Opparounds

The Efficacy and Safety of Cannabinoids for the Treatment of Mental Disorders and Substance Use Disorders: A Systematic Review and Meta-Analysis

Jack Wilson, Olivia Dobson, Andrew Langcake, Janni Leung, Maree Teesson, Tom P. Freeman, Wayne Hall, Gary C. K. Chan, Emily Stockings · The Lancet Psychiatry · April 2026

This review screened 5,774 records and pooled 54 randomized controlled trials covering 2,477 participants in which a cannabinoid was the primary treatment for a mental disorder or a substance use disorder. It is the most complete synthesis of this literature to date.

For the indications patients most often ask about, the answer was no benefit: anxiety, PTSD, psychotic disorders, OCD, anorexia nervosa, and opioid use disorder all showed no significant effect. Cannabinoids increased cocaine use in the relevant trials. There was some signal for cannabis use disorder, tic disorders and Tourette syndrome, autism spectrum disorder, and insomnia — but the authors rate the certainty of that evidence as generally low.

Read the paper →  doi:10.1016/S2215-0366(26)00015-5

Multidomain Lifestyle Intervention for the Prevention of Cognitive Decline in At-Risk Older Adults in Latin America (LatAm-FINGERS): A Single-Blind, Multicentre, Randomised Controlled Trial

Lucía Crivelli, Claudia K. Suemoto, Ana Luisa Sosa, Francisco Lopera, Ricardo F. Allegri, et al., for the LatAm-FINGERS Study Group · The Lancet · July 2026

The FINGER model — combining exercise, diet, cognitive training, social engagement, and vascular risk management — was culturally adapted and tested across 11 Latin American countries. A total of 1,065 adults aged 60 to 77 at elevated dementia risk were randomized to a structured intervention or flexible health advice for 24 months; 82.3 percent completed.

Both arms improved on the global cognitive composite, as expected with repeated testing, but the structured arm improved more: 0.31 SD per year versus 0.20 SD per year, a between-group difference of 0.11 SD per year (95% CI 0.06 to 0.15, p<0.0001). Dropout was lower in the structured arm (15.2 versus 20.2 percent). Serious adverse events occurred in 9 percent of structured and 5 percent of flexible participants, and the eight deaths were unrelated to the intervention.

Read the paper →  doi:10.1016/S0140-6736(26)01278-X

Why the last 5 mg is the hard part: antidepressant tapering is hyperbolic, not linear

Serotonin transporter occupancy does not track dose linearly — it follows a hyperbolic curve that saturates early. PET data show that 20 mg of citalopram already occupies roughly 80 percent of SERT, and going from 20 to 40 mg buys only a few additional percentage points. Read the curve backward and the clinical consequence appears: equal reductions in milligrams produce steeply *unequal* reductions in occupancy. Dropping citalopram from 40 to 30 mg changes occupancy by a couple of percent; dropping from 10 mg to zero changes it by well over 50 percent. This is why patients sail through the top half of a taper and then hit a wall — and why they are so often told, incorrectly, that a reaction at 5 mg cannot be pharmacological.

The practical rule is to taper by a constant fraction of the current dose rather than a constant number of milligrams — roughly 10 percent of the *current* dose per step, with the interval set by the patient's response rather than the calendar — which produces an exponentially decreasing dose schedule and an approximately linear decline in receptor occupancy. That arithmetic requires final doses below the smallest tablet, which is why liquid formulations, compounded doses, and tapering strips exist; splitting tablets stops working exactly where the difficulty begins. Two drugs deserve extra caution for pharmacokinetic reasons: paroxetine, with a short half-life and no active metabolite plus its own CYP2D6 autoinhibition, and venlafaxine, whose 5-hour half-life makes even a missed dose symptomatic. Fluoxetine's long half-life makes it the gentlest to stop and the reason it is sometimes used as a bridge. Finally, distinguish discontinuation from relapse by timing and content: withdrawal begins within days and features dizziness, electric-shock sensations, and nausea, and it resolves within hours of reinstating the drug, whereas a relapse takes weeks to build and looks like the original episode.

Neutrality is not a flat affect — and the real relationship is not a lapse in technique

Neutrality is the most misunderstood word a psychotherapy supervisor uses. Anna Freud's formulation is precise: the analyst stands equidistant from id, ego, and superego — not distant from the patient. It is a stance toward the patient's internal conflict, a refusal to take sides between the wish and the prohibition, so that both can be examined rather than one being reinforced by the therapist's approval. It has nothing to do with warmth, and a therapist who understands neutrality as blankness has substituted a caricature for a technique.

Running alongside this, Greenson's distinction between the transference and the real relationship — the genuine, non-transferential perception of and relatedness to each other as actual people — is now supported by outcome data. Stronger real relationship predicts better outcome across modalities, and it is measurably distinct from the working alliance. So the practical question is not whether to be real, but which register a given moment calls for.

The useful heuristic: when the patient's material is conflictual, stay neutral, because taking a side forecloses the conflict you are trying to open. When the event is real — a death, a diagnosis, an actual injustice, or something you actually did — a neutral response is not technique, it is a failure of contact, and patients experience it correctly as such. The error trainees make most often is reaching for neutrality at the moment of real loss, because it feels professional. The error more experienced clinicians make is drifting into friendliness during conflictual material, because it feels kind. Both are the same mistake: reading the wrong register.