This Week in Psychiatry — Sep 28, 2026

This AudioScholar briefing is generated by artificial intelligence for healthcare professionals and trainees. It is not medical advice.

Welcome to This Week in Psychiatry. This week we're covering 10 notable papers spanning prescribing safety and comparative efficacy, familial and developmental risk across psychosis, substance use and depression, and a cluster of biological work on antibodies, metabolites and brain circuits. Let's dive in.

We start with two papers that speak directly to how we choose a drug. In the BMJ, Ostinelli and colleagues report an individual participant data network meta-regression asking what antidepressants actually do to the QT interval in the first eight weeks of treatment for major depression [1]. They pooled individual level data from thirty-five randomised trials, close to nine thousand participants, covering placebo and ten antidepressants, and then added an aggregate analysis of one hundred and thirty-nine trials with more than fifty thousand participants to look at early cardiovascular events. Compared with placebo, escitalopram was associated with a QTc increase of roughly nine milliseconds and amitriptyline roughly five milliseconds, and when modifiable and non-modifiable risk factors such as age, sex, body mass index, potassium and renal function were taken into account, those two agents again sat at the prolonging end while venlafaxine and vortioxetine sat at the lower end. The variability between individuals was substantial, particularly for escitalopram, fluoxetine and mirtazapine, which is arguably the more useful message than any single point estimate. Critically, the aggregate analysis found no evidence that antidepressant use increased early major adverse cardiovascular events or non-suicidal sudden death relative to placebo, with a risk difference indistinguishable from zero. The authors frame this as support for shared decision making, where individual cardiac risk factors inform the choice, rather than as a signal that antidepressants are cardiotoxic in the short term.

Staying with comparative efficacy, the British Journal of Psychiatry publishes a systematic review and network meta-analysis from Lamsma and colleagues on antipsychotics for hostility and aggressive behaviour in schizophrenia [2]. Forty-one trials, just over twenty-one thousand participants and eighteen antipsychotics, with a median follow-up of eight weeks. Several antipsychotics outperformed placebo, with the largest effects for clozapine, paliperidone and olanzapine, and smaller but still significant effects for asenapine, risperidone, aripiprazole and haloperidol. The important caveat is that confidence in the evidence was rated low or very low, driven by within-study bias and incoherence, and that in nearly nine out of ten studies aggression was captured only as a hostility item embedded in a broader symptom scale rather than measured as a primary outcome. So this is consistent with, but does not firmly establish, a specific anti-aggressive advantage for clozapine, and the authors call for trials designed with aggressive behaviour as the primary endpoint and longer follow-up.

The second theme is risk that runs in families and unfolds over development. In Molecular Psychiatry, Khemiri and colleagues report a Swedish national register study of more than four million people born between 1958 and 1999, examining lifetime substance use disorder and suicidal behaviour [3]. Among the roughly two hundred and fifty-eight thousand individuals with a substance use disorder diagnosis, about twenty-three percent had a lifetime suicide attempt or death by suicide, compared with around one and a half percent of those without such a diagnosis. After adjustment for sex, birth year, socioeconomic factors and psychiatric comorbidity, substance use disorder was associated with roughly a six-fold higher risk of any suicidal event, with the highest risk in people carrying both alcohol and drug use diagnoses. The familial co-aggregation is the novel part: risk in relatives rose as genetic similarity rose, with full siblings of affected individuals at roughly two and a half times the risk and half siblings at about one and a half times. This is observational and cannot separate shared genes from shared environment, but it supports the authors' conclusion that a family history of substance use disorder is itself a marker of suicide risk, across every substance class they examined.

Developmental trajectory is also the theme of a large study in the American Journal of Psychiatry from Mollon and colleagues, drawing on the Early Psychosis Investigation in Mexico cohort [4]. Just over a thousand young people with early-onset psychosis, one hundred and forty-six of their siblings, and more than a thousand psychiatric control participants completed cognitive testing. Those with non-affective early-onset psychosis were impaired across every cognitive domain measured, those with affective psychoses were impaired across most domains but less severely, and siblings showed no overall impairment. What makes the paper interesting is the age-by-group pattern: in non-affective psychosis, impairment in general cognitive ability was small in children, medium in adolescents and large in adults; in affective psychosis, children were unimpaired and adults showed large deficits; and among siblings, only the adults showed large deviations, mainly in processing speed. The authors are explicit that these data are cross-sectional, so they cannot confirm that individual people decline over time, and longitudinal follow-up is needed before anyone concludes that targeted early intervention in specific windows would prevent divergence.

Also in Molecular Psychiatry, Hammes and colleagues take the opposite tack and ask what resilience looks like in the brain [5]. In more than eighteen hundred participants from the Marburg-Münster Affective Disorders Cohort Study, they modelled depressive symptom severity from twenty-two risk and protective variables, which together explained about half of the variance in symptoms, and then defined resilience as doing better than that model predicted. At baseline there were no structural brain correlates at all, in either region of interest or whole-brain analyses. At two-year follow-up, greater resilience at baseline predicted lower grey matter volume in the left inferior orbitofrontal gyrus and temporal pole, with no cortical thickness changes. The authors offer competing interpretations, greater neural efficiency or a delayed biological cost, and the effect size is modest, so this is a hypothesis-generating result rather than a biomarker.

The third theme is biology at the bedside, and it begins with the most directly usable paper of the group. In Molecular Psychiatry, Endres and colleagues, writing for the German network for research on autoimmune encephalitis, set out interdisciplinary expert recommendations for evaluating patients whose psychiatric presentation might be driven by autoimmune encephalitis [6]. The proposed approach is deliberately stepwise: clinical assessment, magnetic resonance imaging, electroencephalography, blood testing, cerebrospinal fluid analysis and neuronal or glial antibody testing in both serum and spinal fluid, with fluorodeoxyglucose positron emission tomography, live-cell-based assays and volumetric imaging reserved for genuinely ambiguous cases. The authors are careful to say they are not defining a new disease entity or new criteria, and they devote real space to the limitations of each modality precisely because the field risks both under-recognition and overdiagnosis. This is consensus guidance rather than trial evidence, but it is the sort of document that gives psychiatric services a defensible pathway.

Two biomarker papers sit alongside it. Bi and colleagues, also in Molecular Psychiatry, combine a case-control study of eleven hundred and fifty-nine patients with schizophrenia and fourteen hundred and forty controls with a meta-analysis of fifty-four studies and nearly ten thousand participants, and find blood homocysteine levels consistently elevated in schizophrenia, by roughly a third on the pooled ratio of means, with consistent results across disease stage, sex, region and assay method [7]. The patient group was considerably younger than the control group, and although the analysis adjusted for age and sex, this remains cross-sectional; the authors themselves call the findings preliminary and note that prospective longitudinal work is needed before homocysteine means anything clinically. Meanwhile in Biological Psychiatry, Gaudette and colleagues used carbon-11 CURB positron emission tomography to measure fatty acid amide hydrolase, the enzyme that degrades anandamide, in thirty-five people with post-traumatic stress disorder and fifty-one controls [8]. Preclinical work predicted higher enzyme levels with stress; they found the opposite, with amygdala levels about ten percent lower in post-traumatic stress disorder and no differences elsewhere in corticolimbic regions. An exploratory analysis found higher corticolimbic enzyme levels in people with childhood trauma regardless of diagnosis, and there was no association with symptom severity. The authors suggest this may help explain why trials of fatty acid amide hydrolase inhibitors have disappointed.

Two conceptual pieces round out the week. A review in Molecular Psychiatry from Bambico and colleagues surveys the mechanisms of transcranial magnetic stimulation, transcranial direct current stimulation, deep brain stimulation and transcranial focused ultrasound, arguing that neurostimulation represents a conceptual shift toward circuit-based models of mental illness while acknowledging that the mechanisms beyond immediate changes in regional excitability remain incompletely understood [9]. And in the British Journal of Psychiatry, Bell and colleagues take on irritability, a symptom that is ubiquitous across diagnoses yet still lacks an agreed definition or a settled place in nosology, and ask what it is, in whom it should be studied, and how it should be modelled [10].

If you only have time for one paper this week, make it the BMJ individual participant data analysis of antidepressants, QTc and early cardiovascular events [1]. It is the largest attempt yet to separate a measurable electrophysiological effect from actual cardiac harm, and it reframes the question from which antidepressant is dangerous to which patient carries the risk factors.

Here is what this week's evidence adds up to in psychiatry. First, antidepressants differ measurably in their effect on the QT interval, with escitalopram and amitriptyline at the prolonging end, but across more than fifty thousand randomised participants there was no signal of early major cardiovascular events or non-suicidal sudden death, which the authors read as an argument for individualised risk assessment rather than blanket avoidance. Second, antipsychotics do reduce hostility in schizophrenia over the short to medium term, with clozapine, paliperidone and olanzapine showing the largest effects, but the certainty of that evidence is low and aggression was rarely the primary outcome, so ranking between drugs remains unsettled. Third, register data at national scale show that substance use disorder of any type carries a markedly elevated risk of suicidal behaviour and that this risk co-aggregates in families in proportion to genetic relatedness, though observational designs cannot disentangle genes from shared environment. Fourth, cognitive deviation in early-onset psychosis appears to widen with age and is detectable even in adult siblings, but the cross-sectional design means within-person decline is not yet demonstrated. And finally, the biological findings, from lower amygdala fatty acid amide hydrolase in post-traumatic stress disorder to elevated homocysteine in schizophrenia to structural correlates of resilience emerging only at two years, are all interesting and none are ready to influence individual patient care; the autoimmune encephalitis recommendations are the one document here written for immediate clinical use, and they are expert consensus rather than trial evidence.

That's your roundup for This Week in Psychiatry. The full transcript and references are available on the episode page in your AudioScholar library. This is an AI-curated summary — for clinical decisions, always consult primary sources and current guidelines. See you next week.

If this weekly briefing is useful, follow the show in your podcast app so new episodes arrive automatically. And if it saves you time, a quick rating in your podcast app helps other physicians find the show.

This is an automated summary generated by artificial intelligence, which can make mistakes. Always review the original source materials.