Skip to content

Not yet independently audited

This page was written from verified sources but has not yet passed the independent citation audit. Treat its claims as provisional.

Statistics

Every row retains population, period and method. Conflicting or non-equivalent estimates are not averaged. All figures were re-verified against the source abstract during the independent audit on 2026-09-02.

Prevalence and incidence

Statistic Estimate Population / method Source
Lifetime AN, US adults 0.80% (SE 0.07%) NESARC-III; 36,306 adults; AUDADIS-5 structured interview; 2012–13 Udo 2018, PMID 29859631
12-month AN, US adults 0.05% (SE 0.02%) Same Udo 2018, PMID 29859631
Lifetime AN, young women 0.8–6.3% Range across DSM-5 studies of young people, 2013–22, mostly Western settings Silén 2022, PMID 36125216
Lifetime AN, young men 0.1–0.3% Same Silén 2022, PMID 36125216
Any DSM-5 eating disorder by early adulthood 5.5–17.9% of young women; 0.6–2.4% of young men Same Silén 2022, PMID 36125216
Lifetime AN prevalence, review estimate Up to 4% of females; 0.3% of males Narrative review of recent epidemiology van Eeden 2021, PMID 34419970
Incidence trend Overall rate stable over decades; increased among those aged <15 Review; earlier detection vs earlier onset not separable van Eeden 2021, PMID 34419970
Incidence trend, earlier review Overall rate stable; increase among 15–19-year-old girls Review Smink 2012, PMID 22644309
Peak age at onset, feeding/eating disorders 15.5 years; median 18 (IQR 15–23) 192-study meta-analysis, n=708,561; ICD-11 diagnostic block, not AN-specific; k=11 Solmi 2022, PMID 34079068
Onset before 14 / 18 / 25, feeding/eating disorders 15.8% / 48.1% / 82.4% Same Solmi 2022, PMID 34079068
Eating-disorder years of life lost, global 17,361.5 (95% UI 15,518.5–21,459.8) GBD 2019; AN and BN were the only mental disorders modelled as underlying causes of death; authors call the YLL estimate "extremely low" and not reflective of premature mortality GBD 2019 Mental Disorders Collaborators 2022, PMID 35026139
Sex of AN patients 5–15% of patients with AN are men Endocrine review; limited data in adolescent boys and men Schorr 2017, PMID 27811940
Lifetime AN prevalence, population twins 1.20% women; 0.29% men Swedish Twin Registry; 31,406 twins born 1935–1958; interview, hospital-discharge or death-certificate ascertainment Bulik 2006, PMID 16520436
Register-diagnosed AN incidence, Denmark 6.4 → 12.6 per 100,000 person-years, 1995–2010 Danish Psychiatric Central Research Registry, ages 4–65; N=5,902 incident AN. A sizeable part of the rise is attributed by the authors to a general increase in first-time psychiatric diagnosis (249,607 over the same period) Steinhausen 2015, PMID 25809026
Register-diagnosed BN incidence, Denmark 6.3 → 7.2 per 100,000 person-years, 1995–2010 Same; N=5,113 Steinhausen 2015, PMID 25809026
Male-to-female ratio, Denmark 2010 1:8 for AN; 1:20 for BN Same Steinhausen 2015, PMID 25809026
Diagnosed eating disorder, US transgender claims 2.43% (95% CI 2.14–2.74); AN 0.84% 10,415 people identifiable as transgender via gender-affirming-care codes, 2018 MarketScan Commercial Database. Authors note this is lower than self-report estimates Ferrucci 2022, PMID 35524487
Same, by age 5.60% (ages 12–15); 0.52% (ages 45–64) Same Ferrucci 2022, PMID 35524487
Eating-disorder hospital admissions, pandemic +48% on average (591 → 876 across 10 studies) Systematic review of 53 studies, 36,485 individuals, Nov 2019–Oct 2021 Devoe 2023, PMID 35384016
Admissions vs public-health stringency Per 10% stringency increase: aRR 1.05 (Quebec), 1.05 (Ontario), 1.08 (Prairies), 1.11 (British Columbia) 11,289 Canadian eating-disorder hospitalizations, Apr 2016–Mar 2023; 77% female aged 12–17; interrupted time series Roumeliotis 2024, PMID 38976259
Excess pandemic admissions at 1 year RR 2.02–2.44 across regions Same Roumeliotis 2024, PMID 38976259
Lifetime prevalence, any eating disorder 2–5% worldwide Narrative review Attia 2025, PMID 40048192
Lifetime depression in AN 49.5% (vs 76.3% BN, 65.5% BED) Same Attia 2025, PMID 40048192

Mortality and outcomes

Statistic Estimate Population / method Source
Mortality rate 5.1 deaths per 1,000 person-years AN; 36-study meta-analysis, literature through 2010; 166,642 person-years Arcelus 2011, PMID 21727255
Standardized mortality ratio 5.86 AN versus population expectation; meta-analysis Arcelus 2011, PMID 21727255
Suicide among deaths 20% Proportion of deaths in AN cohorts, not patient cumulative risk Arcelus 2011, PMID 21727255
All-cause mortality risk ratio, AN 5.52 (95% CI 4.47–6.82) 83 studies; 307,710 ED patients vs 15,719,076 general-population controls; mean follow-up 11.96 yr; 94.35% female Semchishen 2026, PMID 41536100
All-cause mortality risk ratio, any ED 4.92 (95% CI 4.03–6.00) Same; ranges from AN down to a non-significant difference for BED Semchishen 2026, PMID 41536100
Suicide mortality risk ratio, AN 9.86 (95% CI 5.63–17.27) Same; BN 6.15 (2.52–15.04) Semchishen 2026, PMID 41536100
Natural- vs non-natural-cause RR, any ED 3.47 (2.29–5.25) and 6.46 (4.62–9.04) Same Semchishen 2026, PMID 41536100
Standardized mortality ratio, AN (2026 update) 5.06 (95% CI 3.47–7.38) 30 studies, 33,176 patients, 22 pooled; searched to May 2025; studies with zero deaths excluded Lai 2026, PMID 41277145
SMR by sex Male samples 3.47 (1.60–7.52); female 3.86 (1.82–8.20) Same; overlapping intervals, few male samples Lai 2026, PMID 41277145
Cause of death, AN Suicide 21%; cardiac 19% Same Lai 2026, PMID 41277145
Overall ED mortality, pooled proportion 0.4% (95% CI 0.2–0.7) 214 studies; mean follow-up 72.2±117.7 months Solmi 2024, PMID 38214616
Observational-study mortality rate, all EDs 5.2 deaths/1,000 person-years (95% CI 4.4–6.1) 167 studies; mean follow-up 88.7 months; range 8.2 (mixed ED) to 3.4 (BN) Solmi 2024, PMID 38214616
Mortality after involuntary treatment All-cause HR 2.21 (1.54–3.17); external causes 3.88 (1.94–7.77); suicide 5.30 (2.07–13.54) Danish register cohort; 4,425 people with AN diagnosed 2000–2016, 93.3% female, mean age 22; 821 (18.5%) received involuntary treatment, 206 (4.7%) died; adjusted. Confounded by indication Bager 2026, PMID 42383339
All-cause SMR, methodological re-analysis 5.2 (95% CI 3.7–7.5) 41 cohorts from 40 studies, 1966–2010; double data extraction, over-dispersed Poisson log-linear regression; 10% lower than previously reported Keshaviah 2014, PMID 25214371
Suicide SMR, same re-analysis 18.1 (95% CI 11.5–28.7) vs 15–34-year-old females Same; 49% lower than previously reported suicide SMRs Keshaviah 2014, PMID 25214371
Weighted SMR, any eating disorder (2010–2024) 3.39 (95% CI 2.90–3.95); I²=95.1%, k=74 Four databases, studies 2010 to Oct 2024 Krug 2025, PMID 39889307
SMR by diagnosis (same) AN 5.21 (k=30); EDNOS 2.51 (k=8); BN 2.20 (k=18); BED 1.46 (k=3) Same Krug 2025, PMID 39889307
SMR, treatment-seeking cohort 4.37 (95% CI 2.4–7.3) lifetime AN; 7.7 (3.7–14.2) within first 10 years of follow-up 246 women interviewed 6-monthly for median 9.5 years; vital status to median 20 years; 16 deaths (6.5%) Franko 2013, PMID 23771148
SMR by illness duration 3.2 (0.9–8.3) for 0–15 years lifetime AN (4/119 died); 6.6 (3.2–12.1) for >15–30 years (10/67) Same Franko 2013, PMID 23771148
Recovery, 22-year follow-up AN 62.8% at 22 years vs 31.4% at 9 years; BN 68.2% at both 228 women, DSM-III-R/DSM-IV; 77% re-interviewed, multiple imputation; mean follow-up 22.10 (SD 1.10) years Eddy 2017, PMID 28002660
Early recovery predicting long-term recovery AN OR 10.5 (95% CI 3.77–29.28); BN OR 1.0 (0.49–2.05) Same Eddy 2017, PMID 28002660
Remission, large inpatient cohort 30% total sample; 40% in the 20-year subsample 1,693 consecutive inpatients of a specialized hospital over 25 years (mean 10-year follow-up); 112 with 20-year follow-up Fichter 2017, PMID 28644530
20th-century outcome baseline <50% of survivors recovered; ~33% improved; 20% chronically ill 119 study series, 5,590 patients; "no convincing evidence that outcome improved over the second half of the last century" Steinhausen 2002, PMID 12153817
Deaths in Danish AN severity cohort 132/9,167 (17 from AN, 30 suicide, 85 other) Danish register; born 1963–2007, diagnosed 1969–2013; AN-RSI scored 5 years after first diagnosis Larsen 2025, PMID 40670905
Hospitalization, any ED 26% (95% CI 18–36); AN 32%, BN 4% 18 studies; mean follow-up 43.2 months Solmi 2024, PMID 38214616
ANTOP 5-year global outcome Full recovery 41% (33–49); partial 41% (33–49); full-syndrome AN 18% (12–24) 154/242 (64%) reassessed at mean 5.96 years; observed data Herzog 2022, PMID 35294860
Overall ED recovery 46% (95% CI 44–49) 415 studies, 88,372 participants; transdiagnostic Solmi 2024, PMID 38214616
Overall ED chronicity 25% (95% CI 23–29) 170 contributing studies; mean follow-up 59.3 months Solmi 2024, PMID 38214616
Recovery <2 years 42% Transdiagnostic studies grouped by follow-up Solmi 2024, PMID 38214616
Recovery 2–<4 years 43% Same Solmi 2024, PMID 38214616
Recovery 4–<6 years 54% Same Solmi 2024, PMID 38214616
Recovery 6–<8 years 59% Same Solmi 2024, PMID 38214616
Recovery 8–<10 years 64% Same Solmi 2024, PMID 38214616
Recovery ≥10 years 67% Same Solmi 2024, PMID 38214616
Relapse after one-year remission 2/33 (6.1%) Selected 79-person long-term follow-up of adolescent RCT Le Grange 2014, PMID 25440306
New remission after one-year follow-up 10/44 (22.7%) Same selected follow-up Le Grange 2014, PMID 25440306

Genetics

Statistic Estimate Population / method Source
GWAS cases 16,992 ANGI + PGC-ED Watson 2019, PMID 31308545
GWAS controls 55,525 Same Watson 2019, PMID 31308545
Significant loci 8 Genome-wide significant Watson 2019, PMID 31308545
Twin-based heritability 50–60% Range summarized in GWAS background Watson 2019, PMID 31308545
Twin/family heritability, alternative range 33–84% Review of twin and family studies — shown side by side with the 50–60% figure, not reconciled Donato 2022, PMID 36479493
Additional GWAS locus (2026) 1 novel genome-wide significant locus near SOX5 Meta-analysis of European and Finnish AN GWAS data Song 2026, PMID 41927769
MTAG loci (2026) 86 significant, of which 25 novel (incl. VAMP2, LPL, BDNF) Multi-trait analysis borrowing power from correlated traits — not equivalent to loci discovered in AN cases Song 2026, PMID 41927769
Local genetic-correlation regions 185 significant; 100 with pleiotropy across multiple traits Same Song 2026, PMID 41927769

Comorbidity, familial risk and heritability

Statistic Estimate Population / method Source
Population-register heritability AN 36%; BN 39%; other EDs 30% Danish and Swedish registers 1972–2016; >67,000 people with EDs, first-degree relatives and matched controls from populations of 17 million Meijsen 2025, PMID 40615413
Genetic correlation, AN–OCD rg = 0.65 Same Meijsen 2025, PMID 40615413
Genetic correlation, AN–autism rg = 0.36 Same Meijsen 2025, PMID 40615413
Twin heritability, narrow DSM-IV AN a² = 0.56 (95% CI 0.00–0.87); c² = 0.05 (0.00–0.64); e² = 0.38 (0.13–0.84) Swedish Twin Registry, 31,406 twins Bulik 2006, PMID 16520436
Heritability of continuous disordered-eating score 0.65 (95% CI 0.61–0.68) 1,481 female Swedish twin pairs at age 18, EDI-2 Dinkler 2021, PMID 31843035
Twin genetic correlation, EDI-2 score with AN diagnosis 0.26 (95% CI 0.08–0.42) — vs 0.52 (0.39–0.65) for other ED diagnoses Same Dinkler 2021, PMID 31843035
Autistic traits in AN Hedge's g = 0.88 (95% CI 0.65–1.12) 22 studies; 1,172 AN vs 2,747 controls Inal-Kaleli 2025, PMID 39530423
Screening above ADOS cut-off in AN 29% (95% CI 19–38) Same; screening instruments, not diagnostic assessment Inal-Kaleli 2025, PMID 39530423
Set-shifting deficit, AN vs controls Hedge's g = −0.38 (95% CI −0.50 to −0.26) 38 studies, 3,505 participants; BN g = −0.55, not significantly different from AN Keegan 2021, PMID 33305366
Central coherence deficit, AN vs controls Hedge's g = −0.53 (95% CI −0.80 to −0.27) 16 studies, 1,112 participants; BN g = −0.70, not significantly different from AN Keegan 2021, PMID 33305366
Childhood maltreatment and ED odds ORs 1.71–3.29 across four maltreatment types Up to 63,989 women, Norwegian MoBa cohort; no multiplicative interaction with polygenic score Bjørndal 2026, PMID 42471970
ED polygenic score and ED odds ORs 1.05–1.31 Same Bjørndal 2026, PMID 42471970

Medical complications at extreme malnutrition

Statistic Estimate Population / method Source
Anaemia 79% 354 consecutive adults admitted to a French clinical-nutrition ED referral unit 1997–2014; mean admission BMI 12.2 ± 1.6 kg/m² Guinhut 2021, PMID 33023762
Neutropenia 53.9% Same Guinhut 2021, PMID 33023762
Hypertransaminasaemia 53.7% Same Guinhut 2021, PMID 33023762
Osteoporosis 46.3% Same Guinhut 2021, PMID 33023762
Hypokalaemia 39.5% Same; more frequent in binge/purge subtype Guinhut 2021, PMID 33023762
Hypophosphataemia 26% Same Guinhut 2021, PMID 33023762
Infectious complications 24.3% Same Guinhut 2021, PMID 33023762
Hypoglycaemia 13.8% Same Guinhut 2021, PMID 33023762
Cardiac dysfunction 7.1% Same Guinhut 2021, PMID 33023762
Gelatinous bone-marrow transformation (biopsy-proven) 6.5% Same Guinhut 2021, PMID 33023762
ICU transfer / deaths 10% transferred; 5 deaths Same Guinhut 2021, PMID 33023762
Hypoglycaemia at admission, very low BMI 50% detected on point-of-care testing; 20.6% severe (<50 mg/dL) 34 patients, BMI <14.5 kg/m², meal-based rapid weight-gain protocol Fischer 2022, PMID 35994205
Blood glucose <55 mg/dL, BMI <13 32 of 48 patients Median admission BMI 10.51; all patients with poor prognosis were in the hypoglycaemic group Matsunaga 2024, PMID 38702806
Renal impairment at admission (eGFR <90) 33% of 111 StRONG secondary analysis, ages 12–24, AN or atypical AN Downey 2022, PMID 35705423
Pericardial effusion, adults 9/48 vs 0/44 controls (p = 0.003) Prospective consecutive AN adults vs matched controls Scheggi 2022, PMID 35597626
Left-ventricular mass, adults 63 ± 15 g vs 99 ± 30 g controls (p < 0.001) Same Scheggi 2022, PMID 35597626
"AN cardiopathy" at paediatric admission 63% (24/38); effusion + bradycardia + mitral regurgitation together in 26% Children admitted for physical instability, 2015–2019 Borgia 2021, PMID 34050378
QTc prolongation, very-low-BMI inpatients 21% (4/19); 10.5% >500 ms; not correlated with LV mass, BMI or REE Mean BMI 12.3; blinded digital ECG, Fridericia correction; 68% on QT-prolonging drugs Krantz 2012, PMID 21917317
Seizure prevalence, eating disorders 4.5% (75/1,664 charts); PNES 29.3%, substance withdrawal 18.7%, primary seizure disorder 12%, electrolytes/hypoglycaemia 10.7%, presumed Wernicke's 4% Retrospective chart study Gibson 2023, PMID 37092766
Full Wernicke's triad at presentation in AN 8 of 12 published cases (vs ~16% in alcohol-related Wernicke's) Systematic review of case reports Oudman 2018, PMID 29984541
Maladaptive exercise, lifetime 88% pooled prevalence 31,671 people with lifetime EDs across US, Australia, New Zealand and Sweden (EDGI, NCT04378101) Watson 2026, PMID 41115789
Maladaptive exercise, current 40% any driven exercise; 35% compulsive exercise; 12% regular driven exercise Same; prevalence varies substantially by instrument Watson 2026, PMID 41115789

Pregnancy and neonatal outcomes

Statistic Estimate Population / method Source
Preterm birth, maternal AN RR 1.6 (95% CI 1.4–1.8) Swedish Medical Birth Register 2003–2014; 2,769 women with AN vs 1,225,321 without; adjusted Mantel 2020, PMID 31746972
Preterm birth, maternal AN (independent cohort) RR 1.32 (95% CI 1.13–1.55) 2,134,945 Quebec pregnancies 1989–2016; AN requiring hospitalization Ante 2020, PMID 32100355
Microcephaly RR 1.9 (95% CI 1.5–2.4) Swedish cohort Mantel 2020, PMID 31746972
Hyperemesis RR 2.1 (95% CI 1.8–2.5) Same Mantel 2020, PMID 31746972
Antepartum haemorrhage RR 1.6 (95% CI 1.2–2.1) Same; stronger in active than previous disease Mantel 2020, PMID 31746972
Stillbirth RR 1.99 (95% CI 1.20–3.30) Quebec cohort Ante 2020, PMID 32100355
Low birth weight RR 1.69 (95% CI 1.44–1.99) Same Ante 2020, PMID 32100355
Small for gestational age RR 1.52 (95% CI 1.35–1.72) Same Ante 2020, PMID 32100355

Economic burden

Statistic Estimate Population / method Source
Nationwide annual financial cost of eating disorders PPP-USD 70.5 billion Systematic review of 26 studies (11 cost-of-illness), Aug 2013–Jun 2024; all figures converted to 2024 USD PPP; PROSPERO CRD42022358136 Ahmed 2025, PMID 39542867
Share of financial cost that is indirect 70–93% Same Ahmed 2025, PMID 39542867
Intangible costs (burden of disease) PPP-USD 355.6 billion Same Ahmed 2025, PMID 39542867
Five-year median total cost per patient ~EUR 14,000–20,000 German statutory-insurance claims; 1,242 women and 71 men with incident AN, 1,104 women and 64 men with incident BN; 2 years before to 3 years after index diagnosis Bothe 2022, PMID 34599621
Share of AN cost attributable to mental illness ~two thirds (vs ~half in BN) Same Bothe 2022, PMID 34599621
Inpatient / partial-hospitalization charge $2,295 and $1,567 per day 314 consecutive adult first admissions, US programme, 2003–2015 Guarda 2017, PMID 28130794
Cost per pound of weight gained $4,089 (inpatient) and $7,050 (partial hospitalization); 70% reached discharge BMI ≥19 Same Guarda 2017, PMID 28130794
Incremental cost-effectiveness, FBT vs no intervention AUD 5,089 per DALY averted (95% UI dominant to 16,659); 100% likely cost-effective at AUD 50,000/DALY Markov model, ages 11–18, Australian health-system perspective, 6-year horizon Le 2017, PMID 29044637
Incremental cost-effectiveness, adolescent-focused therapy AUD 51,897 per DALY averted (21,591 to 1,712,491); 45% likely cost-effective Same Le 2017, PMID 29044637

Trial and registry landscape

Statistic Estimate Population / method Source
Registered interventional AN studies 337: 159 completed, 55 recruiting, 23 terminated, 22 not yet recruiting, 14 active-not-recruiting, 12 withdrawn, 8 enrolling by invitation, 44 unknown status ClinicalTrials.gov v2 count queries, condition "anorexia nervosa", study type interventional Live API query re-run by independent auditor, 2026-09-02
Phase 2/3 interventional AN records, all time 49 Same API, phase filter Live API query, 2026-09-02
Registered AN trials 2000–2018 reaching publication 201 registered → 101 completed → 41 published → 8 prospectively registered → 7 replicated Systematic audit of ClinicalTrials.gov listings Murray 2020, PMID 31638722
Adult AN psychotherapy randomized evidence base 14,003 reports screened → 16 trials → 13 in network → 1,047 patients (97.4% female) Network meta-analysis Solmi 2021, PMID 33600749
Second-generation antipsychotic RCTs in AN Olanzapine 7; quetiapine 2; risperidone 1; aripiprazole 0 Systematic scoping review, articles 2000–2022 Thorey 2023, PMID 36928656

Bone and fracture

Statistic Estimate Population / method Source
Fracture hospitalization rate, women 47.6 vs 21.0 per 10,000 person-years (AN vs control) Quebec matched cohort; 7,332 AN inpatients vs 73,215 controls; 1989–2023 Cyrenne-Dussault 2026, PMID 41109616
Fracture hospitalization rate, men 74.0 vs 39.9 per 10,000 person-years Same Cyrenne-Dussault 2026, PMID 41109616
Adjusted fracture hazard ratio Women 2.26 (2.02–2.51); men 1.86 (1.34–2.58) Same; Cox regression stratified by sex Cyrenne-Dussault 2026, PMID 41109616
Osteoporotic-fracture hazard ratio 7.50 (5.62–10.01) Same Cyrenne-Dussault 2026, PMID 41109616
Bone pharmacotherapy evidence base 19 studies, 1,119 participants, 10 double-blind RCTs Systematic review; bisphosphonates raised BMD in adult women, transdermal oestrogen in mature adolescents, oral contraceptives did not Robinson 2017, PMID 28554377
Teriparatide trial n=21 (10 vs 11), 6 months; no microarchitecture change; IT cortical thickness +13% (p=0.075) Randomized placebo-controlled; women with AN and T-score ≤ −2.5 Amorim 2026, PMID 41591404
rhIGF-1 added to transdermal oestradiol Negative: lumbar aBMD increased more without rhIGF-1 (p=0.004) 75 women aged 14–22 randomized, 33 completed; 12 months Singhal 2021, PMID 33693703

Treatment

Statistic Estimate Population / method Source
Parent-focused remission at end of treatment 43% n=107 randomized adolescents; 18 sessions/6 months Le Grange 2016, PMID 27453082
FBT remission at end of treatment 22% Same Le Grange 2016, PMID 27453082
Parent-focused vs FBT odds ratio 3.03 (95% CI 1.23–7.46) End-of-treatment remission Le Grange 2016, PMID 27453082
12-month remission, PFT vs FBT 37% vs 29%; OR 1.39 (95% CI 0.60–3.21) Difference not statistically significant Le Grange 2016, PMID 27453082
MOSAIC randomized 142 Adult broadly defined AN; MANTRA n=72, SSCM n=70 Schmidt 2015, PMID 25984803
ANTOP randomized 242 Adult female outpatients; 80 FPT, 80 CBT-E, 82 optimized TAU Zipfel 2014, PMID 24131861
ANTOP attrition, end of treatment 54/242 (22%) Multicentre adult RCT Zipfel 2014, PMID 24131861
ANTOP attrition, 12 months 73/242 (30%) Same Zipfel 2014, PMID 24131861
Refeeding RCT randomized 120 Age 12–24; AN or atypical AN; ≥60% median BMI Garber 2021, PMID 33074282
Higher-calorie start 2,000 kcal/day; +200/day Monitored inpatient arm Garber 2021, PMID 33074282
Lower-calorie start 1,400 kcal/day; +200 every other day Comparator Garber 2021, PMID 33074282
Acute pharmacotherapy RCTs 19 eligible; 8 meta-analyzed; 5 reporting BMI Systematic review; no significant olanzapine–placebo BMI difference Cassioli 2020, PMID 32448045
StRONG time to medical stability HR 1.67 (95% CI 1.10–2.53), p=0.01, favouring higher-calorie 111 in modified ITT of 120 enrolled Garber 2021, PMID 33074282
StRONG length of stay and cost 4.0 days shorter (95% CI −6.1 to −1.9); $19,056 saved (−$28,819 to −$9,293) Same Garber 2021, PMID 33074282
StRONG 1-year clinical remission No group difference (p=0.42) 111 mITT (60 higher-calorie, 51 lower); 12 months post-discharge Golden 2021, PMID 33753542
StRONG 1-year rehospitalization 32.8% (19/58) vs 35.4% (17/48), p=0.84 Same Golden 2021, PMID 33753542
StRONG atypical AN subgroup 43% of 111; heart-rate restoration 8.7±4.0 vs 6.5±3.9 days (p=0.008); weight gain 3.1±5.9 vs 5.4±2.9 %mBMI (p<0.001); hypomagnesaemia 29% vs 11% (OR 3.29) Atypical AN defined as %mBMI > 85 Garber 2024, PMID 38179719
StRONG caloric dose received 32.4±6.9 kcal/kg (atypical AN) vs 43.4±9.8 (AN) Same; per 10 kcal/kg, heart rate restored 1.7 days faster (1.0–2.5) Garber 2024, PMID 38179719
Refeeding-syndrome incidence in high-calorie reviews 1 true clinical case across 20 studies, 2,191 participants Systematic review of paediatric/adolescent refeeding, 2010–Feb 2023 Mosuka 2023, PMID 37351245
Olanzapine BMI trajectory 0.259 (SD 0.051) vs 0.095 (SD 0.053) kg/m² per month 152 adult outpatients (96% women, mean BMI 16.7), 5 North American sites, 16 weeks Attia 2019, PMID 30654643
Olanzapine obsessionality (Y-BOCS obsessions) −0.325 vs −0.017 points/month; not significant Same Attia 2019, PMID 30654643
Intranasal oxytocin phase II No significant difference vs placebo on ED psychopathology or BMI at any timepoint 61 female inpatients, 18 IU twice daily for 4 weeks, assessed to 6 months Maguire 2024, PMID 38520886
OPEN olanzapine feasibility (young people) 52 pre-screened → 35 eligible → 20 recruited → 15 continued ≥16 weeks Open-label, one-armed; target was 55; ages 12–24 Filiz 2026, PMID 42116676
Adult psychotherapy network 14,003 reports screened → 16 RCTs → 13 in network, 1,047 patients (97.4% female) None outperformed treatment as usual; CBT dropout lower than psychodynamic (OR 0.54, 0.31–0.93); confidence low to very low Solmi 2021, PMID 33600749
Family therapy vs treatment as usual (Cochrane) Remission RR 3.50 (1.49–8.23) from 2 studies, 81 participants; not maintained at follow-up 25 trials in review; low-quality evidence; 68% at high risk of selective reporting bias Fisher 2019, PMID 31041816
Family therapy vs other psychological interventions RR 1.22 (0.89–1.67), 5 studies, n=252; long-term RR 1.08 (0.91–1.28), 4 studies, n=200 Same Fisher 2019, PMID 31041816
ANDI day-patient non-inferiority Mean BMI difference 0.46 kg/m² favouring day patient (95% CI −0.11 to 1.02); p_non-inferiority<0.0001, margin 0.75 172 female adolescents aged 11–18, below 10th BMI percentile, first admission; randomized after 3 weeks inpatient Herpertz-Dahlmann 2014, PMID 24439238
DSM-5 severity specifiers vs outcome No significant difference across the four severity groups for weight recovery or good outcome 128 adult women (64 outpatient, 64 inpatient) treated with CBT-E; EOT, 6- and 12-month follow-up Dalle Grave 2018, PMID 30059831
TIARA rTMS, real vs sham BMI d=0.2 (−0.49 to 0.90); ED symptoms d=0.1 (−0.60 to 0.79); mood d=0.61–1.0 34 adults with severe/enduring AN, 20 sessions over 4 weeks; feasibility primary McClelland 2018, PMID 30012789
DBS evidence base, total 36 patients across 11 studies; mean age 38.07, mean BMI at surgery 12.58 kg/m² Patient-level network meta-analysis; subcallosal cingulate best-supported target (P-scores 0.94, 0.98) Shaffer 2023, PMID 36724522
Psilocybin phase 1 n=10 adult females, single 25-mg dose; no clinically significant ECG, vital-sign or suicidality change; 2 asymptomatic hypoglycaemias Open-label feasibility, single site; NCT04661514 Peck 2023, PMID 37488291
Treatment-refractory proportion, EDs 20–30% fail to respond to best available treatments Narrative review Bryson 2024, PMID 38503683
Adolescents needing intensive care One fifth to one third over the illness course Review of intensive treatment models Herpertz-Dahlmann 2021, PMID 33924294

Diagnostic-spectrum evidence

Statistic Estimate Population / method Source
Comparative atypical-AN publications 24 PRISMA review, literature from 2013 to 17 July 2022 Walsh 2023, PMID 36508318
Comparative atypical-AN publications, 2026 update 64 Invited update by the same group; atypical AN shows greater ED psychopathology, comparable non-ED psychopathology, and menstrual disturbance and reduced BMD at lower frequency Lee 2026, PMID 42557659
Atypical-AN course/outcome studies 0 identified Both reviews state that longitudinal course and outcome evidence is absent Walsh 2023, PMID 36508318; Lee 2026, PMID 42557659
SF-36 quality of life across EDs Significantly below population norms in every group; no difference established between diagnoses 7 studies (AN: 5 studies, n=227; BN: 4, n=216; EDNOS: 2, n=166; BED: 4, n=148) Winkler 2014, PMID 24857566
Qualitative studies in patient-meaning synthesis 24 International studies, 1990–2005 Espíndola 2009, PMID 19225241

Known conflicts and caveats

  • Mortality rate, SMR and cumulative mortality are different estimands.
  • The Solmi outcome percentages pool eating disorders and should not be relabelled AN-specific.
  • Recovery rises with follow-up partly because delayed recovery is observed; cohort composition also changes.
  • Trial remission definitions differ, so percentages are not directly rankable across studies.
  • BMI and percentage median BMI are not interchangeable, especially during growth.
  • Registry enrollment is planned for recruiting studies and actual for completed studies only if the record is updated; the clinical-trials table marks each value (E) or (A).
  • The 2011 mortality rate (5.1/1,000 person-years, AN-specific) and the 2024 transdiagnostic rate (5.2/1,000 person-years, all EDs pooled) are numerically close but are not the same estimand: different populations, different eras, different diagnostic mixes. Their agreement is suggestive, not confirmatory.
  • The three mortality meta-analyses (Arcelus 2011, Semchishen 2026, Lai 2026) share primary studies, so their agreement is not fully independent replication.
  • Heritability estimates of 50–60% and 33–84% come from different study sets and assumptions and are listed side by side rather than averaged.
  • Register-based hazard ratios for involuntary treatment measure the vulnerability of the exposed group, not the effect of compulsion; no comparator pathway exists.
  • MTAG-derived loci borrow statistical power from genetically correlated traits and should not be counted alongside case-control GWAS loci.
  • All bone-treatment evidence uses BMD, microarchitecture or bone-turnover surrogates; the fracture hazard ratios come from an observational cohort, not from any trial.
  • The four AN all-cause SMR estimates (Arcelus 5.86; Keshaviah 5.2; Krug 5.21; Lai 5.06) are close, but the same methodological re-analysis that left all-cause SMR nearly unchanged lowered the suicide SMR by 49%. Suicide-specific figures are far more model-dependent than all-cause figures and should always be quoted with their source (Keshaviah 2014, PMID 25214371).
  • Register-based heritability (36%) and twin-based heritability (50–60%) answer different questions on different ascertainment; the register figure is a lower bound because it counts only diagnosed cases.
  • The extreme-malnutrition complication prevalences come from a single national referral unit and describe the most severely ill tail of the illness, not AN generally.
  • Pandemic hospitalization increases are health-system measures. No population-incidence study has separated increased presentation from increased incidence.
  • Economic cost figures are dominated by indirect costs estimated with varying methods; about half the underlying cost-of-illness studies met 60% or more of a standard quality checklist.
  • ClinicalTrials.gov counts reflect records, not trials: registration practice varies by country and era, and mixed-diagnosis records appear under the AN condition query.
  • Maladaptive-exercise prevalence varies from 12% to 40% within the same 31,671-person sample depending on instrument; no single figure should be quoted without naming the measure.