Fatigue is the symptom that most often keeps people with long COVID out of work, and there is still no established drug treatment for it. A trial published this year is one of the more serious attempts to find one, and its result is genuinely positive and genuinely small.

The study, published in the Annals of Internal Medicine on March 31 and led by Gilmar Reis, randomized 399 adults in Brazil who had fatigue persisting at least 90 days after a confirmed coronavirus infection. Participants received fluvoxamine, an inexpensive and long-established antidepressant, at 100mg twice daily; metformin, a common diabetes drug, at 750mg twice daily; or a matching placebo, for 60 days. Enrolment ran across 22 outpatient sites between October 2023 and February 2025.

The result

At day 60, fluvoxamine produced a mean reduction in fatigue score of 0.43 points against placebo, with a 95 percent credible interval from −0.80 to −0.07. At day 90, a month after treatment stopped, the difference was 0.58 points, credible interval −0.98 to −0.16.

Those are credible intervals rather than confidence intervals because the trial used a Bayesian adaptive design. In that framework the researchers report a 99 percent probability that fluvoxamine beat placebo, which is a different statement from a conventional p-value and should not be read as a 99 percent chance the drug works for any given patient.

Metformin showed no meaningful benefit.

How large is 0.43 points

This is the part that matters most and is easiest to lose. The primary outcome was a self-reported fatigue scale. A shift of roughly half a point on such a scale is detectable statistically and sits at or below the threshold most researchers would call clinically meaningful for an individual.

The authors say so themselves, acknowledging uncertainty about whether the improvements translate meaningfully into patients' daily functioning, as reported by CIDRAP at the University of Minnesota.

On safety, adverse events were recorded in about 20 percent of the fluvoxamine group against about 30 percent on placebo, which does not suggest the drug was poorly tolerated in this population.

What it did not measure

The trial looked at fatigue and nothing else. It did not assess post-exertional malaise, the worsening of symptoms after exertion that many patients and clinicians regard as the defining feature of the condition. Nor did it measure cognitive dysfunction, autonomic problems or pain.

There were no biological measures. Outcomes were self-reported, with no biomarker or imaging to indicate how the drug might be acting.

People with existing depression or anxiety disorders were excluded, which narrows how far the result generalizes, since mood disorders are common among long COVID patients.

The independent view

Researchers not involved in the work told the Science Media Centre that the finding needs replication in broader patient groups and with outcomes covering more of the condition.

The ME Association, which represents patients with myalgic encephalomyelitis and chronic fatigue syndrome, noted that a significant proportion of people with long COVID improve over time regardless of treatment, which complicates attributing gains to a drug, and that metformin's failure raises questions about the biological premise the trial was testing.

There is relevant history. Serotonergic antidepressants have been trialled for decades in ME/CFS without establishing a clear benefit, and some clinicians caution that patients with ME/CFS-like presentations can react poorly to them.

Where this leaves things

A single trial, in one country, with a self-reported primary outcome and an effect at the edge of clinical relevance, is a reason for further research rather than a change in practice. It is not evidence that anyone should seek or start this medication, and fluvoxamine is a prescription drug with its own interactions and withdrawal profile.

What the result does is narrow the search. Fatigue in long COVID has resisted almost everything tried against it, and a signal that survives a placebo-controlled design is worth pursuing properly, with larger trials, longer follow-up and outcomes that reflect what the illness actually does to people.