Manaus is the capital and largest city in the Brazilian state of Amazonas (population a little over 2 million). Back in the spring of 2020 it was hit hard by the SARS-Cov2 pandemic, so hard in fact that one study estimated the city had reached herd immunity with an estimated 76% infection rate. In the last two months however the city has been hit hard again with hospitals running out of oxygen and turning infected patients away. On the surface this appears to be a contradiction-in-terms: how can you have an intense second wave if most of the inhabitants have already been infected with the virus? Knowing the answers is crucial because the implications could affect us all.

The current situation is so dire and peculiar (Figure 1) that The Lancet very recently published a summary and suggested several explanations. Here’s my take.

Hospitalizations and deaths due to SARS-Cov2 in Manaus, Brazil from beginning of the pandemic to late January 2021.

(A) Dark lines are the 7-day rolling averages and lighter lines are the daily time series of COVID-19 hospitalisations and excess deaths. Hospitalisation data are from the Fundação de Vigilância em Saúde do Amazonas. Total all-cause deaths for 2020–21 were reported initially by the Prefeitura de Manaus and subsequently in the daily COVID-19 bulletin of the Fundação de Vigilância em Saúde do Amazonas. All-cause deaths from 2019 were from Arpen/AM (Associação dos Registradores Civis das Pessoas Naturais do Amazonas). The compiled excess death data are from Bruce Nelson from the Instituto Nacional de Pesquisas da Amazônia. (B) Rt was calculated using the time series of COVID-19 hospitalisations after removal of the past 14 days to account for delays in notification. Rt was calculated using the EpiFilter method. Lines are median Rt estimates; shaded areas are the 95% CIs. Rt=Effective reproduction number. SARS-CoV-2=severe acute respiratory syndrome coronavirus 2. Figure courtesy Sabino et al, The Lancet, January 27, 2021.

The Manaus and Sao Paulo Seroprevalence Study
The epidemiological study was first published as a preprint in late September and finally published in the journal Science a couple of weeks ago. According to the study, the estimated attack rate (percentage of population infected) by October of last year was 76%, which is certainly approaching the range in which herd immunity effects can show up. (When the percentage of infected people exceeds the herd immunity threshold of 60 to 67%, or 100 × [1 – (1/R0)], each infection generates fewer than one secondary case, and so the epidemic dies out; it’s 67% for an R0 of 3.) Much criticism has been leveled at the study because it used a blood donor population, which has different sex and age parameters compared to the overall population. However, when you take the time to read the study (and its huge supplementary file) you can see that the authors did their best to adjust for these parameters. Indeed, the study was well done. So, what happened?

Social Mobility
The first hypothesis is that blood donor populations are likely to be relatively healthy and more mobile, and hence more socially interactive than the general population. Thus these donors could have experienced a higher attack rate than the general population. While this is certainly possible, the data provided for Sao Paulo from blood donor populations is consistent with how the epidemic played out in that city in 2020 with a vastly lower attack rate (29%). Other studies, notably the one carried out in The Netherlands in the spring of 2020, also suggest that social mobility of blood donors compared to the general population is unlikely to grossly affects seroprevalence numbers. Finally, the poor observance of the first lockdown in Manaus in the spring of 2020 would mean that the general population was not very constrained compared to pre-covid times, and even with lockdown imposed a second time its was not greeting with enthusiasm. Consequently, while seroprevalence results could have been slightly overestimated, I find it hard to believe that the blood-donor population is very unrepresentative of the entire city population.

Waning Immunity
One of the concerns that was voiced early in the pandemic is that immunity acquired from SARS-Cov2 infection could deteriorate over time. Several studies published to date do suggest that while antibody response can decline with respect to time, robust responses are still likely at 6 months post-infection. The average time between the first and second outbreaks in Manaus is around 7-8 months. Although this could be a small factor, it seems unlikely that immunity could have declined so badly that the majority of the population was essentially susceptible to infection.

Herd Immunity Was Never Achieved
The accompanying editorial published in Science suggests that herd immunity was not achieved in the city population by October of 2020, and that is one of the reasons why the second outbreak occurred. With all due respect to the authors I disagree. While it’s certainly possible that herd immunity was not achieved in respect of the existing SARS-Cov2 strains earlier in 2020, given the estimated seroprevalence (76%), a severe outbreak in which peak deaths and hospitalization have exceeded levels in the first outbreak, infection of the remaining susceptible population is simply too small to be responsible by itself.

Could the Culprit be the new P1 SARS-Cov2 Strain?
Over the last few months, we’ve seen several new SARS-Cov2 strains emerge that are quite different. These are strains in which many multiple mutations are present, particularly in the spike protein with others in the non-structural proteins. In southern California, for example, which had a particularly nasty SARS-Cov2 outbreak in late 2020, it appears from genetic analysis that the new CAL.20C strain is in large part responsible. Likewise, the P1 strain has emerged in Brazil in the last several months. It contains a unique 17 unique amino acid changes and 3 deletions, including the E484K, and N501Y mutations in the spike protein. Both these mutations are likely confer better viral fitness because of increased transmissivity and the ability to evade antibodies to early SARS-Cov2 strains although more data are required to confirm these points. To date reinfection by different strains has been rarely reported, but the big problem is that we haven’t conducted any largescale studies to assess reinfection.

Let’s take a closer look at the number of cases and deaths over the past 2 months (Figure 1).

Cumulative SARS-Cov2 cases and deaths December 5 through February 5 in the state of Amazonas, Brazil. Data source: JHU CSSE.

Cumulative SARS-Cov2 cases and deaths December 5 through February 5 in the state of Amazonas, Brazil. Data source: JHU CSSE.

The data show two clearly different infection and mortality rates, and by using linear least squares regression, the intersection of the two lines approximates to January 7, 2021. Similarly, the mortality rate differences suggest a point estimate around January 13 (range: January 6-18). Typically, the time from infection to mortality is around 13-19 days and in the raw Brazilian data appears to be 6 days. Moreover, the case fatality rate changes from 2.2 to 4.7 using the aforementioned points as time breakpoints. Although there are likely to be errors in reporting of cases and deaths, as well as the assumption that the vast majority of cases occurred in Manaus, given that the lowest granularity for available data is the state of Amazonas, both results appear consistent with a more virulent strain eclipsing older less aggressive strains over a breathtakingly short period of time.

My hypothesis is that in Manaus the P1 strain is re-infecting the population because immune responses to prior strains are not cutting the mustard in many individuals. I hope it’s just that—bad as it is—and that we aren’t seeing antibody-dependent enhancement. That would be even worse news.

As the pandemic spreads and many more individuals are infected with SARS-Cov2 we can expect many more “very rare” events to be come more commonplace due to sheer numbers. This includes the rise of much more viable strains, which have higher transmission characteristics and possibly more lethality. We’re seeing a bunch of these types of strains starting to dominate the world right now. Vaccines aren’t likely to be quite as effective against these new strains so we need to get our vaccination programs sorted before viral evolution moves on, otherwise this pandemic will take a long time to control.