Representative image: Testalize.me/Unsplash.
Two weeks ago, the Indian Council of Medical Research (ICMR) published the results of its controversial seroprevalence survey, leading to more questions than answers. The survey reported a seroprevalence – or fraction of antibody-bearing people in the population – of 0.73% among the 28,000 people surveyed in early June. It has been hard to interpret these results.
ICMR said the survey’s findings mean India may have had up to 6.4 million cases of COVID-19 by early May, implying, in comparison to the official tally, that we likely missed 81 other cases per confirmed case.
Other experts compared the relative low seroprevalence to other surveys (e.g. Delhi’s second serosurvey found 29% of participants had antibodies) to conclude that herd immunity for the country remains a distant dream, as infeasible today as it was when the outbreak began.
A bulk of the confusion here is ICMR’s doing – considering the confused reporting, data suppression and obfuscation. However, a considerable part of the problem also lies in the scientific context in which antibody tests operate, specifically what they are and aren’t designed to tell us.
What are seroprevalence surveys?
These are studies designed to identify how many people who were never diagnosed with the disease had antibodies to the disease in their blood.
In order to be indicative, it’s important to get the specific population to be studied right. If the results are to be widely applicable, the subset of the population selected for the survey has to be large (typically in the thousands) and diverse (with respect to the gender, age, ethnicity, place of residence and socio-economic determinants). The closer the study’s population represents the actual population, to which the results are to be applied, the better.
The purpose of a serosurvey, according to the WHO, is to determine the level of susceptibility of a population to a disease. To interpret how close we are to reduced community transmission for a disease (due to herd immunity), we have to know what the threshold for that particular disease is.
For COVID-19, we have no clue. For measles, researchers have estimated this figure to be above 90%.
These surveys are also of limited value in isolation. They are designed to determine temporal trends, i.e. how the value changes over time, as a reflection of disease control efforts. Against the background of various issues being raised with the design and interpretation of antibody testing exercises, the most important use of the first ICMR serosurvey is arguably as baseline data for comparison with the second serosurvey, which was reportedly completed a few days ago.
How accurate are COVID-19 antibody tests?
Not very. There are many reasons for this. A Cochrane review of nearly 16,000 patients from around the world showed that the sensitivity – i.e. the likelihood of a test correctly reporting a ‘positive’ – varied between 30% in the first week after infection to 96% in the fourth week. Since around 15% of COVID-19 patients have no symptoms at all, determining where within this time frame they lie is virtually impossible.
False positives are rare but have been known to occur.
Another obstacle is scale: even with a sensitivity of 99%, when testing 28,000 people, up to 280 patients who have SARS-CoV-2 antibodies will test ‘negative’.
Importantly, the mathematical accuracy of these tests hinges on the prevalence – the number of people who already have the disease – in the population being tested. The predictive value or accuracy of a positive test that is 99% sensitive is 45% when prevalence is 1%; 81% when prevalence is 5%; 90% when prevalence is 10%; and 95% when prevalence is 20%.
Although this sounds complex, the explanation is simple: the accuracy of all tests depends on the ‘pre-test probability’, which is the statistical likelihood that a patient within a population has the disease. This is intrinsically linked to prevalence.
What is the clinical interpretation for antibody tests?
The principles on which these tests are based are straightforward. When accurate, these tests measure antibodies once their levels have crossed a certain threshold, and remain above this threshold and confer immunity against the disease.
For COVID-19, we have found exceptions at every level. False negative results show up when the disease is ‘young’, or later on when the antibody levels are falling.
While studies have shown that the volume of antibodies can rapidly fall within three months of a COVID-19 infection, some experts have argued this is not necessarily a sign of diminishing immunity; it could also represent a normal transition from early immunity to a more prolonged and protracted late response, which relies on immune cells.
Perhaps, most importantly, antibodies detected with most commercial tests don’t test for neutralising antibodies, which are the specific antibodies that bind to the virus and clear it from the blood, preventing re-infection. With researchers having documented the first cases of reinfection among healthcare workers, we know that this is a potential outcome in those without demonstrable levels of antibodies, even if it’s uncommon.
Antibody tests can, however, be very useful in the right setting. They are cheaper and much quicker than RT-PCR tests, so they’re more preferable in emergency settings (i.e. before emergency hospital admission or surgery) and to screen large groups of people.
A true positive antibody test result implies that the person did have a recent SARS-CoV-2 infection. However, it fails to tell us if they have fully recovered and are no longer infectious, or if they are safe from reinfection.
True-negative results have much lower clinical value. They could represent ‘no infection’, ‘early infection prior to an adequate antibody response’, a ‘previous infection with antibody levels having declined beyond the testing threshold’ or ‘previous infection with no antibody response’.
So within the boundaries of what we currently know about the human immune response to COVID-19, antibody tests are more useful as epidemiological tools than as clinical ones.
How far are we from ‘immunity passports’?
Briefly setting aside the moral, ethical and legal issues associated with selectively allowing those who have recovered free passage and travel, to revive economic growth, the scientific premise is weak.
Data obtained recently has indicated that different bodies activate their immune response to COVID-19 to different degrees. Some researchers have speculated that vaccines may also have different efficacies according to geography and ethnicity. So it’s important for us to rigorously and exhaustively test all vaccine candidates.
If this seems laborious – that’s because it needs to be. The race to a COVID-19 vaccine is more a marathon, and the best way forward is to complementarily use different vaccines, protect the vulnerable population (including healthcare workers), expand healthcare capacity and sustain social responsibility.
Dr Narayana Subramaniam is a head and neck surgical oncologist at the Mazumdar Shaw Medical Centre, Bengaluru.