COVID Transmissions for 11-20-2020
Good morning! It has been 369 days since the first documented human case of COVID-19.
Today I basically wrote two in-depth pieces; one is as a headline regarding a Danish study on masks and a blog post about it. The other is a dive into that duration of immunity paper, which is in the proper “in depth” section.
As usual, bolded terms are linked to the running newsletter glossary.
Keep the newsletter growing by sharing it! I love talking about science and explaining important concepts in human health, but I rely on all of you to grow the audience for this:
Now, let’s talk COVID.
Danish mask study
This was sent in by a reader, and I’ve been meaning to cover it anyway.
Recently, a study from Denmark had some findings that have been interpreted as saying “masks don’t work.” I want to begin by saying that this interpretation is flat wrong. I also want to link the Danish study: https://www.acpjournals.org/doi/10.7326/M20-6817
The link I was sent was from some kind of “skeptic” site that clearly has an agenda. It made the following assessment of the study (I will not be linking the article I am quoting; it does not deserve the traffic)
The headline result–wearing a mask makes no difference in cases or in level of transmission in the community. It is obvious to all of us now, that this is true, as we can see the case growth in areas of the country with extremely high mask-wearing rates.
This statement is not true. I have shared numerous studies showing that mask mandates and mask compliance are correlated with reductions in transmission in the community. Additionally, the Danish study did not examine the level of transmission or the level of cases in “ the community.” It only looked at infections in the sample of subjects, who were moving freely in the world.
It studied the number of infections in a group of people who were asked to wear masks. These people did not represent their entire community. They were not an entire community. There was no study here that looked at the wider community. The statement made here is not simply wrong; it is a lie. A person would have to be illiterate to read the study in question and come away with this impression.
Specifically, the authors state the limitations of the study in the abstract of the paper:
Inconclusive results, missing data, variable adherence, patient-reported findings on home tests, no blinding, and no assessment of whether masks could decrease disease transmission from mask wearers to others.
I’ll explain much of these limitations in a minute, but please take a look at the phrase I bolded. I don’t think I have to explain it. The authors explicitly state that their work cannot support the falsehood stated by the blogger I quoted.
Remember that masks can do two things. 1, they can prevent you from getting the virus from someone else. 2, they can stop you from giving the virus to someone else. Definitively, as stated in the limitations, this study cannot look at both of these items. Both of these items must be studied to understand transmission dynamics.
The blogger wants you to disregard these limitations:
Before describing the study and its findings, let me make a couple of preliminary observations. I have been reading medical research for 40 years and science papers for even longer. When you read this study and you know the history of the attempt to get it published, you are immediately struck by what the authors must have been forced to do to get any reputable journal to publish it. They were clearly forced to constantly refer to limitations and caveats about the study and there is a discussion about the confidence intervals around the results that you absolutely never see in the published reports of research.
This is also a bald-faced lie. Every single study of any notable meaning states its limitations clearly. A skeptical inquiry requires an understanding of study limitations. Papers will be rejected for not assessing their limitations because it is an essential part of the scientific process. I am a publications manager. I write publications for a living, on behalf of a pharmaceutical company. Limitations are always stated. Again, this blogger is flat-out lying.
Now, I think I’ve quoted quite enough from that blog post. I wanted to do it to be clear about how wrong it is, because it was sent to me by a reader and is presumably circulating, but I’m done with it now.
Instead, let’s talk about the science in the study. The study was a randomized trial, with a control group, that looked at how likely people who were recommended to wear masks were to get infected with SARS-CoV-2 compared with people who did not wear masks. It had a number of limitations, which I’ve quoted above, but let’s look at its results:
A total of 3030 participants were randomly assigned to the recommendation to wear masks, and 2994 were assigned to control; 4862 completed the study. Infection with SARS-CoV-2 occurred in 42 participants recommended masks (1.8%) and 53 control participants (2.1%). The between-group difference was −0.3 percentage point (95% CI, −1.2 to 0.4 percentage point; P = 0.38) (odds ratio, 0.82 [CI, 0.54 to 1.23]; P = 0.33). Multiple imputation accounting for loss to follow-up yielded similar results. Although the difference observed was not statistically significant, the 95% CIs are compatible with a 46% reduction to a 23% increase in infection.
After reading that, I wish to call your attention again to the limitations, which stated “inconclusive results.” There is a 38% chance that the difference between the two groups is entirely due to random chance. I know that from the part where “P = 0.38” is noted; “p” is a statistic that estimates how likely it is that any observed difference between two groups was recorded due to random chance. The standard in biomedical sciences for a result to be considered “significant” (jargon for meaningfully true) is a p-value of 0.05, implying that there is only a 5% chance the results are due to random variation. Stated another way, we prefer a standard where if you did the same study 20 times, you would only get these results at random 1 of those times. At a p value of 0.38, we would get these results at random 8 out of 20 times; the other 12 times, these results might be real. Note that all this tells us is whether the difference appeared at random or not. A difference was observed; we just can’t be sure that this difference is real. This is not the same thing as saying there is no difference between the two groups.
Based on that, I honestly cannot be sure why this was published; perhaps it is because masks are an interesting question, and an inconclusive result is perhaps informative towards the idea that the effect of masks may be weak. It tells us next to nothing, except that “masks might be able to protect a person from becoming infected with SARS-CoV-2; they also might not be able to. Also, we can’t comment on whether they prevent an infected person from spreading the virus.” Thankfully, that latter part has been demonstrated definitively elsewhere. Masks work to prevent spread of infection from sick people, as described in CDC guidance on this matter. It is also unclear if they prevent someone wearing a mask from catching the virus; this is also described in CDC guidance on this matter. Nothing in this study contradicts that guidance.
Anyway, I would like to translate the remainder of the limitations into approachable language. Earlier, I noted that the “mask” group was “recommended to wear masks.” That’s because there was no confirmation in this study that the participants actually wore their masks or knew how to wear their masks appropriately (like, over their noses). This is what the limitations section means when it says “variable adherence.” These people were given a mask and told to wear it. That’s all. Likewise, the control group wasn’t told not to wear a mask. They just weren’t given any masks and they weren’t told by the researchers to wear masks. At best, then, this turns the study from one about mask-wearing into one that assesses whether being told to wear a mask by medical researchers has an effect on people.
Considering that the world is full of messaging recommending that everyone wear masks right now, it’s not a big surprise that we didn’t see a big effect from this recommendation by the researchers.
“Missing data” means that they didn’t get information on everyone in the trial. Missing data is a problem in all clinical studies, but in this study it was a much bigger problem. Of approximately 3000 patients in each trial arm, approximately 600 in each trial arm did not provide results on the primary endpoint of infection with SARS-CoV-2. That’s a huge chunk of missing data, and we can’t be sure that there isn’t a reason the data are missing—perhaps they are missing because patients were embarrassed when they failed to follow the trial protocols.
The next limitation is related to that; “patient-reported findings on home tests.” This means the patients tested themselves for infection at home and if they didn’t like the results they didn’t need to tell the researchers the honest test results. That’s a big problem, especially if these patients had some kind of agenda when it comes to masking.
Then to round out the list of limitations, there is the statement that there was no blinding. “Blinding” is when the use of a trial intervention is hidden from the patients who received it. This is usually done by giving the control group something that simulates the intervention; typically this is called a placebo. You can’t really make a placebo mask. People know when they’re wearing a mask and they know when they’re not wearing a mask. People wearing masks might behave differently because they’re wearing masks. People not wearing masks might also behave differently during a global pandemic.
Anyway, I think this is an interesting study because it suggests that providing people with free masks and recommending they wear them may not have a big effect compared to not doing that. It doesn’t tell me if masks work or not to prevent the wearer from catching the virus, but that’s fine, because we know that masks do work to prevent wearers from spreading the virus. Spreading the virus is also something we wish to prevent.
Wear a mask.
What am I doing to cope with the pandemic? This:
Miniatures gaming league
In what may be the least adult-appropriate thing that I do, I play a game that involves “flying” little plastic Star Wars ships on a table. It’s a fun game and I like that the community is relatively low-drama; it turns out if you get a bunch of adults together to do something they all know is a little silly, people get rid of a lot of the overly competitive pretenses that you might see in other settings.
Sometimes I paint these things, which is a nice excuse to share a photo:
The game is meant to take place in person, but it’s had to adapt to COVID-19 just like everything else. It is played on 3 foot by 3 foot tables, which is not big enough for social distancing. Instead, a lot of us have turned to adaptations of the game that run on something called Tabletop Simulator, a piece of software that does exactly what its title suggests. You can play all kinds of tabletop games with it.
This is all preamble to saying I’ve joined a league that will be playing online over the next 10 weeks, which is great. It’s a nice way to break the monotony. Now, I don’t expect you all to pick up this particular game, but I do think adding some board gaming online is a great idea! I recommended a casual option yesterday, so today I thought I’d give a sense of the more intense experience.
Estimating the duration of immunity against SARS-CoV-2 — Part 1, antibodies
A few days ago I covered a paper that looked at the duration of immunity against SARS-CoV-2, but I wanted to walk through it in detail because I think it was a really cool piece of work. The reality is, though, that it is going to take multiple days to walk through this paper. It has a fantastic amount of data, and I want to work through some concepts that it explains because it’s a great educational opportunity. I am hoping that by walking through this, I will help explain how to read a scientific paper.
It’s also an opportunity for me to critique this paper, which is important because it is a preprint that has not yet been peer-reviewed. I will attempt to provide a limited peer review, though keep in mind that there are more appropriate experts than myself who would be selected as reviewers in a formal setting. I can give you my opinion, but this paper still isn’t official yet. That said, let’s begin.
If you want to follow along, you can find this preprint here: https://www.biorxiv.org/content/10.1101/2020.11.15.383323v1.full.pdf
First, we’re going to skip the introduction. What it’s going to do is set up the problem, and we already know what the problem is: SARS-CoV-2 is sickening and killing people and we don’t know how long immunity to it may last. That out of the way, we can move on to understanding the study.
Today we will start by examining the study population and then walking through Figure 1, which, broadly, answers two questions:
What kinds of antibodies do we see in the response to SARS-CoV-2 infection?
How long do levels of these antibodies remain in the blood of recovered patients?
The first thing we want to look at is the study population in any kind of clinical study. We want to make sure that these are the right people to be looking at, and I think this was a decently varied sample with a lot of types of individuals included. That’s good. Here’s who they recruited to their trial:
185 individuals with COVID-19 were recruited for this study. Subjects (43% male, 57% female) represented a range of asymptomatic, mild, moderate, and severe COVID-19 cases (Table S1), and were recruited from multiple sites throughout the United States.
Blood samples were taken from these patients, and they form the main source of data in this study. Most patients supplied a single blood sample (this group is called the “cross-sectional” group), though 38 patients provided several samples over time, allowing for analysis of changes in individual patients; these are called the “longitudinal” group. This was important because the researchers were interested in studying the duration of immunity.
I’ll note my first criticism though: the numbers were pretty small. That’s understandable, since research isn’t cheap, but it does limit how much these results may represent the wider population.
Now we want to look at the first figure. When you read a scientific paper, you really want to look at the pictures first. They tell the real story because they contain all the data. Everything else is a guide to the figures. Each figure—each panel—is an experiment meant to answer a specific question.
The first, simple question to ask was this: what were the levels of antibodies in these patients compared with the amount of time that had passed since their infections. This is depicted in Figure 1, of which I’ve grabbed a representative panel:
The graphs above show us that antibodies against the spike glycoprotein get to high levels after a week or two, as expected, and then stay at high levels, both in the “cross-sectional” sample and the “longitudinal” sample. I really like the comparison between the two groups here. The cross-sectional group contains more patients, and is more like what we might expect in antibody tests in people walking into a clinic for an antibody test. The longitudinal sample helps us to confirm that the trend seen bears out when you are comparing internally to one patient. Since it’s possible we could have seen different things in the two groups, it’s great that we see very similar things.
One thing we can assess from these figures is the “half-life” of antibodies here. Half-life just refers to the amount of time that it takes for the level of something that decays to reach half of its starting amount. In the cross-sectional sample the estimated half-life is 140 days; that’s almost half a year. In the longitudinal sample, it’s 100 days, which is a little different but we’re looking at a smaller sample of people. These half-life numbers don’t represent the time until the antibodies are no longer effective—we don’t actually know the real limit for antibody effectiveness in this disease, for one thing—but they do give us some hint that the decay in immune response is slow.
With that estimate of half-life, a person would still have 10% of their antibodies after about a year. For reference to the graph, the blue line starts around 10^3, or 1000. 10% would be around 10^2, or 100. I’ll note that 10^2 is still above that green threshold line. That green line represents the limit of sensitivity, which basically represents a line below which the experiment is no longer very good at detecting antibodies. Since this is an experiment where antibodies are detected by their ability to recognize virus antigens, that line might actually be a hint about a meaningful level of antibody. Also, it might not. But it’s interesting to see that the decay is so slow, either way.
The rest of Figure 1 looks in similar ways at other questions. They see similar trends. Specifically, the researchers looked at antibodies specifically against the “receptor binding domain” (RBD) of the spike glycoprotein, and a similar trend was seen this this specific type of antibody. They also look at antibodies against the nucleocapsid protein, which is a protective protein that the virus makes to coat its genome. The reason they look at this one is that the antibody tests available commercially look for antibodies against the nucleocapsid. Again, we see a similar trend in decay of antibodies against this protein as we did to the spike.
Specifically, the half-life estimates for the cross-sectional group for the RBD and the nucleocapsid are 83 and 67 days, respectively. These are also pretty long periods.
They also looked at the level of antibodies that could neutralize “pseudovirus particles,” which are simulated virus particles. This is very cool because it comes on an even closer approach to the potential immune effects of antibodies to SARS-CoV-2. Interestingly, the half-life for this one, in the cross-sectional group, was 27 days, but there was a wide variation in the estimate of half-life, with the “true” value being 95% likely to lie between 11 days and 153 days. So we can’t be too sure about that one. In the longitudinal sample, we have a different picture. The half-life estimated there is 83 days for neutralization, with the 95% window (called a “95% confidence interval [CI]” in statistics) ranging from 68 to 123 days. It’s surprising that there’s a tighter interval in this smaller group of patients, but I can’t readily explain that.
The rest of the figure looks at something called IgA antibodies. Previously, the work looked at antibodies that are more active in the blood, called IgG. It is unclear how important IgG antibodies are against SARS-CoV-2, because the virus is primarily found in tissues called respiratory mucosa, surfaces that form the interface between outside air and our bodies, rather than in areas accessible to blood-based antibodies. IgA are what are known as “mucosal antibodies;” this means that they are often found in mucosa, and they are important for immunity there. For this reason, it’s suspected they might be important for SARS-CoV-2 immunity.
The researchers also looked at IgA against the spike RBD and against the full spike protein. In this case, again, apparently a slow decay was observed, but there’s much more variation. For example, anti-spike IgA in the cross-sectional group had an estimated half-life of 11 days (95% CI: 5 days to 25 days) but in the longitudinal group the estimate was 214 days (95% CI: 126 to 703 days). These numbers are wildly different! I’m not sure which is the better estimate, but it at least tells us that it’s possible for mucosal immunity in patients who recovered from SARS-CoV-2 to be quite long-lived.
Let’s revisit the questions that I said this figure was meant to ask:
What kinds of antibodies do we see in the response to SARS-CoV-2 infection?
The researchers saw IgG (largely blood-based) and IgA (mucosal) antibodies against the spike protein of SARS-CoV-2 and the receptor binding domain of that protein. They also saw IgG antibodies against the nucleocapsid protein, which are currently measured in the clinic to indicate past SARS-CoV-2 infection.
How long do levels of these antibodies remain in the blood of recovered patients?
The researchers saw a wide range of half-lives for the antibodies that they looked at, but there were some that were particularly long-lived, breaking the 100 or even 200-day marks for half-life. This is suggestive of potential for a long-term immune response to SARS-CoV-2.
Now, I recognize that we saw some low numbers for half-life in some of the measurements here. I don’t want you to worry about these. Antibody levels fall when an infection subsides; this is similar to how after a war, you send a lot of soldiers home. Antibodies are a weapon, and it costs the body to manufacture them. When there is no infection, levels of them are expected to decay.
However, the body does tend to remember certain past infections. This memory is implemented through antibody-producing B cells, which were also examined in this paper. That experiment forms part of another figure, which we will deal with in a future installment.
You might have some vaccine questions! Send them in.
Join the conversation, and what you say will impact what I talk about in the next issue.
Also, let me know any other thoughts you might have about the newsletter. I’d like to make sure you’re getting what you want out of this.
This newsletter will contain mistakes. When you find them, tell me about them so that I can fix them. I would rather this newsletter be correct than protect my ego.
Though I can’t correct the emailed version after it has been sent, I do update the online post of the newsletter every time a mistake is brought to my attention.
No corrections since last issue.
See you all next time.
Always,
JS
Thank you for the in depth on the Danish study. Some people I know have been citing the claims that this study proves that masks are ineffective. I am using your in depth as a non-confrontational apolitical means of getting people to focus on the actual facts. Very helpful.