Q J Med 1998; 91:493–503
Treatment of compulsive behaviour in eating disorders with
intermittent ketamine infusions
I.H. MILLS, G.R. PARK1, A.R. MANARA1 and R.J. MERRIMAN
From the Department of Medicine, University of Cambridge and 1Department of Anaesthetics,
Addenbrooke’s Hospital, Cambridge, UK
Received 15 December 1997 and in revised form 8 May 1998
Summary
We have previously shown that eating disorders are
a compulsive behaviour disease, characterized by
frequent recall of anorexic thoughts. Evidence suggests that memory is a neocortical neuronal network, excitation of which involves the hippocampus,
with recall occurring by re-excitement of the same
specific network. Excitement of the hippocampus by
glutamate-NMDA receptors, leading to long-term
potentiation (LTP), can be blocked by ketamine.
Continuous block of LTP prevents new memory
formation but does not affect previous memories.
Opioid antagonists prevent loss of consciousness
with ketamine but do not prevent the block of LTP.
We used infusions of 20 mg per hour ketamine for
10 h with 20 mg twice daily nalmefene as opioid
antagonist to treat 15 patients with a long history
of eating disorder, all of whom were chronic and
resistant to several other forms of treatment. Nine
(responders) showed prolonged remission when
treated with two to nine ketamine infusions at intervals of 5 days to 3 weeks. Clinical response was
associated with a significant decrease in Compulsion
score: before ketamine, mean±SE was 44.0±2.5;
after ketamine, 27.0±3.5 (t test, p=0.0016). In six
patients (non-responders) the score was: before ketamine, 42.8±3.7; after ketamine, 44.8±3.1. There
was no significant response to at least five ketamine
treatments, perhaps because the compulsive drive
was re-established too soon after the infusion, or
because the dose of opioid antagonist, nalmefene,
was too low.
Introduction
We have previously shown,1 by means of a questionnaire, that patients with eating disorders have the
same compulsive features as those with Compulsive
Personality Disorder as described in DSM-III.2
However, the compulsive patients did not have the
features which are characteristic of those with eating
disorders.
The Coping Questionnaire was constructed from
a collection of 76 questions which had a bearing on
coping behaviour mechanisms, but it was slanted
towards patients with anorexia nervosa. The questions related to depression, abnormal attitudes to
food and compulsive traits such as compulsive
thoughts and compulsive working.
The compulsion score was calculated from 13
questions which were shown to be answered signific-
antly differently by 28 patients defined as having
Compulsive Personality Disorder in DSM-III and 28
controls matched for age and social group. The same
13 questions were answered significantly differently
by anorexics and controls, although the questions
were not specifically related to eating problems.
Scores were also calculated for sub-scales such as
Depression, Compulsive eating, Compulsive starving
and an Alcohol score. The scores for the sub-scales
in a control group of 42 subjects are taken from
Mills & Medlicott.1 Hence, the Coping Questionnaire
was not intended to be comparable to other existing
questionnaires which had been designed only for
patients with eating disorders. The questionnaire was
shown to be a stable instrument with a high coefficient of concordance on retesting and a high success
Address correspondence to Professor I.H. Mills, Department of Surgery, University of Cambridge, Douglas House,
Trumpington Road, Cambridge CB2 2AH
494
I.H. Mills et al.
in allocating depressed patients, compulsive personality patients and those with eating disorders into
correct groups. These groups were significantly different from control patients.
In the studies with our Coping Questionnaire we
found that the question ‘Are you bothered by recurrent thoughts which you have difficulty getting rid
of?’ was answered very significantly differently by
controls and patients with compulsive behaviour
( p<0.001). Long-standing patients with anorexia
nervosa often say ‘I am tired of thoughts about
weight, diet, size and food; can you not help me to
get rid of them so that I can think about other
things’? This recall mechanism is activated involuntarily. The features of the patients with compulsive
behaviour indicate that they are in an almost permanent high arousal state, with sleep disturbance, mental
and physical restlessness.
For recall to take place, the material must first
have been put into long-term memory. For some
time there has been evidence that the memory
mechanism involves the hippocampus. Bilateral
removal of the hippocampus in a patient with
intractable epilepsy left him with only very shortterm memory.3 Several other studies confirmed this.
Repetitive electrical stimulation of hippocampal
CA1 pyramidal cells leads to a large increase in the
excitatory post-synaptic potential which persists for
hours. The more the stimulation, the higher the
potential and the longer the excited state lasts:
indeed, in a long-term study in rabbits it was shown
to last for 16 weeks. This is referred to as long-term
potentiation (LTP). Facilitation occurs, but only on
the stimulated pathway, at the synapse at which
glutamate is the transmitter and the receptor is of
the NMDA (N-methyl-D-aspartate) type. Recent work
has shown that stimulation of the metabolic glutamate receptors, leading to Ca-mobilizing responses,
must also occur at the same synapse. These receptors
have been named metabotropic glutamate receptors
(mGluR).4
The characteristics of hippocampal LTP are those
which would be appropriate for information storage
as in memory. The proof that memory and recall are
related to activation of the NMDA receptor and LTP
was provided by Morris, Davis and Butcher.5 They
used the drug AP5 (2-amino-5-phosphonopentanoate) which blocks the induction of LTP but does not
affect it once it has been established. Infusion in rats
of AP5 into the cerebral ventricle for up to 17 days
blocked LTP when the dose of AP5 was high enough,
and in the same rats prevented learning the method
of escape in a tank of water with a submerged
platform. However, once LTP had been established
and the method of escape learnt, AP5 did not prevent
recall of the previously learnt escape method.
The mechanisms of memory and recall are import-
ant if the repeated, involuntary recall of anorexic
ideas are an essential component of the compulsive
behaviour in anorexia nervosa. Recently Fuster6 has
described in detail the concept and proposed structure of the memory network. This is a complete
divergence from the usual anatomical concept of the
memory process. Fuster summarizes the new ideas
as follows: ‘Our memories are networks of interconnected cortical neurons, formed by association, that
contain our experiences in their connectional structure … Recall, recognition and working memory
consist largely in their reactivation, also by association.’
The neurones involved are in the neocortex. They
include sensory receptive nerve cells in the more
posterior parts of the brain, with extensive association
areas, and motor related functioning cells in the
frontal regions, with appropriate association areas. It
is the very extensive interconnections between cortical neurones in various areas that builds up the
network of related memories. Corticocortical connections between the sensory and motor areas effect
any necessary response to the sensory input.
Electrical stimulation of the posterior association
areas can, in conscious patients, produce the recall
of memories of perceptions. The hippocampus is
crucial in the establishment of memory networks in
the associative areas and the excitatory process (LTP)
ensures the activation of the appropriate pathways.
The cortical cells also have a glutamate excitatory
mechanism including NMDA receptors.
The process of recall involves re-excitement of
the initial memory network and its associations. The
hippocampus appears to play a part in this, since
patients with hippocampal lesions are poorly able to
recall some old memories although well-learned
habits are easily retrievable, but these are more in
the nature of reflex motor responses. Voluntary recall
of old memories normally has to be triggered by
some component in the appropriate network. The
hippocampus appears to be necessary for the
re-excitation of all the original network of that
memory. Once the recollected memory has served
its purpose, it would not be recalled unless a decision
was made that it was again required. The establishment of new memory requires the activity of the
hippocampus in spreading the excitation along the
appropriate network paths. Similarly these memories
would not be recalled until it was felt necessary.
The ease of recall is related to the state of mental
arousal when the material is first presented e.g.
questions asked at an oral examination (when arousal
is high) are readily recalled by the candidate for
many years.
Our concept of involuntary recall stems from the
study we carried out to establish what the girls were
doing when they first embarked on ‘crash dieting’.7
Ketamine treatment of eating disorders
Many of them told the same story. They had started
dieting (to attain the fashionable slim image) and for
some time it was well controlled until ‘suddenly, the
diet took over’. From that point on they were no
longer able to control their food consumption; they
felt compelled to restrict it. Analysis of the data
showed that these were girls who were working for
examinations and all had high aspirations for themselves, that is, they were perfectionists.
They appear to have discovered what the physiologist Benedict described in 1915 in his study of
fasting.8 It raised arousal to a plateau, reached in 2
weeks, and interfered with sleep. In this state the
brain was more efficient in solving problems with
less sense of fatigue. The girls welcomed the
increased brain efficiency, not realizing that it
depended on ‘the diet taking over’.
Involuntary recall might be due to more intense
excitement of the pathways (by starvation) similar to
the raised potential with repeated electrical stimulation. Each recall would then lead to the hippocampus
re-activating the excitation of the memory paths and
networks, increasing still further the potential of the
facilitated pathway and leading to repeated, involuntary recall. In the eating disorder patients, once the
critical point has been reached (when the ‘diet took
over’) the constant recall of ideas about weight, size,
food, diet, etc. would lead, via the corticocortical
connections to the forebrain, to the implementation
of the behavioural response, that is, the eating
disorder.
The NMDA receptor in the hippocampus, which
is a critical component of the excitatory post-synaptic
potential, can be blocked in a number of ways; one
of these is by the action of ketamine, the so-called
dissociative anaesthetic.9 This action of ketamine
also occurs in the presence of naloxone. Prior
treatment of patients with naloxone markedly
decreases the anaesthetic response to ketamine.10
Blocking long-term potentiation can prevent new
memories being put in place, as Morris et al.5 have
shown, but such a block did not prevent recall of
memories which were in place. Hence, ketamine
could not affect memories already established nor
prevent their recall. It could prevent new memories
being established but this would not affect the
compulsive drive. We postulated that the high level
of arousal associated with the ideas of anorexia is
what causes the involuntary recall. This recall, as
Fuster points out, requires the action of the hippocampus to excite all the initial network pathway. In
doing so it would stimulate the already high arousal
still further and facilitate repeated recall. Blocking
this further excitation with ketamine would decrease
the arousal and lessen the drive to repeated recall.
Since ketamine prevents all new memories being
established, it would not be appropriate to use
495
continuous ketamine treatment, so we used intermittent infusions. We postulated that this would prevent
further stimulation of the NMDA receptors which
are essential for LTP to be enhanced. It would then
break the cycle of recall—re-stimulation of longterm potentiation—reinforcement of compulsive
behaviour—further recall. By the use of the orally
active opioid antagonist nalmefene, it was hoped to
prevent loss of consciousness during the ketamine
infusion.
The patients selected were all severely affected
and selected because other forms of treatment had
failed to produce a remission. The patients were
aware of the concept that the basis of their illness
was a form of compulsive behaviour. The reason for
the use of ketamine was explained to them in the
presence of a senior nurse as witness. Each patient
then gave written informed consent, and each time
ketamine was to be repeated, the patient signed
a new consent form in the presence of the nurse.
The study was approved by the District Ethical
Committee.
Methods
The patients studied were not typical anorexics,
because they had failed to respond to various forms
of treatment over many years. They were older than
the normal eating disorder patients (see Table 1) and
were in a chronic refractory state.
The patients were treated in a general Medical
ward. All the patients were on amitriptyline, as
previously described.11 On the morning of treatment,
the patient had no breakfast, but took any essential
drugs and 20 mg nalmefene orally at 8 am (about
an hour before the ketamine). The patients were all
on nalmefene 20 mg b.d. since we had shown that
some patients gain weight on continuous naloxone
infusion by virtue of its anti-lipolytic action.12 They
were not gaining weight at the start of this study.
The dose of ketamine and duration of infusion
was decided by the consultant anaesthetist (GRP) so
that it was similar to that used to produce analgesia
(not anaesthesia) in hospital patients13 and battle
casualties.14 The infusion was delivered at 20 mg per
hour for 10 h. An initial bolus of the infusion was
delivered until the patient appeared sedated. At no
time did the patient lose consciousness. A second
dose of nalmefene was given at 6 pm: the duration
of action of nalmefene is approximately 12 h.
Hallucinations were produced in two patients: in
one who was given only 10 mg of nalmefene, before
we knew that this was insufficient, and in a very
underweight anorexic woman in whom the bolus
was given too fast. Subsequently in 61 treatments,
no patient experienced hallucinations either during
496
Table 1
I.H. Mills et al.
Age, diagnoses and duration: other clinical data in brackets
Patient
Age (years)
Diagnosis
1
39
2
27
3
4
32
23
5
6
33
40
7
8
39
42
9
24
10
25
11
12
34
34
13
42
14
36
15
29
Anorexia nervosa
Bulimia
Anorexia nervosa
Obsessional-compulsive neurosis
Bulimia
Anorexia nervosa
Anorexia nervosa
(Compulsive eating)
Anorexia nervosa
Anorexia nervosa
(Compulsive eating)
(Compulsive vomiting)
Anorexia nervosa
Anorexia nervosa
(Compulsive thirsting)
Anorexia nervosa
Bulimia
Alcoholism
Anorexia nervosa
Bulimia
Alcoholism
Anorexia nervosa
Anorexia nervosa
Diabetes mellitus
(Compulsive eating)
Anorexia nervosa
(Compulsive working)
Anorexia nervosa
Bulimia
Anorexia nervosa
(Compulsive eating)
or after the ketamine infusions. Blood pressure was
recorded throughout the ketamine infusion either
manually or automatically with a Datascope Accutorr
1. Only small changes in blood pressure were found.
Further treatments with ketamine were given,
depending on the clinical response. The intervals
between treatments were usually between 5 days
and 3 weeks.
Nalmefene was obtained from Sanofi, Paris, and
made up into capsules by the hospital pharmacy.
Results
There were some side-effects to ketamine. In about
80% of patients, the initial treatment caused headache, but only one patient suggested stopping the
ketamine. After the first treatment, headache was
uncommon. Nausea occurred in about 30% of
patients but was not severe: it did not occur after
the second treatment. After 1–2 h the patients were
allowed to eat and drink as they wished, and any
Duration (years)
5
10
7
7
7
19
14
10
10
7
21
19
12
8
13
other drug therapy was continued orally. The sensation of sedation was regarded by some people as
unpleasant, but after the first ketamine infusion, most
thought it was a relief from the high arousal state
they had previously been in. Revival of long-distant
memories occurred in some patients, but they saw
them as memories and were not confused by them.
After the ketamine, the patients remained more
sedated, and the dose of other drugs, such as
amitriptyline, had to be reduced. All the patients
had previously been treated with amitriptyline, not
only because fighting their compulsions frequently
led to depression, but also because amitriptyline is
a partial suppressant of the action of excitatory
amino-acids, which would help to block the action
of glutamate in stimulating the NMDA receptors
leading to long-term potentiation.15 The speed with
which a person changed from the high arousal,
compulsive state to the more normal state was
occasionally fast enough to resemble a change in
personality. This was particularly so with patient 8,
whose husband described her as changing back to
497
Ketamine treatment of eating disorders
the personality he knew when he first married her.
The patient found the speed of change difficult to
adapt to in the first week.
A total of 15 patients were treated: diagnoses
based on DSM III and the paper by Russell on
bulimia,16 are shown in Table 1 and some clinical
features are given in brackets. The patients are
divided into responders and non-responders. The
nine responders showed a marked and sustained
clinical response which persisted long after discharge
from hospital. This was shown by a return to normal
eating behaviour and the acceptance of normal
weight relative to height. They found it easier to
maintain social contact, and discussed plans for
their future.
The Compulsion scores, weights and state of
menstruation before and after ketamine and the
number of ketamine treatments are presented in
Table 2, divided into the values in responders and
non-responders. The normal Compulsion score in 42
control subjects (mean±SE) was 13.1±1.1 and is
given for comparison in Table 2 (data from Mills &
Medlicott1). In the responders, the mean±SE values
for Compulsion scores were significantly different
before and after ketamine ( p=0.0016): in the nonresponders they were not significantly different.
Table 2
However, the Compulsion score after ketamine in
responders was still significantly higher than the
value in the controls ( p<0.001).
The changes in weight depended upon whether
the patient had bulimia or compulsive eating or was
a restricting anorexic. The former group showed a
loss of weight with ketamine but the latter had a rise
in weight. In the non-responders, one patient with
bulimia (patient 10) had a marked loss of weight
with ketamine, but she could not sustain this afterwards. The other patient with bulimia had a further
increase in weight.
Complete amenorrhoea occurred in the five anorexics, numbers 3 to 7. All except one had subsequent
return of menstruation but only in patient 5 was it
regular. In patient 7, amenorrhoea continued, probably because her weight only barely reached normal.
In the other responders, irregular menstruation was
present before and after treatment, clearly related to
their weights being well maintained part of the time.
In the non-responders, the two who had amenorrhoea before ketamine had no change after it. The
rest had irregular menstruation before and after
ketamine infusions.
The mean number of infusions in the responders
was 4.1±0.8 and in the non-responders 8.3±1.4,
Compulsion scores, weights and type of menstruation before and after ketamine infusions
Compulsion score*
Weight
No. of
infusions
Before
After
Before
After
Responders
1
2
3
4
5
6
7
8
9
Mean (±SE)
t test
36
38
39
48
55
42
45
56
37
44.0±2.5
p=0.0016
8
16
31
40
33
40
22
28
25
27±3.5
80.1
62.5
44.6
45.4
44.2
45.1
44.6
51.3
47.6
76.3
56.8
50.2
50.1
50.2
50.7
47.2
52.6
47.2
Non-responders
10
11
12
13
14
15
Mean (±SE)
t test
40
52
55
36
32
42
42.8±3.7
p=NS
37
53
52
37
40
50
44.8±3.1
86.7
32.8
48.0
49.4
56.5
37.5
70.7
34.1
50.1
50.0
62.4
38.2
Menstruation
Before
After
2a
2a
3b
9b
5b
4b
2b*
4c
6d
4.1±0.8
Ir
Ir
Amen
Amen
Amen
Amen
Amen
Ir
Ir
Ir
Ir
Ir
Ir
Reg
Ir
Amen
Ir
Ir
7e
15
8
6
8
6
8.3±1.4
Ir
Amen
Ir
Ir
Ir
Amen
Ir
Amen
Ir
Ir
Ir
Amen
* Control value (Mills & Medlicott, 1992)1=13.1±1.1.
a Compulsive eating stopped 2 years. b Weight rose to normal and stable. b* Weight not quite normal but fed baby normally.
c Compulsive thirsting stopped. d Alcohol score fell 19 to 2 and remained low for 6 months. e Weight loss not maintained.
Ir, irregular; Reg, regular; Amen, amenorrhoea.
498
I.H. Mills et al.
which reflects the hope that more infusions might
eventually be effective.
Table 3 shows the mean±SE values for the subscales in controls, responders and non-responders.
The Depression score was highly significantly less
following ketamine in the responders ( p=0.0027),
but the difference was less significant in the
Compulsive eating ( p=0.035) and Compulsive starving ( p=0.043) scores. None of the differences in
scores was significant in non-responders. The Alcohol
scores were often zero, so the change in the responders was not significant. Only patient 10 had a score
in the non-responders, namely 22 (the maximum
possible) before and 20 after ketamine. This was
further evidence of her non-response in spite of the
weight loss in hospital.
In the responders, the Depression score after
ketamine was not significantly different from the
value in the controls, but the Compulsive Eating
score was still significantly different ( p<0.005) as
was the Compulsive Starving score ( p<0.01).
Responses may be seen in both the Compulsive
Eating and Starving scores because two of the five
questions in the Compulsive Eating scale also loaded
on the Compulsive Starving scale. Although the
Compulsive Eating scale was shown to distinguish
reliably between the controls and bingeing anorexics,
it only distinguished the two types of anorexics
(those who binged and those who did not) when
the Eating score was above 9.5 (Mills & Medlicott1).
Overall the responses to ketamine depended upon
the type of anorexia, especially whether they binged
or not. Patients 1 and 2 had bulimia and had not
succeeded in controlling their weight by diet or by
vomiting; they had not responded to drugs or psychotherapy. The second patient had had bulimia and
obsessional-compulsive neurosis for 10 years with
complex rituals in relation to cleanliness. Both
patients responded to two ketamine infusions with
good control of their compulsive eating. The second
Table 3
patient also stopped her obsessional-compulsive
neurosis (Figure 1). In 24 months of follow-up,
neither showed any relapse.
The restricting or starving anorexics are five
patients (numbers 3 to 7). They had been in hospital
and gained some weight, but had reached the point
where they were too afraid to allow further weight
gain because they felt it would get out of control.
This was the pattern of previous treatments in hospital. In patient 3, after three ketamine treatments
she was happy to eat enough to get her weight up
to 51 kg and she held this as an out-patient for a
year in follow-up (Figure 2). She was married shortly
afterwards.
Patient 4 increased her weight after three treatments but agreed to more in the hope that her
compulsion score would come down further. After
a year in follow-up with a normal weight, she then
started work and remained in a stable state while
followed-up for nine months.
Patient 5 was a married woman and reached a
normal weight after five treatments. As an outpatient, her periods returned and she had a successful
pregnancy. Patient 6 had had a long history of
alternating anorexia and bulimia. After four treatments and despite only a small fall in compulsion
score, she became able to control her eating and
her weight. She held a responsible job with no
relapse during 2 years of follow-up.
Patient 7 had anorexia nervosa while at college,
but recovered sufficiently to get married and have
four children. She was unique in our experience, in
that after the birth of her fourth child she again
became anorexic and then starved her baby to the
point of preventing her growing for several months.
With difficulty, she was persuaded to come into
hospital while her mother looked after the two
youngest children. After some weight gain, she
agreed to ketamine treatment, but her response to
two treatments convinced her that she would be
Scores on sub-scales before and after ketamine
Depression
Compulsive
eating
Compulsive
starving
Alcohol
7.9±0.9
2.5±0.3
5.5±0.6
0.7±0.3
Responders (n=9)
Before ketamine
After ketamine
t test
21.1±1.5
10.8±2.3
p=0.0027
12.2±1.0
7.7±1.6
p=0.035
19.2±2.0
12.3±2.4
p=0.043
5.8±2.3**
1.8±0.8
NS
Non-responders (n=6)
Before ketamine
After ketamine
t test
20.7±2.8
19.2±2.3
NS
12.0±1.0
13.3±1.1
NS
17.0±2.4
18.7±2.2
NS
(22) (1 patient not zero)
(20)
Controls* (n=42)
* Control values from Mills & Medlicott (1992)1. ** Four patients had zero values. Data are means±SE.
Ketamine treatment of eating disorders
499
Figure 1. Patient 2. Sequential compulsion scores, depression scores and compulsive eating scores over time in months.
K, ketamine treatment.
Figure 2. Patient 3. Sequential compulsion and depression scores and weight (kg) over time in months. K, ketamine
treatment.
able to cope on return to her family. Though her
weight never became normal she remained stable
and allowed her baby to grow normally during a
year’s follow-up.
Patient 8 suffered from the relatively rare condition
of compulsive thirsting. She would cut her fluid
intake to extremely low levels and on rare occasions
had gone for a week or more with zero fluid intake.
Previously she had had anorexia nervosa with short
phases of bulimia. Compulsive thirsting was always
associated with a sharp fall in food intake. After four
ketamine treatments she drank normally and her
compulsion and depression scores fell sharply
(Figure 3). After discharge, she returned to a
demanding job and relapsed. A short stay in hospital
with two ketamine treatments restored normal eating
and drinking. A month later while she was an outpatient she had a slight relapse and was given one
further ketamine treatment. She then drank enough
fluid each day to remain out of hospital while being
followed for the next 7 months.
The last person with a successful response to
treatment was patient 9. She was a long-standing
anorexic with bulimia. She discovered, as many
compulsive eaters do, that drinking alcohol could
prevent bingeing when it was imminent. With time
500
I.H. Mills et al.
Figure 3. Patient 8. Sequential compulsion and depression scores in three separate admissions; time in months. K, ketamine
treatments.
the amount of alcohol required rose until she was
an alcoholic. At this time she was frequently violent
and in trouble with the authorities. She had a very
high Alcohol Score, namely 19, when the maximum
is 22 on our Coping Questionnaire. After five ketamine treatments her alcohol score had fallen to 2 and
remained about there for 6 months (Figure 4). She
had rare, short binge phases and on two occasions
drank alcohol without any loss of control. With the
change in her personality, she became reconciled
with her family.
The first non-responder (patient 10) was a most
unfortunate woman who at three times in her childhood was subjected to sexual abuse by three men.
She had the highest depression score of all the
patients (31) even after some months on amitriptyline.
She had developed bulimia and was dependent
upon alcohol.
Patient 11 was an extremely severe restricting
anorexic of long standing. Though declaring her
desire to get better, she refused to co-operate with
behaviour therapy and frustrated attempts to use
other forms of therapy.
An insulin-dependent diabetic since childhood,
Figure 4. Patient 9. Sequential compulsion and alcohol scores and weight (kg) over time in months. K, ketamine treatments.
Ketamine treatment of eating disorders
patient 12 had become skillful at manipulating her
diet and insulin to decrease her weight or to binge
without serious metabolic problems. She was highly
intelligent and had no difficulty with examinations
before going to university. She frequently had the
maximum score on the Compulsion scale (55). She
had great difficulty with social relationships and
would cry at the thought of never having had a
boyfriend.
Patient 13 was nearly 30 years old when she
developed anorexia but for many years she had been
a compulsive worker. As a midwifery sister, she had
difficulty delegating duties to her staff and eventually
had a depressive breakdown.
With hindsight, patient 14 was a typical bulimic
but for long periods she totally denied bingeing. She
had friction with her husband when he wanted to
know where the money was being spent.
Patient 15 was a typical anorexic whose problems
began when working for examinations. She was
offered a university place but was unable to take it
up because of her health. She trained as a laboratory
technician but was constantly wanting to work for
other qualifications. When not studying she almost
got her weight to normal, but as soon as she started
studying again she became anorexic and had intermittent bingeing. She did not vomit but took large
doses of phenolphthalein-containing laxatives. She
invariably achieved distinction marks in the examinations.
Discussion
We regard eating disorder (anorexia and bulimia) as
a form of compulsive behaviour in people with a
perfectionist personality, as we previously described.1
In these patients, the drive to cut calorie intake
severely has become beyond normal control.
Psychotherapy as a sole treatment was frequently
unsuccessful in the most severe cases. Although
many patients can be treated as out-patients with a
variety of treatments, the most severely affected need
in-patient treatment. The most intense anorexics have
constant relapses, and may go from one medical
practitioner to another hoping for a magic cure.
Those who refuse to take part in behavioural therapy
are the most resistant to treatment. All the starving
anorexics in this series had had prolonged relapsing disease.
The severity of compulsive behaviours frequently
disrupts whole families. An attempt to find a neurophysiological basis for such behaviour led us to the
idea that long-term potentiation (LTP) in the hippocampus might be related to the compulsive drive.
The neuronal network concept of memory6 which
involves excitation of neo-cortical pathways specific
501
to each memory, could lend support to the suggestion
that LTP plays a part in compulsive behaviour
diseases, especially since the cortical cells also have
NMDA receptors as occur in the hippocampus.
Specific networks are set up for each memory,
even though they may have partial overlap of neurones, as Fuster6 points out. The more intense the
initial excitation of the pathway and network, the
more likely is recall to become involuntary. Only
repeated, involuntary recall would be expected to
potentiate the compulsive drive, and hence the
behaviour of the eating disorder patients, as indicated
in the Introduction.
Decreased LTP would lessen the intensity of
repeated recall. Continuous inhibition of LTP blocks
new memory formation in animals, so we decided
on intermittent ketamine therapy to effect partial
block of the NMDA glutamate receptors, and thus
decrease LTP in relation to the compulsive ideas.
The opioid antagonist nalmefene allowed ketamine
to be infused without loss of consciousness, as shown
previously with naloxone.10 It also prevented hallucinations during and after ketamine when the dose
of nalmefene was 20 mg b.d. These features made
ketamine therapy acceptable to the patients.
The mean Compulsion score decreased significantly after ketamine treatment in the responders ( p=
0.0016). The behaviour of the patients with a fall in
Compulsion score showed a sustained improvement
and a move to a more normal state, in that they
were no longer driven by the compulsive ideas.
Patient 8 (Figure 3) showed the ease with which a
patient might slip back. Prompt action and repetition
of a few ketamine treatments established a more
secure, non-compulsive state.
The dose of ketamine used in these patients was
similar to that used to provide analgesia in both
hospital patients13 and in battle casualties.14 In neither case has haemodynamic instability or unconsciousness been reported as a significant problem.
Despite its safety, certain precautions were deemed
essential. The infusion was always established, and
the initial bolus, given, by an anaesthetist who was
available throughout the infusion. The patients were
fasted overnight, apart from their medication, and
for 2 h from the start of the infusion. Ketamine has
been shown to increase circulating catecholamines17
and initially concern over the concurrent administration of tricyclic antidepressants led to intensive
monitoring using an automated pressure device with
alarms and ECG monitoring. In the early studies, a
nurse was present throughout the infusion. It became
apparent that a drug interaction using ketamine at
this dose did not occur, and less intense monitoring
was sufficient.
Only the most severely affected patients with
compulsive disorders are likely to need ketamine
502
I.H. Mills et al.
treatment. It is expensive because it requires
in-patient treatment and a well-trained team of nurses
as well as dedicated doctors to give the strong
support that some patients, and perhaps their relatives, need in the first one or two infusions. It would
be unwise to embark upon treatment without the
preparation and training of nursing and medical staff
to cope with possible side-effects and the changes
in the patient’s approach to life. Patients who can
be effectively treated with cognitive behavioural
therapy, or in-patient behavioural therapy or amitriptyline, etc. should be excluded. Since some restricting
anorexics put on weight with continuous opioid
antagonist,12 this group should also be excluded
before ketamine is considered. The patients treated
here had not responded to amitriptyline or nalmefene
prior to ketamine therapy.
Six of the patients treated fell into the nonresponder group. Patient number 10 was clearly
much affected by her childhood sexual abuse. Her
bulimia was severe and she had a maximum Alcohol
score (22). Her high and unresponsive Depression
score (31) may indicate a cause for failure to respond.
One of failures was a classical restrictive anorexic,
albeit a very severe one (number 11). She refused to
take part in behavioural therapy and reduced her
calorie intake whenever other treatments led to
weight gain. All this was despite declaring she
wanted to be cured. It is possible that her severe
starvation diet led to excessive mental excitation, as
described by Benedict,8 and of the memory network,
with frequent, involuntary recall reinforcing it. This
might have reactivated the network as fast as ketamine suppressed it.
The other failure who had bulimia (patient 14)
was also severe. Part of the compulsive drive in
bulimia is probably addiction to the release of
endorphin which occurs with very high blood glucose levels generated by carbohydrate bingeing, as
shown by Fullerton and his colleagues.18 The two
with compulsive eating (patients 12 and 15), though
strictly not bulimic, might have been similarly affected. The initial drive to binge is related to craving
something to calm down their high arousal state.
Some of them will use alcohol for the same purpose.
The diabetic patient had the highest possible
Compulsion score (55). It may be that nalmefene in
a dose of 20 mg b.d. was not high enough to control
the addiction to their own endorphin. Alcoholics
can often be prevented from relapse, we now know,
by a much higher dose of naltrexone, namely 100 mg
per day.19
Since the anaesthesia produced by ketamine is
partly prevented by an opioid antagonist,10 it might
be presumed that ketamine in part causes anaesthesia
by release of endorphin, though this is not necessary
for the blockade of the NMDA receptor.9 The nalme-
fene must be high enough to antagonize any endorphin released by ketamine as well as the patients’
already high endorphin.
Why patient 13 failed is not so obvious, since her
anorexia itself was not very severe. However, on the
sub-scale ‘Compulsive work’ (see reference 1 for
description) she had near-maximum levels most of
the time, namely, a score of 17.
After successful treatment with ketamine, it was
important to remind the patients that they still had
to live with their perfectionist personality. It is this
that tends to drive them to achieve perhaps impossibly high goals. If they fail to control this excessive
drive, they will excite the same or some other
memory network to the point where involuntary
recall will again produce the behaviour which made
them suffer an eating disorder. Prolonged follow-up
may be necessary to ensure that they learn and act
on this lesson.
Although nalmefene was the opioid antagonist
used in this study, we would expect a comparable
effect with oral naltrexone.
Acknowledgements
We are grateful to the nursing staff for their dedication
in looking after these patients on a general Medical
ward.
References
1. Mills IH, Medlicott L. Anorexia nervosa as a compulsive
behaviour disease. Q J Med 1992; 84:507–22.
2. DSM III. Diagnostic and Statistical Manual of Mental
Disorders, 3rd edn. Washington DC, American Psychiatric
Association, 1980.
3. Milner B. Amnesia following operation on the temporal
lobes. In: Whitty CMM, Zangwill OL, eds. Amnesia.
London, Butterworths, 1966:109–33.
4. Bashir ZI, Bortollotto ZA, Davies CH, Berretta N, Irving AJ,
Seal AJ, Henley JM, Jane DE, Watkins JC, Collingridge GL.
Induction of LTP in the hippocampus needs synaptic
activation of glutamate metabotropic receptors. Nature
1993; 363:347–50.
5. Morris RGM, Davis S, Butcher SP. The role of NMDA
receptors in learning and memory. In: Watkins JC,
Collingridge GL, eds. The NMDA Receptor. Oxford, IRL
Press, 1989:137–51.
6. Fuster JM. Network memory. T I Neurosciences 1997;
20:451–9.
7. Mills IH, Wilson RJ, Eden MAM, Lines JG. Endocrine and
social factors in self-starvation amenorrhoea. In: Robertson
RF, ed. Symposium—Anorexia Nervosa and Obesity.
Edinburgh, Royal College of Physicians of Edinburgh,
1973:Publication no 42:31–43.
8. Benedict FC. A study of prolonged fasting. Washington DC,
Carnegie Institute of Washington, 1915.
Ketamine treatment of eating disorders
9. Stringer JL, Guyenet PG. Elimination of long-term
potentiation in the hippocampus by phencyclidine and
ketamine. Brain Res 1983; 258:159–64.
10. Stella L, Crescenti A, Torri G. Effect of naloxone on the loss
of consciousness induced by IV anaesthetic agents in man.
Br J Anaesth 1984; 56:369–72.
11. Mills IH. Amitriptyline therapy in anorexia nervosa. Lancet
1976; ii:687.
12. Moore R, Mills IH, Forster A. Naloxone in the treatment of
anorexia nervosa: effect on weight gain and lipolysis. J Roy
Soc Med 1981; 74:129–31.
13. Ito Y, Ichiyanagi K. Postoperative pain relief with ketamine
infusion. Anaesthesia 1974; 29:222–6.
14. Bion JF. Infusion analgesia for acute war injuries. A
comparison of pentazocine and ketamine. Anaesthesia
1984; 39:560–4.
503
15. Evans RH, Francis AA, Watkins JC. Differential antagonism
by chlorpromazine and diazepam of frog motor neurone
depolarization induced by glutamate-related amino acids.
Eur J Pharm 1977; 44:325–30.
16. Russell G. Bulimia nervosa: an ominous variant of anorexia
nervosa. Psychol Med 1979; 9:429–48.
17. Zsigmond EK, Kelsch RC. Elevated plasma norepinephrine
concentration during ketamine anaesthesia. Clin Pharm
Ther 1973; 14:149.
18. Fullerton DT, Getto CJ, Swift WJ, Carson IH. Sugar, opioids
and binge eating. Brain Res Bull 1985; 14:673–80.
19. O’Malley SS, Jaffe AJ, Chang G. Naltrexone and coping
skills therapy for alcohol dependence. Arch Gen Psych
1992; 49:881–7.