The dark side of randomized double-blind studies

While randomized double-blind tests have become the gold standard for testing of various hypotheses, they also raise some ethical questions for which there are no easy answers. In the case of the covid-19 vaccine trials, there were discussions as to when the study should be unblinded so that those in the placebo group could also be vaccinated. Commenter Bethany said that a co-worker was in the Pfizer study and told her that it has been unblinded.

But before that happened, a group of scientists who are members of the WHO Ad Hoc Expert Group on the Next Steps for Covid-19 Vaccine Evaluation urged that the double-blind study continue.

While vaccine supplies are limited, available vaccines are still investigational, or public health recommendations to use those vaccines have not been made, we believe it is ethically appropriate to continue blinded follow-up of placebo recipients in existing trials and to randomly assign new participants to vaccine or placebo. Moreover, under these conditions, we believe that trial sponsors are not ethically obligated to unblind treatment assignments for participants who desire to obtain a different investigational vaccine. People who enroll in clinical trials for altruistic reasons would probably understand the value of gathering data that will further elucidate the safety and efficacy of these vaccines and their appropriate use.

Large, placebo-controlled, phase 3 efficacy trials could provide much of the needed information if they have appropriately prolonged follow-up while random assignments are still blinded. Such continuation would yield unbiased evidence on the duration of protection and on longer-term safety, including assessment of any evidence of the vaccine eventually enhancing the risk of severe disease (as was recently detected by continued follow-up of placebo recipients in dengue vaccine studies).

This opportunity to obtain reliable evidence about longer-term effects would be destroyed by early unblinding and immediate vaccination of participants assigned to placebo. Although each participant has the option to pursue any available intervention, if substantial numbers of participants choose not to do so, continuation of blinded follow-up in a population in which no licensed vaccine is being deployed could yield important and unexpected findings that would be difficult to obtain reliably any other way.

There is a price to be paid when using this method to determine the effectiveness of life-or-death treatments. Vaccine inventor Paul Offit is concerned about the ethics of double blind tests where someone will live because they got the vaccine, or die because they got the placebo. He says that with the covid vaccine trials, one person died during the trial and they all held their breath and then everyone gave a huge sigh of relief when it was found that they were in the placebo group. But what about that person who died?

PAUL OFFIT: Because there has never been a medical breakthrough in history that has not been associated with a price. When Thomas Francis did the polio field trial in the mid-1950s, Jonas Salk had made his vaccine, but he didn’t know whether it worked or not. So they chose to do a big field trial. Four hundred and twenty thousand children were given his vaccine over a year period funded by the March of Dimes. Two hundred thousand were given placebo – first- and second-graders throughout the country. And then after it was over, Thomas Francis stood up on the podium at Rackham Hall at the University of Michigan and said, safe, potent and effective. That’s what he said. Those three words were the headline of every major newspaper in this country. I mean, church bells rang. Synagogues and churches held special prayer meetings. Department stores stopped. Trials stopped, you know, so the judges could hear that announcement. It was announced over the Voice of America.

Well, the question is, how do we know that it worked? We knew that it worked because 16 children in that study died from polio, all in the placebo group. Thirty-six children were permanently paralyzed, 34 in the placebo group. But for the flip of a coin, those children could have been alive and well today. Those were first- and second-graders in the 1950s. I was a first- and second-grader in the 1950s. I mean, those people suffered or died because they just happened to be in the control group. That’s what knowledge takes. And that was – that statistic never really rang. I mean, we were so busy celebrating that that I think we didn’t really stop and take a look at just how one comes to acquire knowledge.

Jonas Salk, inventor of the vaccine for polio, said “I would feel that every child who is injected with a placebo and becomes paralyzed will do so at my hands” and that those who demanded a placebo controlled trial to achieve a statistical end point placed their “values in which the worship of science involves the sacrifice of humanitarian principles on the altar of rigid methodology,”

This article has more on Salk’s objections.

Placebos have been essential to clinical trials for decades. It’s vital that neither the volunteers nor the staff running the trial know who is randomly assigned to get the vaccine or the placebo. This “blinding,” as it’s called, eliminates the chance that people will behave differently depending on which treatment they get, potentially skewing the trial’s results.

Yet the prospect of giving people something useless in the face of a life-threatening disease has always been fraught. Even Jonas Salk balked at the idea of giving people placebos when researchers designed a trial to test his new polio vaccine in 1953.

“I would feel that every child who is injected with a placebo and becomes paralyzed will do so at my hands,” he complained. The study, Dr. Salk declared, “would make Hippocrates turn over in his grave.”

But Dr. Salk lost that fight, and the placebo-controlled trial went forward. It clearly demonstrated that the polio vaccine was safe and effective. Only when the trial was over did the children who received the placebo get the vaccine — along with millions of other children.

When it comes to trials for things that are not life or death, there is no question that the double-blind tests are the best. For example, if one is developing a drug for (say) migraines, the people given the placebo may suffer terrible headaches during the trial, but they will not die in the interim. Perhaps someone will come up with a better and more ingenious method than the double-blind test to measure the effectiveness of treatments when the issues are life and death.

As long time readers of this blog know, I got polio as a child. I am trying to think what my feelings might have been if I had been part of the placebo group when I got the disease. Would I (and my parents) have been accepting and philosophical about it, feeling that my misfortune was for the greater good, that it was this study that gave people the confidence to give the vaccine to everyone, preventing a huge amount of suffering? Or would I have been resentful about not having been given something that might have spared me having to deal with the disease? I do not know. It is the essential problem of utilitarian calculations, that things that sound very reasonable in the aggregate can be problematic when viewed on the individual level.


  1. says

    The difference, for me, is a matter of informed consent.

    If I, as an adult, enter into such a study with full informed consent, then I’m risking my life for the good of humanity.

    When you involve children, or anyone who cannot give informed consent of their own free will, then I have to say no. I don’t know what the solution might be; there might not be a workable solution, but consent must be informed and freely given.

  2. says

    Medicine has a long history of experimenting on patients. I feel that, rather than treating that as shameful, we should emphasize when it has a positive benefit. Yes, some in the placebo group got sick, but they were going to get sick anyway (How do we know that? They got sick on the placebo) what would be problematic would be if patients in the placebo set were deliberately exposed -- which is how the US prison population was, basically, treated. I’m surprised that the prison industrial complex hasn’t realized they can barter time off for being an experimental subject, therefore moving the US one step closed to Dr Mengele’s ideal (Mengele was in love with the US eugenics movement -- it’s safer to say that he was a product of US civilization than nazism)

    I think that given the sizes of the experimental cohorts, the clinical trials were run about as well as they could be.

    A better approach would be to treat covid-denying republicans as a volunteer placebo cohort and just give them little dunce hats to wear. Except that is basically what happened. I guess we should not create a placebo cohort when there is a volunteer placebo cohort standing by.

  3. file thirteen says

    Hindsight is 20/20, and ethical hindsight adds a layer of hysteria on top. If it had been known that the vaccine was safe, potent and effective, then of course all the children on placebos should have been given the real thing. And what about those children who didn’t even get to go on the trial: didn’t they matter? How many of them were killed or paralysed?

    But it wasn’t known that the vaccine was safe, potent, or effective, that is the whole point. To discover whether it was, the randomised double-blind trial had to go ahead. Don’t blame the trial, blame the disease. Even with hindsight, the children who took part had a greater chance of avoiding the disease than those that did not; each had a 50% chance of receiving the vaccine.

    Suppose the vaccine had been found to be impotent, ineffective, or worst of all, unsafe, what then?

  4. xohjoh2n says

    where someone will live because they got the vaccine, or die because they got the placebo.

    Of course, *at the time the trial is being put together*, we don’t know whether it is a vaccine or not. It might be another placebo. It might be a poison with no therapeutic value. You’re not denying the placebo group an effective treatment, because you do not yet have a known effective treatment.

    Well, the question is, how do we know that it worked? We knew that it worked because 16 children in that study died from polio, all in the placebo group. Thirty-six children were permanently paralyzed, 34 in the placebo group.

    And the flipside of that is that without the trial, 32 children would have died from polio, and 68 would have become paralyzed. Not to mention the many many millions who came after. (Peak prevalence in the 40s was half a million dead or paralyzed *per year*.)

  5. DonDueed says

    I can think of only two alternatives to double-blind vaccine trials.

    1. Trials with no control group — vaccinate everyone in the trial. The problem here is that the only control group is the general population, so the result is likely to be ambiguous — you can’t be sure whether the vaccine was effective or had only a placebo effect. Further, you’re relying on whatever statistics may be available on infection rates in the general population, which may be incomplete or significantly inaccurate. The problem is worse for a disease with low rates of infection or transmission, in which case it’s much harder to separate any protective effect from random variation. Also, your trial group may not reflect the general population (for age, gender, ethnicity, and so on) closely enough. Bottom line: fuzzy results at best.

    2. Challenge trials. The ethical issues here are obvious — if your vaccine is ineffective, you’ve probably condemned the entire test group to suffering from the disease and (likely) passing it on to others. On the other hand, you can get valid results using a much smaller test population.

    So unless, as Mano suggests, someone cleverer than me invents a better approach, double-blind tests are the best way to determine the efficacy of a drug or vaccine with confidence and minimal risk.

  6. seachange says

    The real world is as it is. Since the real world is as it is, there are people right now who will not take the vaccine. All y’all can be smug and smart and deride these folks all you like but this will not change the real world as it is. Reality (or the god/godesses) *does not care* about your perhaps entirely genuine and well-earned brilliance or your sense of superiority.

    Therefore, the placebo exists, in this case the placebo not being a injection with adjuvant or a shiny shiny pill but religion/politics/limitations of the human brain. Since the placebo exists, the placebo cohort still exists.

    Nobody who is making this vaccine is making this cohort take their vaccine. But they might be able to get information about this cohort. The useful pratical question is is it ethical to do this?

    I say yes.

  7. cafebabe says

    The last time Mano hinted at this question (see first link in the OP) I described my personal quandary from being on one of the vaccine trials. I shall not revisit those details here.

    However, I would like to draw attention to a thoughtful article in the December 10 edition of the estimable Bulletin of the Atomic Scientists addressing exactly this issue. The most sensible of the variations discussed notes that there are multiple concurrent trials: different potential treatments up against a placebo. Given that a placebo is a placebo you could share the placebo data over all of the trials with comparable circumstances. This would reduce the number of at-risk participants overall.

  8. Mano Singham says


    Thanks so much for that link! So it looks like we may be able to find a better way.

  9. xohjoh2n says


    This would reduce the number of at-risk participants overall.

    Actually, it increases the proportion of at-risk participants.

    (Though not necessarily their absolute number.)

    Look, there’s a definitional problem that I see repeated several times above. The people who end up in the placebo group of the trial at at no more or less risk than had they not joined the trial at all. As I understand it they are not being deliberately exposed to the virus (what I infer from the use of “challenge group” above), nor are they being encouraged to change their behaviour to put themselves at more risk. In fact it should be made very clear to them beforehand that even if they end up in the non-placebo group there is a good chance they are not protected. (Because, this being a clinical trial, we no not know for sure yet that the vaccine actually works.)

    It is in fact the non-placebo group who are “at risk”: they may in the worst case be both unprotected from the virus, *and* might suffer real negative effects from the vaccine that were not noticeable without the power of phase III numbers.

    Now, apart from that, I understand there are statistical problems when you try and use a single experiment to test multiple hypotheses at once: the proper way is to make an adjustment to account for that which reduces the resolving power of the experiment. This would likely not have made a difference in this case where 90% efficacy is going to be noticeable anyway, but for trials on other treatments with much more marginal benefits it could either make them useless or extend their numbers or duration beyond the point of practicality. (And of course, until we did these particular phase III trials, we didn’t know we were going to get 90% efficacy. It could have been much lower, but still clinically useful.)

    Now that’s not to say you couldn’t arrange a trial like that, but you’d have to crunch the numbers very carefully first, and one of the key input numbers to that analysis (margin of benefit above placebo) is one of the very things you’re conducting the trial to find out.

  10. cafebabe says

    @9 No.
    It is quite commonplace to collect evidence of the effectiveness of several interventions against a single do-nothing cohort. And not just in public health!

    In the particular context of, say, phase three trials for Covid-19 vaccine candidates it is already established that the candidate produces antibodies against the virus, all that is in doubt is the possibility of (possibly serious) rare side-effects and the extent of the immunity granted. Running two vaccine trials in parallel and sharing a single placebo group would reduce both absolute and relative total risk.

  11. xohjoh2n says

    @10 Yes.

    Yes, you can, but you have to adjust for statistical power.

    By the start of phase III you have a good idea what will happen, but 50% of drugs still fail at phase III so “already established” is too strong.

  12. Suren Fernando says

    Is there any difference between those who were on the placebo arm of the trial and the 330 million others in this country who did not have access to a vaccine? The WHO Ad Hoc Expert Group was only suggesting not unblinding the trial till the vaccine became available to the general public. One possible option would have been to offer to unblind and vaccinate participants two weeks ahead of when their cohort was to have access to the vaccine.

    The Ad Hoc Expert Group cites the recent (2018) instance of a vaccine trial for dengue fever in which the efficacy of the vaccine wore off after a few months, and then those who had been vaccinated experienced more severe disease. This was only evident by making a comparison between the vaccine group with the placebo group.

    In the view of Ad Hoc Group, four months, which was the duration of the Pfizer and Moderna trials, was too short, and the numbers of participants in the Pfizer and Moderna trials, 43,000 and 30,000 respectively, were too small to tease out rare but serious side effects. If serious side effects do arise down the road, this will give rise to an increase in vaccine hesitancy, which is already a problem in the US.

  13. Sam N says

    I love threads like this. Honest arguments as strong as they can be for different positions.

    I’ve got to say the pro-blind side is far more convincing to me.

    Questions I have. If common well-tested adjuvants and preservatives/stabilizers are used is that factor even necessary to test anymore?

    Follow up, are there mRNAs that would pose significant risks to humans?

    I’m pretty sure the answer to the second question is yes. I should look this up, I’m not a virologist, but if my memory over 20 years is accurate the tobacco mosaic virus is just mRNA. Infects plants not humans.

    Third, what confidence is there that the sorts of mRNA in vaccines could be poisonous.

    In general I’m wondering if mRNA vaccine tech is well enough understood that our caution, using the best standard, isn’t warranted.

  14. Mel says

    My (70+yo) parents are in the AZ/Oxford university trial in the UK where the vaccine has just been approved. They have been told that if and when they are offered a vaccine as part of the UK’s programme, they will be unblinded as to what they received in the trial. Therefore, they will continue to contribute to the trial data (weekly swab testing and symptom diary), until such point as it is their turn to receive a vaccine, if they have not already done so.

  15. xohjoh2n says


    If common well-tested adjuvants and preservatives/stabilizers are used is that factor even necessary to test anymore?

    Whether or not you consider it “necessary”, you do test it again, because what you ship is what got regulatory approval, and that is what went through clinical trials. The full formulation. If you decide you want to change that, you re-test and re-approve.

    are there mRNAs that would pose significant risks to humans?

    Well, some RNAs can themselves be “active”, RNA can have catalytic properties. Ribosomes for example, the machines that construct new proteins, are mostly made of RNA. You have to try quite hard to get there though, and we have not.

    Plain mRNA does very little, and it is constantly being degraded and removed from within cells, which then regenerate it from their original DNA. So the mRNA itself is going to do you no harm, and not going to stick around for any length of time.

    However: an mRNA is an instruction to make a particular protein. Snake venom for example is just a protein. So if you introduced the mRNA encoding a snake venom peptide into your cells, there’s a good chance you’d be in a bit of trouble. But we’re not doing that.

    You mention TMV. (Its RNA core is not *quite* an mRNA, but for simplicity let’s ignore that.) If you’re a tobacco plant, why would introducing just that RNA into your cells be bad for you? Well, the RNA encodes for everything needed to make whole new infectious virus particles. The RNA itself encodes for 4 distinct proteins which the host cell gets right on to manufacturing as soon as the RNA is introduced. One is the coat protein, two others are involved in replicating and organising the viral RNA, and the fourth allows the virus to directly infect neighbouring cells (without waiting for cell lysis). By the time the original RNA is degraded, there are a whole bunch of new copies of it so it just keeps going and going until eventually the cell dies, releasing more virus for longer range infection.

    Coronaviruses are a bit more complex than TMV, but have some similarities. The viral genome is also a single-stranded RNA molecule that engages the host cell machinery on entry to replicate the virus, and encodes everything it needs to do that.

    So if you somehow introduced just the viral RNA into your cells, by some other method without using the normal envelope/spike/other structures, the end result would be no different than just infecting you with the virus. We’re not doing that.

    The mRNA we actually use is just the RNA for the spike protein. All of the other replication and structural stuff is not present. We know that is true, because we made the mRNA that way. So on entry there is a quick burst of making lots of copies of the viral spike protein, until the mRNA is degraded and the process stops. It stops in that cell, and since we included no replication capability, it stops for good everywhere. The spike proteins get noticed by the immune system, which is thus primed to tackle the real virus should it turn up later. The spike proteins themselves are very good when attached to the rest of a virus particle at gaining entry for the virus into a host cell, but as is the nature of these things, pretty much useless at anything else. So not much scope for harm on that front either. (And as soon as the immune system is primed, it’ll start mopping them up and disposing of them.)

    Now that’s the theory as to why an mRNA vaccine should be safer than say a traditional attenuated vaccine -- often containing real live virus! -- but that’s pretty much meaningless, practically speaking. We don’t inject new things because they have a good story behind them. Clinical use requires that to be *proven*, which we have done using a scaled series of clinical trials. So far these new types of vaccine appear to be both safe and effective. *That* is why we consider them safe, though *all* drugs undergo continuing monitoring even after approval, just in case we did miss something.

    It is said that by the time widespread rollout in this country reaches my cohort we’ll be mostly using the locally made Oxford/AZN vaccine (easier to scale up and easier to handle) and won’t be given an individual choice. If I did have a choice though, I’d much prefer one of the mRNA ones.

    (The AZN vaccine is a live virus of a different type that has had just the covid-19 spike protein code inserted and its own native replication machinery broken.)

  16. Sam N says

    Thanks for the lengthy post, xohjoh2n.

    The primary risk I see would be that the ‘spike protein’ could bind to certain classes of receptors and do harm to the organism that way. Although do you know if theses proteins even make it to the outside intact? Or are they being chopped up and presented by MHCs. Or I suppose acting as a transcription factor or activating transcription factors mucking about with normal cell function. Given the nature of the spike I’d be far more concerned about potential receptor interactions though.

    I feel a bit bad about you putting so much thought into remedially educating me. Is there a good high level review or website on mRNA vaccines and everything we know about how they work and potential risks?

    In this case I see the public health utility of going through this rigorous process. I’m thinking about future pandemics. Something with higher lethality or more difficult to control the spread (not that USA managed that well). I’d be willing to accept a lower degree of evidence and thus confidence for speed’s sake. But I can be a bit gambler at times.

  17. xohjoh2n says


    In the vaccine context, the spikes “job” is to get presented via MHC to trigger an immune response, however: I do *not* know, but I’d consider it an extremely safe bet that the biological reality of it is extremely messy and yes, you’re going to get at least locally a certain amount of raw spike protein just swilling around the area until it gets mopped up. That means there’s a good chance it’ll find its receptor and bind. (Though I suspect injection site and receptor prevalence there was given some thought.) At which point it’ll get degraded via that route instead. Without the rest of the virus attached to it, there’s just not a whole lot more it can do. At worst those cells will get killed, but I would guess there’s just not enough and too localised for that to be really noticeable.

    As for the wider cellular effects you mentioned, that sounds extremely unlikely to me given how closely targeted the spike protein is going to need to be to match its preferred receptor. I couldn’t say whether it’s possible or not, you’d need someone actually in the field to say, but I’d personally consider it a very safe bet against.

    But here’s two things:

    First, as I already stated, we do a series of scaled clinical trials before releasing drugs, and continually monitor afterwards. Something like you suggest would likely be noticed already. The longer we go without noticing negative effects, even if they do turn up later the less serious they are likely to be. Really, we have no obvious mechanism, and no empirical observation of the fact.

    Second, consider this: you’re worried about some as yet unknown effect of being injected with the spike mRNA, because it’s all so new. You know what else introduces the spike mRNA into you? Infection with SARS-CoV-2. Exactly the same thing. Along with the whole of the rest of the virus. And given the current rise in fast-spreading variants (which were predicted earlier this year and have now come to pass), combined with many countries’ terrible ineffective responses, without an effective vaccine the odds would be extremely high you’re going to be exposed to that at some point soon anyway.

    Now would you prefer to have your spike mRNA as a component of an uncontrolled replicating agent with a whole load of other junk thrown in as well, with its known mortality risk and long term disablement risk, applied to the parts of you most vulnerable to it, or alone as part of a carefully constructed treatment that has been tested and appears to show no serious side effects and appears to be extremely effective at preventing infection by the former, injected into a site that is not the original viruses’ preferred target area?

  18. xohjoh2n says

    Just to add: the AZN vaccine also includes the spike mRNA coding as a component of its modified viral genome. So with that “traditional” vaccine, you’re *also* getting this so-called “novel” mRNA, but again with a whole bunch of extra monkey adenovirus junk on top. Really, the new vaccine method is just a more efficient delivery mechanism of that, both of them have the main job of getting you to produce spike yourself to provoke the immune system, and as long as it’s proven to work I’d consider not having to include all that extra (active virus) junk a bonus.

  19. Sam N says

    @18, I’d like to clarify, based on my understanding of biology, mRNA vaccine seems incredibly safe and ideal. Especially for someone like me that has rarely shown immune overreaction (some mild swelling from a bee sting).

    It’s why I’m pushing the boundaries wondering if the current validation process is overkill in light of a dangerous pandemic. I’d expect in less pressing situations, for example, visiting an area that potentially exposes you to Yellow fever.

    We seem to be generally on the same page regarding risk of the mRNA vaccines (like you I prefer them).

Leave a Reply

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