Antidepressant Cuts Long COVID Fatigue. Metformin Not So Much.

Some symptoms announce themselves loudly. Long COVID fatigue doesn’t. It accumulates: a heaviness that makes getting dressed feel like a project, that turns the walk from bedroom to kitchen into something requiring planning. For tens of millions of people, this has been the reality for two years, three years, more. Not the dramatic collapse of acute illness, but the slow erasure of ordinary life. And until a few days ago, no medication had been shown in a rigorous clinical trial to meaningfully reduce it.

That has changed. A randomized, placebo-controlled trial published March 31 in the Annals of Internal Medicine found that fluvoxamine (a decades-old antidepressant sold as Luvox, inexpensive and available in most countries) significantly reduced fatigue severity in people with long COVID over a 60-day course, with effects that persisted for another month after the treatment ended.

The drug belongs to the same family as Prozac: selective serotonin reuptake inhibitors, or SSRIs. But fluvoxamine has an unusual secondary profile that made it, perhaps, a more interesting candidate here than other members of the class. It binds strongly to a receptor called sigma-1, which sits at the intersection of immune regulation and cellular stress response. Earlier in the pandemic, multiple trials found that fluvoxamine cut hospitalization risk in people with acute COVID-19, a finding attributed, at least in part, to its anti-inflammatory and immunomodulatory properties. Whether that same biology might help with the chronic aftermath was the question the REVIVE-TOGETHER trial set out to answer.

The trial enrolled 399 adults in Brazil, across 22 outpatient sites in and around Belo Horizonte, between October 2023 and early 2025. All had experienced fatigue persisting for at least 90 days after a confirmed SARS-CoV-2 infection, and all were scoring in the moderate-to-severe range on the Fatigue Severity Scale, a nine-item self-reported measure where a 5.6 out of 7 is about where participants averaged at baseline. They were randomized to receive fluvoxamine (100 mg twice daily), metformin (a common diabetes drug with anti-inflammatory properties that had separately shown promise in preventing long COVID), or placebo for 60 days.

Metformin didn’t help. Its arm was actually stopped early, after the first interim analysis, when it became clear the drug was offering nothing meaningful to people with established long COVID fatigue; a result that sits in interesting tension with earlier evidence suggesting it lowers the risk of developing long COVID if taken during the acute phase. Different problem, it seems, different biology.

Fluvoxamine, though. At day 60, fluvoxamine showed a statistically meaningful reduction in fatigue compared with placebo, with a 99% posterior probability of superiority, the kind of number, in the Bayesian framework this trial used, that is about as unambiguous as clinical trial statistics get. By day 90, a month after treatment had ended, the benefit had actually grown slightly: a 99.7% probability of superiority, mean difference of roughly half a point on the FSS. Participants also reported improved quality of life at every time point measured. And, perhaps counterintuitively, those taking fluvoxamine had fewer adverse events than those on placebo (20% versus nearly 30%).

“This is an important step forward for patients who have been desperate for evidence-based options,” says Edward Mills, a professor at McMaster University’s Department of Health Research Methods, Evidence, and Impact and the trial’s co-principal investigator. Mills also noted that because fluvoxamine is already widely used and well understood, the path to clinical use is relatively short; no new safety infrastructure required, no regulatory novelty to navigate.

The trial’s design is worth dwelling on, not just as a procedural footnote but as part of the result itself. The REVIVE-TOGETHER trial used a Bayesian adaptive framework, meaning it could stop treatment arms early when the evidence crossed prespecified thresholds: futility in metformin’s case, for superiority in fluvoxamine’s. “The trial used a sophisticated adaptive design that allowed it to reach conclusions more efficiently than traditional trials, stopping early when the evidence was clear enough, a design innovation as important as the findings themselves,” says Gilmar Reis, the trial’s lead author and a researcher at Cardresearch in Belo Horizonte. Two of four planned interim analyses were conducted. The approach, borrowed from oncology trial methodology, is increasingly used in complex post-pandemic research where waiting years for a conventional trial result isn’t really an option for the patients involved.

There are real caveats here, and the authors don’t shy away from them. The trial was conducted exclusively in Brazil, in a predominantly female, Hispanic or Latino population, most of them in their mid-40s. Whether the results generalize to men, to older patients, to populations with different vaccination histories and different COVID-19 variant exposure patterns: genuinely unknown. The follow-up was only 90 days; no one knows whether the benefit holds at six months or a year. The trial measured fatigue and quality of life, not brain fog, not autonomic dysfunction, not the sleep disturbances or cardiovascular symptoms that compound long COVID into something more complicated than any single scale can capture. And the mechanism remains speculative. Whether fluvoxamine is working through its antidepressant properties, its sigma-1 receptor activity, its anti-inflammatory effects, or some combination of all three is unresolved. A sensitivity analysis found no evidence that baseline depression levels explained the fatigue improvement, but the question of how mood and post-viral fatigue interact isn’t settled by a single experiment.

Long COVID is thought to affect somewhere around 65 million people globally, though estimates vary considerably depending on how you define it. Most clinical guidelines still recommend only supportive care: pacing, symptom management, multidisciplinary support. Recent trials of nicotinamide riboside showed no improvement in fatigue, cognition or sleep. Antivirals have shown inconsistent effects on established long COVID. Low-dose naltrexone and several other candidates are still working their way through trials. Fluvoxamine is, right now, the thing with the most rigorous positive evidence behind it, even accounting for the Brazil-only enrollment.

“This trial gives clinicians their first strong evidence for a medication that helps reduce long COVID fatigue. Patients want something they can try today, and this finding brings us closer to that reality,” says Jamie Forrest, the trial’s corresponding author and a postdoctoral research fellow at the University of British Columbia.

What happens next is the part that should probably be said plainly. Fluvoxamine is not a cure. The effect size (roughly half a point on a seven-point scale) is real and meaningful, but it isn’t transformative for everyone who takes it. The durability beyond 90 days isn’t established. The subgroup that benefits most isn’t yet identifiable. What this trial has done is prove the concept: that a drug can move the needle on long COVID fatigue in a controlled setting, and that the Bayesian adaptive framework can do it faster than conventional trial design. That creates a template. What followsnotype-informed patient selection, combination approaches, is the actual work of turning a proof of concept into a treatment.

For the millions of people still waiting, that timeline matters enormously. A half-point improvement on a fatigue scale might sound modest. For someone who hasn’t been able to work in two years, it might be the difference between possible and not.

DOI: https://doi.org/10.7326/ANNALS-25-03959