Skip to main content


Vaccine Efficacy Data!

Earlier this morning, Pfizer and BioNTech announced the first controlled efficacy data for a coronavirus vaccine. And the news is good.

You may recall that these vaccine trials are set up to get to a defined number of coronavirus cases overall, at which time the various monitoring committees lock the door and unblind the data to have a look at how things are going. Pfizer’s original plan (as mentioned in that post) was the most aggressive of them all – they were planning to take their first look once they hit 32 cases. But one of the things we learned from this morning’s press release is that the company and the FDA changed that, dropping the 32-case read in favor of a 62-case read. By the time they finished those negotiations, though, the number of cases had reached 94, so we actually have a much more statistically robust look than we would have otherwise. And the split between placebo-group patients and vaccine-arm patients is consistent with greater than 90% efficacy. That number will come into better focus, but I hope that we can continue to take 90% as the lower bound.

The final analysis of the trial is set for a 164-case level, and that should be reached sooner than we might have thought. The higher number of cases in this current readout is surely because the coronavirus pandemic is itself at the “uncontrolled spread” stage in much of the Northern hemisphere. Remember the worries about whether companies would have to move their trials around to find places where the disease was still spreading? Seems like a long time ago, and as things have worked out you can (unfortunately) run your vaccine trial pretty much anywhere you like.

More details: these numbers are from 7 days after the second dose of the vaccine (28 days after the first dose overall). Going forward, they plan to collect data at 14 days after the second dose to make the statistics more comparable to the other ongoing vaccine trials. Pfizer/BioNTech say that the protection looks like it should last at least a year – no numbers on that yet, but it can only be based on neutralizing antibody titers and/or T-cell levels and their change over time. The only way to get better numbers on that is to wait and collect better numbers; there is absolutely no way to tell without waiting to see. But if we’re already out to about a year’s protection, that’s very good to hear. And this is the time to mention that nothing has (so far) been set off on the safety side, “no serious concerns”. But the only way to collect longer-term safety data is to keep watching over that longer term as well. I suspect that we’re still going to see plenty of fevers and very sore arms after injection (with this and the other vaccines), and I say bring them on, then.

What does this mean for the pandemic vaccine effort in general? The first big take-away is that coronavirus vaccines can work. I have already said many times (here and in interviews) that I thought that this would be the case, but now we finally have proof. The worst “oh-God-no-vaccine” case is now disposed of. And since all of the vaccines are targeting the same Spike protein, it is highly likely that they are all going to work. There may well be differences between them, in safety, level of efficacy across different patient groups, and duration, but since all of them have shown robust antibody responses in Phase I trials, I think we can now connect those dots and say that we can expect positive data from all of them.

Having the Pfizer/BioNTech candidate read out first has some other implications. One that may (or may not!) get lost in the excitement is that Pfizer explicitly and publicly did not take US government funding for this effort. In fact, they said at the time that working in the “Operation Warp Speed” framework would likely slow them down, and it looks like they can make a case for that! Not all the victory laps that people are going to try to take for this will be justified (and wasn’t that ever the case, right?)

But the other thing to keep in mind is that this candidate has (so it appears) by far the most challenging distribution of all of them. The Pfizer/BioNTech candidate, last we heard, needs -80C storage, and that is not available down at your local pharmacy. Pfizer has been rounding up as many ultracold freezers (and as much dry ice production) as they can, but there seems little doubt that this is going to be a tough one. I know that the press release talks about getting 1.3 billion doses of this vaccine during 2021, but actually getting 1.3 billion doses out there is going to take an extraordinary effort, because you’re getting into some regions where such relatively high-tech storage and handling becomes far more difficult. Population density is as big a factor as electricity and transport infrastructure. With demanding storage requirements, the more people that are within a short distance of a Big Really Cold Freezer, the better. And the more trucks (etc.) that you have to send down isolated roads to find the spread-out patients, the worse. That’s always the case, but if you’re rushing against dry-ice-pack deadlines the situation is more fraught.

What are the regulatory consequences (as discussed here)? The press release says that the companies expect to reach the required safety data (two months after dosing) in the third week of November, and that they plan to file for an Emergency Use Authorization (EUA) after that. The complex rollout of this vaccine itself (see below) may mean that such an EUA, if granted, will not have as much of an immediate effect on the other trials, but this is still an open question, and there seems to be a good deal of debate at the FDA itself about how to handle things. I would assume that any EUA would be directed at the very highest-need populations; we’re not going to see a country-wide effort to vaccinate the population until sometime in 2021.

Remember as well that the efficacy levels we’re seeing this morning are after the second dose. If everyone magically got the vaccine this morning, we still wouldn’t all be able to breath easy for nearly a month. And we are not all getting it this morning, for sure. It’s obvious that the pandemic is ripping through a long list of countries right now – cases are rising steeply, hospitalizations right behind, and although we’re better at treating the patients than we were, deaths are inevitably going up as well. This is all going to get worse before it gets better – but the good news here is that it really is going to get better. Keeping your head down, wearing masks and keeping your distance is more valuable than ever, because there’s an actual finish line in sight that you have to reach. Hang on for the vaccine – for the vaccines – because they really are coming.

We’re going to beat this. We’re starting to beat it right now. An extraordinary, unprecedented burst of biomedical research – huge amounts of brainpower, effort, money and resources – has come through for the world.

347 comments on “Vaccine Efficacy Data!”

  1. Luis Filipe says:

    Fantastic news,well done to those involved in this trial!

    1. Peter Joseph says:

      What’s the plan to disclose to Pfizer guinea pigs like myself whether they got the real deal or saline? I want to know whether I can go out for dinner tonight.

      1. Ex-Pharma guy says:

        That should be in the informed consent form you signed at the start of the trial. If not, check with your investigator. And good for you, for volunteering for the study!

      2. MagickChicken says:

        They haven’t decided yet. Got my second dose(?) today and asked the nurse.

      3. Nick Sparkman says:

        If you’re really curious you could go get an antibody check and unblind yourself.

        I’m in the study, if you’re asking you’re probably in the control arm. I know a few other people that all got severe chills the night of and 101F fever for 2-3 days after each injection.

        1. Liz says:

          OTOH, my physician’s father is in the trial and reported arm soreness, though his antibody test was positive post-vaccine. Always possible he had an asymptomatic infection… Merely offered as a counterpoint.

          1. David E. Young, MD says:

            So, he was in the actually vaccine arm, right?

      4. Dave says:

        I’m under the impression that the real vaccine caused very noticeable side-effects, so if you didn’t feel significant soreness, fever, aches, etc. then you probably got the placebo. Is that right?

    2. Paul says:

      Seems that the study is seriously flawed since it does not control for behavior. There is no way of knowing if one are was predominantly less cautious than the other.

  2. Project Osprey says:

    Great news!
    It’s almost too good to be true – the most experimental therapy reading out first and hitting the target in a big way – but I’ll take it

    (… the first link to the trial designs article isn’t working BTW)

    1. Roy Little says:

      Perhaps the most significant news is not medical but financial. Bourla sold more than 60% of his stock yesterday, strongly implying that he knows the vax will not make any money. We can also speculate that it is NOT as promising as alleged simply on thte basis of an argument that has been used many times by Derek and others that a publicity release is not trustworthy and that we must wait for the publication of the final report. I first became suspicious witnessing Bourla’s embarrassing performance on BBC News saying that they were “jumping in their chairs” with excitement over the “90% efficacy.”

      1. Michael says:

        It’s more complicated than that. This stock sale was disclosed in, and planned since, August. The sale was legal, if awkward optically. Misrepresenting the interim efficacy findings in a press release would be very, very, very illegal under securities legislation.

      2. Hopeful Layman says:

        This may be naive of me, but it could be possible that Borula sold his stock because he knows that he will be under intense public scrutiny (e.g., the ongoing pressure on Slaoui to divest his Moderna holdings), and he doesn’t want to create the appearance of impropriety if his own company is on the verge of a historic triumph. Also, remember, this is not solely (or even primarily) Pfizer’s “victory.” Is BioNTech a publicly held company? If so, have their executives unloaded any stock recently?

        1. Michael says:

          BioNTech is BNTX on NASDAQ.

          1. Hopeful Layman says:

            Well, if there’s noting “funny” going on there, I’d wager that there’s nothing “funny” going on with Pfizer, either.

  3. JJ Walker says:

    “After discussion with the FDA, the companies recently elected to drop the 32-case interim analysis and conduct the first interim analysis at a minimum of 62 cases. Upon the conclusion of those discussions, the evaluable case count reached 94 and the DMC performed its first analysis on all cases.”

    Interesting the FDA encouraged Pfizer to delay the interim analysis… if they had not done that we would have most likely known that the vaccine worked weeks ago.

    1. Derek Lowe says:

      Them not wanting it to be seen as a political rush-job seems like the obvious explanation.

      1. Derek Lowe says:

        NOTE to the real Derek: this isn’t actually Derek. Sciencemag needs to fix their comments feature and not allow duplicate names.

        1. Hopeful Layman says:

          Point well taken — but I think that poster is probably right, anyway. And thank Jah for it!

    2. debinski says:

      It’s also interesting that they began the analysis on Nov 4th.

      1. Kathryn Storer says:

        Where did you see that? The press release says the interim was conducted yesterday, Nov 8

        1. debinski says:

          At the end of October (when they already had 32+ cases) they stopped analyzing the swabs from potential cases for some strange reason. They didn’t start analyzing them again until Nov 4. Sounds like they were afraid the info would leak. If they knew the swab results, I guess someone could break the blind (even though they weren’t supposed to)? From : “Gruber said that Pfizer and BioNTech had decided in late October that they wanted to drop the 32-case interim analysis. At that time, the companies decided to stop having their lab confirm cases of Covid-19 in the study, instead leaving samples in storage. The FDA was aware of this decision. Discussions between the agency and the companies concluded, and testing began this past Wednesday.”
          Is there some other reason they would stop analyzing the samples? Why not get that done ahead of time, even if you planned the data analysis for later?

      2. Roy says:

        Interesting – CNN are reporting that the DMC only met to unblind the results at 11am yesterday.

    3. Kathryn Storer says:

      I think I heard today that Pfizer chose to delay after learning they couldn’t submit till efficacy data was available till third week of November, and decided to let the efficacy data mature a bit more. FDA blessed it

      1. Kathryn Storer says:

        … I meant to say, they couldn’t submit until SAFETY data was available anyway, so dropped the first interim

  4. Anonymous says:

    Funny how Biontech/Pfizer essentially saving the world is still worth half of Tesla or a 4th of Facebook.
    Here forward, anyone who says drug companies make too much money clearly is probably wrong.

    1. Philgro says:

      Share prices are not the whole picture….

  5. Bruce says:

    I bet Trump is hopping mad that Pfizer didn’t make the announcement a week ago.

    1. WP says:

      Lol, what a racket. Now that the proper candidate has won, a massive cha-ching for the good guys and bye bye crisis!

      Can’t we get a GIF of Derek Lowe tap dancing in a barbershop costume at least? “HCQ — no no no, US vaccine is maxin’; now the onus on Pfizer is bonus, marvellous festivus bonus, pay us, just pay u-u-ussss! Bonus!”

        1. M says:

          The Babylon Bee is hilarious. I just hope most people realize it’s purely satire. Not always sure about that…

      1. Anonymous says:

        A vaccine today would not erase the comedy of errors that has been the current administration’s response. If anything it makes the callousness of Atlas’s attempted push for herd immunity even more stupid looking.

        This news doesn’t change the current situation or reduce the need for effective leadership that’s been lacking. Get out of here with your conspiracy theories. There’s still going to be tens of thousands of additional deaths in the US while distribution is worked out.

    2. Chauncey Gardener says:

      Are you kidding? After the Speaker of the House and Vice-President Elect gave pharma those stern warnings to NOT do anything before the election, and they weren’t kidding. These firms know what side their bread is buttered on.

      1. Steve says:

        Can you provide your source for the statement you made?

        1. BGDavis says:

          Evidently not. No surprise there.

      2. BGDavis says:


    3. MTK says:

      One would like to think that this announcement was made at the appropriate time without regard to the election or politics, but that’s probably naive.

      Having said that this announcement prior to Election Day would probably been damaging to the overall vaccination effort, since many would not have believed the results. Or at least more than don’t believe the results now.

      It’s unfortunate that this is situation we find ourselves in.

      1. M says:

        I’m going to assume coincidence and move on.

        1. Charles H. says:

          It’s not exactly coincidence, since they had apparently planned the info release date long ago…before whether they knew whether they had good news or bad news. But that they did it to avoid the appearance of political interference is quite plausible. Remember, they did decline to participate in the feds “Warp Speed” program. This may, indeed, have been to avoid excessive paperwork, but surely avoiding the appearance of political gaming was a factor.

      2. Hopeful Layman says:

        “. . . . this announcement prior to Election Day would probably been damaging to the overall vaccination effort, since many would not have believed the results. Or at least more than don’t believe the results now.”
        I agree 100%. This issue was obscenely politicized, and it is 100% appropriate for the company to have waited a few extra days to eliminate even the appearance of impropriety. Too much hangs in the balance.

    4. enl says:

      “I bet Trump is hopping mad that Pfizer didn’t make the announcement a week ago”

      I should imagine. Credit has already been claimed by the administration’s most prominent epidemiologist:

      Pfizer has a different take, though:

    5. Dr. Manhattan says:

      Trump offered to have them make the announcement right before the election. Giuliani had even reserved a podium for the announcement at the Four Seasons in Philadelphia…

      1. Hopeful Layman says:

        . . . before they realized that being adjacent to a crematorium would NOT send a very positive message . . .

  6. Matthew says:

    My hope right now is that ALL of the vaccine trials are able to continue fully to the end of their studies.
    How can we ensure that?

    1. guineapig says:

      As a trial participant (first shot of the Pfizer, so don’t know if it was the real thing), I would expect it’s going to be very difficult and ethically-challenging to get the placebo arm to remain un-vaccinated for the duration. Maybe lots of cash would do it.

      1. Ian Malone says:

        Not so sure, will take time to get the vaccine manufactured and distributed, they’ll be prioritising certain groups first (likely the most vulnerable and health workers), so other trials have months before there’s any possible conflict. Even then it’ll be longer before a significant proportion of the population get vaccinated, so allowing those already enrolled in trials to continue is not really an additional risk. (It also carries a possible benefit if a vaccine with less difficult distribution gets approved, and some of the others currently in trials already have stocks ordered, so a positive trial would unlock more doses.)

      2. Kathryn Storer says:

        Pfizer wanted to vaccinate placebo patients if positive efficacy results became available, but I believe the October ACIP recommended against that and for letting the trial run to endpoint. That’s only 164 cases thiugh and at this rate won’t be long

        1. guineapig says:

          Pfizer said they would un-blind and give the placebo arm the vaccine if it was approved. I doubt the placebo arm will complain much if that is about whenever their turn is in the normal distribution schedule, but asking them to wait the full 24 months may be tough.

        2. David E. Young, MD says:

          I have a real problem with that.

          It’s not in the consent. If you are going to force placebo arm enrollees to forgo the real vaccine until the study is entirely over, then please, put that in the consent form and make it clear.

          I am on the Moderna vaccine trial and undoubtely in the placebo group. That’t fine. But I want a vaccine soon after it is available. If Moderna has clearly favorable results in their study I hope that they would offer me that actually vaccine. Maybe not right away but not before long.

          There is nothing in my consent about having to refrain from getting a vaccine if a vaccine comes out. In fact, I asked. The reply was “we understand it perfectly well if you go and get another companie’s vaccine when it comes out.” They would still have me on trial none-the-less and they would still pay me to come for scheduled visits, scheduled phone calls and scheduled phone app questions.

          I would prefer to get the Moderna vaccine, if it works well.

          1. Jay says:

            I’m in the Pfizer group and a lawyer. I read the consent docs back and forth. Anyone can drop out of the trial at any time and a contract to the contrary would be unenforceable. Pfizer has already advised the FDA that it feels it has an ethical duty to unblind and offer trial participants a vaccine as soon as it applies for an EUA, so I imagine I’ll get the notice soon.

    2. Liz says:

      Not going to happen, and it’s not ethical to do so.

    3. Ken says:

      There was a long discussion of that in this very blog just a couple of weeks ago:

    4. doctorp says:

      It’s unlikely that any single vaccine manufacturer will be able to supply enough doses to keep up with demand. More approvals means more doses available in a timely manner and a better chance of ending the pandemic.

      1. Hopeful Layman says:

        Yes, especially remembering that this is a WORLD WIDE pandemic. Another reason to hope for good results for vaccines that don’t require such difficult and complex distribution protocols. (If J&J come out with results anywhere near as positive as these seem to be, I wonder if their “one-shot” vaccine will nudge them to the head of the line?)

  7. li zhi says:

    1. Why would 90% be a lower bound? This seems pollyanna-ish. I hope it is 99+% effective, but 94 cases can’t really be sufficient to be certain of that even if all 94 are in the control arm. Or am I wrong?
    2. The protocol was changed. That is troubling. Why was it changed? Unless there was a math error, the reason would seem likely to be political. If it was the FDA, then it would seem to be another instance of incompetence. If it is Pfizer, then why didn’t they get it (the protocol) right the first time? The peanut gallery speculates the motive was to withhold good news for Trump immediately before the election.

    1. Zambo says:

      Worth noting that the efficacy could be artificially dampened by the trial design, as those who got the saline placebo would likely be more cautious on average (and still got infected at 10X the rate!)

      1. Matthew says:

        Why would they “likely be more cautious?”
        They have no idea if they received vaccine or placebo.

        1. Liz says:

          I mean, I’m in the trial and it would be absolutely trivial to go get an antibody test and see what side of the analysis I ended up on. My better angels have deterred me, but it is what it is.

        2. filialleiter says:

          Some participants would know in which arm they are – the adverse events in the vaccine arm (especially after the 2nd dose) can be pretty severe (headache, fatigue, fever) so there was some “unblinding”. A lot of chatter on Twitter, people reporting on their experience. I don’t know what proportion of participants would be affected by this info going around, but this is one aspect of the whole story. With so much spotlight on these trials, the participants were paying more attention to reports on social media than they would in any other kind of trial.

          1. Bob says:

            Why don’t they put an adjuvant in (some of) the placebos to more accurately simulate being does?

          2. Marko says:

            “Why don’t they put an adjuvant in (some of) the placebos to more accurately simulate being does?”

            Because then the placebo becomes an active intervention. Vaccine adjuvants alone will stimulate innate immunity , which may offer some protective effect for a time.

            It would have made for an interesting experiment , though.

        3. paperclip says:

          No trial participant knows for sure, but if you get jabbed with saline and the only side effect you feel is a sore shoulder, you might correctly guess what you got.

          1. Dan says:

            Hopefully the study subjects that do get side effects and consequently presume they were vaccinated will understand that this is an EXPERIMENTAL vaccine and it may not work. So they would be as cautious as those that presume are on placebo. I would hope this is explained during the IC process.

        4. David E. Young, MD says:

          Nope…. it is pretty obvious. When you get a shot that doesn’t hurt… at all, and you get no redness, no swelling, no aches and pains, no headache, no fever, no fatigue or sleepiness…. at all, you can be pretty sure you are on the placebo arm. And yes, you will continue to be very careful to wear your mask, avoid crowds, wash hands, etc.

          When I go to my vaccine trial site, and see the principle investigator there, it is pretty much, “wink, wink…. I know that you know that I know that I am on the placebo” and there is subtle confirmation of that.

          1. J Tyson says:

            I wish you’d stop interfering with the trial by helping people unblind themselves.

          2. Irene says:

            I have had multiple immunizations that have had no obvious side effects, not even a sore arm. It’s far from impossible.

          3. Paul Jakma says:


            The investigator who deals directly with you surely should have no possible way to know which arm you are on, in any credible pharmaceutical / vaccine trial, precisely to avoid any risk of the investigator letting slip the arm to the subject? I.e., this should be a double-blind study.

            I’ve not been on a large-scale medical trial, but I’ve been on much trials in (less rigorous / much lower stakes) other fields, and they took care there to ensure double-blinding.

            Are you saying you know for sure the person dealing with you knew, and that this Pfizer / BioNTech study was not blinded?


          4. Paul Jakma says:

            Reading the Pfizer study design, the investigator /should/ have been blinded.

            The person administering the dose to the subject is not blinded, but their contact with the subject is meant to be limited as much as possible.

          5. Chris Phillips says:

            The thing is, if some people do decide they’ve had the vaccine because they get a reaction, and as a result behave in a more reckless manner, that will tend to reduce the apparent efficacy of the vaccine, not increase it.

            That behaviour can’t produce a finding of “false efficacy”.

          6. John says:

            What a lot of rubbish. The study is double-blind so the doctor doesn’t know. And that would invalidate practically all the clinical trials ever done. “MD”, really ?

    2. Patrick says:

      I’m going to ignore the commentary on the protocol change – I’m not really sure how to engage with that – and discuss the second comment.

      It’s really easy for this to provide efficacy information to a certain level of statistical confidence. They can bound it with a certain number of cases. The information “this many cases have occurred and this many were in the vaccine arm and the trial arm” can be used – combined with a robust estimate of the overall probability of infection in the trial population – to calculate the probability that the vaccine is a certain percent effective. There is absolutely nothing Pollyanna-ish about it at all. You calculate the *probability* that certain events occurred by chance vs were caused by various degrees of vaccine efficacy.

      The data allows you to calculate the probability of a range of efficacies. This is relatively simple hard science. More data just firms up those estimates. (And the quality of the estimates was dramatically improved by the protocol change to wait for more cases.)

      1. Patrick says:

        To add a little more:
        There is a non-zero probability that the rest of the trial is going to read out horribly, with infections equally in both groups, with the difference between the vaccine and placebo groups disappearing.

        But we *know what that is*, we can calculate it, and when we say we’ve got a readout like this, it means we’ve judged that probability is below a certain threshold. (How strong the evidence is, ie, how low the probability the events occurred by chance, rises as you assume a lower vaccine efficacy. When they report >90% efficacy, it means that some value over 90% efficacy has passed some probability threshold they’ve chosen. Perhaps .05? Perhaps much lower? And, critically, if they’ve got, say, a 95% (1-.05) chance of (say) 96% efficacy, then they’ve got a higher chance of 90% efficacy, and a hugely higher chance of, say, 80% efficacy.)

        1. MathematicallySceptical says:

          The big question is whether the low number of infections in the vaccine arm is in significant part just noise given the overall much larger size of both arms of the trial.

          It’s not so easy to calculate how probable that is. I would hesitate to make any kind of statement like this, myself.

          I am hopeful btw. But wouldn’t rush.

        2. Mike says:

          “There is a non-zero probability that the rest of the trial is going to read out horribly, with infections equally in both groups, with the difference between the vaccine and placebo groups disappearing. …

          But we *know what that is*, we can calculate it,”

          Careful– not quite! We know what the probability is that we’d get a result like this if the vaccine did nothing. This is not quite the same as knowing the probability the result is nothing.

          Say I tell you I’m giving you a vaccine candidate, and you run a trial on it and get similar results to this 90 days in– but I actually just gave you a saline solution. You have a very small chance of continuing to get lucky and continuing to get good results through the rest of the trial, versus if it was a credible effort at a vaccine. That is, the prior matters, and we can’t measure the prior in any kind of rigorous way.

        3. No, the probability of the rest of the study producing 0% efficacy, equal numbers in each arm, is astronomically small, it is essentially zero. The p-value corresponding to the interim analysis is <0.00000000000000001 for the null hypothesis that the 0% efficacy.

          Using Bayesian predictive probabilities, assuming the interim results were 8/94 of infections in vaccine group, 86/94 infections in the placebo group, here are the probabilities at the end of the study:

          1. Probability that remainder of study, 70 events, have a 0% efficacy, 35:35 ratio, is 0
          2. Probability that the remainder of the study, 70 events, have at least 80% efficacy, 12:58 ratio, is 0.95
          3. Probability that at the END of the study, after 164 events, the overall efficacy is at least 50%, which would be at least 55:109 ratio, is 1
          4. Probability that at the END of the study, after 164 events, the overall efficacy is at least 80%, which would be at least a 27:135 ratio, is 0.9993
          5. Probability at the END of the study, after 164 events, the overall efficacy is at least 90%, which would be at least 15:149 ratio, is 0.65
          6. Probability at the END of the study, after 164 events, the overall efficacy is at least 86%, which would be at least 20:144 ratio, is 0.95
          7. Probability at the END of the study, after 164 events, the overall efficacy is at least 84%, which would be at least a 23:141 ratio, is 0.99

          So this is extremely strong evidence of considerable efficacy, and it is almost certain to be north of 80% efficacy.

    3. Steven says:

      They haven’t said what number of cases were in the placebo vs vaccine groups. It may even be possible that all of the cases were in the placebo group, but because of the small sample so far the efficacy can only be said to be statistically significant at the 90% level or better.

      Nothing is 100% effective, so why overstate it and disappoint people later? Better to understate it and hedge a little bit. 90% is still great.

      With more data, they should be able to put a tighter bound on it.

    4. Joseph Severs says:

      Hello Li Zhi: yes, having all 94 cases in the control arm would be plenty sufficient to conclude that the true efficacy is greater than 90%.

    5. Based on the endpoint in the protocol, the 90% efficacy likely means that of the 94 infections, 8 were in the vaccine arm and 86 in the placebo arm, which would give an efficacy of 100 x (1-8/86) = 0.907. This would means that of the 86 infections that would have been expected in those assigned to the vaccine arm had they not been vaccinated, the vaccine prevented 79, or 90.7% of them. This is higher than most dared to hope for and is substantial. BTW, testing against a null hypothesis of 0% effectiveness, i.e. vaccine doesn’t work at all, the p-value is <0.00000000000000001, and testing against a null hypothesis of 50% effectiveness, the lower bound established by the FDA, the p-value is 0.000000015, so this evidence is beyond substantial the the vaccine works, at least in the time frame and setting of the study.

      1. Pat Collins says:

        How do mitigation factors (staying home, mask wearing etc) impact the study? Could some people not show symptoms simply because they haven’t been exposed to covid, regardless of whether they received vaccine or placebo?

  8. Anon says:

    “In fact, they said at the time that working in the “Operation Warp Speed” framework would likely slow them down, and it looks like they can make a case for that!”

    This certainly seems to be the case. I’ve heard rumors from some of the academic institutions that joined Operation Warp Speed, and it doesn’t seem like anything is particularly speedy. In short, it seems that it takes way longer than a month or two to get an academic research group up to a GLP standard, and even with practically unlimited money you can’t instantly equip a lab for high-throughput experiments. For one thing, if you have a working BSL-3 lab, even simple matters of equipment delivery and user training become a lot harder to manage… can’t just call up the vendor rep and have them pop over for every little issue with that shiny new liquid handling robot that none of the bench scientists have ever used.

  9. Theo says:

    Derek, it looks like the link to the post on trial design in the second line of the second paragraph is broken.

  10. artificial chemist says:

    Can anyone help me understand the following? They have reached the number of 94 positives. Is that 94 positives in the placebo group? Or the overall total placebo AND treatment? And is the placebo group equally big to the treatment group?

    1. Michael says:

      From what I understand, 94 participants reported PCR-confirmed symptomatic COVID across the entire study. And the vaccine/placebo ratio was 1:1, with more than 90% of the ill participants coming from the placebo arm.

      1. M says:

        This is my understanding as well.

        1. artificial chemist says:

          Thanks. Indeed promising 🙂

          I wonder if the world put as much effort and money into a malaria vaccination….?

          1. sort_of_knowledgeable says:

            A malaria vaccine is more difficult for a number of reasons such as the fact that it reproduces sexually providing more genetic variability.

  11. exGlaxoid says:

    This is great news, and likely the Moderna vaccine will also be efficatious, based on its similar design, They should be able to release 10’s of millions of doses per month, so even with the logistics issues, that will move us forward quickly. If they can just dose the healthcare workers first, then assisted living staff and patients, then teachers, it would move us a long way forward.

    1. Ken says:

      My cynical side notes that you left “politicians”, “billionaires”, and “investment bankers” off your prioritized list.

    2. John Stamos says:

      I’ll be god damned if teachers get it before age 60+ category.

      1. Brian says:

        The question is: Do you protect more people by vaccinating teachers, supermarket cashiers, bartenders and others whose jobs bring them into contact with a lot of people first or do you protect more people by vaccinating the 60+ crowd first. The answer isn’t simple. The hope is that enough vaccine can be manufactured and distributed fast enough to make the argument moot.

        1. Hopeful Layman says:

          For that matter, some people suggest that the best route is to vaccine potential superspreaders first, which would mean younger people (however we choose to define “younger.” There’s good case, epidemiologically, to be made for all of these suggestions. I think, though, that right now the general consensus is to go with medical care providers first, followed by the highest-risk patients, then “essential workers.” The precise definition of “essential workers” is still being debated, I think, but at least some scenarios definitely include teachers in this category.

          1. Bob says:

            That should include the immuno-comprised. See the recent story of the elderly leukemia patient who was shedding viruses for 70 days.

          2. achemist says:


            That has a great chance of not working.

            A vaccine requires a working immune system to do anything.

        2. Susie S'bar says:

          Regarding Brian’s post, do we know how the effectiveness of the vaccine varies with the age of the recipient? If (say) the vaccine’s 100% effective in 20-year-olds and only 50% effective in 60-year-olds, we might be better off focusing on the former, even if they’re less likely to suffer serious consequences from coronavirus infection.

          1. MrRogers says:

            Phase II data show that the vaccine generates an antibody response in older adults that is just as robust as that generated in younger people. Presumably that means that efficacy will be just as high, but that will be a secondary endpoint in this trial.

        3. bill l says:

          You would like a lot more data first before answering the question about who to vaccinate first.

          Is this protecting against infection or disease? mild-disease and severe disease? How does that differ by age?

          You wouldn’t want to create an army of infectious asymptomatic front-line workers. Not unless you could mitigate that with an effective testing plan.

          1. sort_of_knowledgeable says:

            Sterilizing immunity has not been tested because it is much more expensive to track everybody who had contact with any trial participant, but it is reasonable to expect that some sterilizing immunity is provided. Even if no sterilizing immunity is provided, it is still a win if frontline workers don’t become sick, but just have to isolate themselves for a week.

  12. M says:

    With this and all the other vaccines targeting the spike, I agree with Derek’s statement that we are going to beat this. If there are spike mutations (from what I understand, most of these lead to a sucky virus from the perspective of the virus) that lead to a different epitope that is still potent, we will know what the sequence is quickly, and we’ll have the infrastructure to produce a “strain relevant” vaccine if we need to. There’s still a lot of work to do, and things to better understand, but we’re in a much better place now. Gotta keep fighting.

  13. Kathryn Storer says:

    I wonder if the companies were able to make a preliminary extrapolation of duration of efficacy – at least one year- in part from the phas 1/2 trials that started earlier this year

  14. Jonathan says:

    Can someone address what “more than 90% efficacy” means?

    Does it mean 90-91% efficacy, or between 90% and 100% efficacy, but we are rounding down to not get everyone too excited?

    1. M says:

      > 90% efficacy means somewhere between 90% and 100%. Where, exactly, I don’t know. And it’s not exact anyway, as there’s surely a confidence interval associated with the value, and I don’t know what the CI is.

    2. Just another chemist says:

      Based on my understanding is that this is between 90-100%. Meaning that of the 94 people that got COVID-19 in the trial between 0-8 are in the vaccine arm vs 86-94 in the placebo arm

    3. JIM OTTO says:

      You can think it this way for 94 infections:
      0% efficacy = 47 infections in each arm
      50% efficacy = 24 infections vaccine, 70 infections placebo
      75% efficacy = 12 vaccine, 82 placebo
      87.5% efficacy = 6 vaccine, 88 placebo
      93.75% efficacy = 3 vaccine 91 placebo

      So if you’ve got 90+% efficacy, that is spectacular

      1. anon says:

        Hi Jim OTTO,

        could be so kind an include the calculation? I get to different numbers. Let’s say 6 people are infected in the vaccine arm and 88 in the placebo arm.

        ((Attack Rate Unvaccinated – Attack Rate Vaccinated) / Attack Rate Unvaccinated) * 100

        (((88/22000)-(6/22000))/(88/22000))*100 = 93.18%

        Or am I missing something?

      2. Based on the endpoint in the protocol which is public,

        assuming equal numbers per arm:
        efficacy = 100 x (1- X1/X0) where X1=number of infections in vaccine arm, X0= number of infections in placebo arm.

        Based on that 90% efficacy is X1=8 and X0=86 or so, which would have a p-value of <0.00000000000000001 for testing the null hypothesis of 0% efficacy (equal infections in placebo and vaccine arm) and p=0.00000015 for testing the null hypothesis of no more than 50% efficacy, the threshold the FDA has set as minimal.

        This is very strong evidence.

    4. achemist says:

      It means there is a certain percent (often 95%, but I dont know what was used in this study) chance that the vaccine is between 90 and 100% effective

  15. Roy says:

    Good article in the Wall Street Journal on Pfizer’s work on setting up a distribution network for this:

    1. JasonP says:

      Roy, per 60-Minutes last night the US Army has been working on distribution plans for the vaccine for MONTHS. Not sure why Pfizer proposes to do its own thing?

      1. Dark Day says:

        Maybe the Army distribution strategy will apply only to Operation Warp Speed vaccines?

      2. Irene says:

        Because of the special refrigeration requirements, maybe?

      3. Wilhelm Cody says:

        Pfizer is distribution to many other locations than the USA. They need a plan for all the various countries. Of the initial doses made and ready by approval, approximately 20 million (10 million treatments)are for the USA (out of 100 million paid for) and the rest go elsewhere. They plan to have 1.5 billion does(750 million treatments) by the end of 2021.

    2. Geoff PR says:

      Per the WSJ article –

      Excellent, the distribution network looks close to ready-to-go (if not already at 100 percent).

      Has there been any informed speculation on how long it will be before Pfizer is regularly shipping in quantity?

      (Purely selfish reasons, I have an 84 year-old parent…)

  16. luysii says:

    Trump would have won had this come out a week ago. Such is fate. 20 years ago Gore would have won if the ballot in Florida had been written correctly. As a result a bunch of confused elderly Jews in Palm Beach mistakenly voted for an antiSemite (Pat Buchanan) depriving Gore of their votes and giving Bush the victory. My wife’s cousin’s daughter was a reporter for the Miami Herald and had to watch the counting of the chads. She met the designer of the ballot, describing her as ‘an idiot’.

    1. Peter Joseph says:

      My accountant, who got things right down to the last dime, lived in Palm Beach. When I asked him who he’d voted for (Gore, of course), he replied “I don’t goddammed know.”

  17. Marko says:

    Between the election and vaccine results of the last several days , the end appears in sight of two plagues against humanity. Not a bad week’s work.

    New plagues are inevitable , however , both biological and political. Their extermination will be an ongoing battle , probably forever.

    1. Tom says:

      Thanks to your new China puppet President we will live in eternal lockdown. There will never be an end to the social and business restrictions. It was always supposed to be permanent and vaccine efficacy and uptake are completely irrelevant. It’s about controlling people, not controlling a virus.

      1. confused says:

        Even if Biden wanted that (and he doesn’t) that is not actually within the President’s power (as Biden acknowledged when talking about mask mandates) – the mandatory public health stuff is mostly a state matter.

        Places like South Dakota can keep doing what they’re doing regardless of who’s President.

      2. James says:

        You’re in the perfect place for people with your condition, we have Risperidone, Quetiapince (beautiful molecule), Olanzapine … your choice. Now, back to check monitors, chop chop

  18. Hopeful Layman says:

    Very encouraging — and all this time, I thought Slaoui was being ridiculously over-optimistic, given his self-interest as head of Operation Warp Speed. Regardless of how this plays out in the long run — whether that “90%” ends up being somewhat compromised by further data, margin-of-error considerations, or some other factor– this is definitely a sign that a vaccine CAN work. Of course, now the major question is long-term, both efficacy and safety. My gut feeling is that there’s more of a chance that whatever long-term immunity results will fade over time (hopefully, not less than a year) than that any catastrophic Cutter-like safety incidents will take place.

  19. Walter Sobchak says:

    I would not be worried about dry ice shipping availability overly much. It is routine to order ice cream over the internet, and have the ice cream company pack it in dry ice and have FedEx or UPS deliver it overnight in good condition.

    FYI absolutely the best ice cream in the world is Graeter’s from Cincinnati Ohio. I love the Mocha Chip. If you love chocolate try one of their chip flavors. The Cinnamon is also spectacular.

    1. ice-olation says:

      This seems to be an American thing, I had never heard of overnight ice cream shipping before.

      1. sc says:

        I didn’t either until this year, but I’ve had ice cream and meat delivered that way now. UPS managed to deliver thawed meat with no remaining dry ice twice, but I suspect (and hope!) vaccines will get treated a little better.

      2. FoodScientist says:

        Dry ice is used because it’s ice, but dry. Also the small ice crystals melt lower than 32f due to the sugar content. The smooth and creamy mouth feel would deteriorate.
        Jeni’s is spectacular. $11 for a pint might seem expensive, but it’s very tasty.

    2. Anon says:

      Yeah because shipping a few ice cream packages is the same as shipping millions of doses of vaccine. Lol

    3. debinski says:

      Yes, Graeter’s! I miss it so much.

    4. Chrissy says:

      I’d go with Dietsch Brothers, instead, but either way, ice cream just needs to be a few degrees below 32F/0C. Tons easier than keeping something at -80C. But, yeah, overnight shipping is fairly commonplace here.

    5. Driv—DidSomeOneSayMintChipGraeters? says:

      Nooo, don’t tell people about Graeter’s! They have like one factory with their specific equipment. I don’t wanna hafta bribe the grocery manage to get mine 😉

    6. GamingBuck says:

      Graeter’s is indeed great. If you love great ice cream though you should try Jeni’s. Some odd flavor combos, but it’s truly top notch (probably better than Graeter’s, but at least on par).

  20. Hopeful Layman says:

    p.s. RE: Distribution — Especially if one or more of the other vaccines with less-challenging distribution requirements comes out looking promising, I wonder if the Pfizer/BioNTech vaccine could be distributed in limited numbers to specifically targeted populations (maybe the “first phase” of medical care workers and high-risk patients). That might make things at least a little more logistically manageable.

    One caveat/query: If I recall, early trial participants reported some pretty dauting side effect of the Pfizer/BioNTech vaccine — fevers of over 101 degrees F, wracking chills (one actually chipped a tooth from grinding his teeth while shivering under his electric blanket), serious body aches, etc. They all said they thought it was “worth it,” but of course as volunteers they came into this thing with a different mindset than most people would probably have. Do we have any data on side effects, or does “safety” only include potentially debilitating or fatal effects? I think some transparency here will be essential.

    And yes, the fact that these results came out AFTER the election? Maybe there is a God after all!!!

    1. JIM OTTO says:

      Most people over the age of 50 go through the same thing with the two dose shingles (Shingrix) vaccine. These side effects are not uncommon for two dose vaccines

      1. Hopeful Layman says:

        I’m going to be getting Shingrix II at the end of the month. We’ll see what happens . . . maybe a “dry run” for the COVID jab a few months down the road?

      2. The Moar You Know says:

        55 years old. The second dose of shigrix was brutal. Better than shingles to be sure, but damn, get it on Friday, clear the weekend, and be prepared to call in sick on Monday.

        1. Hopeful Layman says:

          I’ve heard that — from what you’re saying, it might actually be rougher than the COVID vaccine, whose effects apparently don’t last longer than 12- 24 hours.

          1. Sc says:

            The pfizer vaccine made me feel extremely crappy for a bit there. The second injection was a bit weird, I only had arm soreness after the injection in the morning, but got some pretty rough fatigue and general flu-ish feeling in the evening and through the next day. It definitely impacted my productivity, but the next morning after that it was all gone.
            Now I just have the issue that I can’t sleep because my immune system is so powerful I glow in the dark.

          2. Dark Day says:

            Well, I do hope that there’s honesty and transparency about this when the vaccine is rolled out. People have to know what they’re in for, and understand that it is, indeed, “worth it” (for themselves, for others, and for society at large). The last thing we need is social media exploding with raving rants about how unexpectedly horrible the vaccine side effects were, followed (no doubt) by conspiracy theories and made-up anecdotes about “hidden” adverse reactions, deaths, etc. If ever honesty was “the best policy,” this is the time. If that 90% efficacy rate holds up, we really are looking at a strong possibility that this thing could be well under control before the end of 2021, but it will still take widespread uptake for that to happen.

  21. steve says:

    A few things. First, Pence lied about it and Pfizer immediately corrected – the Trump Administration had nothing to do with Pfizer’s success. Second, 90% efficacy in terms of what? Antibody production? They don’t have enough COVID cases to say what the efficacy is in terms of disease prevention. Third, this is a select population, we don’t know the efficacy once it hits the general population and in particular we don’t know efficacy for the at-risk group. Finally, as Derek mentioned, there is the whole -80oC cold chain storage issue. It doesn’t even seem very stable after defrosting and before the patient receives it. How this will work in the real world remains to be seen.

    1. Marko says:

      “…90% efficacy in terms of what? Antibody production? ‘

      No. Efficacy in terms of preventing symptomatic COVID-19 that is verified by PCR.

      “They don’t have enough COVID cases to say what the efficacy is in terms of disease prevention.”

      They do in terms of the disease endpoint described above. The interim data analysis was pre-specified. Significance of effect upon the chosen endpoint is as meaningful in an interim analysis as in a final analysis , except for the longer-term safety and durability of response information that will be gleaned if the trial proceeds to completion.

      1. MathematicallySceptical says:

        I am not sure what are you talking about.

        94 cases are not in any way enough to claim anything about effectivity in quantitative terms. It is ridiculous to claim otherwise. It is enough to start hoping the vaccine is effective, that is all.

        1. steve says:

          Thanks. Exactly my point.

        2. Sc says:

          It’s absolutely enough when the signal is this strong.

          1. Marko says:

            You’re not going to be able teach them about statistical significance in this comment thread. Better to let them just rage against facts.

          2. MathematicallySceptical says:

            Lol wut?

            You mean like a magical signal? Like when you just feel it and bet on the red? Or buy 10000 dollars worth Moderna stock?!

            Surely not anything to do with logic or reasoning or statistic or any kind of that boring stuff…

          3. Marko says:

            ” Let’s say the mean of the distribution is 90 (suppose this is the number of infections in placebo group) giving us the rate of 0,046 ”

            No , the rate is 0.0046. Only wrong by a factor of ten.

            I can see why you’re sceptical of math. In your hands , math doesn’t work.

        3. Ian Malone says:

          Given your chosen name is “mathematically sceptical” I can see I may have trouble convincing you, but no, this is 94 people contracting the virus in a trial of over 43,000 participants. The minimum number of those to be in the control arm consistent with 90% efficacy is 85, with 9 in the treatment arm. The probability of observing that outcome (or better) given a particular efficacy can be calculated trivially in a binomial distribution.

          pbinom(9, 94, c(0.1,0.2,0.3)) # probability of 9 or less in 94 trials at p=0.1, 0.2, 0.3
          [1] 5.321499e-01 5.165619e-03 1.865777e-06

          I.e., with 10% chance of a given infection being in the treatment arm, 90% in the control arm (not exactly the same as 90% efficacy, but close at this limit), there’s about 50-50 chance of these results or better (fewer infections in the treatment arm), rising to 20% chance (~75% efficacy) it becomes about 0.5% chance we’d get these results, at 30% chance (~60% efficacy) these results are about 2 million to one.

          Or… flip a coin 94 times, count how many heads you get. It’s going to be more than 9, I’ll be quite surprised if it’s less than 30.

          1. MathematicallySceptical says:

            Dear Ian,

            I really hope you yourself understand that your reasoning works very well indeed for the case of computing the chance of 96 heads in 100 experiments, but not so well when we talk about population of tens of thousands of whom just short of a hundred (we believe? Or maybe more ?) fall Ill at all with probability of illness being unknown. Does your expertise in mathematics include only discrete probability or also some statistics?

          2. sort_of_knowledgeable says:

            The control group provides the probability and treatment is based off that. If 97 people in the group became ill and 2 people in the treatment became ill then the treatment is at least 90% effective to a high degree of confidence, assuming the 2 groups are the same size and otherwise comparable. There are 10s of thousands of people in the trials so that we can get to over hundred people becoming ill in a short time, but that doesn’t change the relative difference between the groups.

          3. Chris Phillips says:

            I suppose one would need to do a proper Bayesian calculation to work out a confidence interval for the efficacy. As far as I can see, this kind of calculation isn’t specified in the protocol, only calculations of the probability that a (rather low) threshold for efficacy has been surpassed.

            But I would hope that “greater than 90%” will be based on a properly calculated confidence interval, so we can say there is only a small probability that the efficacy will turn out to be less than 90%. Whether it’s worth trying to work out the actual number of infections in the vaccinated group at this stage I’m not sure.

          4. Ian Malone says:

            “I really hope you yourself understand that your reasoning works very well indeed for the case of computing the chance of 96 heads in 100 experiments, but not so well when we talk about population of tens of thousands of whom just short of a hundred (we believe? Or maybe more ?) fall Ill at all with probability of illness being unknown. Does your expertise in mathematics include only discrete probability or also some statistics?”

            Well then, hands up, I’m not a professional statistician, I’m a scientist who uses statistics regularly. And I understand my reasoning here very well thank you. That population of >38,000 included in analysis is actually better for this analysis than if it was only 100:

            1. It lets us assume the groups are identical as they have been randomly assigned in a double blind design.
            2. It means that the draws (infection events) are effectively independent since there’s a much smaller number of them than the group size, for each case it could (under the null hypothesis) have been drawn from either the treatment group or placebo group equally, since the chance of drawing from 19000:19000 is basically the same as from 19000:18999.

            If you prefer to care about the group size, in the controls we have (at least, since Pfizer don’t say), approx 85 of 19000 infections giving a rate of about 0.0045, and probability of getting 9 or less from the control group at that rate is p=7*10^-26, for 0.1 of that rate it is p=0.65 and for 0.2 it’s p=0.026 and falling rapidly. However these figures are problematic since it doesn’t take into account the variation in the control infection rate, while the calculation on the infections ratio has it baked in.

            If you prefer your explanations more “statistical” (but those probability calculations /are/ statistics), then sticking 9 and 85 out of 19000 each into a risk ratio calculator (fmsb) gives 95% confidence intervals 0.05 to 0.210, which is in quite good agreement with David’s analysis elsewhere in the comments. The overall trial size you can vary by a factor of ten either way and barely move those confidence intervals, provided you’ve still got 9 in one group and 85 in the other (of course the reason for the trial size in the first place is the infection rate is out of their control and it needs to be big enough to get those ~100 events).

            (Bonus: that upper confidence interval 0.210, corresponds to p=0.174 of an infection being from the vaccinated group (0.174/0.826=~0.210), pbinom(9,94,0.174)=0.025, i.e. the upper 2.5% tail of the 95% confidence interval, it’s not an accident.)

          5. MathematicallySceptical says:

            Dear Ian,

            Perhaps to just elaborate on what I have said with some numbers:
            In this case you can well assume the distribution to be normal because the number of participants was large and so the binomial distribution is well approximated by the gaussian pdf (btw this is generally the reason for taking large number of participants in studies as you might well know yourself, although from your message it was not completely apparent Another reason is of corse that the main target of this trial was safety). Let’s say the mean of the distribution is 90 (suppose this is the number of infections in placebo group) giving us the rate of 0,046, then the variation is 85.5. This is your probability distribution for infections without any interference. The chance of getting just 9 infections with such distribution is not very small – it is 0,17, or 17 percent. That is if you just injected everyone with saline instead of vaccine you would get with probability of 17 percent just 9 infections. This is not such a surprising result in other words.

            Intuitively speaking it is because 90 and 9 are not that far apart on the scale of 19000…which is clear without any calculations or calculators 🙂 just a healthy common sense with a bit of experience perhaps?

            I am very hopeful for this vaccine (and I am sure every human on this planet is, including antivaxxers), but I am just averse to incorrect use of mathetmatics.

          6. Marko says:

            ” Let’s say the mean of the distribution is 90 (suppose this is the number of infections in placebo group) giving us the rate of 0,046 ”

            No , the rate is 0.0046. Only wrong by a factor of ten.

            I can see why you’re sceptical of math. In your hands , math doesn’t work.

            ( reposted here on correct thread )

          7. sort_of_knowledgeable says:


            Intuitively speaking if a casino got 90 big multi million dollar winners instead 9 among 100 million players, the casino is launching a huge investigation because that is not a 17% chance.

            You can do the calculations on the trial numbers at this website .

            Assuming the probability of being sick is about 0.004734
            probability of success on a single trial = .004737
            number of trials = 19000
            Number of successes (x) =90
            and you will find
            Cumulative probability: P(X x)
            so 90 is about the expected number

            now change the number of successes to 9
            and you will find that
            Cumulative probability: P(X > x) > 0.999999
            The chance of getting 9 instead of 90 is less than 0.00001%

            The difference between 9 and 90 is large whether the trial is sized 100 and 20 thousand

          8. Ian Malone says:

            Others have already commented, but let’s explore why you might have gone wrong here. First a sanity check on the numbers:

            “Let’s say the mean of the distribution is 90 (suppose this is the number of infections in placebo group) giving us the rate of 0,046, then the variation is 85.5.”

            I can’t actually get that variance (variation?) from any of the figures it might be. The binomial distribution variance is Npq = Np(1-p), standard deviation is the square root (n.b. at this point I don’t know which you mean by “variation”). Using the incorrect 0.046 figure this gives σ^2=834, σ=29 (rounding). Using 0.0046 (perhaps it’s a typo) I get σ^2=87.0 (closer to the number you give), σ=9.3. (N.B that standard deviation, not variance, is the relative measure for distance from the mean.) 90/19000=~0.0047, a little higher goes the wrong way, σ^2=89.6, σ=9.5, but rounding errors count for a lot here, so the variance can shift by about 1 depending where you round off on p).

            Let’s assume we want to use the Gaussian distribution. While, yes, we generally like big numbers because it lets us use central limit theorem and Gaussian assumptions (not always correct, and the reason we have non-parametric tests), there really is no reason here to do so, the binomial probability is computable and exact. However the two should agree.
            Starting from the incorrect rate estimate of 0.046 (which corresponds to 874/19000):
            Assuming the mean was 90 (incompatible with the variance estimate, but hey, we’re exploring):
            [1] 0.00261025
            Quite low, this is 2.8 standard deviations away from the mean. Again, standard deviation is the normalising factor for difference from the mean, so we could equivalently write z=(N-μ)/σ=-2.79
            [1] 0.002635402
            What does the binomial distribution say?
            [1] 0
            Essentially impossible. They disagree here because the mean and standard deviation don’t agree; this is not an approximation to that binomial.

            If we use the mean that would actually correspond to that standard deviation then we get a much lower answer, because we’re now at z pnorm(70,90,9.46)
            [1] 0.01725059
            > pbinom(70,19000,0.004737)
            [1] 0.0168282

            Not perfect, but not terrible, two significant figures of agreement.

            Finally, what if I *incorrectly* use the variance? It’s 89.6, so, back to n=9 cases:
            [1] 0.182993

            This is your roughly 17% answer, it’s not correct, as Npq is the variance of the binomial rather than the standard deviation. We already know what the answer should be. What if we use it within a couple of s.d. instead (our 70 cases example)?

            [1] 0.4116844
            Remember this should agree with the binomial answer, which was 0.0168.

            So, first thing, Npq=σ^2, variance, not standard deviation. Intuitively Npq would be unlikely as standard deviation, since a value near the lower or higher end of possible outcomes (p close to 0 or 1) would be about one standard deviation from the limits.

            Next point, and this is the one I wanted to make: μ=Np, so μ/σ = sqrt(Np/q), as the size of the sample increases the ratio of the mean to the standard deviation increases, the binomial distribution gets *relatively* tighter. Also, say we’re dealing with a fixed μ, as in this case (those 90 cases). Well, p=μ/N. σ=sqrt(Npq)=sqrt(μ(1-μ/N)), this places an upper limit on σ, once N gets big enough it sticks to sqrt(μ), sqrt(90)=~9.49. This is why 90 really is a long way from 9 even if it might not seem like it; it’s at least 8 standard deviations out. The result is you *can* make the assessment without a calculator, but the result is actually that an outcome of 9 or less is really improbable by chance.

            I think the point to start questioning your working was when it diverged so far from the binomial solution, it’s not really more “professional” in some sense to be working with the normal distribution.

          9. Ian Malone says:

            Somehow I cut a section out here while editing:
            “If we use the mean that would actually correspond to that standard deviation then we get a much lower answer, because we’re now at z [stuff goes here] pnorm(70,90,9.46)
            [1] 0.01725059
            > pbinom(70,19000,0.004737)
            [1] 0.0168282”

            The missing part was that the calculation corresponding to p=0.046, mean=874 gives z < -29 and extremely low p:
            [1] 8.574011e-196
            Which is difficult to compare to the pbinom(9,size=19000,prob=0.046) = 0 result due to precision, but both extremely low.

            The next bit starting pnorm(70,90,9.46) was to illustrate that agreement between the binomial distribution and normal approximation improves as you move away from the tails, for 70 treatment cases the p results agree quite well, while for 9 treatment cases in 19000 vs mean of 90 the normal approximation gives p~=6*10^-18 vs 8*10^-28 for the binomial result, due to the relatively heavier tails of the Gaussian (both very low, but a big ratio between them, compared to the 2s.f. agreement nearer to the mean).

          10. MathematicallySceptical says:


            Well I have to apologise – I did mistakenly plug sigma^2 rather than sigma (ie variance rather than standard deviation) in my matlab function when computing my estimate which explains the rather large error margin that I got *^^*. Your calculation of magnitude of error is therefore correct (which is good news for Pfizer and all of us of course). In other words, yes this data tells us (assuming our two trial strands are indeed identical in this respect) that the vaccine will be around 90% effective with very reasonable confidence.

            Of course this is if we believe that the two strands of the trial behaved and are exactly the same – heuristically it still seems to me that it would take a very slight difference in the composition of both arms to change the infection numbers quite significantly. However I must concede that my crude attempt at quantifying this has failed 😛

          11. Ian Malone says:

            No worries, we all make mistakes. Fingers crossed for more data.

        4. achemist says:

          Statistics says otherwise.

          Of course there is a confidence interval attached to that, but that goes for every number in clinical trials.

        5. 94 infections provides more than enough evidence in this context. There are ~44,000 patients, and the trial is designed to run to a certain number of total confirmed sarscov2 infections, 94 here, and then testing whether they are equally distributed between placebo and vaccine arms, indicating the vaccine doesn’t work at all, or is mostly concentrated in the placebo arm, indicating the vaccine prevented infections that presumably would have occurred sans vaccine.

          90% effective means of the 94 cases, at least 86 were in placebo group, and no more than 8 in the vaccine group. If the vaccine did not work then we would expect 86 cases in that time in the vaccine group as well, so 8 represents a 90% reduction.

          These numbers are phenomenal, and more than anyone dared hope for (the FDA only requires 50% efficacy for approval).

          The p-value for testing whether the vaccine has any effectiveness is <0.00000000000000001, and is as close to a statistical certainty as possible.

          It is certain that this meets the efficacy bar for FDA approval in terms of efficacy and the probability that at the end of the study (after 164 infections) it will meet the efficacy bar set by the FDA is 1.00 — certain — and comparative safety analysis will be done later this month.

          There are important caveats that cannot be answered right now — how well it works in subgroups like at risk groups including older people or with preexisting conditions, how well it prevents severe disease when exposed to high viral levels, how long the immunity lasts, whether there are long term side effects, whether it might exacerbate severity of disease for those who do get infected after vaccinations … all these remain to be answered by future data.

          And distribution is a real challenge given 2 doses are needed and the vaccine must be stored at -94 degrees F.

          But this is the best news we could have hoped for at this point and should be celebrated.

          And while the company did NOT accept the Operation Warp Speed $ for the phase 3 study, choosing to fund it themselves, they DID accept the Operation Warp Speed offer to guaranteed buy $2 billion worth of vaccine if it was approved and help in distribution, which is non-trivial benefit given the distributional challenges will immediately put this vaccine in line behind all others if they are proved more effective and still safe, so they clearly benefited from the federal program.

  22. Toni says:

    It would be very interesting to see how big the difference between the groups is in terms of virus detection. Can one expect that the immunised ones are also 90% PCR negative?

  23. David E. Young, MD says:

    There may be merit on a second trial….. take those who had the vaccine and randomly assign them to a third booster at 6 months or placebo shot at 6 month and then follow them for 3 years. Might be useful data.

  24. Hopeful Layman says:

    A damper on the enthusiasm?

    “The nonprofit group Public Citizen released a statement on Monday calling for more data. ‘The release of preliminary and incomplete clinical trial data by press release to the public is bad science,’ Michael Carome, director of the group, said in the statement. (From CNN)

    1. Michael says:

      I mean, in a vacuum, yes, but it’s surely material information requiring disclosure for a reporting issuer.

  25. Enufwork says:

    Will the other vaccine candidates also have ultra-cold storage requirements?

    1. Hopeful Layman says:

      Moderna requires -20 C, if I’, not mistaken. Not quite as daunting as this one, but not a walk in the park, either.
      There is a discussion about this on this blog (from a few months ago) . . .

      1. Marko says:

        “Moderna requires -20 C, if I’, not mistaken. Not quite as daunting as this one, but not a walk in the park, either.”

        -20C is hardly a “daunting” storage requirement. You have that in your own home , assuming it’s not a cardboard box.

        Yes -4C would be somewhat better , and RT storage would be ideal , but -20C is still a cakewalk compared to -80C.

        1. Marko says:

          “… -4C would be somewhat better ”

          I meant +4C , i.e. standard refrig temp.

      2. Dan says:

        -20 C is a home freezer (-4 F). That’s more or less trivial even in the rural US. Developing nations, that might still be a challenge, but not nearly as nasty a challenge as -70 C.

  26. David says:

    Interpreting statistics out of a press release is always a bit tricky, but here goes my attempt:

    From the Pfizer press release, this is the closest I can get to the critical data: “the evaluable case count reached 94 and the DMC performed its first analysis on all cases. The case split between vaccinated individuals and those who received the placebo indicates a vaccine efficacy rate above 90%, at 7 days after the second dose. … The Phase 3 clinical trial … has enrolled 43,538 participants to date, 38,955 of whom have received a second dose ”

    So, 85 infections in placebo subjects and 9 in active-dose subjects (rounding case count to make it “above 90%” as 8+86 would be under 90%), with a bit less than 38,955 subjects eligible for the analysis (a bit less since the endpoint in this trial starts 1 week after the second dose). Assume 1:1 randomization was accurate.

    That gives a relative risk of 0.1 (as expected), with a 95% confidence interval of 0.054 to 0.214. The NNT is approximately 250. The p-value has lots of zeros between the decimal point and the first non-zero digit.

    1. Joseph Severs says:

      @David: your analysis sounds correct to me. However, I expect the confidence level would be greater than 95% because this is an interim analysis. But otherwise your explanation is strong.

    2. Dan says:

      To have 95% confidence that it’s above 90% effective (which is how I interpret what they’ve said so far, though maybe I’m wrong), you’d want the actual ratio to be considerably better than 9:1. An 85-9 split would make the best point estimate around 90%, but the statistical uncertainty around that would mean it could well be more like 80% and you just got lucky. My rough guess is you’d need something like a 90-4 split between placebo and real vaccine arms of the trial to call it at least 90% effective with the degree of confidence you’d want.

      1. Chris Phillips says:

        I did a very rough Bayesian calculation. It would be interesting to see a proper one, but it looks to me as though a 90-4 split wouldn’t be quite good enough to ensure that the lower bound of a 95% confidence interval was above 90%.

        So it may be that the split was 91-3 or better. If so the raw measured efficacy (which would be subject to statistical uncertainty) would be 96.7% or more.

        1. Chris Phillips says:

          On second thoughts, maybe a 90-4 would just about be good enough, if that’s what they did (i.e. 95% confidence interval, with lower limit above 90%).

          But at any rate, provided “above 90” refers to the lower limit of a confidence interval, rather than just the raw apparent efficacy, it looks as though the latter must have been something like 95% or better.

  27. David says:

    Second comment, which I’m keeping separated from the data for obvious reasons:

    Anybody who thinks Pfizer entered into a conspiracy to delay the results and help the Democrats hasn’t been watching the drug price wars. Republican administrations defend free-market pharma pricing, and Democratic administrations support price controls, Medicare negotiation, and drug importing.

    If Pfizer had wanted to tip this, it would have been +R.

    1. Hopeful Layman says:

      I wouldn’t suspect any grand plot to “help the Democrats”. It does occur to me, though, that because of the rampant politicization of this issue we’ve seen, Pfizer might have wanted to avoid the appearance of impropriety, both for business reasons (a vaccine perceived as having been used as an election gimmick would be distrusted by a lot of people, and hence not as profitable, especially since there’s a good chance that other, later vaccines might not be burdened with this stigma) and, in all honesty, for reasons of public health: Uptake is essential, and for that, we need trust.

    2. Marko says:

      Agreed. Pfizer was pushing hard for the pre-election data reveal and EUA. The FDA put the kibosh on it , not Pfizer.

  28. MathematicallySceptical says:

    How can that be that out of tens of thousands trial participants only 96 got COVID?
    I am admittedly to lazy to crunch the numbers myself now but does this align with the estimated infection rate in the participating countries?

    1. confused says:

      Probably vaccine trial participants are more careful than the average, as they are pretty much by definition people who take COVID seriously.

      Those who think it’s “just a cold/flu” greatly exaggerated by the media/government/etc. won’t volunteer for trials.

    2. MrRogers says:

      That’s about right for a trial run in the US. It implies an infection rate of 18/100K/day. Over the last week, the US has averaged ~30.

    3. eigenperson says:

      Enrollment in the trial began on July 27th and appears to have been roughly linear over time [1]. A case of COVID-19 doesn’t count unless it happens at least 28 days after the patient received the first dose. That means the first patients entered the relevant population around August 24th (77 days ago), and now there are 39,000 patients in that population, with the average patient being in the population for half that time (39 days). Also, half the patients received an apparently effective vaccine and were very unlikely to get the disease, so the susceptible population is only 19,500. Of these 19,500 or so patients, about 90 got the disease in an average 39-day period. That is about 460 cases per 100,000 people.

      The current rate of officially reported US cases is 33 per 100,000 people per 7 days [2], which is 184 per 39 days. So the placebo population is actually getting diagnosed with COVID-19 at a higher rate than the overall population — which is not surprising because they are much more likely to get tested.

      In short, the overall picture looks consistent to me.



      1. MathematicallySceptical says:

        Thanks a lot for your detailed answer.

        So is the data mostly from the US? I thought their trial sites included Turkey and some other countries as well?

        What about Astra Zeneca who are running trials in Brazil and South Africa? I would assume infection rates there are higher so surprising they haven’t reached their benchmark of infections yet? Or are their benchmarks different?

        1. confused says:

          Is the infection rate still higher in Brazil? I think that was true during a lot of the summer, but may not be now.

          Although testing may not be comparable, so who knows…

      2. liesdamnlies says:

        28 days between doses and 7 days after last dose = 35 days from enrolment to symptoms wth PCR confirmation counting for the end point.

        what happens to an active arm participant who gets COVID symptoms with PCR confirmed in those 35 days? They do not meet the criteria for the active arm. By definition they are either excluded from the original criteria or in placebo.

        your math is that average 39 days enrollment, and out of that 39 days 35 days don’t count for active arm. Just cases in the last 4 days do, but in placebo arm they do. Plus anyone in active in those 35 days who gets covid is excluded from the count of active arm cases.

        is this true?

        1. eigenperson says:

          There were actually 21 days between the two doses, plus 7 days of waiting after the second dose, for a total of 28 days, not 35.

          The 39-day average that I calculated already excludes the time spent in the 28-day waiting period. For example, a patient who received their first dose on the first day of the trial (July 27th) received their first dose 105 days before the press release. Of these 105 days, 28 were in the immunity-building period, and infections during this period would not count. So I only counted this patient as “eligible” for 77 days.

    4. Dan says:

      That’s an infection rate of roughly 0.2-0.25%. In the span of a month or two, and with half of those in the study being at considerably reduced risk because they have been vaccinated (so the infection rate among those susceptible is closer to 0.4-0.5%), that’s a little lower than the general population (at least in recent days, less so in the earlier part of the trial) but not enormously so, and that might be explained by trial participants being more likely to observe the recommended precautions than the general population (I doubt there are many “COVID is a hoax!” or “I don’t need no stinkin’ mask!” folks who are willing to participate in such a trial, though I’m open to being proven wrong on that).

  29. Craig says:

    I assume this does not provide sterilizing immunity and rather prevents serious disease after infected?
    Isn’t the optimism a bit premature though since community spread can/will still happen since infections will still occur?

    1. confused says:

      >>Isn’t the optimism a bit premature though since community spread can/will still happen since infections will still occur?

      Well, obviously people aren’t vaccinated yet, so ‘premature’ in that sense… But if a vaccine is effective enough, it won’t matter much… Even if infection isn’t stopped by ‘herd immunity’, once most people are vaccinated COVID won’t amount to anything.

      If it’s no longer dangerous, getting it doesn’t really matter.

      1. Craig says:

        >>Even if infection isn’t stopped by ‘herd immunity’, once most people are vaccinated COVID won’t amount to anything.

        I suspect this will take much, much longer though if people are still actively spreading the virus. The number of people getting vaccinated will be competing with the number of people getting infected and spreading it.

        1. confused says:

          Maybe I’m missing something, but I think more infection would mean more hospitalizations/deaths, but not necessarily a longer timeline — I think the timeline to vaccinate everyone who wants it would be based on supply chain/distribution issues and vaccine uptake.

      2. Jon Tyson says:

        No, it makes a huge difference: If the vaccine prevents disease (which was what the 90% number was for) by converting all infections to asymptomatic ones then the first people to get the vaccine may get restless and go infect the unvaccinated ones.

        1. Hopeful Layman says:

          Which is why it’s universally agreed upon that public health practices like masking, “distancing,” etc. will remain essential for some time even after a vaccine is rolled out. Loosening these restrictions and inching our way back to “normal” will be a slow, gradual process. The good news is that if we all do these things, if this vaccine (and maybe more to follow) ends up being as effective as these preliminary data suggest, and if there’s sufficient uptake [the wild card in all this], we WILL get there.

    2. Sc says:

      We don’t really know yet, since the trial wasn’t designed to find out and doing so is much trickier. I suspect it will at the very least decrease the spread, but we will probably only find out as it is rolled out based on whether it stops the spread or not.

      1. Dark Day says:

        . . . and, of course, if people continue masking, “distancing,” etc. as the vaccine is being rolled out (as they should), it will be very difficult to tell how much of that (hoped-for) decline in numbers is due to the vaccine, and how much is due to personal behaviors. The obvious answer, of course, is “both,” but we won[t know for sure until/if any existing “mask mandates” are repealed or modified. If the numbers start spiking again, we’ll know that the train hasn’t left the station, after all.

        1. Michael says:

          Dark Day, the disease will become attenuated over time due to natural infection and vaccination. Now we know the latter is possible. Allow yourself some optimism — the “numbers” (in terms of technical infections) will not be relevant once most of us are not immunologically naive to the virus.

          1. confused says:

            Yeah. Given how devastating things like measles were in immunologically-naive populations (America/Pacific Islands), something that’s maybe 0.65% lethal when it’s totally new is probably going to be no big deal once few people are immunologically-naive.

            That’s a totally non-expert guess, but even the far deadlier H1N1 1918-19 flu turned into a “normal” seasonal flu post-pandemic, didn’t it?

        2. confused says:

          And there may be a third factor, the ‘natural’ end of the pandemic… if mass “general population” vaccination (as opposed to specifically at-risk groups) starts in the spring… well that’s already more than a year since this started, not too far off the timeline of the last five respiratory pandemics.

          IIRC the second wave of the 2009 swine flu pandemic was already dying down by the time the vaccine became *broadly* available in ~November.

          I would expect compliance with masks etc. to drop, but for that to be less disastrous than we’d expect now (also — what % of the US do you think will already have antibodies by March or April?)

      2. DataWatcher says:

        Here’s the New York Times’ take on it, but the way I read this, they’re not really being very specific when they say “very few people who were vaccinated got Covid-19.” It could well be, as Craig suggests, that they didn’t show serious symptoms, but they (or at least some of them) did “get” it in terms of actually being infected. I think we simply don’t know yet.

        1. Michael says:

          My understanding is “Covid-19” is the disease, not the SARS-CoV-2 virus. I’m not sure asymptomatic Covid-19 is a thing. So whether or not some in the vaccinated cohort got technically infected, I do think it’s fair to say very few of them got “Covid-19.”

    3. WST says:

      Could this trial say something about sterilizing immunity?
      Would 90% sterilizing immunity be consistent with these results … and the symptomatic in treatment group be just the unlucky ones?

      1. Hopeful Layman says:

        I think 90% sterilizing immunity would be a grand slam home run (think Kirk Gibson with the bases loaded). Probably too much to hope for, but I’m ready to go for it.

    4. wilhelm Cody says:

      The trial design does not specify gathering the kind of data to show whether the vaccine provides sterilizing immunity or not: “Absence of data is not data of absence” . The trial participants had a nucleic acid test only if they showed one or more COVID-19 symptom. Many infected people have few or no symptoms.

      A different trial design would have been to test each participant every 3 or 4 days or so, to catch anyone in a period where they might be producing virus even in the absence of symptoms. The trial logistics probably would have been too difficult and too expensive to look for such cases.

      However, if the vaccine does not provide sterilizing immunity, the pandemic will take longer to end: lots of folks who can produce virus without getting sick spreading the virus before everyone is immunized.

      Note that there is a vaccine which ostensibly does this: diphtheria vaccine protects against the toxin but not necessarily against the bacterium. I do not know of data showing whether the DOT vaccine reduces the rate of carrying or transmitting the bacterium.

  30. TallDave says:

    not sure we can expect *all* spike-targeting vaccines will see efficacy

    but even a bare majority would be great, great news

    of course the first is much more valuable than the 17th

    unless cheaper and more easily delivered solutions can deliver similar results

  31. Pierre says:

    And if I’m not mistaken, what this also means is that research for dedicated molecules to fight Covid-19 has suddenly switched from “get a head start on a locating a solution we pray we will never need, just in case” to “utterly pointless”.

    1. Wilhelm Cody says:

      If the vaccine is only 90% effective, people will still get sick and need treatment. Even with mass influenza vaccination each year, Tamiflu is sold and used. COVID-19 may not be different.

      1. Hopeful Layman says:

        Well, flu vaccine is only about 60% effective, right? And, at least until this year, uptake was somewhere around that same number. I’m guessing that if this 90%-plus efficacy rate (or something close to it) holds up, the situation here might be quite different. Either way, though, there are some very promising therapeutics being developed, and I think that by mid-2021 at least some of them will probably have been approved and rolled out. Looks as if a very interesting year is coming our way.

      2. 90% effective is plenty and if widely dispersed and if these numbers held up while truly preventing infection and transmission, not just symptoms, would effectively end the pandemic.

        Of course, whether preventing 90% of the infections detected by PCR testing of people showing some COVID-19 symptoms ends up implying that 90% of people exposed will not get infected enough to transmit to others remains to be seen, but this is really promising — the most promising news of the pandemic to date.

  32. Eric nandonside says:

    “And since all of the vaccines are targeting the same Spike protein, it is highly likely that they are all going to work”

    I’ve read much here and other places about the frequent Spike protein mutations. Since the vaccines are targeting that, will these vaccines be similar to a Flu vaccine where we’ll never know if this year’s batch will work?

    1. confused says:

      Isn’t flu mutation really unusually fast and not really comparable to other viruses?

      Can someone who knows more answer that?

      1. Michael says:

        Confused, I would recommend this thread from Deepta Bhattacharya, a U of Arizona immunologist. From a week or so ago, but he anticipated that these vaccines would be more efficacious than flu vaccines and explained why quite clearly:

        1. Some idiot says:

          Thanks! Good thread! 👍

  33. Hopeful Layman says:

    Just to clarify — do we know whether this is 90% protection against serious illness, minor symptoms, or both? I know that’s a question that’s been bandied about a lot in much of the commentary on these trials — anyone know the details on this?

    1. RonW says:

      I believe the symptoms need to be significant enough the trial participants self-report them, at a minimum.

      1. Joseph Severs says:

        I would expect protocol would specify that investigator contact subjects weekly (or so) to prompt for recent symptoms consistent with the disease. So a subject who said in reply “Yeah, I had some of those symptoms” would be urged to come to the site for a followup evaluation.

        1. RonW says:

          There’s an app that pops up weekly that asks “did you have any of these symptoms?” If you hit yes, then human interaction begins.

          You can also open the app whenever you want and hit ‘yes’ if you feel ill.

    2. David says:

      Hopeful Layman: “Just to clarify — do we know whether this is 90% protection against serious illness, minor symptoms, or both?”

      According to (look for NCT04368728) the first-listed efficacy endpoint for the larger part of the study is confirmed infections, which seems to be what Pfizer reported out in the press release. There are no listed endpoints there for serious cases, nor for minor symptoms.

      1. Hopeful Layman says:

        Okay, so it’s infections, not just disease. If that’s the case, then does this mean something closer to sterilizing immunity, after all? That would be remarkable in a first-generation vaccine.

        1. JIM OTTO says:

          It’s actually for symptomatic disease. Subjects completed daily logs and were brought in for testing if they had symptoms. The early look would not catch asymptomatic cases (those will get caught by an antigen test at 6 months).

          Secondary endpoints include disease severity, but it will likely require way more than 94 events to read those out.

        2. confused says:

          A completely non-expert thought, but I wonder if our expectations are actually too low.

          Most medical development these days is hard because anything that doesn’t already have good treatments is an inherently hard problem, and if there’s already good treatments there’s a much higher bar to improve on them.

          But the lack of treatments/vaccines for COVID-19 is because it’s new, not because it’s an inherently hard problem like HIV/AIDS or Alzheimer’s.

  34. Steven says:

    How important is this cold chain? I wonder if the -80C stuff is overstated.

    They tested the materials at -80C, so they know that works. Moderna’s vaccine essentially the same and yet they say they “only” need -20C. I wonder if this isn’t a bit like the expiration dates on medicines: They are still good well past those dates, but those are the safe estimates and no one wants the liability from using a longer more realistic time frame.

    If Pfizer wants to test a vaccine that was stored at only -20C, do they need to re-run the whole study? Or would it be enough to do something more like phase 1 testing to show it still provokes the same immune reaction?

    1. Mammalian scale-up person says:

      Emulsion / “lipid nanoparticle” (AKA liposomes) organization and stability depends on the choice of lipids used; some are more thermodynamically stable than others. I believe Moderna has had quite a patent litigation issue over their particular lipid choices; seems very likely BioNTech / Pfizer are using something completely different for lipid (due to IP reasons if nothing else) which is driving the storage issue.

      1. FrankN says:

        Nah, it is much more simple: You can’t dissolve lipids in saltwater, you only can form an emulsion, but that isn’t stable. In order to ensure proper dosage, it is neccessary to keep the emulsion from segregating. The technically (not logistically) easiest way to do this is by deep freezing. Since Pfizer had the required infrastructure already in place, they went for -80C, in order to ensure that the emulsion always stays below its freezing point – the last thing they wanted was to compromise the trial by potentially inconsistent vaccine dosage.
        In the meantime, tests are under way to establish whether the freezing/ cooling regime may be relaxed.

  35. Marko says:

    Lilly’s neutralizing antibody bamlanivimab (LY-CoV555) receives FDA emergency use authorization for the treatment of recently diagnosed COVID-19

    – Bamlanivimab is authorized for treatment of recently diagnosed, mild to moderate COVID-19 in high-risk patients
    – U.S. government will allocate 300,000 doses of bamlanivimab to high-risk patients, with no out-of-pocket costs for the medication

    If the usage is restricted to patients who are efficiently risk-stratified and caught early enough in the disease course , this might have an impact on death rates going forward , even with the limited supply of doses.

  36. DrivingDownTheDragway says:

    Contrary to the bloviations of a certain head of WHO, I think there ought to be exactly zero complaints about the cold-chain complications of this one. It will (or should) go to those places best able to handle and administer it. Not long after, another vx with simpler requirements will be out, and then another and another.

    The existence of this one merely means more supply of the others to places that can’t reliably maintain an 80 below cold chain. As a side benefit, those places will have a much lower cost of administration.

  37. An Old Chemist says:

    I just listened to Derek Lowe talk about the Pfizer vaccine on NPR’s program ‘MarketWatch.”

    1. Hopeful Layman says:

      Do you have a link for that? I can’t find it . . .

      1. An Old Chemist says:

        I just did a Google search but did not find the link for today’s ‘Marketwatch’ program. Marketwatch is a daily program on NPR that discusses the stock markets and economy, but as today’s actions in the stock markets were driven solely by the Pfizer vaccine’s 90% efficacy, they mentioned this and had Derek talk about the vaccine for a few minutes.

  38. Barry says:

    OK, I’m confused. Is Pfizer scoring it a success if a subject had no symptoms of severe Covid 7days after the 2nd dose? Or only if the subject was negative by PCR 7 days after the second dose?
    If they’re blocking severe disease w/o neutralizing mucosal immunity, they’re not blocking the spread of the contagion.

    1. Michael says:

      I think they have determined that protection, if any, can be assumed seven days after the second shot. So I believe that is when they start the clock for whether the particular participant becomes ill and therefore counts as an event.

      But it isn’t necessarily symptoms of “severe” disease – it is any symptomatic COVID at all confirmed by PCR that comprise the 94 events.

      1. Barry says:

        So are they counting asymptomatic infections as “infections”? Or as successes?

        1. DrSAR says:

          Kind of as success if they are truly asymptomatic. All trial participants (it’s blinded, mind) are self-monitoring for Covid-like symptoms for the entirety of the 24month study. They are also followed up at 1, 6, 12, 24month. From the trial protocol at I get in Section 8.13 Covid surveillance:

          “If a participant experiences any of the following (irrespective of perceived etiology or clinical significance), he or she is instructed to contact the site immediately and, if confirmed, participate in an in-person or telehealth visit as soon as possible, optimally within 3 days of symptom onset (and at the latest 4 days after symptom resolution). … A diagnosis of COVID-19, Fever, New or increased cough, New or increased shortness of breath, Chills, New or increased muscle pain, New loss of taste/smell, Sore throat, Diarrhea, Vomiting.”

          This seems reasonable, I’d say. You will presumably miss asymptomatics in both arms. A priori, the efficacy for those could also be upwards of 90%.

          1. bill l says:

            But the most recent FDA guidance requires 5 or more serious cases before submitting for EUA.

            Of course, Pfizer and the FDA are negotiating so anything can happen.

          2. sort_of_knowledgeable says:

            They may have 5 serious cases among the total cases already or may have to wait longer. And maybe after 6 months, 4 serious cases among the control group with none in the treatment group and enough total cases to say with confidence that the vaccine is 95% effective the FDA might go with an EUA anyway.

    2. TabeaK says:

      No, they do not test everyone at all times. If someone in the study develops COVID symptoms, they get a telehealth visit & PCR test – or, if a study participant gets a COVID test for any other reason, they are heavily encouraged to contact the study director with the result.
      The trial will miss asymptomatics – but then so does testing in the general population. The trial is also not designed to find out if the vaccine prevents positive tests – it may well not – but if it turns the vast majority of problematic infections into asymptomatic ones, we are good.

      Any confirmed COVID results from 7 days after the 2nd inoculation count towards the overall case count in the study.

  39. bill l says:

    Before submitting for EUA, the FDA requires a median follow-up duration of at least 2 months after completion of the full regimen.

    Pfizer says they’ll be ready in the 3rd week of November (the also say next week), which means that data set had to have their 2nd shot back in September.

    But they also appear to be saying that this group of 94 is from late October.

    I guess they could be different groups?

    And these samples are from only 7 days after the 2nd shot? Or does it mean they tested negative 7 days after the shot and that’s when the clock started?

    Although this is good news, there are many reasons to think it is still a ways off.
    As cases rise uncontrollably around the country, it sure would have been nice for Pfizer to announce this with some humility instead of going on tour and shamelessly claiming this a great day for humanity.

  40. MM says:

    Could local colleges and universities with -80 freezers be temporarily volunteered for vaccine storage/delivery? I’m no immunologist, but I work at a college with a massive, largely empty -80 freezer located in a rural area. There would be a number of secondary issues to solve (access, storage QC, cross-contamination, etc.), of course, but desperate times…

  41. David Malenchini says:

    Non-scientist here. Can you explain what it means to have 90% effectiveness?

    1. dEePsTaTe says:

      it means 90% of the time it works every time

    2. Dan says:

      Roughly speaking, it means that vaccination reduces your risk of getting the disease by (at least) 90%. (Whether this means “infection”, “symptomatic case”, or “severe case” may vary depending on who’s doing the trial, but my understanding based on what I’ve read so far is that Pfizer’s primary criterion is “symptomatic case”.)

      1. Hopeful Layman says:

        Please forgive yet another “Layman’s question.” If the vaccine arm of the study is telling us, basically, how many cases DIDN’T happen, how can that be quantified except in general terms? Let’s say the control arm showed 90 cases, distributed among mild, serious, and very serious (none fatal, as far as we know). The vaccine arm, let’s say, showed 9 cases (90% efficacy). Of the hypothetical 81 cases that were prevented by the vaccine, how do we know for sure which were mild, serious, etc.? Do we just assume a similar distribution to what arose in the control group?

        1. We don’t know from what was reported what proportion of cases in the placebo (or vaccine) arm were mild, moderate, or serious, or led to deaths.

          If you want to be pessimistic, it is possible that:

          1. the vaccine is only preventing the mild cases, or worse yet transforming mild symptoms to no symptoms, but that it doesn’t prevent more severe cases caused e.g. by exposure to higher viral loads.
          2. the vaccine prevents symptoms but not infections or transmissions, and the indivduals in vaccine arm who would have presented with symptoms sans vaccine are still infected and could still infect others.
          3. the vaccine exacerbates disease in those for whom it doesn’t prevent the disease, something called “immune mediated enhancement”.

          These possibilities will be checked in follow up data, but to focus on them is to be a pessimist and worry-wort in my opinion. These results of 90% effectiveness are more than I dared hope for and extremely promising not just for this vaccine, which is very difficult to distribute because of the cold temperature requirements, but for all vaccines targeting the spike protein.

    3. A treatment is 90% effective if when you see 100 bad things in an untreated population, you see only 10 of those things in an otherwise-identical treated population.

      What Pfizer is reporting is that their vaccinated population did much better than their unvaccinated population. They could have got that result if their vaccine were 0% effective but they were very very very lucky (p ~ 10^-29). Or, if their vaccine were only 80% effective and they were very very lucky, and so on.

      I’m not sure what lower bound they’re using, but when they say that the vaccine was probably at least 90% effective they are probably saying that (a) it appeared to be even more effective than that, but (b) it might be “only” 90% effective, but then they were fairly lucky (p ~0.05) so the results looked as good as they did.

  42. Michael says:

    Considering the efficacy of the Pfizer vaccine, I am *really* excited to see the results of Novavax’s Phase III. They have shown much higher titers than Pfizer and have shown sterilizing immunity in two animal studies.

    How high can the Novavax efficacy get?

    1. Craig says:

      same. I believe they were the only ones to mention sterilizing immunity seen in their primate studies and I believe their vaccine produced many times more the # of antibodies than any of the other vaccine candidates.
      Even in the best case scenario that their vaccine is a game changer, I’d be concerned with their production capacity. I have zero knowledge of this but I’d imagine they don’t have nearly the same resources and production capacity as some of the bigger guys like Pfizer, J&J, etc.

      1. Hopeful Layman says:

        . . . although since they’re Warp Speed participants, might that not avail them of resources (financial, at the very least) that could help?

      2. Wilhelm Cody` says:

        Part of Warp Speed money is going to increase general manufacturing capacity for any of the selected vaccines that might get FDA approval. That means paying up front for 100 million doses as well as paying other companies to increase capacity for making vaccines. If something is in the Warp Speed program and gets approved, 100’s of millions of doses can be made quickly.

    2. Hopeful Layman says:

      I agree. Their early-phase results were very encouraging — I believe their vaccine was the only one to show promise for actual sterilizing immunity — and the presence of the adjuvant seemed to suggest the potential that it might be effective in older patients, as well. I’ve had high hopes for them since the beginning. Given, also, the similarities between Moderna and Pfizer, we might actually see at least two or three excellent vaccines by early 2021.

  43. Dark Day says:

    I think this qualifies as more good news. Biden’s Coronavirus Task Force will be co-chaired by Dr. Marcella Nunez-Smith. She has been a powerful advocate for communities of color throughout the pandemic, and she is also spot-on in terms of advocating and educating for vaccine uptake — along with equitable distribution and access (closely related, actually, to the uptake issue) — for the people of those communities. She has a true “boots-on-the-ground” approach to public health and public health advocacy, and it was a very prescient move to include her among the leaders of the task force. Bodes well for the future.

  44. WST says:

    Pfizer announcement:
    “The Phase 3 clinical trial of BNT162b2 began on July 27 and has enrolled 43,538 participants to date, 38,955 of whom have received a second dose of the vaccine candidate as of November 8, 2020. ”

    Does that mean that 43,538 – 38,956 = 4,581 are in placebo group?

    The Pfizer design calls for 1:1 ratio.
    What did I miss ?

    1. Michael says:

      I think that was a very poorly worded statement. I think they meant to say “second dose of the vaccine candidate or placebo.” The 4,581 was meant to convey those who recently got the Day 0 shot and have not yet received the Day 21 shot.

  45. RonW says:

    Yes, they were recruiting volunteers until the end of October (and maybe beyond for younger groups), so those won’t have their 2nd shot and 7 days following for a few weeks yet.

    1. Hopeful Layman says:

      Personally, I wouldn’t mind waiting until then, both for the sake of more robust data and to further ensure to a skeptical public that this process has been, and continues to be, legitimate. Would hurrying an EUA by a few weeks be too important to let that happen?

  46. Hopeful Layman says:

    Interview with BioNTech CEO Ugur Sahin that addresses (but doesn’t really answer) many of the questions asked here. He basically says that the issues of severity vs. mild cases and efficacy among the elderly are still being studied, and answers should come forth within a few weeks. (The interviewer didn’t raise the question of “infections” vs. “disease” — that’s a subtlety that seems to escape most mainstream reporting.) I’m guessing that means, among other things, that they’re waiting for data on the entire trial population comes through.

  47. LX says:

    Does anyone know whether the readout was based on:
    1. Only cases recorded between 0-7 days after the second dose
    2. Any cases recorded from 7 days on after the second dose

    The language in the press release seemed a bit ambiguous in this regard:
    “The case split between vaccinated individuals and those who received the placebo indicates a vaccine efficacy rate above 90%, *at 7 days* after the second dose. This means that protection is achieved 28 days after the initiation of the vaccination, which consists of a 2-dose schedule.”

    1. I think it is poorly worded — based on protocol outcome definition I think it means that any cases that occur before 7 days after 2nd dose, 28 days after first dose, do not count. So any cases thereafter are counted.

      And 90% means that based on the proportion of detected cases that are in placebo and vaccine arms, the vaccine prevented 90% of the cases that would have occurred sans vaccine.

  48. Charles E says:

    Question I’ve always had about vaccine efficacy: so say a vaccine is listed as 70% effective. Would that mean that if I’m vaccinated, EACH encounter I had with someone with the virus, I’d have a 30% chance of getting it from each encounter? Or does it mean if I get vaccinated, there’s a 30% the vaccine does not work at all for me and a 70% chance that I’m fully “protected”?

    1. Moses says:

      @Charles E: The latter

      1. Michael says:

        I would be careful of thinking of this as a binary issue of protection vs. “does not work at all for me.” Immunity and protection will be on a spectrum and we are all different.

        If, 7+ days after the second dose of the vaccine, you develop any symptoms at all, that’s considered a fail under this protocol. But the vaccine is very likely to prompt an antibody and T-cell response that could, for example, reduce disease severity and/or decrease amount and duration of shedding. In particular, there’s reason to hope that lungs could be significantly protected even if the vaccine doesn’t prevent all symptoms of an upper respiratory illness.

    2. It is not really either of those.

      By study definition, it means that 90% of the cases that would have happened sans vaccine were prevented by the vaccine.

      In this study, case means a PCR-confirmed SARS-CoV-2 at least 7 days after 2nd dose of vaccine/placebo received, in which PCR tests are only given if the subject indicates that they experienced a COVID-19 symptom.

      It is computed by efficacy = 100 x (1-X1/X0) where X1=number of cases in vaccine group and X0=number of cases in placebo group, at a point in the study in which a combined N=X1+X0 cases have occurred (94 for this interim analysis). The study results suggest 90% efficacy would correspond to X1=8 and X0=86.

      If the vaccine was not truly effective, then we would expect 86 cases in the vaccine arm by now, since we saw 8, a reduction of (86-8)/86=90.7% from what was expected, we presume that these were “prevented” by the vaccine.

      Whether it truly confers immunity from infection and infectiousness remains to be seen (it could just make someone asymptomatic but still contagious) and it is possible that it only prevents a certain subset of cases (e.g. mild ones) but that people exposed to high enough viral levels to get severe disease may not be prevented by the vaccine. But all of these and more detailed followed up questions will be addressed by secondary analyses.

      1. Craig says:

        >>>Whether it truly confers immunity from infection and infectiousness remains to be seen (it could just make someone asymptomatic but still contagious) and it is possible that it only prevents a certain subset of cases (e.g. mild ones) but that people exposed to high enough viral levels to get severe disease may not be prevented by the vaccine.

        If that’s the case, isn’t all the optimism incredibly premature? If it only prevents a subset of cases (perhaps mild ones), then that’s essentially of little/no value, right?

        1. Michael says:

          From Eleanor Riley, immunologist:

          “If you were going to have a trial big enough to pick up a statistically significant difference in deaths, that trial would have to be absolutely enormous and it would have to run for a very long time,” she says. “If it stops people getting sick, it probably stops them getting severely ill.”

          I haven’t been able to find the quote just now, but at the congressional vaccine hearings a few weeks ago, someone testified to the effect that never in the history of vaccinology has a vaccine prevented mild disease with no effect on severe disease. But I am out on a limb with my memory on that point. If you scroll through Deepta Bhattacharya’s Twitter (or many other immunologists), the expectation is that the vaccine would make what could have been severe disease more mild.

        2. Hopeful Layman says:

          Michael, I remember that same quote. I wasn’t going to post it here, because, like you, I can’t remember precisely where/how I read it. But I’m guessing that others here can weigh in more authoritatively. My own personal biggest misgiving is the question of whether this vaccine can prevent spread (i.e., infection) as well as measurable symptoms (i.e., disease). If not, it’s still a major step in the right direction, and if that 90-plus percent figure holds AND we get sufficient uptake, this could still play a major role in bringing the pandemic under control, esp. considering that at least one or more of the other vaccines still in the pipeline (e.g., Novavax?) might be able to do an even more robust job of conferring something closer to sterilizing immunity.

        3. There is no evidence that it only prevents the mild cases, I was just stating the caveats that are not possible to know for sure based on the single primary analysis.

          To think that just because it is possible that the cases prevented were only mild, or that it somehow didn’t prevent infections just symptoms, does not mean this is likely.

          The strength of the results, far above the 60% efficacy that was targeted by the design, is indeed incredibly positive news and warrants excitement.

          While these results may be better than most dared to expect, there should not be unbridled optimism that somehow this signals the end of the pandemic — that is extrapolating way out from these preliminary results.

          But if these results are as they appear this is nothing but outstanding news — not just for this particular vaccine but for the increased promise that other similar vaccines may work.

  49. JS says:

    A few questions:

    Are they keeping track of cases in household members? If yes, might we get some idea of the reduction in the ability to pass on the virus?
    Are they collecting case data between the first dose and 7 days after the second? In other words will we know (at too low significance?) if there is good protection after one dose?
    With such a good protection, might we guess that a smaller dose would be better from a public health point of view? For example half as much to twice as many people? As I recall it they did try lower doses in the earlier phases.
    Why did they exclude the age group 56-64?

  50. Bill says:

    I see Sputnik V has now achieved 92% success rate. Good for them. But it was based on 20 people actually contracting Covid? What calculation dealing with 20 cases could yield an answer of .92?

    1. Craig says:

      I want to, in good faith, believe this but all the alarm bells go off. They release it *right after* Pfizer and the number is *slightly* better than Pfizer….seems a bit odd. But for the sake of humanity, I am rooting for it to be true. I’ve read that a lot of countries without immediate access to the Pfizers, Modernas, etc are trying to sign deals for Sputnik vaccine so that’d be great if it was safe and effective so it could fill that need.

    2. sort_of_knowledgeable says:

      To calculate the significance of 20 cases out of a 16,000 person trail I went to binomial calculator. 20 out of 8000 is 0.0025. Walking back the probabilities of fewer cases than 20 using the same rate of .0025, it starts to drop off rapidly pass 15 and by 5 cases in the trail case the probability is 0.00007088373. So it is possible to get a statistically significant result if the two trial groups are comparable.

  51. Gary Cornell says:

    If we had a government that looked to production efforts in the earliest days of WW2 as their model and started *now*, it should be possible and relatively inexpensive to build the infrastructure to easily distribute a -80c vaccine. See the back of envelope calculation I made here back in August:

    1. Hopeful Layman says:

      Let’s hope Biden’s coronavirus task force takes this on. It’s a good group with a diverse array or experience, perspectives, expertise, etc., and they’ve already begun meeting. Let’s hope also that they meet with Gen. Gus Perna, who is overseeing logistics for Operation Warp Speed. With the right focus and opportunity to work undistracted (a tall order in this incredibly, toxic and chaotic political moment), they might well be able to get a viable project of this nature under way within a relatively short time.

      1. Wilhelm Cody says:

        Warp Speed program has already taken it on and has distributed money for manufacturing capacity for many of the possible vaccines. They also have spend money on general increase in capacity, including vials and distribution. The leaders of the program are smart enough to understand what is needed.

  52. An Old Chemist says:

    Pfizer’s Positive Vaccine Data is a Relief for Many, but Logistical Hurdles Remain (Biospace):

  53. An Old Chemist says:

    Moderna’s Personalized Cancer Vaccine Shows Promise in Early-Stage Trial (Biospace):

  54. An Old Chemist says:

    Russia Claims its Sputnik V Vaccine is 92% Effective Following Interim Analysis (Biospace):

    1. Derek Lowe says:

      But they somehow hit “92%” with only 20 cases in the interim analysis? That makes very little statistical sense.

      1. Chris Phillips says:

        The 92% claim puzzled me for a long time, but I think it could be compatible (as the lower end of a confidence interval) with a 20-0 split between the two arms, PROVIDED that two thirds (rather than half) the participants received the vaccine and only one third the placebo (as in the US AstraZeneca protocol).

        Is it known what percentage received the placebo in the Russian trial?

        1. Chris Phillips says:

          Actually Reuters says about a quarter of the participants received the placebo.

          If that’s right, perhaps 92% efficacy just means that 4 of the cases were in the vaccine arm and 16 were in the placebo arm, because the vaccine arm was three times as big, so 3×16 = 48 cases would be expected on the basis of the placebo arm, and 4/48 = 8.3%.

          Or it might be based on a confidence interval if there were fewer than 4 cases in the vaccine arm.

  55. Sharma says:

    I am missing the cause for optimism here but I know I’m in the minority because much smarter people than myself seem optimistic with the Pfizer news. If we still don’t know if 1) severe cases are prevented, 2) transmission is reduced and 3) sterilizing immunity is attained (not likely but not necessary), then we essentially don’t know anything, right? Let’s say that only mild cases are prevented …then what is the value of that? Mild cases would have been mild cases with a vaccine or not. I am sure I am missing something though cause as I said, much smarter people are optimistic about this 🙂

    1. True — all these questions remain.

      But a prevention of 90% of the cases in this time period is better than most dared to hope for and clearly shows the vaccine works in some sense.

      These other questions will be addressed in future analyses — some of them from this study and others will not be knowable until a vaccine is more broadly disseminated via phase 4 studies

      1. Sharma says:

        thanks for the reply. Assuming the answers to these are not what we want to hear (severe cases still occur, transmission still happens and sterilizing immunity is not obtained), would there still be a value to the vaccine then?

        1. It is extremely unlikely that those things are true, so I think fixating on them is pessimistic worry-worsting.

          The high efficacy in preventing cases, with 90% much higher than the study’s targeted 60% rate, suggests it is working to neutralize the virus, and it is unlikely that it is only working on mild cases or only shifting infectious symptomatic cases to infectious asymptomatic cases. But to be cautious, these things are theoretically possible so more study and characterization of results is needed to rule them out.

          Fortunately, many of these questions will be addressed to some extent in secondary analyses that should be done in the not-too-distant future.

          1. Sharma says:

            thanks. Do you suspect sterilizing immunity is a possibility with this based on the efficacy or is that still unlikely?

          2. Hopeful Layman says:

            If I’m not mistaken, the original wording was that among people who received the vaccine, there were more than 90 percent fewer “symptomatic case” than in the control arm. That doesn’t quite sound like sterilizing immunity, but as worded it doesn’t necessarily rule it out, either. I will say, though, that aside from Novavax, I don’t remember any of the vaccine candidates being reported as eliminating both upper and lower respiratory infections in earlier animal trials.

  56. Dark Day says:

    “. . . he expectation is that the vaccine would make what could have been severe disease more mild.”

    Good news in terms of severity and death rates, but is this good news in terms of slowing / stopping the spread? Since not everyone will be vaccinated, a vaccinated person who is “lucky” enough to have “only” a mild case can still spread it to unlucky people who may get severe cases and still die (a scenario not unlike those college students coming home and infecting their parents and grandparents).

    1. Bill says:

      Which is why the at-risk group is a higher priority for vaccination availability than the others.

      If offered and refused…well, that’s another issue. But overall it still has to be a net gain over status quo doesn’t it?

      1. Dark Day says:

        Yes, as long as the vaccine is sufficiently protective in higher-risk populations (esp. the elderly), which we don’t know yet — although intuitively, I guess, one might suggest that even if efficacy levels dropped significantly — from 90% to 75%, let’s say — for these other groups, it would still represent a significant step in the right direction. Also — is there any evidence that people with less severe cases (and, for that matter, the dreaded “asymptomatic carriers”) are carrying reduced viral loads, and so may be less likely to be severely contagious? I read recently that up to 70% of all infected people actually shed very little of the virus – they could almost be considered “non-contagious carriers.” So if it’s that remaining 30% who are spreading most of the cases, then perhaps our efficacy data in terms of “slowing the spread” need to be reconsidered (?)

  57. Benji says:

    Is there any realistic possibility of having a mandated vaccination in America? I am not saying I am in favor of it or against it — just a question. I have a suspicion that even if a vaccine efficacy is 90+%, we won’t have anywhere near the numbers of people we need to take it to achieve a herd immunity.

    1. Dark Day says:

      Little if any. My own personal view is that some private venues where people gather — restaurants, bars, probably performance venues like clubs, theaters, casino lounges, etc. — might well decide to initiate “Proof Of Vaccine Required For Entry” regulations (not unlike having to show an ID to purchase liquor, which in some states is already required for ALL customers, not just those who “look” as if they might be under 21), but I can’t envision any kind of governmental mandate. I’m guessing that it won’t even be required for schoolchildren after a vaccine for them has been approved. American’s don’t like the “gummint” telling them what to do.

      1. Dark Day says:

        By the way, just as the recent spike in cases has been “good” news for vaccine testing, it might also be “good” news for vaccine uptake. If more and more states, counties, cities, etc. find themselves forced to re-impose at least partial shut-down measures, and if hospitals (and morgues) continue to be overloaded, and if people keep waking up every morning and get get slugged in the gut with ever-more depressing reports of increased cases, hospitalizations, and deaths . . . they just might decide that this thing is for real after all, and whatever risk or discomforts they may associate with getting vaccinated are worth it to be able to get their lives, their jobs, and their mental/emotional balance back. One can hope, at least.

        1. Benji says:

          can the uptick in infections also be viewed as if someone got a vaccine? They might not have willingly got a vaccine but now they got infected & have antibodies. When I hear of the uptick, of course it’s sad but part of me says “well, maybe some of those who are antivaxxers are getting infected so this will unintentionally give them antibodies”

          1. Dark Day says:

            Well yes, there’s that, too — but it’s a little too close to the Atlas-style “herd immunity” approach to my liking. That being said, I agree with Michael (if I’m reading him correctly, and if not I apologize) that potential efficacy might actually be getting under-estimated, because it does not take into consideration the percent of people who already have at least some immunity from previous exposure.

          2. confused says:

            Of course immunity by infection isn’t really a good thing – but I do think we need to acknowledge that it does seem to exist/work in most cases (reinfection exists, but seems rare).

            Of course we’d prefer people to become immune via vaccination, preventing all the sickness & deaths. But I don’t think it’s out of line to consider how the baseline level of natural immunity that will be around by the time mass vaccination starts will affect the course of the pandemic.

            That doesn’t mean *advocating* it!

          3. Dark Day says:

            You’re right — this entire scenario has been fraught with tragic ironies (e.g., it was “good” for expeditious vaccine testing that there were so many cases spiking in so many places). This would just be another one.

          4. confused says:

            Ironically, in September (when things were improving in most of the US) I was worried that vaccine development might take forever if they were waiting for a certain number of cases, and thinking they maybe should be doing challenge trials… Looks like they knew what they were doing.

            I really would like to know what % infected the US is at now… CDC suggests 0.65% IFR, so 240,000 deaths would imply something like 37 million infections, which is “only” ~11% of US population – but there is a time lag from infection to death, and from death to reporting of death.

            But I’m not sure how well-constrained that IFR estimate is, as it seems to have significantly changed over the course of the pandemic.

    2. Michael says:

      Herd immunity (or at least herd protection in that the disease becomes typically asymptomatic or mild) will eventually occur through a combination of vaccination and infection. The aim is to maximize the former and minimize the latter in getting there.

      As much as we all don’t want herd immunity to happen through infection alone, I do find it puzzling to see people multiply the vaccine efficacy by the number of vaccinated as the only relevant variables in getting to that threshold, while ignoring post-infection immunity. And yes, as with vaccines, that immunity level is on a spectrum too.

      1. Craig says:

        this is assuming protection (whether natural or vaccine) lasts a decent amount of time in order to obtain a certain percentage though right? If protection lasts 6 months, I’d imagine it’d be near impossible to get 60% + of the population to be vaccinated within that window of time.

        1. Dark Day says:

          Yup — THAT, I think, is the wild card. Indications are that, apparently rare cases of reinfection notwithstanding, COVID antibodies last a pretty long time, and of course antibodies aren’t the only immunity tools in the body’s arsenal anyway. Pfizer/BioNTech seem confident that it “looks” as if their vaccine should provide protection for at least a year, but I honestly don’t know on what evidence or data they’re basing that (salesmanship, perhaps??) —

          1. Craig says:

            thanks. Wouldn’t they have some data on longevity already if the trial started in July? I’d imagine they’d be able to see X amount of people are either infected or negative after their 2nd shot by this time. Maybe not to say “it last THIS amount of time” but at least enough to say “at this point, we see it lasts at least X amount of months and we may extend that as the time goes on”

        2. Michael says:

          There are many, many reasons to assume immune memory doesn’t universally evaporate after six months, including the studies cited by Derek in his T-cell blog post today and the fact that we have not seen hundreds of thousands of severe reinfections 10-11 months in.

          1. Craig says:

            this is true. I know Florian Krammer, a professor of vaccinology at Mount Sinai, released a study recently saying that 90% of their participants had robust antibodies for *at least* 5 months …and “at least” was because it was still ongoing so it could be much longer. ( )

      2. Riah says:

        surely the fact that only 94 out of c 44,000 (0.4%) got any symptoms indicates there is already a lot of herd immunity around? Otherwise, surely, far more would have been infected. I cant believe all these people locked themselves away any more than others.

        the new lateral flow tests (designed to exclude viral fragments picked up by PCR) in the UK give far lower prevalence rates than PCR. Around 44,000 tests in Liverpool yesterday turned up only 220 positives. That’s only 0.5%

        1. Hopeful Layman says:

          I don’t understand such low case-positivity rates in the face of these lockdown-level resurgences, hospital overloads, and death rates, How can both of these things be true at the same time?

          1. confused says:

            That is odd. Maybe random testing vs people with symptoms? Positivity rates are very high in some states – but these aren’t random samples of the population, they are largely people who think they have COVID or were exposed.

            If 0.5% of the population currently has COVID, that’s actually quite a lot!

            (Though some states are probably well above that – South Dakota reports “active cases” over 1.5% of its population, and who knows how many haven’t been tested…)

  58. Meeklin says:

    Any thoughts on the longevity of this? I saw Dr. Fauci stated today that people should expect this only lasts 1 year or 2 years, at max, but then others speculate immunity would be much longer based on the fact that the original SARS antibodies are present 17 years later in people.

    1. Dark Day says:

      Those are all estimates — right now, I don’t think we have data for a truly informed prediction. My own intuitive reaction is that a year is what we should be hoping for; any less will result in chaos and doubt, and may well result in people refusing to get any more boosters, leading to new “surges” a little further on down the line. Any more, and we’ll be lucky.

      1. Meeklin says:

        yikes. Even at a year, I could see that being chaotic. I can imagine many, many people not keeping on schedule and this just continuing on and on. I would think 3 years + would be needed to properly address this.

        1. Dark Day says:

          Well, at least some people manage to get an annual flu shot, so it wouldn’t be unthinkable. I don’t know, in this case, whether a single booster or another two-shot regiment would be required, though.

          1. Craig says:

            well I think we can get some idea of this based on how things are playing out. Figure there are over 50+ million cases of covid since February (and perhaps earlier), there are only a handful of confirmed reinfections. And this is with natural infection so you have to think that vaccine immunity will be a bit stronger but even if not, we’re looking at February – now ( 9 months ) with only a few reinfections…so far, so good?

          2. Dark Day says:

            Yup — which is why I think it’s probably reasonable to hope for at least a year. But the fact is, we won’t know until we know. I hope we “find out” as painlessly as possible; the last thing we need is to suddenly discover, a year or two from January, that immunities are wearing off and new “surges” are occurring, necessitating a re-start of the entire process.

          3. Craig says:

            right. I guess we have will have a reasonable idea from all the trial participants which should give the general public advanced notice.

          4. Dark Day says:

            The participants are all going to keep getting regularly assessed for antibodies, unexpected side effects, etc. for at least another year or two, right?

          5. confused says:

            What would “wearing off” look like? If some partial protection remained (and I don’t see why it would be a sudden, binary switch) it might not be terribly dramatic.

            And is COVID likely to still be “pandemic” in a year or two? The last five respiratory pandemics (1889, 1918, 1957, 1968, 2009) have lasted 2 years at most, and we’re already roughly a year from the first infections.

          6. Craig says:

            @darkdays yes, Pfizer said they will be following up for 2 years

          7. Craig says:

            @confused sure but isn’t the R-rate of this virus significantly higher than any of those previous respiratory pandemics? The 1918 flu was less contagious than this, deadlier than this, and easier to contain spread because was almost always symptomatic (as far as I know).
            SARS-COV-2 has a higher R-rate and can spread asymptomatically which is why I’d be skeptical that this follows along the lines of the other respiratory pandemics. I might be entirely wrong though 🙂

          8. Hopeful Layman says:

            Craig, I think you’re right, esp. about the “asymptomatic carriers” part, but I’ve also become a little confused about the R-rate. I think I’ve inquired about this before, but — I have read that at least according to some studies, up to 70% of all people infected with COVID shed very little of the virus, hence are at low risk for transmitting it; they could almost be classified as “non-contagious carriers.” If this is true, and if roughly 30% of all carriers are responsible for the vast majority of all infection spread, what does this do to the overall R-rate? (And, for that matter, what does it mean when we’re trying to estimate vaccine efficacy, presumably as measured across the entire population)?

          9. confused says:

            But those past pandemics weren’t ended by containment, were they? Social distancing wasn’t done in 2009, and I don’t think it was in 1957 or 1968 either… So I’m not sure the R is relevant to this question.

          10. Craig says:

            @confused The 1918 pandemic certainly used social distancing (although not called that) and it was followed much more stringently than it is now. I can’t speak about pandemics other than that but major cities in 1918 followed this well and it helped flatten the curve. There are a few good articles on this, especially one in Nat-Geo but I believe that is a pay article. Here is a free article though:

          11. Dark Day says:

            The response to the 1918-1919 pandemic did involve variants on what we now call “social distancing” and masking — but they certainly weren’t universal. People were still crowding into nightclubs, concert halls, and other venues to hear live music; professional sports (most notably baseball — this was the height of the fabled “dead ball era”) was filling stadiums, as was college football; movies were on the ascendency, and theater — both Vaudeville and “legitimate” — was in full bloom. And people were congregating outdoors: Look at the photos of New York, Chicago, Boston, etc., and you’ll see packed sidewalks, open-air markets, and all kinds of what we’d now call “superspreader” events. So the analogy between that and what we’re going through now doesn’t seem to be entirely precise.

          12. confused says:

            @Craig: yes, there were social distancing measures (and masks) in 1918 (though inconsistently implemented & with significant opposition – much like today).

            But not in 1957, 1968, and 2009.

            Now, those were flu viruses, not coronaviruses. But if the means of spread are similar…

            Those flu viruses didn’t suddenly become pandemic again once immunity waned, even though (IIRC) less than half the US population gets flu shots.

            For COVID, I think that we will (unfortunately) be at a fairly significant level of “natural immunity” (=antibodies/T-cell response from infection) by the time a vaccine becomes available outside the most high-risk groups. Probably more so than for the 2009 flu (which infected ~20% of the US population).

          13. Dark Day says:

            . . .although that will also no doubt mean that even with a vaccine, it will take a lot longer to slow the spread because we will have, and will continue to have, so many active cases, still spreading, even during the early and middle periods of the rolling-out process.

          14. Dark Day says:

            @Confused, I think a big difference in this case is the “asymptomatic carrier” problem. It’s simply many times harder to identify cases — for treatment, for quarantine — when upwards of 40% of them are basically invisible. I’m not aware that that was an issue in any of the previous pandemics.

          15. confused says:

            But I don’t think anyone really tried to contain it in 2009… I think pandemics do end naturally.

          16. theasdgamer says:

            Fun Flu Facts

            33% of people infected with flu are asymptomatic and can spread flu to some degree, just like covid.

            Between 30-50% of flu tests are false negatives after symptoms have begun. (Between 50-70% of flu cases return correct positive results.)

            Flu can be spread by presymptomatic cases, just like covid. Obviously, the longer the duration of the infection, the greater likelihood of spread, just like covid.

          17. theasdgamer says:


            “I think pandemics do end naturally.”

            Hell no. The Black Plague never ended. It killed a third of Europe last year.

            Remember. Science has never proved that pandemics ever ended naturally.

          18. confused says:

            The thing with the Plague though is that, though it obviously did end, there were repeated outbreaks for centuries. But I don’t think that is comparable to this since it had a very different mode of transmission, with the rats/fleas.

            The more recent human-to-human respiratory pandemics only seem to last 1-2 years.

            I think some of the confusion is due to people inaccurately talking about “herd immunity” which (in the sense of vaccination) can imply total disappearance of the disease. The pandemic flu viruses didn’t disappear… but they became normal seasonal flu viruses, no longer pandemic.

            And IIRC there is some speculation that the 1889 “Russian flu” might actually have been the jump-over into the human population of one of the current “common-cold” coronaviruses?

            Even plague didn’t *disappear* – the bacterium is still in existence, and there are still occasional human cases – but it is definitely no longer *pandemic*.

          19. theasdgamer says:


            I’m sure that you recognized my sarcasm.

            I think that for colds/flu we should be talking about herd resistance, not sterile herd immunity.

  59. Craig says:

    interesting. I’m sure at even 30% though, it’s ‘somewhat’ of a problem still. If we can’t openly see who is infected, then it’s hard to contain them. I think the best thing we can hope for is that immunity is long lasting (many years). If so, then over the span of years, the entire world will be able to be vaccinated and even if you we account for the people who refuse vaccination, we can mostly get rid of it if the efficacy is truly in the 90% range.

    1. confused says:

      Hopefully once people are vaccinated, there won’t really be any need to contain anyone… I mean people should still stay home when sick, but they should do that for flu etc. too.

      Post-vaccine COVID won’t necessarily be more significant than any other standard, seasonal upper-respiratory viral infection.

      1. Craig says:

        I was originally leaning towards no but if we’re truly in the 90% efficacy range then yea, I think you’re right. It’d be great if we can make this become like measles — essentially something that the average person doesn’t go around worrying about.
        There are 3 keys to achieving that (correct me if I’m wrong).
        1) The efficacy, which if truly in the 90% range, I guess would be considered a ‘super vaccine’ in the same range as measles.
        2) The R-rate, which I know measles is notoriously worse than Covid so if we can generally control Measles with it’s incredibly high R-rate, then that bodes well for Covid’s much lower R-rate
        3) Length of immunity — this, I guess, is the last remaining key which we don’t know. If we had a year long immunity, then I think this kills the whole notion of making this an afterthought for most people but if we had 17+ years (which SARS patients apparently have), then that’d be incredible.

        1. confused says:

          It also depends on what “duration of immunity” means — if after a year you’re more likely to be *infected* but serious consequences are still way less likely, I think that would be a win too. As long as it ends up being, post-vaccine, no worse than flu I don’t think it will have any lasting social impacts.

          1. Hopeful Layman says:

            Depends on whether the effectiveness would continue to wane over time until it approached “0”. Isn’t that the case, for instance, with tetanus, which requires a booster every five years or so? if the COVID vaccine remained viable for even five years, I think we’d consider that a victory. Most of the estimates I’ve heard, though, seem to see a year as the most likely optimal timeline.

            Meanwhile, another “Layman’s” query. What would be the implications for any of the three two following scenarios:

            (1) A person contracts COVID, recovers, and then gets vaccinated. Would this even be necessary?
            (2) A person contracts COVID (or at least gets exposed), and gets vaccinated during the early stages of the disease.
            (3) A person gets vaccinated, contracts COVID (or gets exposed), and then gets the booster shot a month later.

            With the disease running rampant as it is right now, it’s likely that at least one of these scenarios will occur, probably multiple times. How might this effect vaccination success?

  60. An Old Chemist says:

    5 important questions about Pfizer’s COVID-19 vaccine (FiercePharma)

  61. El-Hassan Shabana says:

    My impression after reading all this is it is early to conclude any practical information.

  62. theasdgamer says:

    Are there any docs here who see flu or covid outpatients who have mild symptoms who can distinguish between the two based on clinical presentation?

    1. groncowsky says:

      I assume ASD stands for aut*sm spectrum disorder?

      1. theasdgamer says:

        The answer to my question is, “No.” I’ve been asking around.

        You have established that you are rude, so congratulations, you get a cookie.

  63. Chris Phillips says:

    I had missed this news on Wednesday.

    “Moderna, Inc. … today announced that it has completed case accrual for the first interim analysis of the Phase 3 COVE study of mRNA-1273, its COVID-19 vaccine candidate.
    “Moderna has seen a significant increase in the rate of case identification across sites in the last week. As a result, the Company expects the first interim analysis will include substantially more than 53 cases, the targeted trigger point for the analysis. The data on these cases is being prepared for submission to the independent Data Safety Monitoring Board (DSMB) for analysis and recommendation. Moderna remains blinded to whether these participants received vaccine or placebo.”

    According to the protocol, success in this interim analysis would require a 74% efficacy rate.

  64. Senad says:

    This just revealed some other hugely efficient prevention and it is – Placebo!
    Only 0.395% people among the placebo controls was infected and without vaccine side effects risk!

    1. Senad says:

      To further emphasize, the risk as per this trial is that, if you take placebo you have 99.6% chance of NOT getting the COVID-19 and 99.96% if you get vaccinated (with all possible side effects).
      Have we found a new cure – Placebo?

  65. starrisktaker says:

    Without looking at the raw data and longterm side effects i would not believe anything. Do not forget 85-90% do not show symptoms or express mild symptoms.

    1. Tony Campanella says:

      It’s more than just expressed symptoms. Participants (I am one of them) give blood periodically. Blood is tested to make sure asymptotic people are still counted.

  66. Chris Phillips says:

    Pfizer and BioNTech have now released the primary endpoint results.

    It seems that, as anticipated, the “>90%” in the interim results came from a confidence interval. They now have 170 cases, with only 8 in the vaccine arm, therefore an observed efficacy of 95%. In those over 65 the observed efficacy was almost identical – 94% – and efficacy was “consistent across age, gender, race and ethnicity demographics”. There were 10 severe cases, including 1 in the vaccine arm. The safety criterion required by the FDA has been achieved:

    1. Marko says:

      “….Pfizer intends to offer vaccine to people who were randomly assigned to get a placebo injection in their Phase 3 trial; Bourla said the company has a “moral obligation” to the people who stepped forward to volunteer for the trial.

      The FDA has been discouraging vaccine manufacturers from doing that preemptively, arguing that the trials should stay blinded — neither the participants nor the scientists who monitor them know who got vaccine and who got placebo — for as long as is practical to generate as much data out of the trials as possible. Bourla said Pfizer is in discussions with the FDA over when the vaccination of placebo arm participants can take place. ”

      I hope the FDA holds the line here. By necessity , the initial rollout will limited in scope , addressing only select , high-risk groups. There’s no “moral obligation” to offer the vaccine toa placebo participant that doesn’t meet those criteria. Continue the trial with placebo participation as long as possible.

  67. Anthony Campanella says:

    Example: If 50% of positive cases happened in the placebo group and 50% in the vaccine group, that doesn’t make the vaccine 50% effective, it makes it 0% effective.

    Question: So why is it that when 90% of the cases are in the placebo group do we call it 90% effective. Don’t we have to have closer to 95% of the cases happen in the placebo group to be 90% effective?

    Either way this is great news!

  68. John Andersen says:

    A sample size of just a few hundred infections is statistically weak with high error bars to draw conclusions about effectiveness of vaccines.. Error bars are simply to high on small sample sizes like this…

  69. Chris Phillips says:

    Pfizer/BioNTech vaccine approved for use in the UK, with vaccination expected to begin next week:

    1. debinski says:

      After listening to some of the CDC vaccine distribution meeting yesterday, I have doubts that the US is moving at “warpspeed” to get vaccines out. They spent most of the time on niceties, introducing people, thanking people, repeatedly asking if anyone had questions, etc. I kept yelling out loud (even though they couldn’t hear me) to cut to the chase. Very inefficient meeting. We can only hope the FDA has been able to pick up their pace. Why no meeting until Dec 10 – are they doing a more thorough review than the UK or are they just slow?

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload CAPTCHA.

This site uses Akismet to reduce spam. Learn how your comment data is processed.