Thread by @michaelzlin

New study in @NatureMedicine correlating neutralizing antibody levels to vaccine efficacy. Has a nice Fig. 1 that summarizes both data and model well. Suggests that maybe you can predict vax efficacy from neutralizing antibody levels. Quick impressions ...
nature.com/articles/s4159…

Conveniently all studies on vax-elicited nAbs also measured nAbs in convalescent sera, so the authors could normalize the former by the latter. We have to hope convalescent nAb levels don't vary too much from study to study. Not perfect but better than no normalization.

When that is done, and one assumes having nAb levels at ~20% of mean convalescent sera gives you a 50-50 chance of getting symptoms after SARSCoV2 infection, then based on the gaussian distribution of nAb levels among individuals, you can predict % breakthrough infections (Fig 1)

This may be the first peer-reviewed article to try to compare vaccines, so the comparison is interesting to me to adjust my own understanding. I added the manufacturer names as we're more used to those.

The results pretty much line up with impressions from clinical trials so far. The RNA vax and one protein vax tested so far look great. So does the Sputnik 2-dose Ad vax.

In the middle we have Covaxin inactivated virus, J&J 1-dose Ad, and AZ 2-dose Ad. The similarity of J&J 1-dose to AZ's 2-dose is likely because J&J uses prefusion spike. AZ didn't and their 1-dose result is worse (not shown in this chart).

Why is Gamaleya's 2-dose Ad so much better than AZ's 2-dose Ad? One explanation is the use of different Ads for the two doses, so that nAbs generated to Ad26 by dose 1 won't blunt the effect of dose 2 (Ad5). That's smart. (But some controversy over reliability of their findings)

Recombinant vax give manufacturers flexibility over vector and antigen. The other Ad manufacturers used that to their advantage but not AZ. They didn't do either serotype switching or prefusion configuration.

The very different results between inactivated vaccines may be due to the adjuvant, dose amount, or dose spacing. In the Phase 1 trial, the Coronavac dose escalation didn't reach a plateau in nAb generation, so perhaps higher doses would be useful.
thelancet.com/article/S1473-…

Also the Phase 3 result used in this study is the one reporting 51% efficacy in Brazil at papers.ssrn.com/sol3/papers.cf… (really hard to find). It uses a 2-week interval, which is odd as the Phase 1 trial already showed a 4-week interval to elicit 2x higher nAbs.

That trial also was conducted in the presence of the P.2 variant; not clear how that affects the expected protection from the expected distribution of nAbs against original strain. It was also done in HCWs in Manaus so could be they were challenged by higher amounts of virus.

Chile's reported 67% efficacy could be due to a longer average interval between shots (which increases nAbs and thus expected protection), or a more favorable strain profile, but info is hard to find. This discrepancy is still a mystery.
cdn.who.int/media/docs/def…

But overall it seems dosing amount and interval is important for inactivated vaccines. Chile's experience was that Coronavac had very low efficacy in just one dose alone. So that one in particular needs a higher dosage, and then perhaps 2 follow-up boosters spaced 1 month apart.

As for the paper, it looks pretty solid. The derived model even predicts the % protection loss from known fold reduction in nAbs for new strains. This can be a very useful tool to guide vaccine makers in phase I trials toward the right dosing regimens to optimize protection.

Follow-up since this thread is getting renewed attention. New data on Sinovac came to light from Chile. They appear to use an interdose interval of at least 3 weeks there (maybe 4) and, and see 67% overall efficacy. Info below with updated chart.

https://twitter.com/michaelzlin/status/1396222623453827073

Also the final Phase 3 results for Bharat's Covaxin were announced today. Overall efficacy is 78%, similar to the first interim look (the chart above used the 81% of the interim look to plot Covaxin):

https://mobile.twitter.com/michaelzlin/status/1398895920331194368

Unofficially the Chile trial uses a 4week interval between doses compared to a 2week interval in Brazil. The longer interval was found in Phase 1 trials to produce 2x higher neutralizing antibodies. So the Brazil trial results should basically be tossed.