Received: 6 August 2019 Revised: 21 April 2020 Accepted: 30 April 2020

DOI: 10.1002/erv.2745

REVIEW

Duration of untreated eating disorder and relationship

to outcomes: A systematic review of the literature

Amelia Austin1 | Michaela Flynn1 | Katie Richards1 | John Hodsoll2 |

Tiago Antunes Duarte3 | Paul Robinson4 | Jonathan Kelly5 | Ulrike Schmidt1,6

1
Section of Eating Disorder, Institute of
Psychiatry, Psychology and Neuroscience,
Abstract
King's College London, London, UK Objective: This systematic review assesses the average duration of untreated
2
Department of Biostatistics, Institute of eating disorder (DUED) in help-seeking populations at the time of first eating
Psychiatry, Psychology and Neuroscience,
disorder (ED) treatment and investigates the relationship between DUED and
King's College London, London, UK
3 symptom severity/clinical outcomes.
Hospital de Santa Maria, Centro
Hospitalar Universitário Lisboa Norte, Method: PRISMA guidelines were followed throughout. Selected studies pro-
EPE, Lisbon, Portugal vided information on either: (i) length of DUED, (ii) components of DUED,
4
Division of Medicine, University College (iii) cross-sectional associations between DUED and symptom severity,
London, London, UK
5 (iv) associations between DUED and clinical outcomes, or (v) experimental
Beat, Norwich, UK
6
South London and Maudsley NHS
manipulation of DUED. Study quality was assessed.
Foundation Trust, London, UK Results: Fourteen studies from seven countries were included. Across studies,
average DUED weighted by sample size was 29.9 months for anorexia nervosa,
Correspondence
Ulrike Schmidt, Section of Eating
53.0 months for bulimia nervosa and 67.4 months for binge eating disorder. A
Disorders, Institute of Psychiatry, younger age at time of first treatment was indicative of shorter DUED. Retro-
Psychology and Neuroscience, King's spective studies suggest that a shorter DUED may be related to a greater likeli-
College London, De Crespigny Park, PO
Box 59, London SE5 8AF, UK. hood of remission. Manipulation of DUED by shortening service-related
Email: ulrike.schmidt@kcl.ac.uk delays may improve clinical outcomes.
Conclusions: Data on length of DUED provide a benchmark for early inter-
vention in EDs. Preliminary evidence suggests DUED may be a modifiable fac-
tor influencing outcomes in EDs. To accurately determine the role of DUED,
definition and measurement must be uniformly operationalised.

KEYWORDS
anorexia nervosa, bulimia nervosa, duration of untreated illness, early intervention, eating
disorder

1 | INTRODUCTION cancer to cardiovascular disease. In relation to mental

health, these ideas have been most enthusiastically taken
The principles of early intervention and associated stage up by the psychosis field and have influenced research and
models of disease are widely accepted in medicine. Early policy. Early intervention services for psychosis are now
intervention has been defined as early detection followed ‘mainstream’ in many countries (McGorry & Mei, 2018).
by stage-specific or proportionate intervention, for as long These developments are underpinned by a solid evidence-
as necessary and effective (McGorry, Ratheesh, & base from randomised controlled trials, finding that early
O’Donoghue, 2018). This approach has led to improved sur- intervention is superior to treatment as usual (TAU; Correll
vival rates and better outcomes in many disorders, from et al., 2018). In this context, attempts to reduce the duration

Eur Eat Disorders Rev. 2020;1–17. wileyonlinelibrary.com/journal/erv © 2020 John Wiley & Sons, Ltd and Eating Disorders Association 1
2 AUSTIN ET AL.

of untreated psychosis (DUP), that is, the time from onset

of symptoms to the start of treatment, have been a key strat- Highlights
egy of early intervention programmes for promoting • This systematic review is the first to examine
favourable long-term outcomes of first episode psychosis duration of untreated eating disorder (DUED)
(Oliver et al., 2018; Penttilä, Jääskeläinen, Hirvonen, across different eating disorders. Definitions
Isohanni, & Miettunen, 2014; Sullivan et al., 2018). This is and measurement of DUED and its compo-
with the aim of preventing neuroprogression, that is, neuro- nents vary considerably between studies.
biological changes associated with psychosis symptoms • Across different eating disorders average DUED
which unfavourably affect the illness trajectory (Gama, weighted by sample size ranges from approxi-
Kunz, Magalhaes, & Kapczinski, 2013; Moylan, Maes, mately two and a half years (for anorexia nervosa)
Wray, & Berk, 2013). to nearly 6 years (for binge eating disorder).
Many of the arguments made in relation to psychosis • DUED appears to be related to age such that
also apply to eating disorders (EDs). As in psychosis, the younger patients have shorter DUED.
peak onset of EDs spans adolescence into emerging adult-
hood, about age 15–25 (Schmidt et al., 2016), with new EDs
rarely presenting after age 30 (Micali, Hagberg, Petersen, &
Treasure, 2013). It has been known for some time that lon- goes beyond documenting overall DUED and involves
ger illness duration is a key predictor of poor outcome in identifying specific sources of delay in different ED diag-
EDs (Steinhausen, 2002). In addition, and similar to psycho- noses and populations (e.g., adolescents or adults). Previ-
sis and bipolar disorder, converging data support the idea ously, DUP has been divided into three components
that neurobiological changes associated with disordered eat- (Birchwood et al., 2013): (i) a period where an individual
ing unfavourably affect the illness trajectory of EDs experiences symptoms, but does not recognise that they
(O’Hara, Campbell, & Schmidt, 2015; Steinglass & have a problem; (ii) a period where an individual recog-
Walsh, 2016). Provisional staging models for EDs have been nises that they have a problem, but are not yet ready to
developed, suggesting that illness stage at diagnosis is pre- seek help; and (iii) a period where an individual has
dictive of later outcomes (Maguire et al., 2017). Evidence sought help, and is waiting for treatment. Similar compo-
suggests that early stage EDs can be defined as less than nents could plausibly be applied to EDs.
3 years illness duration, beyond which the treatment The overall aim of this review is to better understand
response becomes more muted (Ambwani et al., 2020; Trea- DUED, its component delays and its impact on outcome
sure, Stein, & Maguire, 2015). A service model for early in different ED diagnoses and populations, and to provide
intervention in EDs has been developed and tested in the a benchmark for early intervention.
UK (Brown et al., 2018; Fukutomi et al., 2020; McClelland
et al., 2018; Schmidt, Brown, McClelland, Glennon, &
Mountford, 2016) and is included in national guidance as a 1.1 | Objectives
positive practice example [National Collaborating Centre
for Mental Health (NCCMH), 2019]. This review aims to answer the following questions:
Despite the parallels to psychosis and the availability What is the typical length of DUED for different ED
of a plausible rationale for early intervention, the ED diagnoses [anorexia nervosa (AN), bulimia nervosa (BN),
field has been slow to adopt early intervention binge eating disorder (BED), otherwise specified eating
approaches. Existing efforts have been piecemeal, and disorder (OSFED)], and populations (children, adoles-
hampered by a number of problems, with traditional ser- cents, adults, male, or female)?
vice designs, separating services for children and adoles- What are the different components of DUED?
cents from those for adults, being just one of them [see Is DUED cross-sectionally associated with symptom
Davey & McGorry, 2019 for discussion of these issues in severity?
relation to depression]. Is DUED associated with long-term clinical outcomes?
Nonetheless, national guidance is now increasingly Is experimental manipulation of DUED associated
supporting the need for early intervention approaches in with improved outcomes?
EDs [e.g., National Institute for Health and Care Excel-
lence (NICE), 2017]. However, if successful strategies for
early intervention for EDs are to be developed, a clear 2 | METHOD
understanding of duration of untreated ED (DUED),
pathways into care, and delays in accessing specialist A systematic review (i.e., an organised and transparent
treatments during a first episode ED, are necessary. This method of gathering, evaluating, and synthesizing data
AUSTIN ET AL. 3

applicable to specific research questions) was conducted. which 865 were duplicates. Based on the abstracts, 30 arti-
While a traditional meta-analysis (e.g., calculating a cles were eligible for full-text screening, and 13 of these
pooled measure of effect from multiple RCTs) was not were excluded as they either did not report on DUED or it
conducted, a meta-analytic approach, whereby quantita- was unclear whether study participants were experiencing a
tive results from multiple studies are combined into a first episode. We identified two previous systematic reviews
summary statistic, was used. on early intervention in AN (Schoemaker, 1997) and BN
PRISMA guidelines for conducting systematic reviews (Reas, Schoemaker, Zipfel, & Williamson, 2001). The
were applied (http://prisma-statement.org/). The search Schoemaker (1997) review used duration of illness as a
strategy was designed by two reviewers (M.F. and A.A.). proxy for ‘time between onset and first admission’, stating,
The protocol was registered with PROSPERO ‘duration of illness is the only indirect estimate available at
(CRD42018110884). Relevant literature was identified by this time’ (p. 2). None of the articles from this review met
searching PubMed, World of Science, and PsycINFO. We our inclusion criteria. Likewise, Reas et al. (2001) used
used the following search terms: (‘duration of untreated’ duration of illness as an approximation of DUED and thus
OR DUED OR ‘illness duration’ OR ‘duration of illness’ did not meet our criteria.
OR ‘early intervention’ OR ‘first episode’) AND (eating dis- Table 1 summarises the characteristics of the
order* OR anorexia nervosa OR bulimi* OR binge eat*). 14 included studies. These were from seven countries,
The terms were used to search all fields and no language that is, Australia, Canada, Germany, Republic of Ireland,
or publication date restrictions were imposed at this point. Singapore, Spain, and the UK. All studies reported DUED
Detailed information on search strategy, eligibility using a statistic of central tendency. Four studies
criteria, data extraction, quality assessment, and data syn- (n = 2,246) reported a component breakdown of DUED,
thesis are available in Supplementary Methods. The sea- (Beat, 2017; Brown et al., 2018; Gumz et al., 2018; Schlegl
rch was conducted from inception until December et al., 2019), two (n = 787) reported DUED and its cross-
17, 2019. Average DUED was calculated for each popula- sectional association with symptom severity (Bühren
tion group (i.e., diagnosis, age) in two ways: (i) A simple et al., 2013; Flynn et al., 2020), and one (n = 38) reported
mean weighted by sample size and (ii) meta-analytic esti- associations between DUED and long-term clinical out-
mates weighted by the inverse variance of the DUED. comes (Andrés-Pepiñá et al., 2019). Three studies
The meta-analytic approach allows statistically efficient (n = 721) attempted to experimentally manipulate DUED
95% CI intervals to be calculated for the pooled average (Brown et al., 2018; Flynn et al., 2020; Gumz et al., 2018),
incorporating both sampling and between study hetero- one of which (n = 142) also reported the prospective
geneity. Here, we use the IVHet approach which argu- associations between DUED and clinical outcomes
ably corrects the under-estimation of statistical error (McClelland et al., 2018).
which can result from a random effects model under con-
ditions of high heterogeneity (Doi, Barendregt, Khan,
Thalib, & Williams, 2015). We present both sample size 3.1.1 | Participants
weighted means in addition to meta-analytic means as
estimations of variance are subject to sampling error. The Overall, 5,032 patients were included in the selected
MetaXL plugin for Microsoft Excel was used which is studies. Information on patient diagnosis and age by
freely available to download and install from www. study can be found in Table 1 and Supplementary
epigear.com. As the Beat (2017) report was potentially Material.
less methodologically robust than the other studies
(i.e., used self-report data for illness onset, treatment
start, and diagnosis), used a broader definition of onset 3.1.2 | Methodological characteristics
than the other studies, and had the longest DUEDs across and quality
different EDs, a sensitivity analysis was run to explore it's
influence on the overall pooled estimate. Details on study characteristics can be found in Table 1
and details on methodological quality can be found in
Supplementary Material. Three studies (n = 275) mea-
3 | R E SUL T S sured the onset of EDs, and therefore start of DUED,
through a clinical interview using Diagnostic and Statisti-
3.1 | Characteristics of included studies cal Manual of Mental Disorders (American Psychiatric
Association, 2000, 2013) or International Classification of
The results of the study search are detailed in the PRISMA Diseases (World Health Organization, 1992) diagnostic
diagram in Figure 1. The search produced 1862 articles, of criteria (Gumz et al., 2018; Neubauer et al., 2014; Weigel
4 AUSTIN ET AL.

Records identified through database

FIGURE 1 Prisma flow diagram
Additional records identified
searching through other sources
Identification

n = 1859 n=3

- Pub Med n = 522

- PsychInfo n = 548
- World of Science n = 789

Records after duplicates removed

n = 997
Screening

Records screened Records excluded

n = 997 n = 967
Eligibility

Full-text articles assessed Full-text articles excluded,

for eligibility with reasons
n = 30 n =13

- Did not report DUED

n = 10
- Did not specify if first
episode
n=3
Included

Studies included in
qualitative synthesis
n = 14 (reported in n = 17
publications)

et al., 2014). Two studies (n = 644) used a clinical inter- 3.2 | Length of DUED
view plus an additional research assessment with an
adapted version of the eating disorder diagnostic scale 3.2.1 | Studies comparing DUED by
(Stice, Telch, & Rizvi, 2000) to confirm onset date (Brown diagnosis
et al., 2018; Flynn et al., 2020). Two studies (n = 2,027)
measured onset date using participant self-report Anorexia nervosa
(Beat, 2017; Schlegl et al., 2019), one (n = 285) reported Eleven studies explored the duration of untreated AN
that this was ‘assessed at admission’ (Bühren et al., 2013), (Andrés-Pepiñá et al., 2019; Beat, 2017; Bühren
three (n = 1,298) relied on young person and/or parental et al., 2013; Flynn et al., 2020; Gumz et al., 2018; Kwok
self-report (Kwok et al., 2019; Lieberman et al., 2019; Shu et al., 2019; Lieberman et al., 2019; Neubauer et al., 2014;
et al., 2015), and three (n = 503) did not define how onset Ng et al., 2018; Schlegl et al., 2019; Weigel et al., 2014).
was determined (Andrés-Pepiñá et al., 2019; Ng As shown in Figure 2, the average DUED in these studies
et al., 2018; Nicholls et al., 2011). To measure the start of ranged from 6.39 to 39.96 months, with a simple average
treatment, or end of DUED, six studies (n = 1,508) used of 29.9. Meta-analytic estimation found a mean DUED of
the date of entrance into specialised treatment (Andrés- 14.6 months (99% CI [5.1, 24.0]). Heterogeneity was high
Pepiñá et al., 2019; Brown et al., 2018; Bühren with I2 at 99% (95% CI [98.6, 99.1]. A sensitivity analysis
et al., 2013; Flynn et al., 2020; Kwok et al., 2019; was run excluding the Beat (2017) study (M = 11.4, 95%
Lieberman et al., 2019), five (n = 2,433) used CI [4.5, 18.3], I2 = 97.2%, 95% CI [96.2, 97.9]).
questionnaires (Beat, 2017; Neubaeur et al., 2014;
Nicholls et al., 2011; Schlegl et al., 2019; Weigel et al., Bulimia nervosa
2014), and three (n = 1,091) did not explicitly define how A total of four studies assessed DUED for BN (Beat, 2017;
start of treatment date was measured (Gumz et al., 2018; Flynn et al., 2020; Ng et al., 2018; Schlegl et al., 2019). The
Ng et al., 2018; Shu et al., 2015). average DUED ranged from 23.05 to 58.56 months, with a
TABLE 1 Study characteristics
Assessment Components Symptoms/clinical Other findings/
Author, location N Sample Design Measures of DUED of DUED Length DUED outcomes comments
AUSTIN ET AL.

Studies in AN
Andrés-Pepiñá 38 All females diagnosed Retrospective, Questionnaires: EDI-2, DUI: Time between N/A M = 13.05 months (SD Longer DUED was No other measured
et al., 2019 with AN between cohort BDI, ASQ illness onset and first = 9.80) significantly variables were
(Spain) 1987 and 1993 (age M Clinical interview: contact with services. Remission group M = associated with an significantly related to
= 14.4 years, SD = SCID-I Onset: Not described. 8.4 (SD = 8.4) increased risk of ED ED status at
1.6). At 22-year Current ED group M = status at 22-year follow-up.
follow-up mean age 18 (SD = 10.8) follow-up (OR = 3.3).
was 37.03 (SD = 4.01).
Bühren et al., 285 All females between age Cross-sectional Age-adjusted BMI scores ‘The time between N/A Local sample M = 10.8 DUED was not a Older patients were
2013 11 and 18 years old beginning of weight months (SD = 8.2) significant influence significantly more
(Germany) referred between 2001 loss and admission to Multi-site sample M = on age-adjusted BMI likely to have a longer
and 2009. All hospital’ (p. 396). 11.1 (SD = 8.6) at admission. DUED and lower
diagnosed with AN Included in both age-adjusted BMI at
(DSM-IV) with BMI samples M = 11.9 (SD admission than
below 10th percentile. = 8.8) younger patients.
Local sample-Aachen,
Germany n = 116
(age M = 15.2 years,
SD = 1.7), multisite
sample n = 127 (age
M = 15.1 years, SD =
1.5), included in both
samples n = 42 (age
M = 15.6 years, SD =
1.5).
Gumz et al., 2014; 77 Pre-intervention (n = Pre/post Questionnaires: EDE-Q DUI: Time between AN DUC: Duration until Pre-intervention M = N/A Intervention was a
Gumz, Weigel, 59) and post between-subjects or Ch-EDE-Q, PHQ-9, onset and initiation of first contact with 36.5 months (SD = systemic public health
Wegscheider, intervention (n = 18). intervention GAD-7, date of first ED-specific healthcare system for 68.2) intervention, no effect
Romer, & Löwe, All females between evaluation contact with guideline-based ED related issues. Post-intervention M = was found.
2018 (Germany) 10 and 60 years old healthcare. treatment. 40.1 (SD = 89.4) GP/paediatrician was
(M = 22.2, SD = 7.2). Clinical interview: Onset: Date when all typically the first
All diagnosed with SCID-I AN criteria were first healthcare
AN or atypical AN met. professional consulted
(DSM-IV). All about ED symptoms.
receiving first Study is limited by small
ED-specific treatment. sample size.
Kwok, Kwok, Lee, & 435 Female (n = 415) and Cross-sectional Retrospective chart Duration of illness prior N/A M = 33.60 months (SD Those with Those with
Tan, 2019 male (n = 20) review: to presentation = 34.32) childhood-onset had childhood-onset had
(Singapore) adolescents between Sociodemographic, Childhood-onset M = longer DUED than subsequently longer
age 13 and 18 years clinical 57.0 (SD = 64.68) those with adolescent inpatient stays and
old (M = 16.26, SD = characteristics, Adolescent onset M = onset when more admissions.
1.85). All diagnosed treatment details 31.44 (SD = 29.40) presenting at
with AN adolescent ED
(DSM-IV-TR) services.
between January 1,
2003 and December
31, 2014. Child
5

(Continues)
 6

TABLE 1 (Continued)
Assessment Components Symptoms/clinical Other findings/
Author, location N Sample Design Measures of DUED of DUED Length DUED outcomes comments
(<13 years) onset (n =
36) and adolescent
(13–18 years) onset (n
= 399).
Neubauer et al., 2014 140 All females between 10 Cross-sectional Questionnaires: EDE-Q DUI: Time between N/A M = 25.14 months (SD N/A Longer DUED was
(Germany) and 60 years old (M = or Ch-EDE-Q, PHQ-9, onset and = 36.76) associated with
17.51, SD = 5.81). FTQ) presentation to first Early onset M = 38.35 internal rather than
Early onset Clinical interview: ED-specific treatment. (SD = 45.92) external motivation to
(≤14 years) (n = 40), SCID-I Onset: Date when all Intermediate onset M = initiate treatment.
intermediate onset AN criteria were met 20.57 (SD = 31.97) GP/paediatrician was
(15–18 years) (n = simultaneously for the Late onset M = 19.04 typically the first to
53), late onset first time (or all (SD = 30.44) diagnose AN and
(>19 years) (n = 47). criterion except one provide ED-specific
All diagnosed with for subsyndromal AN) treatment
AN or subsyndromal information.
AN (DSM-IV or
ICD-10) with onset
1990 or later.
Weigel et al., 2014 58 All females between age Cross-sectional Questionnaires: FTQ, DUI: Time between date N/A M = 31.8 months (SD = Significant predictors of
(Germany) 10 and 60 years old PSSIK, of illness onset and 71.4) longer DUED:
(M = 22.3, SD = 7.8). sociodemographic date of first treatment Adolescents M = 14.1 Statutory health
Adolescents <18 years Clinical interview: reported in FTQ. (SD = 34.7) insurance, healthcare
old (n = 19), SCID-I Onset: Date of illness Emerging adults M = system-related factors
emerging adults onset reported in 20.7 (SD = 24.8) (e.g., waiting times),
18–25 years old (n = SCID-I. Adults M = 83.2 (SD = low insight into the
25), adults ≥26 years 122.9) disorder, low
old (n = 14). All self-motivation to
diagnosed with AN or initiate treatment,
atypical AN (DSM-IV higher paternal
or provisional education, having a
DSM-V). romantic partner,
separated parents,
immigrant
background,
one-point lower BMI,
and certain
personality
characteristics
(depressive,
rhapsodic, schizoid,
obsessive-
compulsive, paranoid,
and dependent).
Studies in mixed diagnostic groups
Beat, 2017 (United 1,821 Female (n = 1,741), Cross-sectional Questionnaire (online Time spent waiting for Time waiting for M = 39.24 months (SD N/A Overall, those with AN
Kingdom) male (n = 54), self-report) treatment. treatment = 1. Time = 33.39) had a shorter wait
non-binary (n = 18), between symptom
AUSTIN ET AL.
TABLE 1 (Continued)
Assessment Components Symptoms/clinical Other findings/
Author, location N Sample Design Measures of DUED of DUED Length DUED outcomes comments
AUSTIN ET AL.

and other gender onset and realising AN M = 34.54 (SD = time than those with
(n = 8). Rates of they had an ED, 2. 30.39) other diagnoses.
self-reported Between realising and BN M = 55.06 (SD = The longest component
diagnosis: AN n = seeking help, 3. 36.42) of DUED was the
1,330, BN n = 261, Between first GP visit BED M = 67.39 (SD = time between
BED n = 63, atypical and referral, 4. 39.70) symptom onset and
ED (i.e., OSFED/ Between referral and Atypical M = 42.40 (SD realisation of having
EDNOS) n = 154, assessment, 5. = 36.02) an ED.
unknown n = 13. Between assessment Children & adolescents:
Age: M = 20.77, SD = and start of treatment. M = 117.29 (SD =
8.43. Up to 18 years 95.92)
old (n = 831), 19+ Adults: M = 246.74 (SD
years old (n = 717), = 163.92)
age not given (n =
273). All participants
began treatment in
the UK between 2000
and 2017.
Brown et al., 2018; 142 Female (n = 139) and Historical Questionnaires: EDE-Q, DUED: Time from onset DUSC: Time from ED Audit group (n = 89): M Treatment group had DUED was successfully
Fukutomi et al., male (n = 3) aged controlled CORE-10, DASS-21, to evidence-based onset to assessment. = 19.09 months (SD = significant decrease in reduced using an
2020; McClelland 16–25 years old intervention WSAS, LEE, CIA treatment. 11.67) ED symptoms from 0 early intervention
et al., 2018 (United (control group: M = study Novel structured onset Onset: The time at Treatment group with to 12 months with model/care pathway.
Kingdom) 20.4, SD = 2.0, interview plus life which DSM-5 criteria minimal gate-keeping 70% below clinical
treatment: M = 20.4, chart (including items for an ED was first (n = 14): M = 13.04 cut-off by 12-months.
SD = 2.4). Diagnosed from the EDDS and met. (SD = 9.29) Between assessment and
ED (AN n = 57, BN n EDE) Treatment group with treatment, BMI
= 42, BED n = 5, complex gate keeping decreased in the audit
OSFED n = 38). All (n = 37): M = 17.66 group but increased
with DUED ≤3 years. (SD = 10.20) in the treatment
group.
Flynn et al., 2020 502 Female (n = 475) and Cross-sectional Questionnaires: EDE-Q, DUED: Time from onset DUSC: Time from onset TAU (n = 160): M = DUED was not related DUED was successfully
(United Kingdom) male (n = 27) aged CORE-10, DASS-21, to evidence-based to assessment. 19.98 months, SD = to BMI at assessment reduced using an
16–25 years old (M = WSAS, LEE, CIA treatment. 11.13, AN (n = 84) M for those diagnosed early intervention
20.56, SD = 2.35). All Clinical assessment Onset: The time at = 18.57, SD = 11.27, with AN model/care pathway.
diagnosed with ED Research interview which DSM-5 criteria BN (n = 42) M =
using criteria from (EDDS adapted for for an ED were first 23.05, SD = 9.35,
DSM-5 (AN n = 233, onset, life chart) met. OSFED (n = 29) M =
BN n = 131, BED n = BMI 19.90, SD = 12.64
9, OSFED n = 129)b Treatment (n = 272): M
= 17.85, SD = 10.38,
AN (n = 114) M =
17.50, SD = 10.62, BN
(n = 68) M = 20.26,
SD = 10.45, OSFED
(n = 82), M = 16.30,
SD = 9.84
Optimal treatmenta (n =
153): M = 15.96, SD =
7

(Continues)
 8

TABLE 1 (Continued)
Assessment Components Symptoms/clinical Other findings/
Author, location N Sample Design Measures of DUED of DUED Length DUED outcomes comments
9.74 AN (n = 56) M =
14.02, SD = 9.08, BN
(n = 47) M = 19.72,
SD = 10.76, OSFED
(n = 41) M = 14.05,
SD = 8.37
Lieberman, Houser, 106 Females (n = 88) and Cross-sectional Questionnaires: Duration of illness in N/A ARFID: M = Children with ARFID Patients with ARFID
Voyer, Grady, & males (n = 18) (ChEAT, EDI-C, months since onset 29.28 months (SD = had a significantly were significantly
Katzman, 2019 between the ages of 8 CDI-2, MASC-2) Onset: Symptom onset 40.6) longer DUED than younger and more
(Canada) and 13 years old (M = Clinical interview: AN: M = 6.39 months children with AN. likely to be male.
11.27, SD = 0.9). All Unspecified (SD = 4.7)
assessed between May
2013 and January
2017 and diagnosed
using DSM-V with
either AN (n = 77) or
ARFID (n = 29).
Nicholls, Lynn, & 208 Female (n = 171) and Prospective, cohort Questionnaires: DUI: Time between N/A M = 8.3 (SD = 7.0) N/A
Viner, 2011, male (n = 37) aged Study-specific illness onset and
(United Kingdom 5–12 years old (M = clinician presentation to
and Republic of 11.5, SD = 1.3). All questionnaire secondary care
Ireland) diagnosed with ED Onset: Not described
using criteria modified
from DSM-IV and
ICD-10 (AN n = 76,
BN n = 3, EDNOS n =
89 [including BED n =
6], ‘other’ ED n = 40).
Ng, Kuek, & Lee, 2018 257 Female (n = 238) and Cross-sectional Questionnaires: EDE-Q, Not described N/A Total M = 43.6 months Those with AN had
(Singapore) male (n = CIA, demographic (SD = 63.0) shorter DUED than
19) ≥ 12 years old (M info AN M = 26.9 (SD = those with BN or
= 20.52, SD = 7.14). 35.3) EDNOS.
Diagnosed ED using BN M = 57.8 (SD =
DSM-IV (AN n = 107, 57.4)
BN n = 76, EDNOS n EDNOS M = 53.0 (SD =
= 74). 89.2)
Schlegl et al., 2019 206 Females (n = 200) and Cross-sectional Specially created Treatment latency: From 1. Period from onset to M = 48.24 months, SD = Those with AN had Age of onset did not
(Germany) males (n = 6) questionnaire start to treatment diagnosis 64.32 approx. 1.5 years differ between AN
diagnosed using including 2. Period from onset to AN M = 39.96 (SD = shorter DUED than and BN.
ICD-10 with AN (n = demographics, clinical treatment 54.96) those with BN. A BN diagnosis, a higher
140) or BN (n = 66). characteristics, illness 3. Period from diagnosis BN M = 58.56 (SD = current age and not
All between 12 and course (by GP) to treatment 70.68) perceiving the eating
58 years old at disorder as a problem
diagnosis (AN M = were significant
24.28, SD = 7.74, BN predictors of a longer
M = 25.35, SD = DUED latency.
8.31). Recruited
AUSTIN ET AL.
TABLE 1 (Continued)
Assessment Components Symptoms/clinical Other findings/
Author, location N Sample Design Measures of DUED of DUED Length DUED outcomes comments
AUSTIN ET AL.

between August 2011

and September 2012.
Shu et al., 2015 757 Females (n = 704) and Cross-sectional EDE (adapted versions Self-reported in medical N/A Male DUED N/A No significant difference
(Australia) males (n = 53) under for child and parent assessment by parent median = 9.00 in DUED between
18 years old (male M informant) and child. months ± 5–12 genders.
= 13.57 years [SD = Female DUED median Males had significantly
2.02], female M = = 8.00 ± 5–12 earlier age of onset
14.74 years [SD = than females.
1.56]) diagnosed with
ED using DSM-V (AN
n = 286, atypical AN
n = 179, BN n = 71,
BN low frequency/
duration n = 10, BED
n = 3, BED low
frequency/duration n
= 5, purging disorder
n = 29, UFED n =
174.

Abbreviations: AN, anorexia nervosa; ASQ, autism spectrum quotient; BDI, beck depression inventory; BED, binge eating disorder; BN, bulimia nervosa; CDI-2, children's depression
inventory-2; ChEAT, children's eating attitudes test; Ch-EDE-Q, child eating disorder examination-questionnaire; CIA, clinical impairment assessment; CORE-10, ten-item version of clinical
outcomes in routine evaluation; DASS-21, depression, anxiety, and stress scale 21; DUED, duration of untreated eating disorder; DUI, duration of untreated illness; DUSC, duration of
untreated to specialist service contact; EAT, eating attitudes test; ED, eating disorder; EDDS, eating disorder diagnostic scale; EDE, eating disorder examination; EDE-Q, eating disorder
examination-questionnaire; EDI-2, eating disorder inventory-2; EDI-C, eating disorder inventory for children; EDNOS, eating disorder not otherwise specified; FTQ, first treatment question-
naire; GAD-7, generalized anxiety disorder scale; GP, general practitioner; LEE, level of expressed emotion; MAEDS, multiaxial assessment of eating disorder symptoms; MASC-2, multi-
dimensional anxiety scale for children-2; OR, odd ratio; OSFED, other specified feeding or eating disorder; PHQ-9, patient health questionnaire; PSSIK, personality style and disorder
inventory; SCID-IV, structured clinical interview for DSM-IV; WSAS, work and social adjustment scale.
a
Included as a subgroup of the previous ‘treatment’ group.
b
86 patients in the TAU condition were previously included in the study by Brown et al. (2018) and McClelland et al. (2018).
9
10 AUSTIN ET AL.

simple average of 53.0 months (see Supplementary (≤12 years old) or adolescents/adults (≥12 years old) at first
Figure 1). Meta-analytic estimation found a mean DUED of treatment are included in these figures. The categories of
34.3 months (95% CI [3.6, 65.0]). Heterogeneity was high adolescents and adults were collapsed, as several studies
with I2 at 98% (95% CI [97.3, 99.0]). A sensitivity analysis included participants both below and above 18 years of age.
was run excluding the Beat (2017) study (M = 26.6, 95% CI A simple mean DUED weighted by sample size was calcu-
[−16.1, 69.4], I2 = 96.7%, 95% CI [93.2, 98.4]). lated for children (9.8 months) and adolescents/adults
(34.7 months). Meta-analytic estimation found a mean of
Binge eating disorder 7.5 months for children (95% CI, [4.8, 10.2], I2 = 86.9%, 95%
Only one study analysed DUED for BED (Beat, 2017), CI [62.5, 95.4]), and 21.3 months (Supplementary Figure 4)
which revealed an average of 67.4 months (SD = 39.7). for adolescents and adults (95% CI, [12.3, 30.3], I2 = 96.0%,
95% CI [92.9, 97.2]). DUED appears to increase with age.
OSFED/EDNOS Two studies analysed DUED information for separate
Three studies assessed DUED for OSFED/EDNOS age groups at first treatment [child/adolescent vs. adults
(Beat, 2017; Flynn et al., 2020; Ng et al., 2018). As shown in (Beat, 2017) and adolescents vs. emerging adults vs. adults
Supplementary Figure 2, the average DUED ranged from (Weigel et al., 2014)]: again, DUED increase reflects an age
19.9 to 53.0 months with a simple average of 43.8. Meta- increase.
analytic estimation found a mean DUED of 29.5 months
(95% CI [7.5, 51.6.0]). Heterogeneity was high with I2 at
95% (95% CI [89.2, 97.9]). A sensitivity analysis was run 3.2.3 | Studies comparing gender
excluding the Beat (2017) study (M = 21.5, 95% CI [−20.1,
63.1], I2 = 89.6%, 95% CI [61.6, 97.2]). One study examined the role of gender in DUED (Shu
et al., 2015). Gender was not related to length of DUED
in their paediatric population.
3.2.2 | Studies comparing age at first
treatment
3.3 | Components of DUED
As shown in Supplementary Figures 3 and 4, average
DUED varies strongly between age groups. All studies While all studies measured the time between illness onset
reporting mean duration by age for either children and treatment, the component breakdown of this time

F I G U R E 2 Estimated DUED (in months) for anorexia using the inverse heterogeneity approach with point estimate for simple mean
weighted by sample size
Circle indicates the point estimate for DUED as calculated using a simple mean weighted by sample size (29.9 months).
†Bühren et al. (2013) analysed participants in three separate groups: a local sample in Aachen, Germany, a multisite sample for those outside
Aachen, and a third group for participants included in both samples. The corresponding author could not be reached to provide a combined analysis.
‡Denotes those in the treatment-as-usual (i.e., non-intervention) condition
AUSTIN ET AL. 11

varied (see Figure 3). Three studies explicitly identified com- realisation and seeking help; (iii) time between first GP visit
ponents of DUED. Brown et al. (2018) measured DUED as and referral; (iv) time between referral and assessment; and
the time of onset to the start of evidence-based treatment, (v) time between assessment and start of treatment.
defined as any treatment recommended by the NICE guide- As can be seen in Figure 3, DUED can be con-
lines (NICE, 2017). One component of this was the duration ceptualised in different ways but in all cases broadly
until specialised service contact, which represents the time includes patient-related delays (i.e., time before seeking
between illness onset and assessment. help) as well as service-related delays (i.e., time to
Gumz et al. (2014) defined DUED similarly but starting treatment after seeking help).
included duration until first contact with any healthcare
professional for eating related symptoms as a component.
Schlegl et al. (2019) took a related approach by again 3.4 | Cross-sectional association with
measuring DUED from onset to treatment but further symptom severity
breaking this down into two distinct periods: pre and post
general practitioner (GP) diagnosis. Bühren et al. (2013) investigated the role of DUED in
One study (Beat, 2017) deconstructed DUED most thor- children and adolescents with AN. These authors found
oughly, breaking it down into several components: (i) the that age-adjusted BMI was not significantly influenced by
time before realisation of being ill; (ii) the time between DUED. Flynn et al. (2020) investigated the role of DUED

Brown et al., 2018:

DUED

DUSC
Gumz et al., 2018:
DUI

F I G U R E 3 Different
conceptualisations of the putative
components of DUED

DUED, Duration of untreated

DUC
eating disorder (time from onset to
evidence-based treatment); DUSC, Schlegl et al., 2019:
duration until service contact (time Period from onset to treatment
from ED onset to assessment); DUI,
duration of untreated illness (time
from onset to initiation of ED-
specific evidence-based treatment);
DUC, duration until first contact
with healthcare system for ED
Period from onset to diagnosis (by GP) Period from diagnosis (by GP) to treatment
related issues (measured from
onset). A, time before realising they Beat, 2017:
had an eating disorder; B, time
Total wait-time
between realising they had an
eating disorder and seeking help; C,
time between first GP visit and
referral; D, time between referral
and assessment; and E, time
between assessment and start of
treatment A B C D E
12 AUSTIN ET AL.

in emerging adults with a range of ED diagnoses. Base- network of services providing early intervention) on
line analysis of participants diagnosed with AN suggests DUED in AN patients of all ages across a large metropoli-
that there was no significant relationship between DUED tan catchment area (Gumz et al., 2018). There was no sig-
and BMI at assessment. nificant change in DUED from before (M = 36.5 months,
SD = 68.2) to after (M = 40.1, SD = 89.4) the introduction
of the intervention. There was also no significant differ-
3.5 | Association with long-term clinical ence between BMI and EDE-Q scores for the before and
outcome after participant samples.

Andrés-Pepiñá et al. (2019) retrospectively investigated

the long-term outcome of patients with AN an average of 4 | DISCUSSION
22 years (range, 17–29) after initial diagnosis and treat-
ment. Patients who still had an ED diagnosis at follow-up 4.1 | Principal findings
were significantly more likely to have had a longer
DUED (M = 18 months, SD = 10.8) compared with those The first aim of this review was to determine the typical
in remission (M = 8.4 months, SD = 8.4) with an odds length of DUED for different ED populations. Studies
ratio of 3.33 (95% CI: 1.3–8.7; p < .014). from seven countries indicate that, across different disor-
ders, DUEDs are lengthy. Simple average DUED
weighted by sample size ranged from 29.9 months for AN
3.6 | Experimental manipulation to 38.9 months for BN, with OSFED in between. A meta-
of DUED analytic approach accounting for heterogeneity estimated
a lower mean DUED, giving 14.6 months for AN,
Using a pre-post design, a quasi-experimental study from 34.3 months for BN, and 29.5 months for OSFED. The
the UK compared DUED following the introduction of large difference between the mean weighted by sample
an early intervention service model/care pathway for size and the meta-analytic estimate weighted by inverse
emerging adults with recent ED onset (FREED: First epi- variance for AN is due to low variance in samples of
sode rapid early intervention for EDs) with that of similar younger patients. The meta-analytic estimate for AN
patients (age, illness duration, diagnosis) previously hav- highly weighted smaller studies with children, and youn-
ing received treatment as usual (TAU) in the same ser- ger age is a confounding factor for shorter DUED, thus
vice (Brown et al., 2018; McClelland et al., 2018). For lowering overall estimates for DUED. However, the sim-
details of the FREED model see Allen et al. (2020) and ple average mean weighted by sample size did not
Supplementary Table 2. For patients able to access devalue studies based on large variance, therefore giving
FREED under optimal conditions (i.e., without gate- just as much weight to outliers, and likely being more
keeping barriers) DUED was 6 months shorter than for reflective of clinical reality.
TAU patients. Patients offered FREED took up treatment The lengthy time periods as measured by simple aver-
significantly more than those offered TAU (100% vs. 73%) age DUED found here appear longer than that found in a
and had significant improvement in ED and other symp- recent care pathway study from Italy (Volpe et al., 2019).
toms over time. FREED participants diagnosed with AN This study found that this took approximately 26 months
on average gained weight between clinical assessment across disorders. These patients did not necessarily have a
and start of treatment whereas the comparison group lost first episode of their ED and it is likely that during a later
weight in the same interval. By 12 months, 59% of the illness episode patients may ask for help more quickly.
FREED AN group had returned to normal BMI com- Our findings, both within study comparisons
pared with only 17% of the TAU group. These improve- (Beat, 2017; Weigel et al., 2014) and cross study compari-
ments were maintained up to 24 months after start of sons (Supplementary Figures 3 and 4), suggest that there
treatment (Fukutomi et al., 2020). A second ongoing is variability in DUED relating to age, with a younger age
multi-centre study using a similar design found compara- relating to a shorter DUED. This age-related increase in
ble reductions in DUED between participants receiving length of DUED is likely to be related to several factors.
FREED under optimal conditions and those assigned to First, the younger the child the more likely the parents
TAU (Flynn et al., 2020). are able to identify the illness earlier. Longer DUED as
A study from Germany also used a pre-post design, to age increases may be explained by living independently,
evaluate the impact of a multi-faceted public health inter- leaving the individual in charge of seeking help and navi-
vention (including an awareness raising campaign, gating the health care system. In addition, despite the fact
school-based prevention, treatment resources, and a that peak onset of EDs ranges from age 15 to 25 (Schmidt,
AUSTIN ET AL. 13

Adan, et al., 2016), many countries have separate child/ In relation to our third aim, overall, studies failed to
adolescent and adult ED services, which may add to find cross-sectional associations between BMI at the com-
delays and disruptions in accessing first episode specialist mencement of treatment and length of DUED (Bühren
mental health care. Second, studies of younger partici- et al., 2013; Flynn et al., 2020). This may be explained by
pants tend to mainly include AN, which is a highly visi- the limited variability in DUED in these two studies.
ble disorder, whereas studies of adults often include a However, the Bühren et al. (2013) study found that older
mixture of AN and those with bulimic EDs, which are adolescents had a longer DUED and lower age-adjusted
more hidden. For example, parents are often unaware of BMI at admission than younger adolescents, which the
bulimic symptoms in their adolescent children authors attributed to a lessening of parental influence on
(Bartholdy et al., 2017). Third, it may also matter who older teens.
reports on DUED. Where parents report DUED, a self- The fourth aim of the review was to investigate the
serving bias may be operative, that is, with parents not relationships of DUED and long-term clinical outcomes.
wishing to admit they left symptoms unchallenged for a Andrés-Pepiñá et al.'s (2019) retrospective study suggests
period of time. Conversely, where DUED is defined by that a longer DUED may play a role in persistence of AN
patients, an ‘effort after meaning’ bias may mean that many years after initial treatment. No other studies
people date the onset of their symptoms back to mild assessed the influence of DUED on long-term clinical
body image concerns. outcomes, and thus these findings cannot be generalised
Average DUED weighted by sample size found here to the wider group of patients with EDs, although they
for children was 9.8 months (see Figure 3). DUEDs for do bolster the rationale for early intervention.
adolescents and adults (M = 34.7) were longer than dura- The final aim was to investigate experimental manip-
tion of illness in recent large-scale clinical trials in ado- ulations of DUED. Three studies, all using pre-post
lescents with EDs [e.g., AN: Agras et al., 2014 designs, attempted this, one through an ambitious public
13.5 months; Hodsoll et al., 2017 median 12–15 months; health intervention (Gumz et al., 2018). The other two
Eisler et al., 2016 9.6–11.4 months; Herpertz-Dahlmann- attempted to reduce DUED through a novel service inter-
et al., 2014 9.8–12.4 months; BN: Le Grange, Lock, Agras, vention (FREED) designed to reduce service related
Bryson, & Jo, 2015 18.4–19.6 months; Schmidt et al., 2007 delays in specialist ED services in the UK (Flynn
2.5–2.6 years] and shorter than in trials in adults et al., 2020; McClelland et al., 2018). For details of the
[e.g., AN: Attia et al., 2019 10.5–12.6 years; Schmidt FREED model, see Allen et al. (2020) and Supplementary
et al., 2015 8.3 years; BN/EDNOS/BED: Fairburn Table 2. The public health intervention did not reduce
et al., 2009 9.9 years; BED: de Zwaan et al., 2017 7.9– DUED, whereas the novel service intervention did reduce
10.4 years]. While this is certainly due in part to the con- DUED by several months. Clinical outcomes (weight
flation of the average DUED for adolescents and adults, recovery) for FREED patients with AN were much better
there is still another factor: DUED measures time to first than for those receiving TAU with differences in rate of
treatment whereas duration of illness measures time to improvement maintained up to 24 months (Fukutomi
current treatment, including any previous treatments. et al., 2020). This evidence suggests that FREED is a
For the majority of adolescents, this likely constitutes promising early intervention model for reducing DUED
their first ever treatment. across all EDs, and for improving clinical outcomes in
The second aim of this review was to delineate com- AN. Its impact on clinical outcomes in other EDs is yet to
ponents of DUED. The evidence suggests that the largest be demonstrated. By contrast, efforts to intervene with a
delays are patient-related (i.e., from start of illness to prominent focus on prevention of onset of AN and/or
help-seeking; see Figure 3). Likewise, in the Italian care raising awareness about early help seeking may not be
pathway study by Volpe et al. (2019), the larger compo- enough to reduce DUED, as indicated by the disappoint-
nent delay was prior to starting help-seeking. Nonethe- ing findings of Gumz et al. (2018). Similarly, indicated
less, the time between help-seeking and accessing prevention efforts focusing exclusively on AN have also
specialist care was substantial (28 weeks). had disappointing results (Jacobi et al., 2018).
These findings have implications for early interven-
tion programmes. Waiting for weeks or months from
the point of help-seeking is distressing. As such, reduc- 4.2 | Strengths and limitations
ing service-related delays is important. These efforts
need to be joined with attempts to intervene earlier, for A strength of this review is that it assesses DUED across
example, through indicated prevention in high-risk different EDs, and as such provides a benchmark for
groups. future research, clinical practice, and health policy. The
14 AUSTIN ET AL.

data included were from a range of countries with differ- developed an abbreviated assessment of DUED for clini-
ent health care systems, yet findings seemed to be consis- cal practice. Any such assessment tools should also try to
tent, and thus appear generalisable across high-income delineate the components of DUED, as this would inform
Western countries. decisions about when it is best to intervene.
This review also has several limitations. First, the sea- In terms of clinical practice and policy, long DUEDs
rch excluded articles not written in English, Portuguese, or across different countries with different health care systems
German and many types of grey literature. Second, DUED suggest that we are nowhere near achieving early interven-
was not operationalised in the same way across studies. tion. In psychosis, early intervention efforts have been
Studies differed markedly in their definitions of illness organised around shortening DUP as a key outcome. This
onset and treatment start. A recent systematic review on requires routine measurement of this variable. Likewise, it
the duration of untreated psychosis (DUP) cited similar would be helpful to routinely measure DUED.
difficulties with heterogeneous definitions of DUP (Oliver The longest component of DUED is the time before
et al., 2018). This suggests that other mental health fields, seeking help. Measures to improve early detection of
even those in a more advanced stage of research on dura- EDs may help to shorten this period. A second key
tion of untreated illness, are facing similar problems. Like- component of DUED is the time people wait between
wise, the components of DUED have been conceptualised seeking help (e.g., in the UK an appointment with their
differently by different authors (see Figure 3). For exam- GP) and starting specialist treatment. In England, there
ple, Pinhas, Wong, and Woodside (2014) have segmented are nationally binding waiting times targets for EDs in
DUED into several components for both mental and phys- young people below age 18 (Department of
ical health pathways and have also taken into consider- Health, 2014). In parallel, self-referrals to specialist ser-
ation the role of duration of untreated ED to first vices are now allowed for under 18s (NCCMH, 2015).
psychotropic medication (DUPMed) as being a relevant These measures have successfully increased the pro-
period of time, but do not provide any data. portion of under 18s with EDs starting specialist treat-
Third, we were not able to separate out the influence ment (NHS England, 2019).
of age at presentation and diagnosis, given very limited It is expected that similar waiting time targets will be
data on children and adolescents with bulimic EDs. brought in for adults in the UK (NCCMH, 2019). These
Finally, the variable of DUED may be confounded by are important steps for shortening DUED. However,
other factors. Research in psychosis lists the following measurement of waiting times alone is not sufficient, as a
potential confounders: mode-of-onset, pre-morbid func- substantial proportion of young people are referred
tioning, and acuteness of illness at assessment (Sullivan between services (e.g., child to adult services) without
et al., 2018). These variables may also affect DUED. ever starting specialist treatment.
Emerging data suggest that a service model/care path-
way, such as FREED, can successfully reduce DUED,
4.3 | Implications for research, practice, improve clinical outcomes, and appears to be cost-effec-
and policy tive (Brown et al., 2018; McClelland et al., 2018).

Future studies should aim to adopt a common definition

of DUED, including its components, and standardised 4.4 | Conclusion
measurement tools. For research purposes it is desirable
to use in-depth interview measures allowing retrospective This systematic review is the first to examine DUED
assessment of symptoms, anchored in key autobiographi- across different EDs. Whilst definitions and measure-
cal events. This methodology has been tested across ment of DUED and its components vary between stud-
many different disorders (e.g., depression: Brown, ies, there is agreement across different studies
Adler, & Bifulco, 1988; psychosis: Bebbington et al., 1993; internationally that average DUEDs are long, ranging
EDs: Schmidt, Tiller, Blanchard, Andrews, & from about two and a half years (for AN) to nearly
Treasure, 1997). These studies show that, given biograph- 6 years (for BED). DUED appears to be shorter in chil-
ical anchoring, such details can be accurately remem- dren than in adolescents and adults. Preliminary evi-
bered going back many years. We used an adapted dence suggests DUED may be a modifiable factor
version of this methodology in our studies, using a graph effecting outcome in EDs. Taken together, these data
to create a timeline incorporating patient life events and provide a useful benchmark for early intervention
details of the illness course (Brown et al., 2018; Flynn efforts in EDs. To accurately determine the role of
et al., 2020). However, this methodology is time consum- DUED, definition and measurement must be uniformly
ing and less useful for routine clinical practice. We have operationalised.
AUSTIN ET AL. 15

ACK NO WLE DGE MEN TS Bartholdy, S., Allen, K., Hodsoll, J., O’Daly, O. G., Campbell, I. C.,
US receives salary support from the National Institute for Banaschewski, T., … Schmidt, U. (2017). Identifying disordered
eating behaviours in adolescents: How do parent and adolescent
Health Research (NIHR) Biomedical Research Centre for
reports differ by sex and age? European Child & Adolescent Psychi-
Mental Health, South London and Maudsley NHS Foun- atry, 26(6), 691–701. https://doi.org/10.1007/s00787-016-0935-1
dation Trust and Institute of Psychiatry, Psychology and Beat. (2017). Delaying for years, denied for months. Norwich, UK: Beat.
Neuroscience, King's College London. US is supported by Retrieved from https://www.beateatingdisorders.org.uk/uploads/
an NIHR Senior Investigator Award. The views expressed documents/2017/11/delaying-for-years-denied-for-months.pdf
in this publication are those of the authors and not neces- Bebbington, P., Wilkins, S., Jones, P., Foerster, A., Murray, R.,
sarily those of the National Health Service, the NIHR or Toone, B., & Lewis, S. (1993). Life events and psychosis: Initial
the UK Department of Health. KR is funded by the results from the Camberwell collaborative psychosis study. The
British Journal of Psychiatry, 162(1), 72–79. https://doi.org/10.
Health Foundation, and MF and AA are supported by
1192/bjp.162.1.72
the King's College London International Postgraduate Birchwood, M., Connor, C., Lester, H., Patterson, P.,
Research Scholarship. Freemantle, N., Marshall, M., … Singh, S. P. (2013). Reducing
duration of untreated psychosis: Care pathways to early inter-
CONFLICTS OF INTEREST vention in psychosis services. The British Journal of Psychiatry,
None. 203(1), 58–64. https://doi.org/10.1192/bjp.bp.112.125500
Brown, A., McClelland, J., Boysen, E., Mountford, V.,
Glennon, D., & Schmidt, U. (2018). The FREED project (first
ORCID
episode and rapid early intervention in eating disorders): Ser-
Amelia Austin https://orcid.org/0000-0002-4979-4847
vice model, feasibility and acceptability. Early Intervention in
Michaela Flynn https://orcid.org/0000-0003-0208-1492 Psychiatry, 12(2), 250–257. https://doi.org/10.1111/eip.12382
Ulrike Schmidt https://orcid.org/0000-0003-1335-1937 Brown, G. W., Adler, Z., & Bifulco, A. (1988). Life events and
chronic depression. The British Journal of Psychiatry, 152(4),
R EF E RE N C E S 487–498. https://doi.org/10.1192/bjp.152.4.487
Agras, W. S., Lock, J., Brandt, H., Bryson, S. W., Dodge, E., Bühren, K., von Ribbeck, L., Schwarte, R., Egberts, K., Pfeiffer, E.,
Halmi, K. A., … Woodside, B. (2014). Comparison of 2 family Fleischhaker, C., … Herpertz-Dahlmann, B. (2013). Body mass
therapies for adolescent anorexia nervosa: A randomized paral- index in adolescent anorexia nervosa patients in relation to age,
lel trial. JAMA Psychiatry, 71(11), 1279–1286. https://doi.org/ time point and site of admission. European Child & Adolescent Psy-
10.1001/jamapsychiatry.2014.1025 chiatry, 22(7), 395–400. https://doi.org/10.1007/s00787-013-0376-z
Allen, K. L., Mountford, V., Brown, A., Richards, K., Grant, N., Correll, C. U., Galling, B., Pawar, A., Krivko, A., Bonetto, C.,
Austin, A., … Schmidt, U. (2020). First episode rapid early inter- Ruggeri, M., … Kane, J. M. (2018). Comparison of early inter-
vention for eating disorders (FREED): From research to routine vention services vs treatment as usual for early-phase psychosis:
clinical practice. Early Intervention in Psychiatry. Advanced A systematic review, meta-analysis, and meta-regression. JAMA
Online Publication. DOI. https://doi.org/10.1111/eip.12941 Psychiatry, 75(6), 555–565. https://doi.org/10.1001/
Ambwani, S., Cardi, V., Albano, G., Cao, L., Crosby, R., jamapsychiatry.2018.0623
MacDonald, P., … Treasure, J. (2020). A multicenter audit of Davey, C. G., & McGorry, P. D. (2019). Early intervention for
outpatient care for adult anorexia nervosa: Symptom trajectory, depression in young people: A blind spot in mental health care.
service use, and evidence in support of “early stage” versus Lancet Psychiatry, 6(3), 267–272. https://doi.org/10.1016/S2215-
“severe and enduring” classification. International Journal of 0366(18)30292-X
Eating Disorders. https://doi.org/10.1002/eat.23246 de Zwaan, M., Herpertz, S., Zipfel, S., Svaldi, J., Friederich, H. C.,
American Psychiatric Association (2000). Diagnostic and statistical Schmidt, F., … Hilbert, A. (2017). Effect of internet-based
manual of mental disorders (4th, Text Revision). Washington, guided self-help vs individual face-to-face treatment on full or
DC: American Psychiatric Association. Subsyndromal binge eating disorder in overweight or obese
American Psychiatric Association. (2013). Diagnostic and statistical patients: The INTERBED randomized clinical trial. JAMA Psy-
manual of mental disorders (5th ed.). Washington, DC: Ameri- chiatry, 74(10), 987–995. https://doi.org/10.1001/
can Psychiatric Publishing. jamapsychiatry.2017.2150
Andrés-Pepiñá, S., Plana, M. T., Flamarique, I., Romero, S., Department of Health. (2014). Achieving better access to mental
Borràs, R., Julià, L., … Castro-Fornieles, J. (2019). Long-term health services by 2020. Retrieved from https://www.gov.uk/
outcome and psychiatric comorbidity of adolescent-onset government/publications/mental-health-services-achieving-
anorexia nervosa. Clinical Child Psychology and Psychiatry, better-access-by-2020
1359104519827629, 33–44. https://doi.org/10.1177/ Doi, S. A. R., Barendregt, J. J., Khan, S., Thalib, L., &
1359104519827629 Williams, G. M. (2015). Advances in the meta-analysis of het-
Attia, E., Steinglass, J. E., Walsh, B. T., Wang, Y., Wu, P., erogeneous clinical trials I: The inverse variance heterogeneity
Schreyer, C., … Marcus, M. D. (2019). Olanzapine versus pla- model. Contemporary Clinical Trials, 45, 130–138. https://doi.
cebo in adult outpatients with anorexia nervosa: A randomized org/10.1016/j.cct.2015.05.009
clinical trial. American Journal of Psychiatry, Appi-Ajp, 176, Eisler, I., Simic, M., Hodsoll, J., Asen, E., Berelowitz, M.,
449–456. https://doi.org/10.1176/appi.ajp.2018.18101125 Connan, F., … Landau, S. (2016). A pragmatic randomised
16 AUSTIN ET AL.

multi-Centre trial of multifamily and single family therapy for cognitive-behavioral therapy for adolescent bulimia nervosa.
adolescent anorexia nervosa. BMC Psychiatry, 16(1), 422. Journal of the American Academy of Child & Adolescent Psychi-
https://doi.org/10.1186/s12888-016-1129-6 atry, 54(11), 886–894. https://doi.org/10.1016/j.jaac.2015.08.008
Fairburn, C. G., Cooper, Z., Doll, H. A., O’Connor, M. E., Bohn, K., Lieberman, M., Houser, M. E., Voyer, A. P., Grady, S., &
Hawker, D. M., … Palmer, R. L. (2009). Transdiagnostic Katzman, D. K. (2019). Children with avoidant/restrictive food
cognitive-behavioral therapy for patients with eating disorders: intake disorder and anorexia nervosa in a tertiary care pediatric
A two-site trial with 60-weekfollow-up. American Journal of eating disorder program: A comparative study. International
Psychiatry, 166(3), 311–319. https://doi.org/10.1176/appi.ajp. Journal of Eating Disorders, 52(3), 239–245.
2008.08040608 Maguire, S., Surgenor, L. J., Le Grange, D., Lacey, H., Crosby, R. D.,
Flynn, M., Austin, A., Allen, K., Grant, N., Mountford, V., Engel, S. G., … Touyz, S. (2017). Examining a staging model for
Glennon, D., …, & Schmidt, U. (2020). Does introduction of a anorexia nervosa: Empirical exploration of a four stage model
novel early intervention service reduce duration of untreated eat- of severity. Journal of Eating Disorders, 5(41), 41. https://doi.
ing disorder in emerging adults with first episode illness? Manu- org/10.1186/s40337-017-0155-1
script in preparation. McClelland, J., Hodsoll, J., Brown, A., Lang, K., Boysen, E.,
Fukutomi, A., Austin, A., McClelland, J., Brown, A., Mountford, V., Flynn, M., … Schmidt, U. (2018). A pilot evaluation of a novel
Grant, N., … Schmidt, U. (2020). First episode rapid early interven- first episode and rapid early intervention service for eating dis-
tion for eating disorders: A two-yearfollow-up. Early Intervention orders (FREED). European Eating Disorder Review, 26(2),
in Psychiatry, 14(1), 137–141. https://doi.org/10.1111/eip.12881 129–140. https://doi.org/10.1002/erv.2579
Gama, C. S., Kunz, M., Magalhaes, P. V., & Kapczinski, F. (2013). McGorry, P. D., & Mei, C. (2018). Early intervention in youth men-
Staging and neuroprogression in bipolar disorder: A systematic tal health: Progress and future directions. Evidence Based Men-
review of the literature. Revista Brasileira de Psiquiatria, 35(1), tal Health, 21(4), 182–184. https://doi.org/10.1136/ebmental-
70–74. https://doi.org/10.1016/j.rbp.2012.09.001 2018-300060
Gumz, A., Uhlenbusch, N., Weigel, A., Wegscheider, K., McGorry, P. D., Ratheesh, A., & O’Donoghue, B. (2018). Early
Romer, G., & Löwe, B. (2014). Decreasing the duration of intervention-an implementation challenge for 21st century
untreated illness for individuals with anorexia nervosa: Study mental health care. JAMA Psychiatry, 75(6), 545–546. https://
protocol of the evaluation of a systemic public health interven- doi.org/10.1001/jamapsychiatry.2018.0621
tion at community level. BMC Psychiatry, 14(1), 300. https:// Micali, N., Hagberg, K. W., Petersen, I., & Treasure, J. L. (2013).
doi.org/10.1186/s12888-014-0300-1 The incidence of eating disorders in the UKin 2000–2009: Find-
Gumz, A., Weigel, A., Wegscheider, K., Romer, G., & Löwe, B. ings from the general practice research database. BMJ Open, 3
(2018). The psychenet public health intervention for anorexia (5), e002646. https://doi.org/10.1136/bmjopen-2013-002646
nervosa: A pre-post-evaluation study in a female patient sam- Moylan, S., Maes, M., Wray, N. R., & Berk, M. (2013). The neu-
ple. Primary Health Care Research & Development, 19(1), 42–52. roprogressive nature of major depressive disorder: Pathways to
https://doi.org/10.1017/S1463423617000524 disease evolution and resistance, and therapeutic implications.
Herpertz-Dahlmann, B., Schwarte, R., Krei, M., Egberts, K., Molecular Psychiatry, 18(5), 595–606. https://doi.org/10.1038/
Warnke, A., Wewetzer, C., … Dempfle, A. (2014). Day-patient mp.2012.33
treatment after short inpatient care versus continued inpatient National Collaborating Centre for Mental Health. (2015). Access
treatment in adolescents with anorexia nervosa (ANDI): A mul- and waiting time standard for children and young people with
ticentre, randomised, open-label, non-inferiority trial. The Lan- an eating disorder. Retrieved from https://www.england.nhs.
cet, 383(9924), 1222–1229. https://doi.org/10.1016/S0140-6736 uk/wp-content/uploads/2015/07/cyp-eating-disorders-access-
(13)62411-3 waiting-time-standard-comm-guid.pdf
Hodsoll, J., Rhind, C., Micali, N., Hibbs, R., Goddard, E., National Collaborating Centre for Mental Health. (2019). Adult eat-
Nazar, B. P., … Treasure, J. (2017). A pilot, multicentre prag- ing disorders: Community, inpatient and intensive day patient
matic randomised trial to explore the impact of carer skills care, London, England: Guidance for commissioners and
training on carer and patient behaviours: Testing the cognitive providers.
interpersonal model in adolescent anorexia nervosa. European National Institute for Health and Care Excellence (2017). Eat-
Eating Disorders Review, 25(6), 551–561. https://doi.org/10. ing disorders: Recognition and treatment. NICE Guideline
1002/erv.2540 NG 69. Retrieved from https://www.nice.org.uk/guidance/
Jacobi, C., Hütter, K., Völker, U., Möbius, K., Richter, R., ng69
Trockel, M., … Taylor, C. B. (2018). Efficacy of a parent-based, Neubauer, K., Weigel, A., Daubmann, A., Wendt, H., Rossi, M.,
indicated prevention for anorexia nervosa: Randomized con- Löwe, B., & Gumz, A. (2014). Paths to first treatment and dura-
trolled trial. Journal of Medical Internet Research, 20(12), e296. tion of untreated illness in anorexia nervosa: Are there differ-
https://doi.org/10.2196/jmir.9464 ences according to age of onset? European Eating Disorders
Kwok, C., Kwok, V., Lee, H. Y., & Tan, S. M. (2019). Clinical and Review, 22(4), 292–298. https://doi.org/10.1002/erv.2300
socio-demographic features in childhood vs adolescent-onset Ng, K. W., Kuek, A., & Lee, H. Y. (2018). Eating psychopathology
anorexia nervosa in an Asian population. In Eating and weight and psychosocial impairment in patients treated at a Singapore
disorders-studies on anorexia, bulimia and obesity (pp. 1–6). eating disorders treatment programme. Singapore Medical Jour-
New York, NY: Springer. nal, 59(1), 33. doi:10.11622/smedj.2017042
Le Grange, D., Lock, J., Agras, W. S., Bryson, S. W., & Jo, B. (2015). NHS England. (2019). Children and young people with an eating
Randomized clinical trial of family-based treatment and disorder waiting times. Retrieved from https://www.
AUSTIN ET AL. 17

england.nhs.uk/statistics/statistical-work-areas/cyped- Schmidt, U., Tiller, J., Blanchard, M., Andrews, B., & Treasure, J.
waiting-times/ (1997). Is there a specific trauma precipitating anorexia
Nicholls, D. E., Lynn, R., & Viner, R. M. (2011). Childhood eating nervosa? Psychological Medicine, 27(3), 523–530. https://doi.
disorders: British national surveillance study. British Journal of org/10.1017/S0033291796004369
Psychiatry, 198(4), 295–301. https://doi.org/10.1192/bjp.bp.110. Schoemaker, C. (1997). Does early intervention improve the prog-
081356 nosis in anorexia nervosa? A systematic review of the
O’Hara, C. B., Campbell, I. C., & Schmidt, U. (2015). A reward- treatment-outcome literature. International Journal of Eating
centred model of anorexia nervosa: A focussed narrative review Disorders, 21(1), 1–15. https://doi.org/10.1002/(SICI)1098-108X
of the neurological and psychophysiological literature. Neuro- (199701)21:1<1::AID-EAT1>3.0.CO;2-R
science and Biobehavioral Reviews, 52, 131–152. https://doi.org/ Shu, C. Y., Limburg, K., Harris, C., McCormack, J., Hoiles, K. J.,
10.1016/j.neubiorev.2015.02.012 Hamilton, M. J., & Watson, H. J. (2015). Clinical presentation of
Oliver, D., Davies, C., Crossland, G., Lim, S., Gifford, G., eating disorders in young males at a tertiary setting. Journal of Eat-
McGuire, P., & Fusar-Poli, P. (2018). Can we reduce the dura- ing Disorders, 3(1), 39. https://doi.org/10.1186/s40337-015-0075-x
tion of untreated psychosis? Schizophrenia Bulletin, 44(6), Steinglass, J. E., & Walsh, B. T. (2016). Neurobiological model of
1362–1372. https://doi.org/10.1093/schbul/sbx166 the persistence of anorexia nervosa. Journal of Eating Disorders,
Penttilä, M., Jääskeläinen, E., Hirvonen, N., Isohanni, M., & 4(1), 19. https://doi.org/10.1186/s40337-016-0106-2
Miettunen, J. (2014). Duration of untreated psychosis as predic- Steinhausen, H. C. (2002). The outcome of anorexia nervosa in the
tor of long-term outcome in schizophrenia: Systematic review 20th century. American Journal of Psychiatry, 159(8),
and meta-analysis. British Journal of Psychiatry, 205(2), 88–94. 1284–1293. https://doi.org/10.1176/appi.ajp.159.8.1284
https://doi.org/10.1192/bjp.bp.113.127753 Stice, E., Telch, C. F., & Rizvi, S. L. (2000). Development and valida-
Pinhas, L., Wong, J., & Woodside, B. (2014). Early intervention tion of the eating disorder diagnostic scale: A brief self-report mea-
in eating disorders. In P. Byrne & A. Rosen(Eds.), Early sure of anorexia, bulimia, and binge-eating disorder. Psychological
intervention in psychiatry: EI of nearly everything for better Assessment, 12(2), 123. https://doi.org/10.1037/1040-3590.12.3.252
mental health (pp. 288–304). Chichester, UK: John Wiley & Sullivan, S. A., Carroll, R., Peters, T. J., Amos, T., Jones, P. B.,
Sons Ltd. Marshall, M., … Tilling, K. (2018). Duration of untreated psychosis
Reas, D. L., Schoemaker, C., Zipfel, S., & Williamson, D. A. (2001). and clinical outcomes of first episode psychosis: An observational
Prognostic value of duration of illness and early intervention in and an instrumental variables analysis. Early Intervention in Psy-
bulimia nervosa: A systematic review of the outcome literature. chiatry, 13, 841–847. https://doi.org/10.1111/eip.12676
International Journal of Eating Disorders, 30(1), 1–10. https:// Treasure, J., Stein, D., & Maguire, S. (2015). Has the time come for a stag-
doi.org/10.1002/eat.1049 ing model to map the course of eating disorders from high risk to
Schlegl, S., Hupe, K., Hessler, J. B., Diedrich, A., Huber, T., severe enduring illness? An examination of the evidence. Early Inter-
Rauh, E., … Voderholzer, U. (2019). Pathways to care and dura- vention in Psychiatry, 9(3), 173–184. https://doi.org/10.1111/eip.12170
tion of untreated illness of inpatients with anorexia and Volpe, U., Monteleone, A. M., Ricca, V., Corsi, E., Favaro, A.,
bulimia nervosa. Psychiatrische Praxis, 46, 342–348. https://doi. Santonastaso, P., … Maj, M. (2019). Pathways to specialist care for
org/10.1055/a-0922-5651 eating disorders: An Italian multicentre study. European Eating
Schmidt, U., Adan, R., Böhm, I., Campbell, I. C., Dingemans, A., Disorders Review, 27(3), 274–282. https://doi.org/10.1002/erv.2669
Ehrlich, S., … Zipfel, S. (2016). Eating disorders: The big issue. Weigel, A., Rossi, M., Wendt, H., Neubauer, K., von Rad, K.,
The Lancet Psychiatry, 3(4), 313–315. https://doi.org/10.1016/ Daubmann, A., … Gumz, A. (2014). Duration of untreated ill-
S2215-0366(16)00081-X ness and predictors of late treatment initiation in anorexia
Schmidt, U., Brown, A., McClelland, J., Glennon, D., & nervosa. Journal of Public Health, 22(6), 519–527. https://doi.
Mountford, V. A. (2016). Will a comprehensive, person-cen- org/10.1007/s10389-014-0642-7
tered, team-based early intervention approach to first episode World Health Organization. (1992). The ICD-10 classification of men-
illness improve outcomes in eating disorders? International tal and behavioural disorders: Clinical descriptions and diagnostic
Journal of Eating Disorders, 49(4), 374–377. https://doi.org/10. guidelines. Geneva, Switzerland: World Health Organization.
1002/eat.22519
Schmidt, U., Lee, S., Beecham, J., Perkins, S., Treasure, J., Yi, I., …
Johnson-Sabine, E. (2007). A randomized controlled trial of SU PP O R TI N G I N F O RMA TI O N
family therapy and cognitive behavior therapy guided self-care Additional supporting information may be found online in
for adolescents with bulimia nervosa and related disorders. the Supporting Information section at the end of this article.
American Journal of Psychiatry, 164(4), 591–598.
Schmidt, U., Magill, N., Renwick, B., Keyes, A., Kenyon, M.,
Dejong, H., … Watson, C. (2015). The Maudsley outpatient How to cite this article: Austin A, Flynn M,
study of treatments for anorexia nervosa and related conditions Richards K, et al. Duration of untreated eating
(MOSAIC): Comparison of the Maudsley model of anorexia disorder and relationship to outcomes: A
nervosa treatment for adults (MANTRA) with specialist sup- systematic review of the literature. Eur Eat
portive clinical management (SSCM) in outpatients with
Disorders Rev. 2020;1–17. https://doi.org/10.1002/
broadly defined anorexia nervosa: A randomized controlled
trial. Journal of Consulting and Clinical Psychology, 83(4),
erv.2745
796–807. https://doi.org/10.1037/ccp0000019