This Week in Psychiatry — Aug 31, 2026

Welcome to This Week in Psychiatry. This week we're covering 10 notable papers spanning service models and treatment-resistant illness, the physical health and monitoring burden carried by people with severe mental illness, and trauma, measurement and prediction in everyday practice. Let's dive in.

We start with two studies that test whether ambitious treatment models deliver. In The Lancet Psychiatry, Pilling and colleagues report the ODDESSI trial, the first large randomised evaluation of Open Dialogue — the person-centred, network-based, transdiagnostic model that has generated enormous international enthusiasm. This was a cluster-randomised superiority trial across five National Health Service trusts in London and the South of England, with 30 clusters formed from general practices and 494 adults presenting in crisis to community services. The primary outcome was time to first relapse after initial recovery from the index crisis, followed for two years. There was no difference between Open Dialogue and treatment as usual on that outcome; the hazard estimate sat essentially at one. Recovery rates enabling relapse assessment were broadly similar in both arms, around four in five participants. This is a negative trial on its primary endpoint, and it deserves to be reported as such — a service-wide reorganisation around network meetings did not lengthen time to relapse compared with the standard functional team model. For anyone being asked to fund or implement Open Dialogue at scale, this is the evidence base you now have to work with.

The counterpoint comes from the American Journal of Psychiatry, where Conway and colleagues revisit the RECOVER trial of vagus nerve stimulation in markedly treatment-resistant depression. This was a twelve-month triple-blind randomised comparison of active versus sham stimulation in 493 adults who had failed at least four trials in the current episode — a population averaging more than thirteen lifetime antidepressant treatments and an episode duration of nearly eighteen years. Here too the prespecified primary outcome, percent time in response on the Montgomery-Åsberg scale, did not separate the groups. But secondary measures of symptoms, clinician-rated global improvement, work function and quality of life all favoured active stimulation, with response on the clinician global impression scale in roughly half of the active group versus about two in five with sham. The durability signal is the part clinicians should note: over eighty percent of those who achieved meaningful benefit at twelve months still had it at two years, and about three in ten of those who had not benefited at a year improved during the second year. Two trials, two negative primary outcomes, two very different interpretive conclusions — a useful reminder that in chronic, heterogeneous illness the choice of primary endpoint can determine whether an intervention looks effective or futile.

Our second theme is the physical health of people with serious mental illness, where the numbers this week are large. In JAMA Psychiatry, McIntyre and colleagues report a target trial emulation in the multinational TriNetX electronic health record network, comparing new initiation of a GLP-1 receptor agonist against an SGLT2 inhibitor as an active comparator. After propensity matching, more than 1.5 million adults were included, of whom nearly 200,000 matched pairs had serious mental illness. Over four years, all-cause mortality among patients with serious mental illness was about five percent with GLP-1 agonists versus about six and a half percent with SGLT2 inhibitors — roughly a quarter lower relative risk and an absolute difference of about one and a half percentage points. Among those with type 2 diabetes, semaglutide was also associated with fewer major adverse cardiovascular events, myocardial infarction, stroke and heart failure, and the mortality signal held separately in major depressive disorder, bipolar disorder and schizophrenia. This is observational and residual confounding cannot be excluded — but given that cardiovascular disease drives most of the mortality gap in serious mental illness, this argues for psychiatrists being active participants in metabolic prescribing decisions rather than deferring them.

Monitoring burden is the other side of that coin, and the British Journal of Psychiatry offers a twenty-year population-based study from Hong Kong. Zhou and colleagues followed nearly 4,900 clozapine initiators against more than 38,000 patients starting other second-generation antipsychotics. Clozapine carried a clear excess risk of both minor and serious neutropenia in the first eighteen weeks — roughly a threefold higher rate of serious neutropenia early on — but Bayesian change-point analysis showed that risk falling sharply after about six months and converging with other second-generation antipsychotics by around two years. Adjusted rates of serious neutropenia over the whole follow-up were essentially identical between clozapine and comparators. Notably, no significant early excess was seen in patients under 45. The authors propose weekly monitoring for eighteen weeks and monthly out to two years, with the possibility of personalised schedules by age and sex — an evidence-based argument for relaxing lifelong weekly-to-monthly regimens that deter both prescribers and patients. Alongside these, the BMJ published a broad review by Sawa and colleagues on schizophrenia pathophysiology and management, making the case that progress in drug development is blocked by the biological heterogeneity of a syndrome defined only by clinical criteria, and that objective biomarkers and biological stratification are the necessary next step. That theme carries directly into a Special Communication in JAMA Psychiatry from Musliner and colleagues on genomics and prevention. Their conclusion is appropriately restrained: polygenic scores are not ready for population-wide or universal screening, given modest absolute risk differences, weak individual-level prediction and real potential for psychological harm. The value rises in narrower settings — testing for rare, high-impact variants in patients already entering care where the diagnosis is still unsettled, pharmacogenomics for medication choice and safety, and possibly flagging somatic comorbidity risk to help close that mortality gap. In short, genomics as a targeted clinical tool, not a screening programme.

Our third theme is trauma and measurement. In the British Journal of Psychiatry, Ayers and colleagues report the INTERSECT study, a cross-sectional survey of more than 11,000 women six to twelve weeks postpartum across 31 countries. On average, close to a quarter OF WOMEN described a birth that met DSM-5 criteria for a traumatic stressor, and about seven percent met full criteria for childbirth-related post-traumatic stress disorder — but the country-level range was extraordinary, from single digits to two thirds for traumatic birth and from one percent to over a third for the disorder itself. Many women reported distress and impairment even below diagnostic threshold. The variation points to healthcare and cultural factors rather than biology, and the practical message is routine trauma-informed assessment in perinatal care. Trauma in psychosis is addressed from Psychotherapy and Psychosomatics, where Rickard and colleagues report experience sampling outcomes from the STAR trial, in which trauma-focused cognitive behaviour therapy for psychosis had reduced PTSD symptoms on retrospective clinician measures. Using a smartphone app sampling symptoms multiple times daily, there were no between-group differences in momentary symptoms at nine months, though interaction analyses showed greater decline over time in hyperarousal, negative trauma-related cognitions and avoidance in the therapy arm. The robust benefit seen on retrospective scales was not fully replicated in the moment — a genuinely important methodological caution as our field moves toward ecological momentary assessment.

A companion paper in the same journal from Di Lodovico and colleagues suggests daily self-report can do useful predictive work. In a multicentre naturalistic study of 1,536 outpatients starting agomelatine monotherapy across 388 community centres, daily EuroQol visual analogue ratings of self-rated health over the first fortnight sorted into five trajectory classes. The group showing marked early improvement despite low baseline scores had the highest remission rates at six to eight weeks, while those with severe baseline symptoms and little early movement fared worst. It is naturalistic and single-agent, but it supports something pragmatic: a simple daily rating in the first two weeks may flag who is on track. Finally, from the British Journal of Psychiatry, Khan and colleagues examine the NHS England independent ADHD taskforce report, which documents persistent underdiagnosis, fragmented pathways and long waits, and recommends a needs-led approach spanning health, education, employment and justice — while noting unresolved questions about funding, diagnostic standards and integration between public and private providers.

If you only have time for one paper this week, make it the GLP-1 receptor agonist analysis in JAMA Psychiatry [2]. The mortality gap in serious mental illness is our largest unsolved problem, and this is the strongest signal yet that a drug class already in wide use may narrow it.

Here are the key takeaways from this week in Psychiatry. First, Open Dialogue did not delay relapse compared with usual community care in a properly powered cluster-randomised trial — enthusiasm should now be tempered by evidence. Second, vagus nerve stimulation missed its primary endpoint but showed consistent secondary benefit and unusual durability in profoundly treatment-resistant depression. Third, GLP-1 receptor agonist initiation was associated with lower mortality and fewer cardiovascular events in patients with serious mental illness, which makes metabolic prescribing squarely a psychiatric concern. Fourth, clozapine neutropenia risk is front-loaded in the first eighteen weeks and converges with other antipsychotics by two years, supporting shorter intensive monitoring. And fifth, measurement matters: momentary smartphone assessment did not reproduce retrospective trial findings in trauma-focused therapy for psychosis, while simple daily health ratings in the first two weeks of an antidepressant may predict remission.

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.