covid-19

Don’t Panic, But Breakthrough Cases May Be a Bigger Problem Than You’ve Been Told

Current public-health messaging may understate the scale and risk.

Photo: Mark Felix/Bloomberg via Getty Images

The term itself, perhaps, is a problem. “Breakthrough” sounds bad — implying an immune-escape mutation, likely rare, and therefore alarming.

The vaccines were never tested to prevent transmission, only symptomatic disease, and those who knew the science expected, from the outset, that we would see some number of such cases, and that they would be, overwhelmingly, mild. But Delta appears to have changed things. Not everything: The vaccines are working to suppress severe outcomes from COVID infection — according to a New York Times analysis, by more than a factor of 100 for some states, and at least fivefold for even the states where the effect has been most muted. That is, by the standards of historical vaccines, game-changingly well. But most of the data in that analysis comes from before the arrival of the Delta variant, and during the current surge there does seem to be considerably more “leakage” in the protection that vaccines offer against pandemic spread than has widely been acknowledged. While more severe breakthrough cases remain, in relative terms, very rare, we may be seeing a rise in those numbers with Delta, as well.

Over the last few weeks, in the wake of an attention-getting internal CDC presentation on the severity of the current wave, we’ve heard a lot — from epidemiologists, public-health officials, journalists like me — about how the leaked slides lacked context, implying a much scarier near-term future than was really suggested by the data, which showed that vaccines were working, that breakthrough cases remained rare and mild, that the pandemic was now largely a pandemic of the unvaccinated. On July 30, the Kaiser Family Foundation published a comprehensive-seeming report, much passed-around, which compiled partial breakthrough data from 24 states and the District of Columbia, and declared that the relative risk to the vaccinated of infection, hospitalization, and death was close to — or mathematically equal to — zero, and that in almost all states only about one percent of identified cases were breakthrough events. This reading was echoed by the later Times analysis, and itself echoed earlier reassuring statements by Anthony Fauci, that 99.2 percent of deaths in June were unvaccinated people, and by Surgeon General Vivek Murthy, that, as of July 18, “99.5 percent of COVID deaths are among the unvaccinated.”

These readings are, directionally, correct: The vaccines are performing admirably, particularly in protecting people from getting very sick. The current, bleak Delta wave is being driven primarily by cases in the unvaccinated, and the best tool in attacking the pandemic is more vaccination — a very powerful tool indeed. But nevertheless a closer look at the data reveals that some of the public-health communication may be overstating the vaccine effect on transmission and understating the scale and risk of breakthrough infections, which, while far from predominant, do appear prevalent enough to be helping shape the course of the disease.

“The message that breakthrough cases are exceedingly rare and that you don’t have to worry about them if you’re vaccinated — that this is only an epidemic of the unvaccinated — that message is falling flat,” Harvard epidemiologist Michael Mina told me in the long interview that follows below. “If this was still Alpha, sure. But with Delta, plenty of people are getting sick. Plenty of transmission is going on. And my personal opinion is that the whole notion of herd immunity from two vaccine shots is flying out the window very quickly with this new variant.”

“We’re seeing a lot more spread in vaccinated people,” agreed Scripps’s Eric Topol, who estimated that the vaccines’ efficacy against symptomatic transmission, which he estimated to be 90 percent or above for the wild-type strain and all previous variants, had fallen to about 60 percent for Delta. “That’s a big drop.” Later, he suggested it might have fallen to 50 percent, and that new data about to be published in the U.S. would suggest an even lower rate. On Wednesday, a large pre-print study published by the Mayo clinic suggested the efficacy against infection had fallen as far as 42 percent.

“The breakthrough problem is much more concerning than what our public officials have transmitted,” Topol continued. “We have no good tracking. But every indicator I have suggests that there’s a lot more under the radar than is being told to the public so far, which is unfortunate.” The result, he said, was a widening gap between the messaging from public-health authorities and the meaning of the data emerging in real time. “I think the problem we have is people — whether it’s the CDC or the people that are doing the briefings — their big concern is, they just want to get vaccinations up. And they don’t want to punch any holes in the story about vaccines. But we can handle the truth. And that’s what we should be getting.”

The central distortion reflected in the Kaiser report — and echoed by communicators elsewhere, including in the Times — is the result of a basic error of comparison, one that should have been obvious to anyone familiar with the shape of the pandemic. Almost all of these calculations about the share of breakthrough cases have been made using year-to-date 2021 data, which include several months before mass vaccination (when by definition vanishingly few breakthrough cases could have occurred) during which time the vast majority of the year’s total cases and deaths took place (during the winter surge). That means that to calculate a prevalence ratio for cases or deaths using the full year’s data requires you to effectively divide a numerator of four months of data by a denominator of seven months of data. And because those first few brutal months of the year were exceptional ones that do not reflect anything like the present state of vaccination or the disease, they throw off the ratios even further. Two-thirds of 2021 cases and 80 percent of deaths came before April 1, when only 15 percent of the country was fully vaccinated, which means calculating year-to-date ratios means possibly underestimating the prevalence of breakthrough cases by a factor of three and breakthrough deaths by a factor of five. And if the ratios are calculated using data sets that end before the Delta surge, as many have been, that adds an additional distortion, since both breakthrough cases and severe illness among the vaccinated appear to be significantly more common with this variant than with previous ones.

Unfortunately, more accurate month-to-month data is hard to assemble — because the CDC stopped tracking most breakthrough cases in early May, before the Delta wave had begun, and the states maintaining their own databases often update them irregularly and, in some cases, according to idiosyncratic logic — but over the last week, I’ve tried. And while several states show prevalence rates roughly in line with Kaiser’s ballpark one percent estimate (in Virginia, for instance, breakthroughs represent 2.3 percent of new cases and 5.2 percent of deaths), in others the patterns were divergent. In Delaware, between July 1 and July 22, “breakthrough” cases were 13.8 percent of the total. In Michigan, between June 15 and July 30, the figure was 19.1 percent. In Utah, 8 percent of new cases were breakthroughs in early June, but by late July, as Delta grew, the share grew, too, to 20 percent (even while the total number of cases almost doubled). According to those leaked CDC documents, there were, as of late last month, 35,000 symptomatic breakthrough cases being recorded each week — about 10 percent of the country’s total. Presumably many more breakthrough cases were asymptomatic, which would drive the share up further.

Of course, the share of cases isn’t the most illuminating metric — among other confounding factors, it reflects the relative rate of vaccination not just across the population in general but especially among the elderly who are most at risk. And this data is imperfect and incomplete and limited in various other ways, too, including because the country is testing so inadequately that we can’t really see the present state of the pandemic all that clearly. But the piecemeal data does begin to tell you something, suggesting that breakthrough cases represent a bigger share of disease spread, particularly in the ongoing Delta wave, than has been widely acknowledged — perhaps, overall, somewhere in the range of 5 to 20 percent of current cases, rather than the 0 to 5 percent range. When I ran these figures by Topol, he said, “I think the numbers are right on, and I think they’ve clearly been getting worse as Delta became fully dominant, now approaching 100 percent of all U.S. infections.”

The figures circulating about the relative risks faced by a vaccinated individual are more reliable than comparisons built around relative share of infections, because a risk calculation effectively controls for vaccination rates. But they also lack context without comparisons to the same data for the unvaccinated. The 125,000 known U.S. breakthrough cases through the end of July represent, according to NBC News, only 0.08 percent of the vaccinated population — meaning that less than one-tenth of one percent of all vaccinated people have gotten a confirmed breakthrough infection. That is encouraging, and impressive, and reassuring. But less than one percent of the unvaccinated got a confirmed case in July, too. While you can’t precisely compare the two groups — social behavior differences make that impossible, as does the age skew of the disease and the age skew of vaccination; and the timelines are different, as well — the two figures suggest, possibly, something like a tenfold reduction of risk. That’s extremely good! In California, for the week ending July 31, the average case rate among the unvaccinated, the state reported, was 33 per 100,000, while the average daily case rate among the vaccinated was seven per 100,000 — a fivefold decrease, more or less. That’s still very good! But fivefold or tenfold is not the hundredfold reduction implied by Kaiser, Fauci, or Murthy.

How big an issue are breakthrough cases? Well, cases do still matter: As tempting as it might have been, a few weeks ago, to believe that vaccines would have so well-protected the country’s most vulnerable that the collateral damage from any individual infection was really limited, the growing death totals from the Delta wave show there remain a tragically large number of vulnerable people in the country really threatened by pandemic spread. The day-to-day death totals are noisy right now, but on August 11, more than a thousand new deaths were reported — a level the country hadn’t seen since early April, when less than 20 percent of the country was fully vaccinated. The same day, the less-noisy seven-day average hit 608 — a level the country had last reached in mid-May, with barely a third of the country fully protected.

The vaccine effect is considerably more encouraging when it comes to the risk of severe disease and hospitalization on those 50 percent of the country — and 61 percent of adults — who have gotten the shots. According to Kaiser, the hospitalization rate for vaccinated people is, for most states, at or just below 0.01 percent — meaning one out of every 10,000 vaccinated people has been hospitalized. Over the past year, the hospitalization rate for the country as a whole is about 0.7 percent. That is an enormous difference in protection against hospitalization — about 70-fold. But Delta is likely changing things here, as well. Even in low-breakthrough Virginia, for instance, there were 17 breakthrough hospitalizations, out of 430 total — about 4 percent, implying about a 17-fold reduction. Still very impressive, but notably lower than the effect implied by pre-Delta data.

When the Times calculated the risk of COVID death among the vaccinated, it found that in Texas and Georgia, vaccination seemed to reduce the risk of death by 85x and 87x, respectively, but that in Maine, Vermont, and Indiana, it was only 7x. In Michigan it was 8x. Even at the low end, this is a dramatic, even miraculous-seeming impact. But, of course, most of the data was collected before the Delta surge, and the variant may be darkening the picture at least somewhat here, too. The figures cited by Fauci and Murthy look reassuring — 99.2 percent of deaths were among the unvaccinated in June, according to Fauci, and 99.5 percent according to Murthy, speaking in July. But between June 25 and July 26, 6,973 Americans died in total from COVID-19, according to the CDC. Also according to the CDC, during that same period (June 25 to July 26), 364 died of breakthrough infections. That’s more than 5 percent — more than six times Fauci’s estimate and ten times Murthy’s.

Last week I spoke about all this with Harvard’s Michael Mina.

Over the last couple of weeks, and especially since those CDC slides were leaked, a lot of scientists and epidemiologists and data wonks have seemed to want to reassure the public — to emphasize that the vaccines really are working, that breakthrough cases really are rare, that this is now a pandemic of the unvaccinated and that even Delta isn’t very worrisome for the vaccinated. You seem considerably more concerned. Why is that?
I think what we’re seeing again is the same old story that we’ve seen this entire pandemic.

What do you mean? 
Many people have tried to stay in line with the official public-health message — whether that’s because they don’t have enough confidence in their own understanding of how the virus works or if it’s because they want to stay consistent with the public-health message of the country, I don’t know. It’s never been clear to me why. I think everyone has their own reasons for not being willing to be a little bit more outspoken during this, particularly given that the official message has often been weeks or months behind the virus. But I think the message that breakthrough cases are exceedingly rare and that you don’t have to worry about them if you’re vaccinated — that this is only an epidemic of the unvaccinated — that message is falling flat. If this was still Alpha, sure. But with Delta, plenty of people are getting sick. Plenty of transmission is going on. And my personal opinion is that the whole notion of herd immunity from two vaccine doses is flying out the window very quickly with this new variant. And it’s probably going to fly out the window even more quickly with the variants to come.

And that’s because you see the vaccinated continuing to spread the disease, is that right? That would mean that we can’t count on vaccines to ultimately stop spread, and the thresholds of immunity we might’ve imagined, a few months ago, would get us all the way on the other side of the pandemic probably wouldn’t actually bring an end to things, now that Delta is here. Is that right?
Yes. And that’s exactly right. Breakthrough cases are real. It’s no longer what we were talking about when we talked about the Yankees outbreak.

That was earlier in the summer, when you were saying we shouldn’t worry about breakthrough cases, in fact maybe we should celebrate them, because they were overwhelmingly asymptomatic and offered a kind of free immunity booster.
Those cases were PCR positive but likely very low viral load — that was the point of that discussion. But now we’re actually seeing lots of people getting symptomatic disease — getting fevers, getting sick. I just found out today that my brother has a breakthrough case — he’s fully vaccinated and young and healthy and he is very sick right now. And the data is suggesting more and more that that’s not rare at this point.

I’m sorry to hear that, I hope he’s managing and recovering.
I actually haven’t even talked to him yet, my phone’s just been blowing up about it.

But when we discussed the Yankees cluster, I think I said that my definition of a breakthrough case was something that was actually breaking through what the vaccines were intended to do and tested for. And since they weren’t intended to protect against transmission, or tested to measure their efficacy in stopping transmission, or, in that case, low-virus-load PCR positivity, if all the cases in question were asymptomatic and just PCR positive, I thought they didn’t really count as true breakthroughs. But now we’re starting to see that these cases are really causing illness in vaccinated people. And I think we’re probably under-diagnosing, or under-reporting a lot of it. I think breakthrough cases are probably very high, relative to what we’re detecting at the moment. I think it is much more serious than the CDC has been until very recently willing to suggest. And that’s because they wanted to stay on message.

Why?
They don’t want to panic people. You know, every step they’ve ever taken in this pandemic has been about not causing too much concern in the public. But this is a pandemic. You can’t control things that way. You have to just be honest.

And I think what we’re seeing is that viral loads can get very high with Delta. They could actually get high with some of the other variants too, actually. It’s worth mentioning, We are seeing some new data that shows that clearance of the virus from the body is faster, too —

So the infectiousness ends more quickly, in other words.
Well, that’s how it’s being interpreted. But at the end of the day, I’d be willing to bet that almost all transmission happens when people are at their peak viral load. And so if peak viral loads are not differing between the vaccinated and the unvaccinated, but the slope downward is maybe faster in the vaccinated, I think we’re pretty much going to see unabated transmission amongst vaccinated. It may not be as much as it would have been if nobody had a vaccine, but I think the idea that we’re going to vaccinate our way to true herd immunity — that idea has to be put to bed for a moment. Until we get a more stable sort of immunity and a more stable set of mutations, we should anticipate that anyone who’s been vaccinated can still transmit.

I’ve heard a fair amount of pushback on these studies showing a roughly equivalent viral load between the vaccinated and the unvaccinated. One of the main points of criticism is that, because vaccines reduce symptomaticity, you may be drawing from the most severe breakthrough cases and comparing them to a much broader poll of unvaccinated cases. What do you make of that?
That could be the case. But my lab has probably done as much as anyone in the world at this point to really understand all of the confounding factors with viral loads. Now, I’m no oracle here. But what I do well is I look at, what should we expect from the immune system? What should we expect from viral kinetics and how would that play out in the population data? And if your peak viral load is at a billion or a trillion copies per milliliter, you’re going to transmit.

But couldn’t we be seeing just the most severe cases of breakthroughs and comparing them to less severe cases among the unvaccinated, because of a sampling bias?
Possibly. But to your point, we haven’t done a very good job at testing the unvaccinated, either. Really the only people who go get tested today at a PCR lab are those who are feeling ill.

Right.
So I don’t think we’ve done such exhaustive testing at this point in time that we would really be seeing a major difference in the unvaccinated versus the vaccinated populations.

Now, I could be wrong on that if people were really following CDC guidance, which is essentially “don’t get tested if you’ve been vaccinated.” But I don’t think we’re doing a good enough job of testing the asymptomatic unvaccinated to really make that claim. These days, and around much of the world, PCR testing tends to skew toward the symptomatic, likely making the comparisons slightly more comparable.

We’ll get to the higher concern categories — symptomatic disease, severe disease, hospitalization, and death. But just to stay with cases for a minute, you said a minute ago that you expected transmission rates to be comparable. That assumes an infection to begin with, right? What about that initial infection? How effective do you think the vaccines are there? 
That’s the place where the argument does break down, in terms of the numbers game — vaccines may mean that there may be a lot of people who just are more resistant to getting any infection. I would say, tentatively, that, if we had a crystal ball, and we could say there are X number of people who have been exposed, then the probability of an exposure turning into an infection is, as an example, 30 percent lower amongst vaccinated.

We haven’t seen the data to really show it because it’s extremely difficult to measure. I’d say that there’s a very good chance that vaccines will help prevent an exposure turning into an infection. But amongst those who are getting infected and detected, the viral loads are really high. And we’re clearly seeing transmission happen. We’re seeing outbreaks happen amongst vaccinated groups.

So when the CDC says in their presentation that vaccines reduce “disease incidence” eightfold — it sounds like you’re saying the effect is considerably smaller. 
Well, we have to look at what they’re saying. They’re talking about real sickness, not just a transmission chain event. I think it’s plausible that the vaccines do have an eightfold effect on an exposure turning into a clinically relevant case. But in terms of an exposure turning into a transmission event, I don’t think anyone has that data, but I think an eightfold reduction would probably be very generous.

Because of all the asymptomatic transmission, in other words. What do you make of their ballpark estimates for hospitalization and death, that vaccines reduce the risks there roughly 25-fold?
I do think they are serving to protect people from hospitalization. Absolutely. But I also do think that we need to consider this a very dynamic process. For a year, really, I’ve been saying that we need to be very cognizant that the way that we’re measuring these vaccines has been, still to this day, within the first month post vaccination.

Our most elderly individuals, they started getting vaccinated in February or so. People are now six months out from vaccination. And I think as we start to get further and further out, we’re going to really see those numbers change, and we’ll see that elderly people will become susceptible again.

So far, certainly, I think the CDC are being honest with their data. Most people who are being hospitalized are still the unvaccinated. But I do think though that we will start to see outbreaks, hospitalizations, and deaths happen in nursing homes and senior living facilities amongst vaccinated people.

And that’s because immunity wanes more quickly in older people?
Yes. Our bodies have an immune age, too. And when an elderly person starts to get nine months post-vaccine, you really start losing the benefits, unless you get a booster.

How likely is that, do you think?
I can almost guarantee that our senior citizens will get boosters in the U.S. in the next two months, I think the CDC is going to recommend it — there’s no way they’re not. I think the only reason we haven’t announced it already is that the CDC doesn’t want to concern people. The moment they admit that they’re going to boost people, it sends a message, like, are we in vaccination purgatory here?

But the writing’s on the wall. Senior citizens, as they age, don’t have the cells to retain immunological memory very well. Now, if you keep getting vaccinated over five, six years, maybe they can start to build up a much greater cushion of protection. But so far these senior citizens and elderly people in our country and around the world have only been vaccinated twice. Without boosters, I think we should expect that their immune memory will start to wane by the fall.

What would that look like?
I project that it could be kind of a perfect storm. In some areas, we’ll see that the mitigation strategies that we put in place are going to fend off Delta a little bit, especially in the Northeast. But it’s going to fend it off just enough so that we hold it back until September, October.

And then we’d probably going to see a lot of breakthrough cases in nursing homes. And we’ll probably see these outbreaks in nursing homes turn into hospitalizations and deaths by the end of the year, absent boosters.

It sounds like you take seriously that much-debated data from Israel showing somewhat dramatic declines in vaccine efficacy, over time.
I do.

How does that play out among the young and middle-aged?
You know, the middle-aged and the young — they’re already very rarely susceptible to severe disease now. Regardless of what the headlines say, it’s still pretty rare to see very severe disease in the middle-aged and young.

So even if we see a significant waning of immunity there, I think the young and the middle-aged still have plenty of ability to retain an immune response. In that sense, I’m not nearly as concerned for the younger generations as I am for the elderly.

That’s not really a surprise, given what we know about the age skew of the disease and what we know about the immune systems of the old. 
We know that people start to lose the physical architecture of their immunological memory development system as they age. It’s why we get shingles. It’s why we get all kinds of infectious diseases as we get older. It’s why we get cancer, in large part, because our immune system no longer surveys the way it should.

So I think when we hit eight or nine months post-vaccination, we should expect among the elderly — especially those over 70, 75 years old — a really significant waning of immunological memory.

How significant? Enough to prevent a successful immune response for many of them? 
Well, probably a large fraction would still be able to mount a successful immune response. With almost no data, if I had to put a number to it, I might predict that roughly a third of the people above 70 who are vaccinated might no longer be particularly well-protected. Half might still get symptoms.

And the other half might do just fine.
Something like that. That other half will probably still be able to participate in a transmission, but probably won’t be nearly as susceptible themselves. So we’ll still see a much lower degree of severe disease and mortality in terms of the rates of those people who get infected.

Hearing you talk about the fall, it doesn’t sound like you’re made all that optimistic by the British or Dutch experiences, where the Delta has already crashed, and pretty quickly. 
Well, we’re just a very heterogeneous and massive landmass. And so if we look at individual cities or states — Florida will get through this relatively quickly, because it’s burning through the state so quickly already. But by the time it gets to the Northeast and really takes over here, I think we’ll be pushing into late September, and we still have a lot of cases happening.

And at that point, we’re going to start seeing seasonality really take hold. So I think it will be very similar in fact to what we’ve seen across the world and across this country before — multiple waves, spatially segregated. That’s because we’re a large country with different kinds of attack for this virus and different strategies to control it. And so for the U.S. as a whole that just ends up creating a much longer duration wave than what we see in a smaller country like the U.K., for example.

And when you compound the seasonal effect with immunological waning and even immune escape, from Delta and what will inevitably become the next variant …

How worried are you about that?
The next variants will be worse, because Delta is now the baseline. All the lineages will derive from Delta. And when you compound the two and you say, okay, you have immunological waning at the same time that you have a new variant, then all of a sudden, we’re in a much more dire situation with the elderly, where people who think that they’re protected, just won’t be.

It’s not going to be everyone. There’s going to be plenty of people who have a good “immune age” — despite being 85 years old, your immunological clock might still be doing just fine. But there’s going to be a lot of people, maybe as much as a third, who are really not protected anymore. And I think we have to come to grips with that and we’re going to have to figure out how to deal with it.

Hearing you talk about the fall, it sounds pretty bad: Herd immunity is out the window, you say, so the disease can’t be eliminated. But there are also going to be these flare-ups among the vulnerable elderly, which make it somewhat more threatening than the other endemic diseases people often talk about when contemplating this endgame. It won’t threaten at the scale it did in 2020, but also won’t just be the flu, and we’ll be managing it in a sort of ongoing way without a real expiration date in mind.
I think that’s basically right, at least until we see this virus find a nice equilibrium where it is kind of evolutionarily happy to sit where it’s at and not really evolve much more — until there isn’t really a reason to continue evolving. But who knows when it will run out of tricks. And until we really see it run out of tricks, we have to be concerned that it will keep creating these new baselines.

As Delta is now our baseline.
I think we can’t take our eye off of that reality until the virus figures out a nice valley where it’s energetically stable, it has enough transmission to keep yourself going, and it doesn’t really have any more tricks up its sleeve to really break through any more. But that might be a year or two away still. So this might continue happening. I do think, though, that the more exposures people have will help.

That was your message when we were talking about the Yankees. 
I’ve always said that we’re going to age out of this virus. People are going to keep getting exposed. And whether it’s to Delta or to a variant in five months from now, every time you or I or anyone else gets exposed — they’re really building up a decent cushion of immunity with each of those exposures. So it’s only a matter of time before we actually have not only vaccine-derived immunity but natural infection-derived immunity, too.

When you start coupling all that together, you can picture it kind of like a sandwich, just continuing to stack up. Then we can start to say, okay, now, you know, even if the virus changes a bit, I’ve built up so much protection already. I’ve got all these antibodies that not only recognize the spike proteins from the virus, which is what we see with the vaccines, but I’ve now been exposed three times. And so even if the virus mutates this part, I already recognize these other parts. That’s kind of how we’re going to get out of this. That’s how our immune system learns. It’s going to be maybe a couple-year-long endeavor still, and everyone who’s been vaccinated probably will get exposed — eventually, hopefully, without causing any symptoms. I think we’ll get there, but it’s going to be a longer road than we’d like to believe.

This interview has been edited and condensed.