Long COVID fatigue trial retracted and republished: fluvoxamine result intact, side-effect figures revised

In March, a trial in the Annals of Internal Medicine was presented by its authors as the first strong evidence that a drug can ease the exhaustion of long COVID. This week the journal retracted that paper and published a replacement.

The central result has survived. People given the antidepressant fluvoxamine reported somewhat less fatigue than people given a placebo, by the same margin as before. Other things have changed. Side effects are reported at higher rates in every group, the account of the second drug tested, metformin, is different, and the conclusion is more cautious.

What was retracted and republished

The retraction is recorded in PubMed, the index of medical research run by the US government, as a Notice of Retraction and Republication dated 6 October 2026. It applies to the report that appeared in the journal’s May issue, which had first been released online on 31 March. A new version of the article was published the same day as the notice.

Retraction with republication is a recognized way of correcting the record. The committee of journal editors that sets publishing standards says it can be considered in cases where honest error (e.g., a misclassification or miscalculation) leads to a major change in the results, interpretations, or conclusions. The conditions it gives are that the underlying science appears valid, and the changed version of the paper survives further review and editorial scrutiny.

If the journal applied that standard, something in the first version changed in a major way. Which change prompted the journal to act is not something a comparison of the two versions can establish. What the comparison does show is where they differ.

What changed in the republished trial

The first version’s figures are given below as they were reported at the time by 2 Minute Medicine and EMJ, two medical news services.

Item	As first reported	Republished version
Fatigue, fluvoxamine versus placebo	0.43 points lower at day 60; 0.58 at day 90	0.43 points lower at day 60; 0.58 at day 90
Metformin dose	750 mg twice daily	750 mg twice daily for 71 people, then 500 mg twice daily for 40
Side effects, fluvoxamine	20.0%	24.0%
Side effects, metformin	28.8%	45.1% at the higher dose; 42.5% at the lower
Side effects, placebo	29.7%	35.5%
Verdict on metformin	No benefit	Inconclusive

The fatigue result is the same. In March, 2 Minute Medicine gave the difference at day 60 as 0.43 points, with effects maintained at day 90, when it was 0.58. The republished paper gives identical numbers. In its words, the comparison of fluvoxamine with placebo showed a mean difference of -0.43 at day 60 and -0.58 at day 90.

The side-effect figures are not. The same March report said that adverse events occurred in 20.0% of fluvoxamine, 28.8% of metformin, and 29.7% of placebo participants. The republished paper gives higher rates throughout. Incidence of adverse events was lower for fluvoxamine (24.0%) than for metformin, it says. The metformin rates were 45.1% on the higher dose and 42.5% on the lower, and the placebo rate was 35.5%. Fluvoxamine still has the lowest rate. Metformin, which had looked level with placebo, now sits well above it.

Two things about those figures need saying. An adverse event is any medical problem recorded during a trial, whether or not a pill caused it, and a rate below placebo’s is unusual for an active drug. And on this article’s arithmetic, the old metformin figure matches 32 of the 111 people assigned to that drug, while the new ones match 32 of the 71 on the higher dose and 17 of the 40 on the lower.

The metformin group is described differently. In April the trial was reported with a single metformin regimen: participants received fluvoxamine (100 mg twice daily), metformin (750 mg twice daily), or placebo for 60 days. The republished paper says that 71 people took extended-release metformin at that dose between October 2023 and April 2024. The dose was then cut to 500 mg twice daily for another 40 people. That followed a formal change to the trial’s plan, and it lasted until the group was stopped for futility in September 2024. Futility means that continuing was judged unlikely to show a benefit.

The new version explains the consequence. The results were inconclusive for metformin because of a dose change recommended by the data safety and monitoring committee due to adverse effects causing study withdrawal and early termination. That committee is the independent panel that watches over participants in a trial. Put simply, side effects were leading people in the metformin group to drop out or stop treatment, and the committee recommended a lower dose. The group was later stopped. Neither dose got a clean test.

So the verdict on metformin has softened. In April the reported result was that metformin showed no meaningful benefit on fatigue at either time point. The republished paper says only that its results for metformin were inconclusive, which is a statement that the question was not answered.

The conclusion is more guarded. When the first version appeared, one of its authors, Jamie Forrest of the University of British Columbia, said in a university news release: “This trial gives clinicians their first strong evidence for a medication that helps reduce long COVID fatigue.” The republished paper ends with the statement that fluvoxamine shows potential for improving long COVID-associated fatigue and quality of life.

How the fluvoxamine and metformin trial was run

The study, a trial called REVIVE-TOGETHER, took place at outpatient sites in Brazil. It enrolled 399 adults with fatigue 90 days to 24 months after confirmed infection with the coronavirus. In its first-reported form, 399 participants were randomized to fluvoxamine (n = 150), metformin (n = 111), or placebo (n = 138), and the metformin numbers in the new version add up to the same 111.

Fluvoxamine was given at 100 mg twice a day. Each treatment was taken for 60 days with follow-up to day 90. Neither participants nor the staff assessing them knew who was taking what. The research was funded by The Latona Foundation.

The measure was change in Fatigue Severity Scale (FSS) score. People rate their own exhaustion on it, and seven points on the scale indicates the worst fatigue. Participants started high. Median baseline FSS scores ranged from 5.6 to 5.9 across all study groups.

The design was adaptive, which the university described as allowing researchers to stop treatment arms early when results became clear. Arms are the groups in a trial. The analysis expresses each result as a probability that the drug is better than placebo. For fluvoxamine, the results at day 60 and at day 90 both had a posterior probability of superiority of 99%. That is a 99% chance, on the trial’s statistical model, that the drug did better than placebo. Fluvoxamine also improved quality-of-life scores with a high posterior probability of superiority.

Both drugs are old, cheap and widely available. Fluvoxamine is a selective serotonin reuptake inhibitor (SSRI), which works by changing the activity of serotonin, a chemical that affects mood and other functions. Metformin is a standard diabetes medicine.

Each had been tested before in COVID-19. In an American trial of adults treated within days of infection, outpatient treatment with metformin reduced long COVID incidence by about 41%, an absolute fall of about four cases in every hundred people treated. That was prevention, in people who were newly infected. The Brazilian trial asked a different question, whether the drug helps once long COVID is established. In the American trial, fluvoxamine did not prevent the condition. There was no effect on cumulative incidence of long COVID with ivermectin, a third drug in that trial, or with fluvoxamine.

Few drugs have been tested this way against established long COVID, and at least one earlier attempt disappointed. It was a 15-week blinded, placebo-controlled, randomized clinical trial conducted from November 2022 to September 2023 at Stanford University, in which a 15-day course of the antiviral sold as Paxlovid made no measurable difference to a combined score of six symptoms.

How large the fatigue result is

The difference between fluvoxamine and placebo, 0.43 points at the end of treatment, is less than half a point on a seven-point scale, from a starting point near 5.7.

Alba Azola of Johns Hopkins University School of Medicine, who oversees long COVID clinics and was not involved in the research, was restrained when the first version appeared in March. “I don’t want to be pessimistic, but that particular intervention is not something that is getting me excited,” Azola told CIDRAP News, an infectious disease news service at the University of Minnesota. “Overall, the findings suggest a potential modest benefit,” Azola said.

Azola also said more work was needed to show what a change on a seven-point scale means in patients’ everyday lives. Those comments concerned the fatigue estimate, which is the part of the paper that has not changed.

The effect was measured for a month after the pills stopped and no longer. The paper is plain about it. The benefit of fluvoxamine after 90 days is unknown.

What the republished fatigue trial still cannot show

Durability. Long COVID can last for years. The trial covers three months.

Other symptoms. Fatigue is one feature of the condition among many. The trial did not address other long COVID symptoms.

What is being treated. Fluvoxamine is an antidepressant, and low mood and fatigue overlap. EMJ noted in April that people with major depression were excluded but those with symptoms of depression were not, leaving it unclear if fatigue improvements reflect direct benefit or secondary mood effects.

Where it was done. All participants were recruited in Brazil. Results may differ in other populations and health systems.

Metformin. The metformin group was stopped early because it looked unlikely to succeed, but its dose had been changed along the way, and the authors call the result inconclusive. An inconclusive result is not a negative one. The trial does not show that metformin fails to help established long COVID fatigue. It shows that this trial could not tell.

The paper’s history. The fatigue estimate is the same in both versions. It now rests on a report that had to be replaced, and it comes from a single trial. Readers who saw coverage between March and early October, including a side-effect rate of 20.0% for fluvoxamine, were reading figures from the withdrawn version.

Serious harms were uncommon in the republished data. Grade 3 or higher events, the more severe kind, were rare, at 2.7% or less in every group, with no treatment-related deaths.

Whether fluvoxamine is appropriate for a person with long COVID depends on their other conditions and medicines, which is assessed by their doctor. Do not stop or change a prescribed medicine without talking to your doctor.

A trial reporting that fluvoxamine eases long COVID fatigue has been retracted and republished with the fatigue result unchanged at a little under half a point on a seven-point scale, higher reported side-effect rates in every group and an inconclusive verdict on metformin, and it still covers only 90 days in one country.