Endocrine psychiatric comorbidity in metabolic disorders: Integrative mechanisms and traditional Chinese medicine

Abstract

Metabolic-endocrine disorders frequently co-occur with psychiatric and behavioural disturbances that are clinically consequential yet often under-recognised when care is organised around single-disease targets. In this narrative review, we synthesise evidence across major metabolic-endocrine conditions to propose a closed-loop framework in which immunometabolic stress (low-grade inflammation, insulin resistance, and metabolic toxicity) initiates or amplifies central symptom burden, while neuroendocrine axes and the gut-brain-metabolic interface act as bidirectional bridge modules that embed peripheral dysregulation into sleep disturbance, affective symptoms, appetite dysregulation, and cognitive inefficiency. These downstream changes are not merely comorbid “outcomes”; through impaired executive control and altered reward processing they erode behavioural feasibility, leading to inactivity, emotionally driven eating, reduced adherence, and disrupted follow-up, thereby reinforcing metabolic deterioration and sustaining relapse-prone trajectories. Within this systems view, traditional Chinese medicine offers a pragmatic, potentially complementary approach when positioned as adjunctive, module-targeted care rather than as a parallel diagnostic paradigm. We frame traditional Chinese medicine patterns as clinically useful phenotypic stratifiers that integrate symptom clusters relevant to sleep-stress regulation, eating behaviour, fatigue, and function, and we appraise clinical evidence using a matrix that prioritises dual-endpoint capture (metabolic and mental health/sleep/function outcomes), population relevance, and methodological transparency. Current evidence suggests potential benefit across modalities, including Chinese herbal medicine, acupuncture/electroacupuncture, and mind-body practices, but remains constrained by inconsistent dual-endpoint measurement, limited mediator capture, heterogeneous intervention protocols, and insufficient safety and interaction monitoring in polypharmacy-prone populations. We therefore propose an integrative care pathway that operationalises minimal screening, risk stratification, module-based intervention sequencing, and iterative dual-endpoint follow-up, alongside minimum methodological standards for future trials. Finally, we outline a precision and digital research agenda that leverages clinically feasible phenotyping and longitudinal monitoring of key mediators (sleep regularity, activity, autonomic markers, eating behaviour, and care persistence) to test mechanism, personalise sequencing, and strengthen real-world implementability.

Key Words: Metabolic-psychiatric comorbidity; Insulin resistance; Gut-brain-metabolic axis; Traditional Chinese medicine; Integrated care pathway

Core Tip: Endocrine-psychiatric comorbidity in metabolic disorders is conceptualised as a bidirectional, self-reinforcing state in which sleep and behavioural feasibility are clinically actionable leverage points that can shape metabolic trajectories. Traditional Chinese medicine is framed as adjunctive, module-targeted care, with pattern phenotyping used pragmatically for stratification rather than as a parallel diagnostic system. Credible integration requires paired metabolic and mental health endpoints, mediator tracking, and structured safety monitoring within an iterative care pathway.

INTRODUCTION

The growing emphasis on metabolic-endocrine disorders as conditions where biomedical risk and lived symptom burden co-evolve[1], but still, psychiatric, sleep, and behavioural manifestations of these disorders are commonly considered as secondary effects of metabolic control in both care and research[2,3]. In diabetes, obesity-related phenotypes, polycystic ovary syndrome (PCOS), thyroid dysfunction and the fatty liver disease spectrum, depressive and anxiety symptoms, insomnia, fatigue, appetite dysregulation, and cognitive inefficiency are common among patients and directly inhibits functioning and self-management sustainability[4,5]. These manifestations, even in the subthreshold form, occasionally recorded, or attributionally misascribed, can shape adherence, follow-up persistence, and implementation of day-to-day behaviour in a way that can affect metabolic trajectories in ways that may not be captured by the targets of single domain analysis. It is interesting to note that clinical improvements tend to be discordant across domains, as metabolic biomarkers become better and sleep or distress persists or symptomatic improvement is achieved without long-term metabolic stabilisation, highlighting the importance of considering reinforcing interactions over additive burden.

There is significant evidence of a correlation between metabolic disease and psychiatric and sleep outcomes, but there are gaps between explanatory models in different disciplines and usually focus on individual pathways or end states[6]. Mechanistic explanations tend to highlight immunometabolic stress and insulin resistance but do not adequately specify how upstream perturbation results in persistent, clinically actionable phenotypes (e.g., disrupted sleep timing, stress reactivity, eating behaviour and reduced behavioural feasibility) that mediate relapse-prone pathways. On the other hand, psychiatric models can describe the affective and sleep disruptions without incorporation of the metabolic environment that determines vulnerability, persistence of the symptoms as well as recovery[7]. Such a conceptual division limits the design of integrated interventions and leads to disparate trials where patient outcomes are measured in a single domain, and mediators of clinical importance, regularity of sleep, eating behaviour, and adherence, are measured inconsistently[8]. A more integrated account is therefore needed to link mechanistic modules to observable phenotypes and to support dual-endpoint evaluation in routine endocrine practice and translational research.

Demand for integrative care is rising, and many patients turn to traditional Chinese medicine (TCM) to manage symptom clusters spanning metabolic and psychological domains[9], yet interpretability is limited by inconsistent definitions of interventions, populations, and endpoints. TCM is also frequently positioned as a parallel paradigm, rather than being evaluated within a falsifiable framework that maps symptom phenotypes to modifiable mechanisms and measurable outcomes. Accordingly, a practical synthesis is needed that accommodates clinical heterogeneity, identifies high-leverage targets within reinforcing loops, and supplies a common language for stratification, monitoring, and iterative care.

Here, we integrate current evidence into a closed-loop framework in which peripheral immunometabolic load interacts with bridge modules (neuroendocrine axes and gut-brain-metabolic signalling) and is amplified through sleep/circadian disruption, reward and executive-control circuitry changes, and behavioural feasibility constraints[10]. Using this systems perspective, we appraise TCM as adjunctive, module-targeted care with clinical actionability and scientific caution. TCM patterns are treated as pragmatic phenotypic stratifiers aligned with observable domains, and interventions are judged by their capacity to improve paired metabolic and mental health (or sleep) outcomes while tracking mediators that plausibly sustain the loop. This review therefore links mechanistic models to a pathway-oriented management strategy. It aims to clarify why endocrine-psychiatric comorbidity persists, where the loop may be interrupted, and what minimum evidence is needed for safe and generalisable integration in routine practice.

REVIEW SCOPE AND METHODS

This narrative review synthesises evidence on endocrine-psychiatric comorbidity across major metabolic-endocrine disorders, with a focus on bidirectional mechanisms, clinically observable mediators, and the potential role of TCM within integrated care pathways. The literature search was conducted in PubMed/MEDLINE and EMBASE. The final search was performed on 31 December 2025, supplemented by a manual review of reference lists from the highly cited reviews, major society guidelines, and consensus statements.

The primary search window was 1 January 2015 to 31 December 2025 and priority was given to higher-level evidence published within the last five years (where possible) to capture modern concepts, terms, and models of care. Search terms combined controlled vocabulary and free text for three domains: (1) Metabolic-endocrine disorders (e.g., “type 2 diabetes”, “obesity”, “metabolic syndrome”, “PCOS”, “thyroid disease”, “MASLD/MAFLD”); (2) Psychiatric and behavioural phenotypes (e.g., “depression”, “anxiety”, “sleep”, “insomnia”, “cognitive impairment”, “adherence”, “self-management”); and (3) Mechanistic modules (e.g., “inflammation”, “insulin resistance”, “HPA axis”, “autonomic”, “gut microbiome”, “circadian”, “reward”, “executive function”), with an additional TCM component (e.g., “traditional Chinese medicine”, “herbal medicine”, “acupuncture”, “mind-body”). A representative search string for PubMed was: (metabolic OR endocrine OR diabetes OR obesity OR “metabolic syndrome” OR PCOS OR thyroid OR MASLD OR MAFLD) AND (depression OR anxiety OR insomnia OR sleep OR cognition OR “cognitive impairment” OR adherence OR “self management”) AND (inflammation OR “insulin resistance” OR HPA OR autonomic OR microbiome OR “gut brain” OR circadian) AND (“traditional Chinese medicine” OR acupuncture OR herb OR “Chinese herbal” OR mind-body). Equivalent adaptations were applied to EMBASE.

We included consensus statements and guidelines, population-based observational studies (cohort, case-control, and cross-sectional studies with clearly defined exposure/outcome measures), randomised or pragmatic trials, and mechanistic/translational studies relevant to the following modules: Inflammation-insulin resistance, neuroendocrine and autonomic axes, gut-brain-metabolic signalling, sleep/circadian regulation, and behavioural mediators. Studies were included if they: (1) Involved adults or adolescents with a defined metabolic-endocrine disorder; (2) Evaluated psychiatric symptoms/diagnoses, sleep/circadian outcomes, cognitive outcomes, or behavioural mediators (e.g., adherence, self-care behaviours); and/or (3) Provided mechanistic evidence plausibly linking metabolic and psychiatric phenotypes. We excluded non-peer-reviewed items (e.g., editorials without data), single-patient case reports (unless illustrating safety signals), and studies lacking sufficient methodological detail to interpret effect direction. Non-English reports were not systematically included due to feasibility constraints.

To enhance reproducibility, titles/abstracts were screened for relevance, followed by full-text review for inclusion. Evidence was then charted using a structured template capturing study design, population and diagnostic definitions, exposures/interventions, psychiatric/behavioural outcomes, key mechanistic measures [e.g., inflammatory biomarkers, hypothalamic-pituitary-adrenal (HPA)/autonomic readouts, microbiome signatures, sleep metrics], and principal findings. Given heterogeneity across conditions, interventions, and outcomes, formal meta-analysis was not planned; instead, we applied an explicit evidence hierarchy to guide synthesis: Guidelines/consensus and well-conducted systematic reviews were prioritised, followed by large population-based cohorts, then randomised/pragmatic trials, and finally mechanistic/translational studies. When findings were inconsistent, we highlighted potential sources of heterogeneity (e.g., phenotype definitions, confounding control, reverse causality, measurement differences) and avoided causal language unless supported by study design.

A narrative approach was chosen to integrate mechanistic modules with clinical phenotypes, identify methodological gaps (including measurement and temporality of mediators), and propose a pathway-oriented research and care framework (Figures 1 and 2). Throughout, statements with potential clinical implications were anchored to higher-level evidence where available; where evidence was emerging or indirect, we used appropriately cautious wording (e.g., “may suggest”, “could be consistent with”) and noted uncertainties.

Figure 1 Shared pathophysiology underlying endocrine-psychiatric comorbidity in metabolic disorders. Upstream metabolic, behavioral, and treatment-related triggers engage a self-reinforcing loop linking stress/circadian disruption, metabolic inflammation, insulin resistance/metabolic toxicity, gut barrier-microbiome dysbiosis, neuroimmune activation, and altered neural circuitry, generating depressive/anxious symptoms, sleep disturbance, cognitive dysfunction, disordered eating, and reduced adherence that further feed back to worsen metabolic control and cardiometabolic risk. Central hub targets (autonomic-inflammatory reflex, tryptophan-kynurenine pathway, mitochondrial dysfunction/oxidative stress, and glucagon-like peptide-1-related neuropeptides) highlight key leverage points for integrated interventions. BBB: Blood–brain barrier; GLP-1: Glucagon-like peptide-1; HPA: Hypothalamic–pituitary–adrenal axis; HRV: Heart rate variability; IR: Insulin resistance; LPS: Lipopolysaccharide; SCFAs: Short-chain fatty acids; SNS: Sympathetic nervous system.

Figure 2 Proposed integrative clinical pathway for endocrine-psychiatric comorbidity in metabolic-endocrine disorders. Patients undergo a minimal screening set (Patient Health Questionnaire 9, Generalized Anxiety Disorder 7, and Insomnia Severity Index/Pittsburgh Sleep Quality Index) followed by risk stratification to guide tiered, dual-endpoint intervention packages; traditional Chinese medicine pattern assessment and options are incorporated when appropriate with safety monitoring, and outcomes are reassessed using metabolic and mental health endpoints to iteratively adjust care intensity. PHQ-9: Patient Health Questionnaire 9; GAD-7: Generalized Anxiety Disorder 7; ISI: Insomnia Severity Index; PSQI: Pittsburgh Sleep Quality Index; T2DM: Type 2 diabetes mellitus; MAFLD: Metabolic dysfunction-associated fatty liver disease; PCOS: Polycystic ovary syndrome; TCM: Traditional Chinese medicine; HbA1C: Glycated hemoglobin; CRP: C-reactive protein; IL-6: Interleukin-6; HRV: Heart rate variability.

CLINICAL LANDSCAPE AND BIDIRECTIONAL FRAMEWORK OF ENDOCRINE-PSYCHIATRIC COMORBIDITY IN METABOLIC-ENDOCRINE DISORDERS

Definitions and diagnostic framing

Endocrine-psychiatric comorbidity in metabolic-endocrine disorders can be understood as the co-occurrence of metabolic dysregulation with psychiatric and behavioural phenotypes that may share upstream drivers and, in some patients, reciprocally shape disease course[11,12]. In this review, comorbidity is framed less as a count of discrete diagnoses and more as an interlinked clinical state in which symptoms, behaviours, and treatment exposures can plausibly influence metabolic control, functional capacity, and engagement with care (Figure 3)[13,14]. Such a framing is clinically pragmatic because it foregrounds observable phenotypes and modifiable determinants that are relevant to routine endocrine practice.

Figure 3 Conceptual framework of endocrine-psychiatric comorbidity in metabolic-endocrine disorders. Major metabolic-endocrine conditions are linked to psychiatric and behavioral phenotypes through bidirectional pathways. Metabolic disease may worsen mental health via metabolic inflammation, endocrine-autonomic dysregulation (hypothalamic-pituitary-adrenal/sympathetic nervous system), and gut-brain signaling, whereas psychiatric illness and its management may aggravate metabolic status through behavioral changes, chronic stress biology, and psychotropic metabolic liability. These overlapping patterns (disease-driven, behavior-driven, and treatment-emergent) contribute to poorer metabolic control, higher cardiometabolic risk, reduced adherence, and lower quality of life, underscoring unmet needs for integrated screening and dual-endpoint interventions targeting both metabolic and mental health outcomes. PCOS: Polycystic ovary syndrome; HPA: Hypothalamic-pituitary-adrenal; SNS: Sympathetic nervous system; OSA: Obstructive sleep apnea; MAFLD: Metabolic dysfunction-associated fatty liver disease.

To maintain a clear scope, we focus on five high-burden metabolic-endocrine conditions commonly encountered in clinical settings: Type 2 diabetes or prediabetes, obesity or metabolic syndrome, the metabolic dysfunction-associated fatty liver disease/metabolic dysfunction-associated steatotic liver disease spectrum[15], PCOS, and thyroid dysfunction. On the psychiatric side, depression and anxiety are central[16], yet an exclusive reliance on categorical psychiatric diagnoses risks missing clinically important presentations that frequently accompany metabolic illness. We therefore also consider sleep disturbance, cognitive dysfunction, and disordered eating, not as ancillary symptoms but as domains that can mediate everyday functioning and self-management, and may help explain why metabolic outcomes worsen even when biomedical treatment appears adequate[4,17].

A diagnostic framing that combines categorical disorders with dimensional symptom burden is consequently more informative than either approach alone. Brief, validated symptom measures can serve as a practical entry point to identify clinically meaningful distress or impairment and to trigger stepped assessment, particularly in endocrine clinics where specialist psychiatric evaluation is not always immediately available[18,19]. To improve clarity and avoid overgeneralisation, we use two complementary framings throughout this manuscript. “Categorical diagnoses” refer to DSM/ICD-defined disorders (for example, major depressive disorder or anxiety disorders), whereas “dimensional phenotypes” refer to symptom domains quantified on validated scales (for example, depressive symptoms, anxiety severity, insomnia burden, cognitive inefficiency, or eating dysregulation), which may be present with or without a formal diagnosis. When discussing prior studies, we indicate whether the evidence is drawn primarily from diagnosis-defined cohorts, symptom-defined measures, or mixed populations, and we use qualifiers (for example, “diagnosis-defined depression” or “depressive symptom burden”) when needed to make the framing explicit. This operational definition is aligned with our later emphasis on risk stratification and dual-endpoint management across metabolic and mental health domains (Figure 2), while recognising that causal direction is often heterogeneous and should not be presumed in individual patients.

Epidemiology and clinical burden

Evidence describing endocrine-psychiatric comorbidity in metabolic-endocrine disorders comes from population-based cohorts, real-world health records, registry analyses, and condition-focused evidence syntheses, which collectively point to a recurring clinical pattern despite differences in definitions and measurement[20-22]. Across settings, psychiatric symptoms and behavioural phenotypes are not confined to patients with formally coded mental disorders; rather, they often emerge as clinically salient dimensions of illness burden that shape trajectories of care and outcomes[23]. The differences in the screening tools, diagnostic cutoffs, and ascertainment can lead to heterogeneity across studies, but the overall message is that mental health burden is widespread and clinically meaningful in the metabolic care.

The phenomenon cuts across the key metabolic-endocrine conditions prioritised in this review, namely type 2 diabetes/prediabetes, obesity/metabolic syndrome, the metabolic dysfunction-associated fatty liver disease/metabolic dysfunction-associated steatotic liver disease spectrum, PCOS, and thyroid dysfunction, though the prevailing symptom clusters and their clinical consequences may vary depending on the condition (Table 1)[22]. Depressive and anxiety symptoms are repeatedly mentioned in the different categories of diseases[24] and sleep disturbance and disordered eating are often prominent in the obesity-related phenotypes[25], and cognitive complaints are often reported alongside diabetes and fatty liver disease[26]. Endocrine-related somatic symptoms and changing disease activity in PCOS and thyroid dysfunction can make it difficult to identify concurrent psychiatric morbidity and can increase the likelihood of fragmented, single-domain management[27-29].

Table 1 Endocrine-psychiatric and behavioral comorbidity patterns across major metabolic-endocrine disorders.

Index condition	Psychiatric and behavioral phenotypes (typical)	Evidence framing (psychiatric phenotype ascertainment)	Predominant comorbidity pattern1	Salient clinical burden (why it matters)
Type 2 diabetes/prediabetes	Depression; anxiety; sleep disturbance; cognitive dysfunction	Mixed	Bidirectional; disease-driven + behavior-driven[91]	Suboptimal metabolic control; reduced adherence/self-management; higher cardiometabolic risk (Figure 3)
Obesity/metabolic syndrome	Depression; anxiety; sleep disturbance (incl OSA-related); disordered eating	Mixed	Bidirectional; behavior-driven; treatment-emergent[92,93]	Weight trajectory reinforcement; impaired lifestyle persistence; increased cardiometabolic risk (Figure 3)
MAFLD/MASLD spectrum	Depression; anxiety; sleep disturbance; cognitive complaints	Mixed	Bidirectional; disease-driven + behavior-driven	Reduced quality of life; fatigue-related functional impairment; poorer lifestyle maintenance (Figure 3)
PCOS	Depression; anxiety; disordered eating; sleep disturbance	Mixed	Bidirectional; endocrine-metabolic-driven + psychosocial/behavior-driven[94]	Quality-of-life impairment; barriers to weight and symptom management; treatment non-persistence (Figure 3)
Thyroid dysfunction	Anxiety; depression; sleep disturbance; cognitive symptoms	Mixed	Primarily disease-driven; potential bidirectional feedback[95,96]	Functional impairment; symptom-related healthcare use; diagnostic misattribution risk (Figure 3)

¹Pattern terminology aligns with the bidirectional framework in (Figure 3): Disease-driven, behavior-driven, and treatment-emergent.

Evidence framing indicates whether the cited literature primarily defined psychiatric/behavioural phenotypes using DSM/ICD-based categorical diagnoses (“Diagnosis”), validated scale-based symptom thresholds or continuous scores (“Dimensional”), or a combination of both approaches (“Mixed”). OSA: Obstructive sleep apnea; MAFLD: Metabolic dysfunction-associated fatty liver disease; MASLD: Metabolic dysfunction-associated steatotic liver disease; PCOS: Polycystic ovary syndrome.

Clinically, the burden is best reflected in downstream effects, which are of interest to patients and health systems: Decreased self-management capacity, decreased medication and lifestyle adherence, and decreased longitudinal follow-up. Such aspects may be realistic triggers of suboptimal metabolic control and adverse cardiometabolic trajectories in the daily practice when pharmacotherapy is overall guideline-congruent. Simultaneously, patient-reported outcomes, such as fatigue, sleep quality, cognitive functioning, and overall quality of life, tend to deteriorate and reinforce disability and undermine sustained behaviour change[30,31]. Together, these observations argue for routine detection and integrated strategies that address metabolic and mental health outcomes concurrently, consistent with the bidirectional framework and unmet clinical needs summarised in Figure 3.

Bidirectional relationships and unmet clinical needs

Bidirectional coupling between metabolic-endocrine disease and psychiatric or behavioural phenotypes is increasingly evident in routine practice and across observational evidence, suggesting that comorbidity often reflects dynamic interaction rather than coincidental co-occurrence[32]. Metabolic disorders may contribute to depressed mood, anxiety, sleep disruption, and cognitive complaints through the cumulative burden of symptoms, treatment complexity, functional limitation, and uncertainty surrounding long-term risk. In the opposite direction, affective symptoms, insomnia, and maladaptive eating can reshape daily routines and decision-making in ways that directly impair self-management, including diet quality, physical activity, and medication-taking behaviours, thereby worsening metabolic control[33,34]. Rather than restating the broader closed-loop model here, we emphasise the clinically observable ways in which these influences present and become actionable in endocrine settings.

To make this heterogeneity clinically tractable, we propose three overlapping comorbidity patterns based on dominant drivers: Disease-driven, behavior-driven, and treatment-emergent presentations (Figure 3). Disease-driven presentations are characterised by endocrine and somatic symptom dynamics that coincide with changes in mood, sleep, or cognition, and they carry a high risk of diagnostic misattribution when evaluated within a single specialty[35]. Sleep deprivation, reduced activity, disordered eating, and loss of self-efficacy are the main characteristics of behavior-driven presentations and reduce adherence and increase metabolic risk[36]. A comorbidity which emerges during treatment is also prominent when psychotropic regimens have induced metabolic liability or due to conflicting priorities given to therapy resulting in inconsistent monitoring and fragmented decision making[37]. These patterns are not meant to be mutually exclusive but to be used as pragmatic descriptors, and offer a basis for deciding which mediators should be prioritised early.

Although these bidirectional links have a clinical significance, there are still some gaps in their implementation. Screening for distress, insomnia, and disordered eating is often inconsistent in endocrine clinics, even when subthreshold but functionally impairing symptoms are present[14]. Key mediators of care delivery, such as sleep, adherence, and psychotropic metabolic risk, are often overlooked because care is compartmentalised and focused on endocrinology, psychiatry, and behavioural or integrative services. Importantly, only a handful of interventions are formulated and tested against dual endpoints that jointly capture metabolic and mental health outcomes, thereby restricting actionable evidence for integrated management[7,38]. These gaps drive the systems-level synthesis in the following sections and constitute the leverage points that are summarised in the closed-loop framework (Figure 1).

SHARED PATHOPHYSIOLOGY: A SELF-REINFORCING METABOLIC-NEUROIMMUNE-BEHAVIORAL LOOP

Immunometabolic signaling: Inflammation, insulin resistance, and metabolic toxicity

Immunometabolic stress provides a clinically useful starting point for understanding why psychiatric and behavioural phenotypes frequently accompany metabolic-endocrine disease, particularly in presentations that appear predominantly disease-driven[39]. Low-grade inflammation is not merely a laboratory signature of metabolic dysregulation; it is often accompanied by a recognisable symptom constellation, fatigue, reduced motivation, sleep fragmentation, and cognitive inefficiency, that resembles a sickness-behaviour phenotype and can be observed across metabolic conditions[40]. This constellation may precede, mimic, or amplify depressive and anxiety presentations, making attribution difficult when symptoms are assessed within a single specialty. Peripheral immune signals could plausibly influence central affective and cognitive states via several communication routes, providing a coherent bridge between metabolic disease activity and symptom burden without presuming a uniform causal direction across patients. For pragmatic clinical phenotyping, readily available proxies such as C-reactive protein and the neutrophil-to-lymphocyte ratio can be interpreted alongside symptom and function measures to characterise an inflammatory load that may track with mental and behavioural vulnerability (Table 2).

Table 2 Mechanism modules linking metabolic-endocrine disorders with psychiatric and behavioral phenotypes: Candidate readouts and actionable leverage points (Figure 1).

Mechanism module (Figure 1)	Candidate readouts (clinical-first; research in italics)	Phenotypes most linked	Actionable leverage points
Chronic stress and circadian disruption (HPA/SNS)	Sleep timing/regularity; ISI/PSQI; resting HR/HRV; diurnal cortisol pattern; actigraphy/light exposure metrics[73]	Depression/anxiety; insomnia; fatigue[97]	Sleep-circadian repair; stress regulation; autonomic balancing[98]
Metabolic inflammation (immunometabolic signaling)	CRP; IL-6/TNF-α; NLR; visceral adiposity proxies (waist)[40]	Low mood/anergia; anxiety; cognitive complaints[44]	Weight reduction; anti-inflammatory lifestyle pattern; address inflammatory drivers[99]
Insulin resistance and metabolic toxicity	Glycated hemoglobin; fasting glucose/insulin; HOMA-IR; triglycerides/high-density lipoprotein; waist circumference[100]	Anhedonia/fatigue; cognitive slowing; reduced motivation	Improve insulin sensitivity (diet + activity ± pharmacotherapy); structured exercise; reduce metabolic toxicity[41]
Gut barrier and microbiome dysbiosis (gut-brain-metabolic axis)	GI symptom profile; diet quality (fibre/ultra-processed food exposure); stool microbiome/metabolites (SCFAs, indoles, bile-acid pathways)[101]	Depression/anxiety; sleep disturbance; disordered eating	Diet-based gut modulation; evidence-dependent pre/probiotics; vagal-mediated strategies[102]
Neuroimmune activation (microglia/BBB-related)	Peripheral inflammation proxies (CRP/NLR); symptom-triggered cognitive screening; neuroimaging/CSF markers (research)[39]	Cognitive dysfunction; mood symptoms; stress sensitivity[103]	Reduce upstream inflammatory/metabolic drivers; neuroprotective lifestyle; multimodal modulation
Neural circuitry and neurotransmission (reward/mood/cognition)	PHQ-9/GAD-7; sleep score; eating-behavior screening; adherence indicators	Disordered eating; depression/anxiety; impaired self-management	Psychotherapy/behavioral coaching; motivational support; reinforce adherence loops
Treatment-emergent metabolic liability (psychotropics)	Weight/waist trajectory; BP; glycated hemoglobin/glucose; lipids; adverse-effect checklist[104]	Weight gain; reduced adherence; metabolic deterioration[37]	Psychotropic stewardship; proactive metabolic monitoring; shared decision-making with risk mitigation[105]

HPA: Hypothalamic-pituitary-adrenal; SNS: Sympathetic nervous system; ISI: Insomnia Severity Index; PSQI: Pittsburgh Sleep Quality Index; HR: Heart rate; HRV: Heart rate variability; CRP: C-reactive protein; IL-6: Interleukin-6; TNF-α: Tumor necrosis factor-α; NLR: Neutrophil-to-lymphocyte ratio; HOMA-IR: Homeostasis Model Assessment of Insulin Resistance; GI: Gastrointestinal; SCFAs: Short-chain fatty acids; BBB: Blood-brain barrier; CSF: Cerebrospinal fluid; PHQ-9: Patient Health Questionnaire 9; GAD-7: Generalized Anxiety Disorder 7; BP: Blood pressure.

Inflammation commonly coexists with insulin resistance and metabolic toxicity, and the combination may be especially relevant to motivational and cognitive symptoms that undermine sustained self-management. Insulin resistance and gluco-/lipotoxic exposure may shift energy utilisation and stress responsivity in ways that are consistent with anergia, anhedonia, and slowed information processing, thereby linking metabolic state to behavioural capacity rather than to mood alone[41,42]. Clinically, even modest cognitive and executive inefficiency can translate into missed doses, reduced appointment adherence, and failure to sustain multi-step lifestyle plans, independent of motivation at baseline. Metabolic toxicity may also interact with sleep disturbance and appetite regulation, increasing susceptibility to inactivity and emotionally driven eating, which further reinforces adverse metabolic trajectories. Accordingly, routine metabolic readouts, glycated hemoglobin, fasting glucose/insulin indices, triglycerides-to-high-density lipoprotein patterns, and central adiposity measures, can be viewed not only as cardiometabolic risk markers but also as contextual indicators that shape symptom expression and behavioural feasibility (Table 2).

Within the systems framework proposed in (Figure 1), metabolic inflammation and insulin resistance/metabolic toxicity constitute a core peripheral engine that can couple to gut-brain signalling, neuroimmune activation, and neural circuitry changes. Clinically, this engine often manifests as functional decline, fatigue, poor sleep, low drive, and reduced cognitive bandwidth, rather than as a discrete psychiatric complaint, and it may therefore escape recognition when endocrine encounters focus narrowly on biochemical targets. These observations support an approach in which metabolic optimisation is paired with targeted behavioural scaffolding (notably sleep and activity supports) when functional capacity limits self-management, with concurrent monitoring on both metabolic and symptom/function domains (Table 2)[43,44]. This immunometabolic grounding sets the stage for the bridging roles of endocrine axes and the gut-brain-metabolic interface discussed next, and for identifying hub targets that may yield greater leverage than single-node interventions (Figure 1).

Endocrine axes and the gut-brain-metabolic axis

Neuroendocrine axes and gut-brain-metabolic communication provide the principal interfaces through which peripheral metabolic perturbations can be translated into central symptoms and behavioural change, and vice versa. The HPA axis and sympathetic activation are particularly relevant because they couple perceived stress to both affective state and metabolic regulation, offering a shared explanatory layer for anxiety-like arousal, insomnia, and concurrent metabolic instability[45,46]. Clinically, sustained hyperarousal often presents through sleep disturbance, fatigue, and reduced behavioural tolerance, which can constrain implementation of structured diet and activity plans. These effects are not purely psychological in their consequences: Disrupted sleep-wake timing and chronic stress may be associated with poorer metabolic control through shifts in daily routines, appetite regulation, and adherence behaviours. Practical phenotyping therefore benefits from combining symptom measures with readily obtainable physiological proxies, including validated sleep scales [e.g., Insomnia Severity Index (ISI)/Pittsburgh Sleep Quality Index (PSQI)] and autonomic readouts such as resting heart rate and heart rate variability, while recognising that diurnal cortisol patterns are more often confined to research settings (Table 2).

The gut-brain-metabolic axis offers a complementary bridge in which gastrointestinal physiology, immune signalling, and microbially derived metabolites intersect with endocrine and neural pathways that shape mood, sleep, and eating behaviour[47,48]. A clinically interpretable framework is to consider three interacting layers: Barrier function and mucosal integrity, immune activation and inflammatory mediators, and metabolite signalling that can influence both peripheral metabolism and central neurobehavioural states. Within this framework, dysbiosis is not invoked as an abstract concept but as a potential amplifier of symptom clusters that frequently co-occur in practice, gastrointestinal discomfort, sleep disturbance, appetite dysregulation, and fluctuations in mood or stress sensitivity. Although multi-omics profiling can characterise short-chain fatty acids, tryptophan-kynurenine metabolism, and bile-acid-related pathways, such assays remain largely research tools and are seldom required to recognise clinically meaningful gut-linked phenotypes. In routine care, more accessible indicators, dietary exposure patterns (e.g., low fibre intake or high ultra-processed food consumption) and structured assessment of gastrointestinal symptoms, can help contextualise behavioural risk and guide first-line interventions (Table 2).

In the closed-loop model summarised in Figure 1, stress/circadian disruption and gut barrier-microbiome dysregulation function as bridge modules that determine whether peripheral metabolic stress becomes “embedded” in sleep, affective symptoms, and eating behaviours, thereby enabling feedback back to metabolic outcomes. Importantly, these modules are visible to clinicians through concrete, modifiable levers, sleep timing regularity, stress load, dietary pattern, and gastrointestinal symptom management, rather than through specialised biomarkers alone[49]. Accordingly, module-targeted attention to these levers may be particularly relevant in stress-amplified or behaviour-driven presentations, and it provides a mechanistic bridge to the downstream neural circuitry and behavioural feedback mechanisms discussed next (Figure 1; Table 2)[50].

Neural circuitry, sleep/circadian rhythm, and behavioral feedback

Sleep and circadian rhythm disturbances occupy a pivotal position in metabolic-psychiatric comorbidity because they simultaneously shape affect regulation, stress responsivity, appetite control, and the capacity to sustain effortful self-care[51,52]. In metabolic-endocrine disorders, insomnia and circadian misalignment often present as more than symptomatic accompaniment; they can plausibly function as intermediates that convert physiological stress into day-to-day behavioural risk. Clinically, difficulty initiating or maintaining sleep can translate into daytime fatigue and reduced activity, while delayed sleep-wake timing may shift eating patterns towards late-night intake and higher energy density choices, thereby worsening metabolic parameters through routine-driven mechanisms. Fragmented sleep may further degrade attention and emotional stability, increasing reactivity to stress and lowering tolerance for structured behavioural change[53]. For practical assessment, validated sleep measures (e.g., ISI/PSQI) and indices of sleep regularity provide low-burden entry points, with actigraphy or wearable-derived rhythm metrics reserved for selected contexts where phenotyping or treatment tailoring is required (Table 2). Consistent with a modular view, sleep-circadian disruption may be one of the most efficient clinical “interfaces” through which immunometabolic stress and autonomic load become expressed as functional impairment and reduced behavioural persistence, even when biomedical treatment is otherwise appropriate. Notably, sleep disruption in metabolic-endocrine disorders is not uniform. In obesity and type 2 diabetes, sleep-disordered breathing, particularly obstructive sleep apnea, may contribute to metabolic deterioration through intermittent hypoxia and sympathetic surges, which has different mechanistic and treatment implications than stress-related hyperarousal insomnia characterised by heightened cognitive-affective arousal. Accordingly, when sleep complaints are prominent, clinicians may consider distinguishing suspected sleep-disordered breathing from insomnia phenotypes (for example, through targeted screening and referral for sleep evaluation when indicated), alongside behavioural strategies aimed at stabilising sleep timing and reducing hyperarousal.

Neural circuitry considerations add explanatory depth by linking mood symptoms to the behavioural processes that determine metabolic trajectories. Altered reward processing can present as reduced sensitivity to positive reinforcement and diminished capacity to derive benefit from incremental health gains, which undermines adherence to lifestyle programmes that rely on delayed rewards[54]. Under stress, shifts toward short-term decision-making may favour immediate relief behaviours, including emotionally driven eating, avoidance of physical activity, or inconsistent engagement with care, thereby embedding psychiatric symptoms into metabolic risk patterns. The areas of executive control and cognitive efficiency are most applicable when dealing with endocrine patients due to the complexities of treatment regimens that require planning, working memory, and sustained attention; even modest impairment is likely to result in missed doses, loss of follow-up, and failure to maintain multi-step management of their dietary plans[55]. Due to the hyper-arousal associated with anxiety, it may temporarily elevate the level of monitoring and vigilance but over time, this may lead to exhaustion, avoidance, and loss of engagement particularly where the symptoms seem to persist without any apparent relief. These pathways may be approximated in clinical research and clinical practice by using combined symptom scale [Patient Health Questionnaire 9 (PHQ-9)/Generalized Anxiety Disorder 7 (GAD-7)], structured eating behaviour screening and pragmatic adherence indicators, which work collectively to represent behavioural feasibility over psychiatric diagnosis (Table 2). Further breaking down to specific domains, various psychiatric and behavioural domains can be more inclined to align with certain upstream modules. An example is that chronic anxiety and an increased stress responsiveness are often theorised as a consequence of continued sympathetic-HPA activity, which might further increase glycaemic variability and reinforce avoidance or intermittent involvement in the care. In some patients, depressed mood and anhedonia may be indicative of a combination of an inflammatory-immunometabolic load and a shift in reward learning, which may contribute to less effort meaningfully directed to self-care and less reinforcement due to progressive health improvement. The cognitive slowing and executive ineffectiveness could be considered the emerging consequence of the sleep fragmentation, inflammatory load, and lack of prefrontal control, with downstream effects on medication management, planning, and problem-solving. Disordered eating and craving-driven intake might constitute a link amongst stress systems and reward–executive control balance, in which the dysregulation of affect and peripheral metabolic indicators collectively bias the behaviour towards instantaneous relief. Notably, the mappings are not absolute but probabilistic and can be dependent on conditions, stage, and treatment exposure but they offer an explanation of the mechanism behind the need to target the most salient domain(s) in a particular patient as opposed to assuming a homogenous psychiatric profile across metabolic disorders.

The feedback mechanisms of behaviour in turn make the system self-reinforcing and this explains why comorbidity is persistent and how it recurs among metabolic conditions. Worsening diet quality, sustained inactivity, ongoing sleep disruption, and reduced treatment persistence can otherwise solidify temporary states of symptoms into longer-term metabolic decline that further propagates inflammation and metabolic toxicity and is self-perpetuating with regard to mood, cognition, and sleep. This dynamic also clarifies why improvements may appear discordant, metabolic indices may improve while distress and insomnia remain, or mood may improve while weight and glycaemic control relapse, when key feedback loops have not been interrupted. In the closed-loop framework (Figure 1), sleep/circadian disruption, neural circuitry perturbations, and behavioural pathways constitute the amplification layer that couples upstream biological perturbations to downstream clinical outcomes and provides high-leverage targets for integrated intervention design. Accordingly, the domain-to-module perspective supports the stratification logic proposed later in the clinical pathway: Screening can identify the dominant symptom/behavioural domain(s), which then informs selection of module-targeted interventions and follow-up intensity, while maintaining concurrent attention to metabolic endpoints.

A TCM-INFORMED FRAMEWORK FOR METABOLIC–PSYCHIATRIC COMORBIDITY: PATTERN PHENOTYPING, EVIDENCE SYNTHESIS, AND MODULE-BASED MECHANISTIC MAPPING

TCM pattern perspective on metabolic-psychiatric comorbidity

In this review, TCM patterns are used as a pragmatic phenotyping framework to characterise symptom clusters that bridge metabolic dysregulation with psychiatric and behavioural burden, rather than as a parallel diagnostic system intended to replace contemporary disease classifications[56]. This positioning is clinically relevant because endocrine clinics frequently encounter patients whose distress, sleep disruption, fatigue, or maladaptive eating meaningfully impairs function and self-management even when formal psychiatric diagnoses are absent or unrecorded. By integrating somatic symptoms, functional capacity, and behavioural tendencies into coherent profiles, pattern-based assessment offers a structured way to describe heterogeneity that is otherwise flattened by single-disease labels. The value of such phenotyping lies in its capacity to generate testable, intervention-relevant hypotheses that can be evaluated against dual metabolic and mental health endpoints, consistent with the integrative approach advanced later (Figure 2). Operationally, we treat pattern information as a stratification layer that helps prioritise the most salient symptom/behavioural domain(s) for early intervention, while treatment selection remains anchored to clinical safety, feasibility, and paired metabolic-mental health outcomes.

To avoid conceptual drift, we interpret patterns through observable dimensions that map onto established mechanistic modules within the closed-loop framework (Figure 1)[57,58]. These dimensions include sleep and circadian regularity, stress reactivity and autonomic arousal, appetite regulation and eating behaviour, fatigue and activity tolerance, and cognitive efficiency relevant to planning and adherence. Viewed in this way, patterns can be aligned with bridge and amplification components of metabolic-psychiatric comorbidity, such as stress-circadian disruption, gut-brain-metabolic signalling, immunometabolic load, and the behavioural feedback processes that sustain symptoms and metabolic deterioration. Importantly, different constellations may imply different dominant drivers: A profile centred on insomnia, heightened arousal, and dysregulated routines may indicate a stress-circadian-amplified presentation, whereas prominent fatigue, low drive, and cognitive slowing may be more consistent with an immunometabolic-dominant phenotype. Such distinctions matter because they suggest where clinical leverage might be greatest, and they provide a rationale for sequencing interventions to restore behavioural feasibility before intensifying long-term metabolic targets[59]. For practical use, this mapping can be translated into module-targeted entry points. When sleep timing irregularity and persistent insomnia predominate, a sleep-circadian module may be prioritised (e.g., sleep regularisation and behavioural insomnia strategies), with TCM modalities considered only as adjuncts when feasible and tolerable. When hyperarousal, worry, and stress reactivity are prominent, a stress-autonomic module may be prioritised, emphasising skills-based stress regulation and arousal reduction alongside metabolic care. When low motivation, anhedonia, and behavioural inertia dominate, an executive/reward feasibility focus may be emphasised to support initiation and persistence of self-care, recognising that adjunctive approaches should be judged by functional response rather than pattern labels. When craving-driven intake or emotionally driven eating is prominent, an eating-behaviour interface can be prioritised, integrating appetite regulation and coping skills with clear monitoring of weight trajectory and glycaemic variability. Where gastrointestinal symptoms co-occur with mood and sleep disturbance, a gut-brain interface may be considered, while maintaining cautious interpretation given heterogeneity of mechanisms and evidence.

Pattern assessment can therefore be incorporated into integrative care as a stratification step that complements conventional metabolic risk evaluation, helping to prioritise interventions aimed at sleep, stress load, eating behaviour, and adherence barriers (Figure 2)[60]. To maintain interpretability, patterns should not be treated as exclusive diagnoses or as stand-alone efficacy endpoints; instead, they should be recorded alongside validated symptom measures and metabolic indices so that both domains can be monitored and iteratively managed. In keeping with the adjunctive, module-targeted framing adopted here, the operational question is not whether a pattern “explains” the condition, but whether pattern-informed prioritisation improves dual outcomes: Metabolic indices (e.g., glycaemia, weight, lipids) alongside symptom/function measures (e.g., PHQ-9/GAD-7 and sleep metrics) and feasibility indicators (adherence and follow-up persistence). The pattern categorisation alone should not drive escalation and referral but it should be guided by the severity of the symptoms, functional impairment, safety concerns, and non-response. In terms of the research, credibility is based on transparency and reproducibility such as clear pattern criteria, assessor training and uniform reporting of the symptom characteristics used for classification[60]. In this framing of operations, the evidence of TCM interventions could be evaluated in terms of the ability of studies to measure two clinically relevant dual outcomes, and the use of pattern-informed stratification that supports generalisability and translation to standard care.

Clinical evidence of TCM interventions

The clinical evidence supporting TCM interventions in the setting of metabolic-psychiatric comorbidity is best interpreted through a matrix approach that prioritises population relevance, endpoint structure, and methodological transparency, rather than through isolated claims of efficacy[61]. Because studies vary widely in inclusion criteria, intervention protocols, and outcome ascertainment, the most clinically meaningful question is not whether a given modality “works” in general, but whether it delivers measurable benefit across both metabolic and mental health domains in the same patient population. We therefore emphasise dual-endpoint capture, concurrent assessment of metabolic indices and validated measures of mood, sleep, or behavioural function, as a defining criterion for interpretability and translation (Table 3). In line with the closed-loop framework (Figure 1), symptom change is most informative when evaluated alongside measures that reflect behavioural feasibility and metabolic trajectory, rather than as a single-domain endpoint.

Table 3 Traditional Chinese medicine intervention evidence matrix for metabolic-psychiatric comorbidity, prioritising dual endpoints (Figures 1 and 2).

TCM modality	Target clinical context (metabolic-psychiatric focus)	Intervention use (typical delivery)	Outcomes commonly captured	Dual-endpoint capture1	Evidence maturity and key limitations
Chinese herbal medicine	Depressive symptoms in patients with cardiometabolic comorbidity; adjunctive care when tolerability/acceptability is a concern	Pattern-informed prescriptions; monotherapy or adjunct to antidepressants	Mental health: Depressive symptoms, anxiety, sleep-related complaints; metabolic: Often absent or secondary	Inconsistent (mental-health-dominant in many trials)	Evidence base includes syntheses and condition-specific reviews; limitations: Formulation heterogeneity, variable standardisation, limited integrated metabolic endpoints
Acupuncture/EA (PCOS-focused)	PCOS with negative emotions plus metabolic features (weight/insulin resistance concerns)	Protocol-based acupuncture/EA; often alongside lifestyle/usual care	Mental health: Depression/anxiety; metabolic/PCOS: Weight-related and PCOS clinical features commonly reported	More consistent (mood + metabolic/PCOS features often co-reported)	Recent synthesis suggests potential benefit for negative emotions; limitations: Trial quality and heterogeneity; psychiatric endpoints not always primary[106]
Acupuncture for insomnia (sleep as a mechanistic amplifier)	Insomnia (often coexisting with metabolic disorders), targeting the sleep-stress-metabolic loop	Manual acupuncture; protocol-driven courses; sometimes adjunct to CBT-I/usual care	Sleep: PSQI/ISI and daytime function; mental health: Mood may be secondary; metabolic: Seldom captured	Partial (sleep-centric; metabolic endpoints uncommon)	Systematic reviews support insomnia improvement; limitation: Integrated metabolic-psychiatric cohorts and dual-endpoint designs remain limited[107]
Mind-body therapies (Tai Chi/Qigong)	Adults at risk of MetS/obesity-related risk with psychological distress or reduced self-efficacy	Group-based, adherence-friendly programmes; combined physical + meditative components	Psychosocial: Depressive mood/anxiety, self-efficacy, quality of life; metabolic: Cardiometabolic risk indicators often assessed	Often yes (psychosocial + metabolic outcomes co-reported)	Systematic reviews in MetS-risk populations support multidomain benefits; limitations: Programme heterogeneity, comparator variability, adherence reporting[107]
EA for anxiety/depression (general psychiatric target with metabolic relevance via stress axis)	Anxiety/depressive symptoms where stress biology and autonomic dysregulation are prominent (metabolic impact often indirect)	EA protocols; adjunct to standard care	Mental health: Anxiety/depression; metabolic: Typically not collected	Rare (psychiatric endpoint-dominant)	Evidence from systematic reviews exists; limitation: Metabolic outcomes and psychotropic–metabolic stewardship endpoints are rarely incorporated
TCM-integrated lifestyle counselling (TCM-informed diet/exercise guidance)	Prediabetes/metabolic risk states where behavior change is central and psychological burden may affect adherence	Individualised counselling aligned with constitution/pattern concepts; pragmatic delivery	Metabolic: Lifestyle adherence and metabolic risk markers; mental health: Usually under-measured	Usually no	Evidence supports metabolic prevention approaches, but dual-endpoint capture and standardized mental-health assessment are uncommon (gap aligns with Figure 2)[108]

¹Dual-endpoint capture: Concurrent assessment of metabolic endpoints (e.g., glycaemic/adiposity/cardiometabolic risk measures) and mental health endpoints (e.g., validated mood/anxiety/sleep scales), consistent with the integrated pathway (Figure 2).

TCM: Traditional Chinese medicine; PCOS: Polycystic ovary syndrome; EA: Electroacupuncture; ISI: Insomnia Severity Index; PSQI: Pittsburgh Sleep Quality Index; ISI: Insomnia Severity Index; MetS: Metabolic syndrome.

Evidence for Chinese herbal medicine (CHM) is substantial in breadth, yet in comorbidity-relevant contexts it often remains mental health-dominant in its endpoint selection. Many CHM studies focus primarily on depressive symptoms, anxiety, or sleep-related complaints, whereas metabolic readouts are frequently absent, treated as secondary, or measured without sufficient follow-up to support trajectory-level inference (Table 3). This asymmetry limits causal interpretation within a systems framework, because it becomes difficult to determine whether symptomatic improvement corresponds to reduced immunometabolic load, improved self-management capacity, or durable metabolic stabilisation. Additional constraints arise from formulation heterogeneity and incomplete standardisation, including variability in constituent herbs, dosing schedules, and reporting of pattern-based indications, which together impede cross-study comparability. Safety monitoring and drug-herb interaction reporting are also inconsistently detailed in the comorbid population, where polypharmacy and cardiometabolic risk are common, further constraining immediate clinical generalisation[62]. Accordingly, CHM findings are most interpretable when studies pre-specify paired outcomes and report key mediators (e.g., sleep, fatigue, adherence) that plausibly link symptom change to metabolic control (Figure 2).

By contrast, acupuncture/electroacupuncture and mind-body interventions may be better positioned to operationalise multi-module effects in metabolic-psychiatric comorbidity, particularly where sleep disruption, stress arousal, and behavioural dysregulation are prominent. In contexts such as PCOS with negative emotions or insomnia as a comorbidity amplifier, acupuncture-based protocols have more often been evaluated with both symptom scales and metabolic or condition-relevant readouts, although limitations persist around sham control credibility, protocol heterogeneity, and the consistency with which psychiatric outcomes are treated as primary endpoints (Table 3)[63]. Mind-body exercise modalities, including Tai Chi and Qigong, offer an inherently integrative exposure, combining physical activity with attentional and breathing components, that maps naturally onto bridge and amplification modules in Figure 1, such as stress-circadian regulation and behavioural feedback. Consequently, these studies more commonly report co-occurring changes in psychosocial outcomes and metabolic risk markers, yet interpretation remains tempered by variability in programme dose, comparator selection, and adherence reporting, all of which influence external validity[64]. The translational appeal of these modalities lies in their feasibility as adjunctive interventions that can be delivered and iteratively adjusted in routine care, with outcomes monitored using accessible readouts (Table 2; Figure 2).

Rather than reiterating all structural limitations here, we summarise them and focus in this section on modality-specific interpretive considerations. Briefly, the main gaps remain inconsistent dual-endpoint capture, under-specified phenotyping/stratification, and limited follow-up to test whether symptom and behavioural improvements translate into durable metabolic trajectory change (Figure 1)[65]. These gaps provide a direct rationale for the integrated care pathway proposed later (Figure 2) and for prioritising methodological upgrades that make TCM trials in this domain more interpretable, reproducible, and clinically actionable[66].

Mechanistic insights into TCM: From metabolic modulation to neuroendocrine regulation

Mechanistic interpretation of TCM in metabolic-psychiatric comorbidity is most informative when organised around system modules that are clinically observable and potentially measurable, rather than around long lists of molecular pathways that are rarely assessed in patient-facing studies[67]. We therefore align mechanistic hypotheses with the closed-loop framework in (Figure 1), using readouts that can be captured in routine care or in translational research, metabolic indices, inflammatory proxies, sleep and symptom scales, autonomic measures, and gastrointestinal phenotyping, as anchors for interpretation (Table 2). This approach does not assume uniform causality across patients; instead, it offers a structured way to ask which modules are most likely to be influenced by different TCM modalities and how such changes might propagate through behavioural feasibility and metabolic trajectories. Framed in this way, the mechanism section serves a practical purpose: It identifies testable module-level hypotheses that can guide stratified intervention design and dual-endpoint evaluation within an integrated care pathway (Figure 2).

At the level of peripheral drivers, TCM interventions, particularly multi-component herbal prescriptions and lifestyle-aligned programmes, may contribute to metabolic stabilisation by modulating immunometabolic load and insulin resistance, thereby reducing the physiological context that supports fatigue, low motivation, and cognitive inefficiency. Such effects, where present, would be expected to manifest not only as changes in glycaemic or adiposity-related measures but also as improved functional capacity to engage in sustained self-management, which is often the rate-limiting step in comorbid care. Importantly, metabolic improvement should not be presumed to yield parallel mental health benefit, because symptom persistence can continue to impair adherence and sleep even when biochemical targets improve, and vice versa. For this reason, mechanistic claims are most credible when metabolic readouts (e.g., glycated hemoglobin, central adiposity proxies, lipid patterns) are assessed alongside inflammatory proxies (e.g., C-reactive protein, neutrophil-to-lymphocyte ratio) and paired with patient-centred outcomes capturing fatigue, mood, and behavioural function (Table 2). Within the loop model, reducing peripheral drive would be expected to lower pressure on downstream bridge and amplification modules, but whether this translates into durable trajectory change requires explicit testing in dual-endpoint designs (Figure 2)[68].

Bridge modules offer particularly plausible points of leverage for TCM modalities that combine physiological and behavioural inputs. Acupuncture and related techniques may influence stress reactivity, autonomic balance, and sleep initiation, thereby acting on the stress-circadian module and the autonomic-inflammatory interface that links perceived stress to metabolic and immune signals (Figure 1)[69,70]. Mind-body practices such as Tai Chi and Qigong provide a parallel route to bridge modulation by coupling low-to-moderate physical activity with attentional and breathing components that can regularise routines and potentially improve arousal regulation, which is directly relevant to insomnia-driven and behavior-driven phenotypes[71]. In addition, herbal and diet-oriented approaches may shape gut-linked symptom clusters through changes in dietary exposure, gastrointestinal function, and inflammatory tone, offering an interpretive bridge to the gut-brain-metabolic axis without relying on high-dimensional microbiome assays as a prerequisite. These hypotheses are most readily examined using accessible measures, sleep scales and rhythm regularity, autonomic readouts such as heart rate variability, structured assessment of gastrointestinal symptoms, and pragmatic dietary exposure indicators, rather than through mechanistic inference alone (Table 2). In clinical terms, bridge modulation is valuable because it may restore behavioural feasibility by improving sleep quality, reducing hyperarousal, and stabilising appetite regulation, thereby weakening the behavioural feedback that sustains metabolic deterioration.

The amplification layer of the loop provides a final mechanistic lens with direct translational relevance. If TCM interventions reduce insomnia, improve daytime energy, and support executive functioning, they may decrease emotionally driven eating, inactivity, and disengagement from follow-up, thereby attenuating the behavioural feedback that perpetuates comorbidity (Figure 1). This proposition is inherently testable, but only if trials and real-world programmes treat sleep, eating behaviour, and adherence as mediators to be measured rather than as incidental symptoms, and if they evaluate outcomes against paired metabolic and mental health endpoints over time (Figure 2). Consequently, the most defensible mechanistic narrative for TCM in this domain is not a claim of disease-specific molecular correction, but a module-level hypothesis: Multi-modal interventions may act across peripheral drivers, bridge processes, and behavioural amplification to shift the system toward a more stable trajectory. The next step is to embed these hypotheses within implementable care pathways and methodologically robust studies that can determine which phenotypes benefit, which modules change, and whether those changes translate into sustained improvements across both metabolic and psychiatric domains.

INTEGRATIVE MANAGEMENT MODELS, CHALLENGES, AND FUTURE DIRECTIONS

Integrative care models combining endocrinology, psychiatry, and TCM

An integrative care model for metabolic-psychiatric comorbidity should be designed as an executable pathway rather than as an aspirational multidisciplinary label, because the clinical problem is generated by reinforcing loops that are not contained within any single specialty (Figure 1)[13]. Fragmented care in general practice would usually result in discordant benefits, such as improvements in metabolic indices while insomnia, fatigue, and distress persist, or psychological symptoms improve while weight and glycaemic control relapse, which suggests that the relevant behavioural mediators and bridge modules were not addressed. The main objective of the integration is thus twofold: To optimise cardiometabolic risk and re-establish behavioural feasibility by means of targeted management of sleep, mood, eating habits, and adherence barriers. A pragmatic pathway accordingly prioritises four principles: Low-burden identification, risk stratification, module-based intervention selection, and dual-endpoint monitoring with iterative adjustment (Figure 2).

Pathway implementation can begin in endocrine settings using a minimal screening package that captures the domains most likely to sustain the closed loop and most amenable to early intervention. Symptom screening should include brief validated measures of depression and anxiety, paired with a standardised sleep assessment to detect insomnia and circadian disruption that often drives day-to-day dysfunction[72,73]. Because eating behaviour and behavioural execution are frequent mediators of metabolic outcomes, rapid screening for disordered eating tendencies and for adherence feasibility, such as missed doses, follow-up persistence, or self-reported capacity to implement behavioural change, should be integrated into the initial review. These clinical data are interpreted alongside readily available metabolic readouts (e.g., glycaemic status, adiposity proxies, and lipid patterns), which contextualise immunometabolic load and guide risk prioritisation (Table 2). To operationalise risk stratification and stepped care, we provide a pragmatic interpretation framework that links common screening ranges (PHQ-9, GAD-7, and ISI/PSQI) to risk tiers and suggested clinical actions (Table 4). Screening results should function as triggers for stepped care: Marked distress or functional impairment warrants more detailed assessment and co-management, prominent sleep disruption should prompt early circadian and insomnia-focused intervention, and concurrent psychotropic exposure or high metabolic risk should activate structured metabolic risk monitoring and medication review[74]. In this framework, thresholds are used as practical decision aids rather than definitive diagnoses; escalation may be warranted regardless of total score when functional impairment is substantial, symptoms deteriorate rapidly, or safety concerns are identified. Importantly, psychotropic exposure should be treated as a clinically meaningful context modifier rather than a background detail, because treatment effects can alter both metabolic risk and symptom burden, and may therefore influence interpretation of screening signals and follow-up priorities.

Table 4 Pragmatic interpretation framework for the minimal screening package and stepped-care triggers.

Measure	Score range	Risk tier (suggested)	Suggested stepped-care trigger in metabolic-endocrine clinics
PHQ-9	0-4	Low	Monitor; provide brief education on sleep regularity, activity planning, and stress hygiene; re-screen if symptoms persist or self-management becomes difficult
	5-9	Low-moderate	Consider brief, skills-based support and reassess; explore functional impact and adherence barriers
	10-14	Moderate	Consider structured, module-targeted intervention and closer follow-up; evaluate co-occurring anxiety/insomnia and treatment burden
	15-19	High	Prompt more comprehensive assessment (including function and safety) and consider referral/co-management with mental health services, while continuing metabolic care
	20-27	High	Urgent specialist evaluation should be considered when severe symptoms are present, particularly with marked functional impairment or safety concerns; coordinate integrated management and monitor metabolic effects of psychotropic treatment where applicable
GAD-7	0-4	Low	Monitor; provide brief stress-management and sleep regularity guidance
	5-9	Low-moderate	Consider short, skills-based interventions and reassess
	10-14	Moderate	Consider structured intervention and closer follow-up; assess hyperarousal and its interaction with sleep and glycaemic variability
	15-21	High	Prompt comprehensive assessment and consider referral/co-management; review medication effects and adherence barriers
ISI (preferred)	0-7	Low	Provide sleep regularity counselling; monitor
	8-14	Moderate	Consider brief behavioural insomnia management and reassess; evaluate common contributors (e.g., pain, nocturia)
	15-21	High	Prompt targeted insomnia intervention and closer follow-up; consider differentiating insomnia phenotype from sleep-disordered breathing when clinically indicated, and consider targeted screening for obstructive sleep apnea in higher-risk metabolic phenotypes1
	22-28	High	Consider specialist sleep/mental health referral when severe insomnia persists or functional impairment is substantial, and consider sleep evaluation when sleep-disordered breathing is suspected
PSQI (alternative)	≤ 5	Low	Generally indicates acceptable sleep quality; monitor and reinforce regularity
	> 5	Elevated	Suggests impaired sleep quality; consider further assessment (including insomnia severity and possible sleep-disordered breathing) and module-targeted sleep intervention

¹Cut-offs reflect commonly used screening conventions and should be interpreted in clinical context. Escalation may be warranted regardless of total score when functional impairment is substantial, symptoms deteriorate rapidly, or safety concerns (including suicidality) are identified. Where relevant, sleep complaints should be interpreted with attention to phenotype, as suspected sleep-disordered breathing may warrant targeted evaluation in parallel with insomnia-focused behavioural strategies.

PHQ-9: Patient Health Questionnaire 9; GAD-7: Generalized Anxiety Disorder 7; ISI: Insomnia Severity Index; PSQI: Pittsburgh Sleep Quality Index.

An additional iatrogenic layer should be considered in the closed-loop model, because treatment exposures can intersect with both psychiatric symptoms and metabolic control. Psychotropic medications used for depression, anxiety, or severe mental illness may contribute to weight gain, dysglycaemia, and lipid abnormalities in susceptible patients, and these metabolic changes may in turn worsen fatigue, sleep disruption, and distress, thereby amplifying behavioural feasibility constraints. Conversely, persistent distress or insomnia may prompt treatment escalation without consistent metabolic surveillance, creating a risk of treatment-emergent deterioration that is not captured if care remains compartmentalised. Accordingly, when psychotropic exposure is present, integrated pathways may incorporate structured baseline and follow-up monitoring (e.g., weight trajectory, glycaemic indices, and lipids) alongside symptom/function reassessment, with coordinated medication review and shared decision-making when metabolic deterioration, adherence failure, or functional decline emerges. This is not presented as a mandate for medication discontinuation, but as a rationale for aligning psychiatric stability with metabolic risk mitigation within the same follow-up loop.

Stratification then guides selection and sequencing of interventions organised by modifiable modules rather than by disciplinary silos (Figure 2). A sleep-circadian module is often a high-yield entry point because sleep disruption can amplify mood symptoms, appetite dysregulation, and executive dysfunction; first-line management may combine behavioural sleep interventions and routine regularisation, with adjunctive modalities considered when appropriate[75]. One stress-autonomic module targets hyperarousal and maladaptive stress responses that undermine behavioural capacity and may worsen metabolic control; this may involve psychological strategies and skills-based interventions and acupuncture or mind-body practices may be introduced as an adjunct in circumstances where they are feasible, acceptable, and compatible with the overall management plan of the patient. The eating-behaviour module is based on the regulation of appetite, emotional eating and practical diet application with structured coaching and when needed specialist psychological intervention; any TCM-based adjuncts should be chosen openly and monitored for tolerability, as opposed to being applied instead of standard behavioural support. Indeed, metabolic optimisation will continue to be the cornerstone of treatment, but its administration is reinforced by the alignment with the functional capacity of the patient and the intensification of treatments is supplemented by interventions that directly enhance adherence and day-to-day implementation[76]. In other words, adherence scaffolding (regimens made easier, follow-up, family and structured goal setting) needs to be an essential part of therapy instead of something added on afterwards, as it often decides the nature of trajectory change. Where there is a suspected presence of medication-related appetite change, sedation or fatigue, intervention sequencing might require prioritisation of behavioural feasibility and sleep stabilisation in combination with medication review as opposed to increasing metabolic pharmacotherapy independently.

Monitoring and iteration operationalises the pathway and differentiates integrative care from parallel, uncoordinated interventions. Follow-up should routinely capture paired endpoints, metabolic measurements and symptom/function indicators, focusing on the mediating role of sleep consistency, eating behaviour and adherence maintenance that would explain the presence or absence of outcomes changes (Table 2). There should be clear guidelines regarding the process of iterative adjustment: In case of persistent insomnia or distress, the development of sleep and psychological modules must increase before further intensifying metabolic pharmacotherapy, whereas in the case of metabolic non-response and the lack of adherence, the prominence of behavioural feasibility and support should gain pre-eminence. The safety surveillance needs to be inbuilt in every step, especially where there is the introduction of metabolic liability by the psychotropics or the adjunct administration of herbal medication in a polypharmacy patient, where structured adverse-event reporting and relevant monitoring of the laboratory are mandatory[77]. Here, psychotropic exposure should be recorded and its temporal relationship with appetite, sleep, weight, and glycaemic change may enhance interpretability and inform the timely decisions related to co-management. This paradigm-based model constitutes a direct connection to methodological reform not only because the same constructs that make integration clinically implementable, including dual endpoints, mediator tracking, stratification, and safety also characterize what future TCM-inclusive trials will need to measure to produce interpretable and actionable evidence.

Methodological limitations and how to improve TCM trials

The main methodological limitation of TCM trials in metabolic-psychiatric comorbidity is not the absence of positive findings but a repeated mismatch between study design and the systems nature of the clinical issue, limiting interpretability and translation[78]. Since comorbidity becomes maintained through interactions among biological, behavioural, and treatment-related processes, trials that define populations imprecisely, measure only single-domain outcomes, or omit key mediators cannot be used to test whether an intervention has had a meaningful impact on the reinforcing loop (Figure 2). To reduce this, a practical solution is to follow minimal standards that fit population definition, endpoint framework, capture of mediators, and safety surveillance in the closed-loop framework and in readouts that are clinically feasible (Table 2). Such limitations and solutions are thus presented as requirements to be undertaken as opposed to generic criticism of quality.

Definition of population and comorbidity: In most cases, the patients with a metabolic diagnosis are recruited but concurrent psychiatric or behavioural burden is not clearly established; thus, it is not clear whether the results are reflective of comorbidity or metabolic disease in isolation. Trials must identify comorbidity eligibility on reproducible criteria, such as metabolic disease definitions, validated indicators of both distress and insomnia or behavioural dysfunction, and must provide baseline severity and functional effects to facilitate risk stratification and external validity (Figure 2). This is necessary since the effect of treatment is likely to vary among the severity strata of symptoms and patterns of drivers that are dominant[79].

Dual endpoints as the standard analytic structure: A single-domain endpoint approach, such as metabolic markers alone or symptom scales alone, cannot determine whether an intervention disrupts the hypothesised pathway, that is, whether a change in symptoms results in a change in metabolic trajectory (Figure 1). Trials ought to pre-define therefore to use at least two endpoints balancing at least one metabolic outcome and at least one mental health, sleep, or functional outcome, and should be clear which endpoint is principal and which is a key secondary endpoint to avoid focus on any single outcome[80]. With this structure, interpretable, comparable endpoint profiles can be filled with evidence matrices instead of their disconnected findings (Table 3).

Methods to measure mechanism instead of assume it: Mechanistic accounts usually use the changes in sleep, eating behaviour or adherence, but these are often not measured, and in such a case outcomes cannot be explained by them. It should include a minimum of defined mediators (validated sleep measures and rhythm regularity indices), brief screen of disordered eating behaviour, activity measures and practical adherence or persistence of follow-up measures that reflect behavioural viability (Table 2). The quantification of these mediators enables trials to test whether improvement of symptoms relates to change of behaviour and whether behavioural change plausibly explains metabolic response, which is the core of a closed-loop explanation (Figure 2)[81].

Pattern/phenotype stratification employing reproducible rules: Pattern-based (TCM) stratification has been described often, but is poorly operationalised and so lacks reproducibility, preventing analyses of heterogeneity-of-treatment-effect. Pattern criteria defined in trials should be transparent, assessor training and reliability processes should be documented and the symptom and functional characteristics applied in classifications reported instead of relying on narrative labels. Parallel modern phenotypic descriptors should also be reported (sleep disturbance, stress arousal, appetite dysregulation, fatigue burden) to enable comparisons among studies and also to relate stratification to be connected with the mechanistic modules in (Figure 1). This strategy transforms pattern evaluation into a verifiable stratification instrument that has the ability to guide clinical sequencing in the pathway (Figure 2)[82].

Comparator choice and blinding aligned with the intervention type: In acupuncture and other mind-body therapies, insufficient or improperly described comparators and poor credibility of blinding may be introduced and result in bias and poor attribution of effects, but in CHM, heterogeneity in preparations and poor standardisation make it difficult to replicate and infer dose-effect relationships. Comparator strategies employed in explanatory trials must be practical and clearly explained, allocation concealment must be reported clearly, and when appropriate, blinding assessment must be done, and pragmatic trials should employ higher-quality usual care comparators and objective or trajectory-oriented outcomes to increase their interpretability on a real-world basis[83]. Intervention fidelity (dose, session frequency, adherence) should be assessed and reported across modalities; they have a significant impact on clinical effectiveness and generalisability (Table 3).

Follow-up horizon and time-structured outcome-capture: Many trials are too short to pick up change in metabolic trajectories, or they have time-dependent outcomes that overlook changes in symptoms early and maintenance or relapses over time. Research must embrace time-structured measurements that differentiate between early time windows (sleep, distress, behavioural feasibility) vs intermediate time windows (metabolic control and risk markers) and, where possible, have a maintenance stage where recurrence and persistence that are hallmark features of comorbidity can be assessed. It is also causally interpreted by pre-specifying these time points since changes in mediators can be associated with downstream outcome (Figure 2)[84].

Heterogeneity strategy and management: It has been known that comorbid populations are diverse and that average treatment effects may mask clinically relevant subgroup reactions especially in terms of baseline severity of symptoms, sleep disruption, eating behaviour patterns and psychotropic exposure. Stratification variables and interaction hypotheses should be pre-specified in the trials and sensitivity analyses that cover adherence, missing data and concomitant treatments should be included to prevent post hoc subgroup fishing but still allow the possibility of intuitive effect modification. In situations where possible, designs incorporating stratified randomisation or adaptive features are more consistent with pathway based care, where the level of treatment is adjusted iteratively (Figure 2)[77].

Safety, interactions and pharmacometabolic surveillance: Psychotropics can also lead to weight gain and dysglycaemia in comorbid care and CHM can be administered concomitantly with multi-drug regimens, so safety surveillance is not an add-on but a requirement of translation. Trials must institute systematic adverse-event monitoring, baseline and follow-up liver and renal-function monitoring where feasible and systematic surveillance of key indices of metabolic change, with explicit documentation of concomitant drugs and potential risk of interaction. Safety outcome reporting must be as meticulous as efficacy endpoints, which enhances clinical plausibility and enables its application in an integrated care pathway (Figure 2)[85].

All these minimum standards move TCM trials such that they are conducted to evaluate a pathway, rather than just isolated efficacy testing: Clearly defined comorbid populations, dual endpoints, capturing of any mediators, reproducible stratification, strong comparators, sufficient follow up, pre-specified analytic plans, and safety surveillance, based on clinically reasonable measures (Table 2). Applying these principles would also improve the completeness and interpretability of evidence summaries, enabling matrices such as (Table 3) to reflect not only what was studied, but how confidently findings can inform integrated, iterative care in real-world endocrine and psychiatric practice.

Future research agenda toward precision and digital TCM

Future progress in metabolic-psychiatric comorbidity will depend on moving from modality-centred debates to phenotype-centred strategies that specify who is most likely to benefit, through which modules, and with what measurable readouts[86]. A practical starting point is precision phenotyping anchored in the closed-loop model (Figure 1), in which patients are stratified by dominant drivers rather than by single diagnoses alone. Stress-circadian-dominant profiles, characterised by insomnia, hyperarousal, and irregular routines, may require early prioritisation of sleep and arousal regulation before metabolic targets can be realistically pursued. Immunometabolic-dominant profiles, marked by prominent fatigue, low drive, and cognitive inefficiency alongside adverse metabolic indices, may warrant a stronger emphasis on reducing metabolic load while simultaneously restoring behavioural feasibility. Behavior-driven profiles, typified by emotionally driven eating, inactivity, and poor follow-up persistence, highlight the need for adherence scaffolding and structured behavioural interventions as primary therapeutic levers. Treatment-emergent profiles, in which psychotropic exposure or other therapies introduce metabolic liability, require deliberate pharmacometabolic surveillance and coordinated medication optimisation. For each phenotype, a minimal, reproducible measurement set should be defined using accessible readouts, symptom scales for mood and anxiety, sleep measures and rhythm regularity, brief screening for eating behaviours, pragmatic adherence indicators, and core metabolic indices, so that stratification remains clinically deployable and comparable across studies (Table 2).

Digital monitoring should be positioned as an enabling method to capture the process variables that sustain the loop, rather than as an end in itself. The principal value of wearables and smartphone-based assessments is their ability to quantify mediators that are typically missing from trials, including sleep regularity, physical activity patterns, autonomic indices such as heart rate variability, and signals of care persistence such as appointment attendance and medication-taking behaviour (Table 2)[87]. These measurements directly support mechanism testing by allowing researchers to evaluate whether symptom improvements translate into behavioural change and whether behavioural change plausibly precedes metabolic trajectory shifts, which is the core causal chain implied by the closed-loop framework (Figure 1). In clinical pathways, the same data can operationalise iteration rules: Persistent sleep disruption can trigger escalation within the sleep-circadian module, poor adherence can prompt intensified support structures, and metabolic non-response despite improved process measures can justify careful treatment intensification (Figure 2). Digital tools can also facilitate standardisation of exposure in mind-body interventions by monitoring dose and adherence, improving the interpretability of pragmatic studies without imposing unrealistic laboratory requirements[88].

Evaluation strategies should then be aligned with the adaptive, module-based logic of integrated care rather than constrained to fixed, one-size-fits-all protocols. Because comorbid populations are heterogeneous and because response may depend on dominant drivers, trials that embed stratified randomisation or sequential decision rules are well suited to test pathway-relevant questions, including whether early sleep or stress regulation improves downstream metabolic outcomes in stress-amplified phenotypes (Figure 2)[89]. Pragmatic trials and embedded clinical studies can increase generalisability when they use reproducible intervention packages, a shared minimal dataset, and time-structured follow-up that captures early symptom change, intermediate metabolic trajectories, and longer-term maintenance or relapse[90]. Across designs, transparency in data governance, safety monitoring, and interpretability is essential, particularly where digital phenotyping is used, because implementation hinges on clinical trust and on the ability to translate signals into actionable care decisions. A coherent future agenda should therefore aim to identify which phenotypes benefit, which modules change, and whether those module shifts yield sustained improvement across paired metabolic and mental health endpoints within real-world pathways (Figures 1 and 2).

CONCLUSION

The clinical comorbidity between endocrine and psychiatric disorders in metabolic-endocrine disorders is best theorised as a bidirectional, self-reinforcing clinical state in which modifiable phenotypes (particularly sleep/circadian disruption and behavioural feasibility constraints) influence both symptom burden and metabolic trajectories (Figure 1). This framing facilitates pathway-based care that couples low-burden screening with module-specific sequencing with a series of follow-up in order to iteratively adjust care using concurrent metabolic and symptom/function outcomes (Figure 2; Table 2). In this set of paths, TCM is well justified in its adjunctive, module-directed care, where pattern-based phenotyping is used in a pragmatic way as a means of stratification and efficacy is determined according to paired endpoints and not based on the change in symptoms alone (Table 3). Major methodological weaknesses and minimum standards for credible integration are outlined; explicit phenotyping, dual-endpoint, mediator capture, and structured safety monitoring will have to be key priorities to enhance interpretability and translation. Lastly, the clinically achievable precision phenotyping and longitudinal follow-up, possibly using the digital measures of sleep and behaviour, can reinforce the real-world implementation and clarify which patients benefit from which module-based combinations.