We note the intense interest in the issue of antibody testing for SARS-nCov-2 after the article by the group of pathologists and virologists (Drs Vardas et al), and the responses from Dr Mendelsohn and Prof Mahdi, along with many responses in your comments sections. Dr Kew and I have extensively investigated this issue over the past few months in order to assist employers who have been clamouring for these tests. 

Our initial impression was also that the regulator, the South African Health Products Regulatory Authority (SAHPRA), was remiss in not approving the POC antibody tests when they could apparently be so useful for those wanting to return to work after a quarantine period, or when coming out of lockdown, and as an epidemiological tool to establish and track the extent of the epidemic in the population more widely. However, our research has clearly shown that there are serious limitations of these tests for both diagnostic and epidemiological purposes. This is so for both the rapid POC antibody tests and for the laboratory based enzyme-linked immunosorbent assay (ELISA) antibody tests. 

We have reviewed the use of rapid point of care serological tests for SARS-COV-2 antibodies with a particular focus on their utility in first contact care settings in occupational health clinics. What follows is a summary of a guideline we have developed to assist employers. This guideline can be generalised to all primary healthcare settings, including testing for personal use.

There are two types of test for Covid-19, a viral test, which tells you if you have a current infection, and an antibody test which tells you if you have had a previous infection. Viral tests are usually performed using a swab which is wiped against the back of the throat, or back of the nose. The test can also be done on sputum and fluids obtained directly from the lungs in a hospital setting. Special laboratory techniques (PCR) are applied to detect the presence of viral nuclear material (RNA). Antibody tests are performed on the blood. These may be performed in a laboratory by trained technicians, and also at a clinic or doctor’s surgery or in one’s home, using “rapid point of care” devices.

Antibody tests have been promoted for use in two key applications:

To screen groups (such as the workforce) for the presence of SARS-CoV-2 antibodies, thereby enabling:

• The estimation of Covid-19 prevalence;
• The identification of individuals who have recovered from Covid-19, and who may therefore be at zero or low risk of (re)infection viz. immune.

To be able to exclude Covid-19 in individuals with current symptoms.

However, these applications are highly problematic.

First, all test kits to date, including the Roche test which claimed to have 100% sensitivity and specificity, have poor test performance. This means they have too many false positive and/or false negative results. (FDA approval of antibody test kits was granted as emergency permission without following usual evaluation, and has been criticised. Some tests have already had their approval cancelled.)

The rate of false positives when using rapid point of care devices is very high (around 70%) at prevalences likely to be encountered in clinics or in the community.

Weak antibody responses are likely to be missed by rapid point of care tests, and constitute false negatives.

Furthermore, the tests are unable to determine whether the types of antibodies detected actually confer immunity.  

Second, various types of antibodies to the virus do not always appear in sufficient quantities, if at all, and do not always persist for significantly long periods after infection. 

The detection rate for antibodies during the first five days following symptoms is very low (on average 32%). This improves to 65% after a week, but only reaches a more reliable 90% after 3 weeks. The timing of the appearance of antibodies is too delayed and also results in false negatives.  

Third, and perhaps most important, another key issue that has been missed in all the correspondence on this issue is that the nature of the immune response to SARS-cov-2 is not at all clear. Epidemiological surveys to determine population prevalence of past infection by the virus have shown surprisingly low proportions testing positive (less than 10%), even where the epidemic has been raging. Humoral immunity in which B-cells produce antibodies is not the only form of immunity to this virus. Cellular immunity (via T-cells) is also likely to play a critical role, and this cannot be determined by antibody tests. There are also other forms of immunity, for example in children or other people who do not get Covid, which are unknown, and have been referred to as “dark matter”. These may have to do with biochemical receptors or other unknown mechanisms such as previous immunisations resulting from vaccination with non-specific effects, or previous infections with other coronaviruses which cause the common cold. 

When all of these drawbacks are considered, it is clear that the utility of antibody tests for clinical diagnosis, epidemiological purposes, or relieving anxiety for workers, employers and members of the public at large is severely limited, and that they may potentially give rise to confusing information with attendant safety hazards. There may however be a limited research role for these tests in population surveys and clinical studies, for understanding the role of humoral immunity in Covid-19 infection.

The best approach we have to limiting the damage of Covid-19 is to follow the basic prevention rules and not to put false hope and trust in flawed antibody tests. DM/MC

• Dr Greg Kew is an Occupational Health Specialist in private practice and is an Honorary Lecturer in the Centre for Environmental and Occupational Health Research at the Department of Public Health and Family Medicine at University of Cape Town. Emeritus Prof Jonny Myers is an Occupational Health and Public Health Medicine Specialist in the Centre for Environmental and Occupational Health Research at the Department of Public Health and Family Medicine at University of Cape Town.