The efficacy and safety of transcranial direct current stimulation in patients with ADHD: a systematic review and meta-analysis

ObjectiveThis meta-analysis evaluated the efficacy and safety of transcranial direct current stimulation (tDCS) for treating Attention-Deficit/Hyperactivity Disorder (ADHD).MethodsFollowing PRISMA guidelines, we analyzed 28 randomized controlled trials (RCTs) involving 1,864 participants. Outcomes encompassed core ADHD symptoms, hot and cold executive functions (EFs)—including inhibitory control, working memory, and cognitive flexibility—as well as safety profiles based on adverse events. A multilevel meta-analysis was performed using a random-effects model. Subgroup analyses and meta-regressions were conducted to explore potential moderating factors.ResultsCompared to sham stimulation, tDCS did not significantly improve core ADHD symptoms (standardized mean difference (SMD) = –0.29, 95% CI [–0.59, 0.01], p= 0.05). Similarly, no significant overall effects were observed for cold EFs: inhibitory control (Hedges’ g(g)= –0.11, 95% CI [–0.26, 0.05], p=0.19), working memory (g= 0.13, 95% CI [–0.06, 0.32], p= 0.26), or cognitive flexibility (SMD = –0.42, 95% CI [–1.13, 0.29], p= 0.24). The effect on hot EFs was also non-significant (g = 0.27, 95% CI [–0.14, 0.70], p = 0.19). Exploratory analyses indicated that anode placement at Fp2 was associated with improvement in both inhibitory control (g= –0.52, 95% CI [–0.93, –0.11], p=0.01) and working memory (g = 0.72, 95% CI [0.22, 1.22], p = 0.004), although the overall test for interaction was not significant for inhibitory control (p= 0.19). The most common adverse reactions were mild and transient local skin symptoms, such as itching and redness (RR = 1.42, p=0.04).ConclusiontDCS was well-tolerated but did not demonstrate significant overall efficacy for core ADHD symptoms or executive functions. Anodal stimulation at Fp2 showed potential selective benefits warranting further investigation. tDCS is not currently recommended as a standalone treatment for ADHD. Future research should optimize stimulation protocols and explore combined interventions with behavioral or cognitive therapies.Systematic Review Registrationhttps://www.crd.york.ac.uk/PROSPERO, identifier CRD42024612055.

AI Finds Unreported Side Effects of GLP-1 Drugs in Reddit Posts

A study of more than 400,000 posts in the social media platform Reddit has identified previously unreported side effects from the increasingly popular GLP-1 weight loss and diabetes drugs. Users reported symptoms affecting menstrual cycles and body temperature, which have not yet been described in clinical trials or included in drug labels.  

Published today in Nature Health, the study covers over five years of public online posts from nearly 70,000 Reddit users discussing their personal experience taking the GLP-1 drugs semaglutide and tirzepatide.

“Some of the side effects we found, like nausea, are well known, and that shows that the method is picking up a real signal,” says Sharath Chandra Guntuku, PhD, research associate professor in computer and information science at Penn Engineering and the study’s senior author. “The underreported symptoms are leads that came from patients themselves, unprompted, and clinicians could potentially pay attention to them.”  

Although the study is not representative of the broader population—Reddit users are generally younger, more likely male and based in the U.S.—the symptoms reported collectively match known side effects of semaglutide and tirzepatide. About 44% of users described at least one known side effect, most commonly symptoms of gastrointestinal distress. 

“Clinical trials generally identify the most dangerous side effects of drugs, but they can fail to find what symptoms patients are most concerned about,” says Lyle H. Ungar, PhD, professor in computer and information science at Penn Engineering. “Online patient communities work a lot like a neighborhood grapevine. People who are living with these medications are swapping notes with each other in real time, sharing experiences that rarely make it into a doctor’s office visit or an official report. Even though social media is not necessarily representative, a large collection of posts may reflect additional concerns.” 

The study uncovered a series of side effects that were previously unreported for these drugs. This included discussions of menstrual cycle changes, such as intermenstrual bleeding, heavy bleeding, and irregular cycles. Other users reported chills, hot flashes, fever, and other temperature-related symptoms. In addition, fatigue symptoms ranked as the second most common complaint in these online posts despite rarely being reported in clinical trials. 

“We can’t say that GLP-1s are actually causing these symptoms,” says Neil K. R. Sehgal, doctoral student at the University of Pennsylvania and the study’s lead author. “But nearly 4% of the Reddit users in our sample reported menstrual irregularities, which would be even higher in a female-only sample. We think that’s a signal worth investigating.” 

While efforts to scour the internet for self-reported drug side effects have been ongoing for more than a decade, screening through social media posts at scale remained challenging until the arrival of large language models such as ChatGPT or Gemini. In particular, these tools can prove instrumental in mapping the language users use to describe their symptoms to clinical terminology defined in the Medical Dictionary for Regulatory Activities (MedDRA), used to officially report symptoms in clinical trials. 

“Large language models have made it possible to do this kind of analysis much faster with a level of standardization that could be difficult to achieve before,” says Sehgal. 

The researchers hope these findings will encourage researchers and drug developers to investigate the side effects discussed by users online. In future work, the team plans to expand beyond Reddit and English-language discussions to confirm whether the same symptoms appear across different social media platforms and populations. 

While this approach is not intended as a replacement for clinical trials, screening social media posts for clues on unreported side effects can make a significant difference in terms of speed. This can be especially relevant for drugs like semaglutide and tirzepatide, originally diabetes drugs that quickly became mainstream when the FDA granted them approval as weight loss drugs. 

“Clinical trials are the gold standard, but by design, they are slow,” says Guntuku. “The whole point of this kind of approach is that it can move quickly, and that’s exactly when it’s most valuable.”

The post AI Finds Unreported Side Effects of GLP-1 Drugs in Reddit Posts appeared first on Inside Precision Medicine.

New rules for CDC vaccine panel aim to address lawsuit, empower Kennedy’s allies

After a courtroom defeat, Trump administration health officials have revised the governing documents for a key federal vaccine panel to broaden its membership, increase its focus on potential harms of vaccines, and empower allies of health secretary Robert F. Kennedy Jr.

The new charter for the committee that advises the Centers for Disease Control and Prevention on vaccine use appears aimed at trying to evade the type of legal challenge that has left the currently appointed body in limbo. In addition, the document puts greater emphasis on the role of the Advisory Committee on Immunization Practices in studying injuries possibly linked to vaccination  — though the committee has always paid close attention to any emerging evidence that called into question the safety of individual vaccines.

Read the rest…

Neuroblastoma Tumor Growth in Mice Suppressed by Blocking Enzyme to Inhibit mTOR Signaling

Neuroblastoma is the most common tumor among children under a year of age, and while in its gentlest form neuroblastoma can regress on its own, it can also take an aggressive form, with high-risk neuroblastoma carrying a five-year survival rate of about 40%.

Researchers at The Hebrew University of Jerusalem have now discovered a mechanistic explanation for how neuroblastoma sustains itself and identified a potential approach to severing that mechanism, by inhibiting nitric oxide (NO) production to suppress mTOR signaling. The collective results from work in human neuroblastoma cells and experiments in a mouse xenograft model showed that inhibiting the enzyme neuronal nitric oxide synthase (nNOS) to inhibit NO production suppressed mTOR signaling and slowed tumor growth.

Professor Haitham Amal, PhD, head of The Laboratory of Neuromics, Cell Signaling, and Translational Medicine, is senior and co-corresponding author of the team’s published paper in Brain Medicine, titled “Targeting nNOS suppresses AKT–TSC–mTOR signaling and inhibits neuroblastoma growth.” In their paper the team concluded “Inhibition of nNOS suppresses mTOR signaling, reduces cellular malignancy, and attenuates tumor growth in vivo, identifying the nNOS-mTOR axis as a promising therapeutic target in neuroblastoma.”

Neuroblastoma accounts for roughly 28% of all cancers diagnosed in infants across Europe and the United States. “Neuroblastoma (NB) refers to a spectrum of neuroblastic tumors that originate from the neural crest cells during fetal development,” the authors wrote. “Neuroblastoma is predominantly a pediatric malignancy, with approximately 97% of cases occurring in children.”

NBs can range from spontaneous regression to maturation to an aggressive, deadly metastatic disease. And as the investigators noted, “Despite major advances in multimodal therapy, high-risk neuroblastoma remains associated with poor prognosis, frequent relapse, and therapy resistance, underscoring the need for a better understanding of the signaling pathways that regulate tumor cell survival, differentiation, and metabolic adaptation.”

Nitric oxide (NO) is an essential regulator of carcinogenesis in various tumors, including NB, the authors pointed out. “Nitric oxide (NO) is a ubiquitous free radical signaling molecule produced in multiple organs and tissues), such as those of the central and peripheral nervous systems.” But at elevated concentrations NO becomes reactive, generating nitrogen species that chemically modify proteins through a process called S-nitrosylation. That modification has been implicated in every stage of cancer progression.

The relationship between nitric oxide and tumors is not simple. Very high concentrations can damage DNA and trigger apoptosis. Lower, sustained levels appear to do the opposite, promoting survival and metastasis. Amal and colleagues had previously demonstrated that nitric oxide drives glioblastoma progression. The question that remained was whether the same enzyme, neuronal nitric oxide synthase, was performing a similar service for neuroblastoma, and if so, through which downstream pathway. The answer turned out to be mTOR.

The team attacked nNOS from two directions. They treated human SH-SY5Y neuroblastoma cells with BA-101, a selective pharmacological inhibitor, at 100 μM for 24 hours. Separately, they silenced the nNOS gene with small interfering RNA. The reasoning was that if a drug and a genetic tool produce the same result, you are looking at biology, not pharmacological noise.

The experiments produced the same result. BA-101 reduced NADPH-diaphorase activity, the standard readout of NOS function, by 35-40%. Genetic silencing cut it by 45-50%. Nitrite levels, a stable proxy for nitric oxide production, fell 65-70% with BA-101 and 55-60% with siRNA. Colony formation, the most direct measure of proliferative capacity, dropped significantly after both BA-101 treatment (p < 0.001) and nNOS silencing (p < 0.01). The cells were losing their ability to multiply.

What followed downstream was systematic. Protein tyrosine nitration, measured by 3-nitrotyrosine immunoreactivity, fell sharply after BA-101 treatment (p < 0.01) and nNOS silencing (p < 0.001). The chemical signature of nitrosative stress was fading.

The results then confirmed that AKT phosphorylation decreased (p < 0.01 with BA-101; p < 0.05 with siRNA), while total AKT remained unchanged. Phosphorylation of mTOR itself declined under both conditions (p < 0.01 each). The downstream mTORC1 substrate ribosomal protein S6 followed (p < 0.05 with BA-101; p < 0.01 with siRNA).

And here, the most telling detail, that TSC2, a master negative regulator of mTOR signaling, rose significantly under both treatments (p < 0.05). Removing the nitric oxide signal had allowed the cell’s own braking system to re-engage. In summary, the authors noted, “Pharmacological inhibition of nNOS with BA-101 (100 μM, 24 h) or genetic silencing of nNOS with siRNA caused upregulation of the key negative regulator TSC2 and decreased phosphorylation of AKT, mTOR, and RPS6, indicating suppression of mTOR pathway activity.”

Synaptophysin, a neuroendocrine tumor marker used to gauge the malignant identity of neuroblastoma cells, decreased significantly with BA-101 (p < 0.01) and nNOS knockdown (p < 0.05). The tumor cells were not merely growing more slowly. They were becoming, at a molecular level, less recognizably cancerous. In summary, the investigators noted, “Our results show that inhibition of NO production in the human NB cell line (SH-SY5Y cells), either by pharmacological intervention using the selective nNOS inhibitor BA-101 (41) or by genetic ablation using the specific siRNA, successfully suppressed NB malignancy.”

Schematic model illustrating the NO-mTOR signaling axis in neuroblastoma. Under basal/pathological conditions (left panel), and nNOS inhibition (right panel). [Haitham Amal]
Schematic model illustrating the NO-mTOR signaling axis in neuroblastoma. Under basal/pathological conditions (left panel), and nNOS inhibition (right panel). [Haitham Amal]

But if blocking nitric oxide suppresses mTOR signaling, then flooding the cell with nitric oxide should amplify it. The researchers tested this by exposing SH-SY5Y cells to SNAP, a nitric oxide donor, at 200 μM for 24 hours. This converse experiment produced the converse result. 3-nitrotyrosine rose (p < 0.05), and TSC2 fell (p < 0.01). Phosphorylation of AKT, mTOR, and RPS6 all increased (p < 0.05 for each).

The team then tested their findings in a xenograft mouse model of neuroblastoma, treated with BA-101. “Importantly, to extend these findings to an in vivo context, we further assessed the impact of pharmacological nNOS inhibition on tumor growth in a xenograft NB model,” they stated. The investigators found that while tumors in control animals grew to approximately 1.5 cm in their largest dimension, the treated tumors did not. Final tumor volume and weight were dramatically reduced in the BA-101 group. ‘Quantitative analysis revealed a dramatic decrease in the final tumor volume and weight in the BA-101-treated group (p < 0.001) compared with controls,” they noted.

Body weight did not differ significantly between groups, suggesting that the compound was tolerated without gross systemic toxicity. In summary, the authors wrote, “Our finding demonstrate that the pro-tumorigenic effects of nNOS in SH-SY5Y involve activation of themTOR signaling pathway.” Importantly, both genetic inhibition of nNOS using siRNA and pharmacological inhibition with BA-101 effectively suppressed mTOR pathway activation and reduced malignant properties of NB cells, highlighting the therapeutic relevance of targeting nNOS signaling.  “These findings indicate that pharmacological inhibition of nNOS effectively suppresses xenograft tumor progression, highlighting the critical role of nNOS-derived NO in promoting neuroblastoma growth in vivo.”

“The magnitude of the in vivo suppression caught our attention,” said Amal, the study’s corresponding author, who holds appointments at the Institute for Drug Research, School of Pharmacy, Faculty of Medicine, The Hebrew University of Jerusalem, and the Rosamund Stone Zander and Hansjoerg Wyss Translational Neuroscience Center at Boston Children’s Hospital, Harvard Medical School. “We had demonstrated the role of nitric oxide in glioblastoma previously, but the consistency of the neuroblastoma results across every assay, from protein phosphorylation to colony formation to xenograft growth, points to nNOS as something more than a contributor. It appears to be a central driver of the signaling that sustains this tumor.”

Added first author Shashank Kumar Ojha, PhD, first author of the study and a researcher at the Institute for Drug Research, The Hebrew University of Jerusalem, added, “What convinced me was the concordance between the pharmacological and genetic approaches. When BA-101 and siRNA independently produce the same pattern of effects across NADPH-diaphorase activity, nitrosative stress markers, mTOR pathway phosphorylation, and clonogenic growth, you can be confident the biology is real. That reproducibility is what gives you a therapeutic hypothesis worth testing further.”

The authors acknowledged limitations to their study. The in vitro work relied on a single cell line, SH-SY5Y, which cannot capture the full genetic heterogeneity of neuroblastoma or the complexity of the tumor microenvironment. The chemical identity of BA-101 is currently undisclosed pending patent issuance, which means independent replication by other laboratories must wait. Whether nitrosative stress directly underlies its functional impairment, or whether an intermediary mechanism is involved, remains an open question that the authors explicitly flag for future investigation. “Future studies using patient-derived cells, organoids, or genetically engineered mouse models will be important to further validate and extend these observations,” they stated. Nevertheless, the authors suggest, the limitations do not diminish the central discovery of a druggable nNOS–mTOR axis.

mTOR inhibitors such as rapalogs and catalytic mTOR inhibitors have shown limited efficacy as monotherapies in neuroblastoma, undermined by feedback activation and resistance mechanisms. The present study suggests the potential for a different attack strategy. Rather than targeting mTOR at the lock, intervene upstream at the hand that turns the key. By reducing nitric oxide-dependent mTOR activation, nNOS inhibition may sidestep the compensatory pathways that have frustrated direct mTOR blockade. “Collectively, these results identify the nNOS-mTOR axis as a key driver of neuroblastoma progression and suggest that nNOS inhibition represents a promising strategy for NB treatment,” they concluded.

The post Neuroblastoma Tumor Growth in Mice Suppressed by Blocking Enzyme to Inhibit mTOR Signaling appeared first on GEN – Genetic Engineering and Biotechnology News.

Prefrontal and hippocampal microstructural gray matter following cognitive training under moderate hypoxia in mood disorders: a randomized controlled trial

BackgroundCognitive impairment persists during partial or full remission in 50–70% of individuals with mood disorders and impacts daily functioning and clinical prognosis. Preclinical evidence suggests that extended exposure to moderate hypoxia, combined with motor-cognitive learning, may elevate neuroplasticity and improve cognition. In these individuals with remitted mood disorders, we found that cognitive training under repeated moderate normobaric hypoxia improved executive function, and here investigate neurobiological mechanisms.MethodsParticipants with major depressive disorder (MDD) or bipolar disorder (BD) in partial or full remission were randomized to 3 weeks of 3.5-h daily normobaric hypoxia (12% O2) combined with cognitive training five to 6 days per week or treatment-as-usual (TAU). Participants were assessed with cognitive tests and diffusion-weighted MRI at baseline and 1 month after treatment completion (week 8) as part of the ALTIBRAIN trial (ClinicalTrials.gov: NCT06121206). Prefrontal and hippocampal gray matter microstructure were modelled with Neurite Orientation Dispersion and Density Imaging (NODDI).ResultsFifty-seven participants (mean age 39 years, SD: 13, 70% female) with baseline MRI data were included. No significant effects of hypoxia-cognition training vs. TAU on neurite density index (NDI) or orientation dispersion index (ODI) were observed in either the prefrontal cortex or hippocampus (all p-FDR ≥ 0.832). No significant associations were observed between microstructural changes and changes in cognitive function in either region (all p-FDR ≥ 0.721). At baseline, microstructure in both regions was not associated with executive function or global cognition (all p > 0.40).ConclusionThe absence of detectable microstructural changes, despite selective improvements in executive function, indicates that NODDI-derived metrics did not capture structural correlates of the cognitive response to hypoxia-cognition training. Whether this reflects functional neural mechanisms, measurement insensitivity, or the timing of the single follow-up assessment remains to be determined. Future studies should incorporate multiple imaging time points to capture the dynamic trajectories of putative microstructural brain changes.

Strength of Evidence to Support Decision-Making on the Use of Digital Mental Health Technologies in NICE Evaluations: Cross-Sectional Analysis of Studies

Background: Digital mental health technologies (DMHTs) are playing an increasing role in mental health services. The quality of evidence for DMHTs is variable, and there are concerns that evidence is not sufficient to support decision-making. Objective: This study used a cross-sectional analysis of evidence supporting DMHTs included in National Institute for Health and Care Excellence (NICE) evaluations to examine the strength of evidence available for decision-making. Methods: We identified all NICE evaluations relating to DMHTs by reviewing details of published NICE evaluations on the NICE website. From each of these evaluations, we identified included DMHTs and reviewed committee documentation to identify studies that provided supporting evidence for each of these technologies. We extracted information on a series of items relating to study quality and summarized the characteristics of evidence both at the level of individual studies and across the package of evidence from multiple studies supporting DMHTs. We also identified key evidence gaps in available evidence. Results: We included nine NICE evaluations relating to anxiety, depression, psychosis, insomnia, attention deficit hyperactivity disorder (ADHD), and tic disorders. These evaluations included 30 DMHTs and referenced 78 supporting studies. We identified common evidence gaps relating to effectiveness compared to relevant comparators, use of appropriate outcomes, including health-related quality of life, cost of delivery, and impact on resource use, and reporting of adverse events. Conclusions: Our study highlights that some DMHTs have been supported by high-quality studies and that evidence to support DMHTs is likely to be developed across a series of studies. However, there are often key evidence gaps that need to be addressed to provide a stronger case for adoption. Developers should ensure that they consider these gaps while planning evidence generation, and where possible, address them earlier in the product lifecycle.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/825f13db8cbad54213afa5c433d7adde" />

Analysis of the prevalence of dyslipidemia in early-onset schizophrenia patients and its correlation with clinical characteristics

ObjectiveTo analyze the prevalence of dyslipidemia and related influencing factors in patients with early-onset schizophrenia (EOS).MethodsWe recruited 289 pediatric and adolescent EOS patients from October 2021 to June 2024 in the Third People’s Hospital of Fuyang. Researchers gathered comprehensive demographic and clinical records. Utilizing the 2023 Chinese Guidelines for Lipid Management, they calculated dyslipidemia prevalence and the incidence of irregularities in total cholesterol, triglycerides, LDL cholesterol, HDL cholesterol, and non-HDL cholesterol. Subsequently, differences in dyslipidemia among different genders, body mass index, and antipsychotic medication groups were analyzed. Finally, independent influencing factors of dyslipidemia in EOS patients were explored.ResultsThe overall prevalence of dyslipidemia was 24.9% (72/289), with abnormal rates of TG, TC, HDL-C, LDL-C, and non-HDL-C being 15.9%, 6.6%, 6.6%, 4.2%, and 7.3%, respectively. Male patients, those who were overweight or obese, and those taking two antipsychotic drugs had significantly higher rates of dyslipidemia. Regression analysis showed that male gender (OR = 2.04, P = 0.016), overweight/obesity (OR = 4.55, P < 0.001), body roundness index (OR = 1.53, P = 0.005), and the use of two antipsychotic drugs (OR = 1.90, P = 0.030) were risk factors for dyslipidemia in EOS patients.ConclusionThe prevalence of dyslipidemia in EOS patients is relatively high. When monitoring lipid levels in clinical practice, particular attention should be paid to male patients, those who are overweight or obese, and those receiving combined drug therapy.

Esketamine ameliorates depression-like behavior in mice via modulation of the NRG1–ErbB4 pathway

BackgroundEsketamine has a significant and rapid antidepressant effect. Although studies have shown that Neuregulin 1 (NRG1) and it’s signaling pathway are associated with depression, the possible regulatory relationship of esketamine on the NRG1-ErbB4 pathway is not yet clear.MethodsTo induce depressive-like behavior in mice, a Chronic Social Defeat Stress (CSDS) model was established. Behavioral indicators were then employed to assess depression in these mice, categorized into control, susceptible, and resilient groups. Following intraperitoneal injection of a subanesthetic dose of esketamine, behavioral tests were conducted at 30 minutes and 24 hours post-injection to observe any improvements in depressive-like behavior. Additionally, changes in immunofluorescence and protein expression levels of NRG1-ErbB4 and GAD67 in the prefrontal cortex were evaluated.ResultsCompared with the control group, the CSDS susceptible group mice showed decreases in social interaction ratio in the contact area, sucrose preference ratio, NRG1 immunofluorescence protein expression in the prefrontal cortex and NRG1 expression in tissue homogenate; showed significant increases in immobility time; the expression of NRG1 decreased;no significant change in GAD67 and ErbB4 expression level. in After 30 minutes of intraperitoneal injection of esketamine, the expression of NRG1 in the prefrontal cortex of susceptible mice increased significantly. no significant change in GAD67 and ErbB4 expression level. After 30 minutes and 24 hours of intraperitoneal injection of esketamine, the social interaction ratio of susceptible group improved compared to the control group, and the duration of forced swimming immobility was significantly shortened.ConclusionThe subanesthetic dose of esketamine may regulate the NRG1-ErbB4 signaling pathway and improve depressive like behavior in mice.

Cerebellar dysconnectivity in schizophrenia spectrum: task-based functional connectivity analysis and cognitive stratification

IntroductionSchizophrenia is conceptualized as a disorder of brain network dysconnectivity, yet relationships between neural alterations, cognitive deficits, and genetic risk remain unclear.MethodsWe examined 86 participants: schizophrenia patients (SCZ), unaffected siblings (SCZ-SIB), healthy controls (CON), and control siblings (CON-SIB). We used a multiscale graph-theoretic analysis of task-based fMRI during N-back working memory and unsupervised clinical-cognitive clustering.ResultsWe found that reduced cerebellum-sensorimotor (CER-SM) and cerebellum-cingulo-opercular (CER-CO) connectivity during the 1-back condition robustly discriminated SCZ from CON (AUC = 0.89). Critically, these dysconnectivity patterns were linked to clinical state, present in SCZ vs. SCZ-SIB but absent in SCZ-SIB vs. CON-SIB, suggesting illness expression rather than familial risk. Unsupervised clustering revealed three data-driven subtypes with distinct cognitive- symptomatic profiles: subtype 1 with relative preservation of verbal abilities (predominantly controls), subtype 2 with marked fluid cognitive impairment (enriched in SCZ), and subtype 3 with intermediate performance with working memory sparing (mixed composition). Cerebellar-cortical hypoconnectivity showed graded alignment across these profiles.DiscussionThese findings demonstrate that cerebellar dysconnectivity is most detectable under moderate cognitive load, tracks with clinical state, and covaries with transdiagnostic cognitive profiles, advancing circuit-based understanding of schizophrenia heterogeneity.

It Is the Journey, Not the Destination: Moving From End Points to Trajectories When Assessing Chatbot Mental Health Safety

Large language models are rapidly becoming embedded in everyday life through artificial intelligence (AI) chatbots that people use for practical assistance and companionship, as well as for support with mental health and emotional wellbeing. Alongside clear benefits, clinicians and public reports increasingly describe a minority of users whose interactions seem to drift over days or weeks toward strongly questionable convictions, delusions or suicidal crises. Importantly, clinically meaningful deterioration can occur even without overtly unsafe text outputs, via more insidious processes such as compulsive use and sleep disruption, as well as withdrawal from human contact and progressive narrowing of attention around the chatbot relationship. In this Viewpoint, we argue that risk often arises not at a single tipping point but through trajectory effects that accumulate across extended dialogue, and that prevailing safety evaluation approaches are misaligned with this reality because they primarily score risk at discrete conversational endpoints often reached through scripted dialogues lasting just a single turn or several turns. Mental health benchmarks and safety suites (including clinician-informed efforts) have advanced the field by testing refusal behaviour, toxicity, and adversarial prompting, but they often treat the last message as the unit of analysis and therefore miss when risk-relevant relational cues, signs of validation, contradiction handling, and shifts in certainty first emerge and how they compound. We propose that mental health safety assessment should shift from endpoints to trajectories by 1) treating the whole dialogue, not just the end result, as the focus of evaluation; 2) reporting turn-by-turn dynamics such as delusion confirmation and harm enablement, as well as timing and persistence of safety interventions; and 3) calibrating short multi-turn tests against longer, clinically realistic interaction sequences that can reveal context-length effects and drift. We further argue that transcript-only evaluation is insufficient in mental health contexts. Similar language can reflect very different internal states, and the relationship between expressed psychopathology and real-world harm is non-linear. Safety research should therefore incorporate proximal human outcomes after interactions (e.g., shifts in certainty, openness to counterevidence, arousal, urge to continue, and subsequent sleep or behaviour) and build prospective clinical surveillance infrastructure that supports consented transcript donation and linkage to health outcomes. Together, these steps would enable benchmarks that are clinically relevant and better aligned with the kinds of harms now being observed in real-world chatbot use.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/67a4a0892a39533b2f5d0b2acc0e2689" />