Serological surveys for Covid-19 are being conducted around the globe to understand the rate of prevalence of coronavirus antibodies in people. Blood samples collected from the general population and tested for IgG (Immunoglobulin G) antibodies. If a person is IgG positive, it means they had been infected with the coronavirus in the past, and is protected from future infection as long as a substantial level of the antibody remains in the body.
A major reason for conducting such surveys is the hope that a substantial proportion of the population has already been infected with the virus (possibly in an asymptomatic way) and so is immune to Covid-19. Epidemiologists call this state “herd immunity”
“Herd immunity occurs when enough people become immune to a disease to make its spread unlikely,” says the Harvard Medical School website. “As a result, the entire community is protected, even those who are not themselves immune. Herd immunity, it explains, “is usually achieved through vaccination, but it can also occur through natural infection”.
A stage of herd immunity would allow lockdown and other restrictions to be eased without fear of substantial numbers of people being infected.
On June 11, The Lancet published the results of population-based serosurveys for anti-SARS-CoV-2 IgG antibodies in the Swiss city of Geneva for five consecutive weeks between April 6 and May 9. It was estimated that for every reported confirmed case, there were 11·6 infections in the community.
The estimated seroprevalence for the first week was 4·8% (sample size 341), which increased to 8·5% in the second week (sample size 469). It was 10·9% in the third week (sample size 577), 6·6% in the fourth week (sample size 604), and 10·8% in the fifth (sample size 775). Thus, so far Geneva was concerned, the epidemic was far from coming to an end by means of fewer susceptible people in the population during that time period.
While everybody was expecting an increase in the proportion of seroprevalence, I was, however, particularly amazed at the drop of the estimated seroprevalence from 10.9% to 6.6% from the second week to third week in Geneva. Biologists and medical experts did not raise serious questions about it.
But, being a statistician, I cannot easily explain this by “sampling fluctuations”. This is an event that is very difficult to explain by “chance factor” because such an event has a probability of less than half a percent.
Similarly, the proportion of seropositives in the Midlands in the UK dropped from more than 6% in the 17th week to around 5% in the 20th week), which is again contrary to popular intuition of increasing seroprevalence.
The natural question is: what can and can’t be known from such sero-surveys?
Looking for answers
The Centers for Disease Control and Prevention in the US has an overarching strategy for learning more about how many people have been infected with SARS-CoV-2, the virus that causes Covid-19 and how it is spreading through the U.S. population. It is categorically mentioned that the questions the CDC wants to answer through serology surveillance include how much of the US population has been infected with the virus causing Covid-19 (SARS-CoV-2), how is this changing over time, and how long antibodies can be found after a Covid-19 infection.
However, the questions the CDC cannot answer through Serology Surveillance include how much of the US population is immune to Covid-19 and not able to get infected again, how many antibodies are needed to protect someone from Covid-19, how long someone with antibodies will be protected from Covid-19, can someone be re-infected with Covid-19, and most importantly, can people with antibodies return to work?
In March, Germany found that 14% of Gangelt municipality had antibodies against SARS-CoV-2, where swab tests showed that 2% were sick. According to a rapid serology test in the severely-hit Spain on April 27, only 11.3% of people in Madrid and 7.1% in Barcelona had developed antibodies against Covid-19. Around the same time, 12.3% people of New York and 9.9% of Boston had Covid-19 antibodies.
Does that mean that we are far from herd immunity?
In more technical terms, herd immunity can be better described by the “basic reproductive number”, R0 or R naught, which is the average or expected number of people who get infected by an infected person in her infected’ period.
Specifically, (R0 -1)/R0 is the fraction of individuals in the population who need to be infected to achieve herd immunity. If R0 is 2 (that is, two people persons are infected, on an average, by an infected person), 50% of the population will need to have been infected to achieve herd immunity. If R0 is 3, two-third of the population is needed to be infected before achieving herd immunity. The epidemic will be stopped after the herd immunity is achieved, but people (who are not immune) will still be infected.
How are India’s sero-surveys? In an article in the Indian Journal of Medical Research on June 20, a group of scientists outlined the sampling strategy, sample size, sampling method, study procedures for detecting SARS-CoV-2-specific IgG antibodies using an ELISA-based test in the country.
This serosurveillance is designed as a repeated cross-sectional survey of adults aged 18 years or more. This is in contrast to the Geneva study where subjects were aged 5 years and older.
India’s sero-survey results released on June 10 asserted that only 0.73% people in areas outside containment zones had antibodies. This leaves the major proportion of Indians susceptible to the disease. In a similar follow-up study in West Bengal at the end of June, it was observed that about 14.4% (57 out of 396) of Kolkata residents have developed Covid-19 antibodies. However, the volume of positives was around 1% in other districts of the state.
The sero survey in Delhi to map exposure to coronavirus conducted at the end of June and early in July observed that the positivity rate is 22.86% for the IgG antibodies. Thus, it might safely be guessed that the volume of people infected might now be within 10%-25% in most big cities of the world. In rural and semi-rural parts of India, however, it might still be around 1% mark. Thus, we might be far from achieving herd immunity.
But, the more important point is: how long such immunity would last? In a small study conducted in China in June on 37 asymptomatic and 37 symptomatic patients and published in the journal Nature Medicine, a decline in antibodies within two or three months of infection was observed. Another recent study at King’s College London, longitudinally evaluated antibody responses in SARS-CoV-2 infection of more than 90 patients and healthcare workers, and found steep drops in patients’ antibody levels three months after infection.
If this is true, the sero-prevalence in a community might not become too high. In fact, the Chinese and British studies offer some explanations for the drop in sero-prevalence in Geneva. As immunity does not last long, it is possible that some of the people who were immune in the previous survey might have lost their immunity by the time the next survey was conducted. And this might put the usefulness of sero-surveys to find proportion of immuned people in serious question – everywhere in the world.
Atanu Biswas is a Professor of Statistics at the, Indian Statistical Institute, Kolkata.