Abstract

High impulsivity is common to substance and gambling addictions. Despite these commonalities, there is still substantial heterogeneity on impulsivity levels within these diagnostic groups, and variations in impulsive levels predict higher severity of symptoms and poorer outcomes. We addressed the question of whether impulsivity scores can yield empirically driven subgroups of addicted individuals that will exhibit different clinical presentations and outcomes. We applied latent class analysis (LCA) to trait (UPPS-P impulsive behavior scale) and cognitive impulsivity (Stroop and d2 tests) scores in three predominantly male addiction diagnostic groups: Cocaine with Personality Disorders, Cocaine Non-comorbid, and Gambling and analyzed the usefulness of the resulting subgroups to differentiate personality beliefs and relevant outcomes: Craving, psychosocial adjustment, and quality of life. In accordance with impulsivity scores, the three addiction diagnostic groups are best represented as two separate classes: Class 1 characterized by greater trait impulsivity and poorer cognitive impulsivity performance and Class 2 characterized by lower trait impulsivity and better cognitive impulsivity performance. The two empirically derived classes showed significant differences on personality features and outcome variables (Class 1 exhibited greater personality dysfunction and worse clinical outcomes), whereas conventional diagnostic groups showed non-significant differences on most of these measures. Trait and cognitive impulsivity scores differentiate subgroups of addicted individuals with more versus less severe personality features and clinical outcomes.

Introduction

The term impulsivity refers to behavior that is performed with little or inadequate forethought such that it results in undesirable consequences for self or others (Evenden, 1999; Moeller, Barrat, Dougherty, Schmitz, & Swann, 2001). Accordingly, impulsivity has been trans-diagnostically associated with a number of debilitating psychiatric disorders including substance use, gambling, and personality disorders (Robbins, Gillan, Smith, de Wit, & Ersche, 2012; Steel & Blaszczynski, 1998). This notion has been recently incorporated into the 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) (American Psychiatric Association, 2013), which merges substance use and gambling disorders under the addiction diagnostic category based—among others—on their common impulsive features (Petry, 2001). These commonalities include similarly increased trait impulsive levels and similarly decreased performance on cognitive measures of selective attention and response inhibition (Albein-Urios, Martinez-González, Lozano, Clark, & Verdejo-García, 2012; Goudriaan, Oosterlaan, de Beurs, & van den Brink, 2006; Lawrence, Luty, Bogdan, Sahakian, & Clark, 2009) which have shown sound correspondence with trait impulsivity levels (Cyders & Conskunpinar, 2011; Perales, Verdejo-García, Moya, Lozano, & Pérez-García, 2009).

Despite these commonalities, there is still substantial heterogeneity on impulsivity levels within these diagnostic groups (Verdejo-García, Lawrence, & Clark, 2008). This heterogeneity entails significant clinical implications, since higher levels of impulsivity are significantly associated with more severe clinical symptoms and poorer psychosocial outcomes in cocaine use (especially when co-existing with comorbid personality disorders) and gambling disorders (Bornovalova, Levy, Gratz, & Lejuez, 2010; Ledgerwood & Petry, 2010). More specifically, the emotion-driven dispositions to impulsive behavior (e.g., negative urgency) and the attention/response inhibition aspects of cognitive impulsivity have been uniquely associated with increased severity of symptoms and poorer outcomes in both cocaine (Bornovalova, Levy, Gratz, & Lejuez, 2010; Verdejo-García, Bechara, Recknor, & Pérez-García, 2007) and gambling addictions (Brevers et al., 2012; Torres et al., 2013).

In view of substantial heterogeneity on impulsivity levels within addiction diagnostic groups, and relevance of this heterogeneity to describe higher severity and poorer outcomes, we addressed the question of whether impulsivity scores can yield empirically driven trans-diagnostic subgroups of addicted individuals that will exhibit different clinical presentations and outcomes. These subgroups may contribute to provide better proxies of addiction severity and outcomes compared with traditional diagnostic labels. To test these assumptions, we applied latent class analysis (LCA) to trait and cognitive impulsivity scores in three addiction diagnostic groups (cocaine with personality disorders, cocaine non-comorbid, and gambling) and analyzed the usefulness of the resulting subgroups to differentiate personality beliefs and relevant outcomes: Craving, psychosocial adjustment, and quality of life (Tiffany, Friedman, Greenfield, Hasin, & Jackson, 2012). The LCA approach is ideally suited to uncover unobserved heterogeneity within diagnostic groups with similar impulsive profiles and well characterized to discard state-dependent confounders (e.g., acute or subacute drug effects, Axis I comorbidities). We predicted that the LCA-driven impulsivity subgroups will unveil significant differences on personality dysfunction and outcome measures beyond those detected by conventional diagnostic groups.

Methods

Participants

Ninety-six European-Caucasian individuals with cocaine use and gambling disorders diagnoses participated in this study as part of a larger research on cocaine and personality disorders (COPERNICO project). Thirty-eight individuals were diagnosed with cocaine use and personality disorders from Cluster B and Cluster C, 32 individuals were diagnosed with cocaine use disorders without other current Axis I or Axis II comorbidities, and 26 individuals were diagnosed with gambling disorders without other current Axis I or Axis II comorbidities. The three diagnostic groups were statistically matched for age, education, and IQ distributions and for tobacco, alcohol, and cannabis amount and durations of use.

Cocaine users were recruited as they started treatment in the clinic “Centro Provincial de Drogodependencias (CPD)” in Granada (Spain), which provides behavioral treatment for substance-related disorders in an outpatient setting. Gambling users were recruited as they started treatment in the clinic “Asociación Granadina de Jugadores en Rehabilitación (AGRAJER)” also located in Granada (Spain). This outpatient facility gives self-help oriented interventions for problem gambling and is the main public funded treatment service for gambling problems in the south of Spain.

The inclusion criteria for the cocaine and gambling groups were defined as follows: (a) age range between 18 and 50 years old; (b) IQ levels above 80—as measured by the Kaufman Brief Intelligence Test (K-BIT; Kaufman & Kaufman, 1990); (c) meeting DSM-IV criteria for cocaine dependence (for the cocaine groups) or pathological gambling (for the gambling group)—as assessed by the Structured Clinical Interview for DSM-IV Disorders-Clinician Version (SCID; First, Spitzer, Gibbon, & Williams, 1997); (d) being treatment commencers; and (e) abstinence duration >15 days—as determined by twice weekly urine tests in cocaine users and self- and informant reports in pathological gamblers. Inclusion criteria for cocaine-dependent patients with comorbid personality disorders were restricted to diagnoses pertaining to Cluster B (n = 34) and Cluster C (n = 17), which are the more prevalent among cocaine users (Chen et al., 2011). Axis II disorders were assessed using the International Personality Disorders Examination (Loranger et al., 1994; Spanish version by López-Ibor, 1999). Cluster B diagnoses included antisocial (n = 6), borderline (n = 16), narcissistic (n = 1), and histrionic (n = 11). Cluster C diagnoses included avoidant (n = 11) and obsessive-compulsive (n = 6).

The exclusion criteria were: (a) presence of any other current Axis I disorders—with the exceptions of alcohol abuse, nicotine dependence, and attention deficit and hyperactivity disorder (ADHD)—as measured by the Conners' Adult ADHD Diagnostic Interview for DSM-IV (CAADID; Conners, 1999); (b) history of head injury or neurological, infectious, systemic, or any other diseases affecting the central nervous system; (c) having participated in other behavioral treatments within the 2 years preceding the study onset; and (d) having entered treatment by court request. Comorbid Axis I disorders were assessed with the SCID.

All the diagnoses were conducted by a board certified clinical psychologist, whereas all subsequent tests were administered by an independent (blind to diagnosis) assessor.

Measures

Classification measures: Trait and Cognitive Impulsivity

Trait Impulsivity was measured with the UPPS-P impulsive behavior scale (Whiteside & Lynam, 2001). Specifically, we used the Spanish version of the scale, which has shown sound reliability and internal and construct validity (Verdejo-García, Lozano, Moya, Alcázar, & Pérez-García, 2010). The scale contains 59 items designed to comprehensively assess different personality pathways leading to impulsive behavior: Sensation seeking, lack of perseverance, lack of premeditation, negative urgency, and positive urgency. Sensation seeking (12 items) incorporates two aspects: (a) the tendency to enjoy and pursue activities that are exciting, and (b) an openness to trying new experiences that may or may not be dangerous; lack of perseverance (10 items) refers to an individual's ability to remain focused on a task that may be boring or difficult; lack of premeditation (11 items) refers to the tendency to think and reflect on the consequences of an act before engaging in that act; and finally urgency (26 items) refers to the tendency to experience strong impulses under conditions of negative affect (negative urgency, 12 items) or positive affect (positive urgency, 14 items). The reliability of the different subscales (Cronbach's α) ranged from 0.75 (lack of perseverance) to 0.93 (positive urgency). We obtained the total scores of each of these UPPS-P dimensions for analyses.

Cognitive impulsivity was measured with two neuropsychological tests of selective attention and response inhibition which have empirically demonstrated validity to represent the construct of impulsivity (Cyders & Coskunpinar, 2011; Perales, Verdejo-García, Moya, Lozano, & Pérez-García, 2009).

Stroop Color-Word Interference Test (CWIT; Delis, Kaplan, & Kramer, 2001). This version of the Stroop includes a series of four conditions. The first condition (C1) presents patches of colors and participants have to name them as quickly and accurately as they can. The second condition (C2) presents the words “red,” “blue,” and “green” printed in black ink and participants are asked to read aloud the words written. The third condition (C3) introduces the inhibition demand: The words “red,” “blue,” and “green” are printed in incongruent colors ink and participants have to name the color and ignore the word. In the fourth condition (C4), the items are similar to condition 3 but participants have to switch their response between naming the color of the ink and ignoring the word or reading the word (when the item is framed). Based on our study aims, we used as our performance index the response inhibition score, resulting from the formula Time on C3 minus Time on C1.

d2 Cancellation Test (Brickenkamp, 2002). This test includes 14 different lines of letters including targets (d's with two dashes) and distracters (e.g., d's with less than two dashes, p's, etc.). Participants are asked to cancel each of the targets while ignoring the distracters. Based on our study aims, we used as our performance index the efficiency score: Total number of trials minus total number of errors, indexing selective attention and inhibition of distracters.

Dimensional measures of addictive behavior and personality dysfunction

Substance use behavior: Interview for Research on Addictive Behavior (Verdejo-García, López-Torrecillas, Aguilar de Arcos, & Pérez-García, 2005). This interview collects self-reported information about patterns of drug use including the age at onset of the substances used, monthly use of each substance during regular use and last month (amount per month), and the total duration of use for each substance (duration in months).

Dysfunctional beliefs: Personality Belief Questionnaire (PBQ; Beck & Beck, 1991). The PBQ is a self-report questionnaire that consists of 10 subscales that measure specific beliefs and assumptions associated with the different personality disorders. Here, we used the nine scales corresponding to different clusters of personality disorders: Paranoid and schizoid (Cluster A), antisocial, borderline, histrionic, and narcissistic (Cluster B), and avoidant, dependent, and obsessive-compulsive (Cluster C). The Spanish version of the questionnaire holds appropriate psychometric characteristics (Albein-Urios, Martínez-González, Lozano, & Verdejo-García, 2011) and the reliability coefficients of the different scales in this sample ranged from Cronbach's = 0.71 (narcissistic) to Cronbach's = 0.88 (borderline).

Outcome measures

Craving was measured with the Weiss Craving Questionnaire (Weiss et al., 1997). This questionnaire is composed of five items that measure different aspects of craving during addiction treatment. We computed the total craving scores for our analyses.

Psychosocial adjustment was measured with the General Health Questionnaire (GHQ-28; Lobo, Pérez-Echevarria, & Artal, 1986). This questionnaire is composed of 28 items corresponding to four dimensions of psychological and social adjustment in the context of psychiatric disorders: somatic complaints, anxiety, social dysfunction, and depression.

Quality of life was measured with the Health-Related Quality of Life for Drug Users (Lozano, Rojas, & Pérez, 2009; Zubaran et al., 2012). This 20-item self-report inventory has demonstrated validity to assess subjective quality of life in individuals with substance use disorders. It provides a total score representing global subjective well-being. Due to its specific validity in substance using populations, we did not apply this measure to the gambling group.

Statistical Analysis

We initially conducted one-way analyses of variance to examine differences between diagnostic groups (Cocaine + Personality Disorders, Cocaine, Gambling) on impulsivity, substance use, personality beliefs, and outcome measures. Next, we applied LCA to impulsivity scores in all participants in order to obtain novel subgroups of individuals differing on impulsivity profiles. LCA assumes that the covariation between manifest indicators arises by virtue of their association with underlying classes. LCA facilitates the extraction of distinct and meaningful subgroups based on the unobserved heterogeneity within a population and on the similarity between their response profiles (Reboussin, Young Song, Shrestha, Lohman, & Wolfson, 2006). Here, we employed the trait and cognitive impulsivity indices as indicators for the LCA. Specifically, we used the UPPS-P scales scores (sensation seeking, lack of perseverance, lack of premeditation, negative urgency, and positive urgency), the Stroop Inhibition score, and the d2 Efficiency score. The selection of the number of latent classes that best described our data was conducted using the Bayesian information criterion (BIC) and the Akaike information criteria (AIC). The BIC and AIC are descriptive fit indices wherein smaller values indicate better model fit. In addition, each model was assessed on its interpretability to determine if the classes actually represented different categories (Muthén, 2006). The LCA model yields two types of estimated parameters: (a) class membership probabilities, reflecting the relative size or prevalence of each class and (b) class-specific endorsement probabilities, reflecting the likelihood of endorsement of a given indicator for individuals in a particular class (Uebersax, 1994). We used Latent Gold 4.0 software for model-fitting. The latent class membership was saved for use in subsequent bivariate analyses conducted in SPSS v. 17.

To address the main prediction of the study, we next performed independent-sample t-tests or χ2 tests to analyze the association between LCA-derived membership and diagnostic composition (presence of comorbid personality disorders), personality beliefs, and outcome measures (craving, psychological adjustment, and subjective quality of life). The LCA model was applied to the whole sample (n = 92) but, because questionnaire data were missing, the sample size for clinical and outcome measures was n = 69 for PBQ scores and n = 76 for psychosocial and outcome measures.

Results

Differences Between Diagnostic Groups on Clinical and Outcome Measures

Results are presented in Table 1. The three diagnostic groups (cocaine with personality disorders, cocaine, and gambling) showed statistically significant differences in the UPPS-P negative urgency subscale, F(2, 90) = 9.69, p < .05; and in the PBQ Schizoid, F(2, 67) = 3.73, p < .05, Borderline, F(2, 67) = 3.73, p < .05, and Narcissistic, F(2,67) = 4.09, p < .05, subscales. In all cases, cocaine users with personality disorders had higher scores than cocaine non-comorbid users and gamblers. Diagnostic groups showed non-significant differences in the cognitive measures of impulsivity (Stroop and d2), the four remaining UPPS-P subscales or the six remaining PBQ subscales. As for outcome measures, diagnostic groups showed non-significant differences in craving or psychosocial outcome, although cocaine users with personality disorders showed poorer perceived quality of life than cocaine non-comorbid users.

Table 1.

Descriptive scores of impulsivity, personality beliefs, and outcome measures in the traditional diagnostic groups

 Cocaine + Personality disorders Cocaine Gambling 
Trait impulsivity 
 Negative urgency 35.11 (5.2) 31.3 (6.82) 28.24 (6.06) 
 Premeditation 25.68 (6.56) 25.21 (4.97) 25.56 (4.74) 
 Perseverance 22.77 (4.97) 20.39 (4.41) 21.76 (5.56) 
 Sensation seeking 29.14 (7.62) 29.12 (8.25) 28.8 (6.28) 
 Positive urgency 33.83 (9.73) 31 (10.15) 28.44 (8.01) 
Cognitive impulsivity 
 d2 efficiency 407.42 (83.66) 416.36 (82.47) 434.28 (85.66) 
 Stroop inhibition 25.86 (17.23) 19.84 (10.48) 21.8 (12.47) 
Dysfunctional beliefs 
 Schizoid 24.69 (12.03) 21.31 (10.58) 14.33 (9.53) 
 Paranoid 22.03 (15.08) 17.06 (12.81) 14.5 (13.91) 
 Antisocial 27.1 (10.73) 24.58 (11.43) 20.33 (14.01) 
 Borderline 19.1 (9.89) 12.38 (9.19) 12 (13.57) 
 Histrionic 18.34 (10.35) 15.17 (7.4) 13.83 (10.95) 
 Narcissistic 18.17 (8.12) 15.59 (8.23) 10.25 (7.56) 
 Avoidant 19.34 (8.65) 14.83 (9.26) 13.25 (11.57) 
 Dependent 21.93 (11.22) 17.24 (8.97) 16.83 (15.44) 
 Obsessive-compulsive 26.24 (10.37) 23.31 (11.77) 18.33 (10.2) 
Outcome measures 
 Craving 8.06 (7.36) 6.17 (8.49) 4.31 (6.48) 
 Somatic symptoms 1.81 (1.76) 1.45 (1.78) 1.64 (2.13) 
 Anxiety 3.0 (2.45) 2.1 (2.38) 2.21 (2.36) 
 Social dysfunction 1.56 (2.26) 0.93 (1.58) 1.07 (1.86) 
 Depression 1.97 (2.44) 1.27 (2.17) 1.77 (2.31) 
 Quality of life 83.61 (10.5) 89.96 (11.78) — 
 Cocaine + Personality disorders Cocaine Gambling 
Trait impulsivity 
 Negative urgency 35.11 (5.2) 31.3 (6.82) 28.24 (6.06) 
 Premeditation 25.68 (6.56) 25.21 (4.97) 25.56 (4.74) 
 Perseverance 22.77 (4.97) 20.39 (4.41) 21.76 (5.56) 
 Sensation seeking 29.14 (7.62) 29.12 (8.25) 28.8 (6.28) 
 Positive urgency 33.83 (9.73) 31 (10.15) 28.44 (8.01) 
Cognitive impulsivity 
 d2 efficiency 407.42 (83.66) 416.36 (82.47) 434.28 (85.66) 
 Stroop inhibition 25.86 (17.23) 19.84 (10.48) 21.8 (12.47) 
Dysfunctional beliefs 
 Schizoid 24.69 (12.03) 21.31 (10.58) 14.33 (9.53) 
 Paranoid 22.03 (15.08) 17.06 (12.81) 14.5 (13.91) 
 Antisocial 27.1 (10.73) 24.58 (11.43) 20.33 (14.01) 
 Borderline 19.1 (9.89) 12.38 (9.19) 12 (13.57) 
 Histrionic 18.34 (10.35) 15.17 (7.4) 13.83 (10.95) 
 Narcissistic 18.17 (8.12) 15.59 (8.23) 10.25 (7.56) 
 Avoidant 19.34 (8.65) 14.83 (9.26) 13.25 (11.57) 
 Dependent 21.93 (11.22) 17.24 (8.97) 16.83 (15.44) 
 Obsessive-compulsive 26.24 (10.37) 23.31 (11.77) 18.33 (10.2) 
Outcome measures 
 Craving 8.06 (7.36) 6.17 (8.49) 4.31 (6.48) 
 Somatic symptoms 1.81 (1.76) 1.45 (1.78) 1.64 (2.13) 
 Anxiety 3.0 (2.45) 2.1 (2.38) 2.21 (2.36) 
 Social dysfunction 1.56 (2.26) 0.93 (1.58) 1.07 (1.86) 
 Depression 1.97 (2.44) 1.27 (2.17) 1.77 (2.31) 
 Quality of life 83.61 (10.5) 89.96 (11.78) — 

Note: Numbers represent means (SD).

Extraction of Impulsivity-Based Groups: LCA Results

We explored five different solutions in order to obtain the best-fitting model. The two-class LCA solution provided the most parsimonious and stable model based on BIC values (Table 2). The Wald test yielded statistically significant values for each of the impulsivity measures, indicating that response profiles to these measures contribute to discriminate between classes.

Table 2.

Fit measures for the estimated models

 BIC (LL) AIC (LL) Npar Classification error 
1-Model 5,221.39 5,185.20 14 
2-Model 5,196.07 5,121.10 29 0.07 
3-Model 5,198.05 5,084.31 44 0.10 
4-Model 5,216.42 5,063.90 59 0.08 
5-Model 5,248.73 5,057.44 74 0.06 
 BIC (LL) AIC (LL) Npar Classification error 
1-Model 5,221.39 5,185.20 14 
2-Model 5,196.07 5,121.10 29 0.07 
3-Model 5,198.05 5,084.31 44 0.10 
4-Model 5,216.42 5,063.90 59 0.08 
5-Model 5,248.73 5,057.44 74 0.06 

Notes: BIC = Bayesian information criterion; AIC = Akaike information criteria; Npar = number of parameters.

In order to optimize participants' allocation to classes, we only classified those cases in which probability was superior to 60%. Following this criterion, we classified 93 cases (97%) and only 5 cases (3%) stood unclassified. Class 1, containing 53% of the sample, included individuals with high probabilities of elevated scores on the five dimensions of the UPPS-P (negative urgency = .72; positive urgency = .99; lack of premeditation = .76; lack of perseverance = .76; and sensation seeking = .83), longer Stroop inhibition times (.78), and lower d2 efficiency scores (.72). Class 2, containing 47% of the sample, included individuals with high probabilities of lower scores on the UPPS-P dimensions (negative urgency = .95; positive urgency = .92; lack of premeditation = .73, lack of perseverance = .83; and sensation seeking = .71), faster Stroop inhibition times (.69), and higher d2 efficiency scores (.74). The mean probability of pertaining to Class 1 was of .94 (range .6–1.0) and the mean probability of pertaining to Class 2 was of .93 (range .6–1.0), indicating sound classification for most participants. Fig. 1 display the score profile of both classes on impulsivity measures.

Fig. 1.

Profile of impulsivity scores of the two classes after rescaling scores into a 0–1 scale.

Fig. 1.

Profile of impulsivity scores of the two classes after rescaling scores into a 0–1 scale.

Differences Between LCA-Derived Groups on Clinical and Outcome Measures

Class 1 (n = 50) was formed by 26 participants with cocaine use and personality disorders (52%), 16 participants with cocaine use disorders (32%), and 8 participants with gambling disorders (16%). Class 2 (n = 43) was formed by 9 participants with cocaine use and personality disorders (21%), 17 participants with cocaine use disorder, and 17 participants with gambling disorders (39.5%).

The two classes showed non-significant differences on relevant demographic variables or substance use patterns including tobacco, alcohol, cannabis, and cocaine use (Supplementary material online, Table S1). The two classes showed non-significant differences on adolescent substance use (χ2 = 0.94, p > .1) or ADHD diagnosis (χ2 = 6.09, p > .1). Classes did not either differ on lifetime prevalence of other Axis I disorders (χ2 = 2.56, p > .05). Cocaine users classified in Class 1 versus Class 2 did not differ on cocaine use parameters: Age at onset, monthly use, duration of use, recent use, or abstinence duration.

Results concerning clinical and outcome variables are presented in Table 3. The two classes showed statistically significant differences on the nine subscales of the PBQ, measuring degree of dysfunctional beliefs. The two classes also showed statistically significant differences on all the outcome measures: Craving, GHQ psychosocial measures (somatic symptoms, anxiety, social dysfunction, and depression), and quality of life. In all cases, Class 1 showed significantly elevated clinical symptoms and poorer outcomes.

Table 3.

Descriptive scores and statistics for individuals classified in Class 1 and Class 2 on diagnostic composition, impulsivity, dysfunctional beliefs, and outcome measures

 Class 1 Class 2 t p 
Diagnostic composition CB (34%), CC (18%), CO (16%), PG (32%) CB (14%), CC (7%), CO (39.5%), PG (39.5%)   
Trait impulsivity 
 Negative urgency 36.1 (4.71) 27.04 (4.92) 9.056 .000 
 Premeditation 27.52 (5.54) 23.11 (4.49) 4.164 .000 
 Perseverance 23.74 (4.27) 19.23 (4.72) 4.831 .000 
 Sensation seeking 31.52 (7.22) 26.16 (6.69) 3.689 .000 
 Positive urgency 37.04 (8.29) 24.79 (6.26) 7.932 .000 
Cognitive impulsivity 
 d2 efficiency 394.28 (79.05) 445.18 (81.1) −3.059 .003 
 Stroop inhibition 26.5 (16.76) 18.13 (7.86) 2.997 .004 
Dysfunctional beliefs 
 Schizoid 26.05 (11.01) 15.81 (9.45) 4.112 .000 
 Paranoid 24.69 (13.26) 11.13 (11.3) 4.530 .000 
 Antisocial 27.48 (11.04) 21.64 (11.8) 2.128 .037 
 Borderline 20.67 (9.55) 8.09 (7.59) 5.974 .000 
 Histrionic 19.12 (9.61) 12.65 (7.86) 3.034 .003 
 Narcissistic 18.84 (8.19) 11.83 (7.09) 3.771 .000 
 Avoidant 20.97 (9.03) 10.7 (7.05) 5.190 .000 
 Dependent 22.48 (11.3) 14.87 (9.87) 2.957 .004 
 Obsessive-compulsive 26.67 (10.2) 19.9 (11.29) 2.627 .011 
Outcome measures 
 Craving 9.16 (8.47) 3.56 (5.26) 3.373 .001 
 Somatic symptoms 2.26 (1.99) 0.88 (1.27) 3.651 .001 
 Anxiety 3.51 (2.42) 1.29 (1.78) 4.561 .000 
 Social dysfunction 1.83 (2.27) 0.5 (1.11) 3.307 .002 
 Depression 2.51 (2.58) 0.6 (1.32) 4.108 .000 
 Quality of life 81.38 (10.25) 94.46 (8.99) 5.101 .000 
 Class 1 Class 2 t p 
Diagnostic composition CB (34%), CC (18%), CO (16%), PG (32%) CB (14%), CC (7%), CO (39.5%), PG (39.5%)   
Trait impulsivity 
 Negative urgency 36.1 (4.71) 27.04 (4.92) 9.056 .000 
 Premeditation 27.52 (5.54) 23.11 (4.49) 4.164 .000 
 Perseverance 23.74 (4.27) 19.23 (4.72) 4.831 .000 
 Sensation seeking 31.52 (7.22) 26.16 (6.69) 3.689 .000 
 Positive urgency 37.04 (8.29) 24.79 (6.26) 7.932 .000 
Cognitive impulsivity 
 d2 efficiency 394.28 (79.05) 445.18 (81.1) −3.059 .003 
 Stroop inhibition 26.5 (16.76) 18.13 (7.86) 2.997 .004 
Dysfunctional beliefs 
 Schizoid 26.05 (11.01) 15.81 (9.45) 4.112 .000 
 Paranoid 24.69 (13.26) 11.13 (11.3) 4.530 .000 
 Antisocial 27.48 (11.04) 21.64 (11.8) 2.128 .037 
 Borderline 20.67 (9.55) 8.09 (7.59) 5.974 .000 
 Histrionic 19.12 (9.61) 12.65 (7.86) 3.034 .003 
 Narcissistic 18.84 (8.19) 11.83 (7.09) 3.771 .000 
 Avoidant 20.97 (9.03) 10.7 (7.05) 5.190 .000 
 Dependent 22.48 (11.3) 14.87 (9.87) 2.957 .004 
 Obsessive-compulsive 26.67 (10.2) 19.9 (11.29) 2.627 .011 
Outcome measures 
 Craving 9.16 (8.47) 3.56 (5.26) 3.373 .001 
 Somatic symptoms 2.26 (1.99) 0.88 (1.27) 3.651 .001 
 Anxiety 3.51 (2.42) 1.29 (1.78) 4.561 .000 
 Social dysfunction 1.83 (2.27) 0.5 (1.11) 3.307 .002 
 Depression 2.51 (2.58) 0.6 (1.32) 4.108 .000 
 Quality of life 81.38 (10.25) 94.46 (8.99) 5.101 .000 

Discussion

LCA results showed that, in accordance with impulsivity scores, the three addiction diagnostic groups are best represented as two separate classes: Class 1 characterized by greater trait impulsivity and poorer cognitive impulsivity performance and Class 2 characterized by lower trait impulsivity and better cognitive impulsivity performance. The two empirically derived Classes showed significant differences on personality features and outcome variables (Class 1 exhibited greater personality dysfunction and worse clinical outcomes), whereas conventional diagnostic groups showed non-significant differences on most of these measures. Of note, these profiles emerged in the absence of significant differences on patterns of substance use between Classes. All in all, our results support the value of impulsivity measures to identify trans-diagnostic subgroups of addicted individuals with different personality features and clinical outcomes.

Our first finding indicates that the highly impulsive Class 1 is characterized by greater levels of personality dysfunction. Previous studies on conventional diagnostic groups have yielded controversial results, since comorbidity between addiction and personality disorders has been associated with both increased (Albein-Urios et al., 2013) and decreased impulsivity levels (Vassileva, Gonzalez, Bechara, & Martin, 2007) probably due to inherent heterogeneity within these samples. Conversely, LCA-derived subgroups based on impulsivity are better suited to address this heterogeneity and therefore to provide better proxies of personality dysfunction, represented both by diagnostic composition (there were significantly more personality disorders diagnoses in Class 1) and by dimensional levels of personality beliefs (dysfunctional beliefs were significantly increased in Class 1). Between-class differences on dysfunctional beliefs corresponded not only to the personality disorders represented in the sample (Cluster B and Cluster C) but also to Cluster A's paranoid or schizoid disorders. The fact that classes did not discriminate between Cluster B and Cluster C diagnoses but clearly differentiated between high versus low levels of dysfunctional beliefs (relevant to several disorders) support the notion that these classes represent proxies of global psychosocial functioning (Kuyken, Kurzer, DeRubeis, Beck, & Brown, 2001). These results are also in line with those of previous studies that have successfully applied LCAs to identify subgroups of cocaine-dependent individuals with higher temperamental vulnerabilities and greater psychosocial stressors (Bornovalova et al., 2010).

The second main finding indicates that the highly impulsive Class 1 is characterized by poorer functioning across the different outcome measures: Increased levels of craving, worse psychosocial functioning, and lower perceived quality of life. These findings agree with previous evidence showing that higher impulsivity is significantly associated with greater severity of symptoms and greater psychological and social burden in cocaine use and gambling diagnostic groups (Bornovalova et al., 2010; Ledgerwood & Petry, 2010). We expand these findings by demonstrating that trait and cognitive impulsivity measures are useful to characterize clinically meaningful subgroups of addicted individuals with different clinical profiles. Impulsivity-based classes may therefore contribute to identify addicted individuals in need of tailored interventions (Friedmann Hendrickson, Gerstein, & Zhang, 2004). Because decreases in impulsivity are a mediator of the association between addiction treatment and recovery of psychosocial outcomes (Blonigen, Timko, Finney, Moos, & Moos, 2011) both direct and indirect interventions to tame impulsive behavior are warranted within this group. Due to the link between impulsivity and a number of clinical and psychosocial deficits, the interventions directed to this subgroup should prioritize those areas in which patients experience higher needs and subsequently translate these deficits into meaningful treatment goals (Miller & Miller, 2009).

Our findings should be appraised in the context of several considerations. First, the cocaine group included users with personality disorders, whereas the gambling group had not a comorbid parallel group and this stands as a limitation of the study. The main reason for not including this fourth group was the high rate of comorbid Axis I and Axis II disorders within the gambling population (Giddens, Stefanovics, Pilver, Desai, & Potenza, 2012) and the resulting lower prevalence of personality disorders in the absence of Axis I disorders. Second, we did not use a comprehensive assessment of impulsivity, but we specifically selected those measures previously associated with higher severity of personality dysfunction and poorer outcomes in cocaine and gambling disorders. Moreover, the cognitive measures selected have demonstrated significant associations with trait impulsivity indices in previous research (Cyders & Coskunpinar, 2011; Perales, Verdejo-García, Moya, Lozano, & Pérez-García, 2009). Third, a particular limitation of the sample was the reduced proportion of women, which is reflective of population prevalence, but clearly limits generalization of findings to female addicted populations. Fourth, we had a relatively small sample size. Nonetheless, although LCA models are typically used in larger samples, there is no consensus about the minimum sample size required for these models. In fact, the stability of LCA relies largely on other factors, such as sample homogeneity, number of predictive variables included in the model, and balance between class size and number of classes (Swanson, Lindenberg, Bauer, & Crosby, 2011). Considering these criteria, the LCA approach was well suited for this sample, since participants had homogeneous addiction-related diagnoses; we used theory-driven selected predictors; a small number of cases were discarded due to ambiguity in class membership; and models yielded low estimation errors and high probabilities of cluster fit. Nonetheless, future studies with larger sample sizes are warranted to corroborate the existence of these classes. All in all, the relevance of the study lays in the application of LCA to the impulsivity scores of this carefully characterized clinical sample, and the extraction of impulsivity data-driven subgroups differing on personality features and clinical outcomes. LCA is a model-based approach, such that its results are thought to be applicable to the population from which the data sample is extracted (Magidson & Vermunt, 2002), supporting the robustness of our findings.

Supplementary material

Supplementary material is available at Archives of Clinical Neuropsychology online.

Funding

This study has been funded by the grant COPERNICO from the Drug Abuse Plan (Plan Nacional sobre Drogas Convocatoria 2009) and by the RETICS Program (“Red de Trastornos Adictivos”, Instituto de Salud Carlos III), both from the Spanish Ministry of Health.

Conflict of Interest

None declared.

References

Albein-Urios
N.
Martinez-González
J. M.
Lozano
O.
Clark
L.
Verdejo-García
A.
Comparison of impulsivity and working memory in cocaine addiction and pathological gambling: Implications for cocaine-induced neurotoxicity
Drug and Alcohol Dependence
 , 
2012
, vol. 
126
 (pg. 
1
-
6
)
Albein-Urios
N.
Martinez-Gonzalez
J. M.
Lozano
O.
Moreno-López
L.
Soriano-Mas
C.
Verdejo-García
A.
Negative urgency, disinhibition and reduced temporal pole gray matter characterize the comorbidity of cocaine dependence and personality disorders
Drug and Alcohol Dependence
 , 
2013
Albein-Urios
N.
Martínez-González
J. M.
Lozano
O. M.
Verdejo-García
A.
Estudio preliminar para la validación de la versión española del Personality Belief Questionnaire (PBQ)
Trastornos Adictivos
 , 
2011
, vol. 
13
 (pg. 
144
-
150
)
American Psychiatric Association
Diagnostic and Statistical Manual of Mental Disorders (Fifth ed.)
 , 
2013
Arlington, VA
American Psychiatric Publishing
Beck
A. T.
Beck
J. S.
The Personality Belief Questionnaire
 , 
1991
Bala Cynwyd, PA
Beck Institute for Cognitive Therapy and Research
Blonigen
D. M.
Timko
C.
Finney
J. W.
Moos
B. S.
Moos
R. H.
Alcoholics anonymous attendance, decreases in impulsivity and drinking and psychosocial outcomes over 16 years: Moderated-mediation from a developmental perspective
Addiction
 , 
2011
, vol. 
106
 (pg. 
2167
-
2177
)
Bornovalova
M. A.
Levy
R.
Gratz
K. L.
Lejuez
C. W.
Understanding the heterogeneity of BPD symptoms through latent class analysis: initial results and clinical correlates among inner-city substance users
Psychological Assessment
 , 
2010
, vol. 
22
 (pg. 
233
-
245
)
Brevers
D.
Cleeremans
A.
Verbruggen
F.
Bechara
A.
Kornreich
C.
Verbanck
P.
, et al.  . 
Impulsive action but not impulsive choice determines problem gambling severity
PLoS One
 , 
2012
, vol. 
7
 pg. 
e50647
 
Brickenkamp
R.
Aufmerksamkeits-Belastungs-Test: Manual. Aufl 9
 , 
2002
Göttingen
Hogrefe
Chen
K. W.
Banducci
A. N.
Guller
L.
Macatee
R. J.
Lavelle
A.
Daughters
S. B.
, et al.  . 
An examination of psychiatric comorbidities as a function of gender and substance type within an inpatient substance use treatment program
Drug and Alcohol Dependence
 , 
2011
, vol. 
118
 (pg. 
92
-
99
)
Conners
C. K.
Clinical use of rating scale in diagnosis and treatment of attention-deficit hyperactivity disorder
Pediatric Clinics of North America
 , 
1999
, vol. 
46
 (pg. 
857
-
870
)
Cyders
M. A.
Coskunpinar
A.
Measurement of constructs using self-report and behavioral lab tasks: Is there overlap in nomothetic span and construct representation for impulsivity?
Clinical Psychology Review
 , 
2011
, vol. 
31
 (pg. 
965
-
982
)
Delis
D. C.
Kaplan
E.
Kramer
J. H.
Delis–Kaplan executive function system (D-KEFS)
 , 
2001
San Antonio
The Psychological Corporation
Evenden
J. L.
Varieties of impulsivity
Psychopharmacology (Berl)
 , 
1999
, vol. 
146
 (pg. 
348
-
361
)
First
M. B.
Spitzer
R. L.
Gibbon
M.
Williams
J. B.
Structured Clinical Interview for DSM-IV Axis I disorders (SCID I)
 , 
1997
New York
Biometric Research Department
Friedmann
P. D.
Hendrickson
J. C.
Gerstein
D. R.
Zhang
Z.
The effect of matching comprehensive services to patients' needs on drug use improvement in addiction treatment
Addiction
 , 
2004
, vol. 
99
 (pg. 
962
-
972
)
Giddens
J. L.
Stefanovics
E.
Pilver
C. E.
Desai
R.
Potenza
M. N.
Pathological gambling severity and co-occurring psychiatric disorders in individuals with and without anxiety disorders in a nationally representative sample
Psychiatry Research
 , 
2012
, vol. 
199
 (pg. 
58
-
64
)
Goudriaan
A. E.
Oosterlaan
J.
de Beurs
E.
van den Brink
W.
Neurocognitive functions in pathological gambling: A comparison with alcohol dependence, Tourette syndrome and normal controls
Addiction
 , 
2006
, vol. 
101
 (pg. 
534
-
547
)
Kaufman
A. S.
Kaufman
N. L.
Kaufman Brief Intelligence Test. 1
 , 
1990
Circle Pines, MN
American Guidance Service
Kuyken
W.
Kurzer
N.
DeRubeis
R. J.
Beck
A. T.
Brown
G. K.
Response to cognitive therapy in depression: The role of maladaptive beliefs and personality disorders
Journal of Consulting and Clinical Psychology
 , 
2001
, vol. 
69
 (pg. 
560
-
566
)
Lawrence
A. J.
Luty
J.
Bogdan
N. A.
Sahakian
B. J.
Clark
L.
Problem gamblers share deficits in impulsive decision-making with alcohol-dependent individuals
Addiction
 , 
2009
, vol. 
104
 (pg. 
1006
-
1015
)
Ledgerwood
D. M.
Petry
N. M.
Subtyping pathological gamblers based on impulsivity, depression, and anxiety
Psychology of Addictive Behaviors
 , 
2010
, vol. 
24
 (pg. 
680
-
688
)
Lobo
A.
Pérez-Echevarria
M. J.
Artal
J.
Validity of the scaled version of the General Health Questionnaire (GHQ-28) in a Spanish population
Psychological Medicine
 , 
1986
, vol. 
16
 (pg. 
135
-
140
)
Loranger
A. W.
Sartorius
N.
Anfreoli
A.
Berger
P.
Buchheim
P.
Channabasavanna
S. N.
The International Personality Disorder Examination. The World Health Organization/Alcohol, Drug Abuse, and Mental Health Administration international pilot study of personality disorders
Archives of General Psychiatry
 , 
1994
, vol. 
51
 (pg. 
215
-
224
)
López-Ibor
J. J.
I.P.D.E. Examen internacional de los trastornos de la personalidad
 , 
1999
Madrid
Organización Mundial de la Salud, Meditor
Lozano
O. M.
Rojas
A. J.
Pérez
C.
Development of a specific Health-Related Quality of Life Test in Drug Abusers using the Rasch Rating Scale Model
European Addiction Research
 , 
2009
, vol. 
15
 (pg. 
63
-
70
)
Magidson
J.
Vermunt
J. K.
Latent class models for clustering: A comparison with K-means
Canadian Journal of Marketing Research
 , 
2002
, vol. 
20
 (pg. 
36
-
43
)
Miller
P. G.
Miller
W. R.
What should we be aiming for in the treatment of addiction?
Addiction
 , 
2009
, vol. 
104
 (pg. 
685
-
686
)
Moeller
F. G.
Barratt
E. S.
Dougherty
D. M.
Schmitz
J. M.
Swann
A. C.
Psychiatric aspects of impulsivity
American Journal of Psychiatry
 , 
2001
, vol. 
158
 (pg. 
1783
-
1793
)
Muthen
B.
Should substance use disorders be considered as categorical or dimensional?
Addiction
 , 
2006
, vol. 
101
 (pg. 
6
-
16
)
Perales
J. C.
Verdejo-García
A.
Moya
M.
Lozano
O. M.
Perez-Garcia
M.
Bright and dark sides of impulsivity: Performance of women with high and low trait impulsivity on neuropsychological tasks
Journal of Clinical and Experimental Neuropsychology
 , 
2009
, vol. 
31
 (pg. 
927
-
944
)
Petry
N. M.
Substance abuse, pathological gambling, and impulsiveness
Drug and Alcohol Dependence
 , 
2001
, vol. 
63
 (pg. 
29
-
38
)
Reboussin
B.
Young Song
E.
Shrestha
A.
Lohman
K.
Wolfson
M.
A latent class analysis of underage problem drinking: Evidence from a community sample of 16–20 year olds
Drug and Alcohol Dependence
 , 
2006
, vol. 
83
 (pg. 
199
-
209
)
Robbins
T. W.
Gillan
C. M.
Smith
D. G.
de Wit
S.
Ersche
K. D.
Neurocognitive endophenotypes of impulsivity and compulsivity: Towards dimensional psychiatry
Trends in Cognitive Sciences
 , 
2012
, vol. 
16
 (pg. 
81
-
91
)
Steel
Z.
Blaszczynski
A.
Impulsivity, personality disorders and pathological gambling severity
Addiction
 , 
1998
, vol. 
93
 (pg. 
895
-
905
)
Swanson
S. A.
Lindenberg
K.
Bauer
S.
Crosby
R.
A Monte Carlo investigation of factors influencing Latent Class Analysis: An application to eating disorder research
International Journal of Eating Disorders
 , 
2011
, vol. 
45
 (pg. 
677
-
684
)
Tiffany
S. T.
Friedman
L.
Greenfield
S. F.
Hasin
D. S.
Jackson
R.
Beyond drug use: A systematic consideration of other outcomes in evaluations of treatments for substance use disorders
Addiction
 , 
2012
, vol. 
107
 (pg. 
709
-
718
)
Torres
A.
Catena
A.
Megías
A.
Maldonado
A.
Cándido
A.
Verdejo-García
A.
, et al.  . 
Emotional and non-emotional pathways to impulsive behavior and addiction
Frontiers in Human Neuroscience
 , 
2013
, vol. 
7
 pg. 
43
 
Uebersax
J.
Collins
L. M.
Seitz
L. A.
Latent class analysis of substance use patterns
Advances in data analysis for prevention intervention research. NIDA research monograph, No. 142
 , 
1994
Rockville, MD
National Institute on Drug Abuse
Vassileva
J.
Gonzalez
R.
Bechara
A.
Martin
E. M.
Are all drug addicts impulsive? Effects of antisociality and extent of multidrug use on cognitive and motor impulsivity
Addictive Behaviors
 , 
2007
, vol. 
32
 (pg. 
3071
-
3076
)
Verdejo-García
A.
Bechara
A.
Recknor
E. C.
Pérez-García
M.
Negative emotion-driven impulsivity predicts substance dependence problems
Drug and Alcohol Dependence
 , 
2007
, vol. 
91
 (pg. 
213
-
219
)
Verdejo-García
A.
Lawrence
A. J.
Clark
L.
Impulsivity as a vulnerability marker for substance-use disorders: Review of findings from high-risk research, problem gamblers and genetic association studies
Neuroscience and Biobehavioral Reviews
 , 
2008
, vol. 
32
 (pg. 
777
-
810
)
Verdejo-García
A. J.
López-Torrecillas
F.
Aguilar de Arcos
F.
Pérez-García
M.
Differential effects of MDMA, cocaine, and cannabis use severity on distinctive components of the executive functions in polysubstance users: A multiple regression analysis
Addictive Behaviors
 , 
2005
, vol. 
30
 (pg. 
89
-
101
)
Verdejo-García
A.
Lozano
O.
Moya
M.
Alcázar
M. A.
Pérez-García
M.
Psychometric properties of a Spanish version of the UPPS-P impulsive behavior scale: Reliability, validity and association with trait and cognitive impulsivity
Journal of Personality Assessment
 , 
2010
, vol. 
92
 (pg. 
70
-
77
)
Weiss
R. D.
Griffin
M. L.
Hufford
C.
Muenz
L. R.
Najavits
L. M.
Jansson
S. B.
, et al.  . 
Early prediction of initiation of abstinence from cocaine. Use of a craving questionnaire
American Journal on Addictions
 , 
1997
, vol. 
6
 (pg. 
224
-
231
)
Whiteside
S. P.
Lynam
D. R.
The five factor model and impulsivity: Using a structural model of personality to understand impulsivity
Personality and Individual Differences
 , 
2001
, vol. 
30
 (pg. 
669
-
689
)
Zubaran
C.
Rishi
S.
Emerson
J.
Zolfaghari
E.
Foresti
K.
Lozano
O. M.
Validation of the English version of the Health-Related Quality of Life for Drug Abuser (HRQoLDA) Test
European Addiction Research
 , 
2012
, vol. 
18
 (pg. 
220
-
227
)