Today I’ve got two COVID notes: vaccines for kids, and how to interpret data on share of deaths in vaccinated people.
News dive: Vaccines for kids
What’s going on with vaccines for under 5? At this point, the two key players are Moderna, and the FDA.
The FDA has received an application from Moderna for emergency use authorization for a two-dose COVID-19 vaccine for children under 6 (they submitted their final paperwork on May 9th). The agency expects, but has not yet received, a filing from Pfizer for its three-dose vaccine for this age group. As you may recall, back in February there was some suggestion that the FDA would consider a two-dose version of the Pfizer vaccine, but the agency decided against it at the last minute (see post here).
The next steps are for the FDA to review the filing information from Moderna and make a decision. There are three possible meeting dates in June (June 8 is the first one) at which this could be considered.
When we get closer to the FDA meeting we’ll see more details from the Moderna trial, but a recent press release provides some baseline information.
Moderna’s pediatric trial included 4,200 children ages 2 to 6 years and about 2,500 children ages six months to under 2 years. In terms of safety, the company reported no significant safety concerns and indicated that the “majority of adverse events were mild or moderate.” Fever was seen in 17% of the six-month-to-under-2-year age group and 14.6% of the 2-to-6 age group. This is lower than what researchers there had seen in the 6-to-11 group.
In terms of efficacy, Moderna reported an antibody response in kids similar in magnitude to what was seen in adults. These studies in children are designed as “immuno-bridging” studies: the primary outcome is a sufficient immune response (the fact that immune response = disease protection in adults is then used as the bridge). Symptomatic infection was a “secondary outcome” in the trial. The vaccine efficacy for this outcome in both groups is moderate: 44% in the youngest group, 37.5% in the 2-to-6-years group. If we limit to cases confirmed by PCR testing, this was 51% and 37%.
These numbers may seem low, but they are in line with what we are seeing in older populations in the wake of Omicron. In the trial population, there were no serious illnesses, hospitalizations, or death in either the vaccinated or unvaccinated group, so there is no way to measure whether the vaccine lowered the risk of these outcomes in his age group. Based on what we know about other age groups, we would expect them to, but the small overall risks in this group mean we cannot measure this directly.
Given this Moderna data, I have seen some people express concerns about whether the FDA will approve the vaccine. They’ve previously stated that 50% efficacy at preventing symptomatic infection was a requirement. There has also been some ongoing dicsussion of whether the FDA plans to wait for the Pfizer filing before reviewing Moderna.
We have generally had relatively little information about the FDA’s thinking, but on Monday, Representative James Clyburn released information from a briefing he got last week from the FDA (thank you, Rep. Clyburn). The briefing makes a few key points. First: the FDA representative (Dr. Peter Marks) said they would not wait for the Pfizer filing to review Moderna.
Second, he indicated that the agency would not hold the pediatric vaccine to the requirement that it be 50% effective at preventing symptomatic infection. This makes sense given that the current vaccine efficacy against symptomatic infection in older children and adults is below this, even as serious illness protection has been maintained.
(Dr. Marks also said they could move the meeting earlier than June if documents were reviewed in time, but this seems unlikely to me).
The remaining uncertainty (I think) is around the details of any safety issues in the trial. This is likely to be where the FDA focuses its attention. We will know more details closer to the meetings.
Data dive: Serious illness by vaccine status
A week or so ago, the Washington Post published an article that a number of people flagged for me. The basic message was that vaccinated and boosted people were becoming an increasing share of COVID deaths. The messages that I got approached panic: Did this mean the vaccines were not effective anymore? Does it mean boosters are useless? In fact, it means neither that vaccines are not effective nor that boosters are useless. It reflects expected features of the data, and it is likely to become more true over time.
The basic fact highlighted in the Post story is the following: In October 2021, 1% of deaths occurred in people with boosters, 25% in fully vaccinated people, and 74% in unvaccinated people. In February 2022, 15% of deaths occurred in boosted people, 25% in fully vaccinated, and 60% in unvaccinated. In other words, the share of deaths among vaccinated (and especially boosted) people went up a lot over this period.
This occurs for three reasons.
The first is a simple base-rate argument. Put aside the boosted people for a minute, and imagine that people are either vaccinated or not. Assuming vaccines are not perfectly protective against death, as the share of people who are vaccinated increases, they will also increase as the share of those who die. This is true even though vaccines are very protective. If 100% of people are fully vaccinated, then 100% of deaths would be in fully vaccinated people. (I did a slightly longer discussion of this back in September.) The Washington Post piece made this point by simply noting that the death rate among the unvaccinated was still seven times as high as in fully vaccinated people in March, despite their becoming a smaller share of deaths.
The second issue is slightly more subtle. People vary in their underlying vulnerability to COVID; the most notable predictor is age — older people are at much higher risk for serious illness and death. When vaccines, and then boosters, first became available, more-vulnerable people were given access first, for obvious reasons. These vaccines and boosters lowered the risks of serious impacts of COVID tremendously, but because of the enormous differences in underlying risks, these groups often still remained among the highest-risk. As I wrote about, with some controversy, a year ago: unvaccinated children are lower-risk than their vaccinated grandparents.
What this means is that vaccines early on, or boosters now, are a marker of high underlying risk. When we compare someone who has had four shots with someone who has had two, we are often comparing someone with a high underlying risk to someone with a low underlying risk. The boosters may be extremely protective, but the differences in underlying risk may mean that they do not look as effective as they are when we just compare boosted people with non-boosted people without accounting for other risk factors. This is a form of what’s called Simpson’s paradox, or omitted-variable bias.
A third issue: consistent exposure. At this point in the pandemic, the vast majority of the U.S. population has either had COVID, been vaccinated, or both. Both having had COVID and vaccination are protective against serious illness; hybrid immunity (vaccine + COVID) is likely the most effective. Taken together, this means that nearly everyone in the U.S., even if unvaccinated, has a layer of protection against serious illness and death.
What this means is that, increasingly, it will be the underlying COVID risk that dictates serious illness. As at the start of the pandemic, individual risk profile (age, obesity, diabetes, severely immunocompromised) will come to dominate over vaccination status. This is extremely similar to what we see with the flu or other more familiar illnesses. We all have some baseline layer of protection, and serious illness is largely a function of underlying vulnerability. This doesn’t mean vaccines are not effective — they are! — but baseline risk is also key.
There are two implications here. The first is in how we read the data. The pattern reported in the Washington Post article is expected and will continue. It will increasingly be the case that there are a larger share of deaths among vaccinated and boosted people, for all of the reasons above. (I will say: it is an astonishing comment on the quality of the vaccines that it is still the case as late as February that the majority of deaths are among the increasingly small share of unvaccinated people.)
The second implication is to underscore the need to invest in effective policies that protect vulnerable individuals from serious illness. This means, most importantly: ensuring access to boosters for high-risk individuals; making treatments (Paxlovid and Evusheld) widely available; continuing to invest in testing infrastructure to protect locations like nursing homes; ensuring that high-risk people have access to high-quality masks. These interventions can be key to preventing the worst effects of COVID — and other illnesses — in these groups.
Community Guidelines
