Increased attachment security is related to early therapy drop-out in substance use disorders

Objectives Previous research work suggests a positive association between secure attachment and increased therapy adherence (TA) in different patient groups. However, there is still a strong need for research focusing on the influence of attachment on TA in substance use disorder (SUD) treatment. Hence, this study attempts to investigate the predictive value of different attachment patterns concerning TA in SUD inpatients. Results 122 (34 female) SUD inpatients completed the Attachment Style Questionnaire (ASQ) during the entry phase of therapeutic community treatment. After 6 weeks, subjects who remained in therapy (n = 47) completed the ASQ for a second time. In line with the literature, agglomerative Cluster Analysis suggested a two-cluster solution (Cluster I: increased secure attachment pattern; Cluster II: increased insecure attachment pattern). Notably, inpatients in Cluster I were more likely to drop out of treatment within the first 6 weeks (p < .001). Furthermore, subjects showed less “Confidence in Self and Others” (p < .05) after 6 weeks of treatment. Our findings indicate a negative predictive value of increased attachment security for TA in SUD inpatients. This finding probably mirrors a more realistic kind of self-assessment. More generally, the importance of considering attachment styles in SUD treatment is underlined.


Introduction
Substance use disorders (SUD) have been linked with insecure attachment in several studies [1]. This wellestablished connection is explained by several characteristics observed in individuals displaying an insecure attachment pattern: dysfunctional attributes such as the increased susceptibility and physiological response to stress, as well as the underuse of social support (in conjunction with inadequate help seeking strategies), lead to a predisposition for the frequent use of psychoactive substances as a futile attempt to regulate affects [2][3][4].
Research focusing on the relationship between attachment styles and treatment adherence (TA) is relatively sparse, these studies have been conducted mostly on somatic patients treated for diabetes and lupus [5,6]: the results indicate that insecure attachment, especially dismissing and avoidant styles, correspond to worse treatment outcome and compliance. Insecure attachment might lead to increased difficulties in the relationship between patients and caregivers, due to patients' inability to accept help from others. A study on the role of attachment in TA in SUD inpatients has found anxiouspreoccupied attachment to be a predictor for treatment retention [7]. Moreover, several studies also observed a negative association between TA and parameters such as comorbid personality disorder, cognitive deficits and age [8,9].
In previous research, we observed a higher amount of insecure attachment in different SUD patient groups in comparison to healthy controls [10]. As an extension to this study [10], we intended to examine the role of attachment styles for TA by investigating this group of SUD inpatients for a second time after 6 weeks. Based

Participants and procedure
The initial sample consisted of 122 inpatients diagnosed with AUD (F10.2x) (n = 66) or poly drug use disorder (PUD) (F19.2x) (n = 57) by a licensed psychiatrist according to ICD 10 [11]. As our previous research indicated that patients diagnosed with AUD or PUD do not differ in their attachment style and their personality organization [10], we decided not to further differentiate between AUD and PUD inpatients in the present study. After 6 weeks, the 47 participants remaining in treatment were investigated for a second time. All subjects were abstinent while in treatment. Further inclusion criteria were: no current psychotic symptoms and full command of the German language. Assessment took place at therapeutic facilities of the "Grüner Kreis" society, an Austrian association dedicated to the treatment and rehabilitation of people with SUD.

Assessment of clinical characteristics and attachment dimensions
Participants where asked about age, gender and level of education. Current medication and comorbid psychiatric diagnoses were assessed through the facilities' database.
The Attachment Style Questionnaire (ASQ) [12] is a 40-item questionnaire which measures five dimensions of adult attachment: "Confidence in Self and Others" (15 items), "Discomfort with Closeness" (8 items), "Relationships as Secondary" (7 items), "Need for Approval" (4 items) and "Preoccupation with Relationships" (6 items). Items are rated on a 6-point Likert scale. While higher scores in "Confidence in Self and Others" indicate a more secure attachment, higher scores in the other four scales indicate more insecure attachment facets.

Statistical analysis
We conducted agglomerative hierarchical cluster analysis including a squared Euclidian distance interval and Ward's method to establish a parameter for the different attachment attitudes within the inpatient sample. Furthermore, we performed general linear model multivariate and step-wise hierarchical logistic regression analysis for group comparison and the prediction of treatment drop-out. Additional group comparisons were conducted by means of χ 2 tests. The alpha-level was set to .05.

Cluster analysis of attachment dimensions
All ASQ sub-scales showed satisfying internal consistencies (see Table 1), with the exception of "Discomfort with Closeness", which was therefore not taken into account for further interpretation.
In line with de Rick and Vanheule [13], a two-cluster solution could be accepted after performing agglomerative cluster-analysis using all ASQ subscales. As shown in Table 1 Cluster I, "increased secure attachment pattern", was comprised of inpatients with increased scores in "Confidence in Self and Others" (which equals more secure attachment) and decreased scores in almost all other dimensions representing facets of insecure attachment [11]. Conversely, Cluster II, "increased insecure attachment pattern" consisted of inpatients with lower scores in "Confidence in Self and Others", in combination with increased scores in most insecure attachment dimensions, with the exception of "Preoccupation with Relationships".
Furthermore, Cluster I inpatients exhibited significantly more comorbid disorders than Cluster II inpatients ( χ 2 (1) = 13.49, p < .001). In comparison to healthy controls [9], Bonferroni corrected post hoc analysis revealed that Cluster I inpatients showed lower scores in "Confidence in Self and Others" (p < .001) and higher scores in "Relationships as Secondary" (p < .001), but lower scores in "Need for Approval" (p < .001) and "Preoccupation with Relationships" (p < .001). Cluster II inpatients exhibited lower scores in "Confidence in Self and Others" (p < .001) and higher scores in "Relationships as Secondary" (p < .01), but did not show any differences (p > .05) in "Need for Approval" or "Preoccupation with Relationships" in comparison to control subjects.

Attachment predictors of therapy adherence
Within TR inpatients, "Confidence in Self and Others" decreased during the first 6 weeks of treatment (F (1,46) = 300.89, p < .05, η 2 = .13). No relevant differences were found for any of the other dimensions (p > .05).

Discussion
The primary goal of this study was to explore the impact of attachment on TA, for the entry phase of treatment (6 weeks), in a group of SUD inpatients. An Agglomerative Cluster analysis was conducted whereby a twocluster solution, more secure (Cluster I) versus more insecure (Cluster II) attachment patterns, could be acknowledged. In contrast to our assumptions, Cluster I subjects were more likely to drop out during the entry phase of treatment and also more likely to be diagnosed with a comorbid disorder than participants with a more insecure attachment pattern (Cluster II). Hierarchical Table 2    regression analysis revealed that the influence of the attachment security concealed the effects of the control variables sex and comorbidity on TA. Moreover, the remaining patients showed a significant decrease in the ASQ subscale "Confidence in Self and Others" at the second point of measurement. No other significant changes in attachment dimensions, assessed at the beginning of treatment, were found after 6 weeks. The surprising results regarding the influence of attachment on TA, which contradicts previous findings, might be partially explained by a lack of self-reflection within TD inpatients. In this study, a self-report measure of adult attachment was applied, which might be susceptible to distorted self-images and lacks the capacity to detect repressed information [14]. This problem might be more prominent in SUD patients than with patients with solely somatic condition [15]. By further considering the increased rate of comorbidity within the TD and the more secure attached group, it seems possible that the self-reported high attachment security more likely reflects an unrealistic (or idealized) image of the self in relationships with others, characterized by primitive defense mechanism like splitting and denial [16]. Furthermore, our findings could mirror a specific mechanism taking place specifically within therapeutic communities, which represents a threat to narcissistically biased self-images: The high amount of group cohesion within the community might lead to strong cognitive dissonances in vulnerable individuals, which in turn leads to early treatment drop-out [17]. A reduction in narcissism might also provide an explanation for a decreased "Confidence in Self and Others" score after 6 weeks of treatment. This result fits with the general approach of therapeutic community treatment to support patients in developing more conscious attitudes towards interpersonal deficits during their stay within the community [18]. Either way, the outcome of a decreasing "Confidence in Self and Others" might mirror an initial euphoria combined with the experience of being sober after such a long time of severe substance misuse. This exhilaration possibly vanishes after some time spent in the therapeutic community.

Conclusions
In conclusion, our highly preliminary results indicate that caregivers should be especially attentive with patients exhibiting a high amount of comorbidity in combination with a seemingly secure attachment attitude, especially if attachment was captured via standardized self-report measures. However, in line with the literature [5][6][7], this study confirms the importance of considering the attachment dimension for the treatment of SUD.

Limitations
• A rather small sample size and the use of a selfreport measure for adult attachment styles. • The focus was only on the first short treatment period of 6 weeks. For the description of longterm treatment effects within a therapeutic community setting further research might focus on an enhanced time-span. • The study examined a convenience sample. It is necessary to replicate these highly preliminary findings in future studies using randomized controlled trials. • For a more in depth understanding of our results, further research might especially focus on attachment based therapeutic interventions on TA.
Abbreviations ASQ: Attachment Styles Questionnaire; AUD: alcohol use disorder; PUB: poly drug use disorder; SUD: substance use disorders; TD: patients, who dropped out of treatment; TR: patients, who remained in treatment; TA: therapy adherence.