The impact of poly-traumatization on treatment outcomes in young people with substance use disorders

Recently, researchers and clinicians have increasingly begun to focus on a particularly high-risk sub-group of children and youth who has been exposed to several types of trauma, rather than just a single type of traumatic event [4‐6]. The theoretical foundation of this new focus is the cumulative risk theory which states that the higher the number of risk factors a person is exposed to, the higher the potential for a negative outcome [7] and the underlying dose response model which illustrate that the accumulation of traumatic events is often a stronger predictor of outcomes than single event exposure [8].

Individuals with a history of poly-traumatization have been shown to continue to experience additional victimization [9], severe and persistent biopsychosocial impairment [10, 11], and are at increased risk of developing negative behavioral patterns such as delinquency and substance abuse compared to individuals with few or no trauma experiences [12, 13]. In a study of youth involved in the juvenile justice system, investigators conducted a latent class analysis derived from 19 types of adversity, including different forms of violence exposure produced three unique classes. Among the three classes that emerged, the poly-trauma group was distinct from other adversity groups in the presence of severe emotional and behavioral problems [14]. In another recent longitudinal study that included 1186 adolescents, poly-victimization was linked to lower levels of social support, self-esteem, and mastery [15]. However, in recent studies that examine the accumulative effect of trauma victimization, certain specific types of traumatic exposures, for instance, child sexual abuse have also been shown to account for impairment over and above the cumulative impact of the number of specific exposure types [16‐18]. None-the-less, the established links between multiple types of trauma, negative behavioral patterns, and cognitive functions may carry important implications for substance abuse treatment, particularly in relation to treatment efficacy.

The link between trauma experiences and substance use disorders (SUD) is well established [4]. In surveys of adolescents receiving treatment for substance abuse, between 60% and 87% of patients reported a history of exposure to trauma [5, 6]. Moreover, studies show that 37% to 52% of SUD patients have a lifetime diagnosis of PTSD [7, 8]. In particular, the literature has focused on experiences of childhood abuse as an underlying risk factor for substance use behaviors [9‐11]. Examining the effect of trauma history on SUD treatment outcomes is important, because the treatment of SUD patients with a trauma history may need to be adjusted to the needs of the patient based on their trauma experiences [19, 20]. For instance, treatment may lead to exposure to cues related to the traumatic events. This may, if unaddressed, produce symptoms of anxiety or anger, in turn leading to patient attrition, or reinitiating substance use to cope with intruding memories and avoidance symptoms [21, 22].

However, overall, the influence of previous trauma experiences on the effect of SUD treatment is unclear. A number of studies have indeed indicated that individuals with past trauma experiences, have poorer outcomes of substance abuse treatment including worse treatment adherence and less improvement during treatment [23‐27]. Yet, other studies have found that past trauma experiences are not a predictor of poor prognosis or missing appointments in SUD treatment [28‐30].

The majority of studies that have examined associations between specific trauma experiences and substance use disorder (SUD) treatment outcomes, have focused on single types of childhood trauma such as physical or sexual abuse [30‐33]. Childhood trauma as an overall category has been found to be a predictor for earlier drop-out from SUD treatment [34], higher rates of relapse [35] and less improvement in treatment measured by drug use and psychological functioning [36]. Also, Parolin and colleagues found that young drug users receiving drug use disorder (DUD) treatment who had been exposed to parental substance abuse during their developmental age, had elevated rates of neuropsychological impairments, particularly affecting the attentive and executive functions, compared to drug users who were not exposed to parental substance abuse [37]. However, to our knowledge, existing studies have not controlled for the effect of other trauma exposure on treatment outcomes and therefore it is unclear whether it is indeed the specific type of trauma or an accumulative effect of trauma exposure that is associated with treatment effect.

In light of the inconsistent findings, there is a need for additional research that examines the role of past trauma experiences in substance abuse treatment settings. In particular, very few studies have examined the influence of the number of trauma experiences in relation to outcomes of DUD treatment. To strengthen knowledge in the substance abuse treatment field, this study aimed to examine the effect of poly-traumatization on treatment effect in a Danish treatment for DUD based on cognitive-behavioral therapy.

Since the word trauma holds multiple meanings, it is important to point out that the events referred to as trauma in the present study are only potentially traumatic, meaning that they may or may not have caused trauma symptoms. The same applies for the term poly-traumatization, which in this study covers multiple types of trauma experiences and not the level of trauma symptomatology. Recent trauma studies have used the terms correspondingly [38, 39].

Based on the cumulative risk theory and existing studies that link accumulative trauma to a lower SUD treatment effect, as well as various negative psychosocial and behavioral factors that are relevant for treatment outcome, such as self-esteem, social support and mastery [15], attentive and executive functions [30], and severe emotional and behavioral problems [5, 12], the overall research question for the present paper was whether participants with trauma exposure, and in particular a high number of traumatic experiences, would be less likely to benefit from DUD treatment than participants with no or few trauma experiences. This focus led to the following three hypotheses:

Since similar studies have shown that factors such as gender [40, 41], socio economic status [42], immigrant status/ethnic minority status [43], and age [44, 45] may play a great role in relation to SUD treatment outcome, we controlled for the effect of these factors in the present study.

The study combined data from two sources: one was a randomized controlled clinical trial [46] that examined the effect of a motivational enhancement and CBT-based intervention. Participants were recruited by staff at nine participating treatment centers. The second source was data collected from the same nine sites during the post-trial phase of the clinical trial, when treatment components were implemented in day-to-day practice¹.

All participating sites offered outpatient treatment for youth with illicit DUDs in agreement with the Danish Consolidation Act on Social Services under Part 18, § 101. According to the act, treatment must be offered without co-payment, and within two weeks after clients have contacted the treatment services.

Participants were eligible if they were between the ages of 15 and 25 years, and had contacted the treatment center with the intention of entering treatment for DUD. Clients were excluded if they sought treatment for an opioid use disorder, had known active psychotic disorders, known cognitive deficits, and/or were otherwise deemed not safe to participate (e.g., had recently engaged in violent actions against staff). The reason for excluding participants with opoid use disorders was that treatment for this particular group is often focused on substitution medicine and this focus is very important throughout treatment. It was therefore assessed that the offered treatment would cover this essential focus for this particular group well enough. The demographic characteristics are summarized in Table 1.

Eligible clients were invited to take part in the clinical trial by clinical staff at each of the nine treatment centers.

From September 2014 to May 2016, clients were invited to participate in the clinical trial, and from the end of May 2016 and until October 2017, all eligible clients were asked if they would consent to the use of their assessment data for research purposes and quality assurance.

All participating clients were offered 12 sessions of manual-based counselling incorporating a combination of cognitive behavioral therapy and motivational interviewing (see Table 2).

In addition to this manual based counselling (standard treatment), some participants received vouchers for attendance, reminders of sessions, or both vouchers and reminders [46]. Based on the variety of treatment components, the potential effect of different treatment tracks was controlled for in our analyses (i.e., the four randomized groups and the post-trial phase).

At the first counselling session after trial participation, all clients were assessed using the YouthMap assessment form [47]. The YouthMap assessment form is a comprehensive assessment tool, tailored for young people in treatment for psychoactive substance use disorders. The YouthMap covers background information, current substance use, past and current behavioral and mental problems, victimization experiences, peer and family relations amongst others. Additionally, during the treatment course all participants were assessed in each session with the TEM screening tool [48, 49]. TEM is a routine outcome monitoring tool that consists of eight questions based on the feedback informed treatment approach [50, 51]; five questions related to drug use and three questions related to well-being. All responses were entered into a secure web-based interface at baseline by the involved researchers.

To assess poly-traumatization, the sample was divided into three trauma categories based on the number of reported trauma experiences. The trauma types do not represent a scale, as each is a potentially independent event from the others, rather than different manifestations of the same underlying construct. For this reason, we chose to analyze the number of different traumatic events as an index [55]. That is, a measure of factors that may all individually contribute to the same phenomenon (in this case, being burdened by the impact of traumatic events). We used the eight items measuring trauma victimization. For seven of the eight trauma variables the response “yes” was coded as trauma victimization. For bullying, answers were only coded as trauma victimization if participants had responded “to some extent”, “a lot”, or “extensively”. Individuals were grouped as zero trauma experiences, one to four trauma experiences, or five or more trauma experiences.

A mixed effects Tobit regression was conducted for days of cannabis use. Days of cannabis use was censored at zero days (i.e., past week abstinence) and 7 days (i.e., daily use). The variables of interest were status in terms of trauma exposure measured on a nominal scale (i.e., zero trauma types, one to four trauma types, or five or more trauma types), and the interaction between trauma status and session. The interaction between trauma and session was included in order to assess the slope of days of cannabis use as a function of trauma categories (i.e. the number of different types of victimization).

A mixed effects logistic regression was conducted for session attendance. The variables of interest were again status in terms of trauma measured on a nominal scale (i.e., zero trauma types, one to four trauma types, or five or more trauma types), and the interaction between trauma status and session.

A linear mixed effects regression was conducted for well-being. The variables of interest were again status in terms of trauma measured on a nominal scale (i.e., zero trauma types, one to four trauma types, or five or more trauma types), and the interaction between trauma status and session.

Finally, two graphs were produced to illustrate the course of cannabis use and well-being throughout treatment. For use of cannabis and well-being, a fractional polynomial fit was used to assess days of cannabis use as a function of session, stratified by degree of poly-traumatization.

Missing data can influence the results of a prospective clinical study, when data is not missing completely at random. While there is no one solution to this issue, one way to reduce the influence of missing data is to conduct sensitivity analyses adjusting for factors that are likely to influence missingness [56]. While we cannot guarantee that we can identify all factors that are associated with missing data, we do know that randomization to any of the active treatment groups in our study increased attendance and reduced no-shows [57].

As a sensitivity analyses, all analyses were conducted controlling for treatment group (four randomized groups and post-trial status). Further sensitivity analyses were conducted controlling for the externalizing and internalizing scales of the YouthMap measure [58].

A total of 775 participants were included in the study, of which 599 were male and 176 were female. The mean age was 20.2 years (standard deviation [SD] = 2.6, range = 15 to 25). At baseline, the participants reported smoking cannabis on average 17.6 days in the past 30 (SD = 11.3, range 0 to 30) and 4.04 days in the past 7 days (SD = 2.77, range 0–7). A total of 38% reported use of drugs or substances other than cannabis or alcohol.

Of the participants, 64 (8.3%) had never experienced any of the eight trauma experiences, 540 (69.7%) had experienced between one and four of the events, and 171 (22.1%) had experienced five or more of the eight events. The incidence of each trauma type across groups is shown in Table 3.

Figure 1 shows days of cannabis use in the last week as a function of time.

The mean number of days declined from 4.04 on average at baseline (SD = 2.77) to 2.16 days at the twelfth session for the participants who completed all sessions (n = 134, SD = 2.88). The results of the Mixed Effects Tobit regression are shown in Table 4. The model Wald χ² test was significant (χ²(13) = 268.1, p < 0.001).

In the unadjusted analyses, few or multiple trauma experiences were associated with a higher degree of cannabis use compared with no trauma history (one to four versus zero: p = 0.045; five or more versus zero: 0.010). With each session, days of cannabis declined (p < 0.001). The decline was less for those with a trauma history (one to four versus zero: p = 0.033; five or more versus zero: p = 0.004).

After controlling for confounders, clients who reported one to four types of trauma no longer differed significantly at baseline (coefficient = 1.65, CI − 0.09–3.22, p = 0.064), but clients who reported five or more types of trauma continued to show significantly more cannabis use at baseline (coefficient = 2.06, CI 0.20–3.91, p = 0.030). In terms of reductions in cannabis use over time, both clients who reported one to four of the trauma (coefficient = 0.19, CI = 0.04–0.33, p = 0.010), and clients who reported five or more of the trauma types (coefficient = 0.25, CI = 0.09–0.40, p = 0.002), showed a slower rate of decline in days of cannabis use compared to clients who did not report any of the trauma types, as indicated by significant group by time interactions.

The sensitivity analyses were not meaningfully different from the original analysis. As in the original analysis, both groups that reported trauma were less likely to reduce their cannabis use during treatment compared with those who did not report any of the trauma types.

Clients with zero trauma types attended on average 6.5 sessions (SD = 4.7), clients with one to four trauma types attended on average 6.5 sessions (SD = 4.3), and clients with five or more trauma types attended on average 7.0 sessions (4.3). Table 5 shows the results of random effects logistic regression to predict session attendance.

Compared with clients with zero trauma types, those with one to four (p = 0.036), and those with five to eight types (p = 0.038) were more likely to attend treatment at later sessions as indicated by significant session by group interactions.

The marginal probability of attending sessions over time as a function of trauma victimization is shown in Fig. 2.

A reliability analysis was carried out on the 3-item wellbeing scale. Cronbach’s alpha showed the scale to reach acceptable reliability, α = 0.71.The results of the logistic regression analysis for the association between well-being and the number of trauma types are shown in Table 6 and Fig. 3. In both the unadjusted (the second column of Table 6) and the adjusted models (column 4), trauma history was associated with lower overall well-being, but time by trauma interactions were non-significant, indicating that the rate of change did not differ between more or less trauma types. In all cases, well-being improved substantially over time in treatment.

None of the sensitivity analyses yielded markedly different results.

Firstly, self-reports of victimization experiences and substance use have their limitations. Specifically, questions on victimization that are not behavior-specific have been criticized because they may underestimate incidents since these questions assume a respondent will categorize their experience as abuse or be willing to self-identify as a victim [82]. It is argued, that people who have experienced physical or sexual abuse may not recognize that what they experienced was against the law or may be reluctant to categorize what they experienced as abuse [83]. However, this argument may be more relevant in surveys that examines the general population than in DUD treatment. Individuals in DUD treatment may be more prone to report victimization experiences, firstly because they are already in a developmental process in which they have recognized that they have a serious problem, and secondly, because that information could help their counsellor plan their treatment course. Due to the very high prevalence of victimization experiences in the present study, underreporting of victimization experiences do not seem to be an issue, but indeed more detailed and behavior-specific measures are recommended. Further, the validity of reported drug use has been questioned in multiple studies [84] and the measure we used for substance use has not been validated in a Danish context. It is indeed possible that both recall bias and underreporting may have influenced thhe drug use reports. However, we only asked for drug use within the past 7 days, which has minimized the influence of a recall bias. Even so, further research into the validity of the particular measure used in the present study is warranted. It is furthermore a limitation that we only measure the treatment effect on drug use via cannabis use. However, due to the low frequency of use of other drugs, it was not statistically meaningful to include these in this study. It is indeed possible that the interaction between traumatic experiences and treatment effect varies considerably depending on drug type or the presence of multiple drug use versus single drug use. More research into these variations is recommended. Also, we did not have data to examine the severity of trauma experiences and to which degree the participants were affected by their experiences. Consequently, we do not know if the level of traumatization is a causal factor and perhaps even the most important factor in the relationship between the number of trauma experiences and treatment outcome. Finally, an important limitation is that the findings do not generalize to participants who dropped out of treatment or cancelled multiple sessions. Thus, the results should be interpreted as representing the influence of trauma on progress during treatment, not beyond it.