Seroprevalence Study in Mumbai: In Conversation with TIFR Scientists -- Part 2
Meena Kharatmal: With this introduction, we would like to know some key insights from the seroprevalence study conducted in Mumbai. What risk factors were considered while conducting the study? How does the analysis take into account false positives and false negatives?
Ullas Kolthur: False positives and false negatives are a big concern. In fact, there are multiple ways of doing serological tests. Many people would know that there is something called the rapid antibody test. The rapid antibody test kit consists of a strip, with a prick on a fingertip and a small amount of blood it will tell you if you are positive or negative for the antibodies. There is something called an antigen test which also is based on the fact that you are assessing whether you have circulating viral antigens. You use antibodies to detect the antigen. The third one is the regular serological test which is used in most of the assays for multiple pathogens and antibodies called an ELISA (enzyme-linked immunoassay) test. There are variants of that called as Chemiluminescence assays.
Sensitivity and specificity is an issue because it is a new virus. People have tried to generate reagents that could detect both the antigens and the antibodies and not surprisingly, initial efforts probably gave them the reagents that had only limited specificity and sensitivity. And as we all know, as times progresses, we all learn from the gaps that existed in the field. The kit that we used is CLIA (Chemiluminescence immunoassay) from Abbott. It was chosen because it was shown to have close to 100% specificity. This means that it does not recognize false negatives and it does not cross react with other antibodies which could be raised for any other diseases. Now typically this is done by companies where they take sera from multiple people infected by various pathogens. Of course this is done in what is called as a serobank, wherein they already have created a pool of blood samples from multiple patients and these are tested. The other aspect is sensitivity. We need to know at what level is this kit able to detect the antibodies. We need to know if the kit is able to detect only very high levels of antibodies, which means you will miss out on some positives.
So false positive and false negative are an important aspect. Specificity takes care of false positives. We wanted to ensure that we are not detecting something which is not raised because of antibodies that are specific to COVID-19. False negatives are important for sensitivity. We used this kit which was independently validated by the Public Health England (PHE). Public Health England's independent validation showed the test kit to be close to 93% sensitive, which is actually one of the better kits that is available right now to do this test.
Now having said that, we have also gone back and checked the reports of PHE. The Public Health England report has shown that even seasonal coronavirus antibodies are not picked up by these antibodies, which means it is very specific. In our context, we also ran some pre-COVID-19 samples and we showed that there is close to 97% specificity which means it is not picking something that is endemic in the Indian population. Or in other words, it also means that this kit is only picking up antibodies which is specific for COVID-19. Given it's sensitivity (93%) (Sandeep will elaborate more on that), it indicates that the prevalence is actually higher since it is likely that we are missing out some people who have lower antibodies and the kit is not able to pick it up.
Sandeep Juneja: So we found the results showing 50-60% cases in slum areas that turned out to be positive and this was interesting because it is not just the areas which have lot of infection, but even the emerging areas now. Of course that said, emerging areas have ‘emerged’, but it also says that numbers are plateauing at 50-60% with our test. That is reassuring, that we are beginning to see trends towards herd immunity. We will know more about this in the second round/phase of the study. In the non-slum areas, we found again very similar numbers, from 11/12-18%. That is the big picture. It shows that non-slums are less crowded, and it is easier to maintain social distancing. Slums have all these points where they have to crowd, they have common toilets, common water facilities, while in non-slums people can stay separate. So the social distancing is working and it is slowing down the infection. That we learn from the data.
It is very interesting that when we conducted the survey we were coming up with these numbers. The data were also coming to us from other parts of the country. Sweden with all its policy of minimal regulations, had a prevalence of only about 3%; UK and USA had 5-7%, Spain also showed a low number. Accordingly, our impression was that we were expecting similarly low numbers for Mumbai and suddenly this data of a larger percentage was coming up. It is quite remarkable. We also heard the Delhi numbers of having a bigger percentage (23.48%). Although, we knew the Mumbai numbers before that. So the numbers themselves were quite shocking.
Now what is interesting is that the slums do have much higher prevalence. One thing we did, we asked people about co-morbidities and we got a sense from them that it was not as stark as we expected. So suppose somebody reports in the slums that they have hypertension (blood pressure), that does not mean that, that cohort of population reports a much higher prevalence. That number comes out to be slightly higher 62-63%. This indicates that in terms of getting the infection, the co-morbidities are not playing such a big role, whether for diabetes or blood pressure. On the other hand, it could also be that once you have infection and have severe cases, having these co-morbidities would play a bigger role. That of course we know already now.
The other interesting part is about family size. When we interviewed people, we asked them about the size of their family and what age buckets do they lie in. That gave us an estimate of family size-wise distribution, age-wise distribution, population distribution in slums vis-a-vis non-slums. This is interesting, because we are seeing much higher prevalence in the slums compared to non-slums. From the data of reported cases, reported fatalities, this suggests that slums are actually seeing lesser reported cases per person, lesser fatalities compared to non-slums. One interpretation for this can be that may be the slums are younger. Well, we know already that younger people do not suffer severe consequences that older population does. Age wise distribution broken down by slums and non-slums is not easy to obtain. Our collected data on family sizes corroborates that slums have much younger population compared to non-slums. So this is an important explanation of the lower infection fatality rates observed in slums.
Uma Ramakrishnan: Thanks for that. First of all, as scientists we look for validation through these data, some trends which you may have expected. There is quite a lot of data on differences between males and females. Is that something which you also saw in terms of validating those trends from other places?
Ullas Kolthur: Yes, differences between males and females is interesting. I do not know if Delhi reported the differences between males and females. I certainly know of a very recent Pune data (51.5%) which showed the differences between males (52.8%) and females (50.1%). We found that depending on whether we are looking at just percent positivity or the population at this prevalence, we could say that females were more infected than males. That was a surprise, I do not think anybody would have expected it.
Now this could be because of multiple reasons. It could mean that women, at least in the population that we surveyed, intermingle more, since they are taking care of household chores. This indicates that it is quite likely that the chance of getting infection for a woman in Mumbai is higher. However one cannot rule out whether it has any underlying biological context, which is quite possible. In fact, it might raise interesting questions about whether T-cell versus B-cell immunity might be different in males and females and across ages. Such studies raise very interesting hypotheses. While all of them need not be true, they give interesting pointers. So I would say, it could be a reflection of social behavior. It could also have an underlying biological context.
Sandeep Juneja: I do not have much more to add to that. It is true that, we saw prevalence of the order of 53% in males and of the order of 59% in females in slum areas and is statistically significant in Mumbai. Certainly it calls for deeper understanding.
Uma Ramakrishnan: The fact that these people were in lockdown, they were actually not leaving. For example, if there was more movement, there could have been more spreaders, more prevalence. It was higher, considering that they were at home and in lockdown.
Sandeep Juneja: I think, the issue here is, can you really lock people down in a slum area in Mumbai, given the density? You really cannot. People are still going out to these common places (points), where the mixing occurs, in fact it is more or less uniform mixing. Even if you divide the population by age brackets, you see similar prevalence. It seems like there was a lot more of intermingling going on. Even with our policies of containment, we contain a region in Mumbai. But within the region, there is lot of intermixing going on. On the other hand, we exactly see this point that non-slum areas with much lower numbers tells that what would have happened if there was no lockdown. Then there would have been much higher intermingling and much more infection spread.
Discussion of findings and other studies, continued in Part 3.
About the People:
Dr. Ullas Kolthur is a Professor at the Department of Biological Sciences of the Tata Institute of Fundamental Research, Mumbai. His research interest is in the area of cellular metabolism and energetics.
Dr. Sandeep Juneja is a Professor and Dean at the School of Technology and Computer Science, Tata Institute of Fundamental Research, Mumbai. His research interests lie in applied probability including in sequential learning, mathematical finance, Monte Carlo methods, and game theoretic analysis of queues.
Dr. Uma Ramakrishnan is a Professor at the National Centre for Biological Sciences (TIFR), Bangalore. Her research investigates population genetics and evolutionary history of mammals in the Indian subcontinent, including work to save India’s tigers.
Ms. Meena Kharatmal is a Scientific Officer at the Homi Bhabha Center for Science Education (TIFR), Mumbai. Currently she is contributing articles, resources for the CovidGyan. She is also trying to complete her PhD in the area of Biology Education.
This interview was recorded on 19th August 2020. Since then, the preprint on the findings of the Mumbai seroprevalance study is available as a report published on TIFR website.