Three months after Balram Bhargava, the Director General of the Indian Council of Medical Research, gave skimpy details of the national sero-survey carried out by ICMR institutes at a press briefing on June 11 (“COVID Cover-up”, Frontline , July 3), the results of the same have been finally published in The Indian Journal of Medical Research (IJMR), brought out by the ICMR. The survey was carried out from May 11 to June 4.
The take-home message of the results (notwithstanding their many limitations and issues that are unclear) is that in early May, for every reported infection (confirmed case), as many as 82-130 cases had been missed, which is a huge factor pointing to an extremely lax testing strategy early in the epidemic. As the paper notes, “the high infection-to-case ratio (ICR) in India could be on account of the prioritisation of testing among symptomatic or the variability in testing rates across the States”. It also acknowledges the fact that this prioritisation of allocation of available testing capacity would have missed many asymptomatic and mild infections. It was already known in April that asymptomatic cases could be as much as 40 per cent. Now we know that it is much higher. A recent estimate by the Integrated Diseases Surveillance Programmme (IDSP) put this figure at 80 per cent, while the Union Health Minister, at the beginning of this monsoon session of Parliament, stated that 92 per cent of the cases in India were asymptomatic.
Given that the primary objective of any sero-survey is to estimate the prevalence of infections in the initial stages of a growing epidemic in the general population so that appropriate strategies of interventions to contain the spread are adopted early enough to prevent its escalation to alarming levels, a four-month delay in the release of survey data means that this aim is totally lost. The published sero-survey paper, though, says in conclusion that “…the findings of the sero-survey indicated a low prevalence of SARS-CoV-2 infection in the general population in early May. As most of the population remains susceptible to infection, our public health strategy needs to plan for an inevitable increase in transmission.”
“This pandemic is an avalanche, gathering momentum as it grows,” the eminent virologist Jacob John had written in an editorial commentary in Current Science way back in March. “We are moving a step or two behind the cusp of advance, instead of being wise and proactive, moving two steps ahead. The crucial expertise of epidemiology intelligence is missing in India’s health management system,” he wrote. Unfortunately, that recommendation fell on deaf ears at the higher executive level. Though it came only in June, even the conclusion from the findings of the sero-survey seems to have not been taken seriously and there could also be not enough public pressure because the data were under wraps for over three months. The continuing surge in infections across the country should not be a surprise at all because what we are seeing today is the multiplier effect of those large number of infections that were missed because of the low level of testing.
A cross-sectional community-based sero-survey, like this ICMR’s national survey, is intended to estimate the prevalence of infection during an evolving epidemic by testing a randomly sampled population for the presence of immunoglobulin G (IgG) antibodies in the people. Since IgG antibodies form only about 7-15 days after infection, this seropositivity rate (for IgG antibodies) gives an indication of prevalence of infection in the sampled population in the recent past. From that data, one projects it to an entire population to arrive at an estimate of prevalence of infection at a given point of time slightly in the past for a whole region—a country, for example—to evolve appropriate strategies of intervention for infection control.
At the June press briefing, it was merely stated that the seroprevalence seen in the survey had indicated that 0.73 per cent of the population in the surveyed districts had past exposure to SARS-CoV-2, and since the analysis had been completed only for 65 (of the 71) districts, it was not proper to extrapolate that fraction to the entire country’s population as yet. The details of the survey results now quantify the sparse information that was given out in June.
The published results give seroprevalence data stratified according to the level of incidence—in terms of reported cases—as on April 25. On the basis of the April 25 data of confirmed cases, the sampling template was drawn up and districts were classified for the purpose of sampling into four categories or strata as ‘Zero’, ‘Low’, ‘Medium’ and ‘High’ incidence (Table 1 and box). According to the paper, the pooled and weighted overall seroprevalence (after adjusting for sampling design corrections, appropriate weights for sampled populations and testing accuracy) was 0.73 per cent, which is same as the figure given at the briefing! This is surprising because the number of districts is now given as 70, as against 71 mentioned then, and a comparison of Table 1 and Table 2 (the figure 83 in Table 2 is due to wrong arithmetic) indicates that the omitted district is from a high-incidence category. It is not clear why a high-incidence district was left out in the final analysis.
Further, according to the published results, the survey had sampled 28,000 individuals (from 30,283 households, with one from each household) from 70 districts in accordance with the laid-down protocol of selecting 400 individuals randomly from each selected district (“Denial and deception”, Frontline , June 19). In the briefing on June 11, however, a figure of 26,400 was given as the number of individuals sampled (from 28,595 households) from the 65 (of the 71) districts that had been analysed until then. As against 710 clusters mentioned at the briefing (Table 2; the figure 770 in the table is again due to wrong addition), only 700 had been surveyed and 25 per cent of them were in urban areas. Therefore, the more pertinent issue here is that even after the completion of the analysis of five more districts (with a sampled population greater by 1,600, which is nearly 6 per cent of the total number tested) than on June 11, the overall seropositivity rate is still the same. These are some intriguing questions that hopefully the ICMR will clarify at some point.
Using sero-epidemiological data available from 22 countries, notwithstanding the fact that different serological kits were used by the countries, a pooled seroprevalence figure is 4.76 per cent, ranging from 0.65 per cent in Scotland to 22.6 per cent in Iran. Comparing the Indian seroprevalence value of 0.73 per cent (in May) with the above, the paper says, “the low prevalence [of less than 1 per cent] observed in most districts indicates that India is in [the] early phase of the epidemic [in mid May] and the majority of the Indian population is still susceptible to SARS-CoV-2 infection”. With even the present reported cases already nearly 10 times the total number (64,68,388) of estimated infections in May, only the results of the ongoing second-round survey (in the same geographical areas) will give us the true picture of the current state of infection spread.
Cut-off age
At the June 11 briefing, it was not indicated that a cut-off age had been used in sampling the population for the survey, which the paper has now given as 17. That is, the population in the age group below 18 was not sampled and surveyed, and that is one of the serious limitations of the survey as we know that people of this age group account for around 10 per cent of the reported cases in the country. The comprehensive survey in Tamil Nadu and Andhra Pradesh by a group of Indian and American researchers led by Ramanan Laxminarayanan ( medRxiv preprint posted on July 17) also bears this out. The survey reported in this work found that there was no differential risk across ages (including children) of acquiring or transmitting SARS-CoV-2 infection in both the States. This characteristic would perhaps hold for populations in all the States. It should, therefore, be clear that the sero-survey paper underestimates the infection prevalence significantly. While the paper does not discuss this shortcoming of the survey, it however acknowledges that by selecting only one adult from a household, the prevalence could have been underestimated “as transmission would be expected to be higher within the household”.
The paper has given the seroprevalance data of the sampled population stratified into three age groups: >17-45, 46-60 and >60 years. By projecting the populations in these age groups for 2020 from the 2011 Census data, the survey estimates the infection prevalence for the entire country. The total number of infections in early May (May 3) has been estimated to be about 6.5 million (6,468,388). Table 4 gives the break-up of this number for all the districts in the country distributed according to the four different strata, and their total is this figure.
By comparing with the actual number of reported cases on May 11 (79,230) and May 3 (49,720), the paper estimates the ICR to be between 81.6 and 130.1. That is, for every infection actually detected by the testing strategy in place at the time of the survey, between 82 and 130 infections—a huge number—were missed. A range is given because, by testing for IgG antibodies that form only after 7 to 14 days of infection, a sero-survey actually estimates the number of infections a fortnight to a week before. According to the paper, correspondingly, the dates for estimating the ICR range were taken 7 and 15 days before the initiation of the sero-survey on May 18.
Socio-economic classification
A socio-economic classification of the IgG seropositivity rate data shows that, as compared to the population in the age group 18-45 years, the age groups 46-60 and over 60 years were 1.30 and 1.11 times more at risk of being infected. Similarly, according to the survey data, males were 1.47 times more susceptible than females, perhaps because of occupation. The odds ratio of being IgG positive between a rural, urban-slum and urban non-slum person was found to be 1:1.90:0.93. That is, while an urban-slum person was about twice more likely to have been infected than an urban non-slum person, the odds of a rural individual being IgG positive was about 10 per cent more than an urban non-slum individual. This could be owing to the possible spread of infection in rural areas following the forced return of the migrant workers to their villages in the wake of the sudden nationwide lockdown.
Based on the sero-survey data, the infection fatality ratios (IFRs) for the four different strata of the sampled population were also estimated. The analysis assumed a lag time of three weeks from infection to death and looked correspondingly at the number of deaths in the districts surveyed on May 24 and June 1 respectively (emphasis added). Using the estimated seroprevalence in the different strata and the respective number of positives found in them in the survey, the analysis estimated the total number of infections in the surveyed districts (Table 4). Using the ratios of the number of deaths in the surveyed districts and the number of infections estimated based on the seroprevalence rate in the sampled populations, the plausible IFR ranges for the different strata were estimated. While the IFR per 10,000 infections on May 24 ranged between 0.18 (for zero incidence stratum) and 11.72 (for high incidence stratum), t he corresponding IFR range for June 1 was 0.27-15.04.
Glaring omission
A glaring omission in the published paper is data about the containment zones of “hotspot cities” that were part of the survey and whose analysis was stated to be ongoing on June 11. Only two major cities, Chennai and Bengaluru, have been included in the 70 districts whose seroprevalence data have been presented in the paper (see box). This may be because, at the time of district sampling, Chennai and Bengaluru had not probably been identified as “hotspot cities”. But by the time the survey began, these had been declared as red zones or “hotspot cities”. At the June briefing, Bhargava only cryptically said that “infection[s] in containment zones were found to be high with significant variations”. He neither gave the names of the hotspot cities nor indicated how high the infection rate was found to be and what the nature of the variations was. Only the results of an analysis of this part of the survey can tell, but a lid seems to have been put on them. (See box for a list of “hotspot cities” as learnt from a reliable source.)
It has been reliably learnt that the results of the survey of containment zones had also originally formed part of the paper. But the diktat from above had said that either the paper will be published without the containment zone data or not at all. Apparently, in the wake of this, there was an internal debate among the authors whether to publish the survey results or not. Ultimately, they decided that it was better to get at least a significant part of the survey results out than nothing at all. That could partially explain the delay in the publication of the paper and the other could be a deliberate withholding of its publication at the editor’s level. Interestingly, the paper only notes the following, presumably as an afterthought after the omission of data pertaining to the “hotspot cities”: “We may also underestimate prevalence if our selection missed clusters with higher prevalence including those among most of the metropolitan cities. Only Chennai and Bengaluru were included in the sero-survey on account of the random selection process.”
Coming as late as four months after the survey, the results and the above discussion on them are only of academic interest as they do not serve the basic purpose of a sero-survey any more. The virus is everywhere and, with the country’s health management failing them miserably, people have no other choice but to wait for herd immunity to build in the population to be delivered from the grips of the virus.
However, the findings of the sero-survey do help us to at least understand retrospectively why the country is in such a mess today with regard to COVID control. As of September 17, the total number of infections in the country was 52,12,686, and with the number of cases a day about 50,000 more than in the United States, India will become the No. 1 affected country in the world much sooner than what the trend only early this month had indicated, perhaps by early October. Are we anywhere closer to achieving herd immunity today? Unless we have results of the national sero-survey’s second round soon, there is no way to know that though a few local sero-surveys in cities have indicated that some pockets like slums may be closer to achieving herd immunity.
What is worse is that it cannot even be known when the spread will reach its peak and then begin to decline. As Jacob John told the interviewer from www.thelede.in in April: “What did the bear see when it reached the top of the peak? The other side of the mountain…. [Only] when the curve takes a downturn and stays the down-slope, then we know the peak was passed.”
COMMents
COMMents
SHARE