Vaccines for Children
Are COVID-19 vaccines safe and effective for children?
One would hope so. COVID-19 rates for Americans under 18 have more than quadrupled since July, and children now represent about 27% of new cases (see here). The only vaccine currently available to children is the Pfizer-BioNTech vaccine, authorized for ages 12 and up. On September 28, Pfizer sent the FDA efficacy and safety data for 2,268 5 through 11 year olds, in advance of a request for emergency authorization to use their vaccine with this age group. Pfizer's original hope was that approval would be granted by Halloween, but a combination of on-record and off-the-record sources now suggest November at the earliest. Either way, most experts are optimistic that approval will come soon.
Before examining what Pfizer's data tell us, it's worth mentioning that the company is clearly (and understandably) in a hurry. For instance, Pfizer initially planned to study 1,500 5 through 11 year olds, but the FDA objected that this wouldn't be enough participants to adequately detect adverse side effects. The FDA requested that Pfizer increase their sample size to 3,000. Instead, as I mentioned, Pfizer provided data on just 2,268 children.
Pfizer's eagerness to introduce its vaccine to a new market doesn't mean that the vaccine is ineffective or unsafe. But the small size of the trial sample does raise minor concerns about effectiveness, as well as more serious concerns about safety. (Spoiler alert: I will spell out these concerns, but I won’t be arguing that younger kids shouldn't receive the vaccine. If people stopped taking medicines that aren't 100% effective and safe, there wouldn't be any medicines left to take.)
Pfizer has made public a brief summary of the data it submitted to the FDA. With respect to efficacy, Pfizer reports that one month after the second dose of vaccine, children's antibody levels were comparable to those of a 16 through 25 year old control group (and comparable to levels in the general population that confer 90% percent effectiveness; see here for a definition of effectiveness).
Is Pfizer's sample truly large enough to demonstrate efficacy? In a sense, yes. The FDA requires that drug trials include power analyses, which are statistical techniques that can be used to determine the minimum sample sizes needed to obtain certain effects (in this case, an acceptably large difference in antibody levels between a vaccine group and a control group). However, Pfizer's sample size may not be large enough to adequately disaggregate among key subgroups, such as gender. The FDA will determine whether or not this is a concern (and perhaps reiterate their initial request for a sample size of 3,000).
The possibility that Pfizer's vaccine varies in efficacy across different subgroups of 5 through 11 year olds is realistic but not necessarily more than a minor concern. Although gender differences in both efficacy and effectiveness have been documented for many other vaccines, studies on COVID-19 vaccines have not consistently distinguished such differences, in part because effectiveness rates have been high overall, and in part because individuals of any gender are assumed to be much safer following vaccination (see here for discussion).
As for the possibility of adverse side effects, I believe there's a greater cause for concern. To explain my reasoning, I need to make a quick detour through a broad and seemingly unrelated question: How large should a sample be?
The simplest answer to this question is: the larger the better. You know, for instance, that surveying a million registered voters is better than surveying a hundred, because the goal is to generalize from the sample to a more inclusive population (all registered voters), and one million is a bigger chunk of that population than one hundred would be.
Obviously size isn't everything. If you survey a million Democrats, your results won't generalize well to the population of registered voters, because only 33% of that population currently identifies as Democrats. At minimum, you would want your sample to also include Republicans (29% of registered voters) and Independents (34% of registered voters). In other words, you would want a sample that’s representative of the population.
Sample size and representativeness are both important; when samples are large enough, representativeness may become the more important consideration. (See Appendix 1 for a famous example.) At the same time, what counts as "large enough" or "representative enough" changes from situation to situation. Thus, for emergency authorization of drugs such as COVID-19 vaccines, the FDA relaxes its standards for sampling and other aspects of methodology (more on this in Appendix 2). "Relaxes" is polite way of saying "lowers", but these terms don't necessarily imply shoddy research. The FDA notes that it errs on the side of caution, so that even lowered standards exceed some minimum acceptable level. Still, one could ask, with respect to sample size in particular: Have the standards become too low?
Let's revisit Pfizer's sample of 2,268 5 through 11 year olds. The company claims that the side effects they found are negligible, because they reflect the same patterns previously observed in 12 through 16 year olds. (Pfizer didn't provide the actual stats.) In other words, just like older children, a small number of 5 through 11 year olds react to Pfizer's vaccinate with fevers, headaches, fatigue, and/or other minor, transitory side effects – all acceptable, the argument goes, owing to the benefits of vaccination.
I have no quarrel with this argument. However... Pfizer's sample is small enough that it would be easy to miss serious side effects that happen to be more rare. I consider this a substantive concern.
Of course you could raise this concern, on general principle, about any new drug trial. You could always argue that a sample needs to be larger, in order to increase the chances of spotting some rare, hitherto unknown side effect. In this case, however, there are more specific causes for concern. Here's one:
Two cardiovascular side effects of COVID-19 vaccines are myocarditis (inflammation of the myocardium, or heart muscle) and pericarditis (inflammation of the pericardium, or tissue lining the heart). Both effects are typically mild and temporary, although a small percentage of cases are serious. The CDC has calculated that among 12 through 17 year olds, for every 1 million second doses of COVID-19 vaccine, roughly 76 cases of myocarditis or pericarditis will occur. (Other estimates are higher.) Unfortunately, Pfizer's sample is nowhere near large enough to reliably estimate the incidence of these side effects in 5 through 11 year olds.
To illustrate my point, imagine, for example, that 5 through 11 years olds were three times as likely as older children to experience myocarditis or pericarditis following a COVID-19 vaccine. That would be 228 cases in a million. That's still only about half of one case out of every 2,268 children. You might say that Pfizer has about a 50-50 chance of not identifying a single child with one of these side effects, even if the incidence of these effects is 3 times greater in this age group than among older children.
Are there precedents for side effects of drugs being greater among younger people? Absolutely. One of the most stark and tragic examples is thalidomide, which was prescribed to pregnant women in the late 1950s and early 1960s for nausea. Thalidomide caused serious birth defects if used early in pregnancy, but not if used strictly during the final months. In other words, if the effects of thalidomide on fetal development had been tested before prescribing the drug to pregnant mothers, and only women in their third trimesters had been sampled, researchers would've concluded (incorrectly) that thalidomide is safe for all unborn babies.
The FDA and other experts would doubtless respond to my concerns as follows:
1. There's no scientific reason to suspect that COVID-19 vaccines would affect younger children differently than they affect older children. (In contrast, we know enough now about the mechanisms by which thalidomide causes harm that we could predict effects early in pregnancy but not near the end.)
2. As with all vaccines, COVID-19 vaccines introduce small risks of adverse side effects, but the benefits strongly outweigh the risks. Even 228 instances of inflamed hearts out of a million fully vaccinated children would be a very small number.
I mostly agree with the reasoning here. Every vaccine that young Americans currently receive carries risks of adverse side effects. Some of those side effects are quite serious, but the extent of risk is consistently small, and clearly outweighed by the benefits.
This favorable risk-benefit ratio is nicely illustrated by polio vaccines. During the early 20th century, polio killed or permanently disabled tens of thousands of children in America and elsewhere. In the late 1940s and early 1950s, Americans experienced quarantines, travel restrictions, and warnings against densely crowded places. (Sound familiar?) Now, thanks to the IPV and OPV vaccines, polio has been eradicated in the U.S. Children who receive a vaccine are at risk of a serious allergic reaction, but such reactions occur in only about 1 in one million children.
The stats aren't quite as favorable for COVID-19 vaccines. Pfizer's vaccine won't completely eradicate the disease among 5 through 11 year olds (or any other age group), and the risk of serious side effects is likely to be somewhat higher than one in a million. (Plus, as I've said, the company is clearly in a hurry to sell more vaccines.) All the same, unless the FDA finds something unusual, I would recommend that all parents with 5 through 11 year olds get their children vaccinated as soon as possible, as a way of protecting their kids and contributing to the broader, multi-pronged strategy of reducing the incidence of COVID-19 in the U.S.
(See Appendix 2 for a final comment on the safety data – and a shout-out to one of the under-appreciated heroines of American pharmacology.)
Thanks for reading!
Appendix 1: When representativeness beats size
Here's a famous example of how the representativeness of a sample can be more important than its size:
From 1916 to 1932, the Literary Digest correctly predicted the winner of every U.S. presidential election. In 1936, the Literary Digest polled about 2.4 million people whose names had been drawn from telephone directories, magazine subscription lists, etc., and then predicted that Alf Landon would win by a landslide. In contrast, George Gallup (who was largely unknown at the time) and his colleagues sampled about 50,000 people and predicted that FDR would win by a landslide. Gallup and colleagues were right, and the Gallup Poll became a hit.
Why did the Literary Digest get it wrong, even though their sample was nearly 50 times larger than Gallup's? In a word, sampling bias. In 1936, telephones and magazine subscriptions were luxuries primarily owned by middle- and upper-class citizens, who were more likely to vote for the Republican candidate Alf Landon. Although Gallup's poll was much smaller, it was also much more representative of the U.S. population with respect to socioeconomic status. Including sufficient numbers of lower-income voters, who strongly preferred FDR, allowed Gallup and colleagues to correctly call the 1936 election.
In this instance, Gallup and colleagues wouldn't have benefited much from a larger sample (other than to have refined their prediction – FDR won 62% of the popular vote, 6% more than Gallup projected). However, it's clear that Pfizer's study – indeed, most studies on the development of new drugs – would benefit from larger samples, if only to increase the chances of identifying rare side effects.
Appendix 2: Frances Oldham, thalidomide, and COVID-19 vaccine EUA
I mentioned that it's possible in theory (though unlikely) that Pfizer's vaccine could be safe for children 12 and up yet unsafe for 5 through 11 year olds. By analogy, I noted that studies on the effects of thalidomide on third-trimester fetal development would've indicated no serious side effects, in spite of the fact that the drug is so harmful earlier on. Americans were largely spared the tragic effects of thalidomide, thanks to the vigilance of Frances Oldham (who, incidentally, was able to pursue graduate work in pharmacology at University of Chicago in 1936 because she wrote her future mentor a letter and he assumed that "Frances" was a man). Dr. Oldham went on to serve as a reviewer for the FDA, and in that capacity she withheld approval for thalidomide on multiple occasions between 1960 and 1962, in spite of the fact that it was already approved for use in more than 30 countries – and in the face of relentless pressure from the manufacturer. Her rationale? Not enough data on safety. Kudos to Dr. Oldham!
Unlike COVID-19 vaccines, there was never an emergency need for thalidomide. It was merely pitched as a more effective alternative to other drugs prescribed at the time for pregnancy-related nausea and other maladies. The pandemic, as I noted, has created an urgent need for vaccine development, and we have no specific reason to believe that the vaccines currently in use cause harm (except in a tiny percentage of cases). However, we should remember that all vaccines currently on the market, and all of the rules governing who receives them (and who’s eligible for a booster) reflect emergency use authorization (EUA) by the FDA. None of the vaccines have been approved through FDA's regular review process.
Although the FDA is careful to note that EUA is a rigorous process, it still represents relaxed standards, and the guidelines sound a bit squishy. For instance, EUA can be granted if it is "reasonable to believe" that a drug "may be" effective and safe.
I don't object to the squishiness of those terms, because it’s FDA pharmacologists and other experts rather than just administrators who decide whether it's "reasonable" to believe that a vaccine "may be" effective and safe. When you're not an expert, you may have to trust the experts. Still, as an informed lay person, I would always want assurance that EUA review is sufficiently rigorous given the nature of the emergency. For this reason, I'm curious to know whether the FDA will relax its initial request for 3,000 5 through 11 year olds and allow a full EUA application from Pfizer based on its smaller sample. Stay tuned…