Early Intervention in Ultrahigh Risk for Psychosis Ineffective

Few transition to psychosis anyway, relapse rates were high after treatment, maintenance therapy was ineffective, and no treatment was more effective than any other.


In a new study in JAMA Psychiatry, researchers tested an increasingly intensive, multi-stage approach to early intervention for people at “ultrahigh risk” of psychosis. Their findings? The intervention failed in every way.

“In this sequential multiple assignment randomized trial including 342 individuals, a specialized psychological intervention (cognitive-behavioral case management [CBCM]) and a psychopharmacological intervention (CBCM and antidepressant medication) were not more efficacious than control conditions in improving remission and functional recovery. Relapse rates among individuals who remitted were high.”

Despite being classed as “ultrahigh risk,” few people transition to psychosis anyway. Moreover, the intervention didn’t lead to more remission, relapse rates were high for those who did remit after treatment, maintenance therapy after remission was no more effective than simple monitoring (watchful waiting), and no treatment (for instance, CBCM or drugs) was more effective than any other.

The study was led by Patrick McGorry and other researchers at Australian early intervention nonprofit Orygen, affiliated with The University of Melbourne. Also, it included researchers from Columbia University, UC Davis, and UCSF. Participants were 342 people between 12 and 25 who sought treatment for being at “ultrahigh risk” (UHR) of psychosis.

UHR is a theoretical construct that first appeared in the 1990s. It’s based on the idea that people with schizophrenia or other psychotic diagnoses could be identified early—before they meet the criteria for those diagnoses—based on traits such as subthreshold psychosis (e.g., “unusual thoughts,” “disorganized speech,” “perceptual abnormalities”).

The current study was called Staged Treatment in Early Psychosis (STEP) and had three stages, beginning with six weeks of support and problem-solving (SPS). If that didn’t result in “remission” of UHR symptoms (symptomatic and functional improvement), patients underwent 20 weeks of either continued SPS or CBCM. If that didn’t result in remission either, patients received 26 weeks of CBCM with fluoxetine (Prozac) or CBCM with placebo. If patients did remit, they received either maintenance SPS or simple monitoring.

The theory is that if treatment can cause “remission” of these undiagnosable, subthreshold symptoms, it will prevent people from eventually transitioning to full psychosis. Generally, though, the high-risk model does an extremely poor job of predicting who will eventually graduate to full psychosis. Previous research has also found that interventions for those considered UHR, such as giving them antipsychotic drugs, can lead to worse outcomes.

The researchers had six hypotheses, all predictions about how effective they thought the early intervention treatment would be. But their hypotheses were all proven wrong:

  1. Hypothesis 1 was that stage 1 (SPS) would result in a 50% remission rate. Instead, it resulted in a much smaller remission rate of 8.5%.
  2. Hypothesis 2a was that in stage 2, those who received the more intensive intervention, CBSM, would do significantly better than those who continued SPS. Instead, there was no difference between the groups.
  3. Hypothesis 2b was that in the CBCM group, they could predict outcomes based on specific factors (“cognitive biases or vulnerabilities”). Instead, they found no differences in these factors.
  4. Hypothesis 2c was that those specific factors would improve more for those who did better in CBCM. Instead, they found no differences in these factors.
  5. Hypothesis 3 was that in stage 3, Prozac added to CBCM would be better than CBCM plus placebo. Instead, there was no benefit for Prozac over placebo.
  6. Hypothesis 4 was that in those who remitted, maintenance treatment with SPS would be better than simple monitoring to prevent relapse. Although there was a difference in the early weeks after treatment, there was no difference between groups at the follow-up point of one year.

Ultimately, after one year, 27.2% were said to have “remitted” at some point, although most of them then “relapsed.” Consistent with previous research showing that the UHR model is a poor predictor, only 13.5% actually transitioned to clinical psychosis.

The researchers conclude:

“The findings suggest that enhancing the intensity of treatment with psychological interventions (CBCM) or antidepressant medication in real-world youth mental health services does not produce benefit over continuing simpler care for a longer period.”

One of the major limitations of the study also indicates how much the patients disliked the interventions: By the end of the third stage, 248 out of 342 (72.5%) had dropped out of the study.

The researchers explicitly compare this to the STAR*D antidepressant trial’s enormously high drop-out rate, which led to the researchers in STAR*D fraudulently pumping up their numbers, violating their pre-specified protocol in numerous ways, including considering 606 study drop-outs—and 570 who weren’t depressed enough to be in the study to begin with or never got a baseline score—to be treatment successes.



McGorry, P. D., Mei, C., Amminger, P., Yuen, H. P., Kerr, M., Spark, J., . . . & Nelson, B. (2023). A sequential adaptive intervention strategy targeting remission and functional recovery in young people at ultrahigh risk of psychosis: The Staged Treatment in Early Psychosis (STEP) sequential multiple assignment randomized trial. JAMA Psychiatry. 2023;80(9):875-885. doi:10.1001/jamapsychiatry.2023.1947 (Link)


  1. So basically:

    1.- People with, among others, “unusual thoughts,” “disorganized speech,” “perceptual abnormalities” for the most part are going to be who they are. Not surprising.

    2.- Out of six, 6, hypothesis none was corroborated. Not surprising given mental illeness has no theorical basis. Arguments to formulate hypothesis are not based in facts beyond doubt, per definition of sicentific theory

    3.- An over 70% droput rate might speak of hideousnes of intervention (fluoxetine was used in minors! yikes! and uuuughs!), but given past experience in clinical research, aprioristically, there were most likely lots of harms. Harm is a significant explanation for dropouts, not just unpleasantness or ineffectiveness. Even fatalities concealed from the “clinical monitor”. Or by it.

    4.- Cognitive-behavioural approaches lead, in part, to dropouts that could be explained by harm, and really do not seem to change a whatever, historically associated with “altered” cognition: in this case a most likely hypothezised structural bias towards “psychosis”. Not explictly stated but easily inferred by the hypothesis: otherwise what’s the point of testing the intervention?. Like if psychosis was random?.

    5.- Although a failure of early intervention, I predict most pros are going to premedicate with neuroleptics suspects of “graduating” to “full” psychosis. Reminding myself psychosis only has 2 parts: delusions and hallucinations. Neither of which has mild forms distinguishable from normal behaviour.

    —-Explainer: perceptual distorsions in mental uncare are called among other things illusions, equivocations, misremembering, confabulations or poor testing, like in Bender-Gestalt*.

    —-Delusions require proof that the belief is false, that also has to be shown to the patient, big IF. It is a normal condition for many folks. It’s called among other things stubborness. And as such is part of who many of us are. And in clinical situations, the most common is celotype**, proof is never shown to the patient. A fact admitted in several psychiatrical [sic] articles. So even in clinical practice, unless frankly impossible belief, proof is never, ever, shown, so delusions in practice for the most part can’t be used, according to it’s own definition.

    6.- Simpler care for longer period probably violates human rights or other pieces of the law, and perhaps family “dynamics”. Given these are minors parents will be deputised to surveill, annoy and otherwise participate in such immoral and possibly illegal interventions that, as per reviewed, and aprioristically, will not provide more than anecdotal benefit. Reducing droputs with a less annoying intervention comes to my mind as motive, with the unpleasantness of prolonging surveillance.

    Cool review by PS. Just the STAR*D didn’t have a more than 72% dropout rate, AFIR, it was way less. What does this droput rate tell about the STAR*D oneS might to me go around the lines of: these kids are way more savy, more normal, than the adults enrolled in the STAR*D. They possibly, see above, ran faster, further and farther than adults!.

    *This test was developed by the infamous Lauretta Bender who gave LSD to kids she acussed of being psychotic. Apparently to treat them, with a drug notorious for causing in around 5% permanent psychosis. And she “dropped” out children who got improvement. Some even apparently enjoyed the LSD trip and got kickd out of the “study” because of that, reviewed in a book somewhere else. Provenance from the test and the practice where it originates, the context, gives me the impression of relevance, not character attack, ad hominem.

    **If I am not having cognitive distorsions like confabulating, around half of all delusions are of infidelity in, or by the partner. Which, obviously a pro can’t show in reality is false. So celotype is a quintessential example that delusions are fabricated signs or symptoms since it violates one of the requirements for the concept of delusion. It is not bad practice, it’s a fabrication.

    —-Odd spurious coincidence: CONSTRUCT linguistically is similar, analogous or identical to FABRICATION.

    —-Apt not casual but causal coincidence: Research that violates the definition of scientific theory violates the definition of delusions. Not explictly in this study but one of the metrics was to diagnose less kids with delusions. So both violations seem to me be part of the study.

    —-Irony: to diagnose less delusions in anyone, just stick to the definition of delusion…

    Final thought to suggest the value of aprioristics: If infidelity as understood in celotype is so common, say it was around 50%, ball park guess, and a person came to the office with what passes as celotype, then this individual has 50% chance of being correct in that belief, or 50% chance of being, not deluded, but wrong.

    Beliefs are not subject to argumentation, reasons and explanations are, but beliefs are not. Another violation of the use of language regarding logic, proof, evidence, belief and reason. That’s why they link the “odds” in a mix, to obfuscate the easily demostrable fact that delusions are fabrication used in very bad and pernicious ways to humans, who are just who they are.

    Report comment

  2. Very interesting given that McGorry has made his name out of inducting people into the psychiatric system as early as possible, and has been handsomely rewarded and recognised for doing so. How did he let these results slip through? Surely he could have used some STAR*D – style tactics to come to a more favourable conclusion?

    From Wikipedia: In January 2010 McGorry was named Australian of the Year for his services to youth mental health. In June 2010 he was appointed an Officer of the Order of Australia. In 2013 Professor McGorry was honoured with the National Alliance on Mental Illness Scientific Research Award, the first time the award has been bestowed upon a researcher outside of the United States. In 2015, McGorry was awarded the Lieber Prize for Outstanding Achievement in Schizophrenia Research, given by the Brain & Behavior Research Foundation. In 2018 he was awarded the Lifetime Achievement Award by the Schizophrenia International Research Society. He was the first psychiatrist elected as a Fellow of the Australian Academy of Science.

    Report comment