3 ms·

Two (or more) doses are fairly standard for vaccines. The first to trigger seroconversion, the subsequent ones to exercise the secondary immune response (and fo

by cltby 6y ago

Two (or more) doses are fairly standard for vaccines. The first to trigger seroconversion, the subsequent ones to exercise the secondary immune response (and for those relative few individuals who did not seroconvert after the first jab). The interval between them is typically quite long, often on the order of years [1]. From that pov, the 3-4 week booster interval for the COVID vaccines is tiny. It's likely the minimum possible interval given that seroconversion takes >= 14 days (I'd guess the short interval was chosen to get trial results as soon as possible). There are many reasons to think that alternative dosing schedules would be not only viable but better than the current default. [1] https://www.cdc.gov/vaccines/schedules/hcp/imz/child-adolescent.html https://www.cdc.gov/vaccines/schedules/hcp/imz/child-adolesc...

csense 6y ago
How did they decide on this years-apart schedule though? Presumably they had some data to say "we tried giving group A a booster after 90 days, group B after 365 days, group C after 2 years, group D after 3 years, group E after 7 years, and group D did best in terms of not catching the disease"? Or maybe they said "we measured secondary immune response with a lab test, and group D had the highest number"? If these weren't done for the COVID vaccine specifically, but some other vaccine, couldn't you still put some rough ballpark quantitative guesses on the COVID vaccine to help guide public policy better than "shrug, we have no clue what will happen if you do anything other than 21 days"? Something like "We think 40% will be protected with one shot, 90% will be protected with two separated by 21 days, and we can fit a curve to these two data points which has the same shape as curves for other decades-old vaccines that have tons of data and is supported by biological theory [paper1] [paper2] [paper3] [paper4] [paper5], and that curve continues to monotonically increase, so waiting 5 weeks instead of 3 should be no problem, in fact we think you'd be 91% immune instead of only 90% immune (provided you don't actually get the disease in the extra intervening 2 weeks), r^2 = whatever, but waiting 1 week instead of 3 is probably super bad, you only get up to 50% based on the shape of the curve and what's known from other vaccines, [paper6] [paper7] r^2 = whatever. Disclaimer: This assumes COVID-19 vaccine acts like the vaccines in [paper8] [paper9] [paper10] which collectively cover 80% of known vaccines, but it's always possible since it's MRNA the curve will be something super weird and there's absolutely no reason to think it'll be like that, but it'll take us 10 years to definitively rule it out, but if you need to make public policy and you don't have 10 years to wait, a policy based on a tentative analysis with a 20% chance of being wrong, which 20% we got by polling experts' Bayesian priors of the probability that a new vaccine will have a totally different curve, is probably better than a policy based on a completely random guess [paper11] [paper12] [paper13]." Based on that you could do an expected-value calculation of how many less people will get sick under policies of holding back various percentages of vaccine for second doses (the factor you control), vs some distribution of possible random future production hiccups (the factor you don't). You might find it's better to get shots in arms now to have as many people as possible 50% protected as long as possible if you know it's implausible that it's going to be a disaster if those people wait 5 weeks instead of 3 for the second shot due to a manufacturing hiccup.
- cltby 6y ago
  > If these weren't done for the COVID vaccine specifically, but some other vaccine, couldn't you still put some rough ballpark quantitative guesses on the COVID vaccine to help guide public policy better than "shrug, we have no clue what will happen if you do anything other than 21 days"? You definitely could (and this is exactly what many people are trying to do!) But this approach relies on many assumptions, and the FDA is hesitant to let assumptions (no matter how well-grounded) enter into its evaluations. Thus its near exclusive reliance on RCTs. This is mostly reasonable, but it causes confusion amongst people who don't understand evidentiary standards. Thus all the braying about "no data" or "no evidence."

by cltby 6y ago

csense 6y ago

How did they decide on this years-apart schedule though? Presumably they had some data to say "we tried giving group A a booster after 90 days, group B after 365 days, group C after 2 years, group D after 3 years, group E after 7 years, and group D did best in terms of not catching the disease"? Or maybe they said "we measured secondary immune response with a lab test, and group D had the highest number"? If these weren't done for the COVID vaccine specifically, but some other vaccine, couldn't you still put some rough ballpark quantitative guesses on the COVID vaccine to help guide public policy better than "shrug, we have no clue what will happen if you do anything other than 21 days"? Something like "We think 40% will be protected with one shot, 90% will be protected with two separated by 21 days, and we can fit a curve to these two data points which has the same shape as curves for other decades-old vaccines that have tons of data and is supported by biological theory [paper1] [paper2] [paper3] [paper4] [paper5], and that curve continues to monotonically increase, so waiting 5 weeks instead of 3 should be no problem, in fact we think you'd be 91% immune instead of only 90% immune (provided you don't actually get the disease in the extra intervening 2 weeks), r^2 = whatever, but waiting 1 week instead of 3 is probably super bad, you only get up to 50% based on the shape of the curve and what's known from other vaccines, [paper6] [paper7] r^2 = whatever. Disclaimer: This assumes COVID-19 vaccine acts like the vaccines in [paper8] [paper9] [paper10] which collectively cover 80% of known vaccines, but it's always possible since it's MRNA the curve will be something super weird and there's absolutely no reason to think it'll be like that, but it'll take us 10 years to definitively rule it out, but if you need to make public policy and you don't have 10 years to wait, a policy based on a tentative analysis with a 20% chance of being wrong, which 20% we got by polling experts' Bayesian priors of the probability that a new vaccine will have a totally different curve, is probably better than a policy based on a completely random guess [paper11] [paper12] [paper13]." Based on that you could do an expected-value calculation of how many less people will get sick under policies of holding back various percentages of vaccine for second doses (the factor you control), vs some distribution of possible random future production hiccups (the factor you don't). You might find it's better to get shots in arms now to have as many people as possible 50% protected as long as possible if you know it's implausible that it's going to be a disaster if those people wait 5 weeks instead of 3 for the second shot due to a manufacturing hiccup.

cltby 6y ago
> If these weren't done for the COVID vaccine specifically, but some other vaccine, couldn't you still put some rough ballpark quantitative guesses on the COVID vaccine to help guide public policy better than "shrug, we have no clue what will happen if you do anything other than 21 days"? You definitely could (and this is exactly what many people are trying to do!) But this approach relies on many assumptions, and the FDA is hesitant to let assumptions (no matter how well-grounded) enter into its evaluations. Thus its near exclusive reliance on RCTs. This is mostly reasonable, but it causes confusion amongst people who don't understand evidentiary standards. Thus all the braying about "no data" or "no evidence."