Vous êtes sur la page 1sur 9

Original Research

Subtyping Study of a Pathological Gamblers Sample


Eva Mª Álvarez-Moya, PhD1; Susana Jiménez-Murcia, PhD2; Mª Neus Aymamí, MSc3;
Mónica Gómez-Peña, MSc4; Roser Granero, PhD5; Juanjo Santamaría, MSc6;
Jose M Menchón, MD7; Fernando Fernández-Aranda, PhD8

Objective: To classify into subgroups a sample of pathological gambling (PG) patients


according to personality variables and to describe the subgroups at a clinical level.
Method: PG patients (n = 1171) were assessed with the South Oaks Gambling Screen; the
Temperament and Character Inventory—Revised; the Symptom Checklist–90—Revised;
Eysenck’s Impulsivity Scales, a diagnostic questionnaire for the Diagnostic and Statistical
Manual of Mental Disorders, Fourth Edition (DSM-IV) PG criteria; and the Structured Clinical
Interview for the DSM-IV, Axis I disorders, substance use module. Clinical measures were
collected through a semi-structured interview. We performed a 2-step cluster analysis based
on the above-mentioned personality variables. Clinical data were compared across clusters.
Results: Four clusters were generated. Type I (disorganized and emotionally unstable)
showed schizotypic traits, high impulsiveness, substance and alcohol abuse, and early age
of onset, as well as psychopathological disturbances. Type II (schizoid) showed high harm
avoidance, social aloofness, and alcohol abuse. Type III (reward sensitive) showed high
sensation seeking and impulsiveness but no psychopathological impairments. Type IV
(high-functioning) showed a globally adaptive personality profile, low level of substance and
alcohol abuse or smoking, and no psychopathological disturbances.
Conclusions: At least 4 types of PG patients may be identified. Two types showed a
response modulation deficit, but only one of them had severe psychopathological
disturbances. Two other types showed no impulsiveness or sensation seeking and one of
them even exhibited good general functioning. The different personality and clinical
configuration of these clusters might be linked to different therapeutic approaches.
Can J Psychiatry. 2010;55(8):498–506.

Clinical Implications
· At least 4 subtypes of pathological gamblers, with differing severity level, comorbidity,
and personality profiles, may be distinguished.
· Only some groups show high impulsiveness and (or) high sensation seeking.
· PG is a highly heterogeneous disorder that may require different therapeutic
approaches according to its specific characteristics.

Limitations
· Our sample consisted of treatment-seeking PG patients, which may affect
generalizability of findings.
· Most patients were mainly slot machine pathological gamblers.
· Comorbidity with Axis I disorders other than substance use disorders was not directly
assessed.

Key Words: pathological gambling, subtypes, personality, impulsiveness,


sensation seeking, harm avoidance, substance abuse, alcohol abuse

498 W La Revue canadienne de psychiatrie, vol 55, no 8, août 2010


Subtyping Study of a Pathological Gamblers Sample

he DSM-III1 included PG for the first time as an impulse


T control disorder. SUDs are the most frequent comorbid
features of PG, ranging from 30% to 70%.2,3 Increasing evi-
personality characteristics in PG will be considered in our
study.

dence points to PG and other disorders of impulse control (for


Given the multiple characterization of PG, some authors pro-
example, bulimia nervosa or sexual addictions) as part of an
posed the existence of different subgroups based on
addictive spectrum sharing the same underlying
psychopathological or phenomenological features. Meyer17
biopsychosocial process, together with other types of addic-
identified 5 clusters in a sample of 437 German gamblers
tions.4 In this regard, PG has been considered as an addictive
from self-help groups, namely, emotionally unstable with
disorder4 or behavioural addiction.5 In studies of alcohol- or
depressive-aggressive personality, emotionally unstable
cocaine-dependent patients, the presence of comorbidity with
with depressive personality, pathological gamblers on
other SUDs frequently characterizes specific subgroups,6 and
German-style slot machines, pathological gamblers on clas-
drug of choice is frequently associated with specific personal-
sical games of chance (conspicuous personality), and patho-
ity profiles.7
logical gamblers on German-style slot machines under a
Concerning personality, impulsiveness and sensation seeking subjective strain. Differently, the study by González-Ibáñez
have received much attention and, together with low et al18 identified 3 clusters in a sample of 110 pathological
reward-dependence (sociability and emotional dependence), gmblers seeking treatment in a specialized unit, which dif-
cooperativeness and self-directedness (goal-directed behav- f e r e d ma in ly o n th e s e v e r ity o f th e as s o c ia te d
iour) have been identified as risk factors for the development psychopathology and showed no clear differences on sensa-
of PG and the modulation of gambling behaviour.8–10 Extrava- tion seeking. Ledgerwood and Petry19 identified 3 factors
gance, avoidant, and obsessive–compulsive personality have that derived from a questionnaire assessing the experience of
also been identified as associated personality factors in gambling, that is, escape, dissociation, and egotism (narcis-
PG.11,12 However, there is no agreement about their specific sisticlike traits). Cunningham-Williams and Hong20 applied
personality profile. The role of impulsiveness and sensation latent class analysis to a sample of community-recruited
seeking in PG is unclear, as some authors report low levels of gamblers and identified a 6-class solution reflecting a contin-
both traits in relation to control subjects.13,14 These inconsis- uum of problem gambling risk based on 8 indicators of an
tencies may derive from the use of impulsivity as a unitary instrument, the Computerized-Gambling Assessment Mod-
construct. Recent studies suggest that impulsivity involves at ule.21 Similar results based on the distribution of DSM-IV
least 2 separate factors, that is, rash impulsiveness (acting criteria for PG in community gamblers were reported by
rashly when distressed) and sensitivity to reward (greater Toce-Gerstein et al. 22 From a theoretical viewpoint,
response or activation to rewarding stimuli).15,16 Some Blaszczynski and Nower23 proposed a pathways model that
authors consider rash impulsiveness as a risk factor for included 3 subgroups, that is, behaviourally conditioned
disinhibited behaviour and then for the progression from sub- problem gamblers, emotionally vulnerable problem gam-
stance use to substance dependence, while sensitivity to blers, and antisocial, impulsivist problem gamblers. More
reward would be more associated with motivation to use sub- recently, based on a literature review, Iancu et al24 proposed
stances than with substance dependence per se.15,16 Although 3 subtypes of PG, that is, impulsive, addictive, and
both terms are frequently used indistinctly, in our report we obsessive–compulsive (mainly females).
use the term impulsiveness as closer to rash impulsiveness,
while sensation seeking refers to the sensitivity to reward con-
struct. Impulsivity will be used as a global term including both Most studies agree in the role of emotional instability and
traits. The relevance and interaction of these traits and other Cluster B traits (including impulsiveness and sensation seek-
ing among others) in the differentiation of pathological gam-
blers. To a lesser degree, addictive features are also
considered. In general, subtyping studies include different
variables, sample, and methods, and the subgroups reported
do not coincide and are noncomparable.
Abbreviations used in this article
DSM Diagnostic and Statistical Manual of Mental Disorders
With the aim of shedding light on the subtyping of pathologi-
GSI Global Severity Index
cal gamblers and considering a biopsychosocial approach,
PG pathological gambling our main objective was to obtain an empirical classification
SCID Structured Clinical Interview for DSM-IV of PG patients according to personality variables. Secondly,
SCL-90-R Symptom Checklist–90 Items—Revised we aimed to describe the resultant groups in terms of clinical
SOGS South Oaks Gambling Screen and sociodemographic variables. According to Blaszczynski
and Nower’s23 pathways model, we hypothesized that 3 sub-
SUD substance use disorder
groups would be identified, corresponding to behaviourally
TCI-R Temperament and Character Inventory—Revised conditioned, emotionally vulnerable, and antisocial,
impulsivist problem gamblers.

The Canadian Journal of Psychiatry, Vol 55, No 8, August 2010 W 499


Original Research

Method and k = 0.82–1.0 for other SUD).34 Presence of alcohol use


Participants disorders and SUD were used separately as binary variables
An initial sample of 1576 PG patients attending a PG unit was for our study.
considered. All of them were consecutive referrals for assess-
Psychometric Personality Measures
ment and treatment and were diagnosed according to
DSM-IV-TR criteria.25 The entry into the study was from May TCI-R35 is a 240-item questionnaire measuring 7 personality
2003 to July 2007. This study was carried out according to the factors. Temperamental factors include Harm Avoidance,
latest version of the Declaration of Helsinki. The Ethics Com- Reward Dependence, Novelty Seeking (reflects both sensa-
mittee of our hospital approved this study and informed con- tion-seeking and impulsiveness), and Persistence. Character
sent was obtained from all final participants. dimensions include Self-Directedness, Cooperativeness, and
Self-Transcendence. Reliability of the different personality
Exclusion criteria were the following: presence of severe, dimensions was good in the Spanish adaptation,36 ranging
acute psychiatric disorders (for example, acute psychosis, from 0.77 to 0.84. Given our interest in the configuration of
substance intoxication, or manic episodes) (n = 148); missing general personality across subgroups, scale scores in all 7
information in the main measures (n = 97); and rejecting par- factors were used for cluster analysis. Considering the rele-
ticipation in the study (n = 160). vance of impulsivity in PG, the Novelty Seeking factor
The final sample consisted of 1171 Caucasian PG patients subscales (Exploratory Excitability, Impulsiveness, Extrava-
(7.7% women, n = 90) with a mean age of 39.4 years (SD gance, and Disorderliness) were used a posteriori as
12.4). Most patients had elementary education (76.3%) and descriptors of the subgroups. Exploratory Excitability
were employed (74.2%). More than one-half of the sample reflects sensation novelty seeking and boredom proneness
(54.8%) were married and 31.4% were single. (similar to sensitivity to reward), Impulsiveness reflects
unreflective and careless behaviour (similar to rash impul-
Regarding their main gambling problem, 90.2% were
siveness), Extravagance reflects over-spending behaviour,
slot-machine gamblers, 17.6% were bingo gamblers, 9.7%
and Disorderliness reflects antinormative behaviour.
lottery gamblers, 6.3% casino gamblers, 6.7% had other
games (cards, bets) as main gambling problem, and 10.4% The Eysenck Impulsivity Scale I737 is a 54-item self-report
had problems with several types of gambling. Their mean scale that measures 2 dimensions of impulsivity, that is,
monthly income was €1251.7 (SD 646.7), and 68.2% had Impulsiveness (similar to rash impulsiveness) and Venture-
gambling-related debts (from €50 to €120 000). someness (similar to sensation seeking and sensitivity to
reward), and one dimension of Empathy. Eysenck et al37
Material and Assessment reported test–retest reliabilities of 0.78 and 0.90 for the Ven-
turesomeness and Impulsiveness subscales, which were used
Clinical Measures for cluster analysis in our study.
SOGS26 is a 20-item screening questionnaire that discrimi-
nates among probable PG, problem gambling, and nonprob-
lem gambling. Psychometric properties of the Spanish Procedures
validation27 were the following: test–retest reliability 0.98, First interview (at intake) was addressed to elaborate the clin-
internal consistency 0.94, and convergent validity 0.92. ical history by means of a semi-structured face-to-face inter-
view. Experienced psychologists and psychiatrists
The diagnostic questionnaire for PG according to DSM-IV (specialists in PG) collected sociodemographic and clinical
criteria28 is a 19-item questionnaire reflecting the DSM-IV data. The above-mentioned measures were administered dur-
diagnostic criteria for PG. In the Spanish adaptation,29 the ing a second session before treatment.
standard DSM-IV cut-off score of 5 criteria yielded satisfac-
tory classification accuracy results with high sensitivity (0.92) Statistical Analyses
and specificity (0.99). Statistical analysis was carried out using SPSS version 15.0.1
The SCL-90-R30 is a 90-item scale for assessing self-reported (SPSS Inc, Chicago, IL).
psychological distress and psychopathology. It measures 9 The 7 TCI-R factor scores and I7 Impulsiveness and Venture-
primary symptom dimensions (for example, somatization, someness scales were used for cluster analysis. The identifi-
obsessive–compulsive disorder, and depression). A global cation of empirical homogeneous groups was performed by
index, that is, the GSI was also used for our study. Reliability using a 2-step cluster analysis, which seeks to identify homo-
ranged from 0.81 to 0.90 for the different subscales in the geneous subgroups of cases in a population by both minimiz-
Spanish validation.31 ing within-group variation and maximizing between-group
The SUDs module of the SCID-I32 is a semi-structured inter- variation. Cluster analysis is more often viewed as an explor-
view that establishes the most important DSM-IV Axis I diag- atory procedure rather than a testing hypothesis method such
noses. Its psychometric properties for SUD showed good as latent class analysis. The 2-step method is a
discriminant, concurrent, and predictive validity,33 as well as one-pass-through-the-data approach that addresses the scal-
good interrater reliability (k = 0.94 for alcohol use disorder ing problem by identifying preclusters in a first step, then

500 W La Revue canadienne de psychiatrie, vol 55, no 8, août 2010


Subtyping Study of a Pathological Gamblers Sample

Table 1 Cluster composition (centroids) and results of ANOVA tests (n = 1171)


Cluster 1 Cluster 2 Cluster 3 Cluster 4
n = 341; 29.1% n = 238; 20.3% n = 238; 20.3% n = 354; 30.2%

Scores on personality traits Mean (SD) Mean (SD) Mean (SD) Mean (SD) Fd

TCI-R NS 116.3 (13.0) 105.4 (11.2)a 122.3 (12.0)a,b 98.3 (11.8)a,b,c 234.3
a a,b
TCI-R HA 108.9 (14.6) 114.4 (14.5) 82.8 (13.6) 95.8 (14.2)a,b,c 250.3
a a,b a,b
TCI-R RD 95.4 (14.7) 91.2 (15.3) 110.1 (13.4) 109.8 (14.0) 128.5
TCI-R PS 112.4 (17.8) 87.7 (16.2)a 122.7 (19.8)a,b 119.9 (17.0)a,b 202.6
a a,b a,b,c
TCI-R SD 108.0 (15.2) 126.0 (17.9) 132.9 (16.2) 145.5 (17.7) 297.5
TCI-R CO 121.7 (17.4) 129.8 (16.0)a 141.3 (12.9)a,b 147.3 (11.6)a,b,c 200.2
a a,b a,b
TCI-R ST 71.8 (14.8) 53.8 (11.4) 68.3 (15.6) 68.1 (15.9) 76.5
I7 Impulsiveness 12.0 (2.9) 7.1 (3.1)a 10.3 (2.9)a,b 4.8 (2.7)a,b,c 413.9
I7 Venturesomeness 7.7 (3.8) 4.5 (3.0)a 10.4 (3.5)a,b 5.2 (3.3)a,c 155.5

Schwarz Bayesian Information Criterion = 6060.921; ratio of distance measures = 1.534


CO = Cooperativeness; HA = Harm Avoidance; NS = Novelty Seeking; PS = Persistence; RD = Reward Dependence; SD = Self-Directedness;
ST = Self-Transcendence
a
Statistically significant in comparison to Cluster 1
b
Statistically significant in comparison to Cluster 2
c
Statistically significant in comparison to Cluster 3
d
df = 3,1167; P < 0.001

treating these as single cases in a second step that uses hierar- Results
chical clustering. By default, the SPSS algorithm uses a com-
bination of the Schwarz Bayesian Information Criterion Cluster Composition
(though the Akaike Information Criterion may be selected by The 2-step procedure generated a classification based on 4
the researcher) and log-likelihood distance in auto- clusters. Table 1 shows the centres of every empirical group
determining the number of clusters. SPSS will pick a solution and the results of between-group comparisons. Cluster 1 was
with a reasonably large ratio of Schwarz Bayesian Informa- characterized by the highest scores in I7 Impulsiveness and
tion Criterion changes and a large ratio of distance measures. TCI-R Self-Transcendence, and the lowest scores in TCI-R
Cases are categorized under the cluster that is associated with Self-Directedness and Cooperativeness. They showed rela-
the largest log-likelihood. The SPSS algorithm uses a tively high TCI-R Novelty Seeking and Harm Avoidance
decrease in log-likelihood for combining clusters as the dis- scores.
tance measure.38 The log-likelihood measure is computed by
using the normal density for continuous variables and the Cluster 2 showed the highest scores in TCI-R Harm Avoid-
multinomial probability mass function for categorical vari- ance, and the lowest scores in TCI-R Reward Dependence,
ables.39 We let the 2-step algorithm automatically determine Persistence, and Self-Transcendence. They also showed low
the number of clusters. I7 Venturesomeness and relatively low I7 Impulsiveness,
TCI-R Novelty Seeking, Self-Directedness, and Coopera-
Then, we performed ANOVA and chi-square tests between tiveness scores. This cluster was characterized by difficulties
clusters and sociodemographic (age, employment status, and in impulse control, low planning, and low perseverance.
marital status), clinical (SCL-90-R subscales and GSI, total
SOGS score, number of fulfilled DSM-IV criteria, presence Cluster 3 was characterized by the highest scores in TCI-R
of substance and [or] alcohol abuse, and smoking status), and Novelty Seeking, Reward Dependence, Persistence, and I7
personality (TCI-R Novelty Seeking factor subscales) vari- Venturesomeness. They also showed the lowest TCI-R Harm
ables not included in the cluster analysis. This information Avoidance scores and relatively high scores in TCI-R
was used to externally validate the cluster solution. These Self-Directedness, Cooperativeness, and I7 Impulsiveness.
comparisons were adjusted for age, sex, and duration of the
disorder when needed. Cluster 4 was characterized by the lowest TCI-R Novelty
Seeking and I7 Impulsiveness scores, and the highest TCI-R
To correct for multiple comparisons, we established a stan- Reward Dependence, Persistence, Self-Directedness, and
dard alpha level of 0.01. Cooperativeness scores, as well as low I7 Venturesomeness.

The Canadian Journal of Psychiatry, Vol 55, No 8, August 2010 W 501


Original Research

Table 2 Sociodemographic characteristics of the clusters


Scores on personality Cluster 1 Cluster 2 Cluster 3 Cluster 4
traits n = 341 n = 238 n = 238 n = 354 F or c2(df)a

Age, mean (SD) 37.8 (12.4) 40.6 (11.6) 35.3 (10.4) 43.0 (13.3) 22.3 (3,1167)
Education, %
Elementary 81.5 79.8 74.4 70.0 21.6 (6)
Secondary 16.9 14.2 18.4 24.0
Post-secondary 1.6 6.0 7.2 6.0
Employed, % 66.9 73.6 84.0 75.3 21.2 (3)

Only statistically significant results at the level P < 0.01 are shown.
a
P < 0.001

They also showed relatively low TCI-R Harm Avoidance 4 showed systematically the lowest percentage of alcohol,
scores. substance, and tobacco abuse. Cluster 2 showed the highest
percentage of alcohol abuse and Cluster 1 showed the highest
Sociodemographic Characteristics of the Clusters percentage of substance abuse. Cluster 3 showed the highest
In Table 2, the clusters differed significantly on age, educa- percentage of tobacco smokers.
tion, and employment. Post-hoc comparisons showed that
patients of Clusters 1 and 3 were significantly younger than Distribution of the TCI-R Novelty Seeking Factor
patients of Clusters 2 and 4. Cluster 1 showed the lowest edu- Subscales Scores Across Clusters
cation level, that is, the highest percentage of patients with In Table 4, statistically significant differences were observed
elementary education and the lowest percentage of patients on all Novelty Seeking factor subscales. Cluster 3 showed the
with post-secondary education. Regarding employment, highest Exploratory Excitability, while Cluster 2 showed the
Cluster 1 showed the lowest percentage of people who were lowest scores in this subscale. Clusters 1 and 3 showed the
employed, while Cluster 3 showed the highest percentage of highest scores in Impulsiveness and Cluster 4 showed the
employment. Some tendencies to statistical significance lowest ones. Cluster 4 showed the lowest scores in Extrava-
appeared for sex (Cluster 1: 10.3% women; Cluster 2: 9.2%; gance, while Cluster 3 showed the highest ones. Finally,
Cluster 3: 4.2%; Cluster 4: 6.5%; P = 0.03) and marital status Clusters 3 and 1 showed the highest Disorderliness and Clus-
(Cluster 4 showed the highest percentage of married patients; ter 4 the lowest. Adjustments for age and sex yielded no dif-
P = 0.03). No statistically significant findings were observed ferences in these results.
for the presence of debts, monthly income, or type of main
gambling problem. Adjustment for age yielded no differences Discussion
in the results. The aim of our paper was to present a new classification of
PG patients according to personality variables, as well as to
Clinical Correlates of the Clusters characterize the subgroups regarding clinical aspects and
In Table 3, in general Cluster 1 patients showed the highest specific variants of impulsivity, as measured by the Novelty
scores on all SCL-90-R subscales (including the General Seeking factor subscales.
Symptom Index), followed by Cluster 2, and Clusters 3 and 4,
Cluster analysis automatically generated 4 empirical groups.
which showed similar scores in almost all subscales. Regard-
Cluster 1 was characterized by difficulties with impulse con-
ing the number of DSM-IV criteria for PG fulfilled, Cluster 1
trol (rash impulsiveness), low cooperativeness, and high
showed the highest score and Cluster 4 the lowest. Clusters 1
mysticism or spirituality. This profile was labelled as disor-
and 3 showed the highest SOGS total score and Cluster 4
ganized (schizotypic) by Cloninger40 and describes an illogi-
showed the lowest one. Statistical differences were also
cal, suspicious, and immature personality, which also shows
observed regarding age of onset of gambling problems, with
magical thinking and unconventional behaviour. These peo-
Clusters 1 and 3 showing the earliest age of onset and Cluster
ple were labelled as Type I pathological gamblers and
4 the oldest one. No changes in these results were observed
showed a disorganized and emotionally unstable profile.
after adjusting for age, sex, and duration of the disorder.
They exhibited the most severe psychopathological profile
Percentage of comorbidity with SUDs, including alcohol and (the highest scores in the SCL-90-R) and the most severe
tobacco smoking, were also observed across clusters. Cluster gambling behaviour (as measured by the SOGS), as well as

502 W La Revue canadienne de psychiatrie, vol 55, no 8, août 2010


Subtyping Study of a Pathological Gamblers Sample

Table 3 Clinical characteristics of the clusters


Cluster 1 Cluster 2 Cluster 3 Cluster 4
n = 341 n = 238 n = 238 n = 354

Characteristic Mean (SD) Mean (SD) Mean (SD) Mean (SD) F (df)d

SCL-90-R
Somatization 1.2 (0.9) 0.8 (0.7)a 0.6 (0.6)a 0.6 (0.6)a,b 53.6 (3,1158)
a a,b a,b
Obsessive–compulsive 1.5 (0.9) 1.0 (0.8) 0.7 (0.6) 0.7 (0.6) 83.4 (3,1158)
Interpersonal sensitivity 1.3 (0.8) 0.9 (0.7)a 0.6 (0.5)a,b 0.5 (0.6)a,b 89.6 (3,1158)
a a,b a,b
Depression 1.7 (0.9) 1.3 (0.9) 1.1 (0.7) 0.9 (0.7) 67.4 (3,1158)
Anxiety 1.3 (0.9) 0.9 (0.7)a 0.7 (0.6)a,b 0.6 (0.6)a,b 77.0 (3,1158)
a a a,b,c
Hostility 1.3 (0.9) 0.6 (0.6) 0.6 (0.7) 0.4 (0.5) 88.4 (3,1158)
Phobic anxiety 0.7 (0.8) 0.4 (0.5)a 0.2 (0.3)a,b 0.2 (0.4)a,b 48.8 (3,1158)
a a a
Paranoid ideation 1.3 (0.9) 0.6 (0.6) 0.6 (0.6) 0.5 (0.6) 83.1 (3,1158)
Psychoticism 1.1 (0.7) 0.6 (0.6)a 0.6 (0.5)a 0.4 (0.5)a,b 75.4 (3,1158)
a a,b a,b
General Symptom Index 1.3 (0.7) 0.9 (0.6) 0.7 (0.5) 0.6 0.5) 100.6 (3,1158)
DSM-IV criteria 8.1 (1.6) 6.9 (1.9)a 7.4 (1.7)a 6.2 (2.1)a,b,c 45.2 (3,805)
a b
SOGS 11.6 (3.0) 9.9 (2.7) 11.1 (2.9) 8.9 (2.8)1,2,3 62.4 (3,1161)
Age of onset 31.1 (11.4) 34.0 (11.5)a 30.2 (9.4)b 37.9 (13.0)a,b,c 21.8 (3,909)

% % % % c2 (df)d

Alcohol abuse 22.3 26.0 19.1 13.1 13.3 (3)


Substance abuse 19.3 7.1 11.3 4.2 35.0 (3)
Smoking 81.5 81.7 86.9 70.5 19.9 (3)

Only statistically significant results are shown (ANOVA and c2).


a
Statistically significant in comparison to Cluster 1
b
Statistically significant in comparison to Cluster 2
c
Statistically significant in comparison to Cluster 3
d
P < 0.001

early onset of gambling problems and the highest percentage labelled as Type II pathological gamblers and showed a
of substance abuse (together with a high percentage of alcohol schizoid personality profile. This subtype showed the highest
abuse) as partially suggested in a previous study.41 They also percentage of alcohol abuse and the lowest sensation seek-
showed high extravagance (overspending behaviour) and dis- ing. In this context, considering the high levels of Harm
orderliness (antisocial and uncontrollable behaviour). The Avoidance, the presence of psychopathological disturbances
presence of women was especially high in this cluster, but no (higher than in Clusters 3 and 4 although lower than in Clus-
statistically significant differences were achieved regarding ter 1), and the low level of substance abuse, we hypothesize
sex distribution across clusters. We did not assess personality that these patients may use alcohol (and gambling) to relieve
disorders but we dare to say that their prevalence will proba- emotional distress. The relation between Harm Avoidance
bly be high in this cluster, especially Cluster B (given their and depression is well known,42 as well as the association
high impulsiveness and low self-directedness). Potential between depression and use of alcohol as self-medication.43
treatments for this type of patients should include both These patients might benefit from psychological therapies
psychopharmacological and psychological interventions. focused on their personality deficits (especially their high
Psychological therapies for these patients should address not harm avoidance) and how to cope with negative affect.
only PG and comorbidity with substance abuse but also their
Patients of Cluster 3 showed high sensation seeking and rash
personality disturbances through more intensive and probably
impulsiveness, overspending behaviour, and uncontrolla-
long treatments.
bility. They also showed high sensitivity to reward (including
Cluster 2 patients were characterized by materialistic, con- social reward) and persistence. Their gambling behaviour
trolled, avoidant behaviour, and aloofness. They were was severe (according to the SOGS) in comparison to the

The Canadian Journal of Psychiatry, Vol 55, No 8, August 2010 W 503


Original Research

Table 4 Impulsivity correlates of the clusters: TCI-R Novelty Seeking factor subscales
Cluster 1 Cluster 2 Cluster 3 Cluster 4
n = 341 n = 238 n = 238 n = 354

NS factor subscale Mean (SD) Mean (SD) Mean (SD) Mean (SD) Fd

NS1–Exploratory Excitability 28.3 (5.3) 25.3 (5.3)a 33.1 (5.5)a,b 28.4 (5.6)b,c 83.4
a b
NS2–Impulsiveness 29.9 (5.5) 26.6 (5.8) 29.2 (5.5) 21.9 (5.0)a,b,c 147.4
NS3–Extravagance 35.5 (5.2) 33.7 (5.6)a 36.8 (4.9)a,b 30.3 (5.7)a,b,c 86.0
a b
NS4–Disorderliness 22.7 (4.8) 19.7 (4.8) 23.2 (4.7) 17.8 (4.7)a,b,c 88.4

ANOVA test
NS = Novelty Seeking
a
Statistically significant in comparison to Cluster 1
b
Statistically significant in comparison to Cluster 2
c
Statistically significant in comparison to Cluster 3
d
df = 3,1167; P < 0.001

other groups but they showed no general psychopathological gambling problems. Type I (disorganized and emotionally
disturbances (only a slightly higher hostility than Cluster 4 unstable) showed emotional and personality disturbances
patients). As well, they had early onset of gambling problems (other than impulsiveness), which are expected to be associ-
and the highest percentage of smoking. The literature suggests ated with an eventually poorer outcome. However, Type III
that extroversion and anxiety features are associated with (reward sensitive), even showing high impulsiveness and
tobacco smoking.44 Although this subgroup did not show sensation seeking, had no psychopathological or other per-
higher anxiety than the others, they clearly showed the most sonality disturbances, although their gambling problem was
extroverted personality profile. This subtype of PG was severe.
named Type III and showed a reward sensitivity profile. Ther-
apeutic interventions for these patients should be addressed to These findings may have several implications. It seems that
enhance their abilities of self-monitoring, impulse control, impulsiveness is not necessarily associated with sensation
and consideration of the long-term consequences of their seeking, as the scores in the latter trait were not especially
decisions. high in Type I (disorganized and emotionally unstable)
pathological gamblers. This adds support to the differentia-
Finally, Cluster 4 patients showed the healthiest personality tion between impulsiveness (understood as rash impulsive-
and clinical profile, that is, they showed low levels of impul- ness) and sensation seeking (sensitivity to reward and
siveness and sensation seeking, responsible, goal-directed novelty), as well as to the association of the former with
and cooperative behaviour, and (social) reward dependence. poorer outcomes.15,16 Conversely, both subgroups seemed to
They showed less severe gambling problems (as measured by show a response modulation deficit45 that affected gambling
the SOGS) and low levels of general psychopathology (as behaviour (both subgroups showed severe gambling behav-
measured by the SCL-90-R), as well as the lowest levels of iour) and was independent of the general psychopathological
comorbidity with substance and (or) alcohol abuse or tobacco state (only one subgroup showed psychopathological distur-
smoking. This subgroup showed the latest age of onset of bances). This deficit may also have a role in the presence of
gambling problems. These people were named as Type IV substance and (or) alcohol abuse and tobacco smoking in
pathological gamblers and were considered as a high- Type I (disorganized and emotionally unstable) and Type III
functioning subgroup. This means that PG may also develop (reward sensitive) pathological gamblers. Third, both sub-
in the context of an adaptive personality profile. This subtype groups showed the earliest age of onset of gambling prob-
of PG might benefit from shorter and less intensive interven- lems, suggesting that the response modulation deficit might
tions such as brief psychoeducational groups or short be associated with this clinical characteristic.
cognitive-behavioural treatments.
The classification into 3 subgroups of PG suggested by
Two subgroups showed high (rash) impulsiveness, that is, Blaszczynski and Nower23 was not supported by our data,
Type I (disorganized and emotionally unstable) and Type III given that 4 subtypes were clearly identified. Further, the dif-
(reward sensitive). Type III showed also high sensation seek- ferentiation between emotionally vulnerable and antisocial–
in g . Bo th g r o u p s s h o w e d d if f e r in g lev e ls o f impulsivist subgroups was not confirmed given that both fea-
psychopathological disturbances and comorbidity profiles; tures appeared together in one of the subtypes, suggesting
however, they similarly showed early age of onset of that both emotional and personality disorders may run

504 W La Revue canadienne de psychiatrie, vol 55, no 8, août 2010


Subtyping Study of a Pathological Gamblers Sample

3. Mason K, Arnold R. Problem gambling risk factors and associated behaviours


together in PG, instead of emerging separately. In addition, a and health status: results from the 2002/03 New Zealand Health Survey. N Z
subtype of PG that had not previously been described Med J. 2007;120:U2604.
emerged, that is, Type II (schizoid). However, we have to take 4. Goodman A. Neurobiology of addiction. An integrative review. Biochem
Pharmacol. 2008;75:266 –322.
into account that our sample was limited to slot-machine 5. Holden C. ‘Behavioral’ addictions: do they exist? Science. 2001;294:980 –982.
pathological gamblers who seek treatment for this disorder, 6. Kranzler HR, Wilcox M, Weiss RD, et al. The validity of cocaine dependence
subtypes. Addict Behav. 2008;33:41–53.
while Blaszczynski and Nower’s23 model accounts for patho- 7. Le Bon O, Basiaux P, Streel E, et al. Personality profile and drug of choice;
logical gamblers in general. These subtypes confirm that PG a multivariate analysis using Cloninger’s TCI on heroin addicts, alcoholics, and
a random population group. Drug Alcohol Depend. 2004;73:175–182.
is a highly heterogeneous disorder that may require specific 8. Álvarez-Moya EM, Jiménez-Murcia S, Granero R, et al. Comparison of
therapeutic interventions according to the person’s personality risk factors in bulimia nervosa and pathological gambling. Compr
Psychiatry. 2007;48:452– 457.
characteristics. 9. Janiri L, Martinotti G, Dario T, et al. The gamblers’ Temperament and
Character Inventory (TCI) personality profile. Subst Use Misuse.
Concerning limitations, our patients were treatment-seeking 2007;42:975–984.
pathological gamblers, and our results might not be 10. Saez-Abad C, Bertolin-Guillen JM. Personality traits and disorders in
pathological gamblers versus normal controls. J Addict Dis. 2008;27:33–40.
generalizable to nontreatment-seeking pathological gam- 11. Nordin C, Nylander PO. Temperament and character in pathological gambling.
blers. Second, most PG patients were mainly slot-machine J Gambl Stud. 2007;23:113–120.
gamblers, which also affects generalizability of results. Third, 12. Kerber CS, Black DW, Buckwalter K. Comorbid psychiatric disorders among
older adult recovering pathological gamblers. Issues Ment Health Nurs.
we did not assess comorbidity, other than with SUDs. We 2008;29:1018–1028.
tried to (partially) cover these areas by using psychometric 13. Bonnaire C, Lejoyeux M, Dardennes R. Sensation seeking in a French
population of pathological gamblers: comparison with regular and nongamblers.
measures of psychopathological status and personality. Psychol Rep. 2004;94:1361–1371.
14. Allcock CC, Grace DM. Pathological gamblers are neither impulsive nor
sensation-seekers. Aust N Z J Psychiatry. 1988;22:307–311.
Conclusions 15. Magid V, Maclean MG, Colder CR. Differentiating between sensation seeking
and impulsivity through their mediated relations with alcohol use and problems.
Blaszczynski and Nower’s23 pathways model was not sup- Addict Behav. 2007;32:2046–2061.
ported by our data. Cluster analytic techniques identified 4 16. Dawe S, Loxton NJ. The role of impulsivity in the development of substance
use and eating disorders. Neurosci Biobehav Rev. 2004;28:343–351.
empirical clusters with different clinical, personality, and 17. Meyer G. [Classification of gamblers from self-help groups using cluster
comorbidity profiles. Two types (Type I and Type III) showed analysis]. Z Klin Psychol Psychopathol Psychother. 1991;39:261–282.
a response modulation deficit, substance and (or) alcohol 18. González-Ibáñez A, Aymamí MN, Jiménez S, et al. Assessment of pathological
gamblers who use slot machines. Psychol Rep. 2003;93:707–716.
abuse, and early onset of gambling problems, but only one of 19. Ledgerwood DM, Petry NM. Psychological experience of gambling and
them (Type I) showed psychopathological and other personal- subtypes of pathological gamblers. Psychiatry Res. 2006;144:17–27.
20. Cunningham-Williams RM, Hong SI. A latent class analysis (LCA) of problem
ity disturbances. Another type (Type II) was characterized by gambling among a sample of community-recruited gamblers. J Nerv Ment Dis.
social aloofness and Harm Avoidance, as well as alcohol 2007;195:939–947.
21. Cunningham-Williams RM, Cottler LB, Compton W, et al. Computerized
abuse, while a fourth type (Type IV) showed high general Gambling Assessment Module (C-GAM). St Louis (MO): Washington
functioning. The different types of patients may benefit from University; 2003.
22. Toce-Gerstein M, Gerstein DR, Volberg RA. A hierarchy of gambling disorders
different therapeutic approaches. in the community. Addiction. 2003;98:1661–1672.
23. Blaszczynski A, Nower L. A pathways model of problem and pathological
Future research should include more heterogeneous samples gambling. Addiction. 2002;97:487–499.
in relation to the main gambling problem and consider treat- 24. Iancu I, Lowengrub K, Dembinsky Y, et al. Pathological gambling: an update
on neuropathophysiology and pharmacotherapy. CNS Drugs. 2008;22:123–138.
ment outcome parameters and biobehavioural validators (that 25. American Psychiatric Association. Diagnostic and statistical manual of mental
is, neuropsychological, biochemical) of the subgroups. disorders. 4th ed. Text revision. Washington (DC): APA; 2000.
26. Lesieur HR, Blume SB. The South Oaks Gambling Screen (SOGS): a new
instrument for the identification of pathological gamblers. Am J Psychiatry.
Acknowledgements 1987;144:1184–1188.
27. Echeburúa E, Báez C, Fernández J, et al. Cuestionario de juego patológico de
Dr Álvarez-Moya was supported by a Juan de la Cierva fellowship South Oaks (SOGS): Validación española. [South Oaks Gambling Screen
(Ministry of Science and Technology, Spanish government; (SOGS): Spanish validation]. Anális Modific Cond. 1994;20:769–791.
Reference 1048). This work is part of the PhD dissertation of 28. Stinchfield R. Reliability, validity, and classification accuracy of a measure of
Neus Aymamí at the Universitat Autonoma Barcelona, Spain. DSM-IV diagnostic criteria for pathological gambling. Am J Psychiatry.
2003;160:180–182.
Financial support was received from Fondo de Investigación 29. Jiménez-Murcia S, Stinchfield R, Álvarez-Moya E, et al. Reliability, validity,
Sanitaria-FIS (PI081573; PI081714) and AGAUR and classification accuracy of a Spanish translation of a measure of DSM-IV
(2009SGR1554). It was also partially supported by the European diagnostic criteria for pathological gambling. J Gambl Stud. 2009;25:93–104.
Commission under the Seventh Framework Programme 30. Derogatis LR. SCL-90: administration, scoring and procedures manual-I for the
revised version. Baltimore (MD): John Hopkins University Press; 1977.
(FP7-ICT-215839-2007- Playmancer project). CIBER 31. González de Rivera JL. Version española del 90 Symptoms Check List Revised
Fisiopatología de la Obesidad y Nutricioìn (CIBERobn) and (SCL-90-R). [Spanish version of the 90-Symptoms Check List Revised
CIBER Salud Mental (CIBERSAM), are initiatives of ISCIII. (SCL-90-R)]. Madrid (ES): TEA ediciones; 2001.
32. First MB, Spitzer RL, Gibbon M, et al. Structured Clinical Interview for
DSM-IV Axis I disorders-Clinical Version (SCID-CV). Washington (DC):
References American Psychiatric Press; 1997.
33. Kranzler HR, Ronald MN, Burleson JA, et al. Validity of psychiatric diagnoses
in patients with substance use disorders: is the interview more important than
1. American Psychiatric Association. Diagnostic and statistical manual of mental the interviewer? Compr Psychiatry. 1995;36:278–288.
disorders. 3rd ed. Washington (DC): APA; 1980. 34. Martin CS, Pollock NK, Bukstein OG, et al. Inter-rater reliability of the SCID
2. Petry NM, Stinson FS, Grant BF. Comorbidity of DSM-IV pathological alcohol and substance use disorders section among adolescents. Drug Alcohol
gambling and other psychiatric disorders: results from the National Depend. 2000;59:173–176.
Epidemiologic Survey on Alcohol and Related Conditions. J Clin Psychiatry. 35. Cloninger CR. The Temperament and Character Inventory–Revised. St Louis
2005;66:564 –574. (MO): Center for Psychobiology of Personality, Washington University; 1999.

The Canadian Journal of Psychiatry, Vol 55, No 8, August 2010 W 505


Original Research

2
36. Gutiérrez-Zotes J, Bayón C, Montserrat C, et al. Inventario del Temperamento y Head of Pathological Gambling Unit, Department of Psychiatry,
el Carácter-Revisado (TCI-R). Baremación y datos normativos en una muestra de University Hospital of Bellvitge-IDIBELL, Barcelona, Spain; Member
población general. Actas Español Psiquiatría. 2004;32:8–15. of CIBER Fisiopatología Obesidad y Nutrición (CIBERobn), Instituto
37. Eysenck SBG, Pearson PR, Eating G, et al. Age norms for impulsiveness, Salud Carlos III Barcelona, Spain; Associated Professor, Department of
venturesomeness and empathy in adults. Pers Individ Dif. 1985;6:613–619. Clinical Sciences, School of Medicine, University of Barcelona,
38. Garson GD. Cluster analysis: statnotes from North Carolina State University, Barcelona, Spain.
public administration program web site [Internet]. Raleigh (NC): North Carolina 3
Clinical Psychologist, Department of Psychiatry, University Hospital of
State University; 1998 [cited 2008 Mar 24]. Available from: Bellvitge, Barcelona, Spain; Associated Professor, Department of
http://www2.chass.ncsu.edu/garson/PA765/cluster.htm#2step.
Psychiatry and Clinical Psychobiology, Psychology School, University
39. Fraley C, Raftery AE. How many clusters? Which clustering method? Answers of Barcelona, Barcelona, Spain. Departament de Psicologia Clínica i de
via model-based cluster analysis. Computer J. 1998;4:578–588.
la Salut, Universitat Autònoma Barcelona, Spain.
40. Cloninger CR, Svrakic DM, Przybeck TR. A psychobiological model of 4
Clinical Psychologist, Department of Psychiatry, University Hospital of
temperament and character. Arch Gen Psychiatry. 1993;50:975–990.
Bellvitge, Barcelona, Spain.
41. Jiménez-Murcia S, Álvarez-Moya EM, Stinchfield R, et al. Age of onset in 5
Full Professor in Statistics, Applied-Statistics Laboratory, Department of
pathological gambling: clinical, therapeutic and personality correlates. J Gambl
Psychobiology and Methodology in Health Sciences, Universitat
Stud. 2010;26(2):235–248.
Autònoma Barcelona, Barcelona, Spain.
42. Huezo-Diaz P, Tandon K, Aitchison KJ. The genetics of depression and related 6
Research Psychologist, Department of Psychiatry, University Hospital of
traits. Curr Psychiatry Rep. 2005;7:117–124.
Bellvitge, Barcelona, Spain.
43. Bolton JM, Robinson J, Sareen J. Self-medication of mood disorders with 7
Head of the Department, Department of Psychiatry University Hospital
alcohol and drugs in the National Epidemiologic Survey on Alcohol and Related
Conditions. J Affect Dis. 2009;115:367–375. of Bellvitge-IDIBELL Barcelona, Spain; Head of Group, CIBER Salud
Mental (CIBERSAM), Instituto Salud Carlos III, Barcelona, Spain; Full
44. Munafò MR, Zetteler JI, Clark TG. Personality and smoking status:
a meta-analysis. Nicotine Tob Res. 2007;9:405–413. Professor, Department of Clinical Sciences, School of Medicine,
University of Barcelona, Barcelona, Spain.
45. Patterson CM, Newman JP. Reflectivity and learning from aversive events: 8
toward a psychological mechanism for the syndromes of disinhibition. Psychol Head of Eating Disorders Unit, Department of Psychiatry, University
Rev. 1993;100:716–736. Hospital of Bellvitge-IDIBELL, Barcelona, Spain; Head of Group,
CIBER Fisiopatología Obesidad y Nutrición (CIBERobn), Instituto
Salud Carlos III, Barcelona, Spain; Full Professor, Department of
Clinical Sciences, School of Medicine, University of Barcelona,
Manuscript received April 2009, revised, and accepted February 2010. Barcelona, Spain.
1
Research Psychologist, Department of Psychiatry, University Hospital of Address for correspondence: Dr S Jiménez-Murcia, Pathological
Bellvitge-IDIBELL, Barcelona, Spain; Research Psychologist, CIBER Gambling Unit, Department of Psychiatry, University Hospital of
Fisiopatología Obesidad y Nutrición (CIBERobn), Instituto Salud Carlos Bellvitge, C/Feixa Llarga, s/n, Hospitalet de Llobregat, 08907, Barcelona,
III Barcelona, Spain. Spain; sjimenez@bellvitgehospital.cat

Résumé : Étude de sous-typage d’un échantillon de joueurs pathologiques


Objectif : Classer en sous-groupes un échantillon de patients souffrant de jeu pathologique (JP)
selon les variables de personnalité et décrire les sous-groupes à un niveau clinique.
Méthode : Les patients du JP (n = 1171) ont été évalués à l’aide de l’échelle de jeu de South Oaks;
de l’inventaire des tempéraments et caractères révisé (TCI-R); de la liste des symptômes – 90 —
révisée (SCL-90-R); des échelles d’impulsivité d’Eysenck, d’un questionnaire diagnostique pour les
critères du JP selon le Manuel diagnostique et statistique des troubles mentaux, 4e édition,
(DSM-IV); et de l’entrevue clinique structurée (SCID) pour le module de l’utilisation de substances et
ceux des troubles de l’axe I du DSM-IV. Les mesures cliniques ont été recueillies par une entrevue
semi-structurée. Nous avons effectué une analyse typologique en 2 étapes d’après les variables de
personnalité mentionnées ci-dessus. Les données cliniques ont été comparées entre les types.
Résultats : Quatre types sont ressortis. Le type I (désorganisé et émotionnellement instable)
présentait des traits schizotypiques, une impulsivité élevée, l’abus d’alcool et de substances, et un
âge précoce d’apparition, ainsi que des perturbations psychopathologiques. Le type II (schizoïde)
présentait un évitement du danger, une réserve sociale, et l’abus d’alcool. Le type III (sensible aux
récompenses) montrait une recherche de sensations et une impulsivité élevées, mais pas de
perturbations psychopathologiques. Le type IV (très fonctionnel) présentait un profil de personnalité
généralement adaptatif, un faible niveau d’abus de substances et d’alcool ou de tabagisme, et pas
de perturbations psychopathologiques.
Conclusions : Au moins 4 types de patients du JP ont été identifiés. Deux types présentaient un
déficit de modulation des réactions, mais seulement un d’entre eux avait de graves perturbations
psychopathologiques. Deux autres types ne présentaient pas d’impulsivité ou de recherche de
sensations, et l’un d’eux affichait un bon fonctionnement général. La différente configuration clinique
et des personnalités de ces types peut être liée à différentes approches thérapeutiques.

506 W La Revue canadienne de psychiatrie, vol 55, no 8, août 2010