It has long been thought that people diagnosed with anorexia nervosa (AN), and perhaps those with other eating disorders, have problems properly sensing and interpreting signals from within the body—what researchers call interoception.

Newly published research adds considerable depth to what it currently understood about disruptions in interoception in AN, specifically focusing on signals emanating from within the gastrointestinal (GI) system and their relationship to various behaviors. The research exploits a new technology to identify specific abnormalities in GI interoception, and uses several data types collected from patients combined with advanced mathematical (or “computational") modeling to predict individual relapse risk and symptom severity.

In AN, individuals restrict their food intake relative to the body’s energy requirements by eating far too little, exercising excessively, and/or purging food through laxatives and vomiting. Despite being severely underweight, individuals with AN typically do not recognize this body change and often have distorted body images. Anorexia is among the deadliest of psychiatric disorders and has a high rate of relapse—up to 50%—following successful weight-gain treatment.

A team led by Sahib S. Khalsa, M.D., Ph.D., a 2015 BBRF Young Investigator at UCLA, and including senior member Martin P. Paulus, M.D., a BBRF Scientific Council member and 2000 BBRF Young Investigator, conducted a study of 119 women, 62 of whom were diagnosed with the restrictive subtype of AN, in which individuals lose weight via restrictive eating and/or excessively exercising. (The contrasting AN subtype, not represented in this study, involves binge-eating and/or purging.) The 62 women or girls with restrictive AN were inpatients in the Eating Disorder Program, where they had received weight-restoration therapy, and the study was conducted at the Laureate Institute for Brain Research (LIBR) which is affiliated with the program (Dr. Paulus is LIBR’s Scientific Director and President). The other 57 participants in the study were demographically matched healthy comparisons.

The participants with AN, on average, were in their late teens or early 20s; had a BMI (body/mass index) of 18.5 or more (reflecting successful weight restoration during hospitalization); and had AN for about 4 years. 43% of these women had comorbid generalized anxiety disorder, 23% comorbid major depressive disorder, and 12% comorbid OCD.

To address the paucity of accurate data about GI interoception in AN, the team employed a novel device it had previously validated: an ingestible capsule that can vibrate at various intensities on command and be monitored externally by researchers. Delivering controlled “mechanosensory stimulation” to the stomach and adjacent segments of the gut, the capsule in healthy individuals evokes “non-aversive” sensations in the gut—they can be felt but do not cause pain or discomfort.

The brain’s response to capsule vibrations also generates a neural signature in those who ingest them, measurable via electroencephalogram (EEG), which registers electrical activity from large groups of neurons in several key wavelengths, ranging from very slow to very fast. In the context of the GI system, capsule vibrations as sensed by the body generate data about what researchers call gastric-evoked potentials (GEPs) that correlate with how accurately an individual perceives the signals introduced into the GI tract by the capsule.

The team combined this data from each study subject with a computational-based model of how the body makes inferences based on interoceptive signals. This is based on internal computations that can’t be observed directly but can be inferred from an individual’s behavior. This yields insights about “prior beliefs” (expectations about bodily states before new sensory evidence is noted); and “interoceptive precision,” which registers the degree to which incoming visceral signals are heeded when updating initial expectations. The working assumption, aided by the computational modeling, is that such interpretations of signals generated within the body are in some way distorted or inaccurate in people with AN and perhaps other eating disorders. This would affect perceptions of hunger and fullness (“satiety”), and thus the desire to eat or not eat at a given moment.

The value of these data is explained in an example provided by the researchers in their paper, which appeared in JAMA Psychiatry: An individual with AN who expects that eating will produce unpleasant fullness (a strong “prior belief”) may interpret new but mild gastric sensory signals as strongly aversive if those signals are not assigned proper significance relative to the original expectation. In contrast, higher interoceptive precision—like that in healthy people—“would allow incoming sensory evidence to update or override” a faulty prior expectation.

Each study participant sat for a single experimental session involving ingestion of the vibrating capsule. Stimulation began 3 minutes after swallowing, and was divided in two “blocks,” each of which involved about 60 stimulations each lasting 3 seconds. In one block, the stimulation was “normal,” i.e., similar to routine stimulation of the GI tract, and in the other it was “enhanced.” The participants were randomized to receive these blocks with “normal” stimulation occurring either before or after “enhanced” stimulation.

The tests revealed “numerous disruptions” at the behavioral and computational levels in study participants with AN, compared with healthy controls. The women with AN “showed reduced accuracy in detecting normal-intensity gut sensations,” and computational models based on the data “revealed that those with AN held stronger prior beliefs” that capsule-induced gut vibrations would not be present. The AN participants also failed to adapt, i.e., learn to behave differently, based on new signals that were, nonetheless, registered in their gut when the capsules were vibrating.

The heightened expectation that gut stimulation will either not occur or not be perceived aligns with a view of AN that proposes that abnormally high confidence in prior expectations can “override” conflicting sensory input coming directly from the gut. This may reduce normal “perceptual updating” and “contribute to persistently distorted bodily perception” in AN and perhaps other eating disorders, the team said.

They further wrote: “Misinterpreted or blunted satiety [fullness] cues may also apply to hunger and thirst signals, facilitating the ability of individuals with AN to ignore these bodily cues.” Supporting this was data from the study indicating that women with AN updated their beliefs more in response to the absence of gut sensations than their presence. This, the team suggests, explains data showing increasing expectation among the AN participants that vibrations from the capsule would not occur, even when they did. “This dynamic parallels known cognitive inflexibility in AN,” the researchers said, “and reflects a mechanistic link between perceptual inference and executive dysfunction.”

Future studies will seek to investigate whether this cognitive rigidity reflects interoceptive-specific learning “or general impairments in inhibitory control.” For now, however, data from this study importantly indicates that disruptions in inferences based on interoceptive signals persist even after weight restoration—as all the girls and women with AN in the study had already recovered much of the weight they had lost at the time they ingested the vibrating capsules. This insight, in turn, provides support for “recent clarion calls to move beyond nutritional rehabilitation alone” when developing improved treatments, the researchers wrote. Biomarkers of interoceptive dysfunction could inform development or augmentation of novel brain-gut therapies in AN by focusing on recalibrating maladaptive prior expectations and promoting flexible learning in patients, they said.

Importantly, the current study revealed potential biomarkers, such as interoceptive prior beliefs and stomach/digestive unpleasantness experienced during the capsule-based task, which predicted clinical outcome at 6 months, i.e., whether an individual would relapse. Another interesting signal was a larger increase in hunger following stimulation by the ingested capsule among those with AN. This suggests to the team that “suppressed hunger signaling can be re-engaged through vibratory gut stimulation.” In other words, controlled, capsule-based stimulation could potentially help to restore healthy interoceptive function “and reconnect patients with internal hunger cues” that they are either suppressing or failing to perceive.

Thus, the findings of the study open the way not only to the development of biomarkers for assessing relapse risk, but also of new ways beyond “nutritional rehabilitation” of treating those suffering with the symptoms and relapse risk of anorexia nervosa.