The Hidden Psychosis Risk in Adult Stimulant Prescriptions—Even When Legitimately Prescribed": The Hidden Psychosis Risk in Adult Stimulant Prescriptions—Even When Legitimately Prescribed": Sixty-one people out of 16,125. That is how many adults in Iceland were hospitalized for a first-ever episode of psychosis or mania within a year of starting an ADHD drug, according to a 2025 study in BMJ Mental Health.
It works out to an absolute risk of 0.38 percent. The number sounds small, and it is—but it is roughly eight times the 0.048 percent rate at which Icelanders aged 18 to 67 land in the hospital with a first psychotic or manic episode for any reason.
Before we make that eight-fold figure do any work, look at what has happened to the population it describes.
In 2010, about 1.5 percent of Icelandic women and 1.6 percent of men aged 18 to 44 were prescribed ADHD medication. By 2023, those figures were 11 percent of women and 9.4 percent of men—roughly a sixfold rise in thirteen years.
Adult prescribing in that age band reportedly hit 10.2 percent. The best estimates put true adult ADHD prevalence at 2.5 to 5.2 percent.
So Iceland is now medicating two to four times as many young adults as could plausibly have the disorder. Whatever else that number means, it means the people receiving stimulants today are not the people the safety literature was built on.
The people receiving stimulants today are not the people the safety literature was built on—it is overwhelmingly pediatric, spent freely on adults.
That literature is overwhelmingly pediatric. The efficacy evidence, as the Icelandic authors note dryly, "is not as strongly supported for adults as for children and adolescents." Reassurance assembled from studies of medicated eight-year-olds is being spent freely on thirty-year-olds who arrived via a booming, loosely gated diagnostic pipeline.
That is the setup for the alarm. Now the honest part.
An eight-fold relative risk is the kind of number that produces headlines. The absolute translation is a number-needed-to-harm of 302: you would have to start roughly 302 adults on ADHD medication to see one additional hospitalization for psychosis or mania in the following year that you would not otherwise have seen.
The study also estimated a proportional attributable risk of 87.5 percent—meaning that in eight to nine of those ten cases, the admission likely would not have happened without the drug. That is a real signal. It is not a common event.
And it does not replicate cleanly. A Swedish population-based cohort published in Translational Psychiatry took the population you would expect to be most vulnerable—3,770 adults with a prior history of psychosis—and looked at what happened when they started a stimulant or atomoxetine.
Using a within-individual design, comparing each person to themselves in the six months before and after starting, the rate ratio for psychotic events was 0.95, with a confidence interval of 0.84 to 1.08. No increased risk. The finding reportedly held across sexes, age groups, and specific drug classes.
That is the most powerful design available for this question, because it cancels out the stable differences between people that plague ordinary comparisons. And it landed on nothing.
So we have an eight-fold signal in one register and a flat line in another. The tension is not noise; it is the whole story.
The likely reconciler is confounding by severity—and nobody has resolved it. Adults spiraling toward a first psychotic break often look, in their early presentation, like people who need help concentrating and organizing their lives. Some get diagnosed with ADHD and prescribed a stimulant weeks before the break that was coming anyway.
The pharmacovigilance study of adolescents and young adults (aged 13–25), drawing on the WHO's VigiBase, states the caveat plainly: it "could not exclude the possibility of confounding by disease severity." When the sickest patients are the ones most likely to be both medicated and hospitalized, the drug takes the blame for the disease.
But confounding cuts the other way in the Icelandic data too, which is why the represcription finding matters most.
Of the 61 who were hospitalized, 69 percent—42 people—were put back on their ADHD medication within a year of discharge. Of those 42, one in four, eleven people, were readmitted.
That is the least ambiguous, most actionable number in the entire literature. Whatever the truth about first exposure, restarting a stimulant in someone who has just had a psychotic or manic episode on one is playing a game with a one-in-four loss rate.
Now widen the lens, because the class of drug matters more than most coverage admits.
A JAMA Psychiatry meta-analysis of 16 studies and 391,043 participants found a "heterogeneous but nonnegligible occurrence of psychosis and BD following stimulants"—and, crucially, higher with amphetamines than with methylphenidate. The VigiBase study reportedly put a number on it: amphetamine use carried a reporting odds ratio of 1.61 versus methylphenidate for psychotic symptoms, rising to 1.94 when restricted to ADHD.
This fits the pharmacology. Both drugs block dopamine reuptake, but amphetamine also forces dopamine release—up to four times as much. Atomoxetine, in that same analysis, showed no association.
And the drug can be protective in the right population. A Finnish instrumental-variable study of a national birth cohort found no overall link between sustained methylphenidate treatment and later nonaffective psychosis—and among children diagnosed before age 13, a signal that methylphenidate reduced the risk.
Read the whole body of evidence together and the crude claim collapses. Not "stimulants cause psychosis," but: amphetamines carry more risk than methylphenidate; adults and adolescents differ from children; and the represcription window after an episode is where preventable harm actually concentrates.
That is not a verdict. It is a decision structure. Which means it can be managed—by the exact clinical care that a sixfold prescribing surge tends to erode.
So here are the questions a patient or family should be able to answer before starting, and should ask out loud if the prescriber doesn't raise them.
First: which drug class, and why this one? Given that amphetamines release far more dopamine and show the higher psychosis signal across studies, methylphenidate is the more conservative default—especially for anyone with a personal or family history of psychosis, mania, or bipolar disorder. If an amphetamine is chosen anyway, the reason should be explicit.
Second: how solid is the diagnosis? In a system medicating two to four times the plausible prevalence, the base rate of misdiagnosis is high, and early psychosis can masquerade as inattention. The value of a careful workup is not bureaucratic; it is the difference between treating ADHD and lighting a fuse.
Third: what do dose and monitoring look like in the first year? The risk window in the Icelandic data was the first twelve months. That is precisely the period that warrants scheduled follow-up rather than a prescription refilled by portal—starting low, and watching for the early signs of mania or psychosis, not just for whether focus improved.
Fourth, and most important: what is the plan if an episode occurs? This is where the one-in-four readmission rate should reshape behavior. The Icelandic authors are blunt that if a drug is represcribed after such an event, "regular follow-up must be provided for at least 1 year." The honest default after a stimulant-associated psychotic or manic episode is not automatic represcription—it is a genuine reassessment of whether that drug, or that class, belongs in the picture at all.
None of this argues against treating adult ADHD. It argues against treating it casually.
The reassuring consensus wasn't wrong; it was reportedly answering a different question, about different people, prescribed to more carefully. The new signal doesn't overturn it so much as mark its edge—a small, real, drug- and population-specific risk that lives almost entirely inside the decisions a rushed system stops making.
Three hundred and two people to harm one, if the drug is chosen and monitored well. One in four, if you restart it in the wrong person at the wrong moment. The gap between those two numbers is the whole point—and it is made of attention.