Friday, October 2, 2026

The AI Psych Brief

🔬 Section 1

Psychiatric Research Article

Comparative efficacy of antipsychotics for reducing aggression in schizophrenia: systematic review and network meta-analysis

Lamsma J, Ostinelli EG, Cipriani A, Fazel S · British Journal of Psychiatry · September 2026

The Oxford group pooled 41 randomised trials (21,025 adults, 18 antipsychotics) that measured hostility or aggressive behaviour in schizophrenia and related psychoses, and compared them in a single network meta-analysis. Most trials were double-blind, with a median follow-up of about 8 weeks, and most measured aggression with hostility items or subscales from standard symptom rating scales rather than as a primary outcome.

Against placebo, clozapine had the largest effect (SMD -0.75), followed by paliperidone (-0.62) and olanzapine (-0.60). Asenapine, risperidone, aripiprazole and haloperidol also reduced hostility, with smaller effects (SMD -0.52 to -0.30). The authors rated confidence in the evidence as low or very low, mainly because of bias within studies and incoherence across the network.

💡 Why it matters

When persistent hostility is driving admissions, restraint or forensic risk, this is the best comparative evidence yet that clozapine should come early, not last. But the head-to-head differences are low-certainty: treat the ranking as a reason to consider clozapine, not as proof that one D2 antagonist beats another.

Read the paper → doi:10.1192/bjp.2026.10800

📖 Section 2

Psychotherapy Research Article

The efficacy of modified psychodynamic psychotherapy for patients with schizophrenia-spectrum disorders in Germany: a prospective, single-centre, assessor-blinded, parallel-group, randomised controlled trial

Montag C, Bröcker AL, Bayer S, et al.; von Haebler D · The Lancet Psychiatry · October 2026

At the Charité in Berlin, 130 outpatients with schizophrenia or schizoaffective disorder were randomised to treatment as usual alone or to treatment as usual plus Modified Psychodynamic Psychotherapy for Schizophrenia (MPP-S), a manualised treatment of at least 30 sessions. MPP-S is built around the interpersonal vulnerability typical of psychosis. The primary outcome was psychosocial functioning (Mini-ICF-APP), rated by blinded assessors at 24 and 36 months.

Functioning improved more in the MPP-S group, by an adjusted 3.45 points at 24 months and 4.07 points at 36 months. The gap widened after treatment ended rather than fading. Adverse events were similar between groups. There were two suicides, one in each arm, and the one in the MPP-S arm was a patient who never started therapy.

💡 Why it matters

This is one of the few randomised trials of psychodynamic therapy in psychosis with blinded, long-term outcomes, and its benefit showed up in functioning, the domain antipsychotics treat least well. It is a single-centre trial with unblinded therapists, but it supports offering a relationally focused, adapted dynamic therapy to stabilised patients, not treating psychosis as off-limits for dynamic work.

Read the paper → doi:10.1016/S2215-0366(26)00240-3

💊 Section 3

Psychiatric Fact

A therapeutic total valproate level can hide a toxic free level

Valproate is about 90% protein-bound at usual concentrations, but that binding is saturable. As the total level climbs past roughly 75 to 100 mcg/mL, the free fraction rises out of proportion, so small dose increases near the top of the range deliver more active drug than the total level suggests. The same thing happens at ordinary levels whenever albumin falls or binding sites are occupied. Hypoalbuminemia (older, malnourished or medically ill patients), uremia, late pregnancy, and displacement by aspirin or other highly bound drugs can each double the free fraction. The lab reports a reassuring total of 70 while the patient is sedated, tremulous or hyperammonemic.

The practical move: when a patient on valproate has albumin below about 3.5 g/dL, renal failure, or toxicity that doesn't match the level, order a free valproate level (reference roughly 5 to 15 mcg/mL) and dose to that. Don't raise the dose because the total looks low; in these patients a "subtherapeutic" total may already be a therapeutic free concentration.

🛋️ Section 4

Psychotherapy Teaching Pearl

Pretend mode sounds like insight until you notice nothing moves

In MBT, pretend mode is the opposite failure to psychic equivalence. Mental states are cut off from reality instead of being fused with it. The patient talks fluently about feelings, attachment and their own "patterns", often in borrowed therapy language, but the talk has no affect and never touches anything that changes. It is easy to mistake for good work. The therapist feels interested, even impressed, and the session fills with elaborated theory, but the countertransference is usually boredom or a vague emptiness. Pretend mode is a hypermentalizing state, and joining it makes it worse. Offering a deeper interpretation feeds it more material to intellectualize. The MBT stance is to gently interrupt the flow, name the disconnect without accusing the patient of anything, and bring the conversation back to a specific, affect-laden moment, ideally one between patient and therapist in the room right now.

🗒️ Vignette

A 34-year-old with BPD spends twenty minutes explaining that her rage at her sister is "really displaced abandonment from my father, classic splitting". It is articulate and entirely flat. The resident notices that she is drifting and composing her own interpretation to add. Instead she says: "Can I stop us for a second? We've built a very neat theory, and I realise I've lost track of what you actually felt when your sister hung up on you." The patient laughs and restates the theory. The resident stays put: "Right now, telling me this, what's happening in you?" There is a pause, then: "Honestly, I feel stupid that I still care what she thinks." The resident doesn't interpret the shame. She stays curious about that one moment, because that is the first thing in the session that was real.