A conversation about Convalescent Plasma (CP) – I’m going to give a general overview and then go more into detail about what we do and don’t know (and how it became a political show…)  The bottom line is that it really might help some people, but we need more data, but it’s hard to get good data, it’s getting harder to get good data, and the government made things a lot worse by mischaracterizing the data we do have and politicizing the whole issue… 

Antibodies are little proteins made by the immune system which bind to foreign things (antigens) such as viral proteins, call for backup from other immune cells, etc. Some antibodies, referred to as “Neutralizing antibodies” (NAbs) bind to a virus in such a way that they actually block the virus from getting into new cells. Infected people normally make these antibodies mid-to-late-ish in the course of the infection and they stick around in the blood to keep watch in case the virus tries to strike again. 

So the infected person’s blood contains secreted NAbs which, In addition to preventing that person from getting re-infected, if given to a different person, could prevent that different person from getting infected (or at least prevent more cells in a person’s body from getting infected). This is the basis behind Convalescent Plasma (CP) therapy – give blood plasma (the cell-free part of blood) from someone who recovered from the disease to someone who is currently battling the disease. NAbs in the plasma can then go to work in that person’s body to help ward off increased infection. 

Problem is, (in response to the same virus) different people make different NAbs (some stronger than others) and different amounts of them stick around in the blood.  So each batch is different. To get around this issue, other “passive immunity therapies” (where you give someone antibodies directly as opposed to “active immunity therapies” like vaccines where you teach their body to make antibodies) use lab-made antibodies. The idea with monoclonal antibody therapies is that you isolate the best antibodies, make a ton of them, and give them to a patient. That way you know they’re getting good ones, and a lot of copies of them. 

As I discussed yesterday, one of the downsides to monoclonal antibodies is that you’re only giving a single antibody, against which a person might have a rare mutation in the virus that allows it to escape. This is why some companies, like Regeneron are trying an “antibody cocktail” approach – mixing 2 monoclonal antibodies with the idea that, even if someone has a virus resistant to 1 antibody, it’s unlikely they’d have a virus resistant to 2 antibodies. http://bit.ly/antibodycocktails

A big downside of antibody cocktails, and even plain old monoclonal antibodies, is manufacturing capacity & cost. This is one of the reasons why Eli Lily’s treatment, LY-CoV555, only contains a single antibody, gambling that the increased cost & resources for adding a second would outweigh the benefit. https://bit.ly/3lwicQX 

That may sound like a really crass economics decision, not taking into account patients, but it might actually be able to benefit more patients. It’s really unlikely that even a monoclonal antibody, let alone a cocktail, will be available to patients in developing countries. The monoclonal antibodies currently on the market (for cancer, autoimmune diseases, etc.) are some of the world’s most expensive medications, and 80% of them are sold in the US, Europe, & Canada (despite the fact that those countries only make up 15% of the global population). https://bit.ly/2QDW1d2 

Groups of prominent scientists at the International AIDS Vaccine Initiative (IAVI) and Wellcome are calling for infrastructure, policy, and other practices to be put into place to make monoclonal antibody therapies more widely available to the global population. https://go.nature.com/3gHdHiO 

That’s, unfortunately, a hard sell in a $-centric world. And, in the meantime, people in developing countries do, however, have easier access to convalescent plasma, but doctors and scientists still aren’t sure how well that works. Recent policy changes in the US may make it harder to figure out if it *does* work (and how to make it work better). It seems to help some patients, but it’s not clear yet *who* is benefitting, who is *not* benefiting, and why the difference. Later in the post I will explain the evidence we currently have…

The good news is that it doesn’t seem to be *harmful* (though as with any treatment, there can be rare adverse effects). But to know if it’s really helpful, we need more data – data from controlled trials. The recent EUA (Emergency Use Authorization) from the FDA enabling the use of convalescent plasma outside of clinical trials and something called the Expanded Access Program (EAP) which isn’t a trial but does collect data, likely means that it’s going to be harder to get the data that’s needed. If people are convinced it works they won’t want to enroll in a clinical trial for it where they could just get a placebo (fake treatment like an infusion of salt water). Instead, they’d rather just have their doctors give it to them (which they now can). And those doctors, if they’re convinced it works, would rather give it to them as well because they want to help them. 

So, at least in the US, it’s going to be hard to get the kind of controlled trial data we need. So keep an eye out for reports coming from trials in other countries. And keep in mind that it’s basically impossible to do a truly controlled trial of convalescent plasma because each batch of plasma is different because each donor is different – so the plasma batches have different mixes of different strengths and amounts of neutralizing antibodies directed against the coronavirus. And the levels of these antibodies haven’t been being measured before giving them to patients. So some patients might be getting high levels of strong antibodies (“high titer”) whereas other patients might be getting plasma that barely contains any. 

There have only been 2 of the “gold standard” Randomized Control Trials (RCTs) done – 1 in China & 1 in the Netherlands. They didn’t see a significant benefit, but they were small studies, terminated early, so not deal-breakers https://bit.ly/2EL3apo 

The study that the EUA is largely based off of was *not* an RCT – instead it was from data collected from patients in that Expanded Access Program (EAP). It’s run by the Mayo Clinic, but “any” hospital across the country can choose to participate in it and thus administer CP severely-ill hospitalized patients – and ~2800 hospitals did so. The program was set up to:

  1. help get CP to patients around the country
  2. test for *safety*

It was *not* set up to test for efficacy. It was more a grass roots type of effort that started with a lot of advocacy from a Johns Hopkins scientists named Arturo Casadevall (whom I got to hear talk at last week’s CSHL virtual meeting on SARS-CoV-2 and who seemed incredibly well-respected in the scientific community and genuinely in it for the right reasons – helping patients). Casadevall put together (and chairs) the National Covid-19 Convalescent Plasma Project, a consortium coordinating research into the therapy. https://go.nature.com/2YOfVqr

As I mentioned above, the EAP was about access & safety, *not* efficacy. And they aimed to enroll ~5000 patients. But, as viral spread exploded in the US, so did the demand for CP and over 90,000 patients enrolled. So now you have data from tens of thousands of patients collected without all the rigorous controls (like a placebo group) you really need to truly say something does or doesn’t work. But it would be a waste to not use that data, so the scientists leading the trial did an “after the fact” analysis of it which goes against a lot of “good statistical practices” as I’ll get into later. But they worked with what they had and I really think that some people are being too hard on them. They admit (and detail) the limitations of their studies and it’s really the government response to them that is to blame if you want to blame someone for potentially authorizing it too soon… 

Anyways, one of the things they found was that, in the subset of patients for which they’d measured the antibodies in the CP they’d received, patients who received CP with “high levels” anti-Spike IgG antibody had a higher survival rate than patients who received “low levels” (IgG is just a type of antibody). The EUA requires plasma be tested first to make sure it contains “high levels” but the test is a qualitative “yes/no” test instead of a quantitative test that gives actual numbers, so how far above the threshold of “high” the levels of antibodies in plasma are is unknown. Which can be especially troublesome if you don’t know how high is needed. https://bit.ly/2QCJS8i 

It’s also important to keep in mind that, in addition to having different mixes of of different strengths and amounts of neutralizing antibodies directed against the coronavirus, they have all sorts of other antibodies against past infections, different chemical messengers, etc. (basically a lot of stuff other than the antibodies you want – but there might also be stuff you do want?). 

Bottom line, CP has a lot of variability, so, even with an RCT, in order to really get the “statistical power” needed to say the treatment helps or doesn’t help, you need really big trials with lots of participants. Especially because scientists will want to try to look in that data for hints about who seems to be benefitting and how we can better target the treatment for people most likely to benefit – does age, sex, BMI, etc. matter? If, for example, you only have small numbers of people in each age group, you can’t tell if the effects are just due to chance. 

One trend that seems to be emerging from the data we do have is that, similarly to a lot of antiviral drugs, earlier administration of the treatment is more likely to be of benefit. But, since most people will recover fine it’s harder to get patients at these early time points, when they don’t feel that sick, so why receive a treatment that might not even help them if they’re going to get better on their own? 

There are however, trials underway to test mildly-symptomatic outpatient patients, including a multi-center “randomized, single blind, two arm, placebo controlled phase III trial” (doctors and patients don’t know who’s getting what) with a cool name – C3PO (Clinical Trial of COVID-19 Convalescent Plasma of Outpatients). In this study, over 50 hospitals nationwide are enrolling patients from emergency departments – half get saline, half get plasma. And they’re looking for efficacy! https://bit.ly/2QAmMiJ 

The hope is that if doctors have enough data they can get better at predicting who is likely to benefit and who isn’t. Then they can target those groups most likely to benefit – and if they then do trials in those groups they should see significant evidence of benefit. If they do *not* then, and only then, should they stop researching it. 

The data isn’t hopeless with regards to efficacy, but it’s far from a miracle drug.  It’s pretty clear that CP is far from a “silver bullet” for all – if it were, it wouldn’t be so hard to find evidence that it works. The question now is “is it a silver bullet for some people?” 

Imagine there’s a genetic variant affecting response to a drug so that if a person has a certain genetic variant “a” they respond well to the drug but if they have a different genetic variant “b” they don’t. If you had a group of people, some with each variant, and you treated both with the drug, how “good” the drug will appear depends on how many people in the group had the “a” variant. If you did a trial with mostly “a” people the drug would look awesome, but if you did a trial with mostly “b” people, the drug would look terrible. 

The same type of situation likely applies when it comes to CP – we just don’t know what the “a” vs “b” is – and it’s likely not just a gene variant. A couple of the main questions are, what proportion of people are “a”? And is there an easy way to know they’re “a” so that you don’t waste plasma on people who won’t benefit? Also – if we’re talking about CP having benefit because it can be used in less affluent countries, we can’t also be saying, oh – but you need to do these expensive tests before you give it to them. 

So, we need more data before we can say without a doubt that CP is helpful. But that isn’t what you heard if you turned on the news last week… In giving the EUA (just a day after a group of top scientists called for more evidence), the FDA commissioner really overstated the evidence of the treatment’s benefits, and later had to walk back his enthusiasm as scientists called him out. Scientists weren’t just calling him out, though, they were asking for more evidence. They (at least most of them) weren’t saying it does not work, they were just saying that we don’t have enough proof that it does work – and that there were some problems with the methodology of the studies that the EUA was based off of (as well as problems with how the FDA commissioner kinda played with statistics to make it sound better), both of which we’ll get into. 

But now, the whole issue has become politicized, with one side saying “it works!” the other side saying “it doesn’t work” (with people wondering whether convalescent plasma is just another hydroxychloroquine) and scientists saying “we need more data!.” And this politicization could have the perverse effect of leading to premature abandonment of this treatment strategy, which still might be able to help the word’s most vulnerable, who likely won’t have access to the fancy-dancy treatments of the world’s wealthy. 

On the other side of the coin, some proponents of the EUA argue that it expands access to rural and underserved communities – the Mayo Clinic’s Expanded Access Program in a way meant that patients in wealthier areas with access to fancier hospitals, etc. were essentially being treated as if the EUA were already in place. It was “Expanded Access” to the privileged in some cases. 

So the EUA might be both good and bad for the underserved. 

But how did we even get to this place? Let’s back up and start from the beginning. Actually, let’s start from the 1890s…

That’s when CP was first used, as a treatment for diphtheria. It’s since been used to treat a lot of diseases including Ebola and was even used during 1918 flu pandemic we’ve been hearing so much about lately. Here’s a really good NYT article about its historical and current uses https://nyti.ms/2EBDyeY 

Since it’s more “shelf stable” than whole blood and has less risk of cross-reactions, blood-borne pathogens, etc. human blood plasma is used all the time – for a bunch of different reasons, like if people have lost a lot of blood. The only difference between “normal plasma” and “convalescent plasma” is that you’re taking it from people who have recently recovered from an infection and thus (hopefully) still have high levels of antibodies against the thing (e.g. virus or bacterium) that caused the infection. So there’s lots go evidence that plasma treatment (in the broad sense) is generally safe (though, as with all treatments, there can be “adverse effects” such as Transfusion associated circulatory overload (TACO) in some patients). 

CP is often the first go-to for an emerging pandemic when you don’t yet know the disease’s Achilles’ heels to target with specific therapies – it can even work if you don’t know what’s causing the disease. All that matters is that the recovered patient’s body knew what was causing the disease and produced antibodies to combat it. So it wasn’t a surprise that CP was tried early on. What is a surprise to some is that it’s still being so widely used as it is largely seen as a “stopgap measure”…

But let’s get back to the “early on”… controlled clinical trials to test if it was helpful, where some patients got CP and some didn’t, started in the early days in hot spots, such as New York. But they had trouble enrolling enough patients, especially since the pandemic thankfully subsided. There are lots of people still getting infected across the country, and there are controlled trials attempting to go on, but most people who have been getting CP in the US have done so through the EAP. This has already had a sort of perverse effect of making it harder for clinical trials to get enough enrollees. And now, with the EUA, it’s going to be even harder…. 

NYU Langone Health started this thing called COMPILE (Continuous Monitoring of Pooled International Trials of Convalescent Plasma for COVID-19 Hospitalized Patients) where they’re collecting data real-time data from RCTs taking place around the world (so they don’t need to be finished) & pooling the data to try to get more information. https://bit.ly/3b80uhA  Of course, there are important limitations to this approach to keep in mind (it has a lot of statistical no-no’s that can lead to difficulty with interpretation) but it’s one way we might learn more from what we’ve got.

In the meantime, let’s look at what we’ve got. I want to get into more detail about that paper on the EAP. But first I want to get a little tech talk out of the way. 

Plasma donation is done a lot like whole blood donation, but they take your blood through a machine that filters out the cells & platelets (cell fragments that help with clotting), keeping only the plasma and giving you back the cells. 

When it comes to collecting convalescent plasma, you want to take it from patients who have recently recovered because antibody levels in the serum are higher earlier on and then start to wane. This is totally normal, because you don’t want your blood to get stuffed full of tons of antibodies for every infection you’ve encountered. Instead, your body keeps some in your blood and also keeps “memory B cells” which know how to make it hanging out in your bone marrow ready to make more on demand. But, with CP, the patient receiving it doesn’t get those memory B cells, so they can’t make more copies, so you want there to be high levels of antibodies in there. 

Antibody levels are typically talked about in terms of “titers.” Basically this refers to “how much” antibodies are in the plasma. Typically, when it comes to neutralizing antibodies, they measure how far you can dilute the plasma and still block the virus from infecting other cells. But neutralizing antibody tests are harder to do, so usually tests just look for binding, which can be done using an ELISA. The further you can dilute it, the more (and/or stronger) the starting amount of antibodies. So, for example, a 1:2000 titer would be better than 1:1000, because you can dilute the 1:2000 plasma 2X as much and still have the same antibody effect as you have with the “low titer” one. 

Now let’s dive in. Here’s a link to the paper if you want to follow along: https://bit.ly/32IdaZ4 

It was posted on MedArxiv, a pre-print server, and it has not yet been peer-reviewed (other than on Twitter, etc…). They looked at outcomes for >35,000 patients in the US who received CP at ~2,000 different sites across the country that were participating through the Expanded Access Program. 

In this study, there was no placebo group, so the “35% reduction in mortality” splashed about was comparing people who received plasma with a high concentration of antibodies to plasma with a low concentration of antibodies. And it was a relative reduction, not an absolute reduction – bear with me for a stats-y second… What this means is the difference was between 8.9% mortality at 7 days for patients receiving “high IgG plasma” and 13.7% for patients receiving “low IgG plasma”. The 35% comes from dividing 8.9% by 13.7%  (and subtracting from 1) – and that’s a “relative reduction” – the “absolute reduction” you get by subtracting 8.9 from 13.7 is only 4.8%. 

But the FDA commissioner, Stephen Hahn, expressed it as an “absolute” mortality benefit, saying it would save 35 out of 100 people. That is *not* what the study says. If the numbers from the study are representative of how everyone reacts to the treatment, the study says that, if you treat 100 people with high IgG CP and 100 people with low IgG CP, 4.8 more people will be alive in the high group than the low group. And, since there was no placebo group, we don’t know how that compares to how many people would be alive compared to no treatment – those patients could have even done better than people who received CP but we don’t know.

This mischaracterization and “over-enthusiasm” from Hahn quickly politicized the situation and, in a Tweet on August 24, he walked it back, writing “I have been criticized for remarks I made Sunday night about the benefits of convalescent plasma. The criticism is entirely justified.  What I should have said better is that the data show a relative risk reduction not an absolute risk reduction.”

But the damage had largely already been done, casting shade on the whole study and drawing warranted suspicion that pressure from the White House led to the EUA being granted on shaky data. The whole situation was already seen as somewhat suspicious because of Trump’s pre-hype of the announcement.

I don’t know if the data are strong enough to warrant it. And I’m not going to try to speculate. 

But the data do show some hopeful signs. For one thing, the people in the middle (“medium IgG plasma”) had a survival rate in the middle (11.6%) and the dose-response nature of the effect could be a hint that it actually *is* effective since if it weren’t it would be less likely that, by chance, the more IgG you gave, the better the patients did. 

However, these numbers are not coming from the full >35,000. They only knew the antibody levels for a subset of those patients, about 3000 of them (3,082 to be precise) so only those patients were included in those percentages

As I mentioned before, timing seems crucial with earlier=better. Mortality (7 days out) was 8.7% for patients who got CP within three days of diagnosis & 11.9% for patients who got it after 3 days. note: a big caveat is that “time of diagnosis” is different from “time of symptom onset” which is different from “time of exposure” – basically, “time of diagnosis” isn’t that biologically meaningful so it’s hard to compare, especially as diagnosis depends on testing which depends on test availability, etc. 

That in mind, the earlier=better effect makes sense with CP’s mechanism of action – it can prevent the virus from getting into more cells, but it can’t undo the damage the virus is already wrecking. Nor the damage the immune system is imposing. It appears that the most damage done by the virus is actually the immune system overreacting in the later stage of the infection. This is why the steroid treatment dexamethasone, which dampens the immune system, is helpful for really sick patients but not helpful for less sick patients or patients early on in the infection when you need the immune system to fight the virus off. CP is unlikely to benefit patients later in the infection, at which point it’s really the immune system that’s causing the most problems. At that stage, plasma is unlikely to cause any harm, but it’s also unlikely to cause any benefit.  

Another caveat: I’m not an expert on this, just a biochemist-in-training trying to weed through hype and counter-hype to figure out what the deal is with this treatment. This is how things look to me and I hope it helps. If you want more information…

A really great article by Jessica McDonald at FactCheck.org does a great job explaining things: https://bit.ly/2EL3apo 

John Hopkin’s Bloomberg School of Public Health review of the Mayo paper: https://bit.ly/31D82pt 

more Covid-19 resources: https://bit.ly/covid19bbresources ⠀

more on topics mentioned (& others) #365DaysOfScience All (with topics listed) 👉 http://bit.ly/2OllAB0


Leave a Reply

Your email address will not be published.