Received: 9 December 2019 Revised: 21 April 2020 Accepted: 21 April 2020

DOI: 10.1002/eat.23286

ORIGINAL ARTICLE

An mHealth intervention for the treatment of patients with

an eating disorder: A multicenter randomized controlled trial

Dimitra Anastasiadou PhD1,2 | Frans Folkvord PhD2,3 | Agostino Brugnera PhD4 |

Laura Cañas Vinader PhD5,6 | Eduardo SerranoTroncoso PhD5,6 |
Cristina Carretero Jardí MSc7 | Raquel Linares Bertolin MSc7 |
Rudiger Muñoz Rodríguez MSc8 | Beatriz Martínez Nuñez MD, MSc8 |
Montserrat Graell Berna MD, PhD8 | Jordi Torralbas-Ortega MSc9 |
Lidia Torrent-Solà MSc9 | Joaquim Puntí-Vidal PhD9,10 | Maria Carrera Ferrer MSc11 |
Andrea Muñoz Domenjó MD12 | Marina Diaz Marsa MD, PhD13 |
Katarina Gunnard PhD14 | Jordi Cusido PhD15,16 | Jordina Arcal Cunillera MSc15,16 |
Francisco Lupiañez-Villanueva PhD1,2
1
Department of Information and Communication Sciences, Universitat Oberta de Catalunya, Barcelona, Spain
2
Open Evidence Research Group, Universitat Oberta de Catalunya, Barcelona, Spain
3
Tilburg School of Humanities and Digital Sciences, Tilburg University, Tilburg, The Netherlands
4
Department of Human and Social Sciences, University of Bergamo, Bergamo, Italy
5
Child and Adolescent Psychiatry and Psychology Department, Sant Joan de Déu Hospital of Barcelona, Esplugues de Llobregat, Spain
6
Children and Adolescent Mental Health Research Group, Sant Joan de Déu Research Institut, Esplugues de Llobregat, Spain
7
Eating Disorders Unit, ABB Center, Barcelona, Spain
8
Child and Adolescent Psychiatry and Psychology Service, Niño Jesús University Children's Hospital, Madrid, Spain
9
Child and Adolescent Mental Health Service, Parc Taulí Foundation, Research and Innovation Institute Parc Taulí (I3PT) - Autonomous University of Barcelona,
Sabadell, Spain
10
Department of Clinical and Health Psychology, Autonomous University of Barcelona, Bellaterra (Cerdanyola del Vallès), Spain
11
Eating Disorders Programme IBSMIA, University Hospital Son Espases, Palma de Mallorca, Spain
12
Department of Psychiatry, University Hospital Móstoles, Móstoles, Spain
13
Eating Disorders Unit, San Carlos University Hospital, Madrid, Spain
14
Eating Disorders Unit, Quirón Dexeus University Hospital, Barcelona, Spain
15
Board Member, HealthApp SL, Sabadell, Spain
16
Department of Engineering Projects, Universitat Politècnica de Catalunya, Barcelona, Spain

Correspondence
Dimitra Anastasiadou, Department of Abstract
Information and Communication Sciences, Objective: The current multicentre randomized controlled trial assessed the clinical
Universitat Oberta de Catalunya, Barcelona,
Spain. efficacy of a combined mHealth intervention for eating disorders (EDs) based on cog-
Email: danastasiadou@uoc.edu nitive behavioral therapy (CBT).
Action Editor: Tracey Wade Method: A total of 106 ED patients from eight different public and private mental
health services in Spain were randomly assigned to two parallel groups. Patients of
Funding information
RecerCaixa of "La Caixa" Foundation the experimental group (N = 53) received standard face-to-face CBT plus a mobile
intervention through an application called “TCApp,” which provides self-monitoring
and an online chat with the therapist. The control group (N = 53) received standard

1120 © 2020 Wiley Periodicals, Inc. wileyonlinelibrary.com/journal/eat Int J Eat Disord. 2020;53:1120–1131.
ANASTASIADOU ET AL. 1121

face-to-face CBT only. Patients completed self-report questionnaires on ED symp-

tomatology, anxiety, depression, and quality of life, before and after treatment.
Results: Significant reductions in primary and secondary outcomes were observed
for participants of both groups, with no differences between groups. Results also
suggested that the frequency with which patients attended their referral mental
health institution after the intervention was lower for patients in the experimental
group than for those in the control group.
Discussion: The current study showed that CBT can help to reduce symptoms relat-
ing to ED, regardless of whether its delivery includes online components in addition
to traditional face-to-face treatment. Besides, the additional component offered by
the TCApp does not appear to be promising from a purely therapeutic perspective
but perhaps as a cost-effective tool, reducing thus the costs and time burden associ-
ated with weekly visits to health professionals.

KEYWORDS

eating disorders, mental health, mHealth, randomized controlled trial

1 | I N T RO DU CT I O N mobile interventions (Casasnovas et al., 2007; Kazdin & Blase, 2011). In

addition, empirically supported treatment for EDs is only available to a
In current societies, eating disorders (EDs) are a major cause of physical limited number of patients, making it important to explore ways to
and psychosocial disability and poor quality of life (Keski-Rahkonen & expand the availability and reach of evidence-based psychological treat-
Mustelin, 2016; Klump, Bulik, Kaye, Treasure, & Tyson, 2009), and are ments for EDs (Cooper & Bailey-Straebler, 2015; Kass, Kolko, &
associated with high mortality rates (Arcelus, 2011; Smink, van Hoeken, & Wilfley, 2013; Kazdin & Blase, 2011; Kazdin, Fitzsimmons-Craft, &
Hoek, 2013). In Spain, EDs were the second cause of disability-adjusted Wilfley, 2017; Simon & Ludman, 2009).
life-years (DALY) among women in 2010 after anxiety disorders. This mHealth interventions for EDs hold a great promise for reaching
fact highlights the need to prioritize such disorders in the Spanish public those in need of psychological treatment (Holmes et al., 2018). Multi-
health system (Lara et al., 2015), as well as to evaluate direct and indirect ple studies have shown that both patients and clinicians consider
costs associated with the disease (Kordy, 2005). mHealth techniques that support and facilitate ED symptoms moni-
In general, one of the forms of psychological therapy for patients toring as highly acceptable and feasible—either as a sole or additional
with EDs that is recommended most highly is cognitive behavioral ther- treatment tool for EDs (Anastasiadou, Folkvord, & Lupiañez-
apy (CBT) (National Institute for Health and Care Excellence, 2017). Villanueva, 2018; Darcy et al., 2014; Juarascio, Manasse, Goldstein,
One important behavioral component in CBT for ED is the self- Forman, & Butryn, 2015; Lindgreen, Clausen, & Lomborg, 2018;
monitoring of nutritional intake, as well as the thoughts and feelings Lindgreen, Lomborg, & Clausen, 2018). Up until now, however, the
that go with it (Fairburn, 2008). Despite the increasing adoption of CBT clinical utility of mHealth interventions for people with EDs is not fully
principles and techniques, many patients fail to continue to apply the clear, and high-quality research is largely lacking (Juarascio, Manasse,
skills learned in treatment in their daily lives (Juarascio, Parker, et al., 2015; Lui, Marcus, & Barry, 2017).
Lagacey, & Godfrey, 2018). In specific, although adherence to self- When accurately conducted, app-based treatments may lead to
monitoring is associated with better treatment outcomes in EDs an improved form of traditional CBT by facilitating monitoring of
(Darcy, Adler, Miner, & Lock, 2014), it is sometimes a challenging task symptoms, offering patients the opportunity to communicate and
for this group of patients, with many of them, mostly during adoles- share difficulties and improvements with their therapist wherever and
cence, complaining about difficulties associated with the daily use of whenever, and improving access to psycho-education and skills mate-
paper-and-pen records (i.e., forget to bring paper notes to therapy, fear rials (Fairburn & Patel, 2017; Fairburn & Rothwell, 2015; Juarascio,
of losing them, or do not want to carry them everywhere with them) Goldstein, Manasse, Forman, & Butryn, 2015; Luxton, McCann, Bush,
(Anastasiadou, Folkvord, Serrano-Troncoso, & Lupiañez-Villanueva,- Mishkind, & Reger, 2011).
2019; Lock, 2005). Taking also into account that this group of patients The mobile application that will be tested in this study, called
experiences difficulties associated with their specific condition, such as TCApp, has been specifically developed for people with EDs and is
a low motivation for change and for receiving treatment, it is crucial to based on the general principles of CBT. It represents a tool for con-
investigate alternative methods for delivering self-monitoring tech- necting patients with therapists in the time in-between medical con-
niques based on CBT principles to these patients, such as through sultations using online food records, monitoring of thoughts, actions,
1122 ANASTASIADOU ET AL.

and emotions, and bidirectional messages via chat between patients Committee of the University leading the study (Open University of
and their therapists. Due to the principle 24 hr a day availability to Catalonia, UOC). The study was registered in ClinicalTrials.gov
chat with their therapist and possibility of continuous self-monitoring, (Identifier NCT03197519), and its protocol was previously published
patients may feel more accompanied in their treatment process and elsewhere (Anastasiadou, Folkvord, & Lupiañez-Villanueva, 2018).
may experience more support and self-confidence in dealing with Therefore, we will only describe the most important aspects of the
treatment challenges. Equally, by use of the app, therapists can track study's procedure here.
patients online and have the opportunity to visualize their patients' Informed consent was obtained from all potential candidates for
progress using graphs and reports, as well as to contact them via chat the study. Clinical interviews and assessment for eligibility were then
when the need arises. The TCApp has been developed as a result of carried out using the KSADS-PL or SCID-I interviews (First, Spitzer,
fruitful partnership between technology experts from a company Gibbon, & Williams, 2002; Kaufman et al., 1997). Patients who were
called HealthApp, patients with EDs and ED specialists from different identified as eligible, as well as their families, completed the baseline
mental health institutions in the area around Barcelona (Spain). It is questionnaires (T0). Patients who did not return the documents within
considered to be a patient-centered tool which, through gamification 2 weeks received a reminder call.
elements (i.e., badges, points, scoreboards, alerts) increases patients' After completion of the baseline questionnaires (T0), patients
engagement and adherence with CBT treatment. were randomized across experimental group and control group.
The current study aims to assess the clinical efficacy of a com- Patients were allocated their condition using a computer-generated
bined intervention for EDs that includes a mobile intervention randomization list (allocation ratio 1:1; block size of 10; stratified per
through the TCApp plus standard face-to-face CBT, in comparison to hospital). Patients were notified regarding the outcome of the ran-
standard face-to-face CBT alone. Based on results from previous ran- domization in their next visit to the ED unit. Participants and ED spe-
domized controlled trials (RCT) showing more positive effects of cialists were since that moment aware of the allocated group, while
mHealth interventions for patients with ED as compared to traditional data manager/analyst (DA) remained blinded.
treatment (Bauer, Percevic, Okon, Meermann, & Kordy, 2003; Instructions on how to use the TCApp were given to patients
Hildebrandt et al., 2017), we hypothesized that patients with the from the experimental group by the ED specialist responsible for
TCApp plus standard face-to-face CBT would show more improve- online monitoring. Then, each group of patients received the treat-
ment on primary (ED pathology) and secondary outcomes (depression, ment that corresponded to their condition for a period of 12 weeks.
anxiety, quality of life, total number of visits) than patients receiving At the end of the 12-week treatment, evaluation T1 was carried out.
standard face-to-face CBT alone. Due to high dropout rate from T0 to T1, no T2 evaluation was carried
out (although it had been planned previously). For a detailed descrip-
tion of the study procedure, the definition of the study variables and
2 | METHOD assessment tools, the reader is referred to Figure 1 and Table 1 of the
published protocol study (Anastasiadou, Lupiañez-Villanueva, Faulí,
2.1 | Design and participants Arcal Cunillera, & Serrano-Troncoso, 2018).

A multicentre RCT was carried out with two parallel groups

(an intensive intervention group with standard CBT and TCApp, and a 2.3 | Interventions
standard CBT control group) with a 1:1 allocation.
Participants were recruited between February and September The experimental group received a standard face-to-face CBT (treat-
2018. The sample of patients with EDs was recruited from different ment as usual, TAU) in addition to the mobile health intervention using
public and private mental health services in Spain (Parc Taulí Hospital, the TCApp, for a time span of 12 weeks. During these 12 weeks, the
Balearic Island Health Service-Son Espases University Hospital, Sant patient should use the TCApp daily, completing the self-records and/or
Joan de Déu Hospital, Niño Jesús University Children's Hospital, San contacting his/her therapist via chat, when considered necessary. The
Carlos Clinic Hospital, Quiron Dexeus University Hospital and ABB therapist responsible for the online monitoring should connect to the
Center). All patients were over 12 years of age, had been diagnosed online platform and perform the following actions at least once a week:
with an eating or feeding disorder according to DSM-V criteria, and follow the patient's daily self-records, generate personalized reports or
were receiving a standard CBT treatment, with the support of a multi- graphs and communicate with him/her via chat. After a 12-week
disciplinary team of different ED units (psychiatry, psychology, nutri- period, patients from the experimental group and their therapists will
tion, and nursing). stop using the TCApp (they will be discharged).
The control group received standard treatment based on
face-to-face CBT (TAU), offered by the ED units that participated in
2.2 | Procedure the study, for a similar amount of time. It is worth mentioning that
the frequency and intensity of CBT treatment changed depending
The study was approved by the Ethical Committees of all participating on whether the patient fulfilled the criteria for undergoing outpa-
hospitals. In addition, approval was obtained from the Ethical tient or day hospital treatment. If they so wished, participants in the
ANASTASIADOU ET AL. 1123

Assessed for eligibility (n = 140)

Enrollment Excluded (n = 34)
♦ Not meeting inclusion criteria (n = 2)
♦ Declined to participate (lack of
motivation, lack of time, etc.) (n = 25)
♦ Parents declined to sign informed
consent (n = 4)
T0 assessment (n = 106) ♦ Other reasons (n = 3)

Randomized (n = 106)

Allocation
Allocated to experimental group (TAU + Allocated to control group (TAU; n = 53)
TCApp; n = 53) ♦ Received allocated intervention (n = 53)
♦ Received allocated intervention (app use ♦ Completed T0 assessment (n = 52)
≥1 time) (n = 48)
♦ Completed T0 assessment (n = 51)

Follow-up
♦ Not completed T1 assessment (n = 14) ♦ Not completed T1 assessment (n = 22)
♦ Discontinued intervention (app use ≤4
weeks)

FIGURE 1 Flow of participants through each stage of the study [Color figure can be viewed at wileyonlinelibrary.com]

control group were given access to the TCApp 6 months after the (EuroQol Group, 1990). The variable called “total number of regular
start of the study. visits” was also used as secondary outcome. It was assessed at T0 and
A related point to consider is that all ED units that have been chosen T1 by way of telephone interviews with the clinician responsible for
to take part in the trial employed CBT as a standard treatment for EDs the online monitoring of each patient, and included the number of
and that ED specialists from the seven ED units (five psychologists, one visits to ED specialists (i.e., individual therapy psychologist, group
psychiatrist, and one nursing staff) had previously received a CBT-specific therapy psychologist, psychiatrist, nursing staff, other medical staff of
training during their specialization internship. In addition, ED specialists various departments/specialities, nutritionist, or social worker). An
responsible for the online monitoring (experimental group) were trained additional variable that reflected the total number of emergency visits
on the basic principles of the application and the online platform. was also used as a secondary outcome. Finally, internal consistencies
for all measures at pretreatment were good (see Table 2 for
Cronbach's αs).
2.4 | Measures

Clinical interviews (First et al., 2002; Kaufman et al., 1997) were used 2.5 | Data analysis
to assess ED diagnosis and comorbidities among participants. An addi-
tional interview was employed to assess socio-demographic and Prior to the study, we estimated that a sample size of 200 participants
illness-related characteristics in the sample. (100 patients per study arm) would be recommended, increasing this
The primary outcome of the study concerned ED symptomatol- number to 250 to allow for a 25% loss to follow-up, and assuming an
ogy and was assessed with the Eating Disorder Examination Ques- α of .05 and a power of 0.80 (β − 1).
tionnaire (EDE-Q) (Peláez-Fernández, Javier Labrador, & Raich, 2012), We tested the hypothesis of a higher improvement in primary
and the Short Evaluation of Eating Disorders (SEED) (Bauer, Winn, and secondary outcomes from pre- to posttreatment in participants
Schmidt, & Kordy, 2005). Secondary outcomes were general psycho- with the TCApp plus standard face-to-face CBT compared to patients
pathology and quality of life, measured using the Beck Depression receiving standard face-to-face CBT alone, using two-level hierarchi-
Inventory (BDI-II) (Wiebe & Penley, 2005), the State-Trait Anxiety cal linear models (HLMs). HLMs are considered one the best statistical
Inventory (STAI) (Spielberger, Gorsuch, & Lushene, 1982), the techniques to examine longitudinal changes in nested data (such as in
EuroQoL-EQ-5L (EQ-5D-5L), and the child-friendly EQ-5D version the case of individual participant data hierarchically nested within
1124 ANASTASIADOU ET AL.

TABLE 1 Baseline characteristics of all study participants, and separately for each group

Characteristics Intervention (n = 53) Control (n = 53) Total (n = 106) p value

Demographics
Age (years), M (SD) 17.25 (3.54) 18.88 (7.77) 18.06 (6.04) .557
Sex, n (%)
Male 5 (9.4) 4 (7.7) 9 (8.6) .99
Female 48 (90.6) 48 (92.3) 96 (91.4)
Highest level of education, n (%)
Primary 10 (18.9) 10 (19.2) 20 (19.0) .806
Secondary 22 (41.5) 20 (38.5) 42 (40.0)
Baccalaureate 7 (13.2) 8 (15.4) 15 (14.3)
Vocational training 2 (3.8) 4 (7.7) 6 (5.7)
University 11 (20.8) 9 (17.3) 20 (19.0)
Employment status, n (%)
Full-time job 0 1 (1.9) 1 (1.0) .447
Part-time job 3 (5.7) 5 (9.6) 8 (7.6)
Student 48 (90.6) 45 (86.5) 93 (88.6)
Sick leave 2 (3.8) 0 2 (1.9)
Retired 0 1 (1.9) 1 (1.0)
Currently living together, n (%)
Nuclear family 38 (73.1) 32 (61.5) 70 (67.3) .411
Single parent family—mother 6 (11.5) 9 (17.3) 15 (14.4)
Single parent family—father 2 (3.8) 0 2 (1.9)
Single parent family—joint custody 2 (3.8) 3 (5.8) 5 (4.7)
Grandparent family 0 1 (1.9) 1 (1.0)
Proper family/partner 1 (1.9) 4 (7.7) 5 (4.7)
Living alone 3 (5.8) 3 (5.8) 6 (5.7)
Clinical data
Evaluation center, n (%)
PT 7 (13.2) 9 (17.0) 16 (15.1) .669
BAL 6 (11.3) 3 (5.7) 9 (8.5)
SJD 13 (24.5) 18 (34.0) 31 (29.2)
NJ 11 (20.8) 7 (13.2) 18 (17.0)
SC 4 (7.5) 5 (9.4) 9 (8.5)
DEX 1 (1.9) 0 1 (0.9)
ABB 11 (20.8) 11 (20.8) 22 (20.8)
Diagnosis, n (%)
AN-restrictive 26 (49.1) 25 (47.2) 51 (48.1) .299
AN-purging 3 (5.7) 4 (7.5) 7 (6.6)
BN 5 (9.4) 8 (15.1) 13 (12.3)
BED 4 (7.5) 0 4 (3.8)
OSFED 15 (28.3) 16 (30.2) 31 (29.2)
BMI, M (SD) 20.54 (4.45) 20.14 (3.59) 20.34 (4.03) .808
Comorbidity Axis I, n (%)
Yes 22 (41.5) 23 (44.2) 45 (42.9) .845
No 31 (58.5) 29 (55.8) 60 (57.1)
Duration ED (since onset; months), M (SD) 39.62 (40.29) 47.56 (62.03) 43.59 (52.20) .752
Duration all treatments, M (SD) 25.88 (35.16) 19.04 (20.39) 22.46 (28.81) .772
ANASTASIADOU ET AL. 1125

TABLE 1 (Continued)

Characteristics Intervention (n = 53) Control (n = 53) Total (n = 106) p value

Current treatment type, n (%)
Day hospital 27 (50.9) 25 (48.1) 52 (49.5) .846
Outpatient 26 (49.1) 27 (51.9) 53 (50.5)
Pharmacological treatment, n (%)
Yes 35 (66) 28 (52.8) 63 (59.4) .235
No 18 (34) 25 (47.2) 43 (40.6)

Abbreviations: ABB, ABB Center; BAL, Balearic Island Health Service; NJ, Niño Jesús University Children's Hospital; PT, Parc Taulí Hospital; SC, San Carlos
Clinic Hospital; SJD, Sant Joan de Déu Hospital.

treatment groups; Gallop & Tasca, 2009; Singer & Willett, 2003). Their analyses run with and without transformed variables led to similar
main advantage is the flexibility in handling missing data (Gallop & results. Thus, we reported results with untransformed values for ease
Tasca, 2009), a common occurrence in longitudinal studies. We ran of interpretation (Tabachnick & Fidell, 2007). Because the variable
intent-to-treat (ITT) analyses, examining the longitudinal changes of “total number of visits to emergency departments” was strongly asym-
all participants according to their assigned treatment group and metrical (i.e., less than 15% participants accessed emergency depart-
regardless of actual adherence to the treatment protocol. ments, and among those only 1–2 participants per group had two
We first tested if data were missing at random and if having any visits during the 3 months before), this variable was dichotomized and
missing data were significantly related to outcomes through a pattern data were analyzed through hierarchical generalized linear models
mixture model (Gallop & Tasca, 2009). Because all outcomes were (HGLMs; see Data S1 for the multilevel model).
measured across two time-points (pre- and posttreatment), we were Finally, we tested if the data was missing at random using a pat-
only able to test a linear “time” slope (see Data S1 for the multilevel tern mixture model (Gallop & Tasca, 2009): the nonsignificant effects
model). Analyses were controlled for several individual-level of the missing data pattern (dropouts vs. completers) suggested that
covariates, namely age, the number of times the App was used, dura- all data were missing at random and that the estimates of effects were
tion in months of the eating disorder, diagnosis (dummy coded), the unbiased by the presence of dropouts.
presence of an Axis I diagnosis, and current pharmacotherapy. All
covariates were grand-mean centered.
Effect sizes indicating the proportion of within-person variance 3.2 | Patient characteristics
accounted for by adding the linear parameter and were assessed and
reported using pseudo-R2 (Raudenbush & Bryk, 2002). Their magnitude Of the 250 patients approached for the current study a total of
was interpreted according to guidelines (0.01 = small, 0.06 = medium, 140 were enrolled, 34 of which were excluded or refused to partici-
>0.14 = large; Cohen, 1988). We additionally reported Cohen's d for pate. Patients not interested in participating most often reported
the between-groups differences at posttreatment on all the main and lack of motivation or lack of time (n = 25). In some cases, parents
secondary outcomes. Their magnitude was interpreted according to declined to sign the informed consent (n = 4). Thus, the final study
guidelines (small ≥ 0.20; medium ≥ 0.50; large ≥ 0.80; Cohen, 1988). sample was of 106 patients, 53 of which were randomized to the
Analyses were conducted using SPSS version 26.0 (IBM intervention and 53 to the control group. Figure 1 displays the
Corp., 2013) and HLMs version 7.0.3 (Raudenbush, Bryk, Cheong, & CONSORT diagram.
Congdon, 2011). All statistical tests were two-tailed and maintained a Regarding the frequency of the TCApp use by participants from
5% significance threshold. the experimental group, the mean app use was M = 7.11 (SD = 4.56;
range = 0–12). In specific, five users did not use the application at all,
12 users discontinued intervention (used the app for less than
3 | RESULTS 4 weeks) and 35.8% of users used the app during 12 out of 12 weeks.
We compared the sociodemographic and clinical characteristics
3.1 | Preliminary analyses of (1) patients in the intervention and in the control group and of
(2) dropouts and study completers through chi-square and Fisher
We found few outliers at the two time points (T0 and T1) for frequen- exact test (for frequencies), and Mann–Whitney U test (for continuous
cies of primary and secondary outcomes, so extreme scores were variables). In both cases, no between-group differences were
brought into range (Tabachnick & Fidell, 2007). According to their observed for any of the baseline variables, besides a significantly
skewness and kurtosis values, few variables (SEED BN severity index higher frequency of Axis I comorbidities among dropouts compared to
and BMI) were also non-normally distributed. A square-root or log10 study completers (p = .040). Of note, once the intervention began
transformation corrected the violation of this assumption, however dropout rate was lower in the intervention group (30.2%) compared
1126 ANASTASIADOU ET AL.

TABLE 2 Means, SD, total N, effect sizes (Cohen's d between groups at T1) and Cronbach's αs (computed at pretreatment) for all
psychological variables across the two time points, and separately for the intervention and control groups

Variable N Mean (SD) N Mean (SD) Effect sizea α

T0 T1
EDE-Q total 91 2.70 (1.67) 70 1.93 (1.48) 0.11 .96
Intervention 47 2.83 (1.54) 39 2.01 (1.46)
Control 44 2.55 (1.78) 31 1.84 (1.53)
EDE-Q restriction 91 2.18 (1.80) 70 1.42 (1.43) −0.07 .89
Intervention 47 2.22 (1.67) 39 1.38 (1.36)
Control 44 2.14 (1.94) 31 1.48 (1.54)
EDE-Q eating concern 91 2.20 (1.52) 70 1.51 (1.35) 0.07 .82
Intervention 47 2.27 (1.43) 39 1.56 (1.37)
Control 44 2.13 (1.64) 31 1.46 (1.34)
EDE-Q shape concern 91 3.30 (1.96) 70 2.44 (1.79) 0.18 .95
Intervention 47 3.53 (1.84) 39 2.58 (1.77)
Control 44 3.05 (2.07) 31 2.27 (1.82)
EDE-Q weight concern 92 2.74 (1.79) 70 2.04 (1.63) 0.15 .83
Intervention 47 2.83 (1.72) 39 2.16 (1.62)
Control 45 2.64 (1.82) 31 1.91 (1.65)
SEED AN total severity index 89 0.83 (0.48) 69 0.68 (0.50) −0.09 –
Intervention 48 0.85 (0.47) 38 0.66 (0.48)
Control 41 0.81 (0.50) 31 0.70 (0.53)
SEED BN total severity index 92 0.54 (0.47) 69 0.40 (0.41) 0.09 –
Intervention 48 0.58 (0.50) 38 0.41 (0.39)
Control 41 0.49 (0.44) 31 0.38 (0.44)
BDI-II total 92 23.29 (14.10) 69 15.45 (13.04) 0.07 .91
Intervention 48 23.66 (13.50) 38 15.85 (13.62)
Control 44 22.89 (14.87) 31 14.95 (12.49)
STAI State 92 47.72 (13.80) 69 43.36 (12.84) 0.15 .95
Intervention 48 47.73 (14.36) 38 44.22 (13.65)
Control 44 47.70 (13.32) 31 42.32 (11.90)
EQ-5D-5L (adults) 20 0.86 (0.09) 22 0.88 (0.15) 0.88 –
Intervention 11 0.86 (0.08) 14 0.92 (0.10)
Control 9 0.86 (0.11) 8 0.80 (0.19)
Total visits 100 30.09 (30.38) 102 26.99 (27.87) −0.02 –
Intervention 52 33.54 (33.59) 52 26.77 (24.27)
Control 48 26.35 (26.30) 50 27.22 (31.42)
b
Emergency visits, n (%) 106 15 (14.1) 106 10 (9.4) – –
Intervention 53 8 (15.1) 53 6 (11.3)
Control 53 7 (13.2) 53 4 (7.5)
BMI 106 20.34 (4.03) 106 20.53 (4.05) 0.11 –
Intervention 53 20.54 (4.45) 53 20.75 (4.39)
Control 53 20.14 (3.59) 53 20.31 (3.70)

Abbreviations: α, Cronbach's α at pretreatment; AN, anorexia nervosa; BDI-II, Beck Depression Inventory II; BMI, body mass index; BN, bulimia nervosa;
EDE-Q, Eating Disorder Examination Questionnaire; EQ-5D-5L, 5 level EQ-5D version; SEED, Short Evaluation of Eating Disorders Questionnaire; STAI
State, State-Trait Anxiety Inventory.
a
Cohen's d (between groups, at T1).
b
Due to the extremely low occurrence, participants with more than one emergency visit during the past 3 months were considered as cases, and only one
frequency was counted.
ANASTASIADOU ET AL. 1127

TABLE 3 Results from the hierarchical linear models indicating the growth parameter (β10) for each variable, and the effects of the
interaction between study condition (i.e., intervention and control groups) and the growth parameter (β11) for each variable

Variable β SE 95% CIs t values df p Pseudo R2

EDE-Q total
Time parameter β10 −0.84 0.16 −1.15, −0.53 −5.40 83 <.001 .38
Time × condition parameter β11 −0.73 0.46 −1.63, 0.17 −1.56 83 .122
EDE-Q restriction
Time parameter β10 −0.82 0.18 −1.17, −0.47 −4.51 83 <.001 .34
Time × condition parameter β11 −1.06 0.54 −2.12, −0.002 −1.96 83 .054
EDE-Q eating concern
Time parameter β10 −0.72 0.15 −1.01, −0.43 −4.83 83 <.001 .35
Time × condition parameter β11 −0.33 0.45 −1.21, 0.55 −0.74 83 .463
EDE-Q shape concern
Time parameter β10 −0.99 0.17 −1.32, −0.66 −5.79 83 <.001 .40
Time × condition parameter β11 −0.82 0.52 −1.84, 0.20 −1.59 83 .115
EDE-Q weight concern
Time parameter β10 −0.72 0.18 −1.07, −0.37 −4.09 83 <.001 .28
Time × condition parameter β11 −0.62 0.53 −1.66, 0.42 −1.19 83 .237
SEED AN total severity index
Time parameter β10 −0.18 0.05 −0.28, −0.08 −3.58 82 <.001 .25
Time × condition parameter β11 −0.28 0.15 −0.57, 0.01 1.88 82 .064
SEED BN total severity index
Time parameter β10 −0.13 0.05 −0.23, −0.03 −2.66 82 .009 .21
Time × condition parameter β11 −0.10 0.15 −0.39, 0.19 0.66 82 .510
BDI-II total
Time parameter β10 −8.86 1.61 −12.02, −5.70 −5.50 83 <.001 .41
Time × condition parameter β11 −3.34 4.76 −12.67, 5.99 −0.70 83 .485
STAI state
Time parameter β10 −5.28 1.66 −5.58, −4.98 −3.18 83 .002 .22
Time × condition parameter β11 −0.67 4.92 2.29, −3.63 −0.14 83 .891
EQ-5D-5L (adults)
Time parameter β10 0.01 0.08 −0.15, 0.17 0.08 16 .940 .62
Time × condition parameter β11 0.07 0.12 −0.17, 0.31 −0.58 16 .568
Total visits
Time parameter β10 −3.55 2.64 −8.72, 1.62 1.34 93 .184 .13
Time × condition parameter β11 −18.29 7.57 −33.13, −3.45 2.42 93 .017
Emergency visits
Time parameter β10 −0.90 0.55 −2.31, 0.51 −1.62 97 .110 –
Time × condition parameter β11 −1.43 1.47 −1.92, −0.94 −0.97 97 .333
BMI
Time parameter β10 0.20 0.10 0.004, 0.4 1.94 93 .056 .17
Time × condition parameter β11 0.11 0.30 −0.48, 0.70 −0.39 93 .700

Note: β10 indicates the person-level effect of the time parameter. β11 indicates the interaction between condition (i.e., intervention and control groups) and
the time parameter. Pseudo R2 refers to the amount of within-person variance accounted for by adding the time parameter to level 1 of the completely
unconditional multilevel model. R2 cannot be computed for dichotomous outcomes (i.e., frequency of emergency visits during the past 3 months).
Abbreviations: β, unstandardized regression weight; AN, anorexia nervosa; BDI-II, Beck Depression Questionnaire II; BMI, body mass index; BN, bulimia
nervosa; df, degrees of freedom (df are less than expected due to few missing values at level 2); EDE-Q, Eating Disorder Examination Questionnaire;
EQ-5D-5L, 5 level EQ-5D version; SEED, Evaluation of Eating Disorders Questionnaire; STAI State, State-Trait Anxiety Inventory.
1128 ANASTASIADOU ET AL.

to the control group (43.4%). However, this difference was not statis- 4 | DI SCU SSION
tically significant (p = .227).
Sociodemographic and clinical characteristics for the total sample The current multicenter RCT assessed the clinical efficacy of a com-
and separately for the two study groups are provided in Table 1, while bined intervention for EDs that included standard face-to-face CBT
means and standard deviations for psychological dimensions at the plus the TCApp application when compared to standard face-to-face
two time points (pre and posttreatment) are reported in Table 2. CBT treatment. It was expected that patients in the experimental
group would show more positive effects of treatment compared to
those in the active control group. In contrast to our hypotheses,
3.3 | Effect of the intervention on primary results showed significant reductions in primary outcomes (eating dis-
outcomes order symptomatology) as well as secondary outcomes (anxiety,
depression, and quality of life) for participants in both groups, with
To test our first study hypothesis, we compared the effectiveness moderate effects and no differences between experimental and con-
of the two interventions designed to reduce patients' ED-related trol group. Results also suggested that the frequency with which
symptoms using two-levels HLMs, controlling results for several patients attended their referral mental health institution after the
covariates. intervention was lower for patients in the experimental group than for
Results showed that there was no significant difference between those in the control group. This finding might prove promising in the
the two groups on the longitudinal changes in the EDE-Q total and light of reducing future direct and indirect costs associated with ED
subscale scores, and in the AN and BN total severity index of the treatment (Kordy, 2005).
SEED (Table 3). Of note, the slope parameter β10 was always signifi- These findings are in line with previous research which indicates
cant, suggesting that CBT led to medium-to-large reductions in all that mHealth interventions for patients with EDs focusing on self-
ED-related symptoms (R2 range: 0.21–0.40), regardless the group. At monitoring show limited additional effectiveness compared to active
posttreatment, the between-groups difference in all outcomes was control groups (Mazzeo et al., 2016). In a different study that used a
trivial (Cohen's d range: −0.09–0.18; see Table 2). mobile application based on CBT principles for binge eating, however,
the effectiveness of the intervention in reducing ED symptoms was
indeed underscored (Hildebrandt et al., 2017). Similarly, Ruwaard
3.4 | Effect of the intervention on secondary et al. (2013) found encouraging effects of online CBT for
outcomes BN. Comparing our findings with those in other studies should be
done with caution, however, given the differences between study
To test our secondary hypotheses, we evaluated between-group samples and implemented treatments. In addition, methodological
treatment effects on the secondary outcomes using HLMs, controlling issues surrounding studies examining the efficacy and effectiveness of
results for several covariates. mHealth interventions do not make it any easier to compare such
Results showed that there was no statistically significant difference studies and draw firm conclusions (Lui et al., 2017).
between the two groups on the longitudinal changes in the BDI-II, It should also be noted that, while there were no significant dif-
STAIState, and EQ-5D-5L total scores, in the frequency of “total num- ferences between the two treatment conditions with regard to ED
ber of emergency visits” during the past 3 months as well as in the symptomatology and general wellbeing, we found significant differ-
BMI values. Interestingly, we found a significant effect of the Inter- ences, although with a small effect size, in the total number of times
vention on the longitudinal changes in the variable “total number of patients attended their referral mental health institution to ask for
regular visits”, suggesting that the total number of visits of the experi- help from different health professionals. This included individual or
mental group was significantly lower after treatment compared to the group therapy psychologists, psychiatrists, nursing staff, other medical
control group (Table 3). staff of various medical specialities, nutritionists, and social workers.
The slope parameter β10 of the variables BDI-II and STAI State Findings suggest that the new component offered by the TCApp as
was significant, suggesting that CBT led to large reductions in depres- complementary to the face-to-face CBT intervention does not appear
2
sive and anxious symptoms (R range: 0.22–0.41), regardless the to be promising from a purely therapeutic perspective but perhaps as
group. At posttreatment, the between-groups difference in all out- a cost-effective tool, an important outcome of the treatment condi-
comes was trivial to large (Cohen's d range: −0.02, 0.88; see Table 2). tion that should also be taken into account and judged on its own
The effect size of 0.88 was attributable to EQ-5D-5L total score, merit. An explanation of this finding may be that the group of patients
suggesting that patients in the intervention group reported a higher who were using the TCApp may have perceived increased autonomy
Quality of Life at posttreatment than those in the control group. and self-confidence during their treatment process, which can be con-
All regression coefficients, SEs, t and p values for β10 (i.e., the sidered as an indicator of better adherence to their treatment (Crow
unstandardized regression coefficient for the average rate of growth et al., 2013). In addition, integrating self-management through the app
from pre- to posttreatment) and β11 (i.e., the interaction between in patients' daily life may be associated with reduced stigma and/or
treatment condition and the time parameter) slope parameters are shame associated with seeking in-person treatment and sharing ED-
reported in Table 3, while all Cohen's d are reported in Table 2. related behaviors with the referral professional (Juarascio, Goldstein,
ANASTASIADOU ET AL. 1129

et al., 2015; Juarascio, Manasse, et al., 2015) as well as with barriers TCApp or paper-and-pen records may both be considered valuable
related to transportation and its costs and time barriers (Ali interventions for the treatment of patients with EDs. In addition,
et al., 2016; Juarascio, Goldstein, et al., 2015; Juarascio, Manasse, patients' self-monitoring through the TCApp may be capable of
et al., 2015). Another possibility is that the different outcome vari- increasing patients' sense of autonomy, thus reducing the number of
ables that we have used for clinical assessment were less sensitive to weekly visits to a health institution. This comes on top of the fact that
detect adherence to the CBT treatment. Such an alternative explana- it requires only a minimal online involvement for health professionals.
tion might also take into account the short period of post-assessment Nevertheless, a more detailed evaluation of treatment adherence
compared to a more direct and changing variable such as the patients' should be carried out (Loeb et al., 2005). In addition, the examination
total number of visits to a health professional. of relevant clinical (ED diagnosis, symptom severity, psychiatric
Remarkably, overall adherence of the experimental group with comorbidity, treatment type or duration, referral health institution),
the TCApp presented some problems, taking into account that only demographic (sex, age, economic status), and technical covariables
one third of the users (35.8%) of users used the app during the whole (a.o. smartphone literacy) at different follow-up periods and with a
duration of the intervention. At this point, it is worth mentioning larger sample should be taken into account. Finally, direct and indirect
results from a previous qualitative study by Anastasiadou et al. (2019), costs relating to the use of the TCApp treatment should be examined
which examined TCApp's adoption levels by users. Results showed to provide evidence for its cost-effectiveness.
that the app was deemed easy to use and acceptable by both patients
and clinicians as a complementary tool to regular treatment, although AC KNOW LEDG EME NT S
concerns were expressed about the degree of personalization and the The current study is part of a broader project titled “mHealth: Chal-
overwhelming quantification of symptoms through the app, which lenges and opportunities for health systems” which receives funding
may have led some users to discontinue the online intervention. from RecerCaixa of “La Caixa” Foundation.
The current study has several strengths and limitations. First, one
of its strengths is the inclusion of an active control group and a rigor- CONFLIC T OF INT ER E ST
ous assessment of our sample: through face-to-face diagnostic semi- The authors declare no conflicts of interest.
structured interviews, self-report questionnaires, as well as telephone
interviews. In addition, we recruited a heterogeneous sample with dif- DATA AVAILABILITY STAT EMEN T
ferent ED diagnoses and illness durations, from a variety of private The data that support the findings of this study are available on
and public health care institutions, and with the aim to reflect the situ- request from the corresponding author. The data are not publicly
ation of daily clinical practice in Spain as much as possible. available due to privacy or ethical restrictions.
One of the limitations of the current study is the small sample
size, and the fact that some of our analyses are underpowered. For OR CID
example, the large effect sizes for EQ-5D-5L scores at posttreatment Dimitra Anastasiadou https://orcid.org/0000-0001-7544-523X
suggest that the longitudinal between-group differences in this vari- Eduardo SerranoTroncoso https://orcid.org/0000-0002-4935-7348
able could have been significant with a larger sample size. We should
mention here the difficulty we experienced in recruiting patients and RE FE RE NCE S
the high dropout rate, which makes generalizability of our results Ali, K., Farrer, L., Fassnacht, D. B., Gulliver, A., Bauer, S., & Griffiths, K. M.
decidedly more difficult. Taking into account the characteristics of our (2016). Perceived barriers and facilitators towards help-seeking for
eating disorders: A systematic review. International Journal of Eating
sample (treatment-resistant disorders, as well as dealing with minors
Disorders, 50(1), 9–21. https://doi.org/10.1002/eat.22598
under intensive treatment whose caregivers were not always willing Anastasiadou, D., Folkvord, F., & Lupiañez-Villanueva, F. (2018). A system-
to collaborate due to the risks associated with the exposure of their atic review of mHealth interventions for the support of eating disor-
sons/daughters to app-based treatment), this is a common problem in ders. European Eating Disorders Review, 26(5), 394–416. https://doi.
org/10.1002/erv.2609
this area of research. As a result, the modest sample size and short
Anastasiadou, D., Folkvord, F., Serrano-Troncoso, E., & Lupiañez-
duration of post-assessment prevented us from determining if the Villanueva, F. (2019). Mobile health adoption in mental health: User
clinically meaningful effects of the TCApp may attain greater signifi- experience of a mobile health app for patients with an eating disorder.
cance over time—something that would have been facilitated with a JMIR mHealth and uHealth, 7(6), e12920. https://doi.org/10.2196/
12920
greater sample size. In addition, the small sample size did not allow us
Anastasiadou, D., Lupiañez-Villanueva, F., Faulí, C., Arcal Cunillera, J., &
to conduct specific analyses of subgroups of patients according to Serrano-Troncoso, E. (2018). Cost-effectiveness of the mobile appli-
their specific ED diagnosis, referral institution (public vs. private sec- cation TCApp combined with face-to-face CBT treatment compared
tor) or age (minors vs. adults). For future research, we also suggest to face-to-face CBT treatment alone for patients with an eating dis-
carrying out follow-up assessments at different time intervals, in order order: Study protocol of a multi-centre randomised controlled trial.
BMC Psychiatry, 18(1), 118. https://doi.org/10.1186/s12888-018-
to identify long-term predictors of good outcome or dropout.
1664-4
In conclusion, based on our results and those of previous findings Arcelus, J. (2011). Mortality rates in patients with anorexia nervosa and
(Anastasiadou, Folkvord, & Lupiañez-Villanueva, 2018; Fairburn & other eating disorders. Archives of General Psychiatry, 68(7), 724–731.
Rothwell, 2015), CBT using either online monitoring through the https://doi.org/10.1001/archgenpsychiatry.2011.74
1130 ANASTASIADOU ET AL.

Bauer, S., Percevic, R., Okon, E., Meermann, R., & Kordy, H. (2003). Use of treatment of eating disorders. European Eating Disorders Review, 23(1),
text messaging in the aftercare of patients with bulimia nervosa. 1–11. https://doi.org/10.1002/erv.2327
European Eating Disorders Review, 11(3), 279–290. https://doi.org/10. Juarascio, A. S., Parker, M. N., Lagacey, M. A., & Godfrey, K. M. (2018).
1002/erv.521 Just-in-time adaptive interventions: A novel approach for enhancing
Bauer, S., Winn, S., Schmidt, U., & Kordy, H. (2005). Construction, scoring skill utilization and acquisition in cognitive behavioral therapy for eat-
and validation of the short evaluation of eating disorders (SEED). ing disorders. International Journal of Eating Disorders, 51(8), 826–830.
European Eating Disorders Review, 13(3), 191–200. https://doi.org/10. https://doi.org/10.1002/eat.22924
1002/erv.637 Kass, A. E., Kolko, R. P., & Wilfley, D. E. (2013). Psychological treatments
Casasnovas, C., Fernández-Aranda, F., Granero, R., Krug, I., Jiménez- for eating disorders. Current Opinion in Psychiatry, 26(6), 549–555.
Murcia, S., Bulik, C. M., & Vallejo-Ruiloba, J. (2007). Motivation to https://doi.org/10.1097/YCO.0b013e328365a30e
change in eating disorders: Clinical and therapeutic implications. Kaufman, J., Birmaher, B., Brent, D., Rao, U., Flunn, C., Moreci, P., …
European Eating Disorders Review, 15(6), 449–456. https://doi.org/10. Ryan, N. (1997). Schedule for affective disorders and schizophrenia for
1002/erv.780 school-age children-present and lifetime version (K-SADS-PL): Initial
Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences reliability and validity data. Journal of the American Academy of Child &
(2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Associates. Adolescent Psychiatry, 36(7), 980–988. https://doi.org/10.1097/
Cooper, Z., & Bailey-Straebler, S. (2015). Disseminating evidence-based 00004583-199707000-00021
psychological treatments for eating disorders. Current Psychiatry Kazdin, A. E., & Blase, S. L. (2011). Rebooting psychotherapy research and
Reports, 17(3), 12. https://doi.org/10.1007/s11920-015-0551-7 practice to reduce the burden of mental illness. Perspectives on Psycholog-
Crow, S. J., Agras, W. S., Halmi, K. A., Fairburn, C. G., Mitchell, J. E., & ical Science, 6(1), 21–37. https://doi.org/10.1177/1745691610393527
Nyman, J. A. (2013). A cost effectiveness analysis of stepped care Kazdin, A. E., Fitzsimmons-Craft, E. E., & Wilfley, D. E. (2017). Addressing
treatment for bulimia nervosa. International Journal of Eating Disorders, critical gaps in the treatment of eating disorders. International Journal
46(4), 302–307. https://doi.org/10.1002/eat.22087 of Eating Disorders, 50(3), 170–189. https://doi.org/10.1002/eat.
Darcy, A. M., Adler, S., Miner, A., & Lock, J. (2014). How smartphone appli- 22670
cations may be implemented in the treatment of eating disorders: Case Keski-Rahkonen, A., & Mustelin, L. (2016). Epidemiology of eating disor-
reports and case series data. Advances in Eating Disorders, 2(3), ders in Europe. Current Opinion in Psychiatry, 29(6), 340–345. https://
217–232. https://doi.org/10.1080/21662630.2014.938089 doi.org/10.1097/YCO.0000000000000278
EuroQol Group. (1990). EuroQol–a new facility for the measurement of Klump, K. L., Bulik, C. M., Kaye, W. H., Treasure, J., & Tyson, E. (2009).
health-related quality of life. Health Policy (Amsterdam, Netherlands), 16 Academy for eating disorders position paper: Eating disorders are seri-
(3), 199–208. ous mental illnesses. International Journal of Eating Disorders, 42(2),
Fairburn, C. G. (2008). Cognitive behavior therapy and eating disorders. 97–103. https://doi.org/10.1002/eat.20589
New York: Guilford Press. Kordy, H. (2005). Counting the COST: A European collaboration on the
Fairburn, C. G., & Patel, V. (2017). The impact of digital technology on psy- efficiency of psychotherapeutic treatment of patients with eating dis-
chological treatments and their dissemination. Behaviour Research and orders. European Eating Disorders Review, 13(3), 153–158. https://doi.
Therapy, 88, 19–25. https://doi.org/10.1016/j.brat.2016.08.012 org/10.1002/erv.634
Fairburn, C. G., & Rothwell, E. R. (2015). Apps and eating disorders: A sys- Lara, E., Garin, N., Ferrari, A. J., Tyrovolas, S., Olaya, B., Sànchez-Riera, L.,
tematic clinical appraisal. International Journal of Eating Disorders, 48 … Haro, J. M. (2015). The Spanish burden of disease 2010: Neurologi-
(7), 1038–1046. https://doi.org/10.1002/eat.22398 cal, mental and substance use disorders. Revista de Psiquiatría y Salud
First, M., Spitzer, R., Gibbon, M., & Williams, J. (2002). Structured clinical Mental (English Edition), 8(4), 207–217. https://doi.org/10.1016/j.
interview for DSM-IV-TR Axis I disorders, research version, patient edition rpsmen.2015.10.003
(SCID-I/P). New York, NY: Biometrics Research, New York State Psy- Lindgreen, P., Clausen, L., & Lomborg, K. (2018). Clinicians' perspective on
chiatric Institute. an app for patient self-monitoring in eating disorder treatment. Inter-
Gallop, R., & Tasca, G. A. (2009). Multilevel modeling of longitudinal data national Journal of Eating Disorders, 51(4), 314–321. https://doi.org/
for psychotherapy researchers: II. The complexities. Psychotherapy 10.1002/eat.22833
Research, 19(4–5), 438–452. https://doi.org/10.1080/10503300902 Lindgreen, P., Lomborg, K., & Clausen, L. (2018). Patient experiences using a
849475 self-monitoring app in eating disorder treatment: Qualitative study. JMIR
Hildebrandt, T., Michaelides, A., Mackinnon, D., Greif, R., DeBar, L., & mHealth and uHealth, 6(6), e10253. https://doi.org/10.2196/10253
Sysko, R. (2017). Randomized controlled trial comparing smartphone Lock, J. (2005). Adjusting cognitive behavior therapy for adolescents with
assisted versus traditional guided self-help for adults with binge eat- bulimia nervosa: Results of case series. American Journal of Psychother-
ing. International Journal of Eating Disorders, 50(11), 1313–1322. apy, 59(3), 267–281.
https://doi.org/10.1002/eat.22781 Loeb, K. L., Wilson, G. T., Labouvie, E., Pratt, E. M., Hayaki, J., Walsh, B. T.,
Holmes, E. A., Ghaderi, A., Harmer, C. J., Ramchandani, P. G., Cuijpers, P., … Fairburn, C. G. (2005). Therapeutic alliance and treatment adher-
Morrison, A. P., … Craske, M. G. (2018). The lancet psychiatry commis- ence in two interventions for bulimia nervosa: A study of process and
sion on psychological treatments research in tomorrow's science. The outcome. Journal of Consulting and Clinical Psychology, 73(6),
Lancet Psychiatry, 5(3), 237–286. https://doi.org/10.1016/S2215- 1097–1106. https://doi.org/10.1037/0022-006X.73.6.1097
0366(17)30513-8 Lui, J. H. L., Marcus, D. K., & Barry, C. T. (2017). Evidence-based apps? A
IBM Corp. (2013). IBM SPSS statistics for windows, version 26.0. Armonk, review of mental health mobile applications in a psychotherapy con-
NY: IBM Corp. text. Professional Psychology: Research and Practice, 48(3), 199–210.
Juarascio, A. S., Goldstein, S. P., Manasse, S. M., Forman, E. M., & https://doi.org/10.1037/pro0000122
Butryn, M. L. (2015). Perceptions of the feasibility and acceptability of Luxton, D. D., McCann, R. A., Bush, N. E., Mishkind, M. C., & Reger, G. M.
a smartphone application for the treatment of binge eating disorders: (2011). mHealth for mental health: Integrating smartphone technology
Qualitative feedback from a user population and clinicians. Interna- in behavioral healthcare. Professional Psychology: Research and Practice,
tional Journal of Medical Informatics, 84(10), 808–816. https://doi.org/ 42(6), 505–512. https://doi.org/10.1037/a0024485
10.1016/j.ijmedinf.2015.06.004 Mazzeo, S. E., Lydecker, J., Harney, M., Palmberg, A. A., Kelly, N. R.,
Juarascio, A. S., Manasse, S. M., Goldstein, S. P., Forman, E. M., & Gow, R. W., … Stern, M. (2016). Development and preliminary effec-
Butryn, M. L. (2015). Review of smartphone applications for the tiveness of an innovative treatment for binge eating in racially diverse
ANASTASIADOU ET AL. 1131

adolescent girls. Eating Behaviors, 22, 199–205. https://doi.org/10. Smink, F. R. E., van Hoeken, D., & Hoek, H. W. (2013). Epidemiology,
1016/j.eatbeh.2016.06.014 course, and outcome of eating disorders. Current Opinion in Psychiatry,
National Institute for Health and Care Excellence. (2017). Eating disorders: 26(6), 543–548. https://doi.org/10.1097/YCO.0b013e328365a24f
Recognition and treatment. Spielberger, C., Gorsuch, R., & Lushene, R. (1982). STAI. In Cuestionario de
Peláez-Fernández, M. A., Javier Labrador, F., & Raich, R. M. (2012). Validation Ansiedad Estado/Rasgo. Madrid: Tecnicos Especialistas Asociados (TEA).
of eating disorder examination questionnaire (EDE-Q)–Spanish version– Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th
for screening eating disorders. The Spanish Journal of Psychology, 15(2), ed.). New York, NY: Allyn and Bacon.
817–824. https://doi.org/10.5209/rev_sjop.2012.v15.n2.38893 Wiebe, J. S., & Penley, J. A. (2005). A psychometric comparison of the beck
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical Linear Models: Appli- depression inventory-II in English and Spanish. Psychological Assess-
cations and data analysis methods. (2nd ed.). Thousand Oaks, CA: Sage ment, 17(4), 481–485. https://doi.org/10.1037/1040-3590.17.4.481
Publications, Inc.
Raudenbush, S. W., Bryk, A. S., Cheong, Y. F. & Congdon, R. (2011). HLM
7 for Windows [Computer software]. Lincolnwood, IL: Scientific Soft- SUPPORTING INF ORMATION
ware International, Inc. Additional supporting information may be found online in the
Ruwaard, J., Lange, A., Broeksteeg, J., Renteria-Agirre, A., Schrieken, B.,
Supporting Information section at the end of this article.
Dolan, C. V., & Emmelkamp, P. (2013). Online cognitive-Behavioural treat-
ment of bulimic symptoms: A randomized controlled trial. Clinical Psychol-
ogy & Psychotherapy, 20(4), 308–318. https://doi.org/10.1002/cpp.1767
Simon, G. E., & Ludman, E. J. (2009). It's time for disruptive innovation in How to cite this article: Anastasiadou D, Folkvord F,
psychotherapy. The Lancet, 374(9690), 594–595 https://doi.org/10. Brugnera A, et al. An mHealth intervention for the treatment
1016/S0140-6736(09)61415-X of patients with an eating disorder: A multicenter randomized
Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis:
controlled trial. Int J Eat Disord. 2020;53:1120–1131. https://
Modeling change and event occurrence. Oxford, UK: Oxford University
Press. https://doi.org/10.1093/acprof:oso/9780195152968.001. doi.org/10.1002/eat.23286
0001
Copyright of International Journal of Eating Disorders is the property of John Wiley & Sons,
Inc. and its content may not be copied or emailed to multiple sites or posted to a listserv
without the copyright holder's express written permission. However, users may print,
download, or email articles for individual use.