Interoception and Mental Health: A Roadmap

Abstract

Interoception refers to the process by which the nervous system senses, interprets, and integrates signals originating from within the body, providing a moment-by-moment mapping of the body's internal landscape across conscious and unconscious levels. Interoceptive signaling has been considered a component process of reflexes, urges, feelings, drives, adaptive responses, and cognitive and emotional experiences, highlighting its contributions to the maintenance of homeostatic functioning, body regulation, and survival. Dysfunction of interoception is increasingly recognized as an important component of different mental health conditions, including anxiety disorders, mood disorders, eating disorders, addictive disorders, and somatic symptom disorders. However, a number of conceptual and methodological challenges have made it difficult for interoceptive constructs to be broadly applied in mental health research and treatment settings. In November 2016, the Laureate Institute for Brain Research organized the first Interoception Summit, a gathering of interoception experts from around the world, with the goal of accelerating progress in understanding the role of interoception in mental health. The discussions at the meeting were organized around four themes: interoceptive assessment, interoceptive integration, interoceptive psychopathology, and the generation of a roadmap that could serve as a guide for future endeavors. This review article presents an overview of the emerging consensus generated by the meeting.

Keywords: Biomarker; Computational psychiatry; Interoception; Mental health; Research Domain Criteria; Treatment.

Figure 1

(A) Number of English language publications per year on interoception from PubMed, PsycINFO, and Institute for Science Information Web of Knowledge. The timeline starts in 1905, one year before the publication of Charles Sherrington’s book, The Integrative Action of the Nervous System, which first defined the concept of interoception. Key historical events relevant to interoception science are superimposed. (B) Publications per year on interoception vs. those investigating features of interoception that do not specifically refer to the term. These latter publications are more numerous and arise mainly from basic neuroscience, physiology, and subspecialty disciplines within the biomedical field. Note the use of a logarithmic scale in the second panel. [Figure reproduced and modified with permission from Khalsa and Lapidus (33).]

Figure 2

(A) Cardiac interoceptive processing measures and feature loading. This illustrates how the most commonly used heartbeat perception tasks differentially measure interoceptive features. [Figure reproduced and modified with permission from Khalsa and Lapidus (33).] (B) Multisystem interoceptive processing measures and feature loadings: Example of hypothetical approaches to measuring interoceptive processing across multiple systems. Approaches that perturb the state of the body are recommended, as are convergent approaches such as combined assessments of interoceptive attention and perturbation. (C) Central neural integration of interoceptive rhythms. Interoceptive rhythms vary considerably across the different systems of the body. They exhibit complex characteristics and are hierarchically integrated within discrete regions of the central nervous system. [Figure modified, with permission, from Petzschner et al. (37).] (D) Interoceptive rhythms vary in both amplitude and frequency. The top trace illustrates a hypothetical example of an amplitude modulation signal superimposed on a static frequency. The middle trace illustrates a hypothetical frequency modulation signal superimposed on a static amplitude. The bottom trace illustrates a hypothetical signal change involving both amplitude and frequency modulations that are temporally correlated. GI, gastrointestinal; GU, genitourinary.

Figure 3

(A) Example of one possible form of a general inference–control loop illustrated within a hierarchical Bayesian model. (B) Highly schematic example of illustrating that both interoceptive information and exteroceptive information are concurrently integrated to inform perceptual representations and action selection with respect to internally directed (e.g., visceromotor, autonomic) and externally directed (e.g., skeletomotor) actions. (C) General nodes that comprise a peripheral and central neural circuit for hierarchically integrating afferent interoceptive information into homeostatic reflexes, sensory and meta-cognitive representations, and allostatic regulators (predictions). ACC, anterior cingulate cortex; AIC, anterior insular cortex; MC, metacognitive layer; MIC, midinsular cortex; OFC, orbitofrontal cortex; PE, prediction error; PIC, posterior insular cortex; SGC, subgenual cortex. [Panels (A) and (B) reproduced, with permission, from Petzschner et al. (37). Panel (C) adapted, with permission, from Stephan et al. (36).]

Figure 4

(A) Active inference implementation according to the Embodied Predictive Interoception Coding model. Agranular visceromotor cortices, including the cingulate cortex, posterior ventral medial prefrontal cortex, posterior orbitofrontal cortex, and ventral anterior insula, estimate the balance among autonomic, metabolic, and immunological resources available to the body and its predicted requirements. These agranular visceromotor cortices issue allostatic predictions to hypothalamus, brainstem, and spinal cord nuclei to maintain a homeostatic internal milieu and simultaneously to the primary interoceptive sensory cortex in the mid and posterior insula. The interoceptive sensory cortex in the granular mid and posterior insula sends reciprocal prediction error signals back to the agranular visceromotor regions to modify the predictions. Under usual circumstances, these agranular regions are relatively insensitive to such feedback, which explains why interoceptive predictions are fairly stable in the face of body fluctuations. One hypothesis of the role of interoception in mental illness is that interoceptive input (i.e., posteriors) becomes increasingly decoupled from interoceptive predictions issued by the agranular visceromotor cortex (priors), leading to increased interoceptive prediction error signals. This decoupling may present in the brain as “noisy afferent interoceptive inputs” (97). (B) Proposed intracortical architecture and intercortical connectivity for interoceptive predictive coding. The granular cortex contains six cell layers including granule cells, which are excitatory neurons that amplify and distribute thalamocortical inputs throughout the column. The granular cortex is structurally similar to the neocortex and therefore more recently evolved than the agranular and dysgranular cortices. Within the insula, the granular cortex is present in the mid and posterior sectors. AC, anterior cingulate; PL, prelimbic cortex. [Figures reproduced, with permission, from Barrett and Simmons (45).]