The AI Psych Brief

A Digital Therapeutic Intervention for Negative Symptoms of Schizophrenia: A Randomized Clinical Trial

Abhishek Pratap, Shaheen E. Lakhan, Cassandra Snipes, et al. · JAMA Network Open · September 2026

CONVOKE randomized 457 adults with schizophrenia at 66 US sites to 16 weeks of CT-155 — a smartphone app delivering structured psychosocial strategies for motivation and pleasure — or a sham 'educational' app, both added to a stable antipsychotic regimen. Both arms were presented to patients as 'digital therapeutic A' and 'B', and raters were blinded. Enrollment required at least moderate negative symptoms on the CAINS motivation-and-pleasure (MAP) scale and excluded people already in psychotherapy or with prominent positive symptoms.

CAINS-MAP fell 6.8 points with CT-155 versus 4.2 with the control, a between-group difference of 2.6 points (Cohen d = −0.36, P < .001). Expressive negative symptoms and positive symptoms did not separate, a patient-rated global-improvement benefit at week 8 had faded by week 16, and serious adverse events were rare and none treatment-related; 70% of participants were still engaged at week 15.

Read the paper →  doi:10.1001/jamanetworkopen.2026.35753

Prenatal Acetaminophen (Paracetamol) Use and the Risk of Autism and/or Attention-Deficit/Hyperactivity Disorder Among Sibling-Matched Cohorts

Shan Luo, Qiaowa Gong, Yujie Ai, et al. · JAMA Internal Medicine · September 2026 (online June 29, 2026)

Investigators used Hong Kong's territory-wide electronic health records (2001–2023) covering about 708,000 mother-child pairs, 43% of whom had dispensed prenatal paracetamol. Rather than compare exposed with unexposed children from different families, they built sibling-matched cohorts — 124,333 children for autism and 97,285 for ADHD — so that each exposed child was compared with a brother or sister sharing genes and household.

Within sibling pairs there was no association: adjusted hazard ratio 1.00 (95% CI 0.91–1.11) for autism and 1.01 (0.93–1.08) for ADHD, null across trimester, cumulative dose, and pattern of use. Conventional whole-cohort analyses did show positive associations, and so did a negative-control exposure (paracetamol taken before pregnancy) — the signature of familial confounding rather than a drug effect.

Read the paper →  doi:10.1001/jamainternmed.2026.2215

Paroxetine inhibits its own metabolism, so the dose-level curve bends upward

Paroxetine is cleared mainly by CYP2D6 and is also a potent, mechanism-based (irreversible) inhibitor of that same enzyme. As the dose climbs, it progressively disables its own main clearance route, so steady-state levels rise more than proportionally — the label describes nonlinear kinetics, so each dose increase buys a disproportionately larger rise in level. Two practical consequences follow. First, on a therapeutic dose a normal metabolizer is phenoconverted to a CYP2D6 poor metabolizer: genotype reports showing 'normal metabolizer' no longer predict how this patient will handle metoprolol, atomoxetine, aripiprazole, or a TCA. Second, the curve runs in reverse on taper. Because inactivated enzyme must be resynthesized, clearance recovers as levels fall, so the lower doses drop concentration disproportionately fast — part of why paroxetine has one of the worst discontinuation profiles of the SSRIs. Make the final reductions smallest (liquid formulation helps), and when you start or stop paroxetine, re-check every co-prescribed CYP2D6 substrate.

In TFP, the dyad matters less than the moment it flips

Transference-focused psychotherapy organizes chaotic sessions around the dominant object-relational dyad: a self-representation, an object-representation, and the affect linking them — for example, a helpless victim and a cruel, indifferent caretaker, linked by rage. Naming the dyad is the first step, but the integrative work happens at role reversal. Borderline patients do not stay in one position; within minutes the victim becomes the one who is cold, dismissive, or punishing, and the therapist is now the one made to feel helpless. Because split-off identifications are disowned, the patient experiences only the half they currently occupy. The technical move is to show the patient, in the present tense and with the evidence in the room, that both roles live in them — to interpret the oscillation, not just either role. That is what begins to integrate the split, and it is why TFP waits for the flip rather than interpreting the first role as soon as it appears.