Improvement of Emotional Empathy and Cluster B Personality Disorder Symptoms Associated With Decreased Cocaine Use Severity

Aims: Chronic cocaine users display impaired social cognitive abilities, reduced prosocial behavior, and pronounced cluster B personality disorder (PD) symptoms all contributing to their social dysfunctions in daily life. These social dysfunctions have been proposed as a major factor for maintenance and relapse of stimulant use disorders in general. However, little is known about the reversibility of social cognitive deficits and socially problematic personality facets when stimulant use is reduced or ceased. Therefore, we examined the relation between changing intensity of cocaine use and the development of sociocognitive functioning and cluster B PD symptomatology over the course of 1 year. Methods: Social cognition, social decision-making, and cluster B PD symptoms were assessed in 38 cocaine users (19 with increased and 19 with decreased use) and 48 stimulant-naive healthy controls at baseline and at 1-year follow-up. Cocaine use severity was objectively determined by quantitative 6-month hair analyses. The categorization of the two cocaine user groups was based on a combination of absolute (± 0.5 ng/mg) and relative (± 10%) changes in the cocaine hair concentration between baseline and the 1-year follow-up. Social cognition was assessed using the Multifaceted Empathy Test (MET) and the Movie for the Assessment of Social Cognition (MASC). A combined Distribution/Dictator Game was applied for assessing social decision-making. Cluster B PD symptoms were measured by a Structured Clinical Interview for DSM-IV Axis II Disorders (SCID-II) PD questionnaire according to Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV). Results: Increased cocaine use was linked to worsened empathy, while decreased cocaine use went along with improved emotional empathy. Moreover, whereas decreased cocaine use was associated with reduced severity of self-reported cluster B PD symptoms, these symptoms remained largely stable in increasers. In contrast to a significant reduction of prosocial behavior at baseline in the combined cocaine user group, specifically decreasers were not statistically distinguishable from controls at the follow-up. Conclusions: Sociocognitive deficits and cluster B PD symptoms of chronic cocaine users are adaptable over time as they covary with the increase or decrease in cocaine use. Hence, abstinence orientation and training of social cognition and interaction might improve social functioning, and should therefore be important therapeutic elements in cocaine addiction treatment.

Aims: Chronic cocaine users display impaired social cognitive abilities, reduced prosocial behavior, and pronounced cluster B personality disorder (PD) symptoms all contributing to their social dysfunctions in daily life. These social dysfunctions have been proposed as a major factor for maintenance and relapse of stimulant use disorders in general. However, little is known about the reversibility of social cognitive deficits and socially problematic personality facets when stimulant use is reduced or ceased. Therefore, we examined the relation between changing intensity of cocaine use and the development of sociocognitive functioning and cluster B PD symptomatology over the course of 1 year.
Methods: Social cognition, social decision-making, and cluster B PD symptoms were assessed in 38 cocaine users (19 with increased and 19 with decreased use) and 48 stimulant-naive healthy controls at baseline and at 1-year follow-up. Cocaine use severity was objectively determined by quantitative 6-month hair analyses. The categorization of the two cocaine user groups was based on a combination of absolute (± 0.5 ng/ mg) and relative (± 10%) changes in the cocaine hair concentration between baseline and the 1-year follow-up. Social cognition was assessed using the Multifaceted Empathy Test (MET) and the Movie for the Assessment of Social Cognition (MASC). A combined Distribution/Dictator Game was applied for assessing social decision-making. Cluster B PD symptoms were measured by a Structured Clinical Interview for DSM-IV Axis II Disorders (SCID-II) PD questionnaire according to Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV).
Results: Increased cocaine use was linked to worsened empathy, while decreased cocaine use went along with improved emotional empathy. Moreover, whereas decreased cocaine use was associated with reduced severity of self-reported cluster B PD symptoms, these symptoms remained largely stable in increasers. In contrast to a significant reduction of prosocial behavior at baseline in the combined cocaine user group, specifically decreasers were not statistically distinguishable from controls at the follow-up.

INTRODUCTION
Neurocognitive deficits such as impaired attention, memory, and executive functions related to chronic cocaine use are well documented (1-3) and a risk factor for poor treatment outcomes (4,5). While some studies investigated the linkage between these neurocognitive deficits and cocaine abstinence (6), only one study yet investigated the longitudinal relationship between cognitive impairments and changing cocaine use (7). In sum, these studies indicate that basal cognitive deficits in cocaine users seem to be largely drug-induced, remain stable during the first weeks of abstinence but likely improve after some months (8).
While nonsocial cognitive functions have been studied well during the last two decades, the systematic assessment of sociocognitive functioning in cocaine users has only recently emerged. Per definition, the concept of social cognition comprises not only abilities enabling the dynamic interaction with our social environments and include emotional and mental perspective-taking functions such as emotion recognition, emotional empathy (EE), and Theory-of-Mind, but also interactive abilities such as social decision-making (SDM), moral behavior, and social network behavior (9,10). As daily-life social functioning strongly depends on intact social cognition and as the deteriorative impact of sociocognitive impairments on development, progress, and prognosis on other psychiatric disorders such as schizophrenia is well known (11), a close relationship between sociocognitive functioning and the origin and course of stimulant use disorders has been proposed (12)(13)(14)(15). Accordingly, we previously demonstrated smaller social networks (16), reduced EE (16), altered SDM (17), stronger detachment from social norms (14), and impaired emotion recognition from voices (18) in recreational and dependent cocaine user groups. Moreover, dependent cocaine users made more errors than controls in a videobased Theory-of-Mind task, with recreational cocaine users performing intermediate between the two groups (16). Finally, cocaine users show also blunted neuronal responses to implicit and explicit forms of social reward (19,20). Notably, all these studies were implemented with a cross-sectional design, but no study has investigated the longitudinal development of sociocognitive functioning so far. Thus, it is unclear if sociocognitive impairments are predisposed or drug-induced and if they are reversible upon prolonged abstinence or reduction of drug use.
As social cognition is the sum of those processes that allow individuals to interact in interpersonal contexts (21), disturbed sociocognitive functioning leads to aberrant social behavior and, in excessive forms, to deviant personality characteristics and impaired interpersonal functioning (22,23). Notably, cocaine-addicted individuals show an increased risk for concurrent cluster B personality disorders (PDs), mainly of the antisocial and borderline types (24,25). A cluster B PD comorbidity is largely influential for cocaine addiction severity and treatment outcomes including pronounced executive function deficits (26), more intense cocaine intake, lower rates of treatment applications, and decreased probability of cocaine addiction remission (27,28). Additionally, it was demonstrated that impulsivity and gambling decision-making, which are both closely related to cluster B PD pathologies (22,29), covary with changes in the intensity of cocaine use over 1 year (30). Nonetheless, the longitudinal relation between cocaine use intensity and cluster B PD symptomatology has also not been investigated to date.
In sum, only little is known about the temporal dynamics between cocaine use intensity and sociocognitive functioning. Hence, in order to investigate whether the described sociocognitive impairments and comorbid cluster B PD symptomatology in chronic cocaine users are modulated by the increase or decrease in cocaine abuse, we performed a longitudinal study with an interval of 1 year. Thereby, we compared 48 psychostimulantnaive controls with 19 cocaine users with decreased use (decreasers) and 19 cocaine users with increased use (increasers) after a 1-year interval. To objectively assess the severity and change in cocaine use and to control for co-use of other drugs, we performed quantitative hair and urine toxicology analyses at baseline and follow-up. Considering our previous results from the present sample that changes in basal cognitive functions and impulsivity clearly covary with cocaine use intensity over time (7,30), we hypothesized that escalating cocaine use is also associated with aggravation of sociocognitive impairments and more cluster B PD symptoms within 1 year. Vice versa, we also expected that reduced cocaine use is linked to a reduction of sociocognitive deficits and cluster B PD symptomatology. To test these hypotheses, we expect significant time × group interactions specifically between decreasers and increasers. Given that at baseline cocaine users displayed significant alterations in EE, social network size, prosocial behavior in money distribution games, Theory-of-Mind, and cluster B PD symptoms (14,16,17), the longitudinal analysis was focused solely on these parameters. 3 April 2019 | Volume 10 | Article 213 Frontiers in Psychiatry | www.frontiersin.org

Participants
From a baseline sample of 234 participants (96 healthy stimulant-naive controls, 138 cocaine users) (3,16,17), 48 healthy stimulant-naive controls and 38 chronic cocaine users were included in the present longitudinal study. This subsample has been published twice previously but with different outcome measures (7,30). From the baseline sample, 102 participants could not be measured at the follow-up because of unavailability (i.e., not responding to the invitation, loss of interest, lack of time, death), 27 participants had to be excluded from the final analyses as hair analyses revealed drug use not allowed by our exclusion criteria (e.g., polysubstance use, change in drug preferences), and 19 cocaine users did not meet our cocaine use criteria [see also the cocaine user group assignment below; for further recruitment and selection details, please see Ref. (7)].
At baseline, general exclusion criteria were clinically significant somatic diseases, neurological disorders, head injuries, family history of schizophrenia/obsessive-compulsive disorder/bipolar disorder, or any medication affecting the central nervous system. Additional exclusion criteria for controls were Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV) axis I psychiatric disorders (excluding nicotine dependence) and regular illegal drug use (>15 lifetime occasions, except for recreational cannabis use). Additional exclusion criteria for cocaine users were a history of heroin use, polysubstance use, or DSM-IV axis I psychiatric disorders (except for cocaine, nicotine, cannabis, and alcohol abuse/dependence, attention deficit hyperactivity disorder, and a previous episode of an affective disorder). At baseline, inclusion criteria for cocaine users were cocaine use of >0.5 g per month, cocaine as primary drug, and an abstinence duration of <6 months. Participants were asked to abstain from illegal substances for at least 72 h and from alcohol for 24 h before the test sessions. Compliance with these instructions was controlled by urine screenings (semiquantitative enzyme multiplied immunoassay method). The study was approved by the Cantonal Ethics Committee of Zurich. All participants provided written informed-consent statements and were compensated for their participation.

Procedure
At baseline, self-reported drug use was assessed with a structured and standardized Interview for Psychotropic Drug Consumption (32), attention deficit hyperactivity disorder (ADHD) symptoms were assessed with the ADHD Self-Rating Scale (ADHD-SR) (33), and the Structured Clinical Interview for DSM-IV axis I disorders (SCID-I) (34) was carried out by trained psychologists.
The test battery was assessed at baseline and follow-up and included the Multifaceted Empathy Test (MET) (35) assessing EE, the Movie for the Assessment of Social Cognition (MASC) (36) for the measurement of Theory-of-Mind, a Distribution/Dictator Game (37,38) for the determination of prosocial behavior, the Social Network Questionnaire (SNQ) (39) measuring the social network size, and the SCID-II questionnaire (40) in order to ascertain cluster B PD symptoms. More detailed test descriptions published already in our previous work (16,17) are given in Methods S1.

Statistical Analysis
Effect sizes and power analyses were calculated with G*Power 3.1 (41). As our previous analyses showed an effect size of p η 2 = 0.12 (Cohen's f = 0.37) and a power of 99% for the significant interaction in the domain of working memory between decreasers and increasers (two groups, p < .05, two measurements) for the present sample (7), we assumed a more conservative effect size of p η 2 = 0.06 (f = 0.25) and calculated a still acceptable power of 86% for the detection of significant interactions in sociocognitive functions in the present sample.
In order to reduce data quantity [see also Ref. (17)], we computed an SDM composite score that was derived by averaging z-transformed payoffs for the other player in the Distribution and Dictator Game (payoffs B) according to the means and standard deviations of the control group. Because of a strong correlation of the explicit and implicit EE scores from the MET in the total sample (r = 0.86, p < .001), we further integrated both parameters by adding them up into a single MET EE score. The SCID-II Cluster B symptom score was calculated by summing up the dimensional values from histrionic, narcissistic, borderline, and antisocial PD.
Group differences in demographic data and drug use patterns were analyzed by means of Pearson's chi-squared tests, analyses of variance (ANOVA), or independent Student's t-tests. For the longitudinal analysis and in order to investigate group differences over all groups, we performed a multiple linear regression (forced entry) with the test score change values (Δ = t2 − t1) as dependent variables and four preselected independent variables: age, sex, ADHS-SR score, and dummy-coded (zero/one) group contrasts. The two demographic variables were included because previous findings suggest a linkage between advancing age and fairness in stimulant users (17) and due to known gender effects in social cognition/functioning (42,43). Moreover, because ADHD has previously been linked to cognitive and sociocognitive 4 April 2019 | Volume 10 | Article 213 Frontiers in Psychiatry | www.frontiersin.org performance in cocaine users (3,16,44), this variable was further included as a predictor into the regression model. To compare the groups, cocaine increasers acted as the reference group. To further analyze test score changes within the single groups (value t2 vs. value t1), we applied dependent Student's t-tests (t dep ). To compare the effect of changing cocaine use, we applied independent Student's t-tests (t ind ) between controls and a combined cocaine user sample (CCU = increasers + decreasers) at baseline as well as between controls, cocaine increasers, and cocaine decreasers at the follow-up. Notably, at baseline, cocaine increasers and decreasers showed comparable baseline values in all reported test parameters (MET, MASC, SDM, SNQ, SCID-II Cluster B) differing only with very small effect sizes (t ind (32-35) = 0.05-0.34, p = .99-.74, d = 0.00-0.11). In the test parameter analysis, frequency data were analyzed by the Fisher-Freeman-Halton Exact Test (FET) (45). To test for test-retest effects, we applied the Pearson product-moment correlation analyses. The confirmatory statistical comparisons were carried out on a significance level of p < .05 (two-tailed).

RESULTS
Demographic characteristics and drug use: As shown before (7,30), the three experimental groups did not significantly differ regarding age, sex distribution, verbal IQ, years of education, length of study interval ( Table 1), and socioeconomic status (Table S1). Still, cocaine-using groups showed significantly higher BDI and ADHD-SR sum scores than controls at baseline (7,30). Whereas at baseline both cocaine user groups showed comparable cocaine use severity, the cocaine total hair concentrations for increasers (~3-fold increase) and decreasers (reduction by the factor 3.5) were significantly different at follow-up. Moreover, hair data revealed a clear preference for cocaine use compared to other illegal drugs. Finally, in both user groups, 8 of 19 participants sought psychiatric or psychological treatment during the study interval. The other cocaine users did not report any related treatment between baseline and follow-up.
Emotional empathy: The introduced predictors explained a significant amount of variance of the EE change scores in the multiple regression analysis [F(5,80) = 2.68, p < .05, R 2 = .14; Table 2]. The strongest predictors were sex (β = −0.25, p < .05) and the group contrast cocaine increasers vs. decreasers (β = −0.25, p < .05; Figure 1). Post hoc analyses showed that controls and the combined cocaine user (CCU) group showed a nonsignificant difference in EE with a small to moderate effect size at baseline [t ind (84) Figure 2).
Social interaction: The multiple regression model was also not able to predict the SDM composite change score ( Table 2). From the phenomenological perspective, controls and increasers acted less prosocial (giving less money to the opponent), while decreasers remained stable but, with that, came closer to the controls (Figure 3). Exploratory post hoc analyses confirmed that controls and CCU significantly differed at baseline [t ind (65) = 2.51, p < .05, d = 0.56]. At follow-up, controls and increasers still display a moderate group difference [t ind (65) = 1.92, p = .06, d = 0.50], whereas the group difference between controls and decreasers was reduced to a small effect size [t ind (64) = 0.98, p = .33, d = 0.26]. In addition, we analyzed behavioral changes between baseline and follow-up (more prosocial decisions, more self-serving decisions, similar decision) only in cocaine users and found that about two-thirds of the increasers (58% = 11/19) but only one-third of the decreasers (33% = 6/18) showed more selfserving decisions at follow-up (p = .40; FET; Figure S1).
Social network size: Regarding the SNQ total network size, the multiple regression model could again not substantially predict the change scores (Table 2, Figure 4). Interestingly, during the 1-year interval, all three groups reported a substantial and moderate social network reduction of about 2.  Table 2]. This change score was best predicted by the ADHS-SR score (β = −0.32, p < .01) and the group contrasts cocaine increasers vs. decreasers (β = 0.34, p < .01) and cocaine increasers vs. controls (β = 0.31, p < .01). Importantly, the CCU group showed at baseline significantly more cluster B PD symptoms than the controls [t ind (81) = 4.40, p < .001, d = 0.96;  Figure S2).
Remarkably, the interaction effect on cluster B PD symptoms was mainly driven by changes in the narcissistic and borderline subscores and less by the histrionic and surprisingly also not by the antisocial subscore (see Figure S3a-     Change in alcohol use: As not only cocaine but also alcohol intake was increased in increasers (see Table 1), the change in alcohol consumption was considered in additional multiple regression models. However, alcohol change was not significant in any of the main regression models (p-values ranged from .222 to .659) shown in Table 2, while the interaction effects and also the explained variances remained stable, indicating that changes in alcohol consumption have not impacted our main results.    Test-retest reliability: In the total sample, all dependent variables displayed acceptable to good test-retest reliabilities ( Table 3). Interestingly, in the SDM paradigm, controls and CCU differed significantly in their test-retest reliability (z = −3.25; p < .001): While in controls the SDM score showed hardly acceptable reliability (r = 0.48; p < .001), it was good in the cocaine users (r = 0.85; p < .001).

DISCUSSION
The present longitudinal study investigated the change of social cognition, social interaction, and socially relevant cluster B PD symptoms in healthy controls and relatively pure and non-helpseeking chronic cocaine users who clearly increased or decreased their cocaine consumption during a 1-year study interval. The most striking findings were that i) improved EE correlated with decreased cocaine consumption, whereas increased cocaine use severity was linked to less EE; and ii) cluster B PD symptom burden was lowered in decreasers, whereas increasers showed stable severity in these symptoms. Additionally, during the study interval, we found an approximation between controls and decreasers regarding their prosocial behavior, while the large gap between increasers and controls remained. Moreover, neither the Theory-of-Mind Task (MASC) nor the social network size showed interactions with changing cocaine use, indicating that mental perspective-taking (sometimes also interpreted as cognitive empathy) and the number of social contacts in the last months were not affected by changing drug use during the study interval.
Importantly, the present analysis of smaller (longitudinal) subgroups from our larger cross-sectional ZuCo 2 St sample published previously (14,16,17) still showed significantly reduced prosocial behavior, a smaller social network, and strongly elevated cluster B personality symptoms in the total group of cocaine users at baseline, indicating that these indicators of social functioning were robustly altered in this population. The EE score of the MET showed only a statistical trend between cocaine users and controls at baseline, but the present effect size (d = 0.39) was in the range of the previously   reported effect sizes of the larger cross-sectional sample of recreational and dependent cocaine users (d = 0.39-0.64), suggesting rather a deficiency of power than a lack of reliability. This assumption is further supported by the fact that the MET EE score showed good test-retest reliability scores. Moreover, the MASC did not show any baseline group differences in the present subsample of cocaine users underscoring our previous conclusion that mainly very severe cocaine users with a putative ADHD comorbidity show disturbances in this task (16,44).
While sociocognitive functions represent basic abilities in perspective-taking and interaction, more conventional psychopathology is aiming at the identification and quantification of symptoms in psychiatric disorders (51). As such, the research on the relationship between PDs and cocaine use is of special interest, as the differentiation between predispositions vs. drug-induced effects merges with the question if these pathologies are reversible or not. In our longitudinal investigation, we found that decreasers of cocaine consumption also significantly improved in cluster B PD symptoms during 1 year, whereas the increasers showed a stable PD symptom burden. This is insofar interesting as in both user groups 8 of 19 participants sought psychiatric treatment in the interval, but only decreasers improved in some social functions and socially relevant PD symptoms.
In general, PDs are defined as typical constellations of impaired subjective and behavioral traits that result in suffering of the affected individual and/or society (52). These personality traits are regarded as relatively stable across time and consistent across situations (diagnostically mandatory) (53-56). Moreover, cluster B PDs show a higher stability over 12 to 18 years than the other clusters (57). However, studies also found considerable variability of PD symptoms across individuals over time (58,59), questioning the trait-like character of the disorder. An early study showed changes in PD symptoms related to treatment in substance-dependent patients (60). Interestingly, clusters A and C profited most, while cluster B changes were only observed in patients with borderline PD. In patients with cocaine addiction, cluster B PDs are the most frequent and these patients have the most severe courses of illness including worst treatment outcomes (24, 26-28, 61, 62). Therefore, cluster B PD symptoms are likely personality features that increase the risk for cocaine use and the development of an addiction. However, as seen in the present study, cluster B PD symptom load is nevertheless variable and reduction of consumption leads to a substantial improvement in these symptoms. Consequently, a reduction of cluster B PD symptom burden again increases likelihood of successful treatment, offering the patient an opportunity to leave the vicious circle of addiction.
The suggested consumption-dependent variability of social behavior as well as cluster B symptoms are well in line with our previous analyses from this sample that not only basal cognitive functions such as working memory but also self-reported impulsivity improve with a strong reduction of cocaine use, while they are worsened with increased cocaine consumption (7,30). The present data and the previous analyses from this sample are also in accordance with our recent results from an independent longitudinal investigation showing that decreased cortical thickness (CT) of several regions within the prefrontal cortex of cocaine users can improve after a strong reduction of cocaine use, while sustained use went along with a further decrease in prefrontal CT during the study interval (63). Importantly, the cortical changes were correlated with cognitive changes, i.e., improved CT as associated with enhanced sustained attention (63). Thus, the overall pattern of change shown by longitudinal data supports our assumption that sociocognitive impairments of cocaine users are at least in part drug-induced and that neuroplastic changes in brain regions and neurotransmitter systems involved in social cognition, social interaction, and social reward processing contribute to a further decrease in social contact and social support leading to an increase in social isolation, aggression, and depressive symptoms. This ends in a further reduction of social reward resources, ongoing social withdrawal, and the establishment of cocaine as the main source of reward resulting in the maintenance of stimulant use and recurrent relapses (15).
While EE is more a perceptive social cognition ability, social decision-making (here assessed with a combined Distribution/ Dictator Game) is a form of socially interactive behavior. In our previous cross-sectional analysis sample, cocaine users cared more about efficiency than about fairness compared to healthy controls at baseline (17). This was previously interpreted as predisposition of stimulant use (15), as such fairness preferences and severity of cocaine use were not correlated (17). Intriguingly, utilitarian and opportunistic attitudes assessed with the Machiavellianism Questionnaire (MACH-IV) were also increased in cocaine users compared to controls and were shown to be stable and independent of changing cocaine use (14). However, our data indicated a shift toward improved prosocial behavior in cocaine decreasers indicating space for enhancement potential by treatment. Conclusively, SDM deficits in cocaine users likely have both a trait and a state component, and it might be worse to specifically target the state component in therapy in order to improve the treatment outcome.

Limitations
When interpreting the present results, some limitations of our study have to be considered: i) The total sample size is moderate for a longitudinal analysis. Moreover, the test-retest reliabilities of the applied social cognition tasks and questionnaire have a broad range (in controls: r = 0.48-0.75; in cocaine users: r = 0.60-0.85; in the total sample: r = 0.63-0.80). As a consequence of both, the shown interaction effects are not very strong (in terms of p-values). However, to our knowledge, these are the only existing longitudinal samples of chronic cocaine users with objectively verified increasing and decreasing cocaine use (by hair testing). Moreover, the included individuals were preferably pure cocaine users with little axis-I psychiatric comorbidities. We therefore think that the carefully selected and homogeneous sample has nonetheless sufficient explanatory power. ii) In the context of our hypotheses, we attribute the changes in behavior to the changes in cocaine consumption. However, we cannot rule out if other changes in the lives of 10 April 2019 | Volume 10 | Article 213 Frontiers in Psychiatry | www.frontiersin.org our cocaine users (e.g., positive or negative changes in their social environment) not assessed by our test battery may have impacted both drug use and social functioning. Future studies should therefore assess more information on the social life of cocaine users beyond simple parameters such as social network size (e.g., social media use).

Conclusions
The aim of this longitudinal study was to investigate whether cocaine use is associated with permanent or reversible alterations of social cognition and interaction as well as cluster B PD symptoms. We found that specific social dysfunctions and PD symptoms are variable over time as they seem to depend on variations in cocaine use. Thus, strong reduction of cocaine use within only 1 year seems to positively affect social dysfunctions that are assumed to be crucial factor in the maintenance of stimulant addiction (15,64). From our perspective, the shown positive effects of reduced cocaine use clearly favor abstinenceorientated treatment approaches of cocaine addiction. Furthermore, having the strong impact of social cognitive abilities as well as prosocial behavior and attitudes on the patienttherapist relationships in mind (15), future developments in the psychotherapy of cocaine addictions should consider trainings specifically of social skills and cognitions in order to improve treatment outcome.

ETHICS STATEMENT
The study was approved by the Cantonal Ethics Committee of Zurich, Switzerland.

AUTHOR CONTRIBUTIONS
BQ and MV had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. CE and BQ contributed to the study concept and design. CE, MV, OB, KP, LH, MB, and BQ contributed to the acquisition, analysis, or interpretation of data. OB, CE, MV, and BQ contributed to the drafting of the manuscript. All authors contributed to the critical revision of the manuscript for important intellectual content. MV and BQ conducted the statistical analysis. ES and BQ obtained funding. KP, LH, MV, and ES contributed to the administrative, technical, or material support. ES and BQ were in charge of supervision.

FUNDING
The study was supported by grants from the Swiss National Science Foundation (SNSF; grant Nos. PP00P1-123516/1 and PP00P1-146326/1) and the Olga Mayenfisch Foundation. The funders of the study did not influence the design and conduct of the study; the collection, management, analysis, and interpretation of the data; the preparation, review, and approval of the manuscript; and the decision to submit the manuscript for publication.

ACKNOWLEDGMENTS
This article is in tribute to our esteemed friend and colleague Prof. Christoph Eisenegger, who passed away so suddenly and unexpectedly during the time we worked on this article. We are grateful to Alex Gamma for the critical discussion of the present data. Moreover, we thank Daniela Jenni and Kathrin Küpeli for their excellent support.