Severe enduring illness and compulsory treatment

TL;DR — “Severe and enduring AN” has no universally validated duration, treatment-history or prognosis threshold. Compulsory treatment can be life-preserving during acute incapacity or danger. The comparative long-term evidence remains weak, but it is no longer absent: a Danish register cohort of 4,425 people diagnosed with AN between 2000 and 2016 found that the 821 (18.5%) who received involuntary treatment had markedly higher subsequent mortality — all-cause hazard ratio 2.21 (95% CI 1.54–3.17), external causes 3.88 (1.94–7.77) and suicide 5.30 (2.07–13.54), adjusted for sex, age at diagnosis, birth year, prior involuntary treatment, self-harm and other psychiatric morbidity (Bager 2026, PMID 42383339). Because compulsion is applied after voluntary options are exhausted, this measures the vulnerability of the group, not the effect of the intervention; the authors frame involuntary treatment as an indicator of vulnerability rather than a cause. Capacity is decision-specific and especially difficult when AN shapes the values and reasoning being evaluated. The proposal to recognize “terminal AN” was made in 2022 with four proposed clinical characteristics, was contested immediately and extensively, was empirically tested in 2024, and was formally disavowed by its own lead author in 2025 (Gaudiani 2022, PMID 35168671; Robison 2024, PMID 38619462; Gaudiani 2025, PMID 40361218). The underlying clinical problem — what to offer people whose repeated treatment has not helped — is not disavowed and remains open.

Definitions are interventions

Label	Intended use	Risk
Severe	Describe current symptom/medical burden	Reduced to BMI alone
Enduring	Recognize long duration/repeated treatment	Becomes a prophecy of non-recovery — and NICE NG69 recommendation 1.2.8 explicitly rules out duration of illness, alongside BMI, as a basis for deciding whether to offer treatment (retrieved 2026-09-02)
Treatment-resistant	Describe non-response to adequate care	“Adequate” dose/access often unclear
Palliative approach	Prioritize comfort, harm reduction, quality of life	Misread as abandonment or necessarily end-of-life
Terminal AN	Claim irreversible dying trajectory	Criteria were proposed in 2022, empirically contradicted in 2024 and disavowed by their lead author in 2025 (Gaudiani 2022, PMID 35168671; Robison 2024, PMID 38619462; Gaudiani 2025, PMID 40361218)

The one randomized trial designed for this population deserves separate mention because it shows the category is researchable. Sixty-three adults with an illness history of at least seven years were randomized across multiple sites to CBT-AN or SSCM, both modified for SE-AN, over 30 outpatient visits across 8 months, with quality of life, mood and social adjustment — not weight — as the main outcomes. Retention was 85%; both groups improved significantly; there were no between-group differences at end of treatment, and at 12 months CBT-AN participants had lower EDE global scores (p = 0.004) and higher readiness for recovery (p = 0.013) (Touyz 2013, PMID 23642330). A population repeatedly described as untreatable was recruited, retained at 85%, and improved — on outcomes chosen to matter to them.

Reviews of severe enduring AN identify heterogeneous definitions and small treatment studies, supporting individualized, collaborative goals but not a clean futility threshold (Dalle Grave 2020, PMID 32400903; Zhu 2020, PMID 32265758; Smith 2020, PMID 33488410). Around 20–30% of people with an eating disorder do not respond to the best available treatments and develop persistent treatment-refractory illness — a real and substantial group, whose existence is the legitimate starting point of this debate even though it does not license a prognosis for any individual within it (Bryson 2024, PMID 38503683).

Danish register work offers a different route to the same problem. Rather than defining severity by duration, the AN register-based severity index (AN-RSI) weights early or late onset, number of inpatient admissions and outpatient treatments, cumulative treatment length and illness duration, and is scored five years after first diagnosis. Among 9,167 individuals diagnosed with AN, higher AN-RSI scores predicted mortality from AN, from a somatic anorexia diagnosis, from suicide, from alcohol-related causes and from any cause, with the top 20% of scores carrying especially high mortality; those individuals were also more likely to remain in treatment over the next five years (Larsen 2025, PMID 40670905). This is a prognostic index derived from administrative data, validated against death rather than against clinical consensus — which is exactly the design a defensible severity definition would need — but predicting group-level mortality is not the same as identifying an individual for whom recovery is impossible.

The labelling problem, quantified

The definitional instability described above is not an impression. A PRISMA systematic review of 32 records extracted the explicit labels and defining criteria used for this subgroup and found that "chronic" and "severe and enduring" were the most common adjectives, and that illness duration and number of previously failed treatment attempts were the most common defining features (Broomfield 2017, PMID 28444828). The two consequences the authors name are the ones this page keeps running into: inconsistent recruitment across studies of the same putative population, and misdiagnosis of individuals who may or may not have a durable course. Any statement of the form "X% of people with SE-AN…" inherits whichever definition its source used.

That matters against NICE NG69 recommendation 1.2.8, which rules out duration of illness as a basis for deciding whether to offer treatment (retrieved 2026-09-02) — the most commonly used defining feature of the category is the one the guideline says must not gate care.

Capacity in AN

Decision-making capacity is specific to a decision and time, not identical to diagnosis, agreement, insight or intelligence. Standard elements include understanding, retaining, using/weighing relevant information and communicating a choice. AN creates a distinctive problem when illness-linked valuation of thinness, fear of gain or identity shapes the very weighing process under examination. Jefferson's argument — a philosophical one, not an empirical capacity study — is that the proposed "terminal anorexia" category creates a false sense of certainty about illness trajectory by reasoning from an analogy with physical disorders and from the effects of starvation, and that this conceptualization is in direct tension with the simultaneous claim that people meeting those criteria retain decision-making capacity; on that basis the proposed diagnosis should be rejected (Jefferson 2024, PMID 39463268).

Safeguard	Purpose
Decision-specific assessment	Avoid global declarations
Optimize physiology and communication	Reduce reversible impairment
Explore values over time	Distinguish enduring values from illness-driven narrowing
Independent review	Reduce team desperation and confirmation bias
Least-restrictive alternative	Minimize coercive burden
Reassess repeatedly	Capacity and risk can change

Compulsory treatment

Legal frameworks differ. Evidence on long-term benefit is largely observational and confounded because involuntary patients are generally sicker — a confounding that is now quantified rather than asserted. In the Danish cohort, all-cause mortality after involuntary treatment was more than doubled and suicide mortality more than quintupled relative to those never treated involuntarily, with 206 deaths among 4,425 individuals (4.7%) over the study period (Bager 2026, PMID 42383339). A cohort of this design cannot distinguish "compulsion harms" from "compulsion marks the patients most likely to die", and the authors do not attempt to: their conclusion is that this group needs sustained clinical attention. One small study does compare pathways directly, and its results run opposite to the register signal. Twenty-three patients involuntarily admitted for extremely severe AN were compared with 25 voluntarily admitted patients almost four years after discharge. Weight restoration was higher in the voluntary group (p = 0.01), but quality of life, BMI and mortality did not differ significantly between groups. Within the involuntary group, BMI rose and weight-controlling behaviours fell between admission and follow-up (p < 0.05). Most notably, despite negative experiences of the involuntary treatment itself, participants' perception of its necessity increased from admission to follow-up (p < 0.01) to a level comparable with the voluntary group (p > 0.05). The authors conclude that involuntary treatment does not appear to be a barrier to weight gain, clinical improvement or long-term attitudes towards treatment (Abry 2024, PMID 37690079).

Two features limit how far that carries. The groups were not matched, and 23 versus 25 patients cannot detect a mortality difference of the size the Danish register reports. A larger historical comparison followed 81 compulsorily admitted and 81 other specialist-unit patients for about 20 years. The groups' mortality no longer differed significantly at that horizon, although an earlier five-year excess after compulsory admission had been observed (Ward 2015, PMID 25545619). This adds comparative mortality data but not causal identification: neither historical comparison adequately controls why compulsion was used, and neither measures survival, remission, trauma, trust and quality of life together. What the smaller study does establish is that retrospective endorsement of compulsion by the people subjected to it is measurable and, in this cohort, increased over time.

What Carney and colleagues actually argue is narrower and worth stating precisely: that people with severe and enduring AN have high levels of disability, under- or unemployment and welfare dependence; that AN shows a degree of quality-of-life impairment similar to depression or schizophrenia; and that a rehabilitation model prioritizing harm minimization and quality of life can be cogently defended, with involuntary treatment used judiciously (Carney 2019, PMID 31046931). The claim that coercion causes trauma or damages trust is a plausible and widely voiced concern in the clinical and lived-experience literature, not a finding of that review, and it is measured in no cohort located in this audit.

In minors, authority, safeguarding and emerging autonomy add layers; Ramasamy and colleagues analyze involuntary treatment in severe enduring cases (Ramasamy 2021, PMID 34187872).

The terminal-AN dispute

What was actually proposed

The 2022 paper presented three case studies — one patient asked to be a posthumous co-author — and proposed four clinical characteristics for a terminal eating disorder: (1) a diagnosis of anorexia nervosa; (2) older age, e.g. over 30; (3) previous participation in high-quality care; and (4) a clear and consistent determination, by a patient possessing decision-making capacity, that additional treatment would be futile, knowing that their actions will result in death (Gaudiani 2022, PMID 35168671). The stated motivation was that a minority of people with severe and enduring eating disorders find that further treatment seems futile and harmful, that no achievable level of harm reduction relieves their suffering, and that they and their carers often receive minimal support from professionals hampered by a paucity of end-of-life literature.

Any accurate account of this dispute has to begin there, because most of the subsequent literature is a response to those four criteria rather than to the general idea of palliative care in AN.

Arguments advanced for the proposal

A narrative review of 57 articles identified three recurring arguments in favour: forgoing "excessive" or "futile" treatment; respecting autonomy by heeding patient requests; and harm reduction through access to resources, including medically assisted death (Connor 2026, PMID 42312722). Independently, the futility argument as advanced elsewhere holds that for those unremittingly ill for 8 to 12 years full recovery, while possible, is unlikely, so harm reduction, palliative care and end-of-life care merit exploration (Westmoreland 2021, PMID 34763793).

Arguments against recognizing a terminal category

The same review found four recurring counter-arguments: the difficulty of defining "terminal" AN; the difficulty of assessing capacity in AN; the harm the label itself would cause to patients; and the view that effort should instead go to systemic problems with the quality of and access to care (Connor 2026, PMID 42312722). Its conclusion about the debate's structure is worth recording: both sides raise important issues, but there is a lack of shared understanding, particularly about the value judgements embedded in each position.

Specific published objections include that the criteria lack an evidence base and do not acknowledge the evidence for prolonged timelines to recovery or the nuances of assessing capacity (Riddle 2022, PMID 35710504); that individual course and treatment response cannot be predicted accurately enough to determine futility, and that decisional capacity may be unusually complex here (Crow 2023, PMID 37057340); and that the proposed diagnosis creates a false sense of certainty about trajectory by reasoning from an analogy with physical illness, in tension with the simultaneous claim that those meeting the criteria retain capacity (Jefferson 2024, PMID 39463268).

The empirical test

The dispute is not purely normative, and this page previously implied that it was. An empirical evaluation applied the first three proposed criteria to 782 patients in US eating-disorder treatment facilities, all meeting DSM-5 criteria for AN. Fifty-one met the first three criteria (T-AN) and 16 also met a proxy for the fourth (TD-AN); these were compared with 731 "not terminal" patients and with the 133 aged 30 or over. Against the tight syndromal interconnection and inevitable decline that a terminal designation implies, the results showed substantial variability within the T-AN group and its TD-AN subset and an overall trend of improvement across physiological indices (white cell count, albumin, AST, BMI) and self-reported eating-disorder, depression, anxiety and obsessive-compulsive symptoms. The authors present this as empirical evidence against specifying a T-AN diagnosis (Robison 2024, PMID 38619462).

The design cannot exclude the possibility that the proposed criteria identify a real group whom a treatment-facility sample under-represents — people who have disengaged from care would not appear in it. But it is the first direct test of whether the proposed criteria pick out a population with a terminal trajectory, and it found that they do not.

The disavowal

In 2025 the lead author of the original proposal expressly disavowed the concept and the phrase "terminal anorexia nervosa", acknowledged the original article's inadequacies after several years of listening to personal, clinical and scholarly objections, and called for continued collaboration on care for this group (Gaudiani 2025, PMID 40361218). Collaborative ethical writing published the same year takes that disavowal as its starting point and proposes forward-looking work across terminology, capacity assessment, systemic supports and integrated care, advocating descriptive terms that preserve recovery potential in place of deterministic labels (Bauschka 2025, PMID 40760030).

The label is therefore withdrawn by its proposer and unsupported by the one empirical test of its criteria. The clinical question it was raised to address — how to support people for whom repeated treatment has not worked, without either abandoning them or compelling them indefinitely — is untouched by that withdrawal.

Where lived experience actually sits

Two published lived-experience letters by the same author argue that "terminal anorexia" is a harmful new label, set out problems with each proposed criterion in turn, and warn about ambiguities in how the criteria could be applied (Asaria 2023, PMID 37400874; Asaria 2023, PMID 38082348). This page previously cited those letters among the arguments for the proposal, which inverted their position; that error was corrected in the 2026-09-02 audit.

Evidence that lived experience is genuinely plural comes instead from a mixed-methods stakeholder study of people with lived experience, caregivers and clinicians (up to 219 quantitative and 182 qualitative respondents). Participants were more negative about three of the four criteria — the age threshold most of all, with 38.8% strongly disagreeing and 22.4% disagreeing — while the fourth criterion (capacitous, consistent determination of futility) drew the most agreement (21.6% strongly agree, 25.2% agree). Themes spanned ethical concern, potential for harm, respect for autonomy, difficulty determining prognosis, capacity and futility, treatment problems, and a preference for systemic reform over terminal labelling. Most participants were critical, and a subset supported elements of the framework, particularly around autonomy and relief of suffering (Robb 2026, PMID 42410953). "Patient voice" is not a single position, but on these criteria it leans clearly one way.

Assisted dying: what has actually happened

The terminal-AN debate is not hypothetical at its endpoint. A systematic review of peer-reviewed studies and publicly available government reports identified 10 articles and 20 government reports describing at least 60 patients with eating disorders who underwent assisted dying between 2012 and 2024, in multiple countries including jurisdictions that restrict the practice to terminal conditions. Clinical rationales fell into three domains — irremediability, terminality and voluntary request — with reports emphasizing that these patients had terminal, incurable or untreatable conditions and adequate decision-making capacity. The reviewers' findings on the quality of that record are the substance of the paper: most government reports did not contain data descriptive enough to verify the psychiatric conditions involved, and "in many cases, the clinical rationales that were used to affirm patients with EDs were eligible for assisted death lack validity and do not cohere with empirical understanding" (Roff 2024, PMID 39143961).

Set against the 22-year cohort in which AN recovery rose from 31.4% at 9 years to 62.8% at 22 years, and against the empirical test in which patients meeting the first three terminal-AN criteria showed an overall trend of improvement (Eddy 2017, PMID 28002660; Robison 2024, PMID 38619462), the irremediability rationale is the weakest link in the documented reasoning. This is a description of what has been recorded and of the gaps in that record; it is not a position on assisted dying law.

Open questions

• Does compulsory treatment improve survival or durable recovery relative to feasible alternatives after confounding by severity? The Danish register reports elevated mortality after involuntary treatment (all-cause HR 2.21), while a 162-patient historical comparison found that an early mortality difference attenuated by about 20 years; neither design controls indication well enough to estimate treatment effect (Bager 2026, PMID 42383339; Ward 2015, PMID 25545619).
• How reliably do independent assessors evaluate use/weighing in AN, and what predicts disagreement?
• Can palliative principles be offered without a terminal label and while preserving recovery options? This is now the operative question rather than a hypothetical one, since the label has been withdrawn by its proposer while the clinical problem it addressed remains (Gaudiani 2025, PMID 40361218; Bauschka 2025, PMID 40760030).
• Would the proposed criteria identify a different population outside treatment-facility samples? The one empirical evaluation drew on patients in active treatment and found improvement rather than terminal decline; people who have disengaged from care could not appear in it (Robison 2024, PMID 38619462).
• What minimum prognostic performance would justify any terminal classification? Register-based severity indices now predict cause-specific mortality at group level, which is a step the 2023 critique said was missing — but no published model reports calibration or discrimination adequate for an individual futility judgement (Crow 2023, PMID 37057340; Larsen 2025, PMID 40670905).
• Should the field abandon duration-based definitions altogether? Duration and number of failed treatments are the most common defining features in the literature, and NICE explicitly prohibits using duration to gate treatment (Broomfield 2017, PMID 28444828).
• Does compulsion harm long-term outcome, or mark those most at risk? A national register finds all-cause hazard 2.21 and suicide hazard 5.30 after involuntary treatment, while a 48-patient comparison finds no difference in quality of life, BMI or mortality at four years and rising retrospective endorsement of the treatment's necessity (Bager 2026, PMID 42383339; Abry 2024, PMID 37690079).
• How should the ≥60 documented assisted deaths in eating disorders be reconciled with the recovery trajectories observed at 22 years and with the failure of the proposed terminal criteria to identify a declining group (Roff 2024, PMID 39143961; Eddy 2017, PMID 28002660; Robison 2024, PMID 38619462)?
• Can the Touyz design — long-duration cohort, quality-of-life primary outcome, 85% retention — be replicated at a scale that would support a treatment recommendation for this group (Touyz 2013, PMID 23642330)?

Related pages

• Mortality and long-term outcome — limits of individual prognosis.
• Service models and setting — coercion context.
• Patient experience and advocacy — plural lived experience.

References

1. Westmoreland P, et al. Severe and enduring anorexia nervosa and futility: a time for every purpose? Psychiatr Clin North Am. 2021. PMID 34763793.
2. Crow SJ, et al. Terminal anorexia nervosa cannot currently be identified. Int J Eat Disord. 2023. PMID 37057340.
3. Carney T, et al. Involuntary treatment and quality of life. Psychiatr Clin North Am. 2019. PMID 31046931.
4. Ramasamy RS, et al. Involuntary treatment of minors with severe and enduring anorexia nervosa. J Am Acad Psychiatry Law. 2021. PMID 34187872.
5. Jefferson A. “Terminal anorexia,” treatment refusal and decision-making capacity. Camb Q Healthc Ethics. 2024. PMID 39463268.
6. Dalle Grave R, et al. Severe and enduring anorexia nervosa: no easy solutions. Int J Eat Disord. 2020. PMID 32400903.
7. Zhu J, et al. Psychological treatments for severe and enduring anorexia nervosa. Front Psychiatry. 2020. PMID 32265758.
8. Smith S, et al. Characterizing treatment-resistant anorexia nervosa. Front Psychiatry. 2020. PMID 33488410.
9. Asaria A, et al. “Terminal anorexia”: a lived experience perspective. J Eat Disord. 2023. PMID 37400874.
10. Asaria A, et al. “Terminal anorexia”: lived-experience perspective on proposed criteria. J Eat Disord. 2023. PMID 38082348.
11. Bauschka M, et al. From disavowal to dialogue: collaborative approaches to severe and enduring anorexia nervosa. J Eat Disord. 2025. PMID 40760030.
12. Bager L, et al. Excess mortality among patients with anorexia nervosa treated involuntarily. Int J Eat Disord. 2026. PMID 42383339.
13. Larsen JT, et al. Construction of a register-based severity index for anorexia nervosa in Denmark. Eur Psychiatry. 2025;68:e104. PMID 40670905.
14. Bryson C, et al. Established and emerging treatments for eating disorders. Trends Mol Med. 2024. PMID 38503683.
15. Gaudiani JL, et al. Terminal anorexia nervosa: three cases and proposed clinical characteristics. J Eat Disord. 2022;10:23. PMID 35168671.
16. Robison M, et al. "Terminal anorexia nervosa" may not be terminal: an empirical evaluation. J Psychopathol Clin Sci. 2024;133:285-296. PMID 38619462.
17. Gaudiani JL. Listening and learning: the need for a compassionate and patient-centered approach for longstanding AN. J Eat Disord. 2025;13:81. Formal disavowal of the term. PMID 40361218.
18. Riddle M, O'Melia AM. First, do no harm: the proposed definition of "terminal anorexia" is fraught with danger for vulnerable individuals. J Eat Disord. 2022;10:81. PMID 35710504.
19. Connor C, Kim S. What are the arguments for and against recognizing 'terminal' anorexia nervosa? A narrative review. Int Rev Psychiatry. 2026:1-13. PMID 42312722.
20. Robb S, et al. Stakeholder views on the proposed criteria for 'terminal anorexia': a mixed-methods study. Eat Disord. 2026:1-19. PMID 42410953.
21. Broomfield C, et al. Labeling and defining severe and enduring anorexia nervosa: a systematic review and critical analysis. Int J Eat Disord. 2017;50:611-623. PMID 28444828.
22. Abry F, et al. Longitudinal investigation of patients receiving involuntary treatment for extremely severe anorexia nervosa. Eur Eat Disord Rev. 2024;32:179-187. PMID 37690079.
23. Roff C, Cook-Cottone C. Assisted death in eating disorders: a systematic review of cases and clinical rationales. Front Psychiatry. 2024;15:1431771. PMID 39143961.
24. Eddy KT, et al. Recovery from anorexia nervosa and bulimia nervosa at 22-year follow-up. J Clin Psychiatry. 2017;78:184-189. PMID 28002660.
25. Touyz S, et al. Treating severe and enduring anorexia nervosa: a randomized controlled trial. Psychol Med. 2013;43:2501-2511. PMID 23642330.
26. Ward A, et al. Follow-up mortality study of compulsorily treated patients with anorexia nervosa. Int J Eat Disord. 2015;48:860-865. PMID 25545619.