Thoughts on Placebo-Controlled Trials of Vaccines

Image credit: U.S. Centers for Disease Control and Prevention

The topic of placebo-controlled trials of vaccines, and the question of whether it’s a good idea to conduct them for new and updated vaccines, is suddenly much in the news. My goal in writing this post is simply to register my opinion on this issue, recognizing that policy decisions are in the hands of the new leadership at FDA, CDC, and HHS and, hopefully, their respective staff, who have expertise in this domain. As a career clinical investigator who has led more than 60 clinical trials and then spent years as a federal government regulator, my views on this subject are informed by a long baseline of real-world experience, and I hope that laying those views out as directly as possible will help others think about it.

When it comes to causal inference—that is, determining the relationship between cause and effect in the context of medical intervention intended to improve health—there is no substitute for the tool of randomization. A randomized controlled trial, or RCT, assigns research participants to receive either an experimental therapy or a comparator (an inert placebo, or in some cases, a different active therapy if one exists and has been approved for use, or usual care for the issue under study). Because the process of selecting participants in an RCT to receive a therapy or a placebo/control is random and unpredictable (and in many cases hidden from both patient and physician), it’s more likely that any differences in outcomes between the group—good or bad—are due to the effects of the intervention and not because there were substantial differences between the active group and the placebo/control group that could have affected the result (for example, if there were differences in health status, age, other underlying diseases, etc., between the test groups).

In the absence of randomization, these issues of confounding (mistaking or not being able to determine the likely reason for an observed effect) and calculating time zero (the time when a decision is being made about the intervention and the time at which measurements need to be made in order to adjust for confounding) make nonrandomized, observational estimates of treatment effect a daunting exercise.

However, because of a combination of practical and ethical factors, only a small fraction of important questions about medical interventions can be answered by RCTs in our current system, so we have no choice but to be selective. I’m the first to say that we need systemic reform of our evidence generation system in order to move to a “learning health system” in which high-quality observational and randomized studies are a fundamental cornerstone of healthcare delivery, but we’re far from it now (I’ll have a lot more to say about this is subsequent posts).

As for vaccines, I think of the issue in several dimensions:

• The first is the question is the strength of evidence that the fundamental vaccine in question has been proven to have a favorable benefit-risk balance. For brand-new vaccines, placebo-controlled RCTs are needed to generate the pivotal evidence to inform the key decision about whether the vaccine is “safe and effective,” meaning that the benefits outweigh the risks in a clinically meaningful way. It is important to understand that this is and has been the standard for the approval or authorization of emergency use of novel vaccines by the FDA. At the other end of the spectrum, when a vaccine has been shown to save lives and prevent serious illness with an acceptable risk profile, placebo-controlled trials would place research participants at an otherwise avoidable risk. This is a situation that would not be allowed by ethics boards/institutional review boards (IRBs) or regulatory authorities.

• The second dimension is the degree of underlying risk in the unvaccinated state. For many diseases and health issues, when the risk is very low and there is a proven treatment for symptoms, placebo-controlled trials are entirely reasonable. Even when there is significant risk from long-term withholding of treatment, short-term risk with carefully informed consent and a context-specific study design can be acceptable, as in the case of clinical trials for hypertension.

• The third dimension—the question of who is put at risk—is especially important for vaccines. With some vaccines, the issue is not simply whether getting vaccinated protects oneself or one’s child. For highly contagious and potentially devastating diseases like measles and polio, an unvaccinated person poses potentially serious risks to others (including very young children who cannot be vaccinated against these diseases but can catch them from an infected person). Accordingly, the risk that must be considered is substantially different than the ones that accompany a vaccine like the one given to prevent shingles, which is primarily protecting the individual.

• The fourth dimension is the issue of periodic updates to vaccines with proven efficacy and acceptable safety for the vaccine construct. For influenza vaccines, we accepted many years ago that the vaccine would be updated to match prevalent strains of the virus without the same amount of extensive testing used for wholly new vaccines. While we would of course never stop trying to improve influenza vaccines, this approach to updating an existing “platform” has been shown to be successful based on follow-up surveillance for outcomes in vaccinated and unvaccinated people. The same approach has been used for mRNA vaccines for COVID-19, and there has been considerable enthusiasm in the scientific, medical, and public health communities for using this approach for a broad array of vaccines amenable to the repeatable mRNA construct.

• Finally, and perhaps most complex, is the issue of equipoise. Equipoise refers to a state in which the balance of benefits and risks of an intervention compared with some alternative choice are uncertain enough that there is general agreement that the answer is not known. Several dimensions can be considered, including the amount of uncertainty and who is experiencing it (patients, doctors, experts, etc.). So, what happens when an intervention is proven and recommended by experts, but large proportions of clinicians and potential vaccine recipients decline to be vaccinated? Here, I believe that for vaccines like measles and polio, the risk to others is great enough that the public health system and government should encourage vaccination to the point that the population risk drops below the calculated threshold (this entails a vaccination rate of at least 92% for measles or polio). The situation with vaccines for COVID-19 is different; the vaccine does reduce transmission, but only modestly and for a short duration; however, it has a major effect on death and hospitalization. Nevertheless, the majority of non-high-risk individuals and parents are opting not to take the updated vaccination despite the current (1/7/2025) CDC recommendations.

So, to return to the question of whether placebo-controlled trials for vaccines are appropriate: for measles and COVID-19, I come to different conclusions. In the case of measles, I believe that the vast majority of ethics committees or IRBs that oversee the ethical conduct of research would consider placebo-controlled trials unethical: there is an existing vaccine known to be both safe and effective, and deliberately withholding such an intervention in the context of a clinical trial presents unnecessary risk to the unvaccinated individual and additional risks to others with whom these unvaccinated people would come into contact.

In the case of COVID-19 I believe it would now be quite reasonable, and even advisable, to conduct placebo-controlled trials for “boosters” using updated versions of the vaccine in people who are not high-risk (not elderly or immunocompromised and without major comorbid diseases). On the other hand, for high-risk people, the “boosters” with updated vaccines have been shown to reduce death and hospitalization. The reasoning for this approach is reviewed in my recent JAMA publication. The trials would need to be large, with long-term follow-up to include the recent evidence that the vaccines may prevent long COVID.

Considering this logic, I hope that we can agree that blanket statements like “placebo-controlled trials should always be conducted with vaccines” or “placebo-controlled trials of vaccines are unethical” are both wrong. Each case needs to be considered in detail. That’s why we need unconflicted and non-political civil servants at the FDA to make these decisions as they work with, and oversee the work of, vaccine developers.