Disputed trial weakens evidence for fluoxetine use for childhood depression


Two reviews’ estimates of fluoxetine’s benefits shrank when researchers removed a disputed trial. They seek a guideline review, not an end to prescribing.
A disputed trial involving 40 children in Iran may have overstated the evidence for treating childhood depression with fluoxetine, sold as Prozac. Researchers reran two major reviews without the trial and found that the drug’s unusually large benefits disappeared. Nature reported the findings on 2 October 2026; the analysis was published that month in Cochrane Evidence Synthesis and Methods.
Many national prescribing guidelines recommend fluoxetine for depression in children and adolescents, drawing on those earlier reviews. The researchers question whether the evidence sufficiently supports these recommendations. They want guideline committees to reconsider them, but do not call for clinicians to stop prescribing the medicine.
A meta-analysis combines results from multiple studies to assess the evidence for a treatment. The trial at issue, published in 2006, compared fluoxetine with another antidepressant, nortriptyline—not with a placebo, a treatment without an active drug. It reported an enormous advantage for fluoxetine that influenced the combined results of later reviews.
| Review | 2006 trial included? | Reported finding |
|---|---|---|
| 2016, The Lancet | Yes | Fluoxetine was better than placebo for depression in children and adolescents. |
| 2020, The Lancet Psychiatry | Yes | Supported the 2016 finding. |
| 2021, Cochrane Database of Systematic Reviews | No | Found a “small and unimportant” advantage over placebo; fluoxetine was probably no better than other antidepressants. |
The 2021 review did not omit the Iranian trial because it had judged the data untrustworthy. Its eligibility rules excluded comparisons with the class of antidepressants that includes nortriptyline. By removing that same trial from the 2016 and 2020 reviews, the latest researchers identified its inclusion as the reason for the differing conclusions.
Analysis
No other trial of fluoxetine for childhood depression has reported a treatment effect close to the Iranian trial’s result. The latest researchers also say its reported data imply that every child’s depression score changed by a similar amount within the fluoxetine group. The same pattern appeared within the nortriptyline group. They consider this unusually orderly pattern implausible.
The trial report also contained no description of its ethics-approval process. That is a gap in the reporting, not proof that approval was absent.
Elia Abi-Jaoude, a psychiatrist and clinician at Toronto’s Hospital for Sick Children, praised the reanalysis as a “slam-dunk study”.
Andrea Cipriani, an Oxford psychiatrist and co-author of the 2016 and 2020 reviews, told Nature that he agrees the trial data are problematic. He said the teams discussed the concerns at the time but could not justify excluding the trial without the current process for handling untrustworthy studies. Cochrane has since introduced a process to assess trial trustworthiness and potentially exclude studies from reviews.
According to Cipriani, the earlier reviews flagged poor methods and the risk that flaws could distort the results. The 2016 appendix described confidence in the fluoxetine findings as “very low”, and the 2020 review repeated that warning.
The trial’s researchers did not respond to repeated enquiries from the reanalysis team or to Nature’s request for comment. MedKnow, the trial journal’s publisher, said it had asked the journal’s editors to investigate and expected delays because of conditions in Iran.
Richard Lyus, a Cambridge child and adolescent psychiatrist and co-author of the reanalysis, said guideline committees should review the clarified evidence and weigh whether fluoxetine’s use can still be justified. That proposed reassessment—not a call to stop prescribing—is the clinical next step the researchers seek.

The federal order prioritizes measurable pollution reductions but leaves coordination questions unanswered. Separately, Virginia tribes have withdrawn from formal participation after being denied full partner status.
3 min

Scientific director Steinn Sigurðsson says volunteer moderators check about 1,000 manuscripts a day, with legitimate papers delayed before posting.

The agreement covers some return-to-flight costs. NASA is also starting the process to restore two flights and plans certification support for Vulcan, with only six Atlas V rockets remaining.
3 min