With SARS-CoV-2 prevalence of 66%, Amazonia’s largest city may have reached herd immunity

A cross-border group led by researchers at the University of São Paulo presented findings for Manaus from a study based on mathematical modeling combined with serological analysis. The results showed seroprevalence at 22% in the city of São Paulo.

By Karina Toledo  |  Agência FAPESP – A study reported on September 21 via publication of a non-peer-reviewed paper on the medRxiv platform suggests that about 46% of the population of Manaus, the capital of the state of Amazonas in Brazil, had contracted SARS-CoV-2 when the contagion rate peaked in mid-May. The proportion reached 65% a month later and stabilized at 66% for the next two months.

This “unusually high infection rate suggests that herd immunity played a significant role in determining the size of the epidemic” in the largest city in the Amazon region, according to the authors. 

“Exposure to the virus appears to have led to a fall in the numbers of new cases and deaths in Manaus. However, our results point to a much higher level of seroprevalence than those estimated by previous studies,” said Ester Sabino, a professor in the University of São Paulo’s Medical School (FM-USP) and principal investigator for the study, which was supported by FAPESP. 

The findings reported in the paper were based on a combination of mathematical modeling and serological analysis of blood samples from donors who gave blood to HEMOAM, the Amazonas blood bank, in the period February-August 2020.

“We selected samples from 1,000 donors for each month and analyzed them to detect the presence of antibodies against SARS-CoV-2. Next, we made a number of corrections to the data by means of mathematical modeling,” Lewis Buss, first author of the paper, told Agência FAPESP. Buss is a master’s candidate in FM-USP’s Institute of Tropical Medicine and Department of Preventive Medicine.

This kind of analysis is subject to biases for which adjustments must be made, Buss explained. One is the tendency for blood donors to be younger and healthier than average for the population as a whole, and hence often asymptomatic. Another, in the specific case of Manaus, is that a majority of the population is male. The researchers also had to adjust for the sensitivity of the serological test they used, estimated at 85% for individuals with mild or no symptoms, so that 15% of the results can be false-negatives. The key adjustment, however, corrected for the natural decline in seroprevalence against the novel coronavirus observed in epidemiological studies conducted in several countries.

“Like studies by other groups, our study shows that antibodies against SARS-CoV-2 decay rapidly a few months after infection,” Buss said. “This is clearly happening in Manaus, highlighting the importance of time-series surveys to understand the progress of the disease.”

According to the study, raw seroprevalence (without any kind of correction) rose from 0.7% in March to 5.5% in April, 39.9% in May, and 46.3% in June, slipping to 36.5% in July and 27.5% in August. With adjustments based on mathematical modeling, however, it was estimated at 0.7%, 5%, 45.9%, 64.8%, 66.1% and again 66.1% respectively.

The researchers used a similar approach for the city of São Paulo, analyzing blood donated in the period February-August to the Fundação Pró-Sangue blood bank at Hospital das Clínicas, Latin America’s largest hospital complex. Here, too, they selected 1,000 samples per month. In this case, however, they used sampling weights based on the size of the population in all districts of the city. 

Raw seroprevalence in São Paulo rose from 0.9% in March to 3% in April, 5.3% in May, and 11.9% in June, slowing to 9.6% in July and rebounding to 12.1% in August. The adjusted levels were 0.8%, 3.1%, 6.9%, 16.1%, 17.2% and 22.4% respectively.

“The two cities have very different epidemiological curves,” Buss said. “It’s hard to explain why on the basis of the serological data alone. The new information may be that seroprevalence was already high in Manaus when the death rate began falling, which suggests that herd immunity was a factor. In São Paulo, on the other hand, seroprevalence was far lower and the curve was flatter, so other factors probably influenced the downtrend in the number of new cases there.”

According to Sabino, in São Paulo the death curve peaked in May-June and remained basically flat thereafter. “In contrast with Manaus, the rates fell slowly in São Paulo, and the August levels were similar to those seen in April,” she said. “However, many more people are now wearing masks, and although the shops have reopened mobility is restricted and schools remain closed, as do cinemas and theaters. These factors may have limited the growth of the disease in São Paulo. It’s worth noting that if the curve were similar to that seen in Manaus, mortality in São Paulo would be three times higher.”

Next steps

The group coordinated by Sabino is now conducting serological analyses of blood samples supplied by donors in seven other state capitals: Rio de Janeiro, Salvador, Recife, Fortaleza, Curitiba, Belo Horizonte, and Campo Grande. The results will be published soon.

“Working with blood bank data enables us to measure seroprevalence for a specific disease in less time and for less cost than when we have to go from door to door collecting blood samples,” Sabino said.

The strategy has limitations, however, she noted. The main one is the difference in profile between blood donors and the overall population of a city, so that care must be taken to make the sample representative.

“When we began the study in São Paulo, with FAPESP’s support, one of the first things we did was to use geographic data regarding blood donors to stratify the samples,” Sabina said. “We were able to extend the study to other major cities thanks to support from Todos Pela Saúde [a fund sponsored by private bank Itaú Unibanco] and the serology kits donated by Abbott Laboratories [a US-based pharmaceutical and medical device company].

“The epidemiological curves for Manaus and São Paulo were quite different even though both introduced measures to reduce the transmission rate at about the same time and social isolation levels were similar in both. The next steps will entail analyzing the curves in other cities and then creating models that enable us to understand which factors weighed most in each case.”

The research is conducted under the aegis of the Brazil-UK Center for Arbovirus Discovery, Diagnosis, Genomics and Epidemiology (CADDE), which is funded by FAPESP, the UK Medical Research Council and the Newton Fund.

The article “COVID-19 herd immunity in the Brazilian Amazon” can be retrieved from: www.medrxiv.org/content/10.1101/2020.09.16.20194787v1.full.pdf.