ORIGINAL RESEARCH published: 29 March 2019 doi: 10.3389/fpsyt.2019.00173 Phenotypes in Gambling Disorder Using Sociodemographic and Clinical Clustering Analysis: An Unidentified New Subtype? Susana Jiménez-Murcia 1,2,3*, Roser Granero 2,4 , Fernando Fernández-Aranda 1,2,3 , Edited by: Randy Stinchfield 5 , Joel Tremblay 6 , Trevor Steward 1,2 , Gemma Mestre-Bach 1,2 , Yasser Khazaal, María Lozano-Madrid 1,2 , Teresa Mena-Moreno 1,2 , Núria Mallorquí-Bagué 1,2 , Centre Hospitalier Universitaire José C. Perales 7,8 , Juan F. Navas 7,8 , Carles Soriano-Mas 1,4,9 , Neus Aymamí 1 , Vaudois, Switzerland Mónica Gómez-Peña 1 , Zaida Agüera 1,2 , Amparo del Pino-Gutiérrez 10 , Reviewed by: Virginia Martín-Romera 11 and José M. Menchón 1,3,9 Mercedes Lovrecic, 1 Department of Psychiatry, University Hospital of Bellvitge-IDIBELL, Barcelona, Spain, 2 Ciber Fisiopatologia Obesidad y National Institute for Public Health, Nutrición, Instituto Salud Carlos III, Barcelona, Spain, 3 Department of Clinical Sciences, School of Medicine, University of Slovenia Barcelona, Barcelona, Spain, 4 Departament de Psicobiologia i Metodologia de les Ciències de la Salut, Universitat Xiaohui Xu, Autònoma de Barcelona, Barcelona, Spain, 5 Department of Psychiatry, University of Minnesota Medical School, Minneapolis, Texas A&M Health Science Center, MN, United States, 6 Département de Psychoéducation, Université du Québec à Trois-Rivières, Trois-Rivières, QC, Canada, United States 7 Department of Experimental Psychology, University of Granada, Granada, Spain, 8 Brain, Mind and Behavior Research *Correspondence: Center (CIMCYC), University of Granada, Granada, Spain, 9 Ciber de Salud Mental, Instituto Salud Carlos III, Madrid, Spain, 10 Susana Jiménez-Murcia Departament d’Infermeria de Salut Pública, Salut Mental i Maternoinfantil, Escola Universitària d’Infermeria, Universitat de sjimenez@bellvitgehospital.cat Barcelona, Barcelona, Spain, 11 Department of Psychology, University of Alcalá, Alcalá de Henares, Spain Specialty section: This article was submitted to Background: Gambling disorder (GD) is a heterogeneous disorder which has clinical Addictive Disorders, manifestations that vary according to variables in each individual. Considering the a section of the journal importance of the application of specific therapeutic interventions, it is essential to Frontiers in Psychiatry obtain clinical classifications based on differentiated phenotypes for patients diagnosed Received: 03 September 2018 Accepted: 08 March 2019 with GD. Published: 29 March 2019 Objectives: To identify gambling profiles in a large clinical sample of n = 2,570 patients Citation: seeking treatment for GD. Jiménez-Murcia S, Granero R, Fernández-Aranda F, Stinchfield R, Methods: An agglomerative hierarchical clustering method defining a combination of Tremblay J, Steward T, Mestre-Bach G, Lozano-Madrid M, the Schwarz Bayesian Information Criterion and log-likelihood was used, considering a Mena-Moreno T, Mallorquí-Bagué N, large set of variables including sociodemographic, gambling, psychopathological, and Perales JC, Navas JF, Soriano-Mas C, personality measures as indicators. Aymamí N, Gómez-Peña M, Agüera Z, del Pino-Gutiérrez A, Martín-Romera V Results: Three-mutually-exclusive groups were obtained. Cluster 1 (n = 908 and Menchón JM (2019) Phenotypes in Gambling Disorder Using participants, 35.5%), labeled as “high emotional distress,” included the oldest patients Sociodemographic and Clinical with the longest illness duration, the highest GD severity, and the most severe levels of Clustering Analysis: An Unidentified psychopathology. Cluster 2 (n = 1,555, 60.5%), labeled as “mild emotional distress,” New Subtype? Front. Psychiatry 10:173. included patients with the lowest levels of GD severity and the lowest levels of doi: 10.3389/fpsyt.2019.00173 psychopathology. Cluster 3 (n = 107, 4.2%), labeled as “moderate emotional distress,” Frontiers in Psychiatry | www.frontiersin.org 1 March 2019 | Volume 10 | Article 173 Jiménez-Murcia et al. Clustering for GD included the youngest patients with the shortest illness duration, the highest level of education and moderate levels of psychopathology. Conclusion: In this study, the general psychopathological state obtained the highest importance for clustering. Keywords: gambling disorder, clustering, personality traits, psychopathology, severity INTRODUCTION to be an heterogeneous group, their profile is often the following: male, young, single, high levels of education, good Gambling disorder (GD) is defined as a persistent and economic income, varied gambling preferences, greater severity recurrent maladaptive pattern of gambling behavior associated of gambling behavior (in terms of spending higher amounts), with impaired functioning in the personal, social, and as well as higher rates of alcohol and substance use (17–19). occupational spheres of one’s life (1). Across the globe, GD Likewise, it can be expected that specific gambling activities affects 0.7–6.5% of adults during their lifetime (2). Evidence have a higher representation in certain subtypes of gamblers, points to GD as being more likely to occur among males, singles for instance, online gambling may be more frequent in those or divorcees, unemployed individuals or with a low income, and subtypes composed of younger gamblers with higher education, with a lower level of education (3). Moreover, GD presents high with better incomes and the possibility of betting larger amounts comorbidity with other mental disorders, mainly substance or of money. alcohol dependence (4) and personality disorders (5). Beyond these results, the question is whether it is possible to From an etiological point of view, GD is a multi- formulate the hypothesis that emotional alterations and the type causal disorder involving psychosocial and neurobiological risk of preferred gambling could explain the existence of more specific factors. The interaction between these factors and the specific subtypes. To the best of our knowledge, no studies have explored implication of each can aid in discerning the phenotypes the last question in large clinical samples made up of consecutive of GD patients. One of the psychosocial risk factors of patients who have requested specific treatment for gambling greatest interest has been psychological or emotional distress, problems. For that reason, the objective of this study is to explore which is related to high levels of general psychopathology as the existence of distinct GD phenotypes, in a large sample of measured by the Symptom Checklist-Revised (SCL-90-R) (6, patients seeking treatment for GD, through clustering analysis, 7). In this sense, Blaszczynski and Nower (8) pointed out the using a wide set of variables including sociodemographic, existence of a subtype of individuals with GD characterized gambling-related factors, general psychopathological state and by high emotional vulnerability. These authors hypothesized personality traits as indicators for the grouping. Based on the that emotional vulnerability is a result of different psychosocial available evidence in literature, we hypothesized that different factors, such as childhood disturbances, certain personality traits, empirical clusters would emerge based on emotional distress. mood alterations (depression and anxiety), poor coping skills, Due to the dramatic change in relation to the proliferation of and substance use. In this context, gambling is viewed as a different types of gambling in this last decade (specifically, the means of relieving aversive affective states. Posterior evidence has popularization of online gambling) and the change in profiles of concluded that these “emotionally vulnerable” gamblers exhibit people affected by GD, our secondary hypothesis was that one higher psychopathological and gambling severity than other of these clusters would overlap with the emerging profile of the subtypes of gamblers, as well as a higher prevalence of substance online gambler. abuse and an earlier onset of gambling problems (9–12). On the one hand, problematic or pathological gambling behavior has frequently been associated with certain types of MATERIALS AND METHODS gambling activities. That is, not all gambling activities seem to be addictive to the same degree. Until recently, slot machines Participants were considered the most potentially addictive type of gambling The study sample included n = 2,570 patients treated at the because of their technical and situational characteristics (13). Gambling Disorder Unit at the Department of Psychiatry of However, the emergence and proliferation of online gambling Bellvitge University Hospital, in Barcelona, Spain. This public in the last decade has radically modified the characteristics hospital is certified as a tertiary care center for the treatment of of gambling activities, by which gambling settings have been psychologically addictive behaviors and it oversees the treatment shifted to home and work environments (14). Online gambling of complex cases. The catchment area of the hospital includes is considered potentially more addictive because of its unique over two million people in the metropolitan area of Barcelona. features, such as anonymity, betting speed, accessibility, attractive All patients that met diagnostic criteria for GD, between the design, and marketing, and low cost (15, 16). Indeed, recent ages of 18 and 75, were selected for this study. Patients were evidence has displayed that online gambling increases the recruited between January 2005 and November 2015. Exclusion likelihood of being a problematic gambler when compared criteria included having an intellectual disability or severe with offline gambling (16). Although online gamblers seem mental disorders (schizophrenia or other psychotic disorders, Frontiers in Psychiatry | www.frontiersin.org 2 March 2019 | Volume 10 | Article 173 Jiménez-Murcia et al. Clustering for GD bipolar disorder, etc.), or meeting criteria for another behavioral Symptom Checklist-Revised (SCL-90-R) (26) addiction apart from GD. This questionnaire is a measure of emotional distress and Most participants were men (n = 2,365, 92.0%), born in Spain evaluates a broad range of psychological problems and (n = 2,405, 93.6%), were married or had a partner (n = 1,324, psychopathological symptoms. This questionnaire contains 51.5%) and were employed (n = 1,525, 59.3%). The mean age for 90 items and assesses nine primary symptom dimensions: the total sample was 41.7 years old (SD = 12.8), the mean age of somatization, obsession-compulsion, interpersonal sensitivity, GD onset was 36.8 years old (SD = 12.9) and the mean duration depression, anxiety, hostility, phobic anxiety, paranoid ideation, of GD was 14.0 years (SD = 10.0). The first columns of Table 1 and psychoticism. It also includes three global indices: (1) a global include a description of the study sample. severity index (GSI), designed to measure overall psychological distress; (2) a positive symptom distress index (PSDI), to measure Instruments the intensity of the symptoms; and (3) a positive symptom GD Diagnosis and Severity total (PST), which reflects self-reported symptoms. A validated Spanish version was used (27). The Spanish validation scale South Oaks Gambling Screen (SOGS) (20) obtained good psychometrical indexes, with a mean internal This is a self-report 20-item screening questionnaire that consistency of 0.75 (Cronbach’s alpha). In the study sample, discriminates between probable pathological, problem and non- consistency indices were in the very good (α = 0.83 for hostility) problem gamblers. The Spanish validation used in this work to excellent range (α = 0.98 for the global indexes), with showed excellent internal consistency (α = 0.94) and test-retest exception of paranoia (α = 0.74, good). reliability (r = 0.98) (21). Other Sociodemographic and Clinical Variables Diagnostic Questionnaire for Pathological Gambling Additional demographic, clinical, and social/family variables According to DSM-IV Criteria (22). Spanish Adaptation by related to gambling were measured using a semi-structured, face- Jiménez-Murcia et al. (23). to-face clinical interview described elsewhere (28). Some of the This 19-item questionnaire assesses the DSM-IV diagnostic GD variables covered included the age of GD onset, the mean and criteria for pathological gambling. It should be noted that with maximum amount of monetary spending in gambling episodes, the release of the DSM-5 (1), the term pathological gambling and accumulated gambling-related debts. Moreover, GD severity was replaced with GD. Internal consistency of this questionnaire was assessed according to five different variables: total number of ranged between α = 0.81 for the general population and α = DSM-5 criteria, SOGS-total score, number of gambling activities, 0.77 for gambling treatment samples. Convergent validity with mean bets per gambling episode, and cumulative debts. the SOGS scores was high: r = 0.77 for the general population Table 1 contains the α-values obtained in the sample for and r = 0.75 for gambling treatment groups (22). Due to the fact all the scales in this work. These coefficients ranged between that this questionnaire includes the 8th criterion exploring the good to excellent. presence of illegal acts related to GD, in this study, all patient diagnoses were reassessed and recodified post hoc to avoid the Procedure confounding effect of increased GD severity, in patients with The present study was carried out in accordance with the latest a criminal history. Therefore, only patients who met DSM-5 version of the Declaration of Helsinki. The University Hospital criteria for GD were included in our analysis. of Bellvitge Ethics Committee of Clinical Research approved the study (ref. 307/06), and signed informed consent was obtained Personality and Psychopathological Status from all participants. Experienced psychologists and psychiatrists Temperament and Character Inventory-Revised with more than 15 years of clinical experience in the field (TCI-R) (24) of addictive disorders conducted the two face-to-face clinical This is a reliable and valid 240-item questionnaire that interviews. All the measures used in this study correspond to the measures seven personality dimensions: four temperament assessment carried out prior to the beginning of treatment. (novelty seeking, harm avoidance, reward dependence and persistence) and three character dimensions (self-directedness, Statistical Analysis cooperativeness and self-transcendence). Temperament refers Statistical analysis was carried out with SPSS 20 for Windows. to automatic emotional responses to experiences that are Empirical clusters were explored through the TwoStep- moderately heritable and stable throughout life. In contrast, Clustering Component, entering origin (Spanish vs. immigrant), character refers to self-concepts and individual differences in sex, education level, civil status, employment status, age, age goals and values. Character is moderately influenced by insight of GD onset, duration of the gambling problem, GD severity and learning. All items are measured on a 5-point Likert- and gambling related variables (SOGS-total score, number type scale. A validated Spanish version was used (25). The of gambling activities, mean bets per gambling episode and scales in the Spanish revised version showed adequate internal cumulative debts), comorbid psychopathological state (SCL-90R consistency (Cronbach’s α mean value of 0.87). In the study scales), substances use-abuse (tobacco, alcohol and other drugs) at hand, consistency indices ranged from good (α = 0.70 for and personality traits (TCI-R scales) as indicators. The TwoStep- novelty seeking) to very good (α = 0.84 for persistence and Clustering system constitutes a scalable algorithm designed to self-transcendence). handle large datasets including both continuous and categorical Frontiers in Psychiatry | www.frontiersin.org 3 March 2019 | Volume 10 | Article 173 Jiménez-Murcia et al. Clustering for GD TABLE 1 | Cluster composition and comparison between groups. Categorical variables Total Cluster 1 Cluster 2 Cluster 3 Cluster Pairwise comparisons: p n =2,570 n = 908; 35.3% n = 1,555; 60.5% n = 107; 4.2% n % n % n % n % p-value C1-C2 C1-C3 C2-C3 Origin Spanish 2,405 93.6% 836 92.1% 1483 95.4% 86 80.4% <0.001 0.001 <0.001 <0.001 Immigrant 165 6.4% 72 7.9% 72 4.6% 21 19.6% Sex Men 2,365 92.0% 782 86.1% 1495 96.1% 88 82.2% <0.001 <0.001 0.278 <0.001 Women 205 8.0% 126 13.9% 60 3.9% 19 17.8% Education Primary 1443 56.1% 553 60.9% 873 56.1% 17 15.9% <0.001 0.039 <0.001 <0.001 Secondary 973 37.9% 319 35.1% 597 38.4% 57 53.3% University 154 6.0% 36 4.0% 85 5.5% 33 30.8% Civil status Single 883 34.4% 355 39.1% 486 31.3% 42 39.3% <0.001 <0.001 0.522 <0.001 Married 1,324 51.5% 395 43.5% 887 57.0% 42 39.3% Divorced-widow 363 14.1% 158 17.4% 182 11.7% 23 21.5% Employed Yes 1525 59.3% 445 49.0% 993 63.9% 87 81.3% <0.001 <0.001 <0.001 <0.001 Smoking use-abuse Yes 1,668 64.9% 605 66.6% 1025 65.9% 38 35.5% <0.001 0.718 <0.001 <0.001 Alcohol use-abuse Yes 385 15.0% 155 17.1% 228 14.7% 2 1.9% <0.001 0.112 <0.001 <0.001 Drugs use-abuse Yes 252 9.8% 125 13.8% 121 7.8% 6 5.6% <0.001 <0.001 0.017 0.413 Quantitative variables α Mean SD Mean SD Mean SD Mean SD Age (years-old) 41.73 12.81 42.43 12.45 41.49 13.08 39.30 11.46 0.029 0.079 0.017 0.087 Onset (years-old) 36.76 12.88 36.65 12.90 36.92 12.98 35.36 11.25 0.457 0.611 0.330 0.227 Duration (years) 13.99 9.97 14.86 9.79 13.62 10.03 11.96 10.14 0.001 0.003 0.004 0.095 GD SEVERITY MEASURES Mean bets-episode 129.4 629.7 105.7 196.1 88.8 143.1 920.1 2884.5 <0.001 0.507 <0.001 <0.001 Cumulate debts 10,858 77219 8551 23509 6431 17847 94759 357317 <0.001 0.500 <0.001 <0.001 # gambling activities 1.02 0.22 1.01 0.09 1.00 0.05 1.42 0.92 <0.001 0.454 <0.001 <0.001 SOGS-total score 0.757 10.82 2.96 12.09 2.64 10.05 2.86 11.17 3.15 <0.001 <0.001 0.001 <0.001 DSM-5 total criteria 0.807 6.78 1.99 7.70 1.46 6.24 2.04 6.83 2.28 <0.001 <0.001 <0.001 0.002 SCL-90R SCALES Somatization 0.909 0.94 0.83 1.69 0.81 0.51 0.46 0.83 0.63 <0.001 <0.001 <0.001 <0.001 Obsessive-com. 0.881 1.12 0.83 1.92 0.72 0.65 0.47 1.18 0.63 <0.001 <0.001 <0.001 <0.001 Interp. sensitivity 0.872 1.02 0.84 1.87 0.73 0.54 0.44 0.98 0.53 <0.001 <0.001 <0.001 <0.001 Depressive 0.910 1.48 0.92 2.39 0.67 0.94 0.57 1.57 0.71 <0.001 <0.001 <0.001 <0.001 Anxiety 0.896 1.01 0.83 1.84 0.74 0.52 0.39 0.97 0.55 <0.001 <0.001 <0.001 <0.001 Hostility 0.852 0.91 0.84 1.62 0.89 0.49 0.46 0.89 0.71 <0.001 <0.001 <0.001 <0.001 Phobic anxiety 0.821 0.48 0.69 1.04 0.84 0.17 0.26 0.32 0.49 <0.001 <0.001 <0.001 0.006 Paranoia 0.782 0.90 0.79 1.60 0.77 0.50 0.45 0.86 0.59 <0.001 <0.001 <0.001 <0.001 Psychotic 0.858 0.89 0.77 1.63 0.71 0.46 0.38 0.83 0.56 <0.001 <0.001 <0.001 <0.001 GSI index 0.981 1.04 0.72 1.82 0.55 0.59 0.32 1.03 0.46 <0.001 <0.001 <0.001 <0.001 PST index 0.981 45.90 21.69 67.72 11.84 33.00 15.32 48.35 15.30 <0.001 <0.001 <0.001 <0.001 PSDI index 0.981 1.88 0.61 2.41 0.52 1.58 0.43 1.87 0.45 <0.001 <0.001 <0.001 <0.001 TCI-R SCALES Novelty seeking 0.702 108.89 14.40 111.80 13.95 106.94 14.31 112.58 14.66 <0.001 <0.001 0.592 <0.001 Harm avoidance 0.806 101.09 17.08 110.25 15.96 95.70 15.34 101.78 17.22 <0.001 <0.001 <0.001 <0.001 Reward dependence 0.763 98.54 14.85 95.48 15.25 100.33 14.38 98.46 13.92 <0.001 <0.001 0.047 0.202 Persistence 0.862 108.46 20.03 107.05 20.41 109.44 19.70 106.21 20.69 0.008 0.004 0.679 0.105 Self-directedness 0.846 126.93 21.16 112.13 16.73 135.68 18.60 125.37 19.64 <0.001 <0.001 <0.001 <0.001 Cooperativeness 0.802 130.43 16.43 123.28 16.78 134.52 14.78 131.74 15.75 <0.001 <0.001 <0.001 0.074 Self-transcendence 0.830 64.15 15.38 69.04 15.60 61.46 14.55 61.60 15.17 <0.001 <0.001 <0.001 0.927 SD, standard deviation; α, Chronbach’s alpha in the sample; Bold, significant comparison (0.05 level). p-value includes Finner’s correction for multiple statistical comparison. Frontiers in Psychiatry | www.frontiersin.org 4 March 2019 | Volume 10 | Article 173 Jiménez-Murcia et al. Clustering for GD variables. During the first step, subjects are pre-clustered into avoidance dimensions, and the GD severity measures (number of many small sub-clusters according to a sequential clustering gambling activities, and SOGS-total score). The poorest relevance approach and during the second step the resulting sub-clusters for clustering was obtained for the sample origin (Spanish vs. are considered as inputs and grouped into the desired number immigrant), civil status, substances use-abuse, chronological age, of clusters according to the agglomerative hierarchical clustering illness duration, TCI-R persistence scores, and age of GD onset. method. By default, the TwoStep algorithm uses a combination of The table included in the right panel of Figure 1 contains the Schwarz Bayesian Information Criterion and log-likelihood the centroids for the indicators in the clustering (means for distance in determining the final number of clusters, choosing the quantitative variables and the percentage distribution for a solution with a reasonably large ratio of Schwarz Bayesian categorical variables), which summarizes the clusters patterns for Information Criterion changes and a large ratio of distance this set of variables. measures. The log-likelihood measure is computed by using the Cluster 1 contained 908 patients (35.5% of the sample) and normal density for continuous variables and the multinomial it was characterized by featuring the oldest patients with the probability mass function for categorical variables. In this study, longest duration of GD, the highest GD severity, the most severe we compared the automatic number of clusters selected by the levels of general psychopathology or emotional distress (highest TwoStep Clustering procedure as finalistic candidate solutions, mean scores on the SCL-90R questionnaire), the highest levels in and two additional models: the auto-determined number of the personality traits harm avoidance and self-transcendence and clusters minus one, and the auto-determined number plus the lowest levels in the personality domains of self-directedness one. The final chosen model was based on (29) criteria: (a) the and cooperativeness. This cluster has been labeled as “high highest cohesion and separation index; (b) adequate number of emotional distress.” individuals in each group (to allow for meaningful comparisons); Cluster 2, the largest empirical group with 1,555 patients and (c) the best clinical interpretability. (60.5% of the sample), was characterized by including the The comparison between the derived empirical clusters for highest proportion of men, the lowest GD severity, the the study variables (sociodemographic, clinical, and personality best psychopathological state (lowest means on the SCL-90R measures) was carried out with chi-square tests (χ2 ) for scales), the lowest levels in the personality traits novelty categorical factors and analysis of variance (ANOVA) for seeking and harm avoidance, and the highest mean scores quantitative measures. Finner’s correction (a familywise error in the personality scales reward dependence, persistence, self- rate stepwise procedure which offers more powerful test than the directedness, and cooperativeness. This cluster has been labeled classical Bonferroni correction) controlled for inflation in Type-I as “mild emotional distress.” error due to multiple statistical comparisons (30). Cluster 3, with only 107 patients (4.2%), included the youngest participants with the lowest GD duration, the highest levels of education, the lowest prevalence of substance use-abuse, the most RESULTS severe economical consequences related to the GD (mean bets per episode and cumulate debts due to gambling), and moderate Cluster Composition and Comparison levels in psychopathological state. This cluster has been labeled as Between Groups “moderate emotional distress.” The three-cluster solution was selected as being the most optimal in the sample of n = 2,570 GD patients who Distribution of the Empirical Clusters met inclusion criteria. This structure was auto-determined as During the Period of Data Recruitment the most appropriate by the TwoStep-Clustering procedure. The line-chart in the Figure 2 shows the prevalence of each The Silhouette’s index [a validity measure of consistency cluster in the study sample during the data recruitment period. within clusters, which can be interpreted as the level of As a whole, the most prevalent cluster during the time of data cohesion/separation for the empirical derived groups; (31)] recruitment was cluster 2 (“mild emotional distress”), followed by was moderate (average Silhouette = 0.3, suggesting reasonable cluster 1 (“high emotional distress”). A positive significant linear evidence of cluster structure in the data). The comparison trend for clusters 1 (its presence increased from 29.4% in 2005 to between the largest cluster size (n = 1,555, 60.5%) and the 40.3% in 2015) and 3 (increases ranged from 1.7% in 2005 to 7.3% smallest (n = 107, 4.2%) yielded a ratio of 14.53. in 2015) was found, while cluster 2 obtained a negative significant Figure 1 summarizes the composition of the final solution linear trend (prevalence decreased from 68.9% in 2005 to for the three-cluster model. The first panel contains the bar- 52.4% in 2015). graph with the relative importance of each indicator in the clustering, which reports how well each variable can differentiate Comparisons Between Clusters the different derived clusters (the higher the importance of The last columns of Table 1 contain the results of the comparison the measure, the less likely it is that the variation for the between the groups for the variables analyzed in the study and variable between clusters is due to chance and the more defined as indicators in the clustering procedure. likely it is due to underlying differences). In this study, the Table 2 includes the gambling preferences and the general psychopathological state registered on the SCL-90- comparisons between clusters. The preferred gambling activity R questionnaire obtained the highest relative importance for was slot machines for clusters 1 and 2, while cluster 3 included clustering, followed by the TCI-R self-directedness and harm a higher proportion of participants reporting online gambling, Frontiers in Psychiatry | www.frontiersin.org 5 March 2019 | Volume 10 | Article 173 Jiménez-Murcia et al. Clustering for GD FIGURE 1 | Clustering summary: relative importance of each indicator and centroids. FIGURE 2 | Distribution of the prevalence of each cluster during the data recruitment. LT, linear trend. Frontiers in Psychiatry | www.frontiersin.org 6 March 2019 | Volume 10 | Article 173 Jiménez-Murcia et al. Clustering for GD TABLE 2 | Comparison between clusters based on main gambling preference. Total; n = 2,570 C1; n = 908 C2; n = 1,555 C3; n = 107 Cluster Pairwise comparison: p n % n % n % n % p C1-C2 C1-C3 C2-C3 PRIMARY ACTIVITY Slot machines 2036 79.2% 739 81.4% 1257 80.8% 40 37.4% <0.001 0.736 <0.001 <0.001 Bingo 274 10.7% 139 15.3% 113 7.3% 22 20.6% <0.001 <0.001 0.160 <0.001 Lotteries 236 9.2% 95 10.5% 126 8.1% 15 14.0% 0.031 0.048 0.263 0.034 Casinos 189 7.4% 78 8.6% 89 5.7% 22 20.6% <0.001 0.006 <0.001 <0.001 Cards 92 3.6% 44 4.8% 38 2.4% 10 9.3% <0.001 0.001 0.050 <0.001 Offline betting 103 4.0% 36 4.0% 51 3.3% 16 15.0% <0.001 0.374 <0.001 <0.001 Online gambling 159 6.2% 44 4.8% 85 5.5% 30 28.0% <0.001 0.505 <0.001 <0.001 Other 68 2.6% 16 1.8% 38 2.4% 14 13.1% <0.001 0.265 <0.001 <0.001 SECONDARY ACTIVITY Slot machines 68 2.6% 22 2.4% 39 2.5% 7 6.5% 0.037 0.896 0.016 0.014 Bingo 155 6.0% 67 7.4% 84 5.4% 4 3.7% 0.083 0.048 0.163 0.457 Lotteries 327 12.7% 132 14.5% 186 12.0% 9 8.4% 0.071 0.066 0.083 0.270 Casinos 57 2.2% 25 2.8% 29 1.9% 3 2.8% 0.323 0.146 0.976 0.494 Cards 69 2.7% 23 2.5% 44 2.8% 2 1.9% 0.788 0.663 0.675 0.558 Offline betting 41 1.6% 13 1.4% 26 1.7% 2 1.9% 0.876 0.645 0.723 0.878 Online gambling 19 0.7% 6 0.7% 11 0.7% 2 1.9% 0.375 0.893 0.181 0.187 Other 18 0.7% 9 1.0% 7 0.5% 2 1.9% 0.100 0.107 0.407 0.053 Bold, significant comparison (0.05 level); Bonferroni-Finner’s correction for multiple statistical comparison. but also bingo, casinos, and land-based sport betting as their or emotional distress was harm avoidance. All the mentioned main preferences. personality and emotional vulnerability characteristics make Cluster 1 especially complex from a therapeutic point of view, DISCUSSION given that these patients present greater gambling severity, more vulnerability to stressful situations and specific personality The aim of this study was to explore the existence of distinct characteristics that hinder their ability to cope with adverse GD phenotypes in a large sample of patients seeking treatment situations. These personality traits are characterized by a for GD through clustering analysis, using a wide set of variables, greater tendency to anxiety, worry and insecurity, isolation, and including sociodemographic, gambling-related factors, general disconnection from the environment, poor decision-making psychopathological state, and personality traits as indicators. We skills and planning skills (38). In this profile, therefore, gambling expected that different empirical clusters would emerge based on could be a dysfunctional habit acquired and maintained over emotional distress or psychopathological state, measured using time as a mechanism to avoid problems and difficulties in SCL-90R questionnaire. emotion regulation. Thus, it could represent a group of patients In fact, in keeping with models proposing emotion-regulation more inclined to present poor response to treatment, as various processes to be strongly related to individual differences studies have previously shown (39–41). As such, they may among GD patients (8, 32), the most relevant variable for require treatment programs that incorporate techniques to the clustering analysis was emotional distress (SCL-90-R); improve these aspects, favoring a better response to therapy (8). followed by personality traits (TCI-R), and GD severity measures Another interesting aspect to highlight from these results is (number of gambling activities and SOGS-total score). Thus, that patients with the highest emotional distress and severity three empirical clusters emerged defining psychopathological were the oldest patients with the longest GD duration, most status as their main indicator: (a) Cluster 1 composed of more of them included in the Cluster 1. This result would be in dysfunctional patients who reported high emotional distress; agreement with previous studies that have focused on exploring (b) Cluster 2 grouped patients with low distress levels; (c) the clinical differences of patients diagnosed with GD, depending Cluster 3 characterized by mainly young patients with medium on age (42–44). The findings of these studies suggest that age distress levels. is a strong moderating variable in the evolution of the disorder From a clinical perspective, these findings could be and that it has a relevant impact on the emotional aspects understood taking into account that certain personality traits associated with gambling. In this sense, in the elderly, anxious- or domains could be associated with a greater or lesser degree depressive symptoms would be related to medical illnesses and of psychopathology or emotional distress, as shown in previous health problems, but also to stressful life situations that appear research (33–37). Specifically, according to the results, the at this stage of life (e.g., retirement, reduced purchasing power, personality trait most closely associated with psychopathology loneliness, loss of family or friends, etc.) (35, 45, 46). Frontiers in Psychiatry | www.frontiersin.org 7 March 2019 | Volume 10 | Article 173 Jiménez-Murcia et al. Clustering for GD Patients from our study also displayed different preferences elevated in Cluster 3—have been observed to be among the best in gambling activities, depending on the cluster that they were predictors of low treatment adherence and compliance (51–53). formed part of. Most of patients included in Cluster 1 and This finding is not only relevant to tailoring treatment plans, but Cluster 2 preferred slot machines. In contrast, Cluster 3 was it could also indicate that the low percentage of Cluster 3 patients mainly composed of patients with a higher variety of gambling in clinical sample could be partially due to the fact that a some preferences, even though the most frequent gambling activity potential patients with such a profile are less prone to seek or was again slot machines, these were nearly followed by online remain in treatment. gambling. These results are in agreement with other studies Finally, the progression of the clusters throughout time is having observed that online gamblers usually also engaged in information of clinical and therapeutic interest. It can be seen land-based gambling, combining both modalities (47). Literature how Cluster 2 decreased significantly whereas Clusters 1 and supports that online gambling is mainly practiced by young 3 have increased rates. It could be possible that we are at men (17–19), according to our results where Cluster 3 was a crucial point in terms of a change in the clinical profiles composed of the youngest patients who also prefer new forms of GD individuals, which may require the development of of gambling activities such as online gambling. The increase in new theoretical models and therapeutic strategies that adapt to the use of online gambling may be because of features that make these changes. it more convenient and accessible (15, 16), especially for young people. Relatedly, the rest of features of patients included in Cluster 3 also coincide with profile of online gamblers, being Limitations male, high levels of education, employed with good incomes, The present study has the potential to improve current betting larger amounts of money, as well as diverse gambling knowledge on GD subtypes and provides relevant information preferences (17–19). by describing three different clusters depending on different In more general terms, Cluster 3 seems to largely overlap emotional alterations. However, this research has some with a subtype of gambler at least partially identified in previous limitations that need to be considered. First of all, our sample research. On the one hand, this cluster is likely be responsible was principally made of males referred to a specialized for the pattern of correlations between different types of high unit that usually deals with severe cases of GD. For this arousal, large-stake games (bets, casino games, skill-based games) reason, our results have to be taken with caution before categorized as Type I gambling in Navas et al. (48). On the being generalized to the general population. Future studies other hand, GD patients with this profile (even after controlling should be conducted to explore if the described subtypes for age and other potential confounders) are more likely to are applicable to both sexes and include samples from other present high sensation- or novelty-seeking scores and stronger centers. Second, this is a cross-sectional study; therefore, gambling-related cognitive distortions. At the same time, and no causality can be driven from the results. Future studies somewhat counterintuitively, in these patients, general self- should test if the observed clusters could also predict regulatory, and executive functioning, as well as high-order treatment outcome. cognitive processing, tend to be well preserved (32, 49, 50). Integrity of high-order cognitive and regulatory processes is fully compatible with the fact that Cluster 3 GD patients present a CONCLUSION higher level of education, higher self-directedness scores, and a lower risk of suffering from a comorbid addictive disorder. Cluster formation was carried out mainly taking into account the In spite of these seemingly “favorable” sociodemographic and psychopathological state or emotional distress, using the SCL- neurocognitive features, gambling severity is by no means 90-R. This variable allowed for differentiation of homogeneous reduced in this profile. Actually, the variable cumulated amount groups or subtypes within the spectrum of the disorder. of debts exceeded 94,000 euros in Cluster 3, while in the other In agreement with Chamberlain (54), the identification of groups, it did not reach 10,000 euros. Different studies point differentiated subtypes can contribute to a better understanding out that, in online gambling, spending money is much easier of GD from a clinical perspective. Characterizing phenotypes and faster; therefore, debts are acquired in a shorter period of and endophenotypes will favor the design and implementation time (14, 47). of increasingly specific interventions that improve response Online gambling has brought about the appearance of a new to treatment. GD phenotype, designated as Cluster 3 in our study. Although it includes fewer patients (only 4.2% of the total sample), Cluster 3 represents a completely new and differentiated type of GD AUTHOR CONTRIBUTIONS patient, with respect to the previous subgroup identification studies (8, 10–12). We can postulate, seeing how the online SJ-M, RG, and FF-A designed the experiment and analyzed gambling market is increasing and attracting young people, that the data. SJ-M, RG, FF-A, TS, GM-B, NM-B, NA, MG-P, this group will be more frequent in the near future. Therefore, and ZA conducted the experiment and wrote a first draft of it is necessary to identify indicators of good or bad response to the manuscript. TS, GM-B, RS, JT, ML-M, TM-M, JP, JN, treatment to be able to manage this new and growing phenotype. CS-M, AdP-G, VM-R, and JM further reviewed and modified More specifically, novelty- and sensation-seeking—seemingly the manuscript. Frontiers in Psychiatry | www.frontiersin.org 8 March 2019 | Volume 10 | Article 173 Jiménez-Murcia et al. Clustering for GD FUNDING With the support of the Secretariat for Universities and Research of the Ministry of Business and Knowledge of the Financial support was received through the Ministerio de Government of Catalonia. TM-M and ML-M are supported by Economía y Competitividad (grant PSI2015-68701-R) and a predoctoral Grant of the Ministerio de Educación, Cultura the Investigación subvencionada por la Delegación del y Deporte (FPU16/02087 and FPU15/02911). JP and JN are Gobierno para el Plan Nacional sobre Drogas (2017I067). supported by a grant from the Spanish Government (Ministerio FIS PI14/00290, FIS PI17/01167, and 18MSP001 - 2017I067 de Economía y Competitividad, Secretaría de Estado de received aid from the Ministerio de Sanidad, Servicios Sociales e Investigación, Desarrollo e Innovación; Convocatoria 2017 Igualdad. CIBER Fisiología Obesidad y Nutrición (CIBERobn) de Proyectos I+D de Excelencia, Spain; co-funded by the and CIBER Salud Mental (CIBERSAM), both of which are Fondo Europeo de Desarrollo Regional, FEDER, European initiatives of ISCIII. GM-B is supported by a predoctoral Union), with reference number PSI2017-85488-P. We also want AGAUR grant (2018 FI_B200174), grant co-funded by the to thank CERCA Programme/Generalitat de Catalunya for European Social Fund (ESF) ESF, investing in your future. institutional support. REFERENCES health costs in Germany. Eur J Heal Econ. (2018) 19:967–78. doi: 10.1007/s10198-017-0945-z. 1. American Psychiatric Association. Diagnostic and Statistical Manual of Mental 17. Mallorquí-Bagué N, Fernández-Aranda F, Lozano-Madrid M, Granero R, Disorders. 5th ed. Washington, DC: American Psychiatric Association (2013). Mestre-Bach G, Baño M, et al. Internet gaming disorder and online gambling 2. Calado F, Griffiths MD. Problem gambling worldwide: an update and disorder: clinical and personality correlates. J Behav Addict. (2017) 6:669–77. systematic review of empirical research (2000-2015). J Behav Addict. (2016) doi: 10.1556/2006.6.2017.078 5:592–613. doi: 10.1556/2006.5.2016.073 18. Wood RT, Williams RJ. A comparative profile of the internet gambler: 3. Pavarin R, Fioritti A, Marani S, Gambini D, Turino E, Piazza A. Who are demographic characteristics, game-play patterns, and problem gambling the subjects with gambling-related problems requiring treatment? a study in status. N Media Soc. (2011) 13:1123–41. doi: 10.1177/1461444810397650 northern Italy. J Clin Med. (2018) 7:80. doi: 10.3390/jcm7040080 19. Gainsbury SM, Wood R, Russell A, Hing N, Blaszczynski A. A digital 4. Molinaro S, Benedetti E, Scalese M, Bastiani L, Fortunato L, Cerrai S, et al. revolution: comparison of demographic profiles, attitudes and gambling Prevalence of youth gambling and potential influence of substance use and behavior of internet and non-internet gamblers. Comput Human Behav. other risk factors across 33 European countries: first results from the 2015 (2012) 28:1388–98. doi: 10.1016/j.chb.2012.02.024 ESPAD study. Addiction. (2018) 113:1862–73doi: 10.1111/add.14275 20. Lesieur HR, Blume SB. The South Oaks Gambling Screen (SOGS): a new 5. Ronzitti S, Kraus SW, Hoff RA, Clerici M, Potenza MN. Problem-gambling instrument for the identification of pathological gamblers. Am J Psychiatry. severity, suicidality and DSM-IV Axis II personality disorders. Addict Behav. (1987) 144:1184–8. doi: 10.1176/ajp.144.9.1184 (2018) 82:142–50. doi: 10.1016/j.addbeh.2018.03.003 21. Báez C, Echeburúa E, Fernández-Montalvo J. Características demográficas, 6. Blaszczynski A, Steel Z, Mcconaghy N. Impulsivity in pathological de personalidad y psicopatológicas de los jugadores patológicos de máquinas gambling: the antisocial impulsivist. Addiction. (1997) 92:75–87. tragaperras en tratamiento: un estudio descriptivo. Clínica y Salud. doi: 10.1111/j.1360-0443.1997.tb03639.x (1994) 5:289–305. 7. Steel Z, Blaszczynski A. The factorial structure of pathological gambling. J 22. Stinchfield R. Reliability, validity, and classification accuracy of a measure of Gambl Stud. (1996) 12:3–20. doi: 10.1007/BF01533186 DSM-IV diagnostic criteria for pathological gambling. Am J Psychiatry. (2003) 8. Blaszczynski A, Nower L. A pathways model of problem 160:180–2. doi: 10.1176/appi.ajp.160.1.180 and pathological gambling. Addiction. (2002) 97:487–99. 23. Jimenez-Murcia S, Stinchfield R, Alvarez-Moya E, Jaurrieta N, Bueno B, doi: 10.1046/j.1360-0443.2002.00015.x Granero R, et al. Reliability, validity, and classification accuracy of a Spanish 9. Milosevic A, Ledgerwood DM. The subtyping of pathological translation of a measure of DSM-IV diagnostic criteria for pathological gambling: a comprehensive review. Clin Psychol Rev. (2010) 30:988–98. gambling. J Gambl Stud. (2009) 25:93–104. doi: 10.1007/s10899-008- doi: 10.1016/j.cpr.2010.06.013 9104-x 10. Ledgerwood DM, Petry NM. Subtyping pathological gamblers based on 24. Cloninger CR, Przybeck TR, Svrakic DM, Wetzel RD. The Temperament and impulsivity, depression, and anxiety. Psychol Addict Behav. (2010) 24:680–8. Character Inventory (TCI): A Guide to Its Development and Use. St Louis, MO: doi: 10.1037/a0019906 Center for Psychobiology of Personality, Washington University (1994). 11. Moon M, Lister JJ, Milosevic A, Ledgerwood DM. Subtyping non-treatment- 25. Gutiérrez-Zotes JA, Bayón C, Montserrat C, Valero J, Labad A, Cloninger CR. seeking problem gamblers using the pathways model. J Gambl Stud. (2017) Temperament and Character Inventory-Revised (TCI-R). Standardization 33:841–53. doi: 10.1007/s10899-016-9658-y and normative data in a general population sample. Actas Españolas Psiquiatr. 12. Alvarez-Moya EM, Jiménez-Murcia S, Aymamí MN, Gómez- (2004) 32:8–15. Peña M, Granero R, Santamaría J, et al. Subtyping study of a 26. Derogatis L. SCL-90-R. Administration, Scoring and Procedures Manual. pathological gamblers sample. Can J Psychiatry. (2010) 55:498–506. Clinical Psychometric. Baltimore, MD (1990). doi: 10.1177/070674371005500804 27. Derogatis L. SCL-90-R. Cuestionario de 90 Síntomas-Manual. Madrid: TEA 13. Griffiths M, Hayer T, Meyer G. Problem Gambling in Europe: Challenges, Ediciones (2002). Prevention, and Interventions. Griffiths M, Hayer T, Meyer G, eds. New York, 28. Jiménez-Murcia S, Aymamí-Sanromà M, Gómez-Peña M, Álvarez-Moya NY: Springer (2009). E, Vallejo J. Protocols de Tractament Cognitivoconductual pel joc Patològic 14. Gainsbury SM, Russell A, Hing N, Wood R, Lubman D, Blaszczynski i d’altres Addiccions no Tòxiques. Hospital Universitari de Bellvitge, A. How the internet is changing gambling: findings from an Australian Departament de Salut, Generalitat de Catalunya. Barcelona (2006). prevalence survey. J Gambl Stud. (2015) 31:1–15. doi: 10.1007/s10899-013- 29. Nylund KL, Asparouhov T, Muthén M&, Muthén BO. Deciding on the 9404-7 number of classes in latent class analysis and growth mixture modeling: 15. Bonnaire C. [Internet gambling: what are the risks?]. Encephale. (2012) a monte carlo simulation study. Struct Equ Model. (2007) 14:535–69. 38:42–9. doi: 10.1016/j.encep.2011.01.014 doi: 10.1080/10705510701575396 16. Effertz T, Bischof A, Rumpf H-J, Meyer C, John U. The effect of 30. Finner H. On a monotonicity problem in step-down multiple test procedures. online gambling on gambling problems and resulting economic J Am Stat Assoc. (1993) 88:920–3. doi: 10.2307/2290782 Frontiers in Psychiatry | www.frontiersin.org 9 March 2019 | Volume 10 | Article 173 Jiménez-Murcia et al. Clustering for GD 31. Rousseeuw PJ. Silhouettes: a graphical aid to the interpretation and 45. Dennis CB, Davis TD, Chang J, McAllister C. Psychological vulnerability validation of cluster analysis. J Comput Appl Math. (1987) 20:53–65. and gambling in later life. J Gerontol Soc Work. (2017) 60:471–86. doi: 10.1016/0377-0427(87)90125-7 doi: 10.1080/01634372.2017.1329764 32. Navas JF, Billieux J, Verdejo-García A, Perales JC. Neurocognitive components 46. Parke A, Griffiths M, Pattinson J, Keatley D. Age-related physical and of gambling disorder: implications for policy, prevention, and treatment. In: psychological vulnerability as pathways to problem gambling in older adults. Bowden-Jones H, Dickson C, Dunand C, Simon O, editors. Harm Reduction J Behav Addict. (2018) 7:137–45. doi: 10.1556/2006.7.2018.18 for Problem Gambling: A Public Health Approach. London: Routledge (2019). 47. Gainsbury SM, Russell A, Blaszczynski A, Hing N. The interaction between 33. Aymamí N, Granero R, Penelo E, Fernández-Aranda F, Krug I, Gunnard gambling activities and modes of access: a comparison of internet-only, K, et al. Anger in pathological gambling: Clinical, psychopathological, and land-based only, and mixed-mode gamblers. Addict Behav. (2015) 41:34–40. personality correlates. Span J Psychol. (2014) 17:1–12. doi: 10.1017/sjp. doi: 10.1016/j.addbeh.2014.09.023 2014.40 48. Navas JF, Billieux J, Perandrés-Gómez A, López-Torrecillas F, Cándido 34. Brunborg GS, Hanss D, Mentzoni RA, Molde H, Pallesen S. Problem gambling A, Perales JC. Impulsivity traits and gambling cognitions associated with and the five-factor model of personality: a large population-based study. gambling preferences and clinical status. Int Gambl Stud. (2017) 17:102–24. Addiction. (2016) 111:1428–35. doi: 10.1111/add.13388 doi: 10.1080/14459795.2016.1275739 35. Moragas L, Granero R, Stinchfield R, Fernández-Aranda F, Fröberg F, 49. Perales JC, Navas JF, Ruiz de Lara CM, Maldonado A, Catena A. Causal Aymamí N, et al. Comparative analysis of distinct phenotypes in gambling learning in gambling disorder: beyond the illusion of control. J Gambl Stud. disorder based on gambling preferences. BMC Psychiatry. (2015) 15:86. (2017) 33:705–17. doi: 10.1007/s10899-016-9634-6 doi: 10.1186/s12888-015-0459-0 50. Ruiz de Lara CM, Navas JF, Soriano-Mas C, Sescousse G, Perales JC. 36. Mann K, Lemenager T, Zois E, Hoffmann S, Nakovics H, Beutel M, Regional grey matter volume correlates of gambling disorder, gambling- et al. Comorbidity, family history and personality traits in pathological related cognitive distortions, and emotion-driven impulsivity. Int Gambl Stud. gamblers compared with healthy controls. Eur Psychiatry. (2017) 42:120–8. (2018) 18:195–216. doi: 10.1080/14459795.2018.1448427 doi: 10.1016/j.eurpsy.2016.12.002 51. Aragay N, Jiménez-Murcia S, Granero R, Fernández-Aranda F, Ramos- 37. Vaddiparti K, Cottler LB. Personality disorders and pathological gambling. Grille I, Cardona S, et al. Pathological gambling: understanding Curr Opin Psychiatry. (2017) 30:45–9. doi: 10.1097/YCO.0000000000000300 relapses and dropouts. Compr Psychiatry. (2015) 57:58–64. 38. Cloninger C, Svrakic D, Przybeck T. A Psychobiological model of doi: 10.1016/j.comppsych.2014.10.009 temperament and character. Arch Gen Psychiatry. (1993) 50:975–90. 52. Jiménez-Murcia S, Granero R, Fernández-Aranda F, Arcelus J, Aymamí MN, doi: 10.1001/archpsyc.1993.01820240059008 Gómez-Peña M, et al. Predictors of outcome among pathological gamblers 39. Menchón JM, Mestre-Bach G, Steward T, Fernández-Aranda F, Jiménez- receiving cognitive behavioral group therapy. Eur Addict Res. (2015) 21:169– Murcia S. An overview of gambling disorder: from treatment approaches 78. doi: 10.1159/000369528 to risk factors. F1000Research. (2018) 7:434. doi: 10.12688/f1000research. 53. Jara-Rizzo MF, Navas JF, Steward T, López-Gómez M, Jiménez-Murcia S, 12784.1 Fernández-Aranda F, et al. Impulsividad y conciencia del problema predicen 40. Alvarez-Moya EM, Ochoa C, Jiménez-Murcia S, Aymamí MN, Gómez-Peña la adherencia terapéutica y el abandono del tratamiento en el trastorno por M, Fernández-Aranda F, et al. Effect of executive functioning, decision- juego de azar. Adicciones. (2018) 1041. doi: 10.20882/adicciones.1041 making and self-reported impulsivity on the treatment outcome of pathologic 54. Chamberlain SR, Stochl J, Redden SA, Odlaug BL, Grant JE. Latent class gambling. J Psychiatry Neurosci. (2011) 36:165–75. doi: 10.1503/jpn. analysis of gambling subtypes and impulsive/compulsive associations: time to 090095 rethink diagnostic boundaries for gambling disorder? Addict Behav. (2017) 41. Tominaga Y, Aomori T, Hayakawa T, Kijima N, Morisky DE, Takahashi 72:79–85. doi: 10.1016/j.addbeh.2017.03.020 K, et al. Possible associations of personality traits representing harm avoidance and self-directedness with medication adherence in Japanese Conflict of Interest Statement: The authors declare that the research was patients with type 2 diabetes. J Pharm Heal Care Sci. (2018) 4:16. conducted in the absence of any commercial or financial relationships that could doi: 10.1186/s40780-018-0112-4 be construed as a potential conflict of interest. 42. Granero R, Penelo E, Stinchfield R, Fernandez-Aranda F, Savvidou LG, Fröberg F, et al. Is pathological gambling moderated by age? J Gambl Stud. Copyright © 2019 Jiménez-Murcia, Granero, Fernández-Aranda, Stinchfield, (2014) 30:475–92. doi: 10.1007/s10899-013-9369-6 Tremblay, Steward, Mestre-Bach, Lozano-Madrid, Mena-Moreno, Mallorquí- 43. Jiménez-Murcia S, Granero R, Tárrega S, Angulo A, Fernández-Aranda Bagué, Perales, Navas, Soriano-Mas, Aymamí, Gómez-Peña, Agüera, del Pino- F, Arcelus J, et al. Mediational role of age of onset in gambling Gutiérrez, Martín-Romera and Menchón. This is an open-access article distributed disorder, a path modeling analysis. J Gambl Stud. (2016) 32:327–40. under the terms of the Creative Commons Attribution License (CC BY). The use, doi: 10.1007/s10899-015-9537-y distribution or reproduction in other forums is permitted, provided the original 44. Sauvaget A, Jiménez-Murcia S, Fernández-Aranda F, Fagundo AB, Moragas L, author(s) and the copyright owner(s) are credited and that the original publication Wolz I, et al. Unexpected online gambling disorder in late-life: a case report. in this journal is cited, in accordance with accepted academic practice. No use, Front Psychol. (2015) 6:655. doi: 10.3389/fpsyg.2015.00655 distribution or reproduction is permitted which does not comply with these terms. Frontiers in Psychiatry | www.frontiersin.org 10 March 2019 | Volume 10 | Article 173