More Data Suggesting SSRIs May Provide Some COVID Protection

Could existing, cheap, and widely-available drugs save lives?

The hope that existing medications would prove to be effective treatments for COVID springs eternal, and a host of promising agents have come and, after rigorous study, gone.

These include classics like hydroxychloroquine and ivermectin, and slightly less-celebrated agents like melatonin.

There has even been one undeniable hit — dexamethasone — which in multiple randomized trials now has demonstrated its ability to reduce mortality in individuals with severe COVID.

Over the past year, support for one additional agent — fluvoxamine — quietly built on the backs of several small studies until, in September, we got data from the large TOGETHER trial — which found that the use of fluvoxamine in early COVID-19 was associated with a significant reduction in the risk of prolonged ED visit or hospitalization.

There were numerically fewer deaths in the fluvoxamine arm as well.

Fluvoxamine is cheap and widely available, but, let’s be honest — it does not get prescribed a lot compared to other SSRIs. So the results of the TOGETHER trial prompted a straightforward question. Is the fluvoxamine effect a class effect? In other words, would any SSRI — with their class-level anti-thrombotic and anti-inflammatory effects — work?

The first answers are now trickling in, with the largest study to date examining this question appearing in JAMA Network Open.

Now, this is not a randomized trial. It’s a propensity-score matched cohort study. But it’s big — representing data from 87 health care centers across the US. Here’s how it worked.

Researchers collected data from the electronic health record on 83,584 individuals with COVID-19 from January to September of 2020 — so just to orient ourselves this is the pre-vaccine era and really a time before SSRIs were on the radar as a treatment. The cohort included hospitals, emergency rooms, and urgent care — but this was a pretty sick group — the overall mortality rate was 8.6%.

Now, about 4% of the cohort was receiving SSRIs on or around the time they were diagnosed. And I’ll pause there to note that is quite a bit lower than expected. Around 13% of adults in the US take SSRIs.

Does that mean SSRIs are protective even against COVID infection? The study was not designed to answer that question but… it’s certainly an intriguing possibility.

What the study was designed to answer was whether, among people infected with COVID-19, the SSRIs were beneficial. The raw numbers don’t look good for the anti-depressants. The mortality rate was 14.6% among those taking SSRIs compared to 8.4% among those not taking SSRIs.

But don’t close the book on them just yet. People who take SSRIs are different than those who don’t. In this cohort, those on SSRIs were twice as likely to have coronary vascular disease and COPD, and, unsurprisingly, 3 times as likely to have a mood disorder. 74% of those on SSRIs were hospitalized when they entered the cohort compared to just 44% of those not on SSRIs.

Propensity-score matching tries to compare apples to oranges by finding oranges that look a lot like apples. You build a multivariable statistical model predicting SSRI use — depression would be a major predictor, for example. You then predict the probability of SSRI use for everyone in the cohort — whether they truly used them or not. You then match those that used them to people who don’t, but have a similar propensity score for use. Propensity-score matching evangelists would say this is like a clinical trial — everything about the patients is basically the same after matching, except one person got the drug and one didn’t.

It’s not really a randomized trial, but it is better than looking at the raw data.

And, after propensity score matching, we find that, overall, SSRI use in general was associated with a reduction in mortality — by a modest, but statistically significant 8%.

What about fluvoxamine? Who knows — only 11 out of the 3,401 patients on SSRIs were taking that particular SSRI.

But a strong beneficial effect was seen with fluoxetine, which is sold under the brand name Prozac — with a 9.8% mortality rate compared to 13.3% among propensity-matched controls.

The authors don’t break down the other SSRIs individually, but as a group they didn’t do great. Mortality rate 15.4% compared to 17% in propensity-matched controls.

So — do SSRIs work for COVID-19? Do some work? Does only fluvoxamine work? There seems to be some promise here, but not at the breakthrough level.

And of course, the field upon which we judge these drugs is changing. If Pfizer’s protease inhibitor proves to be as effective as their press release suggests it is, 30% mortality reductions from SSRIs may seem quaint. Still, it’s nice to have some more potential arrows in our quiver, and encouraging to see the search for cheap, effective, and widely available COVID treatments to continue.

A version of this commentary first appeared on