavirus spike protein. That’s one of those three regions mentioned above that such antibodies have been found to interact with. For those into structural biology, there are interesting insights into conserved protein epitopes, and some good points about how such antibodies may be more more relevant in vivo than they appear to be in some in vitro assays.