Brit.J.
Psychiat. (1979), 134,55—59
Behaviour Therapy in Anorexia Nervosa * By ELKE
D. ECKERT,
SOLOMON
REGINA
C. GOLDBERG,
C. CASPER
and JOHN
KATHERINE
A. HALMI,
M. DAVIS
SUMMARY In three collaborating hospitals, 81 anorexia nervosa patients were randomly assigned to behaviour modification or its absence and followed over 35 days. There was no overall significant difference in weight gain in the two groups except in a subset of patients: those with no prior out-patient treatments. Anorexia variety
nervosa
is an illness
of controversial
for which
treatments
have
a
used. Behaviour modification treatment. It seems well suited
is one such for this disorder,
especially
when
in
the
acute
phase
Thompson, 1974; Leitcnberg Garfinkel et al, 1973).
been
abnormal
systematic
de
sensitization to reduce the anxiety and allow normal eating habits to emerge,
thus and
operant
are
suggested:
conditioning
by using
powerful
al,
1968;
Bachrach, Erwin and Mohr (1965) reported an anorectic patient who learned to vomit and stopped gaining weight when food intake was reinforced, but stopped vomiting and gained weight when reinforcement was made con tingent on weight gain. Most later reports indicate use of reinforcements contingent on weight gain rather than on eating behaviour. The latter approach makes sense, not only because it avoids teaching the patient to vomit,
eating behaviour and low weight are strikingly evident. Brady and Rieger (1972) have described anorectics as behaving as though they suffer from an eating or weight phobia—eating or weight gain generates anxiety and failure to eat, or weight loss, represents avoidance of the anxiety. From this analysis two behavioural
procedures
et
but
because
it
reduces
eating by decreasing It is also easier to accurately than to intake, and there is
re
the
anxiety
around
the emphasis on eating. monitor daily weights monitor 24-hour food less chance for power
inforcers contingent on eating or weight gain, thus allowing the anxiety to take care of itself.
struggles
There are very few reports on the use of syste
by using
matic desensitization for the treatment ofanorec tics (Hallsten, 1965; Lang, 1965; Schnurcr et al, 1976) and only one case report where de sensitization is used alone (Schnurer et al, 1976). Operant conditioning procedures have been
the strong effect of regular feedback of weight and caloric intake information on weight gain. Evaluation of behaviour therapy in anorexia nervosa is complicated by several factors. There
extremely
therapeutic effectiveness. Often, medications are used concurrently with behaviour therapy so that it is impossible to see which treatment is the effective one. Diagnostic criteria are often not carefully specified, resulting in the inclusion of patients suffering primarily from phobias, hysteria, or schizophrenia in the treatment studies. Most studies have small numbers of patients. Halmi's report of eight cases success
utilized
recently,
with
powerful
negative
re
such as bed rest, isolation, and tube
feeding, and positive reinforcement such as increased social and physical activities, con tingent on food intake or weight gain (Brady
and Rieger, et al, 1970; * This
study
1972; Halmi Bachrach was
Grants MH26218,
et al, 1975; Blinder
et al, 1965;
supported
MH26310,
by
Public
Bhanji Health
the technique
is virtually
effective and have been increasingly
inforcement
and arguments.
Agras et al (1974) were able to demonstrate,
and Service
fully
and MH26409. 55
treated
of systematic
no controlled
with
a
study
systematic
analysis,
evaluating
behaviour
its
56
BEHAVIOUR
THERAPY
modification programme, using no medication and meeting strict diagnostic criteria remains the largest and purest patient sample described in the literature (Halmi et al, 1975). Recent controversies (Feinstein, 1974; Agras
et al, 1974) over the efficacy and safety of using behavioural contingencies in treating anorexia nervosa certainly
have by provided
their no
lack of objectivity insight but rather
distorted impressions of effective treatment programmes for anorexia nervosa. Thus in order to test the effect of behaviour modification on weight gain, we thought it worthwhile to design and conduct a controlled treatment study comparing an operant conditioning paradigm
with a ward milieu therapy programme containing virtually no behavioural contin gencies related to weight gain. Behaviour modificationprogramme Forty patients were treated with the same operant conditioning programme with negative reinforcement of isolation and positive re inforcement of increased social and physical activities contingent on weight gain. After the 7-day observation period they were isolated and
restricted to their rooms except for the time when they ate meals in the unit dining rooms. While in isolation they were allowed occu pational therapy projects, radio and reading, but no television. Only hospital staff were allowed in the room. Privileges for weight were earned or with drawn at the end of 5-day periods. During the first five days, the patients were required only to maintain their weight to obtain increased privileges. For the next two periods, they were required to gain at least 0.5 kg per five days. Thereafter, they were required to gain 1 .0 kg per 5-day period. If they did not meet the required weight contingency at the end of any 5-day period, they lost some of their privileges. For the first 5-day period, the patients
received no visitors, phone calls, or mail. If the patients obtained the appropriate weight, on the sixth
day
(or the beginning
of the next
5-day
period) they were allowed to make one phone call, receive their mail, have one visitor for one hour, and be out of the room for one hour each day for the next five days. By the end of the
IN ANOREXIA
next
and
NERVOSA
each
succeeding
5-day
period,
if the
weight contingency was met, they could have an additional hour of visiting privileges, receive their mail, make an additional phone call, and be out of their room each day for an additional hour. Phone calls, visitors, and mail were receivedonly on the day privileges were earned.
The patients chose which hours they spent out of their rooms. Each patient received two hours a week of individual re-educative psychotherapy with emphasis on her interaction with others. staff, relatives, and visitors were instructed
The not
to discuss food and weight with the patients. Family counselling was given if appropriate. Programmefor patients without behaviour modification
The 41 patients who were not treated with behaviour modification were treated in the milieu particular to each collaborating hospital. Each patient received the same personal therapy programme as given to patients in the behaviour therapy groups. No privileges were contingent on weight, and each patient could receive visitors twice a week for two hours. Mail and phone calls were not restricted. The staff,
relatives, and visitors were instructed not to discuss food and weight with the patients. Note that by the design of the whole experiment (see paper 1 of this series) half the patients in each programme also received cyproheptadinc. Results
We addressed ourselves to two different questions in this research. The obvious first question was whether patients with behaviour modification
gained
more
weight
over
day treatment period than patients behaviour modification. The second
the 35-
without question
was whether or not there were specificsub
groups
of patients
in which
behaviour
modi
ficationwas more effectivethan no behaviour modification. The effects of the two programmes, behaviour
modification
and no behaviour
modification
on
weight was analyzed by a fixed effects analysis of covarianceusing the post-treatmentweight as a
dependent variable and the pretreatment weight as a covariate. The adjusted post-treatment weight
could
be interpreted
as a measure
of
ELKE D. ECKERT et at
57
TABLE I
Possible predictors of weight gain Means and standard deviations
N =81 VariableMeanStandard deviationPretreatment kg5.23Normal weight35.96 weight for age and height51 kg4.84Normal minus pretreatment weight15.80 kg5.67Percent admission319Hyperactivity below normal weight on (1 = absent, 4 = severe)1 .400.81Deliver@ in childhood complications(score range 6-12, lower score = more complications)11 .680.59Number 15Number of prior hospitalizations0.931 treatments0.530.65Prior of prior out-patient weeks)0.350.66Prior hospitalizations of adequate duration (4 weeks)0.290.51Prior outpatient treatments of adequate duration (12 gain)0.410.67Prior hospitalization failures (less than 2 kg gain)0.400.54Prior outpatient failures (less than 2kg failed0.090.28Prior hospitalization of adequate duration that failed0.220.42Pretreatment out-patient treatment of adequate duration that hyperactivity on ward (1 = no, 2 = yes)1 .420.33Sleep disturbance (self-rated, 1 absent, 4 = severe)2. 190.98Denial* 1 —¿@ absent, 4 = severe)1 .790.68Loss(self-rated, of appetite (self-rated, I absent, 4 = severe)2 .630.63Psychosexual
.880.51Body
immaturity*
From
Anorectic
Attitude
size :actual
Scale
ficant (P = .20). However, it must be noted that patients overall gained a fair amount of weight in both groups over 35 days.
BEHAVIOR @@L@DD@N 32@ NO
Q
.
(self-rated, 1 = absent, 4 = severe)1
size,sumof5measures606.73136.41 image distortion (N 79) perceived
*
.05
In order
MOD(N@?€l
to test whether
of patients modification, pretreatment predictors
@1i@
(Goldberg NONE
ONEORMORE
NUMBER OF PRIOR OUTPATIENT TREATMENTS OF ADEQUATE DURATION
variables, standard
a specific
subgroup
responded selectively to behaviour social and psychiatric history and variables shown to he possible of weight gain in prior research
et al, together deviations,
1977)
were
assessed.
with their are presented
These
means and in Table I.
Tests for homogeneity of regressionbetween FIG 1.—Interaction number of prior
between behaviour modification and outpatient treatments of adequate (12 weeks) duration. (P = .065: N = 81).
change. This analysis revealed the average weight gain for the 35-day treatment period was 5.0 kg for patients treated with behaviour modification, and 4. 1 kg for patients without behaviour modification. Although patients with
behaviour modificationgained somewhat more weight, this resultwas not statistically signi
behaviour modification and non-behaviour modification for the regression of weight gain on each of these variables revealed an interaction approaching significance between behaviour
modification patient
1 shows
and
treatments
the
the
number
of adequate
interaction
of prior duration.
between
out Figure
behaviour
modification and the number of prior outpatient treatments of adequate duration. An outpatient treatment was considered adequate if it was at least 12 weeks in duration. There was a differ
BEHAVIOUR THERAPY IN ANOREXIA NERVOSA
58
ence
approaching
significance
(P
favouring behaviour modification with no prior out-patient treatment
=
.065)
in patients of adequate
duration. Of all patients with no prior outpatient treatment, patients with behaviour modification gained an average of 4.9 kg during the 35-day treatment period, while patients not treated with
effective than a schedule of delayed reinforce ments given only every fifth day. It is known that the longer the time between the completion of a behaviour and the delivery of a reinforcing
consequence,
the less effect the reinforcer
will
have (Rimm and Masters, 1974). A preliminary analysis of an uncontrolled study at the Uni behaviour modification gained an average of versity of Minnesota indicates that anoretic 3.3 kg. For allpatientswith one or more prior patients may gain more weight on a daily reinforcement schedule than on a delayed out-patient treatments of adequate duration, schedule (Eckert, in preparation). there was a negligible difference between In this study there was no difference in the treatments: those on behaviour modification gained 5.3 kg and the non-behaviour modi efficacyof behaviour therapy and the ward fication group gained 5.8 kg. milieu programmes for inducing weight gain. It isconceivablethat our variousward milieu Discussion programmes and isolation may have produced a maximal possible weight gain so that behaviour This is the first controlled randomized therapy could not be expected to do any better. treatment study testing the efficacy of behaviour
modification nervosa.
in
These
preliminary
the
treatment
results
must
of anorexia be
References
considered
since they are based on only 80 per
AGRAS, W. S., BARLOW, D. H. & CHAPIN, H. N. et at (1974)
Behaviour modification of anorexia of General P@yc/natry,30,279-86.
cent of the total expected sample and the
consistency of the results among the three collaborating hospitals has not been examined. Behaviour modification for anorexia nervosa has been widely supported as producing rapid weight gain in the acute phase of the illness. This study shows that there was weight gain in both behaviour modification and non-behaviour modification groups. However, contrary to our expectations, although patients on behaviour modification gained somewhat more weight than patients not on behaviour modification, the difference was not significant. It will be interesting to seeifanalysisof the totalexpected sample of 106 patients will change this differ ence to significant levels in the expected directions.
Possible explanations for our results must be considered. It may be that another behaviour modification programme would be more effec tive than the operant conditioning programme
used in this study. Individualized reinforcers, rather than constant reinforcers in all patients have been used and may
(Brady and Rieger, Bhanji 1973).
and The
be more effective
1972; Blinder et al, 1970;
Thompson, 1974; Garfinkel ci al, schedule of reinforcements can be
varied, and it may be that a schedule of more immediate daily reinforcements would be more
—¿
S@rtmic&iw,
A.
J.
&
BLINDER,
B.
nervosa.
J.
(1974)
Archives Anorexia
nervosa. Journal of the American Medical Association, 229, 1864—5. BACHRACH, A. J., ERWIN, W. J. & MOHR, J. P. (1965)
The control ofeatingbehaviour in an anorexic by operant conditioning techniques. In Case Studies in Behaviour Modification
(eds.
L. P. Ullman
and
L.
Krasner), pp. 153—63. New York: Holt, Rinehart & Winston Inc. BssANJI, S. & THOMPSON, J.
(1974)
Operant
conditioning
in the treatment of anorexia nervosa: a review and retrospective study of 11 cases. British Journal of P4ychial,y, 124,166-72. BLINDER, B. 3., FRww4,
(1970) Behaviour effectiveness
D. M. A. & STUNI
therapy
, A. 3.
of anorexia
of activity as a reinforcer
nervosa:
of weight gain.
An2ericanJownalofP@ychiat7y, 126,1093-8. BR@Y,J. P. & RIEGER, W. (1972) Behaviour treatment of anorexia nervosa. Proceedings of the International S@nposium on Behaviour Mod jflcation. New York: Appleton-Century-Crofts. ECKERT,
E.
D.
(in
preparation)
Treatment
oe anorexia
nervosa with behaviour modification. FEINSTEIN, S. C. (1974) Anorexia nervosa. Journal of the American Medical Association, 228, 1230. GARFINKEL, P. E., KLINE, S. A. & STANCER, H. C. (1973)
Treatmentof anorexia n@rvosausingoperantcon
ditioning techniques. Journal of Nervousand Mental Diseases, 157,428—33. GOLDBERG,
S. C.,
Pretreatment
HALMI,
K. A.
predictors
& CASPER,
of weight
R.
eta!
(1977)
gain in anorexia
nervosa. In AnorexiaNet'vosa (ed. R Vigersky), pp. 31-43. New York: Raven Press.
59
ELKE D. ECKERT cial HALLSTEN,
E.
A.
(1965)
Adolescent
treated by desensitization. Therapy, 3,87—91.
anorexia
LEITENBERG, H.,
nervosa
Behaviour Research and
AGRAS, W.
S. & THOMPSON, L. E.
(1968) A sequential analysis of the effect of selective positive reinforcement
in modifying
anorexia
nervosa.
BehaviourResearchand Therapy,6,211—18. HALMI,
K.
A.,
POWERS,
Treatment
of
P.
&
anorexia
CUNNINGHAM,
nervosa
with
S.
(1975)
RIMM,
D.
C.
modification. Archivesof GeneralPsychiatry,32,93—6.
SCHNURER, LANG,
P.
J.
(1965)
Behaviour
therapy
with
a
case
of
anorexia nervosa. In Case Studies in Behaviour Modi fication (eds. L. P. Ullman and L. Krosner), pp.
217—21.New York: Holt, Rinehart & Winston Inc.
& MMrns,J.
C.
(1974)
Behaviour
Therapy:
Techniquesand Empirical Findings, p. 185. New York and London: Academic Press.
behaviour
A.
Systematic
T.,
RUBIN,
R.
desensitization
R.
&
Roy,
of anorexia
A.
(1976)
nervosa
seen
as a weight phobia. In BehaviourTherapy in Psychiatric Practice (eds. 3. Wolpe, L. 3. Reyna), pp. 29-33. New York: Pergamon Press.
*Elke D. Eckert, M.D.,Assistant Professor of Psychiatry, Department of Psychiatry, University of Minnesota Hospitals, Box 393 Mayo Bldg, Minneapolis, Solomon
C. Goldberg,
Minnesota 55455
Ph.D. Acting Chief, Psychopharmacology
Research Branch,
National
institutes
of
Mental Health, 5600 Fishers Lane, Rm. 9—105,Rockville, Maryland 20852, Katherine
A. Halmi,
M.D., Associate Professor of Psychiatry,
University of Iowa College of Medicine,
500 Newton Road, Iowa City, Iowa 52242, Regina C. Casper, M.D..Assistant Professor of Psychiatry, University of illinois, Illinois StatePsychiatric
institute, West Side Medical Center, 1601 W. Taylor Street, Chicago, Illinois 60625, John
M. Davis, M.D., Professor of Psychiatry, University of Chicago, Illinois State Psychiatric Institute, West Side Medical Center, 1601 W. Taylor Street, Chicago, Illinois 60625
*
Responsible
(Received]
for
reprint
requests.
August 1977; revised 15 March 1978)