Did you get any help? A post-hoc secondary analysis of a randomized controlled trial of psychoeducation for patients with antisocial personality disorder in outpatient substance abuse treatment programs

One intervention designed to improve the outcomes of treatment for SUD in patients with ASPD, the Impulsive Lifestyle Counselling [ILC] program [14], was developed and tested in a pragmatic randomized trial. The trial found that ILC had significant effects on days abstinent from substance use and severity of drug use at the 3-month follow-up, and that aggression declined in both treatment as usual and ILC with no significant differences between interventions [15], and that the ILC program had significant effects on the risk of dropout from substance abuse treatment [16]. However, one question that has not been previously assessed in this study is the perceived helpfulness of treatments for mental health comorbidity. In general population studies, the perceived helpfulness of treatments for depression is sometimes assessed, usually through Likert scale items (e.g. [17, 18]), but research is scarce. Examining perceived helpfulness of treatment offers is important, as individual attitudes and beliefs about treatment can influence future service use and treatment seeking behavior [19]. For example, previous studies show that perceiving treatment as ineffective is one of the commonly cited barriers deterring individuals from seeking further treatment [20]. Perceived helpfulness may be even more important for people with ASPD, since patients with ASPD are prone to anger, and often find themselves in conflicts with others, including clinic staff [21]. Patients with any psychiatric disorder who enter treatment have to face considerable challenges to engage and remain in treatment. Usually, patients with psychiatric disorders have lived a life with critical or concerned family members, while at the same time experiencing a substantial degree of self-criticism. Enrolling in treatment often involves further challenges to the individual patient’s belief systems. In addition, treatment is likely to focus on the need to change coping strategies, which may be perceived by the patient as a form of help for dealing with the present, but may be ineffective in the long-term [22]. In this context, people who are involved with drugs and crime often have strong social bonds with others who use drugs and break the law, and may have value systems that conflict with collaborating with anyone who can be regarded as being authority figures, both factors which further complicate the treatment process [23]. At the same time, patients with DUD need to contemplate getting by without the drugs that have not only helped them cope with daily hassles, and which have also become a central part of their identity. As a consequence, entering into treatment becomes a difficult transition that requires a substantial amount of confidence in treatment; supporting that confidence is a key task for the treatment providers. Feeling that treatment is relevant and helpful may give the patients motivation to engage more in treatment and cooperate better with clinicians.

The ILC study was a Phase I pragmatic randomized controlled trial with single blind assessments that was carried out at community-based outpatient substance abuse treatment clinics in Denmark between January 2012 and July 2014. The details of the study are reported in two previous reports [15, 16]. Inclusion criteria to the study were: between 18 and 65 years old; met criteria for ASPD using the Mini International Neuropsychiatric Interview [24]; and able to provide written informed consent; seeking treatment or already in treatment for a SUD.

Study participants were identified by clinicians at the participating sites from new and existing patients receiving outpatient community treatment for a SUD. After agreeing to be contacted, a trained clinician at each site invited potential participants to take part in a screening to assess the diagnosis of ASPD. Those who met inclusion criteria were told that their responses indicated ASPD, and the clinician would then review their responses to the ASPD MINI module with the patients, and ask if they were willing to talk to a trained ILC counsellor about it. Patients who agreed to see a counsellor and who provided written informed consent, subsequently completed the baseline assessment and were randomly allocated to either one of two treatment groups: treatment as usual (TAU) or the Impulsive Lifestyle Counselling (ILC).

Randomization was stratified by clinic. The randomization schedules were generated by the trial coordinator and kept secure and confidential by the trial coordinator at the study coordinating center in Copenhagen. The trial coordinator informed the referring clinician of the result of randomization immediately after being notified that the patient had been assessed and found to be eligible for study participation. Following this, the clinician informed the patient of the result. In cases in which the patients were randomized to the ILC program, the clinician contacted one of the ILC counsellors at the uptake unit with the participants’ details so that the program sessions could be initiated immediately after randomization.

Because the randomization took place immediately after the screening interview, the trial coordinator was unable to check whether the baseline assessment was complete before randomizing, and patients with incomplete data at baseline were excluded after randomization.

For the three follow-up waves at 3, 9 and 15 months after randomization, patients were contacted first through any phone number they had provided. If it was not possible to establish contact using this procedure, the patient was contacted through the clinic at which he or she had been screened for study participation. If a patient still could not be reached, researchers would ask his or her case manager at the study clinic if they could set up a time for a meeting, or when it would be feasible to meet the patient. Patients who could still not be reached this way were contacted through the contact information that they had provided to significant others, and finally through other accessible services (e.g. prison or psychiatry). In a few cases, the patients were eventually located through the Central Personal Register. Once a patient had been located, and if it was possible to speak with the patient directly, a place and time for an interview was scheduled. If the patient did not show up for a face-to-face interview, a new time would be scheduled, and only after several failed attempts was a telephone interview suggested. If the patients stated that they were not willing to be interviewed, they would be asked if they would agree to be contacted at a later point, and if they refused, they were not contacted again.

Research technicians not affiliated with the study clinics carried out all assessments at the follow-up interviews and were blind to treatment group allocation.

Random-effects regression was used assess to the likelihood of a high versus low endorsement of having received help for ASPD during the three follow-up waves. Random effects were estimated for both patient and study site, and covariates were gender, age, receiving substitution treatment at baseline, treatment satisfaction at baseline and assessment wave (3 months being the reference category, and 9 and 15 months being the dummy codes).

The study applied intent-to-treat analyses, i.e. data were analyzed by randomization arm irrespective of attendance or treatment compliance. Covariates were gender, age, receiving substitution treatment at baseline, and assessment wave (3 months being reference, 9 and 15 months being dummy codes). The predictor was randomization status.

Following this, random-effects regression analysis was conducted for each of the primary outcomes of the trial, to test whether self-rated treatment for ASPD was associated with improved outcomes for the trial. In each of these analyses, the predictors were: perceived help for ASPD, age, gender, substitution at baseline and the baseline value for the outcome of interest, and random effects were estimated for study site and patient.

An identical random effects regression was conducted for global treatment satisfaction, defined in the trial registration as a secondary outcome. Survival analysis was used to assess whether perceived help for ASPD was associated with risk of subsequent dropout from the 3-month follow-up using Cox proportional hazards regression with clustering for site, again adjusted for gender, age and substitution treatment at baseline. Of the 128 patients who were interviewed at the 3-month follow-up, 18 had been discharged from treatment, and were thus no longer at risk of dropping out. Therefore, the survival analysis was conducted using the remaining 110 patients.

Finally, we intended to test whether self-rated help for ASPD mediated the effects of the ILC program on our two significant outcomes, days abstinent and ASI drugs composite score, we conducted two mediation analyses. First, we assessed whether effects of the ILC program on self-rated help for ASPD mediated the effects of ILC on days abstinent, and then we assessed whether the effects of the ILC program on self-rated help for ASPD mediated the effects of ILC on the drugs composite score [15]. In mediation analysis, some of the effect of the independent variable, the IV, is transmitted to the dependent variable, the DV, through the mediator variable, the MV and some of the effect of the IV passes directly to the DV. Mediation occurs when the following four statements are true: (1) the IV has significant effects on the DV; (2) the IV has significant effects on the MV; (3) the MV has significant effects on the DV; and (4) once the effects of the MV on the DV are controlled, the effect of the IV on the DV is significantly reduced [40]. In the mediation analyses, the IV was randomization status (ILC or TAU), the MV was self-rated help, and the DV was days abstinent in the first analysis, and the ASI drugs composite score in the second. In both mediation analyses, baseline values for the DV, substitution treatment, age and gender were included as control variables. We present the results of a power analysis to detect a significant mediation effect based on Thoemmes, Mackinnon & Reiser [41]. The presented analysis is based on the observed effects in this study, and was adapted from Thoemmes and colleagues’ work using MPlus 7.4. Specifically, we retained the a and b coefficients (i.e., the effect of ILC on perceived help and the effect, and the effect of perceived help on days abstinence), as these two are the only ones that are of consequence for the power analysis [42].

To assess the mediation effect, we used the Preacher and Hayes method. The indirect effect was obtained with 5,000 bootstrap resamples [43]. The mediation analyses were conducted using only when both the IV (random assignment) and the MV (self-rated help for ASPD) had significant on the outcome of interest, as mediation cannot occur when either the direct effect of the IV or the MV is not statistically significant.

Characteristics of the patients randomized to each of the two conditions are summarized in Additional file 1: Table S1. The 175 participants were 156 men and 19 women, the mean age was 32.4 years of age (standard deviation [SD] = 9.0), and 38.3% were receiving opioid substitution treatment at the time of randomization.

Of 175 patients randomized, 130 were interviewed at 3 months (74%), 118 at 9 months (67%), and 107 at 15 months (61%). Of these, a small number of respondents (ranging from 2 to 12 respondents at each follow-up wave), did not respond to the question about self-rated help for ASPD. The main reason that respondents did not answer the item was that the informant had to be interviewed by telephone, and in the telephone interviews, we asked only about the primary outcomes for this study. When patients had not responded to an item at a given follow-up wave, they were excluded from the analysis of that variable at that point.

The ratings of help for ASPD are summarized in Fig. 1. As can be seen in Fig.1, regardless of randomization assignment, a very large proportion of patients reported that they did not receive any help for ASPD at all.

The mixed effects regression is summarized in Table 1. The primary aim of this analysis was to test whether the intervention had a direct effect on self-rated help. Higher scores on help for ASPD were associated with randomization to ILC (p < .001), and older patients gave lower scores on having received help for ASPD (p = .016). No other covariates were significant. The intraclass correlation for site was small (ρ = .082, 95% CI = .015–0.337), and the intraclass correlation for patient was moderate (ρ = .415, 95% CI = .295–0.546).

The results of mixed regression analyses for self-reported outcomes are summarized in Table 2. The results of these analyses were to test whether self-rated help was associated with better outcomes.

Perceived help for ASPD was associated with more days abstinent (B = 1.29, z = 2.32, p = .020), treatment satisfaction (B = 0.079, z = 2.50, p = .012), and self-reported general aggression on the BPAQ (B = 0.103, z = 2.13, p = .033). Associations with ASI drugs composite score (B = 0.004, z = 0.81, p = .416), B = -0.002, z = -0.30, p = .766), and interpersonal aggression (B = 0.027, z = 1.38, p = .169) were not statistically significant.

Self-reported treatment for ASPD was associated with higher satisfaction with treatment (B = 0.077, p = .015). The survival analysis was significant (Wald χ ²(4) = 13.83, p = .008), and self-reported treatment for ASPD at the 3-month follow-up was associated with lower risk of subsequent dropout (Hazard ratio [HR] = 0.762, z = -2.27, p = .023). In addition, age was negatively associated with dropout (HR = 0.935, z = -2.02, p = .044) and substitution treatment at baseline was negatively associated with subsequent dropout (HR = 0.464, z = 2.08, p = .037).

The theoretical mediation analyses are summarized Fig. 2. Following the convention in mediation analysis, the a coefficient is the effect of the IV, ILC, on the MV, perceived help for ASPD. The b coefficient is the effect on the MV on the DV, days abstinent or drugs composite score. The c coefficient is the direct effect of the IV on the DV.

Given that only days abstinent satisfied the two first criteria for mediation (direct effects of random assignment and effect of MV on DV), we only analyzed mediation for this variable. The results are summarized in Table 3. Of the 128 patients who answered the question about help for ASPD, four had not answered the items about days abstinent at baseline, and had to be excluded, leaving 124 patients for this analysis. In the first step, days abstinent at the 3-month follow-up was associated with randomization to ILC (Z = 2.06, p = .042). In the second step, perceived help was associated with randomization to ILC (Z = 2.85, p = .005). In the third step, perceived help was associated with more days abstinent (Z = 2.17, p = .032), and randomization to ILC was no longer significantly associated with days abstinent (Z = 1.47, p = .144). For days abstinent, 27% of the effect of randomization was explained by self-rated help for ASPD (observed indirect coefficient = 1.037, bootstrap CI 0.050–2.600) and given that the CI does not overlap with zero, the indirect effect was significant [43].

A post-hoc power analysis was conducted to assess mediation effects for the present study based on the recommendations of Thoemmes, Mackinnon & Reiser [41]. The analysis was based on the Mplus code for a single mediator, and we substituted the values from the present study for the values in their code. The analysis showed that the indirect effect would be significant in 85.3% of replications with a sample of 124 patients. With 90 patients, 62.9% of tests would be significant, with 110 patients, 78.1% would be significant, ant with 140 patients, 91.4% would be significant. The Mplus output is in Additional file 2.