COVID-19 Updates: Part 2
There's good news to share – I'll get to it shortly – but I want to start with two mortality stats that got a lot of attention this week: The number of COVID-19 deaths in the U.S. reached 800,000, and the percentage of Americans over 65 who've died from COVID-19 reached 1%.
These are painful stats, no question. But I have mixed feelings about how to think about them.
On the one hand, there's nothing special about round numbers. From an individual perspective, the 800,000th death is just as tragic as the one that preceded it. From a public health perspective, one additional mortality changes nothing: We’re still mired in a public health crisis.
On the other hand, round numbers are salient. People notice. We're more likely to change, or to press for change, when we see them. On Tuesday, for example, the president released a "Statement by President Joe Biden on 800,000 American Deaths from COVID-19" in which he used this stat as an opportunity to once again urge Americans to get vaccines or boosters. Although I wish the U.S. wouldn't reach the one-million milestone, once we do, that number will turn heads and hopefully prompt safer policies and behaviors.
To the statistically-minded, the most salient numbers usually aren't round numbers, but rather the messier ones that tell more meaningful stories. For instance, consider these mortality stats from a Reuters analysis this week:
—The COVID-19 mortality rate in the U.S. increased from 600,000 to 700,000 in 111 days. The increase from 700,000 to 800,000 took 73 days. Sadly, these stats suggest that we'll reach the one-million mark before Easter.
—The per-capita COVID-19 mortality rate in the U.S. is higher than for any other G7 country. We rank 30th out of 38 OECD countries in this respect. In other words, the richest country in the world is not doing well at preventing the most severe COVID-19 outcome.
The remainder of this newsletter focuses on other news, mostly good, that emerged this week.
Booster rates
More than a million Americans per day continue to get boosters, prompted by fears about omicron as well as a desire for safer holiday travel. I'm encouraged by this trend, given that boosters increase protection against both delta and omicron (more on the latter shortly).
Vaccine donations
Another round number that appeared this week, though it didn't get much media attention, is that the U.S. has now donated over 300 million vaccines to other countries. This is good news, not just for the inhabitants of those 100+ countries, but also for us, because reducing the infection rate worldwide reduces the chances of some future mutation making its way back to the U.S. and contributing to breakthrough infections. (Some risk models predict that fully-vaccinated, healthy adults would actually be safer in the long run if, instead of getting boosters, their vaccines were administered to unvaccinated people worldwide. That's not a very useful prediction, because (a) we don't have to choose between getting boosted and donating boosters, and (b) boosting is clearly desirable, but the point is clear: sharing vaccines is good for everyone.)
Omicron
The CDC announced Tuesday that omicron is now responsible for about 3% of new cases in the U.S. Consistent with findings from last week, several studies reported this week that omicron spreads more rapidly than earlier variants did, and leads to higher breakthrough and reinfection rates, but it also produces less severe symptoms.
Each of these studies is too small in itself to mean much. For example, the CDC reported on December 10 that the first omicron cases in the U.S. reflected high breakthrough and reinfection rates, as well as relatively mild outcomes. This finding got a lot of press, but the sample size was only 43, which is too small for even the simplest statistical comparisons.
Although the individual studies are small, the extent to which their findings converge strengthens the evidence they provide. Analogous to the rationale underlying a meta-analysis, we can assume that as small studies continue to show that omicron is more transmissible but less dangerous, we can be more confident that this is an accurate characterization of the variant.
Vaccines and omicron
A South African study, described at a press conference two days ago, looked at over 211,000 COVID-19 test results obtained up to three weeks after vaccination. The main findings were that a primary course of the Pfizer vaccine (two doses, no booster) is 33% effective at preventing omicron infections, and 70% effective at preventing hospitalizations following infection. Even among the oldest participants (70 to 79), effectiveness against hospitalization remained high (60%).
To put these numbers in context, Pfizer's effectiveness against delta is around 93% for hospitalizations, and roughly 80% for infections. For omicron, those numbers drop to 70% and 33%, respectively.
The good news here is that 70% effectiveness is a relatively high rate of protection against hospitalizations from a variant that seems to produce milder symptoms and fewer hospitalizations to begin with. And, there's reason to believe that a booster would raise that 70% figure considerably. Pfizer presented serological data last week suggesting that among people who receive their booster, the immune response to omicron is roughly the same as what people who received two doses show when exposed to delta. In other words, as I mentioned last week, we can infer that three doses of Pfizer may turn out to be roughly 80% effective at preventing omicron-driven hospitalizations.
The serological data also soften the blow of that 33% figure, because it suggests that a third dose of Pfizer may increase the effectiveness against infections back up to around 80%. (In contrast, routine flu shots are estimated to be 40% to 60% effective.)
Two days ago, Moderna reported similar results from a small serological study, as did several scientists at a World Health Organization meeting yesterday. Here again the studies are small and based on serological data, but the consistency of the findings is meaningful, and the message for you and me is clear: Get a booster if you haven't had one yet!
New treatments
Remdesivir, the first COVID-19 drug, received emergency use authorization for hospitalized patients in October 2020, but the drug is only designed for severe cases, it's expensive, it has to be administered intravenously, and its effectiveness is modest at best. These limitations spurred Merck to develop molnupiravir, a pill that's intended to prevent mild cases of COVID-19 from developing into severe ones. An FDA panel recommended molnupiravir for emergency use on November 30 of this year. Unfortunately, the data Merck submitted to the FDA shows a reduction in risk for severe outcomes (hospitalization and death) of only 30%. This is better than nothing, but nowhere near as good as one would like, and the FDA panel was certainly not too enthusiastic (only 13 out of 23 members voted for approval). As of today, the FDA still hasn't announced whether it will give final approval – a noticeable delay during an otherwise rapid (or rushed) process which hints at internal controversies around the drug. (See Appendix 1 for additional details.)
A third drug, paxlovid, has now shown more promising results. On Tuesday, Pfizer released a report on its use with adults at high risk of developing severe COVID-19 infections (due primarily to pre-existing health conditions). In Pfizer's study, people who tested positive for COVID-19 were randomly assigned to take paxlovid or a placebo. 44 of the 682 patients who received a placebo were later hospitalized (6.5%), as compared to only 5 of the 697 patients who received paxlovid (0.7%). Overall, the risk of severe outcomes (hospitalization or death) was reduced by 88% to 89% when paxlovid was administered within 3 to 5 days after patients first reported symptoms. Apparent side effects were mild and not more frequent than observed in the placebo group. (Pfizer also reported a 70% reduction in risk of severe outcomes among a separate sample of patients not at risk, as well as serological evidence that the drug works well against omicron.)
Paxlovid is currently under FDA review for emergency authorization approval; many experts anticipate approval before the end of the year, with availability to the public by January. This should turn out to be good news for COVID-19 patients – and for our overburdened health care system. (Paxlovid may even reduce patients' infectiousness, a possibility that Pfizer is now studying.) The only downside seems to be that a prescription would be required. This may pose a challenge for the roughly 25% of Americans who don't have a primary care physician, and for the many people who rely on home tests for COVID-19, because they will need to get prescriptions from clinics or PCPs in time to start taking the drug within 5 days of developing their first symptoms.
Bottom line
When I talk with people about COVID-19 stats, the conversation often turns to personal risk. People want to know: What are the chances of me (and those I care about) becoming sick?
As I’ve mentioned before, once you've been fully vaccinated and boosted, it's impossible to calculate personal risk with much precision, because your behavior, age, health status, occupation, place of residence, and luck all influence the calculation in ways that are impossible to completely quantify. But there is some small good news to be gleaned from the stats. To illustrate, I’ll use one of my favorite analogies.
Over the past decade, about 4% of residents in my city (Cambridge) have been the victim of a crime. However, my personal risk in the upcoming decade is probably less than 4 in 100, because I don't live in a high-crime neighborhood, and I don't engage in a lot of the behaviors that increase risk of victimization (e.g., buying or selling drugs, walking around unsafe neighborhoods at night, etc). Likewise, my personal risk of getting COVID-19 is probably less than what city-wide infection rates might suggest, because I'm not among the folks who, by choice or by profession, engage in the riskiest behaviors. (For more on personal behavior, see Appendix 2.)
From an individual perspective, this should be good news for you. Stats on COVID-19 infections, even among those who are vaccinated and boosted, are drawn from samples that contain at least some people who are spending time, unmasked, at crowded bars and restaurants, parties, conferences, etc. If you're not one of those people, the next stat you see on infection rates may (happily for you) overestimate your risk.
Appendix 1: Molnupiravir’s trial(s)
Molnupiravir – a pill taken for five days to prevent mild COVID-19 cases from becoming severe – was supposed to be a game-changer. Then the final data turned up.
The FDA panel that reviewed the data (and recommended approval by the slender margin of 13 to 10), was not just concerned that the drug only reduced the risk of severe outcomes by 30%. Panelists also noted the following:
—Molnupiravir actually performed worse than a placebo during the second half of the trial. Merck offered no explanation for this pattern. In Merck's defense, you could argue that the 30% statistic is an average calculated across the entire trial, and therefore more meaningful than the average for one particular time period. Still...you wouldn't want a drug to be outperformed by a placebo for any substantial length of time. In this instance, the concern is that molnupiravir's stronger performance during the first half of the trial was more or less a fluke.
—Sample size was small. The initial sample consisted of 775 people who were then divided into molnupiravir and placebo groups. Merck would say, in response to a statistician's concerns, that the sample size was large enough to adequately power two-group comparisons, but the statistician would respond that this isn't enough to properly examine covariates like age and health status, and it might not be large enough to identify dangerous but relatively uncommon side effects.
—The trial data that the FDA panel reviewed wasn't Merck's first molnupiravir trial; an earlier trial had been stopped due to lack of benefit. Some of those on the FDA panel who voted against approval pointed out (rightly, in my opinion), that the results of both trials should be considered, rather than just focusing on the one that turned out well.
—Several experts on the panel raised safety concerns ranging from potential toxicity to embryos and fetuses, to the possibility of harmful mutations in the spike protein of the virus. Although no data directly indicated harm, these possibilities were serious enough to cause some panelists to vote against approval.
Quite a litany of concerns! In a future newsletter, I will talk more about the FDA panel's review process, allegations of conflict of interest, and how cases like this inspire conspiracy theories.
Appendix 2: My vaccination story
After discussing vaccines in several newsletters, some of you asked about my personal choices, and to what extent they were statistically-influenced.
To be honest, my initial vaccines were Pfizer because I was in a hurry to get vaccinated and that was the only available option at the site where I was vaccinated. So, there was nothing statistical or even particularly sensible about my taking Pfizer.
As for the booster, there’s a little more to say. On the day of my appointment, I came to a large vaccine site operated by a local hospital, and the person who checked me in asked which kind I wanted. Next to her were three roped off lines, one leading to Pfizer, one to Moderna, one to Janssen. I paused. "Let me think about this for a second", I said.
Inwardly I laughed at myself. There I was, the stats guy, supposedly up on the latest data, and I wasn't sure what to do. Should I stick with Pfizer, or mix it up?
I hesitated because there haven't yet been large, controlled studies on the effectiveness of mix-and-match strategies. Small studies suggest that mRNA boosters (Pfizer and Moderna) outperform Janssen, and a few studies also suggest a slight advantage for mixing.
In theory, then, I should've chosen Moderna, but I didn't. I went with Pfizer. My reasoning was statistical, but not based on specific statistics.
1. Janssen has consistently performed less well than the other two, both as a primary course and as a booster. Janssen’s performance isn’t so weak that you should worry if it was your vaccine – you’ve still gotten substantial protection – but since I needed to make a decision, I narrowed my focus to the two mRNA options.
2. The advantages of switching vaccines for the booster are small (assuming you can trust the results of a few small-sample studies). And, these advantages have only been seen in analyses of antibody levels. We expect some correspondence between serological data and real-world effectiveness, but the correspondence won't be perfect. Personal behavior, for example, has a much stronger impact on personal safety.
3. Homologous boosters are still effective. By sticking with Pfizer I would still get an immune response consistent with about 80% effectiveness against infection from delta or omicron (assuming you can trust the studies). Again, whether or not I get sick was going to depend more on my behavioral choices...
4. I already knew that I didn't have appreciable side effects from my first two Pfizer shots.
In short, instead of basing my decision on small differences observed in small studies that generalize imperfectly to real-word conditions, I relied on the only data I was sure of:
(a) I had no prior side effects with Pfizer (though this doesn't guarantee I wouldn’t have side effects after the booster).
(b) My future behavior would influence my safety more than my choice of booster (though personal behavior doesn't completely guarantee safety).
(That may have been more information than you were looking for, but thanks for asking!)