23. Difficulties in family functioning and adolescent anorexia nervosa, The British Journal of Psychiatry
(C) 1999 The Royal College of Psychiatrists
Volume 174(1), January 1999, pp 63-66
Difficulties in family functioning and adolescent anorexia nervosa
GOWERS, SIMON; NORTH, CLIVE
SIMON G. GOWERS, FRCPsych, Academic Unit, Young People's Centre, Chester; CLIVE
NORTH, MRCPsych, North Essex Child & Family Consultation Service, Clacton-on-Sea,
Background Difficulties in family functioning are often evident when an
adolescent has anorexia nervosa, and the possible causative or contributory role
of such difficulties in the illness is unclear.
Aims To elucidate the relationship between severity of anorexia nervosa and
difficulties in family functioning and whether clinical improvement results in
diminution of self-rated family difficulties.
Method Thirty-five adolescents with anorexia nervosa and their mothers completed
the Family Assessment Device (FAD) while clinicians administered the McMaster's
Structured Interview of Family Functioning (McSIFF). Severity of anorexia
nervosa was rated at baseline and at one year follow-up using the Morgan-Russell
Results Clinicians and patients were more critical of the families' functioning
than parents.There was an inverse association between the extent of family
difficulties and severity of anorexia nervosa. Over time subjects improved
clinically but this was not matched by improvement in family functioning.
Conclusions Difficulties in family functioning do not appear to be directly
associated with severity of anorexia nervosa nor do these difficulties reduce
with clinical improvement, in the short term.
Declaration of interest This study was supported by a grant from the North West
Regional Health Authority.
When an adolescent has anorexia nervosa, family difficulties are often evident
(Waller et al, 1989; Le Grange et al, 1992; Thienemann & Steiner, 1993)
[17,8,16]. Given a worrying condition in which resistance to parental attempts
to increase food intake is usually part of the picture, this is not particularly
surprising. Some family theorists, however, have suggested that certain family
difficulties may antedate the illness and contribute causally, particularly
those involving failures of communication and problem solving (Minuchin et al,
1978; Crisp, 1980) [10,3]. In an attempt to disentangle contributory causative
features of anorexia from effect, we consider the following questions: (a) Are
families less critical than clinicians of their family functioning? (b) What is
the relationship between the severity of anorexia nervosa and the degree of
perceived family difficulty? (c) Are improvements in anorexia nervosa matched by
improvements in family functioning over the course of one year?
This study was part of a larger project evaluating predictors of outcome in
anorexia nervosa. Full details of recruitment and measures are contained in our
earlier publications (North et al, 1995, 1997) [13,14]. These earlier reports
showed that families with a member with anorexia nervosa tended to view their
functioning as healthier than psychiatric controls but where difficulties in
family functioning were reported by subject or clinician rating, they predicted
a poor clinical outcome.
The subjects comprised 36 cases with anorexia nervosa consecutively referred to
a regional adolescent service. Current episode diagnosis (DSM-III-R, American
Psychiatric Association, 1987)  and comorbidity were confirmed at a research
diagnostic interview (Schedule for Affective Disorders and Schizophrenia for
Children and Adolescents (K-SADS); Chambers et al, 1985) . One subject
subsequently refused further assessment leaving 35 participants (31 female, four
male). Their mean age was 14.9 years, mean length of illness 14 months. All were
living at home with at least one biological parent.
The measures used were as follows.
The Schedule for Affective Disorders and Schizophrenia for children and
adolescents (K-SADS), is a semi-structured diagnostic interview (Chambers et al,
1985)  generating diagnoses according to DSM-III-R criteria, based on
interviews with both subject and mother.
The Family Assessment Device (FAD; Epstein et al, 1983)  is a self-report
questionnaire designed to evaluate family functioning. It consists of 60
questions in revised form generating seven subscale scores each rated between 1
(no difficulty) and 4 (severe difficulties). These are problem-solving,
communication, roles, affective responsiveness, affective involvement, behaviour
control and general functioning. The FAD's acceptable reliability and validity
have been demonstrated (Miller et al, 1985; Kabacoff et al, 1990) [9,7]. The FAD
was administered to subjects and their mothers at initial assessment (n=35
subjects/35 mothers) and at one-year follow-up (n=33 subjects/32 mothers).
The McMaster's Structured Interview of Family Functioning (McSIFF) is a
semi-structured clinician-rated family interview, designed to assess family
functioning according to the McMaster model. The interview is divided into six
main sections which cover the first six subscales of the FAD. Each area is rated
on a seven-point scale, the clinical rating scale, a lower score representing
more pathological family functioning (Epstein et al, 1982) . This scale
incorporates detailed descriptions of family functioning to guide assignment of
ratings. An overall rating of family functioning is calculated, the average of
the other six areas. A consensus regarding family functioning was achieved by
comparing the interviewer's rating with an independent judgement by the second
rater on the basis of video or written transcript. Differences of more than one
point were rare and were reviewed until agreement was reached. The McSIFF was
administered at initial assessment only. One family refused to take part in this
The Morgan-Russell Outcome Assessment Schedule (MROAS) is a semi-structured
interview used in the assessment of outcome of anorexia nervosa (Morgan &
Russell, 1975; Morgan & Hayward, 1988) [11,12]. It covers eating behaviour, body
weight, mental state, menstrual functioning, psychosexual adjustment and
socio-economic status. Subscale scores are calculated and averaged to produce
the average outcome score. The measure has been widely used in follow-up studies
of anorexia nervosa. Care was taken over the scoring of the psychosexual and
socio-economic adjustment subscores, to compare the behaviour reported to that
which was developmentally appropriate for an adolescent of the same age and
gender. In order to measure change over time MROAS scores for all subjects were
calculated at initial assessment based on the recorded initial clinical
interview and at one year follow-up by subjects and parental interview. Adequate
follow-up information to complete the MROAS was obtained for 34 subjects at
one-year follow-up, based on 33 face-to-face interviews and one rating based on
medical informants where interview was refused.
Is the clinician more critical than the family of its functioning?
The families did not have a united view of their functioning. The adolescent
sufferers were in general much more critical than their parents as demonstrated
by higher mean scores on the FAD for each subscale (the mean scores have been
previously reported; North et al, 1995) . The difference in the method of
scoring the McSIFF makes a direct comparison between the clinician's and the
family's (FAD) rating difficult but the difference in the level of criticism of
each rater can be judged by the number of ratings given by each which are in the
clinical range proposed by the authors. For the FAD the thresholds for the
clinical range are those suggested for each scale by Miller et al (1985) .
For the McSIFF a rating of
Of the 35 adolescent raters, 17 rated their families to be within the clinical
range for four out of the six subscales (Table 1). Seven of them rated their
families' functioning as abnormal on all six scales. (The general functioning
scale has not been reported as no comparable scale exists for the McSIFF.)
Table 1. Family functioning at presentation. Number of families rated in
the clinical range by each rater
Only seven parents, meanwhile, rated themselves to have difficulties on the
above four scales and only one on all six of them. The clinicians rated 17
families in the former category and seven as poor in all areas.
Thus, it seems that the clinician is no more critical than the adolescent but
much more so than the parents. There was, however, good agreement across the
three raters about which families have the greatest and the least difficulties,
particularly between clinician and parent; five of the seven families rated the
most dysfunctional by the clinician being included in the seven rated moderately
dysfunctional by the parents.
Is family functioning a correlate of severity?
Using the MROAS global score as a measure of severity, there are no significant
associations between the rating of family functioning (by any of the three
accounts: subject-rated and family-rated FAD and clinician-rated McSIFF) and
severity of disorder. However, the global score is a composite of subscales that
correlate in different directions with the FAD scores and more illuminating
results arise from considering the subscales separately. Taking the clinicians'
(McSIFF) ratings first, there are no striking associations with MROAS subscale
scores other than a positive correlation between poor family functioning and
poor social adjustment, which in the MROAS model includes family relationships
(Table 2(a)). (The McSIFF and MROAS yield higher scores for healthier functioning.)
Thus, there is some cross-validation for the two independent ratings of family
difficulty. The lowest correlations (virtually zero), are between the FAD scores
and the scale dealing with the patient's body weight. Thus, the clinician does
not associate family difficulties with the thinness of the patient.
Table 2. Morgan-Russell Outcome Assessment Schedule sub-scales. Correlations
between ratings of family functioning and severity of anorexia nervosa
(Pearson's correlation coefficients)
The mothers (Table 2(b)) also do not show striking associations between the
severity of their son's or daughter's anorexia and their perception of family
difficulties. The second column suggests, however, a trend towards an association
between higher weight and more family problems. (The FAD is scored in the
opposite direction to the McSIFF so a higher FAD score indicates greater family
The most remarkable findings arise from the subjects' own self-report (Table
2(c)). Although we must caution against selecting significant findings from a
large table, the following appear to be consistent findings across the range of
family assessment subscales. The relatively high positive correlation between
nutritional status and FAD scores here indicate that the subject perceives her
family has fewer difficulties where she diets energetically. The sexual
adjustment scale is a difficult scale to rate for young subjects. It probably
equates here to being sexually active in relationships which are not overtly
abusive. Here then, the subject says there are fewer family difficulties where
she is sexually inactive. Finally, the second column indicates that the subject
perceives that while any difficulty her family may have with affective issues is
unrelated to her weight, she judges that her family has difficulty resolving
problems where she is heavier.
Does family functioning change in the direction of clinical progress over one
Over the course of the year, the subjects reported a slight mean deterioration
in their families' functioning on each of the seven FAD subscales. The parents
reported a more mixed picture, although as many reported deterioration as
improvement in family functioning. Meanwhile, there was a mean improvement in
the clinical condition as measured by the MROAS average outcome score and weight
change (mean weight change=6.88 kg). The detailed clinical outcome of this
series has been reported previously (North et al, 1997) .
To address this question, the 32 cases with complete follow-up data were divided
into three categories by change in MROAS average outcome score: those deteriorating
over the course of one year (n=6), those showing no change or a slight
improvement of less than three points (n=17), and those showing a significant
improvement of greater than three points (n=9).
There was no association between change in FAD scores over one year, rated by
either subject or parent, and the global severity change score, or with weight
change. McSIFF change scores were not obtained. Table 3
Table 3. Association between clinical change in one year and change in
self-reported family functioning (n=32)
There are no significant trends across the three groups. A positive value
represents a deterioration in self-rated family functioning. A negative value
represents an improvement. While the FAD change scores were not associated with
change in MROAS score, we have previously reported that presenting FAD scores
predict clinical outcome at one year (North et al, 1997) .
The assessment of family functioning in anorexia nervosa remains controversial
and is fraught with methodological difficulties, despite an increasing number of
reports of family difficulties (Waller et al, 1990a; Dare et al, 1994) [18,4].
This study is based on family members' own report, both by means of questionnaire
and through their own answers and examples given to the interviewer in the
McSIFF interviews. Although the McSIFF is based on the families' own accounts,
the clinicians in this study appear to be more critical than the parents of the
family's functioning, although no more so than the subjects themselves. The
subjects and the clinicians described significant difficulties in family
functioning in many cases. This level of difficulty is higher than has been
previously reported (Stevenson-Hinde & Akister, 1995)  and reflects a view
not shared by parents. The finding that teenagers are more critical of their
family functioning than parents may not be specific to this disorder (North et
al, 1995),  although the possibility that subjects with eating disorders are
more insightful of family difficulties than their parents has been raised by
Waller et al (1990b) . Although the McSIFF and FAD instruments are based on
the same theoretical model and many items from the FAD are included in the
structured interview, it is difficult directly to compare whether the two are
measuring the same phenomena given the confounding issue of a change in rater
and change in form of administration. Although the McMaster items are designed
to measure present functioning (i.e. measure the state of family functioning),
it is possible that some measure trait variables which hence are insensitive to
If difficulties in family functioning were a consequence of having a severely
afflicted adolescent with anorexia nervosa, one might expect an association
between severity of the condition and difficulties in family functioning. This
is not the case. Although agreement between the family members is poor, they
tend to be less critical of their families where the clinical condition is more
Reported family functioning changed, on average, little over the course of one
year despite a very variable clinical course. The nine subjects who made a very
substantial clinical improvement in this time reported a small mean deterioration
in family functioning, while the parents showed no association between change in
the family and clinical change. It is probable that any relief the recovering
person with anorexia brings about for the family does not change its basic
functioning and is tempered by other challenges for the family.
Our present findings are based on an adolescent population with relatively short
histories. Our findings may not apply to older suffers or chronic cases in which
the nature of family functioning may have altered over the course of the
- Family difficulties in anorexia nervosa do not appear to be merely a
consequence of concern arising from the condition.
- Self-reported family difficulties do not remit with the patient's clinical
- A supportive non-critical approach to the assessment of family functioning
should be adopted when planning therapeutic interventions.
- Family factors are only one of a host of contributory factors in a multi-determined
- In the absence of prospective studies of individuals at risk, it is not
possible to conclude a causative role for family difficulties.
- The McMaster model may not be uniformly sensitive to change.