COVID-19 antibody tests are the rage these days, and countries around the world are racing to get their hands on the kits. Experts have touted these rapid diagnostic tests as a game-changer in our efforts to contain the spread of the novel coronavirus. Many have said that these tests could help countries test their people en masse, even help governments hand out ‘immunity certificates’ to those with the corresponding antibodies against the virus in their blood, and that the tests could thus help kickstart a flagging economy by sending those who have recovered from COVID-19 back to work.
These promises, however, oversell what antibody-based tests are really capable of. To understand how, it’s important to know how these tests really work.
Antibody-based tests look for antibodies against the virus present in the blood. The other way to ascertain the virus’s presence in the human body is to look for signs of the virus’s genes in swabs collected from the upper respiratory tract, using a chemical test called the polymerase chain reaction. While swab tests can detect an infection even in its early days, antibody-based tests can only detect anybodies after the body has mounted a considerable immune response against the virus — often five to ten days after the virus has infected the body. As a result, antibody-based tests could yield negative results in the initial days of infection (a.k.a. false negatives).
The objective of testing for COVID-19 is to provide early therapeutic intervention (virtually nothing needs to be done for people with mild or no symptoms) as well as to help isolate positive patients before they spread the infection to others (the average period of time in which a patient infects others is called the serial interval). Early detection also helps public-health officials track down and quarantine the patient’s contacts, and who can be tested later for the presence of an infection. But since antibody-based tests aren’t good at diagnosing infections early, they fail in this objective.
The Indian Council of Medical Research (ICMR) guidelines recommend antibody-based tests in hotspot — areas with an unusually high case load or where a cluster outbreak has occurred — but even there, they can be of little use as every positive antibody test has to be validated by a swab test (if nothing else, to account for the fact that a ‘positive’ could be the result of an older infection from which person has already recovered). Similarly, a ‘negative’ antibody-based test wouldn’t rule out COVID-19 as a person may be in the early stages of infection, so the person has to be kept under quarantine and tested again after 10 days.
Once antibodies enter the blood, they persist for a long time, so antibody-based tests detect current active infections as well as previous infections that have been resolved. Randomly testing the population for antibodies could therefore give us a true picture of the actual number of total cases (active as well as cured). This in turn gives us the true denominator to estimate the infection fatality ratio (IFR) of COVID-19. The case fatality ratio is defined as a fraction of the confirmed cases, so many people infected with COVID-19 but displaying only mild or no symptoms are missed. The IFR includes them as well.
In turn, a more accurate IFR will help assess the effectiveness of public health interventions. The existing case numbers are not very useful because ICMR has changed the testing criteria so many times. So as we test more people, more people are likely to test ‘positive’, giving us the impression that the virus is spreading faster; however, the rate of transmission could really be falling.
In short, the number of new confirmed cases during two different testing regimes are not comparable. On the other hand, the number obtained by random, antibody-based tests doesn’t depend on the testing regimen, allowing officials to compare the rate of development of new infections before and after a particular intervention, and thus help evaluate the effectiveness of the intervention.
More accurate data about the total number of infections, the IFR, etc., will also help epidemiologists build better models of how the epidemic will evolve, which can then inform policy.
Large-scale antibody-based testing will therefore help us understand where we are in the epidemic — and the importance of this can not be overstated.
However, even without antibody-based tests, the actual number of deaths can give us a crude but reasonably accurate data of past infections. It takes an average of 24 days for a person’s health to deteriorate from infection to mortality (‘infection to symptoms’ in five days + ‘symptoms to mortality’ in 19 days). So the actual death numbers can yield an estimate of the total number of infections, if the IFR is known, about 24 days earlier.
Widespread testing onboard the Diamond Princess cruise ship and in Iceland and South Korea suggest the virus’s IFR is close to 1%. Accounting for the age-wise distribution of the different populations as well as cases, we can safely assume the IFR in India is 0.8-1.2%. This in turn would mean that if 40 people died on April 25, approximately 4,000 people were infected on April 1 (assuming an IFR of 1%). On the other hand, antibody-based tests allow us to estimate the actual number of cases with a lag of just one to two weeks (average 10 days).
But this by itself is not really a game-changer, is it?
This leaves us with the issue of granting ‘immunity certificates’ to persons cured of infections so that they may rejoin the workforce. Assuming that antibodies provide lifetime immunity or immunity for a substantial period at least (which is a big assumption!), we will still need to test a large enough group. For example, considering India only intends to test one lakh per day, and assuming all of them will test negative for the virus as well as have acquired immunity to it, we will still need 4,000 days to test the country’s workforce of 400 million people.
Now, assuming the IFR is 1% and the actual number of deaths is 350, and that all 350 died today, there will have been around 35,000 infections 24 days back. (This number underestimates the actual number of infections but not by much.) Assuming the lockdown increased the case-load doubling-time to eight days, we will have 320,000 infections and 3,200 deaths. If the double-time was six days or 12 days, we’d have 640,000 and 160,000 infections, respectively, and 6,400 and 1,600 deaths.
It should be clear at this point that antibody-based tests will help understanding the epidemic better, and the data so obtained will help build better models of the epidemic which will in turn help improve policies. However, the tests by themselves can only modestly improve our chances of besting the epidemic itself. Traditional swab tests of suspects, isolation of those who test positive, quarantine of possible contacts and community-level interventions to improve physical distancing will remain the key to fighting the epidemic, at least until we develop an effective vaccine.
Dr Vipin graduated from AIIMS, New Delhi. He is particularly interested in epidemiology and biostatistics.