Reinfection and Long COVID

My mind, like many of ours, has been on worries other than COVID of late. But it’s also clear COVID is still very present, and for the first time in a while I found myself spiraling a bit on what to do about gathering with friends, mask-wearing, and indoor dining. Part of my challenge, I’ve realized, is that I’m struggling with two aspects of the data. One is the risk of reinfection. I’ve had three shots and I’ve had COVID. And yet it does seem like people in that same boat are getting COVID again. How often? The second aspect is long COVID, discussed everywhere but something I’ve struggled to get my head around. Today I’m going to offer — and sorry in advance that this is so long —- some data on both. The spoiler is that we do not have ideal evidence on either question so I’ll try to surface what we do have. I’m not going to offer much beyond the data today; I’ll try to be back in the next few weeks with some follow-up interpretation for paid subscribers. The long COVID discussion is especially…long. I do urge you to read it all, but I’ll surface a summary paragraph: My sense, putting this all together, is that persistent symptoms of fatigue, loss of smell, and coughing happen to a reasonable proportion of adults post-COVID, even after mild breakthrough infections. In some cases, these take months to resolve. But my sense is also that things like the CDC's claims that COVID causes a dramatically elevated risk for pulmonary embolism and substance abuse disorder — the most significant disability and disease risks that they attribute to long COVID — are likely overstated by the lack of a reliable set of control individuals. This is triangulation, though. It’s informed triangulation, but others may well triangulate differently. Reinfection How often will we get COVID? We do not know for sure. What we do know is that COVID reinfection is possible and not uncommon over longer time scales. We also know that prior infection, vaccination, and (especially) both contribute significantly to protection against serious illness. This remains true over long time scales. But for someone like me with three shots and who had COVID in January, what’s the likelihood of getting it again — and when? Reinfection is hard to measure reliably. It is getting harder because testing is getting less common and PCR testing even less so. Even if people rapid test at home, this may not get into an official system. A tremendously unscientific Instagram poll I ran the other day had 7% of about 20,000 people reporting having had COVID more than once since December. But who knows what that means and I didn’t ask the question well. The most compelling (non-Instagram produced) data we have on this is from the UK through their ONS surveillance system. In their most recent data, they provide estimates of daily reinfection rates in the post-Omicron period. The way their data works is they effectively assume the first three months after infection are protected; they note this is somewhat arbitrary. It is possible to be reinfected within that window, but it looks unlikely (putting aside the Paxlovid rebound issue). They then provide estimates of the daily rate of reinfection after that 90 days, based on the data they see and models they’ve developed. In the graph below, I converted these to monthly risks of reinfection each month post-infection. Because the ONS assumes the first three months are not “at risk,” I’ve put the first three months at zero. This suggests a slightly less than 1% on average chance of reinfection each month, so perhaps a 5% risk of reinfection over six months. In their data, this is higher among unvaccinated people and those with an asymptomatic first infection. To be honest, based on what we’re seeing in the world, these numbers overall seem low (although they are considerably higher than in earlier waves). It would be helpful, of course, to get more data on this going forward. In principle, survey data may be feasible if people are testing reliably. In reality, I’m not sure how much better we’ll do than the ongoing UK data. So thanks, UK, for picking up the slack. Long COVID This is an enormous topic, and I’m going to focus on the chaos of what we know about just one little piece of it — overall long COVID prevalence. Even still, this is not close to a comprehensive review. Interestingly, although in general our evidence of COVID in children is worse than our understanding in adults, this is one area where we actually have slightly better data on kids. An example is one study that compares long-term symptoms of children who tested positive for COVID to those who tested negative in surveillance testing. They found extremely small differences in long-term symptoms, and any symptoms were mild. In general, studies of children that use appropriate control groups do not show a significant long COVID burden in this group. When we turn to adults, it’s clear that long COVID is a more significant issue. But the estimates of the prevalence and the disease burden vary widely, for several reasons. First, there is no standard definition of “long COVID”; definitions differ in what symptoms they include and in how long is “long.” If you define long COVID as any symptoms a month after infection, the rates will be higher than if you require three months. The more symptoms are included, the more likely it is that people have them. When we talk about long COVID, people sometimes imagine this only refers to debilitating cardiac and lung issues that persist for months and months, but a persistent sore throat for six weeks post-infection would also be part of this group. Second, a large share of the studies we have on long COVID rely on selected samples, and most of them do not have controls. To be clear: such studies can be useful in helping us map out the picture of long COVID symptoms. But they aren’t as useful for evaluating prevalence. As an example, I found a quote from one review article stating: “In a very large study (2001 outpatients and 112 inpatients), only 0.7% of the population was symptom-free approximately 3 months after the initial infection.” That’s really scary, but in looking at the original article, we see this data is based on a sample recruited from a Facebook group for individuals who are suffering from persistent COVID symptoms. This is not useful for evaluating population prevalence. A third issue is that even in large studies which do have a comparison group, it can be difficult to avoid the problem that those who get COVID may be different before COVID than others. An example of this is a new CDC study, which suggests approximately 20% of people 18 to 64 who have tested positive for COVID then develop long COVID, and 25% of those over 65. This study uses electronic medical record data and compares the rate of medical problems for individuals who were diagnosed (either inpatient or outpatient) with COVID to a comparison group who were not. The time period is 30 days to 365 days post-diagnosis, so this is intended to capture the long COVID period. The study finds a higher rate of many illnesses in the COVID group: pulmonary embolism, substance abuse disorders, fatigue, heart failure, type 2 diabetes, and so on. Overall, 16% of the control group and 38% of the COVID group had a medical issue sometime in the year post-COVID, and the difference between these groups leads to the one in five long COVID headline numbers. What makes this a bit hard to interpret is the authors are not able to adjust for any other differences across groups — including fairly basic things like gender, race, socioeconomic status, and other preexisting conditions. Since we know that some groups have been more vulnerable to symptomatic and serious COVID than others and that those characteristics are also associated with higher risks for other illnesses, it’s difficult to know how much of this to attribute to COVID. Most notably, obesity is associated with COVID and also with many of the long COVID symptoms identified. This isn’t to say that there isn’t significant evidence of long COVID in this study — notably, a persistent loss of taste and smell seems clearly COVID-associated — but just that the numbers are challenging to interpret. A better version of this approach is this study, from September 2021, which looks for evidence of excess disease among people who had COVID relative to matched controls and compares this to the same approach for influenza. The authors find an excess long-term illness risk of 16 percentage points. This study has a lot to recommend it, although it is from an earlier period of the pandemic. This raises the final issue: we have little sense of how long COVID intersects with vaccination or with COVID variants. There is a general sense that long COVID is less common with post-vaccine breakthrough infections, but we do not have precise information. Recently, a study based on Veterans Affairs data showed that individuals with breakthrough COVID infections were at a greater risk for long-term health issues than controls; the elevated risk was small but still present for those who were not hospitalized. But, again, it’s hard to separate COVID impacts from the other factors that raise the risk of having serious COVID in the first place. Adding to all this complication, last week we got this new study, in Annals of Internal Medicine, which compared a sample of 189 people who had had COVID with 120 controls. Fifty-five percent of their COVID population reported persistent symptoms, relative to only 13% of the controls who hadn’t had COVID. This lines up with high rates of reported long COVID, but when the study looked at laboratory measures of inflammation, pulmonary activity, cardiac activity, cognitive function, and many others, they didn’t show any significant differences. This suggests we have a lot more to understand. The bottom line is that if you’re wondering about the risk of long COVID after a mild to moderate Omicron breakthrough infection, we are in a bit of a data desert. We could triangulate — one study from spring 2021 shows the risk of persistent symptoms is 10 times lower after vaccination — so maybe we take our favorite estimate of 16% pre-vaccine and lower it to 1.6%. But none of this is direct. And we’re even worse off if we want to understand the degree of illness. Vaccination could lower the risk overall and decrease the severity of any long COVID symptoms which do occur. But we don’t really know. My sense, putting this all together, is that persistent symptoms of fatigue, loss of smell, and coughing happen to a reasonable proportion of adults post-COVID, even after mild breakthrough infections. In some cases, these take months to resolve. But my sense is also that things like the CDC's claims that COVID causes a dramatically elevated risk for pulmonary embolism and substance abuse disorder — the most significant disability and disease risks that they attribute to long COVID — are likely overstated by the lack of a reliable set of control individuals. This is triangulation, though. It’s informed triangulation, but others may well triangulate differently. Our lack of data and understanding here is extremely problematic. We need more data, larger samples, and better designs. We need controls and better adjustments for pre-existing differences across groups. An example would be to use populations (i.e. college students) who are tested for surveillance reasons. Combining college-testing data with electronic medical records could provide insight into persistent symptoms in this population (and you could get some older people in there by including faculty and staff). When we work to improve our data, a central goal should also be to distinguish among the forms of long COVID. What puts people at risk for longer-term disability versus persistent but more mild symptoms? Does Paxlovid or Evusheld help limit these risks? What treatments work? I know there are those whose reaction here is to emphasize the need to return to more restrictions — masking, in particular. Without getting into the question of what should or shouldn’t be done, the reality is that COVID is going to continue to circulate. To the extent long COVID is a risk, it will continue to be one. Which means we need to understand it.

This Post Is for Paid Readers Only