microCOVID
Add options for Had Covid and Vaccinated
20% estimated national seroprevalence means that lots of people have some amount of immunity!
Vaccine rollouts add to this.
How long do antibodies provide immunity? One study found 10x risk reduction for 6+ months: https://www.nejm.org/doi/full/10.1056/NEJMoa2034545
To be clear, Had Covid and Vaccinated will need to be modifiers, not Risk Profiles - an average person in California, risk reduced by 95% (best case vaccination) is still 650microCOVID right now!
None of the vaccine trials involved weekly testing, so we don't have a ceiling on the viral load of vaccinated people. We need to build out a model another way. (Pfizer, Moderna)
One hint is neutralizing antibody concentration.
Participants in the Pfizer trial showed 5-40 times the neutralizing antibodies compared to recently recovered COVID patients: https://www.nature.com/articles/s41586-020-2639-4
The antibodies last a good while; Moderna reported phase 1/2 antibodies after 90 days, modest decline, still higher concentration than the average recovered patient. Trend suggested an annual booster "would be good" https://www.fiercebiotech.com/biotech/moderna-s-3-month-data-raise-hopes-for-covid-19-vaccine-durability
We also note that the Lumley et al showed a similar amount of protection in their study from seropositive past infections as the vaccine trials did (90% vs 95% respectively). However, participants in Lumely et al also did not have mandatory weekly testing:
Although health care workers were offered asymptomatic PCR testing every 2 weeks, the workers attended less frequently than that (mean, once every 10 to 13 weeks).
So now we have two ways to do this.
What I actually think: Vaccines are as good as infection at preventing transmission, probably better. Healthcare workers with antibodies were 90% less likely to test positive over 30 weeks than those without (average 2-3 asymptomatic tests / person). Therefore it seems likely that vaccines reduce transmission chance by 90%, or an optimistic effective risk multiplier of 0.1.
A more robust, conservative method
- In Phase 3 Trials, participants were followed up with weekly to check for symptoms. Patients with symptoms lasting 48 hours were tested daily until recovery (daily testing makes false negatives much less likely!). In other words, if a trial participant had even mild COVID symptoms, the trial would almost certainly catch it.
- That is to say, the exact thing that Phase 3 Trials verified was that vaccinated people were 95% less likely to have symptomatic COVID than those in the control group.
- In the worst case, let's say that all of that difference is accounted for by asymptomatic cases, i.e. all cases that would have been symptomatic become asymptomatic. Additionally, let us assume that all asymptomatic cases in the control group were not caught AND there was an equal number of asymptomatic cases in the experimental group.
- https://jammi.utpjournals.press/doi/10.3138/jammi-2020-0030 is a recent meta-study of asymptomatic spread. They found that 17% of cases are asymptomatic and asymptomatic individuals were 42% as likely to transmit.
- Now we back-calculate the total infections:
- If there were N recorded symptomatic infections in the control group, N / (1 - 0.17) = 1.2N total infected
- In the experimental group, there were 0.05 N symptomatic, .95 N who would have been symptomatic but we presume are asymptomatic, and 0.2 presumed asymptomatic.
- We now calculate the adjusted infectiousness by accounting for asymptomatic individuals as 42% as infectious as a symptomatic individual:
- Control:
1 * N + 0.42 * 0.2N = 1.084N - Experimental:
1 * 0.05N + 0.42 * (.95 + 0.2) * N = 0.533N
- Control:
- Finally we divide these to get a worst-case effective transmission risk of 0.49 for vaccinated individuals.
You might instead speculate that the individuals that would have been asymptomatic become non-infectious, which would reduce the effective risk to 0.41
For completeness, there's also a super-pessimistic option which is, the vaccine causes people to become immune to symptoms while spreading just as much virus as a full infection, which would suggest an effective risk of 1.2. I think this is really, really unlikely, because symptoms (and the severity thereof) appear to be correlated with viral load.
I don't have a thought on the validity of the two options above. But I do have an initial reaction that goes like this: if we only give people .49 for vaccines, we are going to need to do a REALLY good job of explaining why it's so low, when people are hearing '95% effective' all over the place.
I asked a friend who is a COVID researcher at MGH and a masters student at London School of Hygiene and Tropical Medicine. Here's what they had to say:
Flu vaccines are really effective at reducing virus concentration in airways in monkeys subjected to 10^6 infectious dose https://www.nejm.org/doi/pdf/10.1056/NEJMoa2024671
Antibodies are found in nasal tissue for COVID patients:
These authors propose that IgG responses in saliva and serum are maintained in Covid patients for about 3 months POS. https://immunology.sciencemag.org/content/5/52/eabe5511.full
IgG antibody response is much more common and durable in patients with severe covid https://www.biorxiv.org/content/10.1101/2020.05.21.108308v1.full
The first dose of the vaccine produces a lot of antibodies, but they are low affinity and should not be considered equivalent to antibodies produced by having covid
https://www.nejm.org/doi/full/10.1056/NEJMc2032195
Disappearance of circulating IgG doesn't mean a loss of immunity (as you know) because immune memory is a thing, just takes a bit of time to mount a secondary immune response and so if you breathe in virus, the particles may be chilling in your nose for a few days before they are neutralized.
I have no idea how you do your meta study combinatorial things. But I'd think it's fair to say that most people who had covid are immune via circulating IgG for 3-4 months (less time if asymptomatic infection) and have a much diminished chance of passing virus to you. And for vax I'd be careful and cautious about carrying over the "antibodies to immunity to virus neutralization and non-communication" paradigm because the antibodies you make between dose 1 and 2 can be large in number but they tend to be low affinity and/or poorly neutralizing. And then of course 14 days after the second shot there's a very small chance of passing on the covid to another, except in cases of exceptional exposure, previously unknown immune conditions and maybe escape mutations
Also, they are concerned that people getting infected between the first two shots will give COVID time and pressure to mutate to evade the vaccine, so folks with one shot maybe should be the most careful for societal reasons.
AstraZeneca released a preprint today with data suggesting asymptomatic infection rates are about the same or slightly lower in trial vs control, both 21 days after the first shot, and 14 days after the second shot. Total efficacy against all infections, asymptomatic and symptomatic, is 50-66% (uncertainty due to the LD/SD shenanigans). This is reasonably strong evidence against the hypothesis "vaccines turn many mild/moderate cases into asymptomatic cases."
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3777268 (see tables 1 and 2)
Given this, there are three hypotheses for mRNA vaccine efficacy at reducing disease compared to the AZ vaccine:
-
mRNA vaccines are more effective at reducing symptoms, but actually less effective at preventing infection.
-
mRNA vaccines are more effective at reducing symptoms, but no more effective at preventing infection.
-
mRNA vaccines are more effective at both reducing symptoms and preventing infection.
-
Here corresponds to the "control" hypothesis above:
1 * 0.05N + 0.42 * (.95 + 0.2) * N = 0.533N -
Implies total infections are reduced to ~50% with vaccination, and only 5% of infections are symptomatic:
1 * 0.05N + 0.42 * (.45 + 0.1) * N = 0.281N -
If the number of asymptomatic infections stays the same before/after mRNA vaccination, infectiousness is then
1 * 0.05N + 0.42 * 0.2 * N = 0.134N
Consider discussion in this ticket as pre-work for actual writeup: https://paper.dropbox.com/doc/Vaccines--BEdshibyYnB5yp1Xw2TAOo0SAg-1xqXJwGf2e5Odh4zt0TIC
Gave feedback, mostly I want:
- More clarity in places I wasn't really sure I was following correctly
- A second triangulation on Byambsuren
- (eventually, looks like it's just not proposed yet) a final topline Pfizer/Moderna number
After doing those, perhaps ask Stephanie Bachar what she thinks - she has been looking into this!
Here's some more studies on asymptomatic spread:
https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2769235 found viral loads in asymptomatic patients 0.7x symptomatic
https://www.medrxiv.org/content/10.1101/2020.05.10.20097543v2.full.pdf early meta-study finding about 1/2 as much transmission from asymptomatic individuals.
AstroZenica's trial itself finds a similar ratio of symptomatic to never-symptomatic cases - in their control group they found about 20% of cases were never symptomatic.
Adding this study here. Lmk if it's in the wrong Github issue. https://www.reuters.com/article/us-health-coronavirus-israel-vaccine/israeli-studies-find-pfizer-covid-19-vaccine-reduces-transmission-idUSKBN2AJ08D
Natural immunity needs to be taken into account for your risk calculation to be vaguely realistic. Both as a choice and in risk calculations if vaccination status is unknown. https://www.medrxiv.org/content/10.1101/2021.08.24.21262415v1 https://nypost.com/2021/09/12/the-us-is-nearing-immunity-from-covid-19/
