Expressed emotion in anorexia nervosa: What is inside the “black box”?

Available online at www.sciencedirect.com

ScienceDirect

Comprehensive Psychiatry 55 (2014) 71–79www.elsevier.com/locate/comppsych

Expressed emotion in anorexia nervosa: What is inside the “black box”?Jeanne Duclosa,b,⁎, Géraldine Dorardc, Sylvie Berthoza,b,d, Florence Curta,

Sophie Fauchera, Bruno Falissardb,d, Nathalie Godarta,b,daPsychiatry Unit, Institut Mutualiste Montsouris, 75014 Paris, France

bINSERM U669, 75014 Paris, FrancecUniversity Paris Descartes, Laboratoire de Psychopathologie et Processus de santé (LPPS), 92100 Boulogne-Billancourt, France

dParis-Sud and Paris Descartes Universities, Paris, France

Abstract

Objective: Expressed Emotion has been called a “black box”, since little is known about contributing factors. The aim of this study was toexamine which parental and which patient/illness-related characteristics contribute to maternal and paternal Expressed Emotion levels.Method: Sixty adolescent girls with Anorexia Nervosa (AN) and their parents completed instruments that evaluate characteristics of theadolescent's illness and patient/parental psychological characteristics (depression; anxiety; obsession–compulsion; social anxiety andalexithymia). The following illness-related characteristics were recorded: age at AN onset, duration of illness, AN subtype (restrictive AN-Rvs. purging type AN-B), current Body Mass Index (BMI) (in kg/m2), minimum lifetime BMI and number of previous hospitalizations,the Global Outcome Assessment Scale total score. Levels of Expressed Emotion were assessed for the two parents using the Five-MinuteSpeech Sample.Results: Less than 30% of the parents in our sample expressed high levels of Critical EE and Emotional Over-Involvement. Our mainfindings indicate that maternal Criticism (Critical EE levels, Critical Comments, Dissatisfaction) and the sub-dimensions of maternalEmotional Over-Involvement (EOI EE) (Statement of loving Attitudes and Excessive Details about the past) were related both to the severityof the daughters' clinical state and to maternal psychological functioning. Only paternal levels of anxiety explained paternal Dissatisfaction,EOI EE and Statement of loving Attitudes.Discussion: Parental psychological functioning and the severity of the daughters' clinical state have an impact on the family relationships.These elements should be targeted by individual treatment for parents where necessary, and by psycho-educational sessions about AnorexiaNervosa for parents generally.© 2014 Elsevier Inc. All rights reserved.

1. Introduction

Family relationships in Anorexia Nervosa (AN) areconsidered as one of the key elements implicated in theevolution of this disorder [1–3] and Family-Based Treatmentis the most widely practiced treatment in adolescents withAN [4,5].

Conflict of interest: none.⁎ Corresponding author. Institut Mutualiste Montsouris, Psychiatry Unit

42 bd Jourdan 75014 Paris, France. Tel.: +33 1566 61 6933; fax: +33 156661 6918.

E-mail addresses: [email protected] (J. Duclos),[email protected] (G. Dorard), [email protected](S. Berthoz), [email protected] (F. Curt), [email protected](B. Falissard), [email protected] (N. Godart).

0010-440X/$ – see front matter © 2014 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.comppsych.2013.10.002

The construct of Expressed Emotion (EE) was originallydeveloped in schizophrenia to assess family relationships[6]. Since 1981, EE has been widely studied in families witha member suffering from AN, and EE appears to be arelevant predictor of treatment compliance, early treatmentoutcome and long-term clinical outcomes of patientswith AN [7].

EE reflects the family climate between a patient andhis/her parents, focusing on 2 dimensions: Criticism(Critical EE) expressed by the parent towards their child,and Emotional Over-Involvement (EOI EE), defined asintrusive, overprotective, excessively self-sacrificing be-havior or exaggerated emotional response to the patient’sillness [8,9]. The underlying mechanisms of EE were for along time poorly understood. In 1985, Leff & Vaughnwere already wondering about the EE dilemma and “the

72 J. Duclos et al. / Comprehensive Psychiatry 55 (2014) 71–79

problem of prediction without understanding” [9]. Thisstatement led Jenkins & Karno to refer the EE as a “BlackBox” [10].

Recently it was hypothesized that the EE construct mayreflect a dynamic interaction between patient and parentaldimensions [11–13]. For example, Criticism could be areaction resulting from the way in which parents perceivethe patient, as determined by their own emotional state[14,15]. However, only a handful of studies have focusedon this theme. One study, concerning schizophrenia, foundthat maternal Critical EE was related to the severity of thepatient’s symptoms, while EOI EE was associated withboth maternal conscientiousness and patient depression[11]. In AN, parental levels of anxiety and depression andthe anorectic behaviors of the patient perceived by theparents as negative/difficult accounted for over 60% of thevariance in parental EOI. For parental Criticism, the mostsignificant variable was the negative/difficult behaviours ofthe patients as perceived by the parents, which accountedfor 50% of the variance at the first step of R2 partitioning.At the second and final step patient rejection of caregiversassistance accounted for a further 2% of the variance [12].Because few studies have sought to determine the contentsof the “Black Box” of EE, the aim of the present study isto examine the contributing factors to maternal andpaternal EE levels in families of adolescent girls withAN. Thus, we consider the parent and patient-relatedcharacteristics, including alexithymia, since difficulties inprocessing emotional states are implicated in the etiologyand maintenance of AN [16]. Furthermore we consider theillness-related characteristics, the socio-economic status ofthe families and the age of the patient, which have notbeen taken into account in previous studies, and becauseall these variables have been found to be related to levelsof EE [7]. We consider maternal and paternal EEseparately since interactions between parent and adolescentchild differ according to the dyad considered (mother/daughter or father/daughter) [17]. Better knowledge of thedeterminants of EE would enable the definition of potentialtherapeutic targets.

2. Materials and methods

This study is part of a comparison of two multidimen-sional post-hospitalization outpatient treatment programs foradolescents with severe AN: treatment as usual versus thistreatment plus Family Therapy (FT) (see Godart et al. [18]for details).

The study received approval from the Ethics Committeeand is in accordance with the terms of the Helsinkideclaration. The Trial Registration is Controlled-trials.comISRCTN71142875.

Prior to inclusion in the study, all participants had beenhospitalized in our care unit for life-threatening physical and/or mental states.

2.1. Participants and recruitment

Inclusion criteria: female subjects 13 to 21 years old, witha DSM-IV diagnosis of AN at admission, having beenhospitalized in our inpatient care unit for AN, age at onsetunder 19 years and AN duration ≤ 3 years at admission tothe hospital, living in the Paris metropolitan area, and neverhaving received FT.

Exclusion criteria were as follows for patients andparents: inability to speak or read French, and/or understandthe interview questions. For patients: any metabolicpathology interfering with eating or digestion (e.g. diabetes)or a psychotic disorder.

Out of the 116 patients for whom eligibility was assessedduring the recruitment period, 40 did not meet our selectioncriteria (10 males: 14 for whom illness onset occurred at age19 or older or illness duration N 3 years, 3 had a parent withschizophrenia, 5 were living outside the Paris area, 8 had hadFT previously). Out of the 76 eligible for participation, 16refused to participate (21%). The patients and parents whorefused to participate did not differ from those included withregard to socio-demographic variables, or clinical status onhospital entry or at discharge (data available on request) [18].

Sixty adolescent girls with severe AN, 60 mothers and 58fathers were included.

2.2. Assessment

The evaluations were conducted during the second part ofthe hospitalization, when refeeding was partially achievedand the patient was in regular contact with the family.

Expressed Emotion was evaluated using the Five-MinuteSpeech Sample (FMSS) [19]. The French version hassatisfactory validity properties [20,21]. EE levels wererated from a recorded five-minute sample of speech providedby the respondent about the patient. Each parent was asked totalk about what kind of a person their daughter was and theirrelationship with her. Two dimensions were rated on thebasis of both content and tone: Criticism (Critical EE) andEmotional Over-involvement (EOI EE). Critical EE is rated"High" if the parent makes an Initial negative Statement and/or expresses a negative Relationship and/or makes one ormore Critical Comments; in other cases, Critical EE is ratedas “Low”. A “High” EOI EE rating is based on Self-sacrificing or Overprotective attitudes and/or EmotionalDisplay (rated as present when, for example the respondentbursts into tears) during the interview, and on presence ofany 2 of the following: Excessive Details, Statements of(loving) Attitudes, at least 5 Positive Remarks; in other casesEOI EE is rated as “Low”.

2.3. Parent and patient-related characteristics

2.3.1. DepressionThe 13-item Beck Depression Inventory (BDI) is a self-

rating scale designed to evaluate cognitive and motivationalsymptoms of depression at the time of evaluation (rated from

73J. Duclos et al. / Comprehensive Psychiatry 55 (2014) 71–79

0 to 3). The scores range from 0 to 39 and higher total scoresindicate more severe depressive symptoms. This scale and itsFrench version have satisfactory psychometric propertiesacross clinical population [22,23].

2.3.2. AnxietyThe 10-item Anxiety Scale from the Symptom Check

List-90-Revised (SCL-90-R) is a clinical self-report scaledesigned to assess the severity of anxious symptoms over theprevious week ranging from 0 (none at all) to 4 (extreme). Itspsychometric properties have been well established ondifferent patient populations [24,25].

2.3.3. Obsession–CompulsionThe Yale–Brown Obsessive Compulsive Scale (Y-

BOCS) is a semi-structured interview designed to rate theseverity, the number and the type of obsessive–compulsivefeatures over the previous week, ranging from 0 (nosymptoms) to 4 (extreme symptoms). The scores rangefrom 0 to 40. This scale and its French version havesatisfactory reliability and validity properties [26,27].

2.3.4. Social AnxietyThe 24-item Liebowitz Social Anxiety Scale (SAS) is a

clinical interview designed to measure a range of socialinteraction and performance situations that patients withsocial anxiety disorder may fear, rated from 0 (none) to 3(severe), and avoidance behaviors (rated from 0 (never) to 3(usually)). The scores range from 0 to 96. The French versionhas satisfactory psychometric properties [28].

2.3.5. AlexithymiaThe 20-item Toronto Alexithymia Scale (TAS-20) is a

self-rating scale designed to capture alexithymic features,with responses ranging from 1 (totally disagree) to 5 (totallyagree). The scores range from 20 to 100. The French versionhas satisfactory validity and reliability properties [29,30].

The following illness-related characteristics wererecorded:

- age at AN onset, duration of illness, AN subtype(restrictive AN-R vs. purging type AN-B), currentBody Mass Index (BMI) (in kg/m2), minimum lifetimeBMI and number of previous hospitalizations.

- the Global Outcome Assessment Scale total score(GOAS) [31,32]: a clinical interview designed toassess the clinical status of the patient with AN overthe previous six months, across 5 sub-scales exploringeating and diet, menstruation, mental state, psycho-sexual functioning and socio-economic status. Thescore varies from 0 (doing very badly) to 12 (doingvery well). Higher total scores indicate a betterclinical state.

2.4. Data Analysis

Quantitative variables were described using mean (m),standard deviation (SD) and range, and qualitative variableswere described using proportions and percentages.

Comparisons between the mothers and fathers were madeusing paired-sample Wilcoxon test, Mac Nemar’s Chi2 testand Spearman's correlation coefficient.

The determination of contributing factors to maternal andpaternal EE levels was performed using univariate andmultivariate analyses.

2.4.1. Univariate analysesWe first tested, in separate analyses, the links between

maternal and paternal Critical EE, EOI EE (two-levelnominal variables) on the one hand and the potentiallyinfluential variables derived from the literature on the other(Psychological functioning: depression, anxiety, obses-sion–compulsion, social anxiety and alexithymia. Patient/illness characteristics: age at onset, duration of illness,current BMI, minimum lifetime BMI, the number ofprevious hospitalizations, the global clinical state, thesocio-economic status of the family and patient age). Wealso separately tested the link between maternal andpaternal EE sub-dimensions (Critical Comments, Dissatis-faction, Statement of loving Attitudes, Excessive Detailsabout the past and maternal Positive Remarks) and thepotentially influential variables, but only when more than 5parents rated the sub-dimensions (i.e. when N N 5). Meanswere compared using Mann–Whitney test or Student's t-tests as appropriate.

2.4.2. Multivariate logistic regression analysesMultivariate models were used to test the associations

between maternal and paternal Critical EE/EOI EE/EE sub-dimensions (when N N 5) and the potentially influentialexplanatory variables identified in the literature. In the firstmodel (Model 1), for each parent, we tested theassociations between Critical EE/EOI EE/EE sub-dimen-sions and parental characteristics (psychological function-ing). In a second model (Model 2), we tested associationsbetween maternal and paternal Critical EE/EOI EE/EEsub-dimensions on the one hand and the patient/illness-related variables on the other (see above). When both theparental characteristics and the patient/illness-related char-acteristics exhibited a significant association in the twoprevious models, we entered them in a third model (Model3) in order to determine which variables contribute themost to the EE dimensions. Forward stepwise multiplelogistic regressions were performed in order to identify thebest models.

Statistical analyses were performed using SPSS 19.0 forWindows. The level of significance was p b0.05.

3. Results

3.1. Participants’ characteristics

3.1.1. Patient/illness-related characteristics (Table 1)The mean age of the 60 patients was 16.6 (SD = 1.6)

years. Clinical information clearly indicated the seriousnessof their condition. The mean age at onset of AN was 14.8

Table 1Psychological functioning.

Patients m (SD)(N = 60)

Mothers m (SD)(N = 58)

Fathers m (SD)(N = 55)

Depression—BDI-13total score

9.9 (7.2) 5.0 (4.9) 3.4 (3.8)

AnxietySCL-90-R AnxietyScale

0.7 (0.7) 0.6 (0.6) 0.4 (0.4)

Y-BOCS total score 12.6 (9.5) 4.3 (6.2) 4.4 (4.9)LSAS total score 48.3 (38.3) 22.6 (21.7) 14.6 (15.8)

Alexithymia—TAS-20 total score

58.2 (11.7) 41.8 (11.7) 44.1 (9.1)

BDI-13: Beck Depression Inventory; LSAS: Liebowitz Social AnxietyScale; m: mean, N: sample size; SD: standard deviation; SCL-90-R:Symptom Checklist-90-Revised; TAS-20: 20-item Toronto AlexithymiaScale; Y-BOCS: Yale–Brown Obsessive Compulsive Scale.

able 2arental levels of EE.

evels of EE (FMSS) Fathers N (%) Mothers N (%)

igh Critical EE 15 (27.3) 14 (24.1)Initial Statement (−) 4 (6.7) 2 (3.4)Relationship (−) 3 (5.5) 4 (6.9)Critical Comments (P) 11 (20) 14 (24.1)Dissatisfaction (P) 21 (38.2) 21 (36.2)igh EOI EE 19 (34.5) 21 (36.2)Emotional Display (P) 3 (5.5) 3 (5.2)Statement of (loving) Attitudes (P) 8 (14.5) 6 (10.3)Self-Sacrifice/Overprotective/Lackof Objectivity (P)

1 (1.8) 2 (3.4)

Excessive Detail about the Past (P) 21 (38.2) 24 (41.4)Positive Remarks (P) 4 (7.2) 8 (13.8)

E: Expressed Emotion; EOI: Emotional Over-Involvement; FMSS: Fiveinutes Speech Sample; LSAS: Liebowitz Social Anxiety Scale; m: mean,: sample size; P: present; SD: standard deviation; (−): negative.

74 J. Duclos et al. / Comprehensive Psychiatry 55 (2014) 71–79

(1.6) years and the mean AN duration was 16.6 (6.8) months.The mean minimum lifetime BMI was 13.0 (1.1) and themean current BMI was 16.9 (1.1). Fifty-two patients (86.7%)were AN-R. The mean duration of hospitalization was 21(13.9) weeks. The mean Total score at GOAS was 4.3 (1.1).Numbers of previous hospitalizations varied from 0 to 4, anda quarter of the patients had been previously hospitalizedonce. Regarding education, 20.8% of the patients were inmiddle school, 75.5% were in higher secondary school and11.7% were attending higher education.

3.1.2. Parents-related characteristics (Table 1)The mean age of the 58 mothers was 47.6 (4.8) years and

that of the 55 fathers was 49.4 (4.9). According to the Frenchnomenclature, 38.3% of the mothers were white-collarworkers, 45.0% were in ‘intermediate professions’, 15%had ‘no activity’ and 1.7% were self-employed or smallbusiness owners. Among the 58 fathers engaged inprofessional activities (3 were retired), 11 (20%) wereartisans, shopkeepers or small business managers, 10(18.2%) were blue-collar workers or low-grade salariedworkers and 34 (61.8%) were in high-level professional jobs.

Mothers displayed significantly more depressive symp-toms (p = 0.048), more anxious symptoms (p = 0.045) andmore social phobia symptoms (p = 0.012) than the fathers.

3.1.3. Parental levels of EE (Table 2)Regarding parental levels of EE, no significant differ-

ences were noted between fathers and mothers. Levels of EEwere generally low. Maternal Critical EE derived only fromCritical Comments. Paternal Critical EE derived for the mostpart from a combination of Critical Comments, negativeInitial Statements and negative Relationships. Therefore tosimplify the paper, we will not present maternal CriticalComments. Parental EOI EE derived for the most part from acombination of 2 of the following three sub-dimensions:Excessive Details about the past, Statement of (loving)Attitudes and Positive Remarks.

TP

L

H

H

EMN

3.2. Relationships between parental EE levels andpatient/illness characteristics, patient/parentalpsychological functioning

3.2.1. Parental EE dimensions

3.2.1.1. Critical EE3.2.1.1.1. Parental characteristics. Where mothers

scored high for Critical EE (N = 44), they had significantlyhigher levels of anxiety in comparison with mothers scoringlow for Critical EE (N = 14) [SCL-R-90 scores: m (SD): 0.9(0.7) vs. 0.5 (0.6); Z = −2.1, p = 0.039]. Converselymothers who scored high for Critical EE did not differfrom mothers scoring low for critical EE for their other levelsof psychological functioning.

There was no link between paternal EE levels and theirlevels of psychological functioning.

3.2.1.1.2. Patient/illness-related characteristics. Com-pared to patients with mothers scoring low on Critical EE,those with mothers scoring high had significantly lowercurrent BMI [BMI: m (SD): 17.2 (1.0) vs. 16.3 (1.1);Z = −2.4, p = 0.016]. We found no group effect (high vs.low Critical EE) for none of their daughters’ characteristics.

Compared to patients with fathers scoring low for CriticalEE (N = 40), those with fathers scoring high for Critical EE(N = 15) had significantly lower levels of depression [BDI-13 scores: m (SD): 7.5 (6.8) vs. 11.2 (7.4); Z = −1.9, p =0.048]. Conversely, fathers did not differ in othersdaughters’ characteristics.

3.2.1.2. EOI EE3.2.1.2.1. Parental characteristics. For the mothers,

there was no group effect (High vs. Low EOI EE) for theirlevels of psychological functioning.

Compared to fathers scoring low for EOI EE (N = 36),fathers scoring high for EOI EE (N = 19) had higher levelsof depression [BDI-13 scores: m (SD): 2.9 (3.9) vs. 4.5 (3.3);Z = −2.1, p = 0.034] and anxiety [SCL-R-90 scores: m

75J. Duclos et al. / Comprehensive Psychiatry 55 (2014) 71–79

(SD): 0.3 (0.4) vs. 0.6 (0.5); Z = −2.6, p = 0.009], but didnot differ for their other levels of psychological functioning.

3.2.1.2.2. Patient/illness-related characteristics. Wefound no group effect for either parent for any of theirdaughters’ characteristics.

3.2.2. Parental EE sub-dimensions

3.2.2.1. Critical Comments. Exploration of this sub-dimension among the fathers revealed no group effect eitherfor the father’s characteristics, or for the patient/illness-related characteristics.

3.2.2.2. Dissatisfaction3.2.2.2.1. Parental characteristics. For the mothers,

the only group effect observed is that those who scored onDissatisfaction had higher obsessive–compulsive levels thanthe other mothers [Y-BOCS scores: m (SD): 7.1 (7.9) vs. 2.8(4.6); Z = −2.4, p = 0.015].

For the fathers, there was no group effect for their levelsof psychological functioning.

3.2.2.2.2. Patient/illness-related characteristics. Wefound no group effect for either parent for any of theirdaughters’ characteristics.

3.2.2.3. Statement of (loving) Attitudes3.2.2.3.1. Parental characteristics. For the mothers’

statement of (loving) attitudes, there was no group effect formaternal levels of psychological functioning.

Compared to fathers who did not express Statements of(loving) Attitudes, fathers who did had higher levels ofanxiety [m (SD): 0.3 (0.4) vs. 0.7 (0.6); Z = −2.0, p = 0.052].There was no group effect for other paternal levels ofpsychological functioning.

3.2.2.3.2. Patient/illness-related characteristics. Theonly group effect observed for mothers who did not makeStatements of (loving) Attitudes, was that their daughtersdisplayed a better global clinical state than the others [GOASscore: m (SD): 7.5 (6.8) vs. 11.2 (7.4); Z = −1.9, p = 0.048].

For the fathers, we found no group effect for any of theirdaughters’ characteristics.

3.2.2.4. Excessive Details about the past3.2.2.4.1. Parental characteristics. Compared to

mothers who did not express Excessive Details about thepast, mothers who did had higher levels of obsessive–compulsive symptoms [Y-BOCS scores: m (SD): 2.3 (4.2)vs. 5.8 (7.1), Z = −2.1, p = 0.033]. Conversely there was nogroup effect for their levels of depression, anxiety, socialphobia and alexithymia symptoms.

For the fathers, there was no group effect for their levelsof psychological functioning.

3.2.2.4.2. Patient/illness-related characteristics.Daughters of mothers whose discourse was characterizedby Excessive Detail concerning the past had had a longerduration of illness than the other daughters [19.1 (7.2) weeks

vs. 14.5 (5.5), Z = −2.6, p = 0.008]. They did not differ forother characteristics.

For the fathers, we found no group effect for any of theirdaughters’ characteristics.

3.2.2.5. Positive Remarks. The only group effect observedfor mothers who made Positive Remarks about theirdaughters was that the daughters had a significantly betterglobal clinical state than the other patients [GOAS score: m(SD): 5.0 (0.8) vs. 4.1 (1.2), Z = −2.1, p = 0.035].

3.3. Contributors to parental EE levels (Table 3)

3.3.1. Parental EE dimensions

3.3.1.1. Critical EE. Maternal Model 1: higher maternalCritical EE was associated with higher maternal anxietylevels. Maternal Model 2: higher maternal Critical EE wasassociated with lower patient BMI. Maternal Model 3:higher maternal Critical EE was associated with both highermaternal anxiety levels and lower patient BMI.

Paternal Models 1 and 2: paternal Critical EE was notassociated with any of the paternal or patient/illness-relatedcharacteristics.

3.3.1.2. EOI EE. Maternal Models 1 and 2: maternal EOIEE was not associated with any of the maternal or patient/illness-related characteristics.

Paternal Model 1: higher paternal levels of EOI EE wereassociated with higher paternal levels of anxiety. Model 2:paternal EOI EE levels were not associated with any of thepatient/illness-related characteristics.

3.3.2. Parental EE sub-dimensions

3.3.2.1. Critical Comments. Paternal Models 1 and 2:paternal Critical Comments were not associated with any ofthe paternal or patient/illness-related characteristics.

3.3.2.2. Dissatisfaction. Maternal Model 1: maternalDissatisfaction was not associated with any of the maternalcharacteristics. Model 2: maternal Dissatisfaction wasassociated with both a poorer global clinical state andlower levels of alexithymia in the daughters.

Paternal Model 1: lower levels of paternal alexithymiaexplained paternal Dissatisfaction. Model 2: paternalDissatisfaction was not associated with any of the patient/illness-related characteristics.

3.3.2.3. Statement of (loving) Attitudes. Maternal Model1: maternal Statements of (loving) Attitudes were notassociated with any of the maternal characteristics. Model2: maternal Statements of (loving) Attitudes were associatedwith a better global clinical state of the patient.

Paternal Model 1: paternal Statements of (loving)Attitudes were associated with higher paternal anxietylevels. Model 2: paternal Statements of (loving) Attitudes

Table 3Contributors to maternal and paternal EE: multivariate regression.

OR CI [95] R2 ROC p-value

MOTHER Critical EE Critical EE Model 1: Maternal Anxiety (SCL-R-90) 2.81 [1.06–7.39] 0.11 72.4 0.037Model 2: current BMI 0.42 [0.21–0.84] 0.18 79.3 0.015Model 3:- Maternal Anxiety (SCL-R-90) 3.76 [1.18–11.9] 0.025- current BMI 0.35 [0.15–0.08] 0.30 77.6 0.013Model 2:- Daughter TAS-20 total score 0.93 [0.88–0.98] 0.008

Dissatisfaction - GOAS total score 0.48 [0.26–.089] 0.24 70.7 0.019EOI EE Statement of Attitudes Model 1: GOAS total score 2.73 [1.21–6.98] 0.24 89.7 0.016

Excessive Detailsabout the Past

Model 1: Maternal Y-BOCS total score 0.88 [0.79–0.99] 0.13 62.1 0.038Model 2: AN duration 1.15 [1.03–1.29] 0.25 70.7 0.012Model 3:- AN duration 1.14 [1.10–1.28] 0.028- Maternal Y-BOCS total score 0.81 [0.77-0.99] 0.36 70.7 0.031

FATHER Critical EE Dissatisfaction Model 1:- Paternal Y-BOCS total score 1.12 [0.99–1.28] 0.069- Paternal TAS 20 total score 0.91 [0.83–0.99] 0.20 70.4 0.026

EOI EE EOI EE Model 1: Paternal Anxiety (SCL-90-R) 4.82 [1.21–19.1] 0.14 66.7 0.026Statement of Attitudes Model 1: Paternal Anxiety (SCL-90-R) 5.20 [1.10–24.4] 0.14 85.2 0.037

AN = Anorexia Nervosa; BMI = Body Mass Index; CI = Confidence Interval; EE: Expressed Emotion; EOI: Emotional Over-Involvement; GOAS: GlobalOutcome Assessment Schedule; OR = Odds Ratio; R2 Nagelkerke; ROC = Receiver Operating Characteristic; SCL-90-R: Symptom Checklist-90-Revised;TAS-20: Toronto Alexithymia Scale; Y-BOCS: Yale–Brown Obsessive Compulsive Scale. Model 1: association between parental EE levels and parent-relatedcharacteristics, Model 2: association between parental EE levels and patient-related characteristics and Model 3 when necessary.

76 J. Duclos et al. / Comprehensive Psychiatry 55 (2014) 71–79

were not associated with any of the patient/illness-relatedcharacteristics.

3.3.2.4. Excessive details about the past. Maternal Model1: maternal Excessive Details about the past were associatedwith lower maternal obsessive–compulsive levels. MaternalModel 2: maternal Excessive Details about the past wereassociated with a longer duration of patient AN. MaternalModel 3: maternal Excessive Details about the past wereassociated with lower maternal obsessive–compulsive levelsand a longer duration of AN.

Paternal Models 1 and 2: paternal Excessive Detailsabout the past were not associated with any of the paternal orpatient/illness-related characteristics.

4. Discussion

The originality of our study lies in four points. First, weexamined the contributing factors of Expressed Emotion(EE) levels and EE sub-dimensions in fathers and mothersseparately. Second, we considered parental and adolescentpsychological functioning together, including AN illness-related characteristics, socio-economic status of the familiesand the age of the patient, as potential contributors toparental EE. Third, we objectively evaluated the daughters’AN symptomatic behaviors, since parental perceptions of thedisorder are highly subjective. Four, we took into accountintrospection and regulation (i.e. alexithymia), a dimensionthat may play a role in emotional coping.

Less than 30% of the parents in our sample (mothers andfathers alike) expressed high levels of Critical EE and

Emotional Over-Involvement (EOI) EE. Our main findingsindicate that maternal Criticism (Critical EE levels, CriticalComments, Dissatisfaction) and the sub-dimensions ofmaternal Emotional Over-Involvement (EOI EE) (Statementof loving Attitudes and Excessive Details about the past)were related both to the severity of the daughters' clinicalstate and to maternal psychological functioning. Onlypaternal levels of anxiety explained paternal Dissatisfaction,EOI EE and Statement of loving Attitudes.

While we did not find any significant difference betweenEE levels in mothers and fathers in our population, as in thatstudied by Kyriacou et al. [12], we nevertheless found adifference concerning those elements contributing to highpaternal and maternal EE levels.

Direct comparisons of our results with those derived fromother studies on factors related to EE in Eating Disorders(ED) are not feasible, since in these other studies parental EEwas not differentiated for mothers and fathers [7].

4.1. Maternal EE levels

High maternal criticism was explained by both theseverity of the daughters’ malnutrition (BMI) and greatermaternal anxiety. A higher level of maternal Dissatisfactionwas associated with the severity of the daughters' globalclinical state and lower levels of alexithymia. In AN most ofthe maternal comments were generally related to eatingattitudes that mothers could not understand [33,34]. Failureon the part of mothers to understand the mechanismsunderpinning their daughters' AN, coupled with the anxietygenerated by the severity of the illness – in particular the life-threatening emaciation – could contribute to rejection of, and

77J. Duclos et al. / Comprehensive Psychiatry 55 (2014) 71–79

critical attitudes, towards their daughters. These results are inline with previous studies, where Critical EE has beentheorized as the result of inadequate understanding of thesymptoms of the illness [14] or as a response to the severityof the illness [35]. In line with the two previously mentionedauthors, we recommend setting up therapeutic educationgroups which could provide explanations about ANsymptoms, which in turn could contribute to reducingmaternal criticism. Indeed, two studies have shown thatparental EE levels are modifiable (i.e. reduced EE) bypsycho-educational interventions [36,37]. In addition,these group interventions would enable parents to meetother parents coping with the same pathology, whichcould have a positive impact, since it has been shown –albeit in individuals with traumatic brain injury – that socialsupport is a significant moderator of caregiver distress infamilies [38].

Maternal Criticism has been previously hypothesized as areaction resulting from the way in which parents perceive thepatient, determined by their own emotional state [14,15]. Ourdata support this hypothesis: maternal anxiety favorsmechanisms of criticism towards the affected daughter, andlowers the tolerance threshold towards the illness. Theassociation between maternal anxiety and criticism couldreflect maternal failure to cope with the illness and itschronic nature, making relations in daily life morecomplicated [39]. This is compatible with the modelproposed by Schmidt & Treasure [1] where the EE constructis seen as reflecting dynamic interaction between patient andparental dimensions [13]. Thus these elements may form avicious circle in which the daughter's emaciation (BMI)increases maternal anxiety, leading to criticism, so that acaregiver with High EE (in reaction to the illness) mightinadvertently fuel the disorder [13].

Maternal Statements of loving Attitudes were associatedwith better clinical status in the daughters. Our results are inline with a previous study of adolescents with AN whereinmothers were found to be more likely to make PositiveRemarks about adolescents with ED when the adolescentshad a “better illness status” at intake [40].

4.2. Paternal EE levels

We did not identify any patient/illness-related character-istics explaining paternal EE levels. However, some paternalcharacteristics were highlighted.

Paternal Dissatisfaction was positively linked to lowerlevels of paternal alexithymia. Fathers who are not markedlyalexithymic have been shown to be better able to identify andexpress their emotions [41]. It can thus be thought that thesefathers would more readily express their discontent abouttheir daughters' behaviors. This could also reflect their lack ofunderstanding of their daughters' pathology [14] and showhow they seek to deal with the stress that it generates [41].

Higher paternal EOI EE and Statements of lovingAttitudes were associated only with higher levels of paternal

anxiety: anxious fathers more readily express love fortheir daughters (« I love her », « she’s everything to me”).More generally, it has been hypothesized that altruisticbehavior (for the fathers in this study, this corresponds toexpressing their love for their daughters) could be mediatedby anxiety [42].

Generally speaking, these results underline the need toinvolve fathers in their daughters' treatment programs. Thisis all the more relevant because recent studies have shownthat there are clear differences between fathers and motherswith regard to the role they play in therapeutic settings: whilemothers seem to maintain their former parenting style,fathers for their part appear more inclined to becomeinvolved and mobilise their resources to obtain a change inrelationships with their children [43,44]. This could have aneffect on treatment outcome for their daughters. Indeed, in aprevious study, paternal "warmth" was found to be apredictor of better outcome in AN patients [45]. Theconstruct of EE originally comprises five dimensions, oneof which is “warmth”, reflecting the respondent's sympathy,concern and interest for the patient [9]. Walters et al. [46] hadalready observed that parental involvement in their daugh-ter's therapy for AN benefitted outcome.

4.3. Limitations

Certain aspects of this work limit the scope of itsresults. First, the evaluations in our sample wereconducted at the end of hospitalization, generating bothlow levels of parental EE and lower levels of patientpsychopathology. It is known that the presence of activeeating disorder symptoms is associated with a morenegative perception of the family functioning [47]. Thismay have reduced the statistical power of our results.Second, as the design of the study was cross-sectional,our results provide a measure of associations betweenEE and the variables under study, but no causal linkscan be identified. Third, as there was neither a normalcomparison group nor a group with another chronicillness, our conclusions are restricted to adolescent girlswith severe AN who have been hospitalized. Fourth, theclinical severity of AN in our patients needs to beunderlined and taken into account in the interpretation ofthe results. Finally, the size of the sample restricted thescope of the study of the different sub-dimensions ofEE. It would therefore be worthwhile replicating theseresults in a larger sample, also including control groups(i.e. schizophrenia).

5. Conclusion

Low BMI in adolescents with AN is associated withhigher levels of maternal criticism. Parental anxiety andpaternal alexithymia have an impact on family relationships.It is therefore important to involve parents so that they cangain a better understanding of the disorder and its

78 J. Duclos et al. / Comprehensive Psychiatry 55 (2014) 71–79

consequences in terms of symptoms and attitudes, so as to beless critical towards behaviors that they find difficult tounderstand, but that are in fact symptoms of the illness. It islikewise important to assess the mental state of the parents soas to offer individual support or treatment where necessary.Interventions for parents that target their anxiety wouldbenefit the parents themselves (by reducing anxiety anddepression), and could also alter their perceptions of theirdaughter’s difficulties and enable them to be moreempathetic and supportive, and to use warmer and lesscritical discourse. This could contribute to a better evolutionof AN, since communication, mutual respect and under-standing would certainly improve their relationship withtheir daughters. Further research is required to ascertain howfar interventions of this nature actually alter parental EE andoutcome in AN.

Acknowledgment

First we would like to thank all the patients andtheir parents who shared their experiences with us forthis research. We would like to thank Aminata Alifor her statistical back up and Alexandre Apfel andAnne-Solène Maria for their help with the manuscript.

This study was funded by the Projet Hospitalier deRecherche Clinique (CRC, PHRC, 1997, AOM97133 AP-HP) and promoted by Assitance Publique des Hôpitaux deParis (AP-HP).

J. Duclos receives grants from the Convention Industriellepour la Formation par la REcherche (CIFRE).

The funding source had no involvement in the writing ofthe article or in the decision to submit the paper forpublication.

We do not have specific acknowledgements for thepresent article, as all the persons who contributed to thiswork are listed as authors.

References

[1] Schmidt U, Treasure J. Anorexia nervosa: valued and visible. Acognitive–interpersonal maintenance model implications for researchand practice. Br J Clin Psychol 2006;45(3):343-66.

[2] Treasure J, Sepulveda AR, MacDonald P, Whitaker W, Lopez C,Zabala M, et al. The assessment of the family with people eatingdisorders. Eur Eat Disord Rev 2008;16(4):247-55.

[3] Le Grange D, Lock J, Loeb K, Nicholls D. Academy for EatingDisorders position paper: the role of the family in eating disorders. Int JEat Disord 2010;43(1):1-5.

[4] American Psychiatric Association (APA). Treatment of patients witheating disorders, third edition. APA. Am J Psychiatry 2006;163(7):4-54.

[5] HAS 2010, Haute Autorité de la Santé (HAS). Recommandation debonne pratique. Anorexie mentale: prise en charge. Collège de la HAS.http://www.has-sante.fr/.

[6] Hooley JM. Expressed emotion and relapse of psychopathology. AnnuRev Clin Psychol 2007;3:329-52.

[7] Duclos J, Vibert S, Mattar L, Godart N. Expressed emotion in familieswith patients with eating disorders: a review of the literature. CurrPsychiatry Rev 2012;8(3):183-202.

[8] Rutter M, Brown GW. The reliability and validity of measures offamily life and relationships in families containing a psychiatricpatient. Soc Psychiatry 1966;1(1):38-53.

[9] Leff J, Vaughn C. Expressed emotions in families: its significance formental illness. New York: Guilford Press; 1985.

[10] Jenkins JH, Karno M. The meaning of expressed emotion: theoreticalissues raised by cross-cultural research. Am J Psychiatry 1992;149:9-21.

[11] King S, Ricard N, Rochon V, Steiger H, Nelis S. Determinants ofexpressed emotion in mothers of schizophrenia patients. PsychiatryRes 2003;117:211-22.

[12] Kyriacou O, Treasure J, Schmidt U. Expressed emotion in eatingdisorders assessed via self-report: an examination of factors associatedwith expressed emotion in carers of people with anorexia nervosa incomparison to control families. Int J Eat Disord 2007;41:37-46.

[13] Treasure J, Crane A, McKnight R, Buchanan E, Wolfe M. First do noharm: iatrogenic maintaining factors in anorexia nervosa. Eur EatDisord Rev 2011;19(4):296-302.

[14] Treasure J, Tchanturia K, Schmidt U (2005). Developing a model ofthe treatment for eating disorder: using neuroscience research toexamine the “how” rather than the “what” of change. CPR 2005; 5:191–202.

[15] Kyriacou O, Treasure J, Schmidt U. Understanding how parents copewith living with someone with anorexia nervosa: modelling the factorsthat are associated with carer distress. Int J Eat Disord 2008;41:233-42.

[16] Hambrook D, Oldershaw A, Rimes K, Schmidt U, Tchanturia K,Treasure J, et al. Emotional expression, self-silencing, and distresstolerance in anorexia nervosa and chronic fatigue syndrome. Br J ClinPsychol 2011;50(3):310-25.

[17] Cummings EM, Davies PT, Campbell SB. Developmental psychopa-thology and family process: theory, research and clinical implications.New York: Guilford Press; 2002.

[18] Godart N, Berthoz S, Curt F, Perdereau F, Rein Z, Wallier J, et al. Arandomized controlled trial of adjunctive family therapy and treatmentas usual following inpatient treatment for anorexia nervosa adoles-cents. PLoS ONE 2012;7(1):e28249.

[19] Magaña AB, Goldstein MJ, Karno M, Miklowitz DJ, Jenkins J,Falloon IRH. A brief method for assessing expressed emotion inrelatives of psychiatric patients. Psychiatry Res 1986;17:203-12.

[20] Rein Z, Perdereau F, Curt F, Jeammet P, Fermanian J, Godart N.Expressed emotion and anorexia nervosa: the validation of the five-minute speech sample in reference to the Camberwell FamilyInterview. Int J Eat Disord 2006;39(3):217-23.

[21] Rein Z, Duclos J, Perdereau F, Curt F, Apfel A, Wallier J, et al.Expressed emotion measure adaptation into a foreign language. EurEat Disord Rev 2011;19(1):64-74.

[22] Collet L, Cottraux J. Inventaire abrégé de la dépression de Beck (13items). Etude de la validité concurrente avec les échelles de Hamiltonet de ralentissement de Widlöcher. Encéphale 1986;12:77-9.

[23] Beck AT, Steer RA, Garbin MG. Psychometric properties of the BeckDepression Inventory. Twenty-five years of evaluation. Clin PsycholRev 1998;8:77-100.

[24] Derogatis LR. Symptom Checklist-90-R (SCL-90-R): administration,scoring and procedures manual (3rd ed). Minneapolis: NCS Pearsons,Inc.: 1994.

[25] Pariente P, Guelfi JD. Inventaires d’auto-évaluation de la psycho-pathologie chez l’adulte. Première partie: Inventaires multidimension-nels. Psychiatry Psychobiol 1990;5:49-63.

[26] Woody SR, Steketee G, Chambless DL. Reliability and validity of theYale–Brown Obsessive–Compulsive Scale. Behav Res 1995;33(5):597-605.

[27] Mollard E, Cottraux J, Bouvard M. French version of the Yale–BrownObsessive Compulsive Scale. Encéphale 1989;15:335-41.

[28] Yao SN, Note I, Fanget F, Albuisson E, Bouvard M, Jalenques I, et al.Social anxiety in patients with social phobia: validation of theLiebowitz social anxiety scale: the French version. Encéphale1999;25(5):429-35.

79J. Duclos et al. / Comprehensive Psychiatry 55 (2014) 71–79

[29] Loas G, Fremaux D, Marchand MP. Factorial structure and internalconsistency of the French version of the twenty-item TorontoAlexithymia Scale in a group of 183 healthy probands. Encéphale1995;21:117-22.

[30] Loas G, Corcos M, Stephan P, Pellet J, Bizouard P, Venisse JL,et al. Factorial structure of the 20-item Toronto Alexithymia Scale:confirmatory factorial analysis in non clinical and clinical samples.J Psychosom Res 2001;50:255-61.

[31] Morgan HG, Hayward AE. Clinical assessment of anorexia nervosa.The Morgan–Russell outcome assessment schedule. Br J Psychiatry1988;152:367-71.

[32] Jeammet P, Brechon G, Payan C, Gorge A, Fermanian J. The outcomeof anorexia nervosa: a prospective study of 129 patients evaluated atleast 4 years after their first admission. Psychiatr Enfant 1991;34:381-442.

[33] Szmukler GI, Eisler I, Russell GFM, Dare C. Anorexia nervosa,parental "expressed emotion" and dropping out of treatment. Br JPsychiatry 1985;147:265-71.

[34] Szmukler GI, Berkowitz R, Eisler I, Leff J, Dare C. Expressed emotionin individual and family settings: a comparative study. Br J Psychiatry1987;151:174-8.

[35] Blair C, Freeman C, Cull A. The families of anorexia nervosa andcystic fibrosis patients. Psychol Med 1995;25(5):985-93.

[36] Uehera T, Kawashima Y, Goto M, Tasaki SI, Someya T. Psychoedu-cation for the families of patients with eating disorders and changes inexpressed emotion: a preliminary study. Compr Psychiatry 2001;42(2):132-8.

[37] Sepulveda AR, Todd G, Whitaker W, Grover M, Stahl D, Treasure J.Expressed emotion in relatives of patients with eating disordersfollowing skills training program. Int J Eat Disord 2010;43(7):603-10.

[38] Ergh TC, Rapport LJ, Coleman RD, Hanks RA. Predictors of caregiverand family functioning following traumatic brain injury: social supportmoderates caregiver distress. Br J Clin Psychol 2002;17:155-74.

[39] Macdonald P, Murray J, Goddard E, Treasure J. Carer's experience andperceived effects of a skills based training programme for families ofpeople with eating disorders: a qualitative study. Eur Eat Disord Rev2010:28.

[40] Van Furth EF, Van Strien DC, Martina LM, Van Son MJM, HendrickxJJ, Van Engeland H. Expressed emotion and the prediction of outcomein adolescent eating disorders. Int J Eat Disord 1996;20(1):19-31.

[41] Espina A. Alexithymia in parents of daughters with eatingdisorders: its relationships with psychopathological and personalityvariables. J Psychosom Res 2003;55(6):553-60.

[42] Cialdini RB. Personal influence: being ethical and effective. In:Oskamp S, & Spacapan S, editors. Interpersonal processes: theClaremont symposium on applied psychology. Newbury Park, CA:Sage; 1987. p. 95-107.

[43] Jauregui Lobera I, Garrido O, Santiago Fernandez MJ, AlvarezBautista E. Social comparison as a coping strategy among caregivers ofeating disorders patients. J Psychiatry Ment Health 2010;17:775-82.

[44] Jauregui Lobera I, Bolaños Rios P, Garrido Casals O. Parenting stylesand eating disorders. J Psychiatry Mental Health 2011;18:728-35.

[45] Le Grange D, Hoste RR, Lock J, Bryson SW. Parental expressedemotion of adolescents with anorexia nervosa: outcome in family-based treatment. Int J Eat Disord 2010;44(8):731-4.

[46] Walters J, Tasker F, Bichard S. Too busy? Father’s attendance forfamily appointments. J Fam Ther 2001;23(1):3-20.

[47] Woodside DB, Lackstrom J, Shekter-Wolfson L, Heinman M. Long-term follow up of patient-reported family functioning in eatingdisorders after intensive day hospital treatment. J Psychosom Res1996;41(3):269-77.