Process-Based Therapy for ARFID

A nine-year-old eats only seven foods. A twenty-six-year-old has not tried a new food since high school. A forty-year-old woman loses fifteen pounds after a choking incident at a restaurant and cannot bring herself to eat anything with a texture that reminds her of the food that lodged in her throat. All three carry the same diagnosis: Avoidant/Restrictive Food Intake Disorder. All three arrived at that diagnosis through entirely different psychological pathways, and the processes maintaining food avoidance in each case share almost nothing in common.

This is the central problem with ARFID treatment, and it is the reason that applying a single protocol to all ARFID presentations produces inconsistent results. ARFID is a diagnostic category, not a mechanism. Process-based therapy starts from the mechanism.

Three Presentations, Three Process Networks

The DSM-5 defines ARFID through three recognized presentations, though in clinical practice these frequently overlap.

Sensory sensitivity is the most common presentation in children and often persists into adulthood. The person experiences certain food textures, temperatures, colors, smells, or tastes as intensely aversive, not as a preference but as a near-involuntary rejection. The maintaining processes typically include heightened interoceptive sensitivity (the body’s internal signals about food are amplified and experienced as threatening), low distress tolerance for sensory discomfort, rigid categorization of foods into safe and unsafe categories, and anxiety-driven avoidance that narrows the food repertoire over time as previously tolerated foods become associated with distress.

Low appetite or lack of interest in eating presents differently. These individuals do not find food aversive. They simply do not experience hunger in a way that motivates eating, or they lose track of meals because food does not occupy cognitive space for them. The maintaining processes here are interoceptive: weak or absent hunger cues, difficulty interpreting internal states, and sometimes a general pattern of disconnection from bodily signals that extends beyond food to thirst, fatigue, and pain. Executive function deficits (common in ADHD, which co-occurs with ARFID at elevated rates) may compound the problem, as the person lacks the organizational scaffolding to eat regularly in the absence of natural hunger cues.

Fear of aversive consequences is driven by a specific learned association. A choking episode, a vomiting incident, an allergic reaction, or an episode of severe gastrointestinal distress creates a conditioned fear response that generalizes from the specific food involved to broader categories of food or eating situations. The maintaining processes here are classical conditioning (the pairing of eating with a traumatic outcome), catastrophic misappraisal of somatic sensations (any throat tightness becomes a choking signal, any nausea becomes an imminent vomiting episode), safety behaviors that prevent corrective learning, and avoidance patterns that progressively restrict the person’s diet and social life.

A person whose ARFID is maintained by sensory amplification and rigid categorization needs different interventions than a person whose ARFID is maintained by conditioned fear and catastrophic appraisal. This is not a subtle clinical distinction. It is the difference between targeting a sensory processing pattern and targeting a phobic response. Using the same protocol for both is roughly equivalent to prescribing the same medication for a headache caused by tension and a headache caused by elevated intracranial pressure: the symptom is the same, the generator is different, and the treatment that helps one may do nothing for the other.

Why CBT-AR Does Not Fit Every Presentation

CBT-AR (cognitive-behavioral therapy for ARFID), developed by Jennifer Thomas and Kamryn Eddy at Massachusetts General Hospital, is currently the most structured treatment manual available for ARFID. It is a valuable resource, and for patients whose ARFID is driven primarily by fear of aversive consequences, its exposure-based components work well. For patients with significant cognitive distortions about food safety, its psychoeducation and cognitive restructuring modules are appropriate.

But CBT-AR applies the same modular structure to all three presentations, and its modules do not map cleanly onto the maintaining processes of each. A patient whose ARFID is maintained by interoceptive disconnection and absent hunger cues does not benefit much from exposure hierarchies designed for fear-based avoidance. A patient whose sensory sensitivity is rooted in neurobiological differences in sensory processing does not respond to cognitive restructuring about the safety of new foods, because the problem was never a cognition about safety. It was always a sensory experience of aversiveness that cognitive reframing cannot override.

The manual is aware of these differences and attempts to accommodate them through module selection. In practice, though, the accommodation is limited. The clinician still works within the CBT-AR framework and selects from its menu. If the maintaining processes in a particular case require interventions that sit outside that menu (interoceptive training for a patient with absent hunger cues, acceptance-based work for a patient whose sensory responses are neurobiologically fixed and require accommodation rather than elimination, or RO-DBT skills for a patient whose food restriction is embedded in a broader overcontrol profile), the protocol does not provide them.

How PBT Assessment Works

Process-based therapy begins with a functional assessment that maps the specific processes maintaining food avoidance in this particular person. The assessment draws on multiple data sources: self-report measures of experiential avoidance, cognitive flexibility, interoceptive awareness, and distress tolerance; behavioral observation during food-related tasks; developmental history; and, in my practice, idiographic clinical network analysis using the ICNS system I built for this purpose.

ICNS asks the patient to rate a set of variables (anxiety, sensory distress, interoceptive awareness, hunger, avoidance urges, food variety, mood, social eating comfort, and others specific to their presentation) repeatedly over time through brief daily assessments on their phone. The resulting data generates a network map: a visual representation of which processes are most strongly connected to which outcomes in this person’s actual daily experience. The map shows which process, when it shifts, produces the largest downstream effects on the rest of the network.

This is the keystone process: the variable that, if targeted successfully, will produce the most change across the entire system. In one patient, the keystone might be distress tolerance. In another, it might be interoceptive awareness. In a third, it might be rigid self-rules about contamination. The network map makes the target visible rather than requiring the clinician to guess based on the diagnostic label.

How Treatment Adapts

Once the keystone process is identified, PBT selects interventions from across therapeutic traditions that have evidence for changing that specific process. If the keystone is experiential avoidance (the person organizes their life around avoiding the distress that new foods produce), acceptance and commitment therapy techniques targeting willingness and values-based action are indicated. If the keystone is conditioned fear, graduated exposure with response prevention, drawn from the behavioral tradition, is the primary tool. If the keystone is interoceptive disconnection, interoceptive training exercises (body scanning, hunger-fullness rating scales, somatic attention practices) from multiple traditions are appropriate. If the keystone is rigid overcontrol, RO-DBT skills targeting flexibility and openness enter the treatment.

The treatment plan is not fixed at the outset. Because the process network is dynamic, the keystone process can shift as treatment progresses. A patient might begin treatment with conditioned fear as the primary maintaining process. As exposure work reduces the fear, interoceptive disconnection (previously masked by the fear) emerges as the new bottleneck. PBT adapts by shifting the intervention target. The ICNS network map, updated every two to three weeks, makes these shifts visible in real time rather than requiring the clinician to notice them retrospectively.

This adaptive quality distinguishes PBT from protocol-driven treatment. A protocol says: deliver these modules in this order for this many sessions. PBT says: target the most influential process right now, monitor what changes, and redirect when the network reorganizes.

Clinical Experience with ARFID at Multiple Levels of Care

I have treated ARFID across outpatient, intensive outpatient, and partial hospitalization settings. The patterns are consistent. Patients referred from higher levels of care have often gained some food variety through structured exposure and meal support but relapse when they return to environments without that structure. The relapse occurs because the treatment addressed the behavioral output (what the person eats) without changing the processes that generated the restriction (how the person experiences sensory input, interprets bodily signals, relates to distress, or applies rules to food).

The patients who sustain gains are the ones whose treatment reached the generating process. A teenager whose sensory sensitivity is neurobiological may never find certain textures pleasant, but if treatment helps her build distress tolerance, increase psychological flexibility around food categories, and develop interoceptive awareness of hunger signals that motivate eating despite the sensory cost, her food repertoire expands and stays expanded because the processes maintaining the restriction have shifted.

PBT makes this targeting possible by refusing to assume that the diagnosis tells you what to treat. ARFID is the label. The process network is the map. Treatment follows the map.

Frequently Asked Questions

How does process-based therapy treat ARFID?

PBT identifies the specific psychological processes maintaining food avoidance in each individual case: sensory sensitivity, interoceptive confusion, anxiety-driven avoidance, low distress tolerance, or any combination. Treatment targets those specific processes rather than applying a one-size-fits-all protocol.

Why does ARFID need a personalized approach?

ARFID has three recognized presentations (sensory sensitivity, low appetite, fear of aversive consequences) that involve different maintaining processes. A person whose ARFID is driven by sensory sensitivity needs different interventions than someone whose ARFID is driven by a choking phobia. PBT maps the individual process network and treats accordingly.

Get Started

For questions about whether Process-Based Therapy is the right fit for your situation, or to schedule a consultation:

Schedule a consultation →