Deep brain stimulation (DBS) efficacy for psychiatric disorders remains heterogeneous due to incomplete mechanistic insights and insufficient individualized treatment planning. A biologically grounded computational framework is urgently needed to simulate DBS-related differences for treatment-resistant patients.

Resting-state fMRI data from 11 patients with obsessive-compulsive disorder (OCD) and 11 with major depressive disorder (MDD), scanned in DBS-ON and DBS-OFF states, were used to derive functional connectivity (FC) via network deconvolution. Whole-brain dynamic modeling incorporated stimulation targets, amplitudes, and frequencies to simulate DBS effects in silico.

The model recapitulated syndrome-specific DBS-related network patterns in exploratory analyses. OCD stimulation primarily altered frontal-subcortical connectivity, whereas MDD stimulation predominantly affected default mode network hubs (p < 0.05). Under leave-one-out cross-validation, correlations between predicted and observed clinical outcomes were r = 0.625 in OCD (p = 0.040) and r = 0.659 in MDD (p = 0.028). Applying the same FC signatures to simulated FC yielded r = 0.465 in OCD (p = 0.149) and r = 0.627 in MDD (p = 0.039). In a proof-of-concept application to a separate OCD cohort (N = 92), three individuals showed high model-predicted NAc-response profiles and predicted improvement exceeding 35%.

This hypothesis-generating framework retrospectively models DBS-related network effects using individualized resting-state FC and virtual perturbations. Linking simulated FC states with empirical DBS-ON/OFF patterns may help characterize how baseline functional architecture shapes DBS-related network responses and inform future individualized DBS modeling.

Major depressive disorder (MDD) is associated with increased stroke risk and can be conceptualized in Traditional Chinese Medicine (TCM) as emotional constraint or “Qi Stagnation” that may contribute to “Blood Stasis.” However, the country-level ecological relationship between stroke burden and MDD prevalence remains unclear, and high-stroke settings may have under-recognized MDD because of differences in data availability, diagnostic practices, mental health services, and cultural symptom expression. This study quantifies this ecological association and interprets it through an Integrated Medicine framework.

Using Global Burden of Disease (GBD) 2023 data, we analyzed age-standardized stroke Disability-Adjusted Life Year (DALY) rates for all stroke and age-standardized prevalence rates for the GBD cause category of MDD across 204 countries and territories from 1990 to 2023. An exploratory Diagnostic Gap Score (ZStroke—ZMDD) was developed to identify countries and territories with disproportionate stroke burden relative to reported MDD prevalence.

A negative ecological correlation emerged globally and strengthened over time (r = −0.059 in 1990 versus r = −0.218 in 2023, P = 0.002). The highest stroke burden quartile (Q4) reported lower MDD prevalence (2639.9/100 000) than the lowest quartile (3067.0/100 000; mean difference 427.1/100 000, 95% CI: 121.52 to 732.58), despite an 8.3% increase in stroke burden in Q4. The score identified clusters with disproportionate stroke burden relative to MDD prevalence in Eastern Europe, Southeast Asia, and Pacific Island settings.

Low reported MDD prevalence in high-stroke countries and territories suggests a potential diagnostic, reporting, or data-quality mismatch rather than biological protection. These ecological findings do not prove that undiagnosed MDD causes stroke burden, but they support culturally adapted mental health screening in primary care, stroke-prevention clinics, and community health settings, with referral pathways for those who screen positive.

Explore/exploit decision-making is a crucial component of motivated behavior. Explore/exploit behaviors are related to disparate cognitive and reward-related functions and have been implicated in multiple neuropsychiatric conditions, including major depressive disorder (MDD). Several mediators of explore/exploit decision-making have been proposed; metabolic mechanisms have been theorized to potentially affect its underlying processes. Here, we examined the effects of a GLP-1 receptor agonist (GLP-1RA), semaglutide, on explore/exploit behavior in patients with MDD, using a four-armed restless bandit task.

In this secondary analysis of a randomized clinical trial, 72 participants with MDD were randomized either to oral semaglutide (n = 35) or placebo (n = 37), and completed the bandit task at baseline and following 16 weeks of treatment. Explore/exploit decision-making was assessed using computational modeling of choice behavior, with a trial-by-trial reinforcement-learning model.

Treatment with semaglutide resulted in increased exploitation (i.e. choosing the option with the highest expected reward) (Average Treatment Effect [ATE] = 0.0242, p < 0.001), directed exploration (i.e. exploring a choice with higher uncertainty) (ATE = 0.0619, p = 0.013), and perseveration (i.e. the value-independent tendency to repeat the previous action) (ATE = 0.3017, p = 0.013). These effects were moderated by, but ultimately independent of, baseline characteristics and weight loss during the trial.

Treatment with semaglutide modulated explore/exploit behaviors in patients with MDD. These results highlight the role of metabolic mechanisms in decision-making processes in MDD and expand knowledge of the potential neuropsychiatric effects of GLP-1RAs.

Immune-inflammatory dysregulation and cortical structural vulnerability have been increasingly implicated in major depressive disorder (MDD). However, how they interact with each other, and whether immune heterogeneity in MDD is systematically organized by structural networks, remain unclear.

Genome-wide association studies data of MDD, gray matter volume (GMV) phenotypes, and 19 immune-related traits were utilized to assess causal relationships. A prospective cohort comprising 122 patients with MDD and 66 healthy controls underwent structural magnetic resonance imaging and assessment of 40 inflammatory and immune-related markers. Individualized structural covariance network analysis, machine-learning approaches, and survival analyses were used to identify neurobiological subtypes and evaluate treatment outcomes.

Genetically predicted MDD was associated with lower levels of circulating IGF-1 (pFDR = 0.013). Levels of IL-5 and leptin were negatively associated with GMV alterations in the right cingulate gyrus and bilateral putamen, respectively (pFDR < 0.05). Structural covariance abnormalities were predominantly localized to connectivity between the default mode and limbic networks and within the limbic network in MDD, and they were classified into a hypoconnectivity subtype (n = 49) and a hyperconnectivity subtype (n = 73). The hypoconnectivity subtype exhibited greater anxiety and anhedonia symptoms, whereas the hyperconnectivity subtype showed broadly elevated inflammatory and immune state and achieved antidepressant response more rapidly during follow-up (p = 0.032).

This study supports a bidirectional immune–brain framework in MDD and suggests that structure-derived subtypes classify different immune signaling heterogeneity, which may provide a biologically informed basis for stratification and prognosis in MDD.

Bipolar disorder (BD) is characterized by mood fluctuations and is prone to misdiagnosis as major depressive disorder (MDD). Quantifying the relationship between symptom fluctuations and conversion risk is essential for early identification.

In a national prospective cohort of patients with MDD (enrolled from May 2022 to February 2025), depressive and manic symptoms were assessed using clinician-rated and self-report instruments. Two measures were calculated: the time in target range (TTR), the proportion of numbers of evaluations with scores below predefined clinical cutoffs; and Statistical Process Control (SPC) stability, evaluated with a person-specific approach defining each individual’s normal fluctuation range via their 25th–75th percentiles. The outcome was the incidence of manic or hypomanic episodes that met the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition criteria. Cox regression models were used to evaluate the associations of the TTR and SPC with the conversion to BD.

Among the 2,367 eligible patients with MDD (mean [SD] age, 29.88 [9.79] years; 792 [33.46%] male), 69 (2.92%) converted to BD in the 52-week follow-up. During weeks 0-8, higher TTR of the Quick Inventory of Depressive Symptomatology-Self-Report (QIDS-SR16) (cutoff set at ≤10) and the 33-item Hypomania Checklist (HCL-33) (cutoffs set at ≤7 and ≤15) were significantly associated with a lower risk of conversion between Weeks 8 and 52. Additionally, a greater SPC-evaluated stability of the HCL-33 scores was significantly associated with lower conversion risk over the same period.

While further validation in independent samples is still needed, continuous monitoring of hypomanic symptoms, especially evaluated with the HCL-33, may be of clinical significance for the early identification of conversion to BD.

Treatment-resistant depression (TRD) is a debilitating condition occurring in major depressive disorder (MDD) and bipolar II disorder (BP-II). Intermittent theta burst stimulation (iTBS) is an established neuromodulation approach for TRD, but standard delivery requires once-daily sessions over several weeks. Accelerated iTBS (aiTBS) may reduce treatment burden by temporally compressing delivery. We evaluated the implementation, tolerability, and clinical outcomes associated with a 5-day aiTBS schedule for adults experiencing TRD in MDD and BP-II.

In this single-center, open-label, uncontrolled, prospective naturalistic cohort study, adults with MDD or BP-II experiencing TRD received 30 iTBS sessions (600 pulses/session; total 18 000 pulses) over 5 days with 50-minute inter-session intervals. HAMD-17 was assessed at baseline, post-treatment, and 2-month follow-up. Secondary outcomes included Patient Health Questionnaire-9 (PHQ-9), Generalized Anxiety Disorder-7 (GAD-7), baseline Clinical Global Impression-Severity (CGI-S), post-treatment/follow-up CGI-I, response/remission rates, and adverse events. Longitudinal changes were evaluated using linear mixed-effects models.

Forty-six participants contributed longitudinal data (mean age 44.3 years; 54.4% male). HAMD-17 response rates were 28.3% (n = 13) post-treatment and 37.0% (n = 17) at follow-up; remission rates were 26.1% (n = 12) and 32.6% (n = 15), respectively. Improvements were also observed on PHQ-9 and GAD-7. On CGI-I, 52.2% were rated improved post-treatment and 60.9% at follow-up. The most common adverse event was transient, non-serious stimulation-site pain.

The 5-day aiTBS schedule was logistically feasible, well-tolerated, and associated with reductions in depressive and anxiety symptom scores in a real-world TRD cohort. Randomized sham-controlled trials are needed to confirm these findings.

Mental and physical disorders are leading contributors to global disability. However, estimates of their health-related burden often overlook the pervasive comorbidity among conditions. This study aimed to quantify the independent and combined effects of these disorders on self-rated health in nationally representative Chinese adults, explicitly accounting for comorbidity.

Data were drawn from the China Mental Health Survey (CMHS), a cross-sectional nationally representative survey. Mental disorders were assessed using the Composite International Diagnostic Interview 3.0, and physical disorders were measured through structured self-report. Health-related burden was assessed using a 0–100 visual analogue scale (VAS) of self-rated health over the past 30 days. Nested regression models examined disorder-specific and multimorbidity effects, followed by interaction-based counterfactual simulations to estimate disorder-specific burden net of comorbidity at individual and population levels.

Among 27,505 respondents included in multivariate analyses, comorbidity was highly prevalent. In crude models, all 20 disorders were associated with lower VAS scores, with the largest decrements observed for mental disorders such as generalized anxiety disorder (GAD), bipolar disorder and major depressive disorder (MDD). After accounting for disorder type and comorbidity structure, many associations were attenuated and the independent effect of disorder count became non-significant. Counterfactual simulations showed the largest individual-level burdens for GAD (–14.71), cancer (–9.02), bipolar disorder (–8.59) and MDD (–8.53). At the population level, insomnia symptoms (–1.56), arthritis (–0.81) and cardiovascular disorders (–0.64) contributed the greatest burden.

Accounting for comorbidity substantially alters estimates of disorder-specific health burden. Mental disorders impose the greatest impairment at the individual level, whereas prevalent physical conditions dominate population burden.

Genome-wide association studies in East Asian populations have identified the gene variant, rs12415800, of the Sirtuin 1 gene (SIRT1) as a major depressive disorder (MDD) risk locus with the A allele conferring increased risk. Accumulating evidence suggests that SIRT1-related genetic effects extend beyond MDD.

To investigate the association between rs12415800 and bipolar disorder, with additional analysis of MDD.

We analysed data from the Taiwan Biobank linked to the National Health Insurance Research Database, including 145 780 participants of Han Chinese ancestry. Clinically diagnosed bipolar disorder and MDD were identified using standardised ICD-9-CM and ICD-10-CM codes. Multivariable logistic regression models were applied with the GG genotype as the reference, adjusting for age and sex. Stratified analyses were conducted by age group and sex.

Both the rs12415800 AG and AA genotypes were associated with an increased risk of bipolar disorder. Compared with GG genotype carriers, AA genotype carriers had a significantly higher risk of bipolar disorder overall (adjusted odds ratio (aOR) = 1.20, 95% CI = 1.05–1.37), particularly among participants aged ≥50 years (aOR = 1.38, 95% CI = 1.14–1.68). The overall and age-stratified associations remained materially similar after excluding participants with schizophrenia or schizo-affective disorder, although sex-specific findings were inconsistent.

Our findings suggest that genetic variation near SIRT1, particularly the rs12415800 risk A allele, may be associated with bipolar disorder susceptibility, especially among older individuals. Given prior evidence linking this allele to reduced SIRT1 mRNA expression, further investigation of SIRT1-related pathways in the pathophysiology of bipolar disorder is warranted.

When deciding on antidepressant maintenance with clinicians, remitted patients must weigh the burdens of continued use against the benefit of less time depressed. This benefit is often inferred qualitatively from relapse-frequency data, even though expected time depressed also depends on episode duration, antidepressant re-initiation, and how evidence from highly recurrent populations applies to patients with fewer prior episodes.

We developed a health-state transition model to combine evidence on recurrence, episode duration, re-initiation, and treatment effects into expected time depressed and time taking antidepressants under maintenance versus active monitoring. Adults in remission were simulated for 5 years across groups based on depression history. We summarized the trade-off as additional antidepressant use required to avert 1 month of depression. The decision framework links this estimate to patient preferences by comparing required with acceptable use, allowing patients who expect greater consequences of relapse or experience fewer treatment burdens to consider more antidepressant use acceptable.

Maintenance efficiency varied substantially with recurrence history. Each month of depression averted required 1.5 additional years of antidepressant use among patients with 4–5 prior episodes; 2.8 years with 2–3 prior episodes, 5.1 years after one moderate-to-severe episode, and 11.8 years after one mild episode. When these trade-offs are acceptable, patients are modeled to prefer maintenance.

The model expresses the maintenance trade-off as a ratio of benefits and burdens. Preferences and clinical judgment determine for whom the ratio is acceptable. Empirical work can determine whether this improves maintenance decisions in practice.

Real-world outcomes of repetitive transcranial magnetic stimulation (rTMS) for depressive disorders vary, and generalizable predictors of response and remission remain unclear.

To identify demographic, clinical, and treatment-related predictors of rTMS outcomes using a multinational registry.

We pooled registry data from six countries between January 2017 and April 2026. Patients with depression treated with rTMS were included. Primary outcomes were response and remission. Predictors were evaluated using hierarchical logistic regression. Model 1 was the primary complete-case core model, with sensitivity analyses using multiple imputation and excluding clinical-trial-derived patients. Model 2 incorporated treatment-dose and medication variables, whereas Model 3 explored clinical and sociodemographic variables with substantial missingness and was considered hypothesis-generating. Clinician-rated and patient-reported outcomes were also analyzed separately.

Among 2,015 cases, 1,691 were analyzed. Response and remission rates were 43.6% and 28.7%, respectively. More treatment sessions were associated with higher odds of response (adjusted odds ratio (aOR) = 1.02 per session, p = 0.002), whereas lower baseline depression severity was consistently associated with higher odds of remission (aOR = 0.33, p < 0.001). Intermittent theta-burst stimulation was associated with lower odds of remission than high-frequency rTMS (aOR = 0.39, p < 0.001), although this association varied by outcome rater and protocol effects for response were non-significant. Country-level differences were observed, whereas medication-related and sociodemographic associations were inconsistent.

Treatment exposure and baseline severity were the most stable predictors of rTMS outcomes. Protocol and regional variations warrant standardized prospective studies to define personalized rTMS strategies.

Maladaptive eating behaviors in mood disorders are diverse and clinically relevant but have only been characterized into phenotypes based on theoretical evidence. The objective of the current study is to assess the prevalence of maladaptive eating behavior phenotypes (EBP) in mood disorders based on empirical evidence.

An online survey package including five questionnaires was administered to assess the prevalence and proportion of maladaptive eating behaviors in a mood disorder population (n = 100). Principal component analysis (PCA) and mixed measures analysis of variance (mANOVA) were conducted to characterize maladaptive EBPs.

Numerous maladaptive eating behaviors were identified in MDD and BD, including poor dietary quality, emotional eating, binge eating, and night eating behaviors. Differences between diagnostic groups are discussed. Three EBPs were characterized according to the PCA: (1) food impulsivity; (2) emotional eating; and (3) mindfulness in eating, with a significant influence of age and medication on the characterizations according to the mANOVA.

Maladaptive eating behaviors occur in mood disorders and can be characterized into distinct EBPS. These findings should be interpreted as preliminary and hypothesis-generating given the sample size, absence of a healthy comparator group, and potential confounds. As psychiatry moves toward a more comprehensive, personalized approach to diagnosis and treatment, EBPs require prospective validation in larger, controlled investigations to determine clinical relevance.

Major depressive disorder (MDD) is a complex disorder caused by genetic and environmental factors. Previous evidence implicates a potential depression subtype from interaction between childhood maltreatment (CM) and kynurenine pathway (KP) gene. Here, we investigated the top-down multi-omics KP alterations for this subtype.

A total of 142 patients with MDD receiving monotherapy were included. Individuals carrying IDO2 rs2340953 TT genotype and reporting CM were classified as IDO2-CM group, whereas the remaining patients were categorized as non-IDO2-CM group. Plasma KP metabolites, DNA methylation, KP-related protein, and inflammatory protein were profiled. Group differences were examined using regression models adjusting for relevant covariates.

Thirty patients met criteria for the IDO2-CM group. Compared with non-IDO2-CM, IDO2-CM individuals exhibited significantly poorer response rate to antidepressant at weeks 4 and 8. Metabolite analyses revealed a significantly elevated quinolinic acid (QA)/kynurenine (KYN) ratio and QA/kynurenic acid (KYNA) ratio in IDO2-CM group (β = 0.086, SE = 0.042, P = 0.023; β = 0.164, SE = 0.045, P = 0.0006). Epigenetic analyses of KP genes demonstrated coordinated methylation shifts favoring QA production (hypomethylation of KMO and AFMID; hypermethylation of ACMSD) and suppression of KYNA related enzymes (hypermethylation of CCBL1, CCBL2, AADAT, and GOT2). Proteomic profiling identified reduced circulating levels of KYNU and VCAM1 in IDO2-CM group.

CM interacting with IDO2 genotype defines a biologically distinct MDD subtype characterized by poor antidepressant response and convergent multi-omics evidence of KP bias toward neurotoxic metabolism. These findings support a stratification strategy and highlight the KP as a potential target for precision treatment in depression.

Population-based longitudinal evidence using administrative healthcare registries remains limited. We aimed to estimate the burden and temporal trends of major depressive disorder (MDD) among individuals aged ≥65 in Lombardy, Italy, 2013 to 2023.

We conducted a retrospective population-based longitudinal study using multi-source administrative healthcare registries (Lombardy region, Italy), including an open cohort of individuals aged ≥65 years. Incident and prevalent MDD cases were identified from 2013 to 2023, through an algorithm combining hospital discharge records, administrative exemptions, long-term care facility admissions and community psychiatric services. Age-standardised incidence rates per 100,000 person-years and annual prevalence (%) were estimated overall, by sex and age, using the age-truncated (≥65) WHO standard population. Annual percent changes (APCs) and 95% confidence intervals (CIs) were estimated by Poisson regressions.

We identified 19,851 incident MDD cases during the study period. Age-standardised incidence declined from 126.5/100,000 person-years (95% CI 121.7–131.3) in 2013 to 50.6 (47.6–53.7) in 2023, with an APC of 10.8% (−11.8 to −9.7). Age-standardised annual prevalence decreased from 0.94% (0.93–0.95) in 2013 to 0.63% (0.62–0.64) in 2023, corresponding to an annual decline of 3.0% (−4.7 to −1.3). Rates were higher among females. Burden and trends did not differ significantly by sex but across age groups, with the largest reductions among individuals aged 80 years or older.

Administrative data capture clinically diagnosed MDD within the healthcare system, rather than the broader burden in the community. Comparable epidemiological estimates are needed to interpret trends across different settings.

Age of onset is considered a key factor for the heterogeneity in major depressive disorder (MDD). Differential cortical thickness (CT) alteration patterns between youth-onset and adult-onset MDD have been evident, yet the neurobiological factors associated with such divergence remain unknown.

CT alteration patterns in youth-onset and adult-onset MDD were derived from the Enhancing NeuroImaging Genetics through Meta-Analysis. We interrogated whether the structural connectome constrains MDD-related CT alterations, followed by identifying likely epicentre regions and examining the hub vulnerability hypothesis. Using the neurotransmitter, transcriptome, and mitochondria atlases, we further explored the neurochemical basis, genetic architecture, and molecular energetic landscape related to the CT alterations.

The structural connectome constrained CT alterations in both youth-onset and adult-onset MDD, but the constraint effect was greater in adult-onset MDD. The epicentres were divergent for youth-onset (lateral prefrontal cortex) and adult-onset (orbitofrontal cortex, anterior and middle cingulate gyrus) MDD. The association between CT alterations and connectome degree centrality confirmed the hub vulnerability hypothesis in youth-onset MDD. Moreover, we found common and distinct neurotransmitter systems linked to CT alterations in youth-onset and adult-onset MDD. Transcriptomic contextualization demonstrated that CT alterations in adult-onset MDD were related to gene categories primarily involving synapse, neuron, synaptic transmission, and ion channel activity, yet those in youth-onset MDD were not related to gene expression. Additionally, associations between mitochondrial features and CT alterations were found in adult-onset MDD but not in its youth-onset counterpart.

Our findings may illuminate the neurobiological factors related to the spatial patterning of cortical abnormalities in youth-onset and adult-onset MDD.

Selective serotonin reuptake inhibitors (SSRIs) represent an effective treatment for major depressive disorder (MDD). This study employed glutamate chemical exchange saturation transfer (GluCEST) magnetic resonance imaging (MRI) to compare glutamate-sensitive signals in the prefrontal cortex and hippocampus of MDD patients before and after SSRI treatment.

The study included 35 unmedicated MDD patients and 39 healthy controls (HCs) who underwent baseline GluCEST imaging, using magnetization transfer ratio asymmetry (MTRasym) values to reflect regional glutamate levels. Twenty MDD patients completed follow-up GluCEST imaging after 8 weeks of SSRI treatment. Depressive symptoms were assessed using the Hamilton Rating Depression Scale (HAMD).

At baseline, the MTRasym values in the left hippocampus and bilateral prefrontal cortex were significantly lower in MDD patients compared to the HCs. Following SSRI treatment, the 20 MDD patients who completed follow-up showed significantly increased MTRasym values in the bilateral hippocampus and bilateral prefrontal cortex relative to baseline. Uncorrected correlation analyses showed that treatment-related differences in MTRasym values in the left hippocampus and left prefrontal cortex were associated with changes in HAMD-17 total score, sleep disturbance, and retardation scores.

GluCEST imaging suggested abnormal glutamate-sensitive signals in the prefrontal cortex and hippocampus of patients with MDD. SSRI treatment was associated with increased GluCEST signals across all regions of interest. The observed associations between regional GluCEST signal changes and symptom changes further suggest a potential role of glutamatergic-related modulation in the therapeutic effects of SSRIs.

Cognitive dysfunction is a core feature of major depressive disorder (MDD). Although both repetitive transcranial magnetic stimulation (rTMS) and vortioxetine have demonstrated procognitive effects, direct comparative evidence remains limited. This study compared the clinical and cognitive outcomes of rTMS and vortioxetine in patients with MDD.

In this 8-week naturalistic study, 40 patients with MDD received either rTMS (n = 25) or vortioxetine (n = 15). Depressive symptoms were assessed using the Montgomery–Åsberg Depression Rating Scale (MADRS). Cognitive performance was evaluated with the Perceived Deficits Questionnaire–Depression (PDQ-D), Trail Making Test Part B (TMT-B), Stroop Test, and Wisconsin Card Sorting Test (WCST). Longitudinal changes were analyzed using linear mixed-effects models. Week-8 outcomes were examined using ANCOVA adjusted for baseline cognitive performance and relevant clinical covariates.

Both groups showed significant improvements in depressive symptoms and cognitive performance over time, with no significant treatment × time interactions. Baseline cognitive performance strongly predicted week-8 outcomes. Domain-specific differences were observed in executive functioning: vortioxetine showed a modest advantage in WCST performance, whereas rTMS showed a non-significant trend toward better Stroop interference performance. No significant differences were detected in processing speed.

rTMS and vortioxetine were associated with comparable improvements in depressive symptoms and cognitive functioning in patients with MDD. Baseline cognitive status appeared to be a stronger predictor of outcome than treatment modality.

Peripartum depression is a common complication of pregnancy and the postpartum period, with adverse consequences for maternal and infant well-being. Perceived social support may be a modifiable protective factor, but South African data linking social context and depressive symptoms remain limited.

To estimate the proportion of peripartum women screening positive for clinically significant depressive symptoms; describe perceived social support and selected social determinants of health; examine associations between social support and screening outcome; and explore whether findings differed between ante- and postpartum participants.

This cross-sectional analytical study was conducted at Tshwane District Hospital, Gauteng, South Africa. Two hundred peripartum women (100 antenatal clinic, 100 postnatal ward) completed a sociodemographic questionnaire, the Edinburgh Postnatal Depression Scale (EPDS), and the Oslo Social Support Scale. Screening positivity was defined primarily as EPDS ≥13, with a sensitivity analysis using EPDS ≥11.

Using the primary EPDS threshold (≥13), 36.5% (n = 73) screened positive for clinically significant depressive symptoms and 48.0% (n = 96) screened positive in the sensitivity analysis (EPDS ≥11). Poor, moderate and strong perceived social support were reported by 29.5, 45.5 and 25.0% of participants, respectively. Lower perceived social support was significantly associated with screening positivity at EPDS ≥13 (p = 0.0062) and remained significant at EPDS ≥11 (p = 0.0008). At EPDS ≥13, lack of access to electricity (p = 0.0252) and a prior depressive disorder diagnosis (p = 0.0251) were also associated with screening positivity. In the sensitivity analysis (EPDS ≥11), reported discrimination/violence/abuse (p = 0.0077) and antepartum status (p = 0.0106) were additionally associated with screening positivity.

Lower perceived social support showed a robust association with screening positivity, supporting the value of integrating brief screening for depressive symptoms and support needs into routine antenatal and postnatal care, with clear referral pathways for psychosocial support.

Obesity is a well-established risk factor for major depressive disorder (MDD), yet the risk is not uniform, highlighting the need for precise risk stratification. This study aimed to develop a metabolomics-based prediction model to identify high-risk metabolic phenotypes among obese participants and to elucidate the causal metabolic pathways involved.

Forty-one-thousand-four-hundred-fifty-nine obese participants were followed for a median of 14.4 years. We integrated multiple machine learning (ML) algorithms to develop predictive models for 3-, 5-, and 9-year MDD risk, with a temporal validation within the same biobank. Furthermore, we investigated the potential causal relationships within the obesity-metabolite-MDD using mediation Mendelian randomization (MR).

During follow-up, 3,642 incident MDD cases were documented. The optimized LightGBM model demonstrated superior predictive performance, achieving AUCs of 0.844 (95% CI: 0.773–0.914), 0.824 (95% CI: 0.771–0.875), and 0.834 (95% CI: 0.796–0.871) for 3-, 5-, and 9-year intervals, respectively, significantly outperforming existing clinical models. Temporal validation confirmed the model’s robustness (AUCs: 0.738–0.776). MR analysis confirmed that key metabolites causally mediate the pathway from obesity to MDD (mediation proportions: −11.5% and −35.3%, all PME < 0.05).

These findings challenge the notion of a uniform obesity–MDD association, demonstrating that metabolomic signatures can effectively stratify MDD risk. We present a validated ML framework for the early identification of high-risk individuals with obesity, offering a precision medicine approach to guide targeted metabolic and psychiatric interventions.