On multiple occasions I have heard the innocent, if vaguely loaded, question posed:

Why not just do a double-blind placebo controlled trial (DBPCT) of the vaccine schedule? Wouldn’t that just end the argument?

For the sake of your attention span, and mine, I will answer the latter first because it’s a much shorter answer: no, we have absolutely no reason to believe that this would end the argument regarding vaccines because the criticisms regarding vaccines have never been based in reality and this is most apparent in the claim that vaccines cause autism (this is a very thorough review of the history of the nonexistent link between MMR and autism). I won’t attempt an exhaustive history at the evolution of that canard here, but the abridged version is pertinent:

• The myth did not begin with Wakefield, but with his publication discussing autistic enterocolitis in autistic children in 1998 and the insinuation that the combination of the three vaccine antigens together provoked autism, the lie gained a lot of steam among terrified parents. Certainly, some children will be recognized as autistic following receipt of the MMR vaccine. The vaccine is given between the first and second year of life where the signs of autism would generally become apparent to parents. The claim was also utterly devoid of scientific validity as the condition that Wakefield invented has never since been substantiated to exist (autistic people can develop enterocolitis, but there is no clear evidence supporting that there exists a specific form of enterocolitis distinct from others in autistic people), there has been no detected difference in the levels of MMR viral genomes among autistic and neurotypical children, and a vaccine delivered as an intramuscular injection containing highly attenuated pathogens making it into the intestines and causing inflammation is science fiction as this would require that the vaccine antigens cause a viremia which is impossible given that they have been successfully attenuated (assuming immunocompetence on the part of the host). Furthermore, the study had a sample size of 12 children, not all of which were autistic and of those who were some were known to be autistic prior to receipt of the MMR vaccine and whose purported symptoms of autistic enterocolitis preceded their receipt of MMR. The paper can easily stand in as a paragon for bad science even without its rampant undisclosed conflicts of interest and overt lying.

• Given what Wakefield’s publication did for vaccine confidence, despite its lack of biological plausibility and the absence of an association between autism and MMR vaccination, multiple epidemiological studies were conducted to allay parental concerns. No association has ever been found by a well designed study. For a sample of such epidemiological studies looking at MMR and autism you can look at this, this, this, and this. For vaccines other than MMR, I can refer you here, here, here, here, and here. Of course, many other studies exist that came to the same conclusion and I do not want to belabor the point or turn this into a gish gallop. Suffice it to say, we have exhaustively examined the situation and found no link, which suggests that none exists, or that the link is so small as to be undetectable. In any case, the absence of any epidemiologic association between vaccination and autism should have been on its own sufficient to quash any further lies about vaccines as a cause of autism.

• After MMR was exonerated (though it was never really implicated according to scientific standards to begin with) the goalpost shifted to malign thimerosal, an organomercury compound used as a preservative in some vaccines to prevent bacterial and fungal contamination at doses measured in micrograms. Many claim that only the link between MMR and autism has been examined, based on the erroneous notion that thimerosal is contained within MMR (you’ll note that some of the studies I presented above examine thimerosal-containing vaccines). Live vaccines, like MMR, do not contain thimerosal as that would inactivate the live components which are fundamental to the immune response the vaccine produces. It was insinuated that the microscopic quantities of thimerosal from the vaccine schedule cause mercury poisoning which leads to (resembles?) autism. Firstly, autism and mercury toxicity have absolutely no resemblance to one another (see table 1), and secondly mercury toxicity does not occur until well above the quantity of mercury that one would obtain from the entire vaccine schedule and occurs primarily through inhalation. Nevertheless, the AAP made a decision in 1999 remove thimerosal from all childhood vaccines and today it is contained only within the multi-dose influenza vaccine and people have the option of requesting a thimerosal-free flu vaccine, though there is no scientific basis to justify this. Still, the incidence of autism has continued to rise.

Then there was the claim that autism is brought about because children receive too many vaccines too soon. This became popularized after the case of Hannah Poling, a young girl who had a mitochondrial enzyme deficiency which resulted in profound disability, who then received several vaccines and subsequently developed autism. However, the reality is that scraping one’s knee likely represents a greater challenge for the immune system than the entire vaccine schedule combined. In fact simple math suggests that a child could receive an antigenic load equivalent to about 10,000 vaccines with no issue.

Today we know that autism is primarily genetic in nature (about 80%). Vaccines have thoroughly been excluded as a cause. Nonetheless, the myth persists.

There is no controversy regarding vaccines. The scientific consensus is that they are safe, effective, and necessary and we have explicit guidelines for when they would not be. Of all the pharmaceuticals in existence, vaccines are among the safest and they have been among the greatest public health innovations in history. They have saved more lives than anything else except clean water. They save society huge costs and they save lives. There is no debate among experts about these things. Unfortunately, the anti-vaccine movement has presented the illusion of a controversy through FLICC, firehosing, gish galloping and otherwise preying on the anxieties of parents who cannot be expected to know better and proving repeatedly that Brandolini’s law holds true (you need only look through this partial answer to see a demonstration of that- compare the amount of effort it must have taken to propose these links between autism and vaccines and how much time it took me to address them, even cursorily).

Now we arrive at the former question: why not do the DBPCT of the entire vaccine schedule? The Institute of Medicine (now the National Academy of Science, Engineering, and Medicine) wrote:

The committee supports the National Vaccine Advisory Committee Safety Working Group statement that “the strongest study design, a prospective, randomized clinical trial that includes a study arm receiving no vaccine or vaccine not given according to the current recommended schedule, would be unethical and therefore cannot be done” (NVAC, 2009, p. 38). In Chapter 6, the committee presents the formidable ethical and feasibility problems associated with the conduct of randomized controlled trials of children who receive all recommended immunizations and children who receive none of them and randomized controlled trials of children who receive all recommended immunizations and children who receive the recommended immunization on an alternative schedule. There are very low observed rates of adverse events with vaccination, which is another factor affecting feasibility of a randomized controlled trial. Because of these problems, the committee concludes that a randomized controlled trial comparing the recommended schedule with any alternative schedule would be unethical and infeasible and could increase the risk of vaccine-preventable diseases in individuals and in the community.

There is much to dissect in this remark. Firstly, let’s start with some definitions: a double-blind placebo controlled trial is a type of study in which neither the participants nor the researchers know whether the participants are in the control or treatment group. Once the experiment is complete, it is revealed who was in which group and the data is analyzed via statistical methods. For completeness: a single-blind study involves the participants not knowing which group they are in (but the researchers do), and a triple-blind study involves the individuals analyzing the data (i.e. statisticians) also not knowing which group is which in addition to the researchers and the participants. It belongs to a broader group of studies called intervention studies. From a scientific perspective, these studies are extremely useful because they allow the researchers an incredible degree of control to remove confounding variables. The study participants are randomized to be part of the treatment or control groups and the members of each group are matched as closely as possible to eliminate the effects of any confounding variables. Furthermore, in observational studies it can be very difficult to tell which direction causality goes, assuming it is a causal relationship. For example, one outmoded hypothesis regarding the etiology of Alzheimer’s suggests that the disease is brought on by aluminum. This is based on the observation that the brains of deceased patients who had Alzheimer’s have, on average, higher aluminum content than control brains. Seems reasonable, right? However, we are now coming to understand that Alzheimer’s has a significant component to it involving dysfunction of the blood-brain barrier, which could allow aluminum, which is ubiquitous in our environment, to enter the brain. Hence it is arguably more likely that the increase in aluminum levels observed in Alzheimer’s patients brains is an EFFECT rather than a cause of the disease and further evidence for why the aluminum hypothesis should be abandoned. The design of an RCT allows you to ensure that your cause precedes your effect, so to speak. Ideally, the sizes of the control and treatment groups should be similar. Assuming that we have a sufficiently large sample size, and we match our control and treatment groups well, these kinds of studies have excellent internal validity, meaning it is very likely that significant (in the statistical sense) differences between the control and treatment groups in the study population are causally related rather than just correlation. For this reason, these studies are commonly called the gold standard of scientific design. However, this is perhaps an oversimplification.

The hierarchy of evidence (which is by no means an absolute hierarchy but offers a valuable framework to begin considering which studies to prioritize in rendering conclusions) affords the honor of “gold standard” to systematic reviews and meta-analyses: the studies of the studies.

These aim to examine what multiple studies looking at the same research question have found and rendering a conclusion based on the sum total of those findings. This is accomplished through a search strategy and inclusion and exclusion criteria- which often exclude observational studies. Of course, garbage in means garbage out- a systematic review of weak studies does not somehow anoint those findings are robust. Still, the best type of study design depends entirely on the question being asked, and must be concordant with the principles of research ethics.

Another key point is that all intervention studies- like an RCT- must have specific endpoints in mind, meaning specific things we can measure to evaluate whether or not the treatment works. I don’t really think I have ever seen concrete endpoints to go along with this question in any of the permutations it’s been asked. The reason that we cannot simply do the study and then select the endpoints later is because the endpoints directly affect the sample size we need (see here for an explanation of statistical power) and which members of the population we use, as different endpoints are subject to different confounders and must be adjusted accordingly. Certain subjects will be inappropriate to include in the study because they may have a condition that would confound the results of a particular endpoint analysis. For instance if I want to see if a particular drug raises the risk of lung cancer, I should make an effort to exclude people with a history of smoking cigarettes from my study so that I can accurately compare my control and treatment groups (I ask that we set aside for the moment the considerations of ethics in such an RCT as I will discuss them in extensive detail shortly). The issue however is that the more response variables we examine, the larger our sample needs to be to prevent a Type 1 error from occurring as the higher the probability becomes that an association between our variables occurs solely due to random chance. When the endpoint is presented as health (as it so often is because of course), which we cannot quantify, the recourse is to examine the litany of conditions contained within the ICD-10 that would require an impossibly large sample size of literally millions. From a purely logistical standpoint the sheer cost in funding such a study would be astronomical.

I am quite confident that if you asked any career bioethicist if they could support such a study being performed (which is effectively what they must do when evaluating a study as a member of an IRB), they would ardently refuse. The design of this study is so egregiously unethical in its concept that it’s truthfully hard to know where to begin in explaining why, especially given that discussing whether a study is ethical requires substantial foundational knowledge in bioethics. Unfortunately, I lack both the means and time to provide readers with this, so I have liberally included hyperlinks with additional reading for clarification and can also reply to explain anything unclear in the comments. Firstly, a bit of background:

Clinical research refers to research that is done on humans or on human tissue (typically referring to Phases 1 through 4 of the clinical trial process), as distinct from pre-clinical research which refers to “Phase 0” studies which occur in animals or in vitro.

There is no question that principlism is overly reductionist for answering bioethical questions, but it has its uses, especially as a foundation for considering research ethics, and they are historically imbricated via the Belmont Report. As proposed by Beauchamp and Childress, none of the 4 principles is supposed to be inherently more important than another (though our culture is very autonomy centric overall- in some cases excessively so but that’s a personal opinion of mine and should be regarded as such), but beneficence, non-maleficence, and justice warrant special consideration. The reality is that most participants in clinical research will not benefit from it, despite what they may believe (when individuals believe that their participation in clinical research will result in tangible benefits to their own health it is termed therapeutic misconception). Most research does not make it to the bedside. Only a small minority of studies ever change clinical practice by the time we winnow them down through the stages of clinical trials. In some cases, the individual may be in the placebo group (in the case of randomization). And most importantly: anything that exerts an effect on your biology can also have adverse effects on your biology. Participation in clinical research has the potential to cause subjects significant harm. This is ostensibly at odds with both non-maleficence (do no harm) and beneficence (act in your patient’s best interests). Then there’s the matter of justice (the equitable distribution of benefits and burdens) which is a sticky matter indeed when it comes to research:

If you were after a soundbite and went around asking most people what single quality makes clinical research ethical, I doubt that you would hear anything more frequently than “informed consent.” However, while it may shock you, informed consent is NEITHER NECESSARY NOR SUFFICIENT for clinical research to be ethical. Firstly, there are considerable limitations on the ability to provide a clinical research subject with informed consent by the very nature of the fact that the clinical research is innovative in nature. The effects of the pharmaceutical on the human body, while we can make reasonable speculation based on preclinical findings and mechanistic reasoning, cannot be known in advance. Unforeseen harms and benefits are inevitable. Ergo, how does one provide the subject in question with an adequate understanding of the relative risks and benefits? To what extent informed consent is feasible at all is a hotly debated subject in the field of bioethics, even outside research. Of course that issue is putting the cart before the horse a bit. Not infrequently, the research subjects in clinical research may lack decision-making capacity as the intended target patient population may not have it as well e.g. the pediatric population, as is proposed in this study. Indeed, when it comes to decision making at the pediatric bedside, informed consent is nonexistent (but rather informed permission is which focuses on using the best interest standard, shared decision-making with parents, harm thresholds, the Rule of Sevens, and other standards to ensure that children receive the best possible care). Minimizing harms to vulnerable populations (such as those described by the Common Rule), particularly those who cannot give informed consent, is a central tenet to upholding justice. At the same time, excluding these groups altogether from these studies means that the quality of their care will stagnate from lack of innovation, which is a harm itself too as on a broader systemic level access to care will be unequal. The voluntariness aspect of informed consent is also complicated because many patient cohorts enter into clinical research through their clinicians, and they may naturally be concerned that refusing entry into a trial could compromise the fiduciary relationship between the patient and their provider.

Principlism may offer a base foundation for evaluating the ethics of clinical research, but it offers little pragmatically in direct application. Emanuel et al have put forth eminent criteria that are largely still in use specifically to that end in the form of 8 (originally formulated as 7) principles and benchmarks to evaluate them which have sometimes been called the Emanuel criteria. I won’t discuss each of these principles as they are thoroughly examined in both those links and this word count is already somewhat unwieldy, but I will be explicit here on where this proposed DBPCT fails.

Foremost, for a study to be ethical it must be scientifically valid. We do not simply jump to clinical research as soon as we have a question and we do not go directly to an RCT without a solid basis to justify that study design scientifically. Contrary to what many believe, there have been a number of studies examining certain health issues of vaccinated vs. unvaccinated individuals. For instance:

1. This study finds that there is a small but statistically significant difference in asthma rates between the vaccinated and unvaccinated in Germany, with vaccinated children having lower asthma rates.

2. The KiGGS study finds no relationship between being vaccinated and atopy (a genetic predisposition towards allergy) but notes that those who were unvaccinated had much higher burdens of vaccine-preventable diseases.

3. This meta-analysis finds that being vaccinated is associated with about half the risk of SIDS.

4. This study which finds that refusal of the influenza vaccine results in a significantly higher probability of severe health outcomes requiring ICU admission and hospitalization (even in previously health people).

5. This study which finds that the immune systems of vaccinated as compared with unvaccinated children do not differ much at all.

6. This study which compares the risk of vaccination with hospitalization due to infections not targeted by the vaccine schedule that finds no difference between vaccinated and unvaccinated except for Hib which is so marginally significant (the confidence interval very nearly contains 1) that I would say the result is very likely a type 1 error.

7. This study on rotavirus vaccination which found that those who completed the rotavirus series had a 33% risk reduction for developing type 1 diabetes (and several other interesting findings I invite you to read about).

I could go on but I think you get the idea and my goal is not to turn this post into a gish gallop. We have plenty of studies examining vaccines and their nonspecific health effects. In fact, part of conducting a clinical trial on vaccines involves the solicitation of adverse events from study participants which are then analyzed for statistical significance. Hence one could say this constitutes a rudimentary analysis of vaccinated vs. unvaccinated individuals and the nonspecific health effects of vaccines with randomization and blinding- for that specific vaccine, assuming it is completely novel as that is the only circumstance in which a placebo would be an appropriate comparator, ethically. Based on the sampling above (which I grant you is not comprehensive, but concordant with the bulk of the literature), arguing that there is good scientific basis to justify a study of the entire schedule vs. placebo despite the clear absence of evidence suggesting we might find any significant difference is untenable. It may be tempting as a corollary to conduct a retrospective study of the health of fully vaccinated and unvaccinated children (perhaps even partially vaccinated children to examine whether there is an existence of a biological gradient) but there’s a crucial problem with that. There are significant systematic disparities between unvaccinated and undervaccinated children and those who receive the entire schedule:

Undervaccinated children tended to be black, to have a younger mother who was not married and did not have a college degree, to live in a household near the poverty level, and to live in a central city. Unvaccinated children tended to be white, to have a mother who was married and had a college degree, to live in a household with an annual income exceeding $75 000, and to have parents who expressed concerns regarding the safety of vaccines and indicated that medical doctors have little influence over vaccination decisions for their children. Unvaccinated children were more likely to be male than female.

These are significant disparities in the social determinants of health that would greatly compromise the internal validity of such a study which makes rendering a conclusion accurately impossible.

More broadly, randomization requires equipoise on multiple levels. Clinical equipoise refers to a state in which there is genuine uncertainty on the part of the community of expert researchers about the relative benefits and burdens of a treatment compared with its comparator. There is no question that a substantial size of unvaccinated children would suffer considerably greater burdens of vaccine-preventable disease, which already sets up a very high barrier to the level of harm a vaccine must cause to meet the condition of clinical equipoise (unless we decide to surround the unvaccinated individual with highly vaccinated ones but this already creates an issue with randomization and introduces bias and the herd immunity effects could obfuscate other pertinent data). There is a clear absence of scientific basis to justify this. Another standard is known as personal equipoise, which states that the clinician conducting the study must not have a preference regarding the treatment vs. its comparator. In practice, this basically disqualifies any RCT as it is virtually impossible to achieve this condition. However, in this case, I think personal equipoise would be especially difficult to attain.

Irrevocably tied to the idea of clinical equipoise is the notion of standard of care. As science discovers new treatments, some inevitably perform better than others, and so they take the place of the old treatments because they work at least as well. Per non-maleficence and beneficence, clinicians have an obligation to provide the standard of care, meaning the best therapy available at the time for that patient’s condition. This means that when we have an effective way to prevent vaccine-preventable diseases (vaccines) we cannot ethically justify the use of a placebo. This is why in vaccine trials when a next-generation vaccine comes out, it is compared with the previous generation. The safety and efficacy of the previous generation one are already established. There is no point in licensing a new vaccine against the same thing unless it works at least as well. For instance, Gardasil initially contained only 2 strains of HPV antigens. Then, after it gained approval a quadrivalent vaccine came out covering 4 strains. What is the appropriate basis for comparison? Are we expected to leave the placebo group unprotected against HPV-associated cancers without their knowledge? That would stand in stark opposition to the role of the clinician and be categorically unacceptable. Hence it was compared to the bivalent vaccine. Then the nonavalent came out which was compared to the quadrivalent one, and so on. I have seen a popular meme call this a pyramid scheme. This is the epitome of agenda-driven cluelessness. This is the ONLY ethically appropriate way to do randomized controlled trials of vaccines. Forcing subjects to receive a placebo and be unprotected against diseases we can protect them against is not appropriate.

Another pertinent point to consider is that the blindness of the study would be readily compromised after a child enrolled became ill with multiple vaccine-preventable diseases. Additionally, for this study to be acceptable, participants must be permitted to withdraw from the study at any point for any reason.

The principle of favorable risk-benefit ratio requires a great deal, but of chief importance is the question of whether or not the risks incurred in this study justify the benefits societally. Each child randomized to the placebo group would be at risk for every single vaccine-preventable disease on the schedule. Tetanus is ubiquitous. Any open wound they get is a potential risk. Measles requires such a high level of immunity in the population that a study of this nature would likely cause outbreaks given the sample sizes it would require (on the order of hundreds of thousands to millions depending on the disease being examined). The same goes for pertussis. Given the body of evidence suggesting the safety of vaccines and the case-control and cohort studies done it is incredibly difficult to argue that the harm that the subject population would face is justified by the social value this study would provide.

Circling back, as this study would involve very young children, who per the rule of sevens are not even able to assent to treatment, there can be no informed consent here. As stated earlier, this is not the model used in pediatrics. However, I think you would be hard pressed to find parents willing to enroll their child in this study regardless of their position on vaccines. If I had children I absolutely would not. From the perspective of the anti-vaccine parent: there is a ~50% chance that their child would receive the vaccine and they (the parent) won’t know if this is the case, nor will their clinician- per the double-blind nature. On the other hand, pro-vaccine parents risk allowing their child out into the world with no protection other than herd immunity and given today’s climate on the issue of vaccines, that seems woefully inadequate. Coupled with the need of a gigantic sample required, this seems to be not only unethical but unfeasible.

Respect for participants for such a study is especially complicated. Children, especially those this young, are a protected group per the Common Rule. Presumably this would mean nestling the placebo group in communities with high vaccine uptake, but this would compromise the scientific validity of the study as it would limit the ability to randomize, and the presence of high levels of herd protection could obfuscate the impact of being unvaccinated. Another point of concern is the fact that this study would not occur in a vacuum. Participants would be out in the outside world where they could serve as the nidus for infection in vulnerable individuals. Indeed, minimizing harm to the subject population seems virtually impossible. The participants are either immunized or not. The study must generate data generalizable to the population at large to be informative regarding the risks and benefits of vaccination and hence tampering excessively to protect the subject population compromises the sterility of the experiment.

These things taken together suggest that, ultimately, it is impossible to justify a study of the vaccine schedule against placebo ethically. Each vaccine has undergone staging through clinical trials and passed to make it to the market and is subject to phase 4 (post-marketing surveillance) of pharmaceutical research. Retrospective studies continue to be actively done and continue to be reassuring. Please: stop asking for this study. It is absolutely NOT an accident that it has not been done.

Key References

1. Bernard L. Resolving Ethical Dilemmas: A Guide for Clinicians. 5th ed. Philadelphia: Lippincott, Williams & Wilkins; 2013.

2. Emanuel E. The Oxford Textbook of Clinical Research Ethics. Oxford: Oxford University Press; 2008.

3. Hawkes D, Dunlop R, Benhamu J. Calls by alternative medicine practitioners for vaccinated vs unvaccinated studies is not supported by evidence. Vaccine. 2016;34(28):3223–3224.

4. Rid A, Saxena A, Baqui A, Bhan A, Bines J, Bouesseau M, Caplan A, Colgrove J, Dhai A, Gomez-Diaz R et al. Placebo use in vaccine trials: Recommendations of a WHO expert panel. Vaccine. 2014;32(37):4708–4712.

5. DeStefano F, Shimabukuro T. The MMR Vaccine and Autism. Annual Review of Virology. 2019;6(1):585–600.