The true prevalence rates are most likely to be lower than the rates reported, because of the inflation of false positives on mass screening.
Even though Covid-19 has shut down schools and colleges, the virus has stimulated mass education programmes that cover the globe through conventional media as well as social media, besides scientific journals. Starting with virology and clinical medicine, this educational upsurge has inundated knowledge domains ranging from ecology, epidemiology, statistics, mathematical modelling and public health practice to behavioural sciences, economics, health technologies, drug trials and international relations. The latest course to be launched, courtesy of this viral don, is one with a high focus on immunology. The course is replete with reports pouring in on antibody-based serological surveillance, vaccine trials and predictions of herd immunity.
Even as Spain claimed to have conducted the largest population survey for Covid-19 antibodies, Delhi and Ahmedabad (Amdavad) released reports of their surveys of over 20,000 and 30,000 persons respectively. Indeed, the Amdavad Municipal Corporation report points out that, in terms of antibody tests per million population, city’s sample survey outdistanced all international and Indian surveys, including what it pointedly describes as a “minuscule” ICMR survey.
The Delhi survey reported an antibody prevalence of 23%, while Ahmedabad reported 17%. While acknowledging these rates to be high in comparison to similar international surveys, both reports clearly caution that herd immunity is still far away. The countrywide Spanish study reported a prevalence of 5%. Between May and June, Tokyo reported only 0.1%, cities in California reported between 2-4.5%, London observed 17% and New York City variably reported 6.7% and 20% at different times. Thyrocare, a private lab in India, claimed that its non-randomly collected blood samples showed a 9% antibody positivity rate in the country.
Every laboratory test is conventionally judged for its diagnostic accuracy on the basis of its ‘sensitivity’ (ability to correctly detect the disease of interest in all truly affected persons) and ‘specificity’ (ability to correctly exclude those without the disease). These ‘true positive’ and ‘true negative’ rates are the test characteristics that are usually reported, based on comparison to a clinical, laboratory or composite gold standard which is judged to be the most accurate at that time. Thumb rule number 1 is that higher the sensitivity, lower the specificity. There is usually a tradeoff. We usually try to select tests, with the least margin of tradeoff.
Unfortunately, while these attributes well describe how the test performs in a mixed group under certain conditions, they do not tell us how well the test identifies the probability of the disease in a specific individual. The test characteristics are called positive and negative ‘predictive values’. Unfortunately, these values are very susceptible to the rate of actual prevalence in the population. Thumb rule number 2 is that lower the true prevalence of disease in a population and larger the sample surveyed, more the number of false positive persons who are wrongly labelled to have the disease. That in turn affects the reported prevalence.
Suppose a test has been evaluated in a hospital or laboratory with 100 persons known to have the disease (by the gold standard criteria) and 100 persons who do not have the disease (by the same criteria). That gives a prevalence of 50% in a sample size of 200. If a test with 95% sensitivity and 90% specificity is employed, the positive predictive value will be 90.4%, which gives the probability that a person who tested positive actually has the disease.
Then, the same test is employed to screen 20,000 persons in the general public where the true prevalence in that population is only 10%. The test performs similarly, but the truly infected cases are only 2,000, out of whom 1,900 are correctly identified (95% of 10% of 20,000). Of the remaining 18,000, the same specificity rate of 90% now yields 1,800 false positive results. If these are combined with the true positives as ‘cases’ identified in the survey, the estimated prevalence rate rises to 18.5% (3,700/20,000) even as the positive predictive value falls to 51%.
This problem is compounded by the fact that false positive tests are known to occur due to cross-reactive antibodies from other coronaviruses, including those which cause the common cold. We do not know what their prevalence is in the population of Delhi or Ahmedabad. However, the true prevalence rates are most likely to be lower than the rates reported, because of the inflation of false positives on mass screening.
That applies to the Thyrocare report too. Even with this noise obscuring the signal, antibody surveys are still useful for comparing different populations at a similar time, or the same population at different times. However, the problem of false positive results deters us from labelling individuals as infected or immune. Hence, the WHO warning against using antibody tests as ‘immunity passports’ for selectively permitting people for re-entry to their worksites.
The starting estimates of Herd Immunity Threshold (HIT) for the Covid-19 virus were 60-80%. Despite some optimistic estimates of 20-40%. the HIT is unlikely to be lower than 50%. That is far away, unless we give a free rein to the virus to race across the country. If we do that in a rush to embrace herd immunity, it carries the human cost of a cascade of cases and deaths suddenly descending upon us. Some argue that all persons are not susceptible and herd immunity thresholds should be estimated only for the rates of infection and transmission among the susceptible persons. Epidemiological and immunological “dark matter” has been proposed to suggest that such a susceptible fraction of the population may be small.
Others suggest that ascribing equal probabilities of infectivity to all persons in a population does not take into account the dynamics of ‘super spreader’ individuals and events. However, since we do not yet know how well these dynamics apply to this new virus, we have to assume that a large proportion of the population is susceptible, vulnerable and not yet within easy reach of herd immunity. A resurgence of the epidemic in many countries which had declared victory suggests that herd immunity may have been a mirage and the oasis of fulfilled hope is still some distance away.
It must also be recognised that herd protection is offered by the many, who were infected, recovered and are now immune, shielding the minority who have not been infected. However, if any member of that uninfected minority travels to another location where only a small fraction of the population has been exposed so far, the new herd cannot offer protection to the visitor. So, even if Delhi crosses HIT, an uninfected person from Delhi cannot feel safe in Raipur or Ranchi if they are nowhere near that threshold.
The other problem with Covid-19 antibodies is that their levels have been reported to decline by three months. What will that do to herd immunity or vaccines? Is there a relationship between antibody protection time and virulence of the virus? Antibodies to relatively innocuous common cold-causing coronaviruses disappear in a few weeks while the sinister SARS-CoV-1 had antibodies lasting 2-3 years? The SARS-CoV-2 intermediate in virulence. Is that why antibodies that last a few months and not weeks or years? We do not know the answers to these intriguing questions of how human-virus battles shape the strategic immune response. It has also been suggested that both infection acquired immunity and vaccine evoked immunity may not only depend on the antibodies produced by B lymphocytes alone, but maybe served better and longer by cellular immunity marshalled by thymus-derived T lymphocytes. How these two Gemini twins of immunity fare in our defence against Covid-19 is still being unravelled.
Even as the immune mechanisms are under intense study, excitement about vaccines continues to mount. Over 200 candidates vaccines have entered the race, with some of them moving to human trials. They will have to provide evidence of safety, efficacy and duration of protection. Some of the under-trial vaccines are reported to have not only evoked strong antibody formation but also stimulated T cell mediated immune response. Though the usual gestation period for new vaccines to complete the journey from molecules to markets is usually long, with rigorous multi-phase evaluation required in animals and humans in between, there have been claims that the vaccine will become available for public use anywhere between September 2020 and March 2021. As with everything else with this novel and nasty virus, we cannot lay sure bets on this, but wait in hope.
The author is President, PHFI and author of the book Make Health in India. Views are personal