Fluvoxamine is an anti-depressant drug that belongs to the class of drugs known as SSRI’s (selective serotonin reuptake inhibitors). It’s normally used to treat depression and anxiety. Lately, there’s been a lot of discussion over whether fluvoxamine might be effective as a treatment for covid-19.
Yes, that was my initial reaction. Why would an anti-depressant have any role in treating covid-19?
Well, apart from its effects on serotonin signalling in the nervous system, fluvoxamine has an activating effect on a receptor called S1R, which is involved in regulating the immune system. Since the most frequent reason for bad outcomes in patients suffering from covid-19 is that the immune system goes berserk, the thinking is that fluvoxamine might help to prevent such bad things from happening. And since fluvoxamine is a cheap, generic drug that’s been around for decades, that would be great if it turned out to be true. Unlike the new anti-covid drugs coming out from Pfizer and Merck, it wouldn’t destroy countries’ health budgets if used widely. Roughly 10-15% of most western countries’ populations are already on an SSRI.
Back in November 2020, a trial showing promising results was published in the Journal of the American Medical Association (JAMA). 152 people diagnosed with covid, with mild symptoms not requiring hospitalization at the point of inclusion in the study, who had experienced symptoms for less than a week, were randomized to receive either fluvoxamine daily for two weeks or a matching placebo. Four patients ended up being hospitalized in the placebo group, while not a single patient was hospitalized in the fluvoxamine group.
This was a small study, so the results were suggestive of benefit, but they hardly constituted proof of anything. Which is probably why the study ended up being ignored. Until now, that is.
Another study investigating fluvoxamine as a treatment for covid-19 recently published its results in Lancet Global Health. Like the earlier study, it was a double-blind randomized controlled trial. But it was bigger, much bigger. 1,497 patients were randomized to receive either fluvoxamine or placebo. For a randomized trial that is a big number.
The trial was carried out in Brazil. It recruited adults with symptoms suggestive of covid-19 and a positive covid antigen test, who had started to develop symptoms less than a week earlier. As in the previous trial, participants could not be sick enough that they required hospitalization on inclusion in the study, since the whole point of the trial was to see whether the drug decreases risk of hospitalization. The average time after symptom onset at which participants were included was four days.
In order to be included in the study, participants also had to have some underlying health condition that would predispose them to more severe disease – this is a sensible decision, because very few people with covid-19 actually end up requiring hospitalization, so even with 1,500 people it’s questionable whether you would find a statistically significant difference in hospitalizations, even if a drug works, if you didn’t restrict yourself to high risk individuals.
Participants in the intervention group received 100 mg of fluvoxamine twice per day for ten days, while those in the placebo received an identical placebo. So, how did things turn out?
10.3% of participants in the fluvoxamine group ended up being hospitalized, compared with 13.1% of participants in the placebo group. If the effect is real, that would mean that for every 35 people with covid that you treated with fluvoxamine, you could prevent one hospitalization. Which doesn’t sound extremely impressive, but it’s not nothing. The relative risk reduction is 22%, which would mean that roughly one in five hospitalization could be prevented – for a stretched health care system in the middle of a big covid wave, that could be a meaningful differnce.
But is the effect real?
Well, that’s hard to say. The result isn’t statistically significant (p-value 0.09), which could mean that the study was simply too small to find a difference, or it could mean that the difference found in the study was due to chance and there is in fact no real world difference.
It’s worth noting that the authors themselves claim that there is a statistically significant difference, and that 11% in the fluoxetine group suffered a “primary event”, compared with 16% in the control group. However, this is based on some statistical shenanigans, where the researchers have combined hospitalizations with emergency room stays longer than six hours. From my perspective, spending an extra hour or two in an emergency room is a completely different thing from being hospitalized, and the two shouldn’t be conflated. They certainly shouldn’t be combined in to a composite end point.
Additionally, the primary endpoint was changed to its final form (hospitalizations + emergency room visits longer than six hours) in March, when the authors had already started gathering data, which is very suspicious, and suggests that they chose the primary end point after they’d looked at the data and seen what would generate a “statistically significant” result. This is a big no-no – you’re supposed to choose the primary end point before you start gathering data, and then stick with it.
Otherwise it’s too easy to cheat and mine the data until you get a result that seems important, and then claim that what you discover is what you were looking for all along. So, as usual, the best thing to do is to ignore what the authors claim about their results, and instead look directly at the data. And what those data show is, as mentioned, a 10% vs 13% difference in hospitalizations that isn’t statistically significant.
What about deaths?
2.2% of participants in the fluvoxamine group died, compared with 3.3% in the placebo group. If real, that would mean that you could prevent one death for every 99 high-risk adults with early stage covid that you treat with fluvoxamine. That might not sound extremely impressive, but it would also mean that 31% of covid deaths could be prevented simply by giving people fluvoxamine at an early stage in the disease course. As before, however, the difference doesn’t reach statistical significance (p-value 0.24). Whether that means the result is due to chance, or that the study was simply too small to detect a difference, even though one exists, is impossible to say.
So, what can we conclude?
This study has generated a lot of buzz, but that buzz is based on a composite outcome of questionable validity. The results are certainly trending towards a benefit, but more data is needed before a firm conclusion can be drawn as to whether fluvoxamine has any role in treating covid-19 or not.
Sebastian Rushworth M.D. is a physician from Stockholm, Sweden. This article was originally published on his blog.