Why did they give antidepressants to COVID patients?

Follow me on Twitter if you want to discuss this essay, my other essays, or anything else having to do with science and biotech. If you’d like to get notified when I publish new essays, sign up for my mailing list, which I literally only use to alert people when I publish new essays.

You may have seen a headline recently that says something like “Cheap, generic antidepressant may reduce severe Covid-19 disease, study finds”.

If you click through to the news story, you’ll see a story about how doctors gave fluvoxamine, a somewhat uncommon SSRI, to COVID 19 patients in Brazil. 11% of the patients given fluvoxamine ended up in the ER, while 16% of the patients given placebo ended up in the ER. You’ll also see some handwaving about how maybe fluvoxamine works in COVID 19 because it reduces inflammation, or maybe because it reduces blood platelets, or maybe just because it’s an SSRI.

It’s a neat story, and a great example of drug repurposing living up to its promise. But, the way it’s presented, you’ll also likely be left with a bunch of questions like:

1. Why were these scientists giving an uncommon SSRI to COVID 19 patients in Brazil?

2. Why are they so uncertain on how it works?

3. Who paid for all this?

4. If trying random drugs on random diseases works so well, why don’t we do it more?

These are all excellent questions that are thoroughly unanswered by the news article. Not coincidentally, the answers to these questions are both interesting and have a lot of relevance to how drugs are developed today, especially drug repurposing development efforts like this fluvoxamine one. I thought I’d take a stab at answering them in the essay below.

Alternatively, if you want the short answers of these questions, they are as follows:

1. Mouse trial + case studies + no better ideas.

2. Too many cooks.

3. Billionaires.

4. Money.

But these short answers aren’t nearly as fun or interesting as the long ones, and don’t give me the chance to talk about the clever way that my company is doing our own drug repurposing project. So, I recommend you read on.

If you look up fluvoxamine on Wikipedia, it’ll tell you that fluvoxamine is an SSRI and it works the same way all SSRIs do. Namely, when one nerve cell squirts out serotonin to a receiving nerve cell, fluvoxamine blocks the reuptake of serotonin. This leads to an increase of serotonin in the synaptic gap, stuff happens, and hopefully the patient’s anxiety or depression is cured.

Selective Serotonin Reuptake Inhibitors (SSRI) - www.medicoapps.org
A misleadingly simplified version of how SSRIs work. This author comes from the classic “gumdrop” school of drawing SSRIs.

Wikipedia will also tell you that fluvoxamine, like some other SSRIs, has the “off-target” effect of activating, or agonizing, the sigma one receptors, mysterious receptors that modulate other receptors to, uh, act differently. The quotation marks and vagueness here is deliberate. Sigma one agonism may contribute to the antidepressant effects of fluvoxamine (like it does for opipramol, an old anxiety drug only used in Europe), and agonism of the sigma one receptors has a confusing grab bag of effects that don’t fit neatly into a single classification.

One of these effects in the grab bag seems to be blocking a specific inflammation pathway that’s very important in COVID and other infectious diseases. If you’re interested in the science behind it, follow the footnote [1]. Otherwise, you should just know that this effect of sigma one agonism was theorized and tested in a mouse study of septic shock in 2019. They found that mice who were given an injection of fecal matter (gross) and then given fluvoxamine did better than mice who were given the injection and not given fluvoxamine.

This mouse study of septic shock was supposed to turn into a human trial of septic shock, which is a thing we were worried about in 2019. To be fair to yesteryear’s worries, septic shock does kill an estimated 20 million people a year, and has a 25-50% mortality rate even in developed countries. So, not a crazy thing to worry about.

So far, this all is normal. Trials where drugs are repurposed to test on mice happen all the time. Every single drug you can think of and more has been tested on every single sick mouse in existence. Moreover, pretty much all those mouse studies end with a great result for the mice, the author saying “let’s test this on humans!”, and then nothing happening. This was a better run mouse study than most, but, still, this isn’t the interesting part of the story.

Have you ever wondered why, if scientists can cure cancer in mice, mice still get cancer?

The interesting part happened the next year after COVID hit, as most interesting things in 2020 did. Early on in COVID it was noticed that the “cytokine storm”, which is the body’s inflammatory response to COVID, was one of the major reasons for COVID’s relatively high fatality rate. Halting or inhibiting this cytokine storm was seen as one of the top priorities for COVID researchers, especially as the drugs that were usually used to tamp immune responses, like the steroid dexamethasone, only worked ok and had serious side effects. 

The panic of COVID meant that science and medicine happened a lot more so than it did in the pre-pandemic days. One of those ways it happened more was in very aggressive collection, sharing, and analysis of patient data, including data on the cytokine storm. In this frenzy of analysis, one Parisian hospital soon noticed something peculiar: COVID patients who coincidentally received antidepressants soon after admission to the hospital had diminished risk of intubation or death. In other words, they seemed to have better outcomes.

In normal times and for normal disease, I’m not sure if anything would have come of an analysis like this one. Even if this analysis had been run (a big if), it would have taken a while for any sort of resulting trial to be run. But, as the bank advertisements say, COVID is a new normal. So, before the Parisian’s hospital’s data was even formally published, an American doctor took their findings, connected them with the mouse septic shock study, then did a 128 person study where she explicitly examined whether fluvoxamine could prevent clinical deterioration in COVID. And, what do you know, it could!

128 people isn’t a lot of people, though, especially as only 6 people in the placebo group experienced clinical deterioration. So, a much larger study was organized in Brazil to get a better sense of whether this result was real or a fluke. That was the TOGETHER trial, mentioned at the beginning of the post.

Whew! That’s question 1 in the bag: how we ended up giving fluvoxamine to COVID patients in Brazil (correction: Patrick Collison, who partially funded the TOGETHER trial through Fast Grants, tweets that why the trial was funded involved more small trials/evidence than I mentioned above).

By the way, it’s worth noting that from the Parisian hospital analysis to the American study was only about 4 months, and the publication of the American study to the start of the Brazilian study was only like 2 months. By trial standards and scientific standards, that is crazy fast. I just want to emphasize one part on how exceptional these studies were before moving onto the next question. Don’t worry, the second part of how exceptional these studies were is coming soon.

Question 2, on why the scientists were so unclear on how fluvoxamine works in COVID, is much quicker to answer and less interesting, in my opinion. This was a large, collaborative study that the scientists were trying to get out as quickly as possible. There was probably a diversity of opinions on why the drug worked, and it was easier just to skip over that and focus on the takeaway that the drug did work. Hammering out the argument was not worth the time.

Question 3 (“who paid for all this?”), on the other hand, is a much more interesting question. The American study was paid for by the COVID 19 Early Treatment Fund (CETF). The Brazil study was paid for by Fast Grants and the Rainwater Charitable Foundation.

That’s interesting because literally all of those foundations are funded by multimillionaires and billionaires. CETF received a lot of funding from the first president of Ebay, Jeffrey Skoll. Fast Grants received a lot of funding (and free publicity) by Patrick Collison, the cofounder of Stripe. The Rainwater Charitable Foundation is funded by the fortune of the late Richard Rainwater, a billionaire investor.

Do you notice something about that list of funders? Well, who’s missing? You guessed it: any governmental or nongovernmental organization who had any prior major investments in public health. These COVID trials were not funded by the NIH, the FDA, the CDC, the Red Cross, or Mass General Hospital. Nope. Just random billionaires.

Staff honored for improving patient care and the workplace environment |  Clinical Center Home Page
Pictured: the NIH giving each other medals. Not pictured: the NIH funding COVID trials.

And now we have the second piece of why these trials were so exceptional: they didn’t just have capable, organized people pushing them to occur rapidly, they were also funded by generous donors who were willing to rapidly deploy capital. That is incredibly rare.

If the rarity of easy funding of worthwhile drug trials seems weird and troubling to you, then I agree. Unfortunately, this is the way the nation’s funding infrastructure is set up. These large trials require a lot of funding and organization. Doctors and hospitals need to be recruited, instructed, and paid; materials need to be supplied; and data needs to be gathered and analyzed. The NIH (or NSF, or whoever) not only won’t provide all the money necessary to run the trial (or sometimes any money at all), they’ll also give any prospective investigator a ton of grief and paperwork, making it difficult to run the trial properly.

This issue ties into the answer to the last question: why don’t we do this more often? Why aren’t we constantly trying random drugs on random diseases? After all, it’s not like there’s any lack of studies like that mouse septic shock study. Hell, there’s an idea right there: fluvoxamine in a trial for septic shock in humans. Done.

The short answer is that we don’t do this more often because it’s really tough to fund these studies. The generosity of billionaires is always a tricky thing to rely on, and, as mentioned, public funds won’t do. Large non-profits can step in occasionally, but they’re usually disease specific (like the Michael J. Fox Foundation for Parkinson’s) and don’t love putting this much of their budget into one trial or study.

The other other option for funding is, of course, through drug manufacturers. You might even expect these would be the most natural funder for these sorts of trials, given that they’ll make a lot of money off of the new indications for the drug. What’s to stop a drug manufacturer from funding a bunch of trials and then making money for whatever diseases the trial works for?

Well, it’s not so simple. If we look at fluvoxamine, for instance, it’s generic, or off-patent. So, while there used to be one manufacturer of fluvoxamine, now there are many. 

This has been great for reducing the prices of fluvoxamine for its original indication, depression. However, this introduces a massive free-rider problem for any new indication that someone might try to come up with for fluvoxamine. No generic manufacturer is going to fund a trial of fluvoxamine for COVID on their own, because the other generic manufacturers would also benefit from the results of this trial and not be out millions of dollars.

This isn’t to say there’s no way to have a drug manufacturer fund a drug repurposing trial. If there were, my company, Highway Pharmaceuticals, would be dead out of the starting gate, as that’s our modus operandi. It just means that drug repurposing trials by for-profit entities have to be more clever than those by non-profits in order to avoid free-rider problems.

So, let’s talk about the clever ways that drug repurposing trials can be funded by for-profit entities. Not only will it be informative, but it’ll give me a chance to call myself clever, which I always enjoy.

The first and most obvious way is just to try to patent the drug that’s being repurposed, and then use the courts to fight any free riders. Contrary to popular belief, you can patent a novel use for a generic drug, as long it’s “novel and non-obvious”, which I’ve discussed before. If you get the patent, you can hopefully carry out the trial on the actual drug that’s already been successful in a mouse trial, without having to modify it or manufacture it yourself. Then you can simply sue anyone who threatens your patent, and laugh all the way to the bank. For a current example, this is what a startup called PharmaTher is doing with ketamine for Parkinson disease dyskinesia, based on case reports of Parkinson’s patients who received ketamine for anesthesia and then found their dyskinesia alleviated.

How club drug ketamine fights depression | Nature
This is what I got when I Googled ketamine, so I have to assume these are Parkinson’s patients celebrating the alleviation of their dyskinesia.

However, while this is the most obvious way to fund and make money off of drug repurposing, it’s actually the hardest to carry out in practice. Patents, as my patent lawyer likes to remind me, are only worthwhile if you’re going to enforce them. And the problem is that you won’t just have to enforce your patents against other drug manufacturers. You’ll have to enforce your patent against any doctor who reads your trial and decides to try out the generic version of the drug on their own patients. So, PharmaTher is going to need to be willing to sue any doctor who uses generic ketamine on a Parkinson’s patient. If you imagine a newspaper headline like, “Pharmaceutical company sues local doctor for trying to help Parkinson’s patients”, you’ll understand why this is a tough path to follow.

The second possible way is to slightly modify the generic, get a patent on the slightly modified version, and then sell that. As the only manufacturer of the slightly modified version, you avoid the issue of off-label prescriptions. This is what Johnson and Johnson did with esketamine for depression, as Scott Alexander eloquently describes

In this case, you do reintroduce some of the risks of a new drug (e.g. difficulty in manufacturing, risk of the drug not working, risk of the drug causing toxicity), but hopefully it’ll be much less of a risk than a completely new drug. You also will likely have some medical professionals who realize they can just prescribe the generic instead of your slightly modified version, like Scott in the link, but there won’t be too many of those and they’ll be fighting uphill against your marketing and relationships with insurance companies. Overall, this is a reasonably safe way for a company with chemistry and manufacturing expertise to repurpose a drug, as long as they’re willing to get a little bit slimy with their marketing.

The third way is for those of us, like myself, who have little manufacturing expertise and a dislike of slime. This way is to combine two generics for some sort of synergy in a new indication. Doctors aren’t likely to prescribe two drugs combined off label, especially if one or both of those drugs in the combination is in a non-standard dose (imagine a doctor telling you to take ⅓ of one pill and ½ of another). You also avoid almost all of the risks of a new drug, as manufacturing and toxicology are way easier, and it’s much easier to predict whether the drug will work in this new indication. Finally, you’re less likely to annoy people if you can argue that this is a genuinely surprising synergy in a new indication, like Amylyx has argued with sodium phenylbutyrate/taurursodiol combination in ALS, given that neither of those drugs could have been predicted to work well in ALS.

The main danger in this approach is that people tend to attack the synergy part, and argue that these are just two drugs which both work separately, or even just one of the drugs work. Amylyx got that charge in a letter to the journal they published their trial in, arguing that taurursodiol works on its own. This can escalate to more serious consequences, too, as the FDA can and will reject drug combinations that they don’t see as synergistic. Overall, this is a slightly riskier but much cheaper way to repurpose a drug, perfect for startups with shallow pockets and a dislike of fake drug differentiation.

So, tying this back to fluvoxamine for COVID, a for-profit company could have conceivably funded this trial by either using a drug that’s still a sigma one agonist but not generically available (like opipramol), or by coming up with some clever, synergistic drug combination to combine with fluvoxamine, like something that suppresses cytokine production by another way.

Neither of these would have been as fast or as efficient as what actually happened with fluvoxamine, and I think it’s truly a great thing that this fluvoxamine trial was able to be carried out and funded the way it was. But, they are much faster and more efficient than the usual alternative path, in which the drugs remain in mouse trials for years until some magical combination of non-profit money comes together to fund a small human trial. You gotta take what you can get. 

[1] One of the effects of sigma one agonism seems to be inhibition of cytokine production and the IRE 1 pathway, part of the unfolded protein response. 

Cytokines are small proteins that can serve as very localized alarms. The unfolded protein response is a cellular response to situations in which the cell notices that its protein factory, the endoplasmic reticulum, is consistently churning out misshapen proteins. When this occurs, the cell immediately stops making proteins and either fixes the factory or burns it down by killing the cell.

Cytokines and the unfolded protein response are very related to each other, as one can cause the other, or both can be caused by the same thing. Anytime that the UPR is activated is an emergency, and massive cytokine production is a very stressful event that can cause the cell to start misfolding proteins.

They are also, for the most part, good things. It’s good for the body to have localized alarms, and it’s good to shut down cells that are consistently malfunctioning. 

The problem comes when this becomes a cascade. Imagine the body as a factory, and each cell as a person. If one person is consistently messing up production, it’s good for the foreman to try to fix the situation, or fire the person if it doesn’t work. It’s also good for the foreman to tell that person’s coworkers, so the coworkers know they’ll have to step up their own production for a bit and watch out for whatever element caused the first person to slack. If, however, the foreman fires 100 workers at once and tells all their coworkers at the same time, that can have unpredictable effects in that section of the factory, as it’s unclear who steps up to cover the load and there’s also going to be a lot of noise. This can lead to chaos and a total factory shutdown.

This is where fluvoxamine comes in. Fluvoxamine, as a sigma one agonist, inhibits the foreman firing people. This can turn a cascade of 100 people being fired at once into only 10 at a time, which is easier for the factory to manage.

Fluvoxamine’s ability to do this was theorized for a while, but the best confirmation came in a mouse study in 2019. In this study, they injected mice with a slurry of fecal matter, which is both disgusting and a great way to induce septic shock. 

Septic shock is a classic example of a cascade: a million cytokine alarms go off at once, all the cells start messing up because of the noise and stress, the foreman freaks out and fires a bunch of people/cells at random, and the organism often dies. However, this study found that mice who genetically lacked sigma one receptors died more often after injection, probably because nobody was there to inhibit the foreman. It also found that genetically normal mice who were given fluvoxamine after getting injected did better than normal mice who didn’t. Finally, they found hints for a similar role of fluvoxamine in human blood samples.