Behavioural and Cognitive Psychotherapy, 2011, 39, 1–19
First published online 1 September 2010doi:10.1017/S1352465810000469
Competitive Memory Training (COMET) for Low
Self-Esteem in Patients with Personality Disorders:
A Randomized Effectiveness Study
Kees Korrelboom, Marlies Marissen and Tanja van Assendelft
Parnassia-Bavo Psychiatric Centre, The Hague, The Netherlands
Background: Self-esteem is a major concern in the treatment of patients with personality
disorders in general. In patients with borderline personality disorder, low self-esteem is
associated with factors contributing to suicidal and self-injurious behaviour. At the moment
there are no well-proven interventions that specifically target low self-esteem. Recently, a
new approach, Competitive Memory Training or COMET, aimed at the enhancement of
retrieving beneficial information from memory, appeared to be successful in addressing low
self-esteem in different patient populations. Aims: To assess whether COMET for low self-
esteem is also an effective intervention for patients with personality disorders. Method: 91
patients with personality disorders who were already in therapy in a regular mental health
institution were randomly assigned to either 7 group sessions of COMET in addition to
their regular therapy or to 7 weeks of ongoing regular therapy. These latter patients received
COMET after their “7 weeks waiting period for COMET”. All patients that completed
COMET were contacted 3 months later to assess whether the effects of COMET had remained
stable. Results: Compared to the patients who received regular therapy only, patients in the
COMET + regular therapy condition improved significantly and with large effect sizes on
indices of self-esteem and depression. Significant differential improvements on measures of
autonomy and social optimism were also in favour of COMET, but had small to intermediate
effect sizes. The therapeutic effects of COMET remained stable after 3 months on three out
of the four outcome measures. Conclusion: COMET for low self-esteem seems to be an
efficacious trans-diagnostic approach that can rather easily be implemented in the treatment
of patients with personality disorders.
Keywords: Personality disorders, self-esteem, memory retrieval, group treatment.
Some debate exists about the value of pursuing high self-esteem in general. While many take
it for granted that self-esteem is a pervasive motivator for adaptive and desirable behaviour
(Pyszczynski, Greenberg, Solomon, Arndt and Schimel, 2004), some researchers found
dissociations between high levels of self-esteem and good performance in several domains
of functioning (Crocker and Park, 2004). Ellis takes this argument even further, stating that
self-esteem is a myth, driving people to chase unremittingly the approval of others. He
Reprint requests to Kees Korrelboom, Head of the Department of Research and Development, PsyQ, Parnassia-Bavo
Psychiatric Centre, Stadhoudersplantsoen 2, 2517 JL The Hague, The Netherlands. E-mail: email@example.com
© British Association for Behavioural and Cognitive Psychotherapies 2010
K. Korrelboom et al.
suggests that it could better be replaced by “self-acceptance” (Ellis, 2005). Notwithstanding
these critics, high self-esteem seems to contribute to overall health and well-being (DuBois
and Flay, 2004), the more so when self-esteem is derived from self-determined standards
(Pyszczynski and Cox, 2004). While most research on self-esteem concerns the level of
(explicit) self-esteem, self-esteem encompasses many different aspects and types, such as
contingent vs. non contingent, stable vs. unstable, global vs. domain specific, and explicit vs.
implicit. It is not unlikely that part of the debate on the significance of self-esteem is due to
different definitions used for different aspects of the concept.
Apart from those general considerations, self-esteem is an issue in clinical psychology and
psychiatry. Although low self-esteem is not an emotional disorder in itself, for many patients
it is the main reason for seeking therapeutic help. In several emotional disorders low self-
esteem is one of the defining symptoms for fulfilling the formal criteria of the disorder. As far
as axis-I diagnoses are concerned, low self-esteem is part of the clinical picture of some of the
eating disorders and depression. Apart from being a formal criterion for these diagnoses, low
self-esteem is considered an aetiological factor for the development of different psychiatric
problems, its maintenance, and a predictor of relapse following treatment (Stice, 2002; Wilson
and Rapee, 2005; Schmitz, Kugler and Rollnik, 2003).
While disturbances in “self” and “identity” are often seen as core issues in all personality
part of the criteria, i.e. avoidant, borderline and narcissistic personality disorders. A diagnosis
of avoidant or borderline personality disorder contributed to having low self-esteem beyond
the level of self-esteem that would have been accounted for by co-morbid depression (Lynum,
Wilberg and Karterud, 2008). In borderline personality disorder low self-esteem appears to be
strongly associated with shame, which is considered to be one of the core characteristics in
this disorder and as being a major determinant for suicidal behaviour, self-injurious behaviour,
anger and impulsivity (Rüsch et al., 2007). Moreover, in a group of 542 psychiatric and non-
psychiatric adolescents, low self-esteem was closely related to depression, hopelessness and
suicidal tendencies (Overholser, Adams, Lehnert and Brinkman, 1995). In summary: low
self-esteem seems to be a major concern in the aetiology, manifestation, persistence in and
vulnerability for diverse axis-I and axis-II pathologies.
It is often assumed that successful treatment of the primary disorder of the patient will
automatically lead to amelioration of feelings of low self-esteem. However, this is not always
the case as many patients continue to report low self-esteem after successful treatment of their
main disorder. In these circumstances, some clinicians suggest that low self-esteem must be
tackled in its own right (Fennell and Jenkins, 2004).
Possibly due to its status of not being a specific disorder in itself, only a few methods
While many consider low self-esteem to be the manifestation of fundamental difficulties in
self-representation and identity that, if exaggerated and inflexible, can move a person from
“a personality type” to “someone with a personality disorder” (Beck et al., 2004), these
methods are time limited and straightforward. Fennell has developed a treatment procedure
that is mainly based on well-known standard cognitive-behavioural principles (Fennell, 1997;
McManus, Waite and Shafran, 2008). According to her model, experiences in life make
people formulate a “bottom line” opinion about their own self-worth. When this bottom line
is negative, this leads to low self-esteem. The bottom line has the status of a basic belief
in cognitive theory. Based on this basic belief people develop behavioural and cognitive
COMET for low self-esteem
strategies to find their way in daily life, without being bothered too much by their perceived
inadequacies. Most of these strategies have an avoidant or compensating character, which
tends to make things worse in the long run by confirming the maladaptive self-evaluations.
Fennell’s therapy consists of four phases: 1) assessment, goal setting and psycho-education;
2) application of cognitive techniques and behavioural experiments to break into maintenance
cycles; 3) re-evaluation of the conditional assumptions and “rules for living”; 4) re-evaluation
of the bottom line and promoting self-acceptance. While Fennell’s approach seems to have
found its way into clinical practice, no trials have been performed to put the efficacy of
Fennell’s method to the test (McManus et al., 2008).
On the other hand, Tarrier’s intervention for influencing low self-esteem (Tarrier, 2001) has
been tested in a trial. A convenience sample of 25 psychotic patients was randomized between
therapy as usual (TAU) and a combination of TAU and a specifically developed intervention
to enhance self-esteem. At termination of the intervention and 3 months later at follow-up,
the specific intervention had increased self-esteem, decreased psychotic functioning, and
improved social functioning (Hall and Tarrier, 2003). The method is aimed at increasing
the patient’s awareness of positive characteristics. In Tarrier’s intervention, first, positive
qualities of the patient are sampled and specific instances of these qualities are indicated.
The patient then needs to rehearse these instances by verbal descriptions or mental imagery.
Next, the patient monitors positive behaviours in daily life that are indicative for the identified
qualities. This monitoring stimulates an increase in the actual performance of these activities.
During treatment, patients repeatedly re-rate their belief that they really possess these positive
attributes (Tarrier, 2001).
Independent from and unaware of Tarrier’s method, another (albeit in several aspects
rather similar) intervention for enhancing self-esteem has been developed and put to the test
(Korrelboom, 2000). In this approach, named Competitive Memory Training (COMET) for
low self-esteem, patients are trained to make memories of actual occurrences of positive
and worthwhile characteristics better retrievable from long-term memory. The method is
a practical elaboration of Brewin’s suggestion that each concept has several meaning
representations attached to it in long-term memory and that a competition exists between
these different meaning representations to be retrieved. In psychopathology, dysfunctional
representations too often win this retrieval competition. Psychological interventions should
influence this competition in such a way that functional representations more often win
COMET is considered to be one such intervention. It is set-up as a training program and
consists of four phases encompassing 6–10 treatment sessions. First, the dysfunctional self-
opinion is identified. Second, actual instances (behaviours, characteristics, experiences) that
contradict this negative self-opinion are sampled. Third, these counter-examples are made
better retrievable from long-term memory by making them more emotionally salient and
by retrieving them frequently. Emotional salience is enhanced by focusing attention on the
positive characteristics by writing them down in short, self-referent stories, by imagining
them, and by supporting these images with a congruent body posture, facial expression,
positive self-verbalization and music that is identified by the patient with positive mood
and feelings of personal strength. The last phase of COMET is concerned with association.
Based on the principles of counter-conditioning, patients have to imagine difficult situations
that function in daily life as cues that trigger self-deprecating feelings and cognitions. While
imaging these triggers, patients have to activate their positive feeling-state with posture, facial
K. Korrelboom et al.
expression, positive self-verbalization, and music. These phases, as well as the other phases
in COMET, are repeated regularly during sessions and in homework assignments, while their
progress is monitored on a daily basis. The COMET method for treating low self-esteem
appeared to be effective in a baseline-controlled study with hospitalized eating-disordered
and personality-disordered patients (Korrelboom, van der Weele, Gjaltema and Hoogstraten,
2009) and in a randomized clinical trial (RCT) with eating-disordered patients (Korrelboom,
de Jong, Huijbrechts and Daansen, 2009). The same COMET protocol is applied in either a
group or an individual setting.
Being cognitive-behavioural interventions, the three methods described above have much
in common. Looking at differences, Fennell’s method seems to be the most traditionally
cognitive, featuring Socratic dialogue and behavioural experiments as its main therapeutic
tools. Tarrier’s method, while focusing on cognitions about self-worth, seems to have a
more behavioural approach by stimulating patients to practise those behaviours from which
they extract positive self-esteem. COMET seems to place most emphasis on experiencing
positive self-feelings. Both COMET and Tarrier’s interventions are mainly practised as
additions to ongoing regular treatment for the main disorder for which the patient seeks
help, whereas Fennell’s is also practised as a stand-alone treatment. By emphasizing positive
personal experiences and characteristics, all three methods share similarities with therapeutic
approaches such as Gilbert’s Compassionate Mind Training (Gilbert, Baldwin, Irons, Baccus
and Palmer, 2006) and Padesky’s Building Resilience (Kuyken, Padesky and Dudley, 2009,
Below, first the practical application of COMET for low self-esteem is illustrated by an
individual case example. Then we describe an RCT of this procedure applied in groups of
Conny1was a 40-year-old professional dog trainer referred for anxiety problems. Conny met
the criteria for avoidant personality disorder. She was extremely discontented about herself
and experienced low self-esteem almost continuously. Being dissatisfied with her social
behaviour (“boring and clumsy”) and her intelligence (“dumb”) she was primarily ashamed
of her body. She had conspicuous scars on her face and her legs due to a car accident as a
child, and she thought her breasts were too small and too flabby. In her youth Conny was
shy and withdrawn; later, she turned to alcohol abuse to cope with her insecurities. Still later
she fled into a more socially active style of coping by constantly making unpleasant jokes
about herself. However, she always avoided (in a more passive fashion) situations in which
it would be impossible for her to completely conceal her appearance, such as sports and
going to the beach or the swimming pool. Apart from a short period in her early twenties,
Conny has never had an intimate relationship. Asked to describe the way she felt about
herself, Conny stated that she found herself “unacceptable”. When the therapist asked her
whether she was really totally convinced that she was “a 100 percent unacceptable person”,
she replied that she “knew with her brains” that she wasn’t, but that she “felt in her heart”
that she was. Then the therapist asked how she “knew by her brains” that she was not
1Names and identifying details have been changed to preserve anonymity.
COMET for low self-esteem
totally unacceptable. Conny related that she knew from television and newspapers that some
people in the world were “really bad and unacceptable”. Moreover, she could mention several
positive characteristics of herself as a person. She was good with animals, accurate in her
work, and helpful and loyal to her family and the few friends she had. The rationale of the
treatment was explained to her: i.e. “People with low self-esteem are inclined to focus on
those characteristics they consider negative, whilst ignoring the positive aspects of themselves
that really do exist. COMET is aimed at restoring a more fair balance between accessing
positive and negative self-opinions, by making the accessibility of positive opinions more
Taking “being unacceptable” as a starting point of the COMET treatment, Conny and the
therapist identified characteristics that contradicted this opinion. As a homework assignment,
Conny was asked to monitor and write down as many actual examples of these positive
characteristics she could think of. In the next session Conny was to elaborate on these
examples by writing short self-referent stories about them. The most convincing of these
stories were used in the next few sessions and homework assignments to make these instances
more vivid and better (emotionally) recognizable. Imagery, facial expression and self-
is “an acceptable person”. Among Conny’s most convincing examples of being “acceptable”
and “worthwhile” was an instance where she successfully completed the training of a very
difficult dog that would have been slaughtered because of bad behaviour had the dog training
failed. Conny wrote this story down, focusing on the gratitude of the dog owner, the radical
behaviour change of the dog and, most of all, on her own feelings of pride in her achievement.
Later, she had to imagine these scenes repeatedly, taking a self-assured posture and assuming
a confident facial expression, meanwhile saying to herself that “taking everything together”
she is an “acceptable and worthwhile person” and that “it’s not only appearance that counts”.
During imaging she played Aretha Franklin’s I will survive. Meanwhile other examples of
being an acceptable and worthwhile person were identified and practised in the same fashion.
Finally, as a last step in COMET, she activated her feelings of positive self-esteem again by
imaging, posture, facial expression, self-verbalization and music. Thereafter she replaced the
positive “dog training” and other images by problematic images of situations in which she
normally felt insecure, while keeping her positive feelings active with the remainder of the
positive ingredients: posture, facial expression, self-verbalization and music. She repeated
this again and again until she was able to imagine herself in these difficult scenes feeling self-
confident and acceptable as a person. After COMET (which lasted 9 sessions) Conny reported
that she accepted herself as she was and that her feelings of social anxiety and depression had
A randomized effectiveness study of COMET for low self-esteem for a convenience
sample of patients with personality disorders
After some isolated positive experiences (e.g. as with Conny) in regular clinical practice,
COMET for low self-esteem was put to the test in more sophisticated ways. Until now
COMET for low self-esteem has proven successful in a baseline-controlled study with
hospitalized eating and personality-disordered patients (Korrelboom, van der Weele et al.,
2009) and in an RCT with eating-disordered outpatients (Korrelboom, de Jong, et al., 2009).
At the same time, COMET’s possible effectiveness was put to the test in an RCT with a
K. Korrelboom et al.
convenience sample of personality-disordered patients in a routine treatment setting. This
latter study is described in more detail below.
Overview of the study
The study was performed at the Program for Personality Disorders (PPD) of PsyQ,
Haaglanden. The PPD is part of the Parnassia-Bavo Psychiatric Centre (PBPC), one of the
largest mental health organizations in the Netherlands. The PPD specializes in the treatment of
personality-disordered patients with cluster C pathologies and borderline personality disorder
(BPD). The PPD distinguishes three lines in which its treatments are organized: the “Cluster
C line”, the “BPD line” and the “miscellaneous line”, in which most patients with mixed
personality disorders and personality disorders not otherwise specified (PDNOS), as well as
some incidental patients with other cluster B of cluster A personality disorders, are treated.
Most cluster A and anti-social patients of the PBPC are treated in other departments than
the PPD. However, these three lines of treatment are not rigidly separated and patients of
different lines often share common treatment modules during some periods of their treatment.
As part of their total treatment program, the PPD organizes specific treatment modules aimed
at different aspects of personality problems in which patients from the department itself (as
well as from other departments of the PBPC) can participate.
The COMET group treatment for patients with personality problems and low self-
esteem is among these specific treatment modules. Patients who are considered by their
regular therapists to have such personality problems (as manifested by “enduring, pervasive,
stable and inflexible patterns of inner experiences and behaviour that deviate markedly
from expectations of the individual’s culture and that lead to distress and impairment”) in
combination with low self-esteem (as manifested by enduring feelings of being “inferior”,
“worthless”, “ugly”, or “stupid”, or by considering themselves to be “failures”) can be
referred to these COMET groups. In addition to these criteria, patients should be able to
identify at least one positive personal characteristic. Exclusion criteria are suicidal risk and
not being able or willing to participate in group treatment, as judged by the COMET assessors.
Being an effectiveness study, the current study was conducted in the midst of a “working
outpatient psychiatric department”. Between October 2005 and October 2007 all regular
referrals for COMET were more thoroughly screened by the COMET team to assess whether
they complied with the inclusion and exclusion criteria of the current study.
All patients who fulfilled these criteria and gave informed consent were randomized into
one of two treatment conditions. The experimental group received 7 weekly sessions of
COMET in groups during their (ongoing) regular therapy as usual (COMET + TAU), while
the control group received 7 weeks of (ongoing) regular therapy as usual (TAU) only. After
7 weeks of TAU the control group then received 7 weeks of COMET. All patients who
completed COMET (whether in the experimental group or later in the control group) were
approached 7–10 weeks after completion of their COMET to assess the stability of the
COMET effects over time.
To be eligible, referred patients had to: 1) fulfil the criteria of the DSM-IV-TR (4thed., text
rev.; American Psychiatric Association, 2000) for having a personality disorder as the main
COMET for low self-esteem
diagnosis; 2) low self-esteem had to be manifested in a score ?28 on the Rosenberg Self-
esteem Scale (RSES; Rosenberg, 1965); 3) patients had to fulfil the customary COMET
criteria of being able to indicate at least one positive personal characteristic (e.g. “being
honest”, or “being a faithful friend”); and 4) be able and willing to participate in a
group treatment. Contra-indications for participation were suicidal risk and not being able
to complete the measurements. A DSM-IV-TR diagnosis of a personality disorder was
established by comparing (in a non-standardized clinical interview) the presenting problems
of the patients and the information given by the referring therapists with the formal DSM
criteria for personality disorders. The cut-off score of 28 on the RSES as an inclusion criterion
to objectify reported low self-esteem was based on a large survey where this score was
found to be 1 standard deviation (SD) below the mean of a functional Dutch population
(Schmitt and Allik, 2005). All other inclusion and exclusion criteria were based on the
clinical judgement of the COMET assessor. Patients who were eligible had to give informed
In previous studies on COMET for low self-esteem, large within-effects sizes (Olij et al.,
2006, Korrelboom, van der Weele et al., 2009) and intermediate to large between-effects
sizes on several measures of self-esteem were found (Korrelboom, de Jong et al., 2009).
Therefore, allowing for dropout, a minimum of 90 patients was considered sufficient to
demonstrate differences between groups, as long as similar differences also existed in this
group of personality-disordered patients.
A total of 119 patients were screened. Of these patients, 28 did not fulfil the criteria
for participation (24 did not have a formal personality disorder, and 4 scored above the
cut-off score of 28 on the RSES). Thus, of the 91 patients enrolled in the study, 48 were
randomized to COMET + TAU (experimental group) and 43 to TAU alone (control group).
In the COMET group 45 patients started treatment while only 31 did so in the control group.
After randomization, a total of 15 patients (3 in the experimental group and 12 in the control
group) ended their participation, whether or not with formal withdrawal of informed consent,
implying that they did not submit their pre-treatment measurements or did not participate at
all in the treatments under study. Finally, 76 patients entered the study (45 in the experimental
group and 31 in the control group). All further calculations and analyses pertain to these 76
Of those 76 patients, 12 (16%) were male and 64 (84%) female. Mean age was 36.1
(SD = 8.7) years, 34 patients (45%) were diagnosed with a borderline personality disorder
and another 34 patients (45%) with “another” personality disorder (“mixed” or “personality
disorder NOS”). Finally, 8 patients (10%) were considered to have a cluster C personality
disorder. Figure 1 and Table 1 present the results of the inclusion and randomization
During the study, all patients were assessed several times. The first measurements were taken
within 2 weeks prior to the start of COMET + TAU for the experimental group and within
2 weeks prior to the start of the “waiting for COMET period” for the control group (pre-
treatment measurements). The second measurements were taken 7 weeks later at the end of
COMET + TAU (experimental group), which was at the same time as the end of the “waiting
for COMET period” for the control group (post-treatment measurement). Since COMET for
K. Korrelboom et al.
Table 1. Pre-treatment status for the two treatment groups
COMET + TAU
(n = 45)
(n = 31)
Age in years
Social optimism (POL-socop)
COMET = Competitive Memory Training (experimental group), TAU = Therapy as
usual (control group), n.s. = not significant, BPD = Borderline Personality Disorder,
PD (oth) = Other personality disorder, Cluster C = Cluster C personality disorder,
RSES = Rosenberg Self-esteem Scale, BDI = Beck Depression Inventory, POL-
aut = Positive Outcome List, autonomy, POL-socop = Positive Outcome List, social
the control group started immediately after their waiting period, this second measurement
was at the same time the pre-COMET measurement for the control group. The control group
was then assessed after they had finished their COMET (post-COMET measurement for the
control group). Finally, all patients who had completed COMET (whether in the experimental
group or later in the control group) were approached 7–10 weeks after the completion of
COMET to fill in measurements again to assess the stability of COMET effects over a longer
time period (follow-up). Thus, the experimental group was assessed three times (start of
COMET, end of COMET, and follow-up), while the control group was assessed four times
(start of waiting period, end of waiting period/start of COMET, end of COMET, and follow-
up). The following measures were assessed at all specified moments:
Rosenberg Self-Esteem Scale (RSES; Rosenberg, 1965).
item scale, items have to be answered on a 4-point Likert scale, ranging from “strongly
agree” (1) to “strongly disagree” (4). Half of the items are positively formulated, the other
half negatively. After recoding, a high total score (range 10–40) means higher self-esteem.
The RSES assesses “global self-esteem” and is sufficiently reliable and valid (Blascovich and
Tomaka, 1991). The RSES was considered the primary outcome measure.
On a Dutch version of this 10-
Beck Depression Inventory (BDI: Beck, Ward, Mendelson, Mock and Erbaugh, 1961).
Dutch translation of this 21-item self-referent 4-point Likert scale has proven to be reliable
(Bouman, Luteijn, Albersnagel and van der Ploeg 1985) and valid (Bouman, 1989). Low
scores are favourable. The BDI (range 0–63) was considered a secondary outcome measure.
COMET for low self-esteem
Figure 1. Flow chart of the study
Positive Outcome Scale (POS; Positieve Uitkomsten Lijst or PUL: Appelo, 2005).
10-item Dutch self-report instrument assesses resilience. Seven items pertain to autonomy
(range 7–28) and three to social optimism (range 3–12). Scores on the POS are strongly
associated with self-efficacy. There are Dutch norms for a normal population and a psychiatric
population. Reliability and validity are sufficient (Appelo, 2005). High scores are favourable.
The POS was considered another secondary outcome measure.
K. Korrelboom et al.
A total of 9 therapists participated in the study: 5 clinical psychologists and 4 nurses. All had
trained and supervised in COMET by the first author (KK). Each COMET group was led by
two therapists. In all instances, at least one of these therapists was a clinical psychologist and
at least one had more than 2 years experience in treating patients with personality disorders.
COMET was carried out in small groups (5–9 patients) as an additional treatment to the
ongoing regular treatment program (TAU). After referral to COMET by their therapists,
patients were interviewed by one of the COMET therapists. After inclusion and informed
consent, patients were randomized over the two research conditions. Randomization was
performed in separate blocks of 10–18 patients, by opening blinded envelopes in which both
treatment conditions were concealed in advance. Pre-treatment measurements were made at
most 2 weeks prior to the start of the treatments (COMET + TAU or TAU alone). When
intake took place later, patients filled in their pre-treatment measurements during intake.
When the period between intake and start of treatments was longer than 2 weeks, they took
the measurements home and filled them in later. Second measurements of all patients were
taken immediately after the experimental group had finished COMET, which was 7 weeks
later. Patients in the control group received COMET as soon as the experimental groups had
finished. Around the same time as these control patients were assessed for their post-treatment
results, patients from the experimental groups who had completed COMET were assessed for
the third time for their follow-up results. Another 7–10 weeks later, COMET completers from
the control group were approached to fill in their fourth and last measurements, i.e. the follow-
COMET was performed according to a manual that all patients received at the start of the
training. COMET lasted 7 sessions of 2 hours each and consisted of the following steps:
Identifying the negative self-image.
about herself (session 1).
The patient describes what she thinks is negative
Identifying a credible positive self-image that is incompatible with the negative self-image.
The patient is asked whether she really believes this negative image of herself is totally true
and, if not, which personal characteristics and experiences contradict the negative self-image
Strengthening the positive self-image.
competition of the contradictory positive self-image is enhanced by repeatedly strengthening
its emotional load. This is realized by: a) writing small self-referent stories of instances in
which the positive qualities were manifest, and distilling positive self-statements of these
instances (session 2 and 3); b) imagining oneself in these positive personalized scenes (session
3); c) the purposeful manipulation of body posture and facial expression (session 4); and d)
listening to music chosen by each patient because it is felt to be congruent with a positive
Then, the competitiveness in the retrieval
COMET for low self-esteem
self-image (session 5). These exercises are to be practised during sessions 2–5 as well as
during daily homework assignments.
Forming new associations between “risk cues” and positive self-image by counter-
In the last sessions of COMET, patients are trained to associate their new
positive self-image with cues that normally provoke uncertainty. Positive self-esteem has to
be activated with the aid of imagination, posture, facial expression, music and positive self-
statements. Then, the positive image is replaced by the image of a situation in which she
normally feels negative about herself. Now, however, by keeping her positive feeling state
activated she tries to feel self-confident while “being in the imagined difficult scene”. Again,
thishastoberepeated severaltimesandalsopractisedindailyhomework assignments.Oncea
difficult scene can be tolerated while retaining positive self-esteem, other scenes are practised
(session 6 and 7).
TAU (treatment as usual) cannot be specified in a detailed manner since included patients
had different psychopathologies, while some of them had their regular treatment in different
teams of the PPD or in different departments of the PBPC. In general, patients in the PPD and
the PBPC are treated on an outpatient basis, with treatment methods that can be considered
evidence-based or, at least, consensus-based.
First, possible pre-treatment differences between completers and dropouts, between both
groups in general, as well as between completers in both groups were tested with regression
analyses, taking age, gender, diagnosis and pre-treatment scores on all outcome variables
as predictors. Then, when no pre-treatment differences between completers and dropouts or
randomization status could be identified, missing data were imputed with the Expectation
Maximization (EM) algorithm of SPSS 17. This imputation procedure was checked by
comparing both groups again, this time with inclusion of the imputed data.
Thereafter, differences at post-treatment between the two groups were tested for the main
outcome variable (self-esteem) with an analysis of covariance (ANCOVA), with the post-
treatment score as dependent variable, the pre-treatment score as the covariate, and the
treatment allocation as a fixed factor. Then, to control for inflated type 1 error, possible
differences at both POS subscales and the BDI were tested in a MANOVA, with the baseline
data as covariates. When this MANCOVA yielded a significant result, a step-down was
made using separate ANCOVAs of the post-treatment data with the pre-treatment data as
covariates and treatment allocation as the fixed factor. Except for the regression analyses
where a p-value of 0.1 was applied, a p-value of 0.05 was considered significant in all other
To give an impression of the meaning and sizes of these differences, Cohen’s d based on the
mean squared error and the sizes of both the experimental and the control group (Thalheimer
and Cook, 2002), as well as 95% confidence intervals, and estimates of clinical significance
Finally, to assess the stability of the effects of COMET, the results of all COMET
completers who also filled in follow-up measurements were analyzed with ANOVA repeated
measures on the main outcome measure and with MANOVA repeated measures on the three
other outcome measures, defining three moments for time (pre-COMET, post-COMET, and
K. Korrelboom et al.
Table 2. Interaction effects between pre- and post-treatment: intention-to-treat
Exp: 45 20.7 (3.4)
(19.7 – 21.8)
(19.0 – 22.1)
(24.1 – 26.9)
(20.4 – 23.9)
Exp: 45 24.7 (10.2)
(21.7 – 27.8)
(20.7 – 28.2)
(14.0 – 20.5)
(19.0 – 27.4)
(15.0 – 16.8)
(15.3 – 18.3)
(17.8 – 19.8)
(16.4 – 19.2)
Exp: 457.3 (2.0)
(6.7 – 8.0)
(7.1 – 8.4)
(7.9 – 8.9)
(7.2 – 8.5)
95% CI = 95% Confidence Interval, RSES = Rosenberg Self-esteem Scale, BDI = Beck
Depression Inventory, POL-aut = Positive Outcome List, autonomy, POL-socop = Positive
Outcome List, social optimism, Exp = experimental group (COMET + TAU), Ctr = control
group (TAU alone).
Of the 76 patients included in the study, 19 (25%) dropped out during the first study period
(COMET + TAU versus TAU): 11 in the experimental group (24%) and 8 (26%) in the
control group. No predictors for dropout in general could be identified. Also, no differences
in pre-treatment status were detected concerning allocation to treatment condition or between
the completers in both treatment conditions. Thus, missing values could be considered as a
random collection of the research data. To be able to report the results on an intention-to-
treat basis, all missing scores at post-treatment were imputed by use of the SPSS 17 EM
algorithm by taking all available pre- and post-treatment outcome measures as well as age,
diagnosis, gender and treatment condition as predictors. After imputation, still no predictors
were found for allocation to one of the treatment conditions, indicating that randomization
was still successful.
On an intention-to-treat basis, patients in the COMET+TAU condition performed better
on all outcome measures post-treatment. Table 2 gives an overview of the main results. There
was a significant interaction effect (treatment x time) on the ANCOVA for self-esteem (RSES:
F (2, 73) = 12.54, p = .001). On the combined scores of depression (BDI), autonomy
COMET for low self-esteem
(POL-aut), and social optimism (POL-socop), MANCOVA showed better effects for
COMET + TAU than for TAU alone: F (3, 69) = 6.15, p < .000. A step-down showed that
COMET + TAU was better than TAU alone on all three separate variables (BDI: F (1, 71) =
11.29, p = .001; POL-aut: F (1, 71) = 6.15, p = .02; POL-socop: F (1,71) = 6.75, p = .01).
The size of these differences between COMET and control were large on self-esteem and
depression, intermediate on social optimism, and small on autonomy.
Within-effect sizes from pre- to post-test for the COMET group were large on two variables
[1.2 (RSES) and 0.9 (POL-aut)] and intermediate on the other two [0.7 (BDI) and 0.6 (POL-
socop)]. For the control group there were two small within-effect sizes [0.4 (RSES) and 0.3
on the POL-aut].
To assess the clinical significance of the changes on the RSES (the primary outcome
variable) that completers realized during COMET + TAU and TAU alone, the procedure
described by Jacobson and Truax (1991) was adopted. According to this procedure a patient
has to fulfil two criteria to make a clinically significant change: a) he/she should progress
from the problematic population to a normally functioning population, and b) he/she should
realize a “reliable change score”, i.e. the difference between his/her pre-treatment and post-
treatment score should surpass the standard error of difference between these two scores. As
a cut-off point between problematic and normal functioning, a score of 28 on the RSES was
chosen, which was one of the inclusion criteria in the present study. A post-treatment score
of 28 or above on this scale would bring the patient within the range of 1 SD under the mean
of a functional Dutch group (Schmitt and Allik, 2005) and thus was considered indicative of
normal functioning on self-esteem. Based on the reliability index of the RSES of 0.87, found
in that same study, an increase of at least 6 points between pre- and post-measurement on this
scale was considered necessary to fulfil Jacobson and Truax’ second requirement. According
to these criteria 12 patients that completed COMET + TAU (35% of the COMET + TAU
completers) progressed from the problematic to the functional population, while 2 patients
did so in TAU alone (9% of the TAU alone completers). In COMET + TAU 13 patients (38%
of the completers) realized a reliable change by progressing 6 points or more between pre-
and post-treatment measures on the RSES, while 3 patients in TAU alone did so (13% of the
completers). Finally, 8 patients in COMET + TAU fulfilled both Jacobson and Truax criteria
and made a clinically significant change during treatment (24%), while only 1 patient in the
TAU alone condition reached this point (4%).
To get an impression of the possible influence of the imputation procedure on outcome,
we finally calculated tests of significance for those patients that completed both treatments.
In these analyses separate ANCOVAs on all outcome measures also yielded significant
differences in favour of COMET, with roughly the same p- values: RSES: F (2, 54) = 11.87,
p = .001; BDI: F (2, 54) = 12.74, p = .001; POL-aut: F (2, 54) = 6.25, p = .015; POL-socop:
F (2, 54) = 10.28, p = .002. These outcomes seem to justify the application of our imputation
Stability of COMET effects over time
After their waiting period, patients in the control condition also received COMET. Of these
patients 23 started COMET, and 20 of them completed the treatment. In general, these patients
improved during COMET on three of the outcome measures as was shown by paired t-tests;
RSES: t (18) = −2.4, p = .03; BDI: t (19) = 3.30, p = .004; POL-aut: t (19) = −1.54,
K. Korrelboom et al.
p = .14 (n.s.); POL-socop: t (19) = 0.009. Cohen’s d for the three significant changes were
intermediate: 0.6 on all three variables.
All patients who had completed COMET (from both the experimental and control
condition) were then approached 7–10 weeks after the completion of COMET to assess the
stability of the COMET effects over a longer time period. Of the 76 patients who started
follow-up measurements; they were not markedly different from the 76 patients who started
the study, i.e. 17% male, 83% female; mean age 37.2 (SD = 8.4) years; 53% BPS, 43% other
PD and 3% cluster C; pre-COMET measures on the RSES: 20.8 (SD = 4.2), on the BDI: 24.9
(SD = 11.6), on the POL-aut: 16.6 (SD = 3.6) and on the POL-socop: 7.3 (SD = 1.6). In 4
separate ANOVAs for repeated measures (with three moments for time: start of COMET, end
of COMET and follow-up) effects remained stable on 3 of the 4 outcome measures; RSES: F
(2, 56) = 14.10, p < .0002; BDI: F (2, 58) = 24.53, p < .000; POL-aut: F (2, 58), p = .003.
As indicated by contrasts, in all instances these significant changes were realized during pre-
COMET to post-COMET, while there were no significant changes between post-COMET
and follow-up of COMET. These findings suggest that patients changed on the specified
measurements during COMET and that these changes remained stable, at least during the
following few months. This was different for POL-socop. On this measure the ANOVA for
repeated measures was also significant (F (2, 56) = 45.93, p = .003), but this time contrasts
showed that changes were significant in both time periods: from pre-COMET to post-COMET
and between post-COMET and follow-up. While the change at POL-socop during COMET
had deteriorated again on this outcome measure. Figures 2–5 give an impression of the follow-
Since low self-esteem is a major problem in the manifestation of different emotional disorders
(including personality disorders), as well as a risk factor for the development of and
relapse into some of these disorders, and since low self-esteem is not always automatically
enhanced by the successful treatment of the main presenting disorder, it is important to
have adequate specific treatment methods for this condition. COMET for low self-esteem
might be such a treatment. As the first randomized study into COMET for low self-esteem
as a treatment procedure for patients with personality disorders and low self-esteem, the
current study confirms earlier findings on its efficacy as an add-on to regular therapy
(Korrelboom, van der Weele et al., 2009; Korrelboom, de Jong et al., 2009). Also in this
specific group of personality-disordered patients, COMET seems to enhance self-esteem (at
least in combination with regular therapy) and reduce depression, both with large effect sizes.
Different from the COMET study on eating-disordered patients (Korrelboom, de Jong et al.,
2009), in the current study a differential effect with small and intermediate effect sizes on
autonomy and social optimism, both related to self-efficacy, in favour of COMET was also
found. All these changes were realized in a short period of time with patients generally
considered difficult to treat and many of whom had been in therapy for extended periods
2For 1 patient post-COMET measures on the RSES and the POL-socop were missing.
COMET for low self-esteem Download full-text
Figure 2. RSES follow-up
Figure 3. BDI follow-up
of time. Moreover, most effects seem to be stable over a certain period of time. Only on social
optimism, patients relapsed after termination of COMET within 7–10 weeks.
on the RSES was still below 1 SD of the functional population. The impact of COMET might