Interoceptive aberrations in major depressive disorder: Neural mechanisms, clinical correlates, and targeted interventions

Abstract

Interoception, the sensing and interpretation of internal bodily signals, supports homeostatic regulation and emotional experience. Findings in major depressive disorder (MDD) remain inconsistent across studies using non-equivalent measures, heterogeneous clinical samples, and predominantly cross-sectional designs. Accordingly, this narrative review compares methods for assessing interoceptive accuracy, sensibility, and awareness; synthesizes behavioral, neurobiological, and clinical findings; and evaluates emerging interventions relevant to interoceptive processing. Current evidence indicates that interoceptive dysfunction in MDD is multidimensional, encompassing task performance, subjective bodily appraisal, metacognitive calibration, and interoception-related neural processing. Variability across these domains appears to reflect not only differences in measurement but also variation in clinical state and illness stage. Consistent with this view, neuroimaging studies suggest stage-related differences in large-scale networks, including lower default mode network activity in early-stage samples and higher activity in chronic-stage samples. From an active-inference perspective, selected findings may be understood in terms of an altered balance between bodily predictions and incoming sensory evidence, whereby negatively biased predictions persist when sensory signals are insufficiently weighted. Emerging intervention studies indicate that selected interoceptive processes are modifiable in some clinical and nonclinical samples, providing a rationale for mechanism-based testing in MDD.

Key Words: Interoception; Major depressive disorder; Active inference; Brain-body regulation; Somatic symptoms; Neural networks; Therapeutic interventions

Core Tip: This narrative review reframes the apparently inconsistent interoceptive findings in major depressive disorder (MDD) as a multidimensional pattern, with part of the apparent heterogeneity across studies related to measurement choice and illness stage. By comparing major assessment approaches and integrating behavioral, neurobiological, and clinical evidence, we identify where findings across interoceptive domains converge and diverge, with active inference serving as a heuristic perspective for interpreting selected links across levels of analysis. This synthesis points toward future stage-sensitive, multimodal assessment and more mechanism-informed intervention research in MDD.

INTRODUCTION

Interoception refers to the sensing, integration, and neural representation of visceral and somatic signals and contributes to homeostatic regulation, bodily awareness, and emotion[1]. Major depressive disorder (MDD) encompasses affective and cognitive symptoms, along with prominent neurovegetative and somatic disturbances, such as insomnia, appetite change, and fatigue[2,3]. The prominence of these bodily symptoms raises the possibility that altered processing of internal physiological states contributes to their expression in MDD.

Research in MDD distinguishes interoceptive accuracy (IAc), interoceptive sensibility (ISe), and interoceptive awareness (IAw), which are related but non-equivalent constructs[4,5]. Performance-based studies suggest reduced cardiac IAc in some samples[6], whereas self-report findings are more variable: Hospitalized patients may report lower bodily trust[7], and apparent reductions in ISe may partly reflect alexithymia[8]. From an active-inference perspective, these differences may reflect the persistence of negative bodily predictions when sensory input is unreliable or insufficiently weighted[3,9-11]. Complementary evidence from somatic disease-related depression implicates interoceptive neural circuits (INC) in pathways linking bodily dysfunction with depressive symptoms[12].

It remains unclear whether interoceptive alterations precede depression, emerge during illness, or vary across clinical states. Interpretation of the existing literature is complicated by the use of non-interchangeable measures of IAc, ISe, and IAw, incomplete integration of behavioral, neural, and clinical evidence, and the predominance of cross-sectional, small-sample, and single-center studies. This review therefore compares major approaches to interoceptive assessment, synthesizes behavioral, neurobiological, and clinical evidence, and draws on active inference as a heuristic perspective to interpret selected links across levels of analysis. It further examines possible stage-related network differences and interventions relevant to interoceptive processing, together with their potential implications for clinical assessment and symptom monitoring in MDD.

REVIEW METHODOLOGY

Review design

This narrative review synthesized evidence on interoceptive dysfunction in MDD, focusing on assessment methods, neurobiological mechanisms, clinical correlates, brain–body interactions, and therapeutic implications. The preparation of this review was informed by the quality criteria of the Scale for the Assessment of Narrative Review Articles[13].

Literature search strategy

PubMed and Web of Science were searched from database inception to May 31, 2026. Search terms covered interoception-related concepts, such as “interoception”, “interoceptive accuracy”, and “body awareness”, and MDD-related concepts, such as “major depressive disorder”, “MDD”, and “major depression”. Synonymous terms within each concept were combined with OR, and the two concept groups were combined with AND. The complete database-specific search strategies are provided in Supplementary Table 1.

The searches yielded 297 records (PubMed, n = 98; Web of Science, n = 199). After duplicate removal using EndNote and additional manual verification, 213 unique records remained for title and abstract relevance review, of which 103 were considered potentially relevant for further examination. Reference lists of relevant publications were also examined. Supplementary targeted searches and citation tracking were conducted during manuscript preparation and revision to identify additional methodological, conceptual, intervention-related, and recently published articles relevant to specific sections of the review. Articles identified through these supplementary searches were not included in the count of 297 records from the primary database search. Overall, 69 publications were included in the final narrative synthesis.

Literature selection and scope

Only English-language publications were considered. Empirical studies examining interoceptive function in individuals with MDD were prioritized. Reviews, theoretical articles, and studies involving related affective, psychiatric, or somatic populations were also considered when they directly informed the assessment, mechanisms, clinical interpretation, or treatment of interoceptive dysfunction in MDD.

Non-human studies, conference abstracts, corrigenda, and publications without substantive relevance to interoception or depression were excluded from the main synthesis. Titles and abstracts were reviewed for relevance, and potentially relevant publications were examined in full and retained when they directly informed the conceptual, methodological, mechanistic, or clinical questions addressed in the review. The overall process of literature identification, relevance-guided selection, and narrative synthesis is presented in Figure 1.

Figure 1 Flow diagram of literature identification, relevance-guided selection, and narrative thematic synthesis. MDD: Major depressive disorder.

Evidence synthesis

Given the heterogeneity of study designs, clinical populations, interoceptive constructs, and measurement approaches, the evidence was synthesized narratively. Findings were organized around assessment methods, neurobiological mechanisms, clinical correlates, brain-body interactions, and therapeutic strategies. Interpretation considered study design, sample characteristics, measurement validity, relevance to MDD, and consistency across studies. Accordingly, methodological strengths and limitations were appraised qualitatively rather than using a single formal risk-of-bias instrument across the full evidence base.

Supplementary Table 2 summarizes the study design or evidence type, sample characteristics, key findings or contributions, and principal limitations of each of the 69 publications included in the final narrative synthesis. Figure 2 provides an integrative schematic summary of the major assessment domains, interoception-related neural circuits, clinical phenotypes, brain–body mechanisms, and intervention targets discussed in this review.

Figure 2 Proposed integrative schematic model of interoceptive dysfunction in major depressive disorder. This schematic integrates the major assessment dimensions of interoception, related neural circuits and networks, clinical phenotypes, brain–body mechanisms, and intervention targets discussed in the review. It is intended as an integrative visual summary rather than an established causal model. MDD: Major depressive disorder; ACC: Anterior cingulate cortex; DMN: Default mode network; VAN: Ventral attention network; SMN-VN: Somatomotor-visual networks; IAc: Interoceptive accuracy; ISe: Interoceptive sensibility; IAw: Interoceptive awareness; OMT: Osteopathic manipulative treatment.

NEUROANATOMICAL AND NETWORK FOUNDATIONS OF INTEROCEPTION

Core neural architecture

Interoceptive processing depends on a distributed neural system involving brainstem and subcortical relays, together with cortical regions such as the insula, anterior cingulate cortex, somatosensory cortices, and prefrontal cortex[1,14]. These regions support the detection, integration, evaluation, and regulation of internal bodily signals.

Large-scale network alterations in MDD

Functional neuroimaging studies have identified altered activity and connectivity across several large-scale networks in MDD[2,15,16]. These include the somatomotor-visual networks (SMN-VN), ventral attention network (VAN), and default mode network (DMN), which collectively contribute to sensorimotor processing, visual perception, attentional and salience-related processing, and self-referential cognition. An umbrella review reported reduced intrinsic activity in the combined SMN-VN and increased activity in the VAN[2]. The insula and anterior cingulate cortex contribute to both interoceptive representation and salience processing, and abnormalities in these regions may affect the evaluation of bodily states[4,14]. From an active-inference perspective, greater VAN engagement can be interpreted as increased salience or estimated precision of selected internal signals, whereas reduced SMN-VN activity may diminish the influence of current sensory input on higher-order bodily predictions[2,9,17].

Stage-related network alterations in MDD

Cross-sectional evidence suggests stage-related reorganization of interoception-relevant networks in MDD. The clearest difference involves the DMN, with lower intrinsic activity in early-stage samples and higher activity in chronic-stage samples, whereas reduced SMN-VN activity and increased VAN activity appear more consistent across stages[2]. First-episode, treatment-naive MDD also shows increased anterior medial but decreased posterior medial DMN connectivity[18], supporting disrupted coordination between DMN subsystems.

These findings are consistent with a stage-sensitive pattern in which early MDD involves altered coordination among sensory, salience, and higher-order systems, whereas chronic MDD shows greater engagement of self-referential and ruminative processing[2,19]. Figure 3 summarizes these proposed stage-related network patterns and the corresponding research priorities. Longitudinal multimodal studies will be important for determining whether these stage-related patterns track persistence, recurrence, or treatment response and whether they can ultimately inform stage-sensitive clinical assessment.

Figure 3 Proposed stage-related model of interoception-related network dysfunction in major depressive disorder. This schematic summarizes a hypothesized dynamic pattern of interoception-related network alterations across clinical stages. Early-stage or first-episode major depressive disorder (MDD) may involve inefficient integration of bodily signals and altered salience processing, whereas chronic or recurrent MDD may be characterized by greater self-referential processing, disrupted network coordination, and persistent negative interpretations of bodily states. The model generates testable hypotheses for longitudinal and mechanistic studies. MDD: Major depressive disorder; ACC: Anterior cingulate cortex; DMN: Default mode network; SMN-VN: Somatomotor-visual networks; VAN: Ventral attention network.

Structural and neurochemical alterations

Cross-sectional studies also suggest stage-associated structural differences in MDD, with greater gray matter measures in the SMN-VN and DMN and relatively localized white matter abnormalities in early-stage samples, vs more widespread gray and white matter reductions in chronic-stage samples[2]. Longitudinal evidence in remitted MDD further links relapse vulnerability and residual symptoms to region-specific cortical variation, including the insula, anterior occipital sulcus, and precentral gyrus[20]. In currently depressed patients, greater symptom severity was associated with lower glutamate-to-creatine ratios in the pregenual anterior cingulate cortex (pgACC), with a similar trend for glutamate-glutamine ratios. Altered pgACC-anterior insula connectivity was also most evident in more severely depressed patients, supporting an association between regional glutamatergic alterations and network dysfunction[21]. Because the insula and anterior cingulate cortex are central to interoceptive and salience processing, these structural and neurochemical findings provide a plausible neural context for altered representation and evaluation of bodily states in MDD[14].

Active inference as a heuristic perspective on interoceptive dysfunction in MDD

Active inference provides a heuristic perspective for interpreting how altered bodily predictions and prediction-error weighting may contribute to MDD[9,10]. Bodily states are inferred by integrating prior expectations with incoming visceral and somatic signals, whose relative influence depends on their estimated precision[11,17]. When corrective sensory evidence is unreliable or insufficiently weighted, negatively biased expectations about bodily states may persist and contribute to negative affect and somatic symptoms[3,10].

This perspective helps connect findings across interoceptive domains and levels of analysis. Reduced IAc can be interpreted as reflecting weaker or less reliable sensory evidence, whereas lower bodily trust and negatively biased self-report may relate more closely to maladaptive expectations about bodily sensations; discordance between task performance and confidence may instead indicate altered metacognitive calibration. At the network level, greater VAN and insular engagement can be viewed as increasing the salience assigned to selected internal signals, whereas reduced SMN-VN activity may weaken sensory constraints on bodily predictions and DMN abnormalities may shape the self-referential context in which bodily sensations are interpreted[2,9,19]. Inflammatory and metabolic disturbances may further alter the physiological signals available for prediction updating, providing a potential bridge between peripheral bodily states and central interoceptive processing[10,22,23].

Viewed in this way, active inference does not imply a single mechanism underlying all interoceptive abnormalities in MDD, but offers a way to relate otherwise heterogeneous findings. It also generates testable predictions: More persistent depressive symptoms should be associated with stronger negative expectations about bodily states, weaker updating following controlled physiological feedback, and greater imbalance among sensory, salience, and self-referential networks[9,10]. Longitudinal and experimental studies that manipulate physiological feedback while measuring behavioral, autonomic, and neural responses could directly test these predictions.

INTEROCEPTIVE CONSTRUCTS AND ASSESSMENT TOOLS

Distinguishing IAc, sensibility, and awareness

IAc, ISe, and IAw represent distinct measurement domains. IAc refers to performance in detecting or discriminating bodily signals, ISe to self-reported attention to and appraisal of bodily sensations, and IAw to the correspondence between task performance and confidence[4,5]. The Multidimensional Assessment of Interoceptive Awareness (MAIA) and its revised version, MAIA, Version 2 (MAIA-2), are self-report instruments. Despite the term “awareness” in their names, they are treated here as measures of ISe rather than metacognitive IAw[4,5,24,25]. Table 1 compares the definitions and major assessment methods of IAc, ISe, and IAw, together with their key advantages and limitations and target populations.

Table 1 Comparison of major interoceptive dimensions and commonly used assessment tools.

Dimension (definition)	Tool/method	Measurement	Key advantages/limitations	Target populations	Ref.
IAc: Performance in detecting or discriminating bodily signals	HCT	Agreement between reported and recorded heartbeats	Simple and widely used; influenced by time estimation, heart-rate knowledge, physiology, instructions, and scoring	Clinical and nonclinical adults	[6,26,29-31]
IAc: Performance in detecting or discriminating bodily signals	Heartbeat discrimination tasks	Trial-based discrimination of cardiac signals	Less reliant on explicit counting; sensitive to attention, cardiac signal strength, and task design; limited standardization	Experimental and clinical samples	[5,28,33]
ISe: Self-reported attention to and appraisal of bodily sensations	MAIA/MAIA-2	Multidimensional bodily attention, appraisal, and regulation	Broad multidimensional assessment; reliability and factor structure vary; interpretation may be influenced by alexithymia and cultural context	General and clinical adults, including MDD	[24,25,34,40,68]
ISe: Self-reported attention to and appraisal of bodily sensations	BPQ-SF	Body awareness and perceived autonomic reactivity	Brief and easy to administer; may overlap with somatic or anxiety-related symptoms; does not assess objective accuracy	General, clinical, and cardiovascular samples	[35-37]
IAw: Correspondence between task performance and confidence	Performance-confidence correspondence	Calibration of performance and confidence	Assesses metacognitive insight; depends on task validity and confidence metrics	Studies combining interoceptive tasks with confidence ratings	[4,5,69]

BPQ-SF: Body Perception Questionnaire-Short Form; HCT: Heartbeat counting task; IAc: Interoceptive accuracy; IAw: Interoceptive awareness; ISe: Interoceptive sensibility; MAIA: Multidimensional Assessment of Interoceptive Awareness; MAIA-2: Multidimensional Assessment of Interoceptive Awareness, Version 2; MDD: Major depressive disorder.

Cardiac IAc in MDD: Conflicting evidence and heartbeat counting task limitations

Cardiac IAc is frequently assessed using the heartbeat counting task (HCT), in which reported heartbeat counts during fixed intervals are compared with recorded heartbeats[5,26]. Evidence from MDD studies, however, has been inconsistent. A systematic review of six studies found that reduced HCT performance was most apparent in moderately depressed samples, whereas some severely depressed samples did not show the expected impairment. Sex, comorbid anxiety or panic disorder, and selective serotonin reuptake inhibitor use were identified as possible confounders[6]. A later study of middle-aged and older patients found no group difference in HCT-derived IAc or heart rate variability despite altered insula-related functional connectivity[27]. More recently, a meta-analysis found only a small negative association between depressive symptoms and HCT performance and no significant association for heartbeat discrimination tasks; none of the examined moderators explained the HCT association[28]. Clinical differences in illness severity, age, sex, anxiety, medication exposure, cardiovascular health, and autonomic function may contribute to this heterogeneity but do not fully account for it[6,28].

Measurement-related factors provide an additional explanation for these inconsistent findings. The construct validity of the HCT remains disputed because performance may reflect processes beyond conscious perception of cardiac sensations. Under the original instructions, participants may rely on elapsed-time estimation or prior knowledge of their heart rate when estimating heartbeat counts. Scores decreased by approximately 50% when participants were instructed to report only heartbeats that they could feel[29]. Subsequent work further demonstrated contributions from time estimation and heart-rate knowledge and showed that relatively small changes in task wording could alter what the task measures[30].

The conventional scoring method introduces additional uncertainty. HCT scores predominantly reflect underreporting, correlate weakly with actual heartbeat counts, vary across intervals, and are structurally related to heart rate[31]. A reanalysis suggested that some of these problems arise from the ratio-based scoring method and the use of bivariate correlations[32]. However, HCT performance does not appear to be explained entirely by non-interoceptive factors. During orthostatic manipulation, heartbeat discrimination performance was the strongest and most stable predictor of HCT performance, whereas time-estimation accuracy was not an independent predictor[33]. The HCT may therefore retain an interoceptive component while remaining sensitive to physiological, cognitive, instructional, and analytical influences.

Given the methodological limitations of the HCT, conclusions regarding objective IAc in MDD should be interpreted cautiously. Current evidence is better understood as indicating altered cardiac interoceptive performance in some depressive populations rather than a uniform deficit in objective IAc across MDD. The HCT may remain informative when considered alongside complementary measures, but it should not be interpreted in isolation. Studies using the HCT should report task instructions and participant strategies, record heart rate and cardiovascular status, and account for medication exposure and anxiety symptoms.

Self-reported ISe: Strengths and limitations of the MAIA and Body Perception Questionnaire-Short Form

ISe is commonly assessed with the MAIA or MAIA-2, which cover eight dimensions of subjective bodily experience: Noticing, not-distracting, not-worrying, attention regulation, emotional awareness, self-regulation, body listening, and trusting[24,25]. Their multidimensional structure allows domain-specific patterns of bodily attention, appraisal, and regulation to be distinguished rather than reduced to a single overall score. This is particularly relevant in MDD, where alterations in subjective bodily experience may not occur uniformly across dimensions. MAIA-2 has also shown applicability and sensitivity to change in hospitalized patients with MDD, although some subscales showed suboptimal pretreatment reliability[25]. Subscale performance and interpretation may also vary across cultural settings[34].

The Body Perception Questionnaire-Short Form (BPQ-SF) provides a complementary self-report approach by assessing body awareness and perceived autonomic reactivity[35,36]. Its brevity supports use in clinical and large-sample research, and it has been adapted for cardiovascular clinical populations[37]. However, its subscales may capture overlapping aspects of bodily awareness, perceived autonomic reactivity, and psychopathology-related tendencies[36,37]. As with the MAIA, the BPQ-SF indexes subjective bodily experience rather than objective IAc.

Self-report nevertheless depends on an individual’s ability to identify, attend to, interpret, and describe internal experience. Questionnaire responses may therefore be shaped by attentional and interpretive tendencies in bodily appraisal[4], while cultural context may influence how bodily sensations are understood and reported[34]. Alexithymia is particularly important in MDD because difficulties identifying and describing emotions overlap conceptually with difficulties identifying and describing bodily states. Across studies, alexithymia has been associated with atypical interoception, although the strength and direction of this relationship vary by construct and measurement method[38-40]. A recent study found that reduced ISe was concentrated in depressed patients with alexithymia, whereas patients without alexithymia did not differ significantly from healthy controls; alexithymia also mediated the association between ISe and depressive symptoms[8].

These limitations constrain interpretation but do not negate the value of ISe measures. MAIA and BPQ-SF scores are best interpreted as indices of subjective bodily attention, appraisal, and regulation rather than proxies for objective IAc. In MDD, they may still reveal clinically relevant dimension-specific patterns that are not captured by performance tasks alone. Future studies should therefore account for alexithymia, cultural context, and related cognitive-affective influences when interpreting self-report findings and consider them alongside performance-based measures.

Toward multimodal and ecologically valid assessment

Given these limitations, future assessment should move beyond reliance on any single measure because different methods capture distinct components of interoceptive processing. In MDD, cortical processing of cardiac signals can be partly dissociated from peripheral autonomic dysfunction[41], further supporting the need to characterize interoception across complementary behavioral, subjective, physiological, and neural levels.

Heartbeat-evoked potentials (HEPs) provide temporally sensitive indices of cortical responses to cardiac afferent signals[14,42]. By capturing neural processing of cardiac information, HEPs complement perceptual tasks and can help distinguish alterations in central cardiac processing from changes in overt behavioral performance. Their interpretation nevertheless depends on cardiac physiology and methodological factors such as artifact correction, electrode selection, and analytical parameters[42].

Functional magnetic resonance spectroscopy (fMRS) extends this multimodal approach to the neurochemical level by quantifying metabolites such as glutamate-glutamine and gamma-aminobutyric acid within interoception-related regions. In healthy young adults, anterior insular glutamate-glutamine concentrations have been associated with error sensitivity[43], showing that fMRS can characterize neurochemical variation within a region central to interoceptive processing. Although evidence in MDD is still sparse, fMRS offers a means of linking regional neurochemistry with behavioral and neural measures of interoception.

Digital and wearable technologies provide a complementary route toward more ecologically valid assessment by extending measurement beyond the laboratory. Smartphone-based symptom reports can be temporally aligned with passive physiological or behavioral data collected from smartphones and wearable devices[44]. Smartwatch-based monitoring of physiological signals has also been incorporated into mindfulness training[45], providing a practical framework for pairing continuously recorded bodily signals with practices that direct attention toward ongoing bodily experience. When combined with momentary self-reports or interoceptive assessments, wearable physiological monitoring may help characterize how fluctuations in physiological states are perceived and interpreted and how they relate to symptoms in daily life.

Taken together, multimodal assessment can combine performance-based tasks, confidence ratings, multidimensional self-report, autonomic recordings, neurophysiological measures, neurochemical imaging, and digital monitoring to characterize complementary levels of interoceptive processing. Such designs may help distinguish alterations in physiological signaling from differences in perception, metacognitive calibration, and bodily appraisal. Interpretation should additionally consider medication exposure, cardiovascular status, age, anxiety, alexithymia, and cultural background, all of which may influence specific interoceptive measures.

INTEROCEPTION AND CLINICAL CORRELATES IN MDD

Emotional processing and alexithymia

Interoceptive abnormalities in MDD have been linked to altered emotional processing, particularly a bias toward negative information. An electroencephalography study found attenuated HEPs and altered late positive potential responses during emotional face processing in patients with MDD. Within the MDD group, prestimulus HEP amplitudes were associated with subsequent neural responses to sad faces[46].

Self-reported ISe also appears to vary with alexithymia status in MDD, suggesting that subjective interoceptive alterations are not uniform across patients and that alexithymia may contribute to heterogeneity in self-reported interoceptive experience[8].

Interoception, social support, and psychosocial functioning

Subjective interoceptive dimensions have also been examined in relation to perceived social support and depressive symptoms. In a cross-sectional study of patients with MDD, women reported lower perceived social support and more severe depressive symptoms than men. Sex-stratified analyses identified a partial indirect effect through noticing in men and partial indirect effects through self-regulation and trusting in women[47]. These sex-specific patterns may inform nursing assessment by drawing attention to perceived social support and individual differences in subjective interoceptive experience in MDD.

Another study examined perceived social support, interoception, and sociability in patients with MDD and healthy controls. Within the MDD group, body listening, emotional awareness, self-regulation, and trusting showed significant indirect effects in the association between perceived social support and sociability[48]. Lower MAIA-2 Trusting scores were also associated with suicidal ideation in an observational study of hospitalized patients with MDD[7].

Distinct interoceptive dimensions have been linked to negative emotional bias[46], alexithymia[8], perceived social support[47], sociability[48], and suicidal ideation[7], indicating that the clinical relevance of interoception in MDD depends on the specific domain being assessed.

INTEROCEPTION, COGNITIVE VULNERABILITY, AND SOMATIC SYMPTOMS IN MDD

Cognitive-behavioral vulnerability

Interoception may link bodily experience with attentional, evaluative, and behavioral-regulatory processes in depression[4]. In a cross-sectional structural equation modeling study of 308 patients with MDD, self-reported interoceptive function occupied a central position in a model that also included environmental, cognitive, behavioral, and emotional factors[49]. These findings suggest that altered interoceptive processing may be embedded within a broader cognitive-behavioral vulnerability profile, whereby the perception and interpretation of bodily states interact with maladaptive attentional and evaluative tendencies and patterns of behavioral regulation.

Somatic symptom burden and bodily interpretation

Somatic symptoms are common in depression and may be influenced by how bodily signals are detected, interpreted, and regulated. Indirect evidence from youths with bipolar disorder showed that greater somatic symptom burden was associated with more severe depressive and anxiety symptoms and lower scores on several MAIA dimensions. Not-distracting remained associated with somatic symptoms after adjustment for mood symptoms[50]. INCs have also been proposed as a potential route through which somatic diseases may increase vulnerability to comorbid depression[12].

Studies of chronic physical illness provide additional context. In chronic heart failure, MAIA-2-based self-report scores were associated with depressive symptoms, perceived social support, and symptom burden[51]. In chronic obstructive pulmonary disease, higher scores on a Body Perception Questionnaire-based measure of interoceptive sensitivity were associated with a lower odds of hospitalization for acute exacerbations, whereas anxiety sensitivity was associated with a higher odds[52].

Functional somatic presentations and related disorders

Functional neurological disorder (FND) provides a related clinical context in which bodily symptoms, psychiatric comorbidity, and neural alterations converge. Across FND subtypes, including functional cognitive disorder, functional movement disorder, and psychogenic non-epileptic seizures, psychiatric comorbidities such as depression, anxiety, post-traumatic stress disorder, and dissociative symptoms are common[53]. Greater physical symptom severity has also been associated with reduced white matter microstructural integrity[54]. Together, these findings illustrate the convergence of bodily symptom burden, affective disturbances, and neural alterations in disorders characterized by prominent physical symptoms.

Overall, the available evidence suggests that interoceptive processing is associated with cognitive-behavioral vulnerability and somatic symptom burden in MDD and related clinical conditions. Findings directly derived from MDD indicate that interoceptive function is embedded within broader environmental, cognitive, behavioral, and emotional processes[49], whereas studies of bipolar disorder[50], chronic physical illness[12,51,52], and FND[53,54] extend this brain–body perspective to related clinical contexts. Together, these findings support a multidimensional view involving bodily awareness, symptom interpretation, and physiological regulation rather than a single interoceptive deficit. Clinically, assessment of somatic complaints in MDD may therefore benefit from considering not only symptom presence and severity but also how patients perceive and interpret bodily sensations.

BRAIN-BODY MECHANISMS OF INTEROCEPTIVE DYSREGULATION IN MDD

Neural circuitry and brain-body regulation

Interoceptive dysregulation in MDD has been linked to changes in neural circuitry and in the wider inflammatory, autonomic, and metabolic systems that shape bodily signaling[1,2]. Structural and functional alterations in interoceptive circuitry form one component of this broader pattern. In remitted MDD, structural variation in the insula, sensorimotor cortex, and associative cortex has been associated with relapse vulnerability, residual symptoms, and decentering[20], suggesting that some cortical alterations may persist beyond acute depressive episodes.

Studies of somatic diseases with comorbid depression provide complementary evidence. The INC has been proposed as a pathway linking systemic somatic illness to depression[12]. Across ulcerative colitis, hypertension, chronic pain, and type 2 diabetes, shared morphological alterations were identified within the INC, including reduced volume of the left ventral diencephalon. Additional analyses implicated ventral diencephalic and thalamic regions in pathways associated with depression risk[12]. These findings extend candidate brain-body circuitry beyond the insula and sensorimotor cortex to subcortical structures.

Immune-inflammatory mechanisms

Immune dysregulation and chronic inflammation have been associated with neural circuit abnormalities and specific symptom dimensions in MDD[22]. An umbrella-review-based model links inflammatory processes to increased activity in the insula and VAN, reduced activity in the SMN-VN, and stage-related changes in the DMN[2]. In this model, early functional changes may reflect compensatory reorganization, whereas chronic illness involves more extensive structural abnormalities. From an interoceptive perspective, this network pattern could heighten the salience of selected internal signals, weaken sensory constraints on bodily-state estimates, and reinforce self-referential interpretations of bodily sensations[2,9,10].

Metabolic regulation

Energy-homeostasis-related genetic liability has been associated with depressive profiles characterized by increased appetite and hypersomnia[23], highlighting metabolic heterogeneity within MDD. Inflammatory and metabolic disturbances may influence the peripheral physiological signals available for interoceptive processing[22,23]. Peripheral autonomic dysfunction can also be partly dissociated from central cardiac interoceptive deficits in MDD[41], suggesting that peripheral physiological changes and central interoceptive processing do not necessarily vary in parallel.

Taken together, these findings suggest that alterations in the reliability of afferent physiological signals may influence higher-order interoceptive processing.

Computational integration

The inflammatory, autonomic, and metabolic processes described above can also be viewed from an active-inference perspective[10,17]. Changes in the reliability of afferent physiological signals may alter the precision assigned to sensory evidence and thereby affect the updating of bodily predictions.

As a theoretical extension of active inference, the Active Intersubjective Inference framework proposes that interoceptive predictions are reciprocally coupled with inferences about how one is perceived by others[55]. Applied to MDD, the model suggests that maladaptive interpersonal and self-related expectations may reinforce negatively biased bodily predictions and contribute to disturbances in self-representation and social-affective experience.

Overall, current models and available evidence are consistent with an interacting-systems account of interoceptive dysregulation in MDD. Neural circuitry, inflammatory and metabolic signaling, autonomic regulation, and predictive inference may interact across multiple levels of brain-body regulation rather than operate as isolated mechanisms.

POTENTIAL CLINICAL AND PROGNOSTIC IMPLICATIONS OF INTEROCEPTIVE MEASURES IN MDD

Candidate indicators of clinical outcomes

Interoceptive assessment may complement conventional clinical evaluation in MDD by characterizing how patients perceive, interpret, and regulate bodily signals[1,4]. These measures are most informative when considered alongside symptom severity, clinical history, psychosocial factors, and physiological measures.

Bodily trust has shown potential clinical relevance. In hospitalized patients with major depression, lower MAIA-2 trusting scores at admission predicted suicidal ideation at discharge after adjustment for baseline suicidal ideation, depression severity, and somatic comorbidity[7]. The Trusting subscale showed moderate discrimination, with an area under the curve of approximately 0.70[7]. Predictive models should also account for alexithymia and difficulties identifying emotional states, which are related to subjective bodily appraisal[8,38].

Interoception-related variables are also being incorporated into prediction models in medical settings. The PsyCor multicenter study is designed to develop and validate prognostic models of clinically significant depressive symptoms after coronary artery bypass grafting using demographic, clinical, psychometric, inflammatory, and interoception-related variables[56].

Prognostic and translational implications

Beyond their associations with current symptoms, interoception-related features may also be relevant to vulnerability and treatment-related change in MDD. In remitted depressed outpatients, residual symptoms and vulnerability to depression were associated with structural variation in the insula and precentral gyrus[20], suggesting that interoception-related circuitry may remain clinically relevant beyond the acute depressive state. Mindfulness studies further suggest that neural and self-reported interoceptive measures can change with training and relate to bodily trust or health behavior[57,58]. A key translational question is whether such measures provide prognostic information beyond established clinical and physiological indicators.

INTEROCEPTION-RELATED INTERVENTIONS AND THEIR RELEVANCE TO MDD

Mechanistic perspectives across interventions

Interventions potentially relevant to interoceptive processing act at different levels of the brain-body system. Cardiac-musical synchronization alters the temporal relationship between cardiac signals and external sensory feedback, whereas aerobic exercise modifies cardiorespiratory physiology[59,60]. Osteopathic manipulative treatment (OMT) and facial functional electrical stimulation (FES) engage somatic afferent pathways[61-63], mindfulness-based interventions (MBIs) target attention to and appraisal of bodily experience[57,58,64], and deep brain stimulation (DBS) modulates neural circuits involved in affective and interoceptive processing[65,66]. From an active-inference perspective, these approaches can be viewed as acting on different components of how bodily evidence is generated, weighted, and integrated with affective and cognitive states[9-11].

Peripheral afferent and somatic modulation

Evidence for peripheral and somatic approaches has largely come from healthy, nonclinical, or medically comorbid populations.

Cardiac-musical synchronization has been examined as a method of aligning cardiac activity with external sensory feedback. In healthy adults, an interactive intervention combining cardiac feedback, mindful attention, and music listening improved cardiac interoceptive performance. Heart rate and negative affect decreased, and heart rate variability increased across all three single-session conditions[60]. The intervention therefore altered the temporal correspondence between cardiac and auditory information, whereas the accompanying physiological and affective changes were not specific to the interactive condition.

Aerobic exercise provides another approach to modifying cardiorespiratory signals relevant to interoceptive processing. In physically inactive young adults, long-term aerobic exercise improved HCT-derived interoceptive performance and confidence and reduced self-reported depressive and anxiety symptoms[59]. Unlike cardiac-musical synchronization, which alters the temporal relationship between cardiac signals and external sensory feedback, exercise repeatedly modifies the physiological state from which cardiac signals arise.

OMT may influence interoceptive processing by altering somatic afferent input and subjective bodily experience. In a randomized sham-controlled functional magnetic resonance imaging study of adults with chronic low back pain, four weekly OMT sessions were associated with fewer errors during a heartbeat-tracking task and reduced blood oxygen level-dependent responses in the bilateral insula, anterior cingulate cortex, left striatum, and right middle frontal gyrus relative to sham treatment[61]. These findings suggest that OMT may influence heartbeat-tracking performance together with neural responses in interoception-related regions. A separate open-label randomized pilot study further reported increased body awareness together with reduced body dissociation, depressive symptoms, and stress following a standardized myofascial and lymphatic OMT protocol[62]. A registered sham-controlled trial in patients with chronic low back pain and comorbid depressive symptoms includes depression, pain, disability, heartbeat-tracking performance, MAIA-2 scores, and confidence-accuracy correspondence as outcomes[67]. The protocol extends this work by examining whether changes in depressive symptoms are accompanied by changes across behavioral, subjective, and metacognitive dimensions of interoception[68,69].

Facial FES targets somatic afferent feedback from facial muscles, providing a peripheral route through which bodily signals relevant to affective and interoceptive processing may be modulated. In an uncontrolled pilot study of 10 individuals with moderate-to-severe MDD, facial FES was feasible and well-tolerated and was accompanied by reductions in clinician-rated and self-reported depressive symptoms[63]. As interoception was not directly assessed, the extent to which these effects involve interoceptive mechanisms remains unclear.

Attentional and circuit-level modulation

MBIs engage interoceptive processing through sustained attention to bodily experience and changes in bodily appraisal. Following eight weeks of mindfulness training for primary care, adults with anxiety and/or depression showed increased left anterior insula responses during interoceptive attention. Among participants with moderate-to-severe depressive symptoms, greater increases in insular activity were associated with larger gains in body trusting[57]. In a larger primary-care study, interoceptive appreciation, operationalized using the MAIA body listening and trusting subscales, partially mediated the association between mindfulness training and the initiation of planned health behaviors, with the pattern differing according to baseline depression severity[58]. A controlled mindfulness teacher-training study also found a greater increase in self-reported awareness of mind–body integration in the training group than in the control group. Within the training group, increases in self-regulation were associated with reductions in depressive symptoms and overall emotional distress[64]. Across these studies, mindfulness-related changes were observed across subjective and neural dimensions of interoception, including bodily trust, self-regulation, and anterior insula responses.

DBS provides more direct evidence that modulation of neural circuits may alter interoceptive processing in treatment-resistant depression. In patients receiving subcallosal cingulate DBS, HEP amplitude increased over six months of treatment, and larger increases were associated with a shorter time to antidepressant response[66]. Complementary circuit-level evidence from ventral anterior limb of the internal capsule DBS showed long-term changes in limbic functional connectivity and insula-centered effective connectivity differences between active and sham stimulation[65].

Mindfulness and DBS provide evidence at subjective and neural levels, but through different routes: Mindfulness training has been associated with changes in bodily appraisal and anterior insula responses[57,58], whereas DBS has been associated with changes in HEPs and interoception-relevant neural connectivity[65,66]. Table 2 summarizes the evidence base, key findings, and evidence appraisal for each intervention.

Table 2 Evidence for interventions with potential relevance to interoceptive dysfunction in major depressive disorder.

Intervention	Evidence base	Key findings	Evidence appraisal	Ref.
Cardiac-musical synchronization	Single-session study in 24 healthy adults	Improved cardiac interoceptive performance; lower HR and negative affect; higher HRV	Small healthy sample; indirect relevance to MDD	[60]
Aerobic exercise	RCT in inactive young adults without MDD	Improved HCT performance and confidence; reduced depressive/anxiety symptoms	Nonclinical sample; HCT-based outcome	[59]
Osteopathic manipulative treatment	Pilot and sham-controlled studies; RCT protocol in chronic low back pain	Improved body awareness/heartbeat tracking; altered interoceptive-region activity	Direct interoceptive evidence outside MDD; depression evidence preliminary	[61,62,67]
Functional electrical stimulation	Uncontrolled pilot in 10 patients with MDD	Reduced depressive symptoms; interoception not assessed	Feasibility evidence in MDD; mechanism untested	[63]
Mindfulness-based interventions	Clinical and controlled studies in depression/anxiety or nonclinical samples	Changes in anterior insula activity and selected MAIA dimensions	Heterogeneous evidence; mainly neural/self-report outcomes	[57,58,64]
Deep brain stimulation	Small mechanistic/longitudinal studies in treatment-resistant MDD	Altered limbic/insula connectivity; HEP increase linked to faster response	Direct neural evidence; very small samples; no performance-based IAc	[65,66]

HCT: Heartbeat counting task; HEP: Heartbeat-evoked potential; HR: Heart rate; HRV: Heart rate variability; IAc: Interoceptive accuracy; MAIA: Multidimensional Assessment of Interoceptive Awareness; MDD: Major depressive disorder; RCT: Randomized controlled trial.

CONCLUSION

Current evidence supports a multidimensional conceptualization of interoceptive dysfunction in MDD, encompassing cardiac signal processing, subjective bodily appraisal, metacognitive calibration, and interoception-related neural processing. Differences in measurement and illness stage may help explain why findings have varied across studies, while active inference offers a heuristic perspective for relating selected findings across these levels of analysis[3,10]. However, the evidence base remains constrained by predominantly cross-sectional designs, relatively small and often single-center samples, and inconsistent control of medication exposure, cardiovascular status, and autonomic function[6,28]. It also remains unresolved whether interoceptive measures provide clinically meaningful information beyond established symptom and physiological measures. From a clinical perspective, interoceptive measures may complement conventional assessment by capturing aspects of bodily processing that are not fully represented by symptom severity alone.

Future research should move beyond cross-sectional associations toward designs that clarify the temporal and mechanistic significance of interoceptive alterations in MDD. Longitudinal studies following patients across acute episodes, remission, and relapse could determine whether lower accuracy on performance tasks, reduced bodily trust, and poorer metacognitive calibration precede symptom worsening and partially normalize with recovery. Such a pattern would support their potential roles as indicators of vulnerability or clinical state rather than merely concurrent correlates of symptom severity. Repeated assessments before and after treatment initiation, with careful characterization of medication exposure, cardiovascular status, and autonomic function, could further distinguish illness-related alterations from pharmacological and physiological influences. If these alterations primarily reflect illness processes, they should be detectable before pharmacotherapy and remain evident after relevant confounders are considered.

Developmental studies should examine whether the profile of interoceptive dysfunction varies across the lifespan, including whether the relative contributions of subjective, metacognitive, cardiac, and autonomic alterations differ across adolescence, adulthood, and later life. Multimodal studies integrating neuroimaging or electrophysiology with behavioral tasks, multidimensional self-report, inflammatory or metabolic markers, and wearable physiological monitoring should test whether combined measures identify clinically meaningful subgroups and predict symptom course or treatment response more effectively than any single modality. Cross-cultural studies should also evaluate the measurement invariance of the MAIA/MAIA-2, BPQ-SF, and performance-based paradigms after appropriate linguistic and cultural adaptation. These studies would help clarify which interoceptive dimensions are altered across different patients and stages of illness, and whether these alterations are clinically informative and modifiable.

ACKNOWLEDGEMENTS

The authors thank the Affiliated Mental Health Center of Jiangnan University for institutional support.