Bio: Scientist @fredhutch, studying viruses, evolution and immunity. Collection of #COVID19 threads here: bedford.io/misc/twitter/
Location: Seattle, WA
Currently, the US is reporting about 54k daily cases of COVID-19 (16 per 100k per capita) and the UK is reporting about 4k (6 per 100k). This seems comfortingly low compared to even this summer's BA.5 wave and let alone last winter's BA.1 wave. Figure from @OurWorldInData. 1/16
However, at this point, nearly all infections will be in individuals with prior immunity from vaccination or infection and this combined with a roll back in testing makes it unclear how to interpret current case counts compared to previous time periods. 2/16
We're interested in the case detection rate or the ratio of underlying new infections compared to reported cases. Throughout much of 2020 and 2021, I had a working estimate of 1 infection in ~3.5 getting reported as a case. 3/16
Although there are things like wastewater surveillance in the US, I believe the gold standard to assess case detection rate is the @ONS infection survey (ons.gov.uk/peoplepopulationa…) that has continued to test swabs every week in the UK regardless of individual's symptom status. 4/16
Here, I'm showing per capita case counts for England for Aug 2020 to Nov 15, 2022 using data from @UKHSA (coronavirus.data.gov.uk/deta…). The spike from the initial Omicron surge is quite evident and nothing in 2022 compares to this. 5/16
While here I'm showing estimated prevalence (proportion of individuals testing positive) from the @ONS infection survey (ons.gov.uk/peoplepopulationa…). The initial Omicron surge results in the most concurrent infections, but later waves in 2022 result in comparable prevalence. 6/16
If we compare daily cases to prevalence we see an interesting result, where there is a consistent relationship in 2020 and most of 2021 (blue and yellow points), and a shift in 2022 to a steeper relationship where fewer cases correspond to greater prevalence (red points). 7/16
If we take a ratio of ONS prevalence to daily incidence, we get the following where the ratio is fairly constant at ~30-fold throughout 2020 and 2021, but increases sharply throughout 2022 and is now perhaps ~300-fold. 8/16
The 30-fold ratio can be converted into case detection rate by dividing by the span of time during which the average infection tests positive. I'll use 10-days for the pre-Omicron period (nejm.org/doi/full/10.1056/NE…). This gives a detection rate of 1 case per ~3 infections. 9/16
Likewise, the current 300-fold ratio can be converted assuming the average infection tests positive for 8-days in 2022 (medrxiv.org/content/10.1101/…). This would imply a current case detection rate of 1 case per ~38 infections. 10/16
Unfortunately, I'm not sure how best to extrapolate this number to the US, as I believe that testing has declined more in the UK than in the US. However, the ~10-fold difference between a year ago and today is absolutely striking and may suggest a roughly similar decline. 11/16
This said, the silver lining here (for what it's worth) is that if there are now many more infections per case it means that looking at the ratio of cases to deaths is not easily comparable to earlier in 2020 or 2021. 12/16
Using the above ratio of ONS prevalence to daily per-capita cases along with a 10 day testing positive duration we can estimate new daily per capita infections in England as the following. 13/16
If we sum the number of infections in England from June 1 to present we get an estimate of 25.8M or 46% of the population infected. During this time, there were 17,514 deaths within 30 days of confirmed COVID infection. 14/16
This would imply a crude infection to fatality rate (IFR) of 17514/25.8M = 0.07% or 1 death in ~1500 infections, which fits with broad expectation for current IFR. If instead we use deaths with COVID on death certificate we get an IFR of 0.04%. 15/16
We have a situation of continued substantial circulation, but with per infection risk of death similar to influenza. Though this is not accounting for long COVID which alongside greater circulation places health burden of COVID substantially higher than influenza. 16/16
I was invited by @VandammeAm and other organizers of the #VEME2022 Virus Evolution and Molecular Epidemiology Workshop to present my perspective on applying pathogen phylogenetics for decision making. Slides from my talk are available here: bedford.io/talks/phylogeneti….
So, if we could halt further evolution of the virus, we'd expect that BA.2.75 would globally displace BA.5, but do so relatively slowly, over the course of months and drive modest epidemics while doing so. 8/10
However, during these months there will likely emerge sub-lineages of BA.5 bearing additional mutations that make these viruses more competitive. And so the "winning" variant will be the one that happens to collect the right mutations and generate a highly fit subvariant. 9/10
These analyses of BA.2.75 relied on open data sharing from scientists in India, Japan, Singapore and the US. We're still seeing remarkable evolution of SARS-CoV-2 with new variants emerging and spreading rapidly making continued global genomic surveillance essential. 10/10
Largely through partial immune escape, lineage BA.5 viruses resulted in sizable epidemics throughout much of the world. However, in most countries these epidemics are now beginning to wind down. What do we expect after BA.5? 1/10
Lineage BA.2.75 (aka 'Centaurus') has been high on watch lists due to sustained increase in frequency in India combined with the presence of multiple mutations to spike protein. We now have enough sampled BA.2.75 viruses from outside India to make some initial conclusions. 2/10
If we look at frequency data we see sustained logistic growth of BA.2.75 in India, Japan, Singapore and the US. Critically, in India it is clearly displacing BA.5. 3/10
However, the rate of logistic growth of BA.2.75 is lower than initially seen for BA.5, where BA.2.75's initial rate of growth in the US is 0.07 per day compared to BA.5's 0.14 per day estimated on May 28. 4/10
If we look at variant-specific Rt estimated by @marlinfiggins, we find initial estimates of Rt of ~1.3 for BA.2.75 in the US, which is greater than BA.5's current Rt of ~1.0, but is lower than BA.5's initial Rt of ~1.6. 5/10
Broadly, we expect variants with higher initial Rt to cause larger epidemics. Finding initial Rt of ~1.3 for BA.2.75 suggests it will drive a smaller epidemic in the US than what we experienced with BA.5 (ignoring potential effects of seasonality). 6/10
This can perhaps be seen at the moment in India, where despite BA.2.75 becoming predominant, there has not been a coincident increase in total cases. 7/10
Based on the experience in winter 2020/2021, seasonal influence on SARS-CoV-2 transmission is quite clear, but much of the Northern Hemisphere is currently experiencing large summer epidemics driven the spread of evolved BA.5 viruses. 1/11
It's necessarily fraught to try to make predictions of seasonal circulation patterns going forwards, but we can gain some intuition from simple epidemiological models. 2/11
In particular, we can use an SIRS system in which individuals go from Susceptible to Infected to Recovered, and then return to the Susceptible class due to immune waning / antigenic drift of the virus. 3/11
As @adamjkucharski expresses here, with SARS-CoV-2-like intrinsic transmissibility (with R0 of circulating viruses of perhaps ~8), infections will be largely gated based on speed of immune waning / viral evolution. 4/11
Including seasonal forcing for seasonal influenza where R0 is a bit lower and rate of waning / antigenic drift restores R→S every ~5 years, we get slow build up of susceptibility over the summer that followed by winter forcing results in winter epidemics and summer troughs. 5/11
However, if waning / viral evolution is fast enough, then this can drive summer circulation and reduce build up of susceptibles. We can see this by running SIRS models with different rates of waning, but otherwise largely SARS-CoV-2-parameters. 6/11
Here, we see that with flu-like ~5 year rate of waning (in blue), we get winter epidemics and summer troughs, while with faster waning we see greater levels of circulation and less variation between winter and summer (in yellow and red). 7/11
Even outside of emergence of wildly evolved Omicron virus, we've so far seen faster evolution of SARS-CoV-2's spike protein relative to influenza H3N2's equivalent HA1 protein. 8/11
Although it's difficult to estimate long-term rate of R→S for SARS-CoV-2, we can extrapolate from @HebAltarawneh et al's finding of 76% effectiveness of BA.1 infection to prevent BA.4/BA.5 reinfection (medrxiv.org/content/10.1101/…). 9/11
If what we've seen with Omicron evolution in 2022 becomes largely the norm, then this result would imply waning of ~24% in the span of ~6 months, or very roughly waning from R→S on a ~1.8 year time horizon, ie close to the yellow curve in the above SIRS model. 10/11
I'm not trying to make a prediction here, but just illustrate a scenario where we end up in a regime of year-round variant-driven circulation with more circulation in the winter than summer, but not flu-like winter seasons and summer troughs. 11/11
Follow up #1: I had missed this thread by @C_Althaus previously. It nicely demonstrates similar results in terms of dampening seasonal oscillations with increasing waning / antigenic drift.
This level of circulation and severe outcomes (including long COVID in addition to death) combine to make "endemic" COVID-19 a serious ongoing health burden. Here, @dwallacewells describes this viscerally as "looking pretty brutal". 5/5 nytimes.com/2022/07/20/opini…
