Available online at www.sciencedirect.

com

Comprehensive Psychiatry 53 (2012) 554 – 561

www.elsevier.com/locate/comppsych

Neither bipolar nor obsessive-compulsive disorder: compulsive buyers

are impulsive acquirers
Tatiana Zambrano Filomensky a,⁎, Karla Mathias Almeida b , Marcelo Campos Castro Nogueira a ,
Juliana Belo Diniz c , Beny Lafer b , Sonia Borcato c , Hermano Tavares a
a
Impulse Control Disorders Outpatient Unit, Institute and Department of Psychiatry, University of São Paulo Medical School, São Paulo, Brazil
b
Bipolar Disorder Research Program, Institute and Department of Psychiatry, University of São Paulo Medical School, São Paulo, Brazil
c
Program for Obsessive-Compulsive Spectrum Disorders, Institute and Department of Psychiatry, University of Sao Paulo Medical School, São Paulo, Brazil

Abstract

Introduction: Compulsive buying (CB) is currently classified as an impulse control disorder (ICD) not otherwise classified. Compulsive
buying prevalence is estimated at around 5% of the general population. There is controversy about whether CB should be classified as an
ICD, a subsyndromal bipolar disorder (BD), or an obsessive-compulsive disorder (OCD) akin to a hoarding syndrome. To further investigate
the appropriate classification of CB, we compared patients with CB, BD, and OCD for impulsivity, affective instability, hoarding, and other
OCD symptoms.
Method: Eighty outpatients (24 CB, 21 BD, and 35 OCD) who were neither manic nor hypomanic were asked to fill out self-report
questionnaires.
Results: Compulsive buying patients scored significantly higher on all impulsivity measures and on acquisition but not on the hoarding
subdimensions of clutter and “difficulty discarding.” Patients with BD scored higher on the mania dimension from the Structured Clinical
Interview for Mood Spectrum scale. Patients with OCD scored higher on obsessive-compulsive symptoms and, particularly, higher on the
contamination/washing and checking dimensions from the Padua Inventory; however, they did not score higher on any hoarding dimension.
A discriminant model built with these variables correctly classified patients with CB (79%), BD (71%), and OCD (77%).
Conclusion: Patients with CB came out as impulsive acquirers, resembling ICD- rather than BD- or OCD-related disorders. Manic symptoms
were distinctive of patients with BD. Hoarding symptoms other than acquisition were not particularly associated with any diagnostic group.
© 2012 Elsevier Inc. All rights reserved.

1. Introduction reference of modern psychiatry, emphasized this disorder's

Compulsive buying (CB) was first proposed as a
psychiatric diagnosis in 1915 by Kraepelin [1], who named
it Oniomania (from the Greek oné, to buy, and mania,
frenzy). As early as 1924, Bleuler [2], another classical
Conflicts of interest: Dr Almeida is the recipient of the 2008
International Society for Bipolar Disorders-GlaxoSmithKline Travel
Fellowship Award. Dr Diniz has been the main investigator in a trial
partially funded by Novartis Pharmaceutics and has received travel grants
from Janssen Pharmaceutics in the last 5 years. Dr Tavares has received
research support from Cristalia, Roche, and Sandoz in his role as President
of Brazil's National Association on Pathological Gambling and Other
Impulse Control Disorders.
⁎ Corresponding author. Department of Psychiatry, University of São
Paulo, São Paulo—SP CEP 05590-120, Brazil. Tel.: +55 11 3031 5381.
E-mail address: tatizf@usp.br (T.Z. Filomensky).
0010-440X/$ – see front matter © 2012 Elsevier Inc. All rights reserved.
doi:10.1016/j.comppsych.2011.09.005
impulsive character, which is characterized by excessive
preoccupations and wishes related to the acquisition of objects
and inability to manage buying and financial expenditure. In
1994, McElroy et al [3] proposed diagnostic criteria for CB,
now adopted by many studies and reviews. The prevalence of
CB is around 6% of the American population [4], with a male-
to-female ratio close to 1:1. However, in clinical settings,
women are more numerous, with the sex ratio reaching 1:4
[5]. There is frequent comorbidity with other psychiatric
disorders, usually organized into 5 major groups: mood
disorders, anxiety disorders, substance use disorders, other
disorders of impulse control, and personality disorders
(mainly cluster B) [5,6]. Evidence suggests that cognitive-
behavioral therapy and treatment with selective serotonin
reuptake inhibitors may be effective for CB [7-9], although for
the latter, there are no controlled studies available so far.
 T.Z. Filomensky et al. / Comprehensive Psychiatry 53 (2012) 554–561
Despite the growing interest, CB remains excluded from
modern psychiatric classifications, except for potential
classification as a residual category: as an impulse control
disorder (ICD) not otherwise classified in the Diagnostic
and Statistical Manual of Mental Disorders, Fourth
Edition-text revised (DSM-IV-TR) [10]. A better classifi-
cation of CB requires elucidation of its true psychopath-
ologic nature. Besides being related to impulse control,
excessive shopping was reported as a common feature of
manic episodes, yielding the hypothesis that CB could
result from subsyndromal bipolar disorder (BD) rather than
being a discrete diagnosis in itself [11].
The uncontrollable and repetitive nature of the buying
behavior may answer for the use of the term compulsive in
naming the syndrome, and it has become its most popular
designation. Likewise, clinical reports have focused on
symptoms shared with obsessive-compulsive disorder
(OCD), especially the feeling of being compelled to shop
against one's own judgment and the careful storing of
purchased objects [6], thus regarding CB as a member of an
impulsive-compulsive spectrum of disorders [12], akin to the
compulsive hoarding commonly observed in patients with
OCD. Despite the uncertainty surrounding the classification
of CB, we have embraced the term compulsive shopping to
avoid the introduction of yet a different term, which could
add substantial confusion to the matter. However, this
decision should not be mistaken by an early acceptance of
CB as being necessarily related to OCD, which indeed
deserves investigation. Considering its classical origin,
Oniomania would be an even better term; unfortunately, it
still remains largely unknown.
Impulse control disorders and BD share several phe-
nomenological characteristics such as impulsivity, danger-
ous behavior, and affective symptoms. In addition, McElroy
et al [11] emphasize that both disorders are likely to
manifest during adolescence or young adult life, leading to
chronic evolution if not properly treated; however, this can
be said about most psychiatric disorders and sounds rather
unspecific. No data about the prevalence of BD among
compulsive buyers whether from clinical settings or from
the general population are available. However, BD is
reported as a frequent comorbidity among ICDs. Indeed, in
a large population survey, about 23% of individuals
fulfilling the diagnostic criteria for pathologic gambling
(PG) also fulfilled criteria for at least 1 lifetime manic
episode [13]. The reverse seems also true, and among a
clinical sample of euthymic patients with BD type I (BD-I),
ICDs were found in comorbidity in 27.4% of patients, CB
being the second most common disorder after skin picking
[14]. Whether ICDs, among them CB, remain as diagnoses
apart from the bipolar spectrum of disorders relies on
verifying if poor impulse control may occur independently
from mood symptoms.
Previous reports have found that OCD is more frequent
among compulsive buyers than in noncompulsive buyers
[15]; the contrary seems also true that CB is more frequent
555
among patients with OCD than in control individuals. It has
been suggested that the resemblances between CB and OCD
revolve around the hoarding behavior usually observed in
both cases [6]. The patient needs to acquire and store specific
objects, generally useless or in extreme amounts, that
completely exceed any reasonable need. Such stocks are
carefully kept, and patients abominate the idea of anyone
meddling with them, which leads to frequent cluttering and
space compromise. In fact, Frost and Steketee [16] found
that, in a sample of compulsive buyers, the correlation
between a CB score and OCD symptoms was significant
only if not controlled for hoarding symptoms. Conversely,
when compared with control individuals, the same sample of
compulsive buyers showed significantly higher scores on all
OCD symptom dimensions, thus leaving open the question
whether CB is related to other OCD symptoms beyond
hoarding. Furthermore, studies investigating the structure of
OCD symptoms have reported hoarding as a discrete
dimension with obsessive fears of discarding objects that
could later be needed along with storing rituals [17,18]. In
OCD, hoarding has been associated with poorer prognosis.
When present alone and apart from other symptom di-
mensions, hoarding may not respond to the usual treatment,
to the point that some authors have questioned whether
hoarding is actually an OCD dimension or a separate
nosologic category [15,16].
Finally, Tavares et al [6] have underlined similarities
between CB and PG, another ICD. First symptoms occur
during adolescence and are usually accompanied by a host
of behavioral problems marked by impulsivity and social
deviance. The 2 have the same comorbidity and personality
profiles: reward-seeking and stress-relief dynamics. There
is frequent cross-comorbidity between CB and PG;
therapeutic programs for both syndromes are based on
cognitive-behavioral therapy with many components
imported from addiction treatment models. Moreover,
ICDs are a complex and heterogeneous group of disorders
with seemingly subdivisions, one of them being the realm
of behavioral addictions best represented by PG, which
includes CB [19]. Both PG and CB present frequent
comorbidity with substance addiction (and vice versa) and
familial aggregation and may share genetic and psychobi-
ologic factors [20]. As matter of fact, it has been
anticipated that, in the Diagnostic and Statistical Manual
of Mental Disorders, Fifth Edition, PG may be removed
from the ICD section and included in a broader section
comprehending both behavioral and substance-related
disorders [21]. Besides, clinical trials have found promising
evidence of the utility of naltrexone in PG treatment [22].
In fact, μ-opioid receptor antagonism has been explored as
a strategy for the psychopharmacology of addictions,
and improvement in CB with naltrexone in case series
has been reported [23].
The objective of the present study was to compare a group
of outpatients who sought treatment of CB, OCD, or/and
BD. Our initial hypothesis was that if CB was indeed an ICD,
556 T.Z. Filomensky et al. / Comprehensive Psychiatry 53 (2012) 554–561
it would distinguish itself from both BD and OCD by
pronounced impulsive trait-related behaviors. However, if
CB was related to BD, we would expect matching traits and
symptom levels associated with the affective instability or at
least a higher level of similarity between CB and BD than
between CB and OCD. Lastly, if CB was an OCD-related
disorder, we would expect matching levels for obsessive-
compulsive traits and symptoms, especially those related to
hoarding, or at least a greater similarity between CB and
OCD than between CB and BD.
2. Method
2.1. Sample selection
The subjects for this research study are patients, 18 years
or older, who spontaneously sought treatment of CB, OCD,
or BD in the specialized units of the Institute of Psychiatry,
São Paulo, Brazil, which is located in a public university
hospital that provides treatment free of charge. Patients could
also be referred by psychologists and physicians from other
specialties. Patients who had completed at least 6 months of
outpatient treatment during 2009 and the first half of 2010
were invited to participate.
All patients were evaluated according to the specified
criteria in the DSM-IV TR [10] by a trained psychiatrist. At
this preliminary study, a choice was made to include only
patients with BD-I because of its clearer diagnostic
boundaries in comparison with BD type II [24], thus
reducing the risk of misdiagnosis. Patients with CB were
assessed according to the following diagnostic criteria
proposed by McElroy et al [3]:
1. Maladaptive preoccupation with buying or shopping or
maladaptive buying or shopping impulses or behavior,
as indicated by at least one of the following:
a) frequent preoccupation with buying or impulses to
buy, experienced as irresistible, intrusive, and/or
senseless; or
b) frequent buying of more than can be afforded,
frequent buying of items that are not needed, or
shopping for longer periods than intended.
2. The buying preoccupations, impulses, or behaviors
cause marked distress, are time-consuming, signifi-
cantly interfere with social or occupation functioning,
or result in financial problems (eg, indebtedness or
bankruptcy).
3. The excessive buying or shopping behaviors do not occur
exclusively during periods of hypomania or mania.
The exclusion criteria adopted were clinical condition
requiring emergency treatment or hospitalization; mental
retardation, cognitive deficits, psychotic symptoms, or any
other condition that could compromise the completion of the
self-report scales; current episodes of mania or hypomania
(clinically assessed by one of the authors, KMA); and
patients who refused to take part in the research protocol by
not signing the free informed consent form.
Eighty-three outpatients were initially assessed; 3 patients
with BD-I were excluded for experiencing manic episodes at
the time of the assessment. The final sample, with 80
subjects, was composed of 24 subjects with CB, 35 subjects
with OCD, and 21 subjects with BD-I.
2.2. Instruments and assessment
The Structured Clinical Interview for DSM-IV [25]
interview, translated into Portuguese and validated [26],
was used for diagnostic assessment. The diagnostic criteria
for CB were checked with a special instrument modeled
on the Structured Clinical Interview for DSM-IV and
checked by 2 independent raters. After confirmation of
diagnosis, patients answered a sociodemographic ques-
tionnaire [27] including sex, age, ethnicity, place of birth
(1 indicates state capital; 0, other), marital status, job
status professional classification according to the Social
Position Index of Hollingshead and Redlich [28], family
monthly income (converted into US dollars according to
the current currency exchange rate), years of education,
and religion of origin.
After the clinical interview, the individuals filled out the
following self-report scales:
Barratt Impulsiveness Scale (BIS) version 11: the BIS-11
[29] is a 30-item self-report questionnaire. The total score
ranges from 30 to 120, resulting from the sum of 3
subfactors—impulsivity deriving from attention and
cognitive instability (BIS-attention), impulsivity deriving
from motor restlessness and lack of perseverance (BIS-
motor), and failure to account for future consequences and
avoidance of cognitive complexity (BIS-nonplanning).
Padua Inventory (PI) for Obsessive-Compulsive Disor-
der Symptoms: the PI (Washington State University
revised form [30]) is a 39-item self-report scale
intended to identify obsessive and compulsive symp-
toms in both patients with OCD and nonpatients. Its
total score ranges from 0 to 156, deriving from the sum
of 5 subscores: contamination obsessions and washing
compulsions subscale (PI-washing), dressing and
grooming compulsions subscale (PI-dressing), checking
compulsions subscale (PI-checking), obsessional
thoughts of harm to self or others subscale (PI-harm),
and obsessional impulses to harm self or others
subscale (PI-impulse).
Structured Clinical Interview for Mood Spectrum (SCI-
MOODS) Self-report format: the SCI-MOODS is an
instrument that assesses both overt and subtle components
of depression and mania symptoms along a continuum of
mood regulation and affective instability [31]. It encom-
passes depressive, manic, and hypomanic symptoms
included in the classification of mood disorders, such as
International Statistical Classification of Diseases, 10th
Revision [32] and DSM-IV TR [10], besides including
 T.Z. Filomensky et al. / Comprehensive Psychiatry 53 (2012) 554–561 557

subsyndromal, atypical symptoms, and temperamental
predispositions related to mood disorders [33]. The SCI-
MOODS is a 161-item scale. Its total score ranges from
0 to 154 and is the sum of 3 subscores: vegetative
functions, circadian oscillation, periodicity, and season-
ality (SCI-MOODS–rhythm); energy, mood, and cogni-
tion symptoms related to depression (SCI-MOODS–
depression); and mania (SCI-MOODS–mania) [34].
Saving Inventory-Revised (SI-R): the SI-R, in its
reviewed version [35], is a 23-item self-report question-
naire intended to identify pathologic storing symptoms.
The total score, ranging from 0 to 92, is obtained as the
sum of the scores of its 3 subscales: SI-R–hoarding, SI-
R–difficulty discarding, and SI-R–acquisition.
2.3. Statistical analysis
To circumvent the risk of inflating the probability of the
α error by too many comparisons, a multivariate analysis
was conducted. However, given the relatively small sample
size, a univariate preanalysis was carried out to select
variables for the regression models [36]. First, the
demographic profiles of the diagnostic groups were
compared. χ 2 Tests and Kruskal-Wallis tests were used
for categorical and continuous variables, respectively.
Furthermore, impulsivity (BIS-11), compulsivity (PI),
affective instability (SCI-MOODS), and hoarding scores
were compared among the diagnostic groups through the
Kruskal-Wallis test, thus comprising the univariate phase of
data analysis. Variables reaching a significance level equal
to or lower than 0.10 at this point were selected for the
second multivariate phase of the data analysis.
A multinomial logistic regression model was built, having
the selected variables as either factors (categorical) or
covariates (continuous) and diagnostic group as the
dependent variable. A backward stepwise procedure was
adopted, removing at each stage the least significant variable
until all independent variables remaining in the model had a
significance level lower than .05.
Finally, the covariates remaining in the final multinomial
logistic model were used in a discriminant analysis. The
covariates' distributions were previously tested using the
Kolmogorov-Smirnov test. Those presenting a significance
level equal to or below 0.10 went through logarithmic
transformation and were retested to ensure normality. Box's
M test was used to test for equality of group covariance matrices.
The software used for statistical analysis was SPSS,
version 16.0 (IBM Corp, Chicago, IL, USA) [37].

Table 1
Sociodemographic profile of carriers of CB, OCD, and BD-I
Characteristics CB OCD BD-I Total Test P
n= 24 n = 35 n =21 N = 80
Sex
Male 7 (29.2%) 21 (60%) 9 (42.9%) 37 (46.2%) χ 22 = 5.577 a .062
Female 17 (70.8%) 14 (40%) 12 (57.1%) 43 (53.8%)
Age (y)
Average (SD) 39.7 (9.5) 35.1 (9.2) 42.1 (9.2) 38.3 (9.7) χ 22 = 6.710 b .035
Ethnic group
White 17 (70.8%) 18 (51.4%) 9 (42.9%) 44 (55%) χ 22 = 3.862 a .145
Nonwhite 7 (29.2%) 17 (48.6%) 12 (57.1%) 36 (45%)
Place of birth
São Paulo (capital) 17 (70.8%) 22 (62.9%) 14 (66.7%) 53 (62.2%) χ 22 = 0.407 a .816
Others 7 (29.2%) 13 (37.1%) 7 (33.3%) 27 (33.8%)
Marital status
With partner 12 (50%) 13 (37.1%) 9 (42.9%) 34 (42.5%) χ 22 = 0.965 a .617
Without partner 12 (50%) 22 (62.9%) 12 (57.1%) 46 (57.5%)
Years of education
Average (SD) 14.3 (3.5) 12.9 (3.0) 12.5 (3.5) 13.2 (3.3) χ 22 = 4.920 b .085
Professional classification c
Average (SD) (variation: 1-7) 3.1 (2.0) 4.2 (1.8) 3.7 (1.9) 3.7 (1.9) χ 22 = 5.203 b .074
Job status
Employed 15 (62.5%) 21 (60%) 10 (46.7%) 46 (57.5%) χ 22 = 1.174 a .556
Unemployed 9 (37.5%) 14 (40%) 11 (52.4%) 34 (42.5%)
Family monthly income (in American dollar) d
Average (SD) 2621.32 (2957.61) 1502.52 (1064.89) 2130.25 (1620.06) 2002.94 (1982.59) χ 22 = 4.260 b .119
Religion of origin
Catholic 19 (79.2%) 25 (71.4%) 16 (72.6%) 60 (75%) χ 22 = 0.476 a .788
Noncatholic 5 (20.8%) 10 (28.6%) 5 (23.8%) 20 (25%)
a
χ 2 Test.
b
Kruskal-Wallis test.
c
According to the Hollingshead scale (Hollingshead and Redlich [28]).
d
Converted according to the current rate of exchange, approximately US $1.00 = R $1.70.
558 T.Z. Filomensky et al. / Comprehensive Psychiatry 53 (2012) 554–561

3. Results backward stepwise procedure excluded the variables that did

Table 1 presents the sociodemographic profile description
of the sample. Patients with BD-I were older than the 2 other
groups. Compulsive buying presented a higher frequency of
women and a slightly higher sociocultural level than the
other 2 subgroups, although the difference is not statistically
significant. In an additional analysis, professional classifi-
cation and years of formal education seemed to be highly
correlated (Spearman ρ = −0.675, P b .001); therefore, to
avoid the risk of covariance, only the professional classifi-
cation was selected for the next multivariate analysis stage.
Table 2 describes the univariate analysis that compared
the diagnostic groups with regard to impulsivity, obsessive-
compulsive features, affective instability, and storage.
In the multinomial logistic regression model, having the
diagnosis as the dependent variable, age was introduced as a
covariate and sex as a cofactor to control for differences in
the samples' profiles. The remaining variables that reached
significance at .10 or lower were introduced as a block. The
not reach a significance level lower than .05 in the following
sequence: BIS-attention (P = .976), BIS-motor (P = .557),
and SCI-MOODS–rhythm (P = .183). Table 3 describes the
final regression model.
The Kolmogorov-Smirnov test showed that, among the
variables that remained in the final regression model, PI-
washing (P = .050), PI-checking (P = .058), and SI-R–
acquisition (P = .032) variables did not have normal
distributions. We then conducted a logarithmic transforma-
tion and repeated the Kolmogorov-Smirnov test for the
transformed variables. The significance levels of log(PI-
washing) (P = .592), log(PI-checking) (P = .337), and
log(SI-R–acquisition) (P = .421) show that the distribution
of the transformed variables was normal.
The transformed variables, in addition to the other
variables present in the final multinomial model, were
inserted into a discriminant analysis model, except for sex
because the discriminant analysis does not admit categorical
variables. Age was introduced into the model to control for

Table 2
Impulsivity, compulsivity, affective instability, and hoarding comparison for CB, OCD, and BD-I; Kruskal-Wallis test
Variables CB OCD BD-I Total Test P
n= 24 n = 35 n =21 N = 80
BIS-attention 21.4 (5.1) 20.3 (3.5) 20.0 (3.4) 20.6 (4.0) χ 22 = 7.534 .023
Average (SD)
BIS-motor 26.2 (7.6) 22.1 (6.8) 23.5 (5.7) 23.7 (7.0) χ 22 = 8.058 .018
Average (SD)
BIS-nonplanning 31.5 (7.2) 27.1 (4.5) 26.8 (4.9) 28.3 (5.8) χ 22 = 20.905 b.001
Average (SD)
BIS total score 79.1 (18.3) 69.5 (13.1) 70.3 (12.2) 72.6 (15.1) χ 22 =14.156 .001
Average (SD)
PI-washing 8.7 (10.3) 16.1 (10.9) 7.9 (7.6) 11.7 (10.6) χ 22 = 13.758 .001
Average (SD)
PI-dressing 1.6 (1.7) 3.4 (3.6) 2.2 (2.5) 2.5 (3.0) χ 22 = 3.362 .186
Average (SD)
PI-checking 8.0 (8.0) 15.5 (10.4) 6.8 (5.4) 11.0 (9.4) χ 22 = 13.241 .001
Average (SD)
PI-harm 5.0 (5.7) 7.8 (7.5) 5.9 (5.2) 6.5 (6.5) χ 22 = 2.533 .282
Average (SD)
PI-impulse 3.1 (3.7) 4.8 (7.4) 2.3 (3.5) 3.6 (5.6) χ 22 = 1.325 .516
Average (SD)
PI total 26.5 (25.0) 47.6 (29.8) 25.0 (18.0) 35.3 (27.6) χ 22 = 12.414 .002
Average (SD)
SCI-MOODS–rhythm 12.9 (5.3) 14.2 (5.0) 16.3 (4.9) 14.4 (5.2) χ 22 = 5.973 .050
Average (SD)
SCI-MOODS–depression 35.0 (15.5) 39.1 (12.0) 39.6 (14.7) 38.0 (13.8) χ 22 = 1.446 .485
Average (SD)
SCI-MOODS–mania 27.0 (11.8) 24.2 (13.0) 38.5 (12.0) 28.8 (13.6) χ 22 = 16.226 b.001
Average (SD)
SCI-MOODS total 74.7 (27.4) 77.5 (26.5) 94.4 (28.4) 81.1 (28.1) χ 22 = 6.759 .034
Average (SD)
SI-R–clutter 9.1 (8.3) 6.7 (5.9) 8.8 (6.1) 8.0 (6.8) χ 22 = 2.085 .352
Average (SD)
SI-R–difficulty discarding 10.8 (8.0) 8.4 (7.0) 9.2 (6.1) 9.3 (7.1) χ 22 = 1.413 .493
Average (SD)
SI-R–acquisition 16.4 (7.5) 7.6 (5.5) 8.1 (6.2) 10.4 (7.4) χ 22 = 18.670 b.001
Average (SD)
SI-R total 36.3 (20.1) 22.8 (16.1) 26.1 (16.7) 27.7 (18.3) χ 22 = 6.916 .031
Average (SD)
 T.Z. Filomensky et al. / Comprehensive Psychiatry 53 (2012) 554–561 559

Table 3
Multinomial logistic regression for CB, carriers of OCD, and BD-I (category
of reference: compulsive buyers, N=80, final model)
Diagnosis: OCD Wald Pa OR 95% confidence
χ2 interval of OR
Lower Upper
bound bound
Age 5.353 .021 0.853 0.745 0.976
BIS-nonplanning 3.446 .063 0.853 0.721 1.009
PI-washing 5.934 .015 1.194 1.035 1.378
PI-checking 2.896 .089 1.148 0.979 1.345
SCI-MOODS–mania 2.822 .093 1.108 0.983 1.248
SI-R–acquisition 15.825 b.001 0.636 0.508 0.795
Sex b 7.061 .008 19.344 2.176 171.991
Constant 5.887 .015 – – –

Diagnosis: BD-I
Age 0.220 .639 1.032 0.903 1.180
BIS-nonplanning 7.018 .008 0.824 0.714 0.951
PI-washing 1.115 .291 1.085 0.932 1.263
PI-checking 2.319 .128 0.827 0.648 1.056
SCI-MOODS–mania 12.310 b.001 1.321 1.131 1.543
SI-R–acquisition 11.769 .001 0.682 0.548 0.848
Sex b 0.196 .658 1.665 0.175 15.886
Constant 0.065 .798 – – –
Model information: χ 214= 104.7, P b .001, Nagelkerke R2 = 0.826 (P b .001).
OR indicates odds ratio.
a
Degrees of freedom = 1. Fig. 1. Canonical discriminant functions.
b
Category of reference: male.

Table 5 compares the original diagnostic classification

differences among the diagnostic groups. The significance with the reclassification obtained by the discriminant
level of Box's M test assured us that the variables inserted functions. Cohen κ coefficient (κ = 0.639) presented a
into the discriminant model showed no different covariance substantial and significant concordance [38] (P b .001)
among the diagnostic groups (P = .364). between the 2 classifications.
The discriminant analysis generated 2 functions that In an alternative approach, we kept the same discriminant
significantly differentiated the diagnostic groups. Function 1 analysis parameters; however, this time, we applied a
accounted for 62.4% of the variance (Eigenvalue = 0.837, stepwise procedure. The results were quite similar, but,
χ 212 = 75.7, P b .001), and function 2, for 37.6% of the nevertheless, the age and PI-washing variables were ruled
variance (Eigenvalue = 0.504, χ 25 = 30.4, P b .001). Table 4 out of the model. The 2 discriminant functions generated
describes the correlation between the model variables and the corresponded, respectively, to 56.6% and 43.4%, both highly
scores of functions 1 and 2. Acquisition and nonplanning are significant (P b .001). Function 1 presented higher
directly related with function 2, whereas the mania score is correlations with PI-checking (values adjusted by logarith-
inversely related with it. It can be noted from Fig. 1 that mic transformation, R = −0.598) and SCI-MOODS–mania
function 1, on the horizontal axis, discriminates patients with (R = 0.480) score. Function 2 presented higher correlations
OCD from the rest of the sample, whereas function 2, on the with SI-R–acquisition (values adjusted by logarithmic
vertical axis, separates patients with CB (upper right corner) transformation; R = −0.634) and BIS-nonplanning (R =
from patients with BD-I (lower right corner).
Table 5
Table 4 Diagnostic classification and rediagnostic classification by means of
Correlation among the variables and the discriminant function analysis discriminant analysis of the CB, carriers of OCD, and BD-I
Variables Function Diagnostic Rediagnostic classification
classification
1 2 CB OCD BD-I Total
log(PI-washing) −0.510 −0.098 Original CB 19 3 2 24
log(PI-checking) −0.516 0.064 OCD 2 27 6 35
Age 0.346 −0.123 BD-I 5 1 15 21
log(SI-R–acquisition) 0.337 0.578 (%) CB 79.2 12.5 8.3 100.0
SCI-MOODS–mania 0.381 −0.481 OCD 5.7 77.1 17.1 100.0
BIS-nonplanning 0.207 0.464 BD-I 23.8 4.8 71.4 100.0
560 T.Z. Filomensky et al. / Comprehensive Psychiatry 53 (2012) 554–561
0.499). Function 1 discriminated better between patients
with OCD and BD-I. Function 2, represented by a score
composed of acquisition and impulsivity, segregated the
patients with CB. The concordance rate in the classification
was slightly lower for this model (71.3%) but was still highly
significant (κ = 0.562, P b .001).
4. Discussion
In this sample, we found a male-to-female ratio of 1:2.4
for patients with CB, which is in keeping with previous
studies reporting a preponderance of women in clinical
settings [5]. The cross-sectional nature of this study does not
yield causal inferences; nonetheless, the high level of
unemployment among patients with CB could be an
indication of the social maladjustment associated with this
syndrome as compared with OCD and BD-I. The remaining
sociodemographic and socioeconomic data are compatible
with the profile of Brazil's middle class [39].
The most distinctive feature of patients with CB is their
higher impulsivity, whereas OCD patients presented exces-
sive scores in the obsession and compulsion measures, as
expected, particularly in the syndrome's more typical
subfactors, that is, contamination/washing and checking.
Equally coherent, bipolar patients presented higher scores in
mania and rhythm but not for depression, which reinforces
the previous accounts of the high frequency of depressive
symptoms in CB and OCD and the frequent comorbidity of
these syndromes with a depressive disorder. Thus, depres-
sive symptoms can be considered a feature equally shared by
the 3 syndromes.
Our hypothesis that patients with OCD would score
higher on storing variables was not confirmed. On the
contrary, the more obvious factors, hoarding and difficulty
discarding, classically related to OCD, were not different for
any of the diagnostic groups. This finding reinforces
arguments that hoarding is not particularly related to OCD,
at least not any more than it is to CB and BD-I, and that
pathologic hoarding could configure a separate syndrome or
symptom cluster [16,35].
On the other hand, patients with CB scored higher on the
acquisition subfactor, and combined with their high level of
impulsivity, these findings strongly suggest that CB is
characterized by 2 closely related dimensions: impulsivity
expressed as an inability to anticipate the future (nonplan-
ning) and an above-average drive to acquire/purchase
objects, which is in keeping with the clinical accounts of
the syndrome. Interestingly, the acquisition subfactor from
the hoarding construct, thought to be a link between CB and
OCD, was related to the impulsive features of compulsive
buyers. This finding of a relationship of CB only to the
acquisition subdimension of hoarding is in keeping with
results from Mueller et al [15]. Although speculative, in
contrasting the current findings with previous ones, it seems
that there could be an interesting division resulting from the
interaction between an acquisition drive and personality
features: patients with preponderant impulsive traits would
fall into the CB category, whereas patients with predominant
compulsive features would fall into the classic description of
the hoarding syndrome presenting greater similarities with
OCD. An investigation of the relationship between CB and
primary hoarding syndrome is a needed complement to the
current study and would come timely as we approach the
publication of the fifth edition of the DSM because proper
classification for CB has not been devised yet.
The multivariate analysis reinforces the impressions
derived from the univariate analysis. The multiple regression
model, controlled for variations in the sociodemographic
status, proved to be highly significant, with roughly 86% of
the variance explained by the final model. The discriminant
analysis, which was also statistically significant, correctly
differentiated and reclassified nearly 80% of the sample
based on the variables selected. Examination of the functions
generated suggests that the obsessive-compulsive symptoms
and traits were typical of the OCD group, just as mania was
for the BD-I group, and that CB was not closely aligned with
any of these disorders but, rather, represented a group apart,
distinguished by the impulsive desire for acquisition.
Nonetheless, the reclassification of the sample and its
graphic representation both based on the discriminant
analysis suggest that there may be some heterogeneity in
the CB group and that some compulsive buyers may share
similarities with patients with OCD or BD.
This study presents some important limitations, such as a
clinical and convenience sample, that limits the possibility of
generalization to individuals from the community who are not
either receiving or looking for treatment. Despite the current
evidences favoring the classification of CB as an ICD and a
possible behavioral addiction, its definite classification must
be ascertained after further investigation including an in-depth
examination of features shared between CB, substance
addictions, and other ICDs, especially behavioral addictions;
comparison with patients with OCD with predominant
hoarding symptoms or pure hoarders; and, finally, comparison
with other diagnostic categories within the bipolar spectrum of
disorders, especially BD type II. Furthermore, it is important
to acknowledge that, given the relatively small sample size
and the need to avoid inflating the number of statistical
comparisons, only single measures of impulsivity (BIS-11)
and compulsivity (PI) features were applied in this study.
Hence, conclusions about psychopathologic nature of the
diagnostics included in this study may endure the limitations
inherent to the shortcomings of the scales applied.
On the other hand, this sample, derived from specialized
outpatient services, shows cardinal differences between
patients whose primary motivation for seeking treatment is
their loss of control over buying and patients with OCD and
BD-I; this distinction provides useful insights for a clinical
approach. As in other behavioral addictions, such as
gambling, impulsivity, and loss of control, were revealed
as the most striking features [12] rather than obsessions and
 T.Z. Filomensky et al. / Comprehensive Psychiatry 53 (2012) 554–561
controlling rituals or mood fluctuations and related phenom-
ena, and this new insight reveals the need for a reorientation
of clinical approaches toward addictions, ICD, or both.
Subsequent studies must help foster a better understanding of
the psychopathology of CB, finding a more accurate
classification that will allow full acknowledgment of a
disorder that is prevalent and severe but that, nonetheless,
has been surprisingly neglected until now.
References
[1] Kraepelin E. Psychiatric. Leipzig, Germany: Verlag Von Johann
Ambrosius Barth; 1915. p. 409.
[2] Bleuler E. Textbook of psychiatry. New York, NY: McMillan; 1924.
[3] McElroy SL, Keck Jr PE, Pope Jr HG, Smith JM, Strakowski SM.
Compulsive buying: a report of 20 cases. J Clin Psychiatry 1994;55(6):242-8.
[4] Koran LM, Faber RJ, Aboujaoude E, et al. Estimated prevalence of
compulsive buying behavior in the United States. Am J Psychiatry
2006;163:1806-12.
[5] Black DW. Compulsive buying disorder: a review of evidence. CNS
Spectr 2007;12:124-32.
[6] Tavares H, Lobo DS, Fuentes D, Black DW. Compulsive buying disorder:
a review and a case vignette. Rev Bras Psiquiatr 2008;30(1):16-23.
[7] Filomensky TZ, Tavares H. Cognitive restructuring for compulsive
buying. Rev Bras Psiquiatr 2009;31(1):76-81.
[8] Mueller A, Mueller U, Silbermann A, Reinecker H, Bleich S, Mitchell
JE, et al. A randomized, controlled trial of group cognitive-behavioral
therapy for compulsive buying disorder: posttreatment and 6-month
follow-up results. J Clin Psychiatry 2008;69:1131-8.
[9] Koran LM, Chuong HW, Bullock KD, Smith SC. Citalopram for
compulsive shopping disorder: an open-label study followed by
double-blind discontinuation. J Clin Psychiatry 2003;64:793-8.
[10] American Psychiatric Association. Diagnostic and Statistical Manual
of Mental Disorders. 4th ed., text revised. 2004. Washington, DC:
American Psychiatric Association.
[11] McElroy SL, Pope H, Keck P, Hudson J, Phillips K, Strakowski S. Are
impulse-control disorders related to bipolar disorder? Compr Psychi-
atry 1996;37(4):229-40.
[12] Tavares H, Gentil V. Pathological gambling and obsessive compulsive
disorder: towards a spectrum of disorders of volition. Rev Bras
Psiquiatr 2007;29(2):107-17.
[13] Petry NM, Stinson FS, Grant BF. Comorbidity of DSM-IV
pathological gambling and other psychiatric disorders: results from
the National Epidemiologic Survey on Alcohol and Related Condi-
tions. J Clin Psychiatry 2005;66(5):564-74.
[14] Karakus G, Tamam L. Impulse control disorder comorbidity among
patients with bipolar I disorder. Compr Psychiatry 2011;52(4):378-85.
[15] Mueller A, Mueller U, Albert P, Mertens C, Silbermann A, Mitchell
JE, et al. Hoarding in a compulsive buying sample. Behav Res Ther
2007;45:2754-63.
[16] Frost RO, Steketee G, Williams L. Compulsive buying, compulsive
hoarding, and obsessive-compulsive disorder. Behav Ther 2002;33(2):
201-14.
[17] Pertusa A, Fullana MA, Singh S, Alonso P, Menchón JM, Mataix-Cols
D. Compulsive hoarding: OCD symptom, distinct clinical syndrome,
or both? Am J Psychiatry 2008;165:1289-98.
[18] Saxena S. Is compulsive hoarding a genetically and neurobiologically
discrete syndrome? Implications for diagnostic classification. Am J
Psychiatry 2007;164:380-4.
561
[19] Grant JE, Potenza MN, Weinstein A, Gorelick DA. Introduction to
behavioral addictions. Am J Drug Alcohol Abuse 2010;36(5):
233-41.
[20] Grant JE, Brewer JA, Potenza MN. The neurobiology of substance and
behavioral addictions. CNS Spectr 2006;11(12):924-30.
[21] O'Brien C. Addiction and dependence in DSM-V. Addiction 2011;
106(5):866-7.
[22] Grant JE, Kim SW, Hartman BK. A double-blind, placebo-controlled
study of the opiate antagonist naltrexone in the treatment of
pathological gambling urges. J Clin Psychiatry 2008;69(5):783-9.
[23] Bullock K, Koran L. Psychopharmacology of compulsive buying.
Drugs Today (Barc) 2003;39(9):695-700.
[24] Vieta E, Gastó C, Otero A, Nieto E, Vallejo J. Differential features
between bipolar I and bipolar II disorder. Compr Psychiatry 1997;38(2):
98-101.
[25] First MB, Spitzer RL, Gibbon M, Williams JBW. Structured clinical
interview for DSM-IV axis I disorders—clinician version (SCID-CV).
Washington (DC): American Psychiatric Press; 1997. p. 84.
[26] Del-Bem CM, Vilela JAA, Crippa JAS, Hallak JEC, Labate CM,
Zuardi AW. Reliability of the Structured Clinical Interview for DSM-
IV—Clinical Version translated into Portuguese. Rev Bras Psiquiatria
2001;23(3):156-9.
[27] Tavares H, Zilberman ML, el-Guebaly N. Are there cognitive and
behavioral approaches specific to the treatment of pathological
gambling? Can J Psychiatry 2003;48:22-7.
[28] Hollingshead AB, Redlich FC. Social class and mental illness: a
community study. New York: John Wiley & Sons; 1958.
[29] Barrat ES. Impulsivity: integrating cognitive, behavioral, biological
and environmental data. In: McCowan W, Johnson J, & Shure M,
editors. The impulsive client: theory, research and treatment.
Washington (DC): American Psychological Association; 1993.
[30] Burns GL, Keortge SG, Formea GM, Sternberger LG. Revision of the
Padua Inventory of obsessive compulsive disorder symptoms:
distinctions between worry, obsessions, and compulsions. Behav Res
Ther 1996;34(2):163-73.
[31] Fagiolini A, Dell'Osso L, Pini S, Armani A, Bouanani S, Rucci P,
et al. Validity and reliability of a new instrument for assessing
mood symptomatology: the Structured Clinical Interview for Mood
Spectrum (SCI-MOODS). Int J of Methods in Psych Research
1999;8(2):71-82.
[32] World Health Organization (WHO). ICD-10 Classification of Mental
and Behavioral Disorders: clinical descriptions and diagnostic
guidelines. Porto Alegre: Artes Médicas; 1993.
[33] Cassano GB, Dell'Osso L, Frank E, Miniati M, Fagiolini A, Shear K,
et al. The bipolar spectrum: a clinical reality in search of diagnostic
criteria and an assessment methodology. J Affect Disord 1999;54(3):
319-28.
[34] Dell'Osso L, Armani A, Rucci P, Frank E, Fagiolini A, Corretti G,
et al. Measuring Mood Spectrum: comparison of interview (SCI-
MOODS) and self-report (MOODS-SR) instruments. Compr Psychi-
atry 2002;43(1):69-73.
[35] Frost RO, Steketee G, Grisham J. Measurement of compulsive
hoarding: saving inventory-revised. Behav Res and Ther 2004;42:
1163-82.
[36] Tabachnick BG, Fidell LS. Using multivariate statistics. 3rd ed.
Northridge: Harper Collin College Publishers; 1996.
[37] SPSS. Statistical Package for Social Sciences for Windows, v.16.0.
Chicago: SPSS Inc.; 2008.
[38] Landis JR, Koch GG. The measurement of observer agreement for
categorical data. Biometrics 1977;33:159-74.
[39] Brazilian Institute of Geography and Statistic (I.B.G.E.). National
sample survey of households. Brazil Volume; 1998.