ABOVE: Molecular structure of fluvoxamine © ISTOCK.COM, STEVEN_MOLINA

Update (December 16): The National Institutes of Health in updated COVID-19 treatment guidelines today states that “There is insufficient evidence for the [panel] to recommend either for or against the use of fluvoxamine for the treatment of COVID-19.” The agency lists a number of difficulties in interpreting the TOGETHER trial data, including differences in adherence between the treatment and placebo groups, a risk of bias associated with the per-protocol analysis, and an endpoint with unclear clinical relevance. Meanwhile, the STOP COVID 2 trial has “been stopped for futility by a data safety monitoring board after lower than expected case rates and treatment effect were observed,” according to the update.

A highly publicized new study of the antidepressant fluvoxamine as a potential COVID-19 treatment suggests that the drug could reduce hospitalizations among people with the disease, but has prompted calls for caution in interpreting the findings.

A report on the randomized, placebo-controlled trial, published last Wednesday (October 27) in The Lancet Global Health, included data from 1,497 people infected with SARS-CoV-2, 741 of whom were given fluvoxamine, and concluded that the drug “reduced the need for hospitalisation defined as retention in a COVID-19 emergency setting or transfer to a tertiary hospital.” 

Some of the authors, and many news stories reporting the findings, have been extremely positive about the results. In a tweet posted Thursday, study coauthor Eric Lenze, a geriatric psychiatrist at Washington University in St. Louis (WUSTL), shared the “Great news!!!” with a link to the journal article and a WUSTL-branded image containing simplified bar charts implying that the drug reduced hospitalization by “as much as 66%” and slashed mortality by “as much as 91%.” The tweeted image also contained a URL for a WUSTL page with “prescribing instructions.” 

In another tweet, WUSTL child psychiatrist and study coauthor Angela Reiersen called fluvoxamine “an inexpensive pill that works against Covid-19.” 

However, other researchers involved in designing and analyzing clinical trials urge caution in interpreting the data. 

In a comment published in the Lancet Global Health the same day as the study, Otavio Berwanger, a clinical researcher and director of the Academic Research Organization at Hospital Israelita Albert Einstein in Brazil, tempered praise of the research with the conclusion that “the definitive answer regarding the effects of fluvoxamine on individual outcomes such as mortality and hospitalisations still need addressing,” and called “the inconclusive effects on patient-important outcomes such as hospitalisation and mortality” one of the study’s main limitations.

James Watson, a statistician and data analyst at the Mahidol Oxford Tropical Medicine Research Unit who has been involved in the design and analysis of clinical trials for drugs tackling COVID-19 and other infectious diseases, is more critical, telling The Scientist that claims circulating online that the drug reduces mortality have been made by “cherry-picking” parts of the results. He remains skeptical that fluvoxamine truly shows efficacy against the disease, he adds.  

“This is a drug where there’s no clear mechanism of action, so a priori there’s a very low chance that this random molecule . . . might actually do anything useful in COVID-19,” he says. The new study, while suggestive of a benefit to fluvoxamine, is far from conclusive, he argues. 

Fluvoxamine as a COVID-19 drug candidate

The Lancet Global Health paper, which was initially posted as a preprint in August, was completed as part of the international TOGETHER trial, which is investigating multiple potential treatments for COVID-19 and updating its protocols as evidence accumulates. The fluvoxamine part of the study, which built on earlier work by Lenze and Reiersen, was carried out across 11 clinical sites in Brazil.

Cherry-picking the per-protocol odds ratio for mortality is really misleading. . . . It’s not ok.

—James Watson, Mahidol Oxford Tropical Medicine Research Unit

It’s one of several trials of the antidepressant, a type of selective serotonin reuptake inhibitor (SSRI), that Reiersen and Lenze are involved in. In an article titled “Following Your Instincts” in Cerebrum magazine earlier this year, Reiersen describes how she started wondering whether fluvoxamine might treat COVID-19 when she herself was unwell with the disease in 2020. 

She, Lenze, and colleagues carried out a small trial of fluvoxamine, published in JAMA last Novemberthat reported an apparent benefit of the drug in reducing the likelihood of clinical deterioration in COVID-19 patients, although its limited sample size and missing data due to some people failing to fill out follow-up surveys made it impossible to draw definitive conclusions about fluvoxamine’s efficacy, notes Berwanger, whose research is partly supported by pharmaceutical companies, in his commentary. 

Lenze and Reiersen—who are both named as inventors on a WUSTL patent application for fluvoxamine as a COVID-19 treatment, according to the new Lancet Global Health study—are also involved in another fluvoxamine trial at WUSTL called “STOP COVID 2.”

Potential mechanisms for the drug as an anti-COVID-19 treatment are currently unclear, although Reiersen and colleagues have speculated that the drug may have anti-inflammatory action that could influence the immune response to the SARS-CoV-2 virus.

Measuring success

The current study combined two measures of treatment success into what’s known as a composite endpoint: first, the drug’s ability to reduce a COVID-19 patient’s “need for emergency care AND observation for more than 06 hours due to the worsening of COVID-19,” and second, its ability to reduce a patient’s “need for hospitalization due to COVID-19 progression.” The study also measured secondary outcomes including mortality, time spent in intensive care, and change in viral load.

Maarten van Smeden, a medical statistician at University Medical Center Utrecht in the Netherlands, notes that it’s common to choose composite endpoints and/or somewhat subjective outcomes in trials such as these, as on their own, deaths are too rare an event to provide enough statistical power to detect differences between a treatment and a placebo. 

He says that in his view the researchers have done a good job of analyzing their data, adding that they provide solid statistical support to indicate that the drug is beneficial in stopping deterioration in people with COVID-19. Specifically, the paper reports that while 16 percent of people in the placebo group needed hospitalization or observation, only 11 percent of those who received fluvoxamine did. 

But Watson, who is currently working on the PLATCOV trial investigating early treatments for COVID-19, questions the viability of relying on a measure such as a “need for observation,” which he says “is starting to get pretty weak in terms of objectivity,” not least because it’s a metric that can easily be affected by patient and physician behavior.

He notes that the study’s main result—that people were more likely to experience one of the outcomes if they were taking the placebo rather than fluvoxamine—is largely driven by differences in this “need for observation” measure rather than differences in the likelihood of being hospitalized. King’s College London visiting professor and pharma consultant Penny Ward raises the same point in comments to The New York Times and in a statement to the Science Media Centre.

Reiersen and colleagues have defended the use of the metric, noting that, at the sites included in the study, continued medical observation was essentially equivalent to being hospitalized. “These were not standard emergency rooms but instead were COVID-19 emergency centers that were set up due to hospitals being overloaded,” Reiersen writes in an email to The Scientist. “A stay in these centers >6 hours was an indication that the patient was receiving care equivalent to hospitalization.”

Watson says that his concerns are compounded by the fact that the researchers observed differences in adherence between the treatment and placebo groups: people were significantly more likely to stop taking their pills if they were on fluvoxamine—a difference that he says suggests that participants, and perhaps also their physicians, could have figured out what treatment group they were in. (Like other SSRIs, fluvoxamine is associated with various side effects, particularly at the beginning of treatment—which may have tipped off patients and their doctors.) 

Awareness of whether patients were receiving treatment or placebo could have affected how long people received medical observation, Watson argues, adding that studies typically try to rule out or at least account for these differences in order to reduce the possibility of bias. Overall, he says, “that’s the thing I’m really worried out—that all this is due to differences in behavior rather than true clinical outcomes.” 

Lenze and Reiersen agree that the differences in adherence between the two groups are likely related to the drug’s side effects. However, Reiersen tells The Scientist that while it’s possible some people could have figured out what they were taking, “some of the common side effects of fluvoxamine (such as nausea and other gastrointestinal symptoms) can also occur as symptoms of COVID-19, so it could still be difficult for participants to guess which treatment they were getting.”

Other outcomes

While some of the discussion of the study’s findings has focused on hospitalizations and medical observations, both Reiersen and Lenze, as well as many news outlets, have implied that the study also demonstrates that fluvoxamine reduces mortality among COVID-19 patients—a claim that is not explicitly highlighted in the paper itself.

This represents an estimate of the best case scenario that could be seen in highly adherent patients.

—Angela Reiersen, WUSTL

To study possible effects on mortality, the trial investigators used two types of analysis. First, in a so-called intention-to-treat analysis, they used data from all randomized participants and found that 25 people died in the placebo group, while 17 died in the fluvoxamine group—a statistically insignificant difference, according to the researchers’ analysis. 

Then, using what’s known as a per-protocol analysis, they included only those people who completed more than 80 percent of the course of fluvoxamine or placebo, and found that 12 people died in the placebo group, while 1 died in the fluvoxamine group. This analysis did show a significant treatment effect, and it’s this second set of data that led to the calculation of a 91 percent reduction in mortality that was included on the WUSTL-branded graphic tweeted by Lenze and others.

Van Smeden notes that there’s often debate among clinical researchers about when and whether to use per-protocol analyses. However, he adds that in this case, the numbers of patients are too low to draw meaningful conclusions, and calls the mortality claim “a bit optimistic.”

Watson argues that per-protocol approaches can be highly prone to bias—the risk being that a participant’s probability of completing a treatment course is related to their probability of getting a particular treatment outcome. He adds that there’s a “massive difference in the number of deaths”—17 versus 1—in the fluvoxamine group depending on the type of analysis, a difference that isn’t deeply explored in the study. 

In his view, all these factors combine to make the 91 percent mortality reduction claim “dodgy,” he says, adding that it shouldn’t be presented on its own without broader context. “Cherry-picking the per-protocol odds ratio for mortality is really misleading. . . . It’s not ok.” 

Asked about the accuracy of the 91 percent claim, Reiersen writes that the team explicitly used the phrase “as much as” when citing that statistic in the WUSTL graphic. “This represents an estimate of the best case scenario that could be seen in highly adherent patients,” she adds.

What’s next for fluvoxamine?

Watson tells The Scientist that one effect of the new fluvoxamine study has been to encourage other trial organizers to include SSRIs in their own studies, noting a trial he is involved with in sub-Saharan Africa as an example. “You get this positive result and then it becomes this new fashion and then all the trials want to try it out,” he says. While he supports the idea of a larger trial of fluvoxamine, for now, he says, “I definitely think this does not support changing [treatment] guidelines.”

Reiersen and Lenze express a different view, saying that they want to see fluvoxamine being prescribed as a COVID-19 treatment based on current evidence. Lenze, who calls the drug “life-saving” in an email to The Scientist, notes that fluvoxamine is easy to take and has a good safety record. “Doctors can prescribe it now to reduce their patients’ risk of hospitalization and death,” he says.

“I think based on the current evidence it makes sense for physicians to go ahead and prescribe fluvoxamine for certain high risk outpatients with symptomatic COVID-19, after considering the current evidence, any potential drug interactions, and the risks & benefits specific to their patients,” agrees Reiersen, who has previously also voiced her support for the use of ivermectin, an antiparasitic drug that has yet to show efficacy against COVID-19 in well-controlled randomized clinical trials.

See “Ivermectin (Still) Lacks Support as a COVID-19 Drug

“I look forward to seeing whether [the National Institutes of Health] adjusts their guidelines to support this,” she says of fluvoxamine, adding that she is currently collaborating with researchers at various institutions to work toward clarifying its mechanisms of action in lab studies. There could be other benefits of the drug, she suggests: “I also hope to start a randomized controlled trial using fluvoxamine to treat long COVID.”