Schizophrenia

Machine learning approaches may support individual-level classification in psychiatry, but many EEG-based schizophrenia studies have relied on small samples or conventional summary features. We evaluated whether trial-level auditory oddball event-related potential (ERP) waveforms could support schizophrenia versus healthy-control classification using deep learning. The study included 258 patients with schizophrenia and 142 healthy controls. EEG recordings from an auditory duration oddball paradigm were segmented into -100 to 500 ms epochs, and trial-level mismatch waveforms were generated by subtracting each participant's mean standard response from accepted deviant trials. Models were trained using a fixed participant-level training, validation, and test split, with demographic residualization fit only in the training set. Five deep learning architectures were trained on full residualized ERP waveforms and compared with classical machine learning models trained on 18 conventional ERP summary features. Deep learning models achieved higher test set discrimination than classical feature-based models, with AUROC values ranging from 0.797 to 0.857 versus 0.705 to 0.720. Benchmark analyses suggested that performance depended on the combination of waveform-level input and deep learning architecture. These findings support trial-level auditory oddball ERP waveforms as promising classification inputs and candidate electrophysiological biomarkers of schizophrenia-related neural information processing.

Recent computational approaches to speech in psychosis generate large multidimensional feature spaces capturing semantic and acoustic aspects of language production. However, the clinical relevance of individual measures often becomes difficult to interpret due to redundancy, interaction effects, and high intercorrelation among features. Building upon previous work constructing a single composite index derived from semantic features based on language model embeddings, we here build a second acoustic index derived from speech acoustic features. Our aim was to evaluate the differential performance of both indices in conjunction in PANSS symptom prediction in psychosis in a cross-linguistic setting, including positive symptoms (P1, P2, P3), negative symptoms (N1, N4, N6), and general measures (G5, G9). The dataset comprised five languages and 221 patients with schizophrenia spectrum disorder (SSD). Both indices showed predictive power for individual PANSS scores, while also demonstrating clinically important complementarity: semantic indices were more strongly associated with positive symptom dimensions (P2, P3, Total Positive), whereas acoustic indices showed stronger relationships with negative and general symptoms (N1, N4, G5, G9). Both domains shared predictive overlap for global measures, such as PANSS Total scores. These findings suggest that both indices capture complementary and partially overlapping dimensions of psychopathology. The proposed composite index framework contributes to the advancement of low-dimensional speech-derived markers of symptom severity variation, potentially informing vulnerability to relapse and remission in psychosis.

Large Language Models (LLMs) are proposed as tools for high-throughput, deep phenotyping of psychiatric disorders. Applied to electronic health records, LLMs could in principle extract patient symptoms, outcome trajectories, risk factors, and treatment history at scale and these, when combined with increasingly available biological data, such as genomics or neuroimaging, could provide powerful resources for health research. Although proof-of-principle LLM-based symptom extractions have been carried out for some medical conditions, the heterogeneous nature of psychiatric disorders, including schizophrenia, requires extensive domain-specific evaluations. Here, we evaluate 14 general-purpose LLMs for extracting eight symptoms of schizophrenia from clinical summaries in a severe mental illness cohort (N = 704). Performance across symptoms was poor to moderate (macro F1 = 0.500 - 0.647), with positive symptoms more accurately extracted than negative symptoms. Few-shot prompting, a common strategy for improving LLM task performance, did not significantly improve these results. Nevertheless, LLM-predicted and gold-standard positive symptoms demonstrated comparable associations with clinical variables in regression analyses. Individual-level extraction errors attenuated group-level associations but did so unevenly across symptom domains. This indicates that general-purpose LLMs may not be able to extract psychosis-related phenotypes from clinical summaries with the quality required for clinical use but might nevertheless be useful for exploratory research on large cohorts. However, closing the performance gap between positive and negative symptom extractions seems essential groundwork to prevent a systematic bias in any LLM-derived characterisations of psychotic symptoms.

Background and Hypothesis: First-line antipsychotics are ineffective in ~40% of people with first episode psychosis (FEP). Electroencephalogram (EEG)-based predictive biomarkers could be a scalable solution to reduce trial-and-error. Study Design: We acquired EEG in 56 minimally treated FEP individuals before treatment initiation, and 34 matched healthy controls (HC). Treatment response was defined as >50% improvement in BPRS scores by 12 weeks of treatment. Modified k-means clustering of resting-state EEG yielded six microstates including canonical topographies (A-D). The first principal component (PC1) of microstate properties derived from HCs served as a normative reference onto which FEP patients were projected. Baseline PC1 was used to predict treatment response in adjusted regression analyses. To test stability of results, an independent dataset of HCs (n=92) was fitted using the same microstate solution and compared to our native HC dataset. PC1 scores derived from this independent normative dataset were also used for predicting response in FEP. Study Results: Individual microstates replicated established patient-control differences (e.g., higher presence of microstate C compared to D, p<0.001). PC1 explained 34.8% of variance. Baseline PC1 predicted subsequent treatment response (adjusted OR=0.63, p<0.001), yielding AUC=72.3% in leave-one-out cross-validation (pperm<0.001) when combined with clinical information. PC1 from our HC transported well to the second healthy control EEG dataset (n=92) (r=0.9,p<0.001; r=0.89,p<0.001 ). PC1 in FEP derived using PC1 weights from a second normative dataset also predicted response in FEP (OR=0.70, p=0.03). Conclusions: EEG microstate architecture differing from the expected norm is associated with lower odds of treatment response in FEP.

Background and Hypothesis Identifying which patients with schizophrenia will remain stable and which will relapse is relevant to both trial design and clinical practice. We hypothesized that higher negative and lower positive symptoms at baseline would predict longer time to relapse. Study Design This analysis pooled individual participant data on 557 patients with schizophrenia randomized to placebo from five placebo-controlled antipsychotic withdrawal relapse-prevention studies, each with a double-blind phase of at least 6 months. Relapse was defined as a [&ge;]12-point increase in Positive and Negative Syndrome Scale (PANSS) total score from double-blind baseline at any time during the double-blind phase. Cox regression analyses modeled time to relapse with the three PANSS subscales, age and sex. Study Results PANSS Negative and Positive symptom subscales predicted relapse in opposite directions and at almost equal magnitude in the pooled model. Higher negative symptoms predicted lower relapse risk (hazard ratio (HR)=0.94 per point, 95% confidence interval (CI)=0.90-0.99) and higher positive symptoms predicted higher relapse risk (HR=1.06, 95% CI=1.00-1.13). PANSS General Psychopathology did not predict relapse (HR=0.98, 95% CI=0.95-1.02). Based on the pooled estimates, a 4-point higher baseline PANSS Positive score corresponded to a 27% higher relapse risk, while a 4-point higher baseline PANSS Negative score corresponded to a 21% lower risk. Conclusions These results may help predict relapse risk in clinical care and in clinical trial samples but require prospective validation.

Objective: Child and early adolescent onset psychosis (COP) is associated with subtle changes in language linked to thought disorder, a key contributor to functional impairment. Large language models (LLM) can detect deviations from expected language patterns by jointly analyzing sentence structure and word choice. This integrated information is captured by measures such as (pseudo)-perplexity, which quantify how difficult it is for an LLM to predict individual words given the surrounding linguistic context. The objective was to determine if perplexity measures change with age and whether they are altered in COP. Method: This study tested for a difference in perplexity in COP cases (N = 23) mean age 12.78 years as compared to controls (N = 15) mean age 11.67 years. Extensive manually transcribed interviews were analyzed (controls: 3,829; cases: 6,579 mean words). Results: Perplexity derived from Large Language Model Meta AI (LLaMA) had a significant negative correlation with age in cases but not in controls. Pseudo-perplexity derived from Bidirectional Encoder Representations from Transformers (BERT) did not have a significant correlation with age in either group. Group differences were evaluated using generalized linear models with (pseudo)-perplexity as the dependent variable, case status as the predictor and age and number of words as covariates. The model predicting perplexity was significant and case status significantly predicted perplexity. In contrast, the model predicting pseudo-perplexity was not significant. Conclusion: These differences in perplexity are interpreted as reflecting an altered developmental trajectory in the real-time semantic and syntactic planning that directs the flow of language in individuals with COP.

**Importance** Average gray matter volume (GMV) loss in early psychosis may conceal clinically relevant differences in brain change. **Objective** To identify longitudinal GMV-change patterns, their cross-site correspondence, and clinical associations. **Design, Setting, and Participants** Prospective cohort with T1-weighted MRI acquired during 2011-2026 at 2 Czech sites; median follow-up, approximately 13 months. Participants were 222 patients with first-episode schizophrenia-spectrum psychosis and 178 healthy controls frequency-matched by site and baseline age. **Exposure** First-episode psychosis vs healthy control status. **Main Outcomes and Measures** Diagnosis-blind clustering by site of image-quality-corrected, subsequently annualized GMV change across 333 cortical parcels; longitudinal voxel-based morphometry (VBM); Global Assessment of Functioning and consensus 5-factor Positive and Negative Syndrome Scale negative-symptom scores. **Results** Among 400 participants (mean [SD] age, 28.2 [7.5] years), 208 (52.0%) were male. Each site independently yielded 3 clusters: C1 (near-stable GMV), C2 (pronounced decrease), and C3 (increase). Cross-site centroid correlations across 333 parcels were strong for C2 and C3 (Spearman rho = 0.865 and 0.861); C1 correlation was weak (rho = 0.138), as expected for near-zero profiles with little regional variation. C2/C3 correspondence supported pooling matched site-specific clusters for post hoc analyses. Pooled clusters contained 77 patients and 125 controls in C1, 66 patients and 16 controls in C2, and 79 patients and 37 controls in C3. Suffixes -P and -HC identify patient and control subgroups. To situate patient changes relative to controls, post hoc VBM revealed an unanticipated gradient of increasingly pronounced GMV reduction across C1-HC to C1-P to C2-P, prominently involving association cortex, and bilateral ventral temporal GMV increase in C3-P (familywise error-corrected P < .05). The opposite contrast along this ordering showed periventricular changes consistent with progressively greater ventricular enlargement. C2-P had the lowest mean follow-up functioning and highest mean negative-symptom severity (adjusted omnibus P = .005 and .02); clinical changes from baseline did not differ significantly among patient groups. **Conclusions and Relevance** Early GMV change in psychosis was not uniform but comprised two contrasting directions: a clinically relevant, control-anchored pattern of increasingly pronounced GMV loss involving default mode regions and a distinct, directionally opposing ventral temporal increase.

Background and aims Over the past decade, there has been increased lifetime use of illicit and medical cannabis, underscoring the need to examine public health impacts. This included reports of increased cannabis-related psychosis. In this context, this study aimed to determine all cannabis-related hospital admissions over a ten-year period to characterise poisonings and mental health harms. Design This is a retrospective observational study of the Victorian Admitted Episodes Dataset (VAED) records, which include all hospital admissions in the state of Victoria and are supplied by the Victorian Department of Health. Setting Victoria, Australia. Cases Hospital records for poisoning (T40.7) and mental health and behavioural disorders (F12.0-F12.9) related to use of cannabis for the period July 2013 to June 2023 were selected for this study. Measurements Crude and age-standardised rates per 100,000 population are reported. Net percentage change over the ten-year study window was estimated using generalised linear models with a log-population offset, contrasting the final year (2022/23) against the baseline (2013/14). Findings There were 39,565 cannabis-related admissions in the ten-year period, including 7263 admissions where cannabis-related harm was the principal diagnosis. The age-standardised rate per 100,000 population was 63.6 for all cannabis-related harm admissions and 11.7 for principal diagnosis cases. The peak annual rate for all cannabis-related harm was observed in 2020/2021 (77.3 per 100,000). There was a 49.0% (p<0.0001) increase in the rate of hospital admissions with a cannabis-related principal diagnosis, when comparing 2022/23 with baseline (2013/14). Overall, harmful use (F12.1) was the most common cannabis-related mental and behavioural disorder. Psychotic disorder (F12.5) was the most common diagnosis (45.3%; n=2,909/6,421). When considering principal diagnosis only, the rates were highest in males aged 15-24 years (36.9/100,000), followed by females in the same age group (20.5/100,000). While most principal diagnosis cases involved males, a greater percentage change between the first and last years of the study period was observed for females (+75.1%, p<0.0001) compared with males (+34.6%, p=0.003). Conclusions A 49% increase in the rate of cannabis-related hospital admissions, particularly mental and behavioural disorders due to use of cannabinoids, has occurred over the past decade. These trends are non-linear, rates in the last two years were lower than the peak in 2020/2021.

Background: Schizophrenia has been conceptualised as a disorder of brain disconnectivity, with white matter abnormalities contributing to altered neural communication. However, fibre-specific white matter alterations in early illness and their potential sex-related differences remain insufficiently characterised. We applied fixel-based analysis to investigate sex-stratified white matter microstructural alterations in first-episode schizophrenia (FES). Methods: Diffusion MRI data from individuals with FES and healthy controls were stratified by biological sex. The male cohort included 79 patients and 29 controls, and the female cohort included 48 patients and 46 controls. Fixel-based analysis was performed separately in males and females to assess fibre density, fibre-bundle cross-section, and their combined measure. Head motion was included as a covariate in the fixel-based analyses. A descriptive follow-up voxel-scale analysis extracted mean fibre density from a binary mask derived from significant male fixels and applied the same region to the female cohort. Results: Males with first-episode schizophrenia showed significantly reduced fibre density compared with male controls, primarily in bilateral frontal white matter and a thin segment of the anterior corpus callosum. The voxel-scale follow-up confirmed lower mean fibre density in males within the significant fixel-derived mask. In contrast, females showed no significant group differences in any fixel-based metric or within the male-derived region. Conclusions: Early white matter microstructural alterations were detected in males with FES but not in females. These findings highlight a differential pattern emerging from sex-stratified analyses and support further investigation of sex-related differences in white matter alterations during the early stages of schizophrenia.

Background. Clinical meaningfulness in schizophrenia trials requires evidence that symptom change is recognizable in clinical practice. Established anchors include a 1 point minimal improvement on the CGI-S and a 7 to 10-point minimal improvement on the Personal and Social Performance scale (PSP), alongside a minimal improvement of 20% on the PANSS negative symptom factor score (NSFS) conventionally applied in negative-symptom trials. Methods. These clinically meaningful anchors were applied as benchmarks to pooled 12 week completer data from two randomized, placebo-controlled trials of roluperidone in adults with schizophrenia and moderate to severe negative symptoms. Results. Roluperidone was associated with higher responder rates than placebo on all three endpoints: at least 20% NSFS improvement (66/176 [38%] vs 39/182 [21%]; RR=1.75, 95% CI: 1.25 to 2.45; p=0.0008), at least 1 point CGI-S improvement (68/167 [41%] vs 48/177 [27%]; RR=1.50, 95% CI: 1.11 to 2.03; p=0.0077), and at least 10 point PSP improvement (63/176 [36%] vs 44/181 [24%]; RR=1.47, 95% CI: 1.06 to 2.04; p=0.018). Effects were consistent across symptoms, global-severity, and functional outcomes, with no evidence of heterogeneity between studies. Conclusions. Across all three anchor criteria, roughly 1.5 to 1.75 times as many patients treated with roluperidone as with placebo reached thresholds of improvement that clinicians and patients would recognize as clinically meaningful, supporting its relevance for persistent negative symptoms in schizophrenia.

In about a quarter of people with schizophrenia-spectrum disorder (SSD), the illness is unresponsive to standard antipsychotic treatment, yet the biological mechanisms underlying this remain poorly understood. Although such treatment-resistant schizophrenia (TRS) shares a substantial genetic liability with treatment-responsive schizophrenia, the limited efficacy of dopamine antagonists in TRS indicates that mechanisms beyond dopamine signalling likely contribute to treatment-resistance, requiring the identification of alternative biological pathways. This is the first cross-ancestry genetic study to investigate the genetic architecture of TRS, by directly comparing patients with treatment-resistant and treatment-responsive schizophrenia in two independent Hong Kong (N=798) and STRATA-G consortium (N=1243) cohorts. Using an integrated multi-level analytic framework, we conducted a genome-wide association study (GWAS) with gene-based and gene set-based analyses, pathway polygenic-risk-scores, and transcriptome-wide association study (TWAS). We further conducted gene-set enrichment analysis focusing on expert-curated synaptic pathways and brain tissues. Genetic signals at the gene, pathway, and genetically predicted expression levels were identified within each ancestry. Whereas limited power constrained individual loci discovery and cross-ancestry concordance, enrichment analyses indicated heterogeneous signals across cohorts, including differences in effect direction, but highlighted cohort-specific, synapse-related biology, particularly pathways involved in presynaptic vesicle dynamics, neurotransmission, and synaptic organization. Collectively, these findings highlight synaptic biology as one potential pathway-level signal from common-variant genetic effects associated with treatment resistance in SSD, despite minimal SNP-level discovery. Our work suggests there is promise in pathway-level and multi-omics approaches to elucidate biologically meaningful heterogeneity within SSD and provides support for synaptic mechanisms as potential targets for understanding and stratifying treatment-resistance.

Apathy is a major driver of long-term disability in schizophrenia, yet existing accounts focus on how patients evaluate the cost of action while neglecting their capacity to generate options for action in the first place. Using a brief, culture-fair option generation task (OGT) within the international CHANSS study, we examined this process in 150 patients with different stages of schizophrenia and 100 healthy controls across the UK, Spain, and China. The OGT requires drawing as many different paths as possible between two points on a touchscreen, yielding two measures: fluency (the number of paths that they generate) and uniqueness (the distinctiveness of those paths). Patients drew as many paths as controls but produced options roughly half as unique. Across patients, uniqueness was positively associated with verbal fluency and depressive symptom severity, but negatively associated with self-reported behavioural apathy. This suggests that increased apathy is associated with reduced ability to generate unique options (paths on the task). To examine the cognitive mechanisms further, we decomposed uniqueness into four exploratory measures: roaming entropy (spatial breadth of exploration), option elaboration (within-path development), perseveration (local repetition), and novelty maintenance (sustaining diversity over time). Together these explained 69% of the variance in motor-residualised uniqueness and attenuated the group difference by 80%, with roaming entropy as the dominant contributor. K-means clustering revealed three distinct patient profiles, each with a different clinical signature. A small subgroup generated options much like the control group. A larger subgroup retained spatial exploration but developed each option less and reported more depressive symptoms. A third subgroup showed broadly impaired exploration and the lowest verbal fluency. These findings establish option generation as a measurable behavioural phenotype in schizophrenia that relates to apathy and can be decomposed into specific cognitive mechanisms. They suggest that an impoverished option space, marked by reduced generation of unique possibilities for action, may be a remediable component of apathy.

We conducted an open-label pilot study of bilateral nucleus accumbens-anterior limb of the internal capsule (NAc-ALIC) deep brain stimulation (DBS) in three patients with treatment-resistant schizophrenia (TRS). Clinical symptoms, global functioning, cognition, stimulation parameters and intracranial local field potential (LFP) were assessed over 24 weeks. Two patients met the prespecified clinical response criterion, defined as a [&ge;]40% reduction in Positive and Negative Syndrome Scale (PANSS) total score, by week 8 and showed PANSS reductions of 69.0% and 69.4%, respectively, at week 24. The third patient showed a 18.7% reduction at week 24. Responders showed improvements in global functioning and selected cognitive domains. Longitudinal LFP recordings revealed increased gamma-band relative power and reduced theta-band activity in responders, whereas these changes were less consistent in the non-responder. DBS was generally well tolerated, with transient stimulation-related adverse effects resolving after reprogramming. These findings provide preliminary clinical evidence supporting further investigation of NAc-ALIC DBS in TRS and suggest that individualized programming and intracranial physiological measures may contribute to the identification of treatment-response signatures.

The potential role of the cerebellum in the pathophysiology of schizophrenia (SCZ) has received relatively insufficient attention. Prior studies characterized cerebellar abnormalities mainly at the group level, without fully considering the heterogeneity of the disease. Here, we applied a machine learning approach (Subtype and Stage Inference, SuStaIn) to cross-sectional 3D volumetric MRIs of the cerebellum, including 1,588 individuals with SCZ (638 females; mean age: 31.9 +/- 12.0 years) and 2,341 healthy controls (HC) (1,041 females; mean age: 33.3 +/- 13.6 years), from 17 sites worldwide. SuStaIn identified three distinct spatiotemporal trajectories in cerebellar gray matter volume (GMV) reduction, respectively originating in lobule X (subtype 1), lobule III (subtype 2) and lobule VIIb (subtype 3), with subtypes 1 and 3 corresponding to the posterior lobe and subtype 2 to the anterior lobe. These cross-sectionally inferred trajectories were replicated in two independent samples of 1,334 and 530 patients, respectively. Multimodal analyses using neuroimaging, transcriptomic and behavioral data revealed subtype-specific biological characteristics in brain morphological patterns, cerebellar-cortical connectivity, gene expression and clinical symptoms. Specifically, subtype-related genes were enriched in metabolism-related processes and immunity-related processes, respectively, in subtype 1 and subtype 2; subtype 3 showed more severe brain abnormalities and worse cognitive symptoms. Treatment data from 381 patients, with up to 12 months of follow-up, revealed poorer response to antipsychotic medications (APM) in subtype 2 and subtype 3, but better response to transcranial magnetic stimulation (TMS) in subtype 1, which also had worse emotion-related symptoms. Together, our findings offer a comprehensive characterization of the heterogeneity of cerebellar pathophysiological progresses in SCZ, which may help in developing future clinical stratification and targeted intervention strategies.

Background: Nicotine dependence is a major contributor to early mortality in individuals with schizophrenia, yet effective treatments remain limited. Drug cue-reactivity is a key contributor to nicotine use. Although transcranial magnetic stimulation (TMS) can reduce nicotine use and cue-reactivity, conventional left dorsolateral parietal (DLPFC)-targeted TMS is less effective in those with schizophrenia, highlighting the need for alternative, circuit-based targets. Because default node network (DMN) function is implicated in schizophrenia, cue-reactivity, and nicotine use, we investigated whether DMN-targeted TMS modulates cue-elicited brain activity in nicotine-using individuals with schizophrenia. Study Design: 63 nicotine-using individuals (schizophrenia: n=31, non-psychosis control: n=32) participated in a randomized, crossover study comparing DMN-targeted and DLPFC-targeted TMS with pre/post-TMS neuroimaging and craving assessment. A nicotine cue-reactivity task was used concurrently with neuroimaging. Mixed-effect models were used to determine effects of TMS target and diagnosis, and model relationship between baseline craving and cue-reactivity change. Study Results: At baseline, we found a Cue Reactive Network composed of regions overlapping with the DMN, including clusters in the posterior cingulate cortex, medial prefrontal cortex, and bilateral lateral occipital cortex (voxel-wise p<0.005, cluster p<0.05). Baseline cue-reactivity in schizophrenia did not differ from controls. DMN-targeted TMS decreased cue-reactivity in schizophrenia across Cue Reactive Network (estimate=-0.180, p=0.009). DLPFC-targeted TMS did not impact cue-reactivity. We did not observe an effect of either TMS target in controls. Higher baseline unprovoked craving predicted larger decrease in cue-reactivity across the Cue Reactive Network (estimate=-0.034, p=0.004). Conclusion: DMN-targeted TMS is a promising intervention that modulates nicotine cue-elicited brain activity in schizophrenia.

Individuals with schizophrenia show well-documented impairment in metaphor comprehension, often exhibiting a bias toward concrete, literal interpretations. While this tendency has traditionally been linked to psychopathological and cognitive factors, the contribution of perceptual processes remains underexplored. Here, we tested the hypothesis that figurative language impairment reflects altered perceptual processing, whereby the visual representations evoked by metaphors remain abnormally active and hinder abstraction. A sample of 143 individuals with schizophrenia and healthy controls was administered a novel paradigm where metaphors (e.g., Wisdom is a flashlight) served as primes for target words related to the metaphor vehicle based on visual (e.g., microphone), action (e.g., remote), or semantic features (e.g., lamp). While in healthy participants metaphors activated semantically associated words, individuals with schizophrenia showed sustained visual priming, emerging 1000 ms after metaphor presentation and persisting up to 1400 ms, with both groups showing reverse priming for action targets. Critically, greater visual priming predicted lower metaphor comprehension in patients, whereas greater semantic priming was correlated with better metaphor skills in controls. These results suggest that visual-perceptual representations are not only overactivated in patients compared to controls during metaphor processing but may also interfere with figurative comprehension. We argue that concretism arises from an imbalance between bottom-up sensory signals and top-down contextual priors, leading to the persistence of the visual representations and impaired abstraction. More broadly, these results support multimodal and predictive accounts of metaphor processing and point to altered perceptual dynamics as a previously unappreciated mechanism contributing to pragmatic impairment in schizophrenia.

Objective. NMDA-receptor hypofunction has been implicated as a core pathophysiological mechanism in schizophrenia, but conflicting evidence exists regarding the locus of the hypofunction--i.e. predominantly affecting receptors on pyramidal neurons or interneurons. To investigate the locus of NMDA-receptor hypofunction we integrated imaging data from humans and mice with biophysical model simulations. Methods. Translational imaging measures--intrinsic neural timescales (INT) and spontaneous co-activation pattern similarity--were investigated in three schizophrenia-relevant mouse models (chronic ketamine administration, n=6; Df(16)A+/-, n=7; and SETD1A+/-, n=9) and matched control mice (n=5; n=7; and n=8, respectively) using two-photon calcium imaging, and individuals with schizophrenia (n=127) and matched non-clinical controls (n=152) using functional MRI. The stabilized supralinear network biophysical model was used to simulate NMDA-receptor hypofunction and its impact on the translational imaging measures. Results. We identified convergent neural phenotypes of shorter intrinsic neural timescales (INT) and reduced spontaneous co-activation pattern similarity in both the schizophrenia-relevant mouse models and the individuals with schizophrenia compared to matched controls. In silico simulations using the stabilized supralinear network biophysical model recapitulated the convergent neural phenotypes through reduced excitatory-to-excitatory connection strength (i.e. NMDA-receptor hypofunction on pyramidal neurons) and predicted that INT and pattern similarity are correlated; this prediction was confirmed in both the mouse and human data. Conclusions. These results identify NMDA-receptor hypofunction on pyramidal neurons as a general and translationally valid mechanism for NMDA-receptor hypofunction in schizophrenia. These convergent translational findings suggest prioritizing the development of treatments for schizophrenia that increase recurrent excitation.

Objective: To determine whether the semaglutide-NAION association reflects true toxicity or reporting bias. Research design and methods: We performed temporal disproportionality analysis of FDA Adverse Event Reporting System (FAERS) data (2018Q1-2026Q2) to detect reporting anomalies and class-level spillover, and GLP1R cis-eQTL drug-target Mendelian randomization (MR) of NAION-relevant ocular phenotypes to test causal effects of GLP1R activation on semaglutide-relevant ocular toxicity. Results: Semaglutide ROR was 116.8 (95% CI 106.8-127.7). FAERS rates rose from below 8 per 10,000 annually (2018-2023) to 52 (2024), 177 (2025), and 203 (2026H1) after July 2024 publication, while comparators remained stable. Median time to onset was 246 days (Weibull {beta} = 1.09), a random-type profile inconsistent with cumulative toxicity. These anomalies were complemented by GLP1R-targeted MR evidence: no causal effect of genetically proxied GLP1R activation (all P [&ge;] 0.11), while confirming glucose-lowering (P = 7.6E-6). Conclusions: Pharmacovigilance and genetic evidence converge to show that the semaglutide-NAION signal reflects notoriety bias rather than toxicity.

Introduction: Auditory verbal hallucinations (AVH) affect 60-80% of people with schizophrenia, yet existing assessment tools inadequately capture their phenomenological complexity. Aim: To determine whether individuals with schizophrenia spectrum disorders could accurately match auditory verbal hallucination loudness to an external audio track, and to explore participant perspectives on this approach. Methods: Eight participants with schizophrenia spectrum disorders and active AVHs rated loudness via a Likert scale and by adjusting a headphone audio track to match their experience. Structured interviews and thematic analysis captured participant viewpoints. Results: No significant correlation was found between audio tool and Likert scale scores. Seven of eight participants reported the audio tool provided greater precision in quantifying AVH loudness and better enabled them to convey their internal experience to others. Discussion: Audio-matching tools may offer meaningful advantages over traditional scales for quantifying AVH loudness, even where statistical convergence with existing measures is absent. Limitations: Small sample size; loudness alone cannot fully capture the qualitative experience of hearing voices. Implications: This tool shows promise for longitudinal tracking of AVH loudness. Recommendations: Further investigation of digital approaches to AVH assessment is warranted.

Abstract Background Negative symptoms are persistent determinants of disability in schizophrenia and often respond incompletely to antipsychotic treatment. Objective To compare the efficacy of pharmacological and non-pharmacological add-on treatments for prominent or persistent negative symptoms using network meta-analysis. Methods This systematic review followed PRISMA 2020 and PRISMA-NMA and was registered in PROSPERO (CRD420261422218). PubMed, Europe PMC, Semantic Scholar and Crossref were searched from inception to 29 July 2026, supplemented by citation chasing. Eligible studies were randomized controlled trials in adults with DSM/ICD schizophrenia or schizoaffective disorder, prominent or persistent negative symptoms, stable antipsychotic treatment and an adjunctive intervention. Outcomes were Positive and Negative Syndrome Scale negative subscale or Scale for the Assessment of Negative Symptoms scores. Random-effects frequentist networks estimated standardized mean differences (SMDs) with 95% confidence intervals (CIs); negative values favoured add-on treatment. Risk of bias was assessed with Cochrane RoB 2. Results Forty-nine unique RCTs met the clinical and design criteria, of which 28 (2,067 randomized; 1,933 analysable participants) contributed to the locked quantitative dataset. The primary connected network included 24 trials, 25 treatments and 31 comparison estimates. Fourteen add-ons had CIs excluding the null versus a broad control node. The highest P-scores were observed for mirtazapine (SMD -2.38, 95% CI -3.55 to -1.21), granisetron (-1.96, -2.74 to -1.17), tropisetron (-1.82, -2.59 to -1.05), minocycline (-1.78, -2.55 to -1.01) and memantine (-1.54, -2.28 to -0.80). Heterogeneity was low ({tau} = 0.119; {tau}2= 0.014), but the predominantly star-shaped network had zero inconsistency degrees of freedom. Overall RoB 2 judgements were low for eight trials, some concerns for 15 and high for five. Conclusions Several pharmacological and non-pharmacological add-ons showed potentially important efficacy signals. Because most nodes were informed by single small trials, direct active comparisons were scarce, inconsistency could not be evaluated and risk-of-bias concerns were common, the treatment hierarchy should be considered hypothesis-generating rather than a basis for firm clinical recommendations. Registration: PROSPERO CRD420261422218 Keywords: schizophrenia; negative symptoms; adjunctive treatment; network meta-analysis; randomized controlled trial; neurostimulation Key Points This review compares pharmacological and non-pharmacological adjuncts in adults selected for prominent or persistent negative symptoms while receiving stable antipsychotic medication. Mirtazapine, granisetron, tropisetron, minocycline and memantine had the highest P-scores, while tDCS, rTMS and body-oriented psychotherapy also showed efficacy signals versus broad control. The evidence network was sparse and star-shaped, most interventions were supported by one small trial, and inconsistency was not estimable; rankings therefore require cautious interpretation.