The short version
- 399 adults in Brazil with fatigue lasting at least 90 days after a confirmed SARS-CoV-2 infection were randomly assigned to fluvoxamine, metformin, or placebo for 60 days.
- Fluvoxamine won. The analysis put the probability that it beats placebo at 99 percent.
- The margin shrank as time passed: about 50 percent more likely than placebo to report low fatigue at day 30, 36 percent at day 60, 19 percent at day 90.
- Metformin did nothing at all.
- Published in Annals of Internal Medicine on 18 July 2026, DOI 10.7326/ANNALS-25-03959. The trial is called REVIVE-TOGETHER and McMaster University co-led it.
What was actually tested
Two generic drugs, head to head against placebo, in people who were already sick and had been for months. That last part matters more than it sounds. Most of the early COVID drug trials, including the earlier TOGETHER studies fluvoxamine came out of, tested whether a drug could stop an acute infection from getting worse. This one asked something harder: can anything help a person whose infection cleared a long time ago and whose exhaustion did not.
Participants took two pills a day for 60 days. The main measure was the Fatigue Severity Scale, where a score of 3 or below counts as being in decent shape. Researchers checked in at day 30, day 60, and then again at day 90, a month after everyone had stopped taking anything.
The number that most coverage skipped
Look at the three timepoints in order.
Day 30: fluvoxamine patients were nearly 50 percent more likely than placebo patients to be at 3 or below.
Day 60: 36 percent more likely. At that point the drug had also pulled average fatigue ratings down by close to half a point on the scale.
Day 90, thirty days after the last pill: 19 percent more likely.
The effect did not vanish, but it clearly leaned on the drug being in the person’s system. That reads less like a cure and more like symptom control, the way an inhaler controls asthma. Nobody has tested what happens if you keep taking it for six months, or a year, or what happens when you stop after that.
Why metformin failing is the interesting half
Metformin had a genuine case behind it. There is a well-argued hypothesis that long COVID involves metabolic dysfunction, and metformin already had trial evidence for preventing long COVID when given during an acute infection. So this trial was a fair test of a real idea.
It came back flat. Metformin did not beat placebo on fatigue in people who already had the condition.
That is a useful negative. Preventing something and treating it once established are different problems, and this result says the metabolic route does not obviously work for the second one. Negative arms rarely make headlines, which is exactly why they end up being repeated at great expense.
What “99 percent probability” means and does not mean
The trial reported a Bayesian probability, not a p-value. Plain version: given the data collected, there is a 99 percent chance fluvoxamine is better than placebo on this measure.
That is a statement about direction. It is not a statement about size. A drug can be almost certainly better than nothing and still deliver a benefit too small for a patient to notice on a Tuesday. Half a point on a seven-point fatigue scale is real, and it is also not the same thing as getting your life back. Both are true.
Why a cheap drug is the good news and the problem
Fluvoxamine is off-patent. It costs very little and it is stocked almost everywhere, including in health systems that will never see a bespoke long COVID biologic. For a condition that hits working-age adults hardest and that most health systems have quietly given up on, an available drug beats a theoretical one.
The flip side is structural. No company earns anything from proving fluvoxamine works, so the confirmatory trial that would normally follow a result like this has no obvious funder. Cheap generics with promising single trials have a long history of sitting in that limbo for years.
The honest gaps
- Every participant was in Brazil. Generalisability to other populations is untested.
- The coverage I worked from does not state the fluvoxamine dose. The published paper does; the news write-ups do not.
- Fatigue only. Long COVID also involves breathlessness, pain, autonomic problems and cognitive symptoms, none of which this endpoint captures.
- Fluvoxamine is an SSRI, and the fatigue improvement could in principle run partly through mood. Whether the trial separated those is a question for the paper.
- Fewer adverse events on the drug than on placebo is a genuinely odd finding. It happens, but it deserves a look at the full safety table rather than a shrug.
What to do with this if you have long COVID
Talk to a doctor. Fluvoxamine is a prescription antidepressant with real interactions, including with a long list of common medications, and it is not approved for this use anywhere. One trial in one country is a reason to have the conversation, not a reason to self-treat.
Sources:
- Annals of Internal Medicine, REVIVE-TOGETHER trial, published 18 July 2026 (https://doi.org/10.7326/ANNALS-25-03959)
- CIDRAP, University of Minnesota, “Antidepressant relieves fatigue in people with long COVID, study finds” (https://www.cidrap.umn.edu/covid-19/antidepressant-relieves-fatigue-people-long-covid-study-finds)
- ScienceDaily, “Common antidepressant may ease long COVID’s crushing fatigue” (https://www.sciencedaily.com/releases/2026/07/260718010158.htm)
Disclaimer: This article is for general information only and does not constitute medical advice, diagnosis, or treatment. Always consult a qualified healthcare provider about any medical condition or before making health decisions.