Immunity Passports? Uh, No! Confusion with Antibody tests

Nice article in USA Today on questions and confusion with antibody tests.


Antibody, or serology, tests measure whether a person has COVID-19 antibodies in their bloodstream, indicating a likely current or previous COVID-19 infection. They measure two main types of antibodies, or immunoglobulins (Ig), IgM and IgG. IgM are the earliest detected when a person is infected, while the IgG remain in the bloodstream long after infection, conferring potential immunity against future infection since the body has the "targeted weapons" for fighting the infection should it be encountered again.


What they are good for or not good for:

* They are not generally good for determining who is infected for two reasons: (1) having the antibodies doesn't mean the person is CURRENTLY infected, but infected at some time in the past, and (2) some infected people are not producing an antibody response yet so would be missed if this were used as a diagnostic test.


So, if not useful for diagnosis, what is it good for? Three main things -- one for sure and two potentially:


1. Applying these tests broadly in the population provide insights into what proportion of the population has been previously infected. This is very important in a virus like SARS-CoV-2 for which many infected people never show symptoms or only show mild symptoms, and for which viral (RNA-based) testing is limited. Official incidence counts only include those who have been confirmed infected by a viral test, and these tests have typically only been given to people with moderate or severe symptoms, and even many of those individuals have not been able to be tested either because they stayed at home or testing shortages. Thus, the official incidence is only the "tip of the iceberg". Antibody tests can help us understand the rest of the iceberg, which is useful for various reasons -- two are:

(1) how close are we to herd immunity (when 50-70% of population has antibodies from

infection or vaccine the virus can't spread anymore and is effectively defeated.

(2) what is the true "case fatality rate", meaning the proportion of infected who will die

from the disease. When using the official incidence as the denominator of this rate,

it tends to be severely overestimated, since incidence is only measuring a select

(not random) subset of those with severe disease. However, it is not enough to only

fix the denominator -- the numerator is also difficult to estimate accurately since

official death counts may not accurately capture all who actually died of infection.

Note: in using these tests to characterize the population, the key statistical principle of "representative sampling" must be considered. If a biased sample is obtained that captures people who think they were sick so are more likely to have the antibodies, e.g., then use of these studies to estimate true infection rate or case death rate without proper adjustment for this systematic bias will yield misleading results. This is the criticism of the famously flawed "Santa Clara" study. Nonetheless, across many of these studies that have been reported, it appears safe to conclude that the "true infection rate" is somewhere between 10 and 15x the reported incidence rate. These studies will keep coming out in the coming week and provide more insights.


2. The presence of SARS-CoV-2 antibodies may confer immunity against future infection. This is the idea that sparked discussion of "immunity passports" that some have suggested could be required for travel or going back to work "safely" without worry of getting COVID-19. As this article mentions, the WHO has (wisely) recommended against such measures, for several reasons.

The primary is that we don't know for sure whether these antibodies confer immunity or not and, if they do, for how long that immunity will last, and whether this immunity would only be against certain strains of SARS-CoV-2. One unfortunately detail of viruses is that they are always mutating, and if mutating enough and in certain ways, previously obtained antibodies or even a vaccine can lose effectiveness. Influenza strains mutate very rapidly, which is why we need to have a new formulation of Flu vaccine every year -- it is reformulated to include new variants. Fortunately, evidence suggests SARS-CoV-2 mutates more slowly, which makes it less of a moving target, but it does mutate and it is not clear whether an antibody test is enough to guarantee immunity.


3. One key potential treatment strategy for advanced patients involves blood transfusions from previous COVID-19 patients who have recovered and thus presumably have effective antibodies against the virus. This is the "oldest trick in the book" in terms of treating viruses, being used well over 100 years in other viruses. There are currently 52 ongoing studies assessing this strategy and antibody tests can be used two ways: (1) to identify good donors with high levels of IgG and (2) to identify the patients most likely to benefit, e.g. those not yet having high levels of anti-SARS-CoV-2 antibodies in their blood stream.


Another key limitation of current antibody tests that is highlighted by this article is their potential inaccuracy. After the testing fiasco at the beginning of this crisis that resulted from the viral test that the CDC was developing not initially working, the FDA streamlined procedures to allow private companies to quickly develop and disseminate tests before they were validated in the usual rigorous fashion -- an important and necessary move during "wartime medicine", but not without its drawbacks. Well, this also enabled the dissemination of >150 different antibody tests whose accuracy is not known. Some of these antibody tests can yield a high number of "false positives" that suggest someone was infected with SARS-CoV-2 but actually were not. These are not technical false positives, meaning the test didn't work right, but are from antibody tests that are not targeted enough to the specific virus SARS-CoV-2 and can indicate previous infection with a previous coronavirus. Since what we call the "common cold" contains a combination of viruses, including some coronaviruses, a poorly formulated test can capture some of those, for example. Most well-formulated tests will not have high rates of nonspecificity, but all will have some degree of non-specificity. One could expect such a "false positive rate" of about 1% for a well-designed test, potentially higher for one not as well designed. This makes interpretation of these tests even more difficult. BTW, 8 of the antibody tests on the market have been approved by the FDA and have studied and characterized error rates.


We will see broad, population based studies with better scientific and statistical designs appearing in the coming weeks and months, and these should give us the information we need to effectively use these tests to gain insights about infection and case death rates, and to assess what degree of immunity is conferred by a positive test.



 

©2020 by Covid Data Science. Proudly created with Wix.com