عوامل پیش بینی و ناظران رفتاردرمانی شناختی مبتنی بر اینترنت برای اختلال وسواس: نمایش نتایج از یک محاکمه تصادفی

Abstract Internet-based cognitive behavior therapy (ICBT) for obsessive–compulsive disorder (OCD) has shown efficacy in randomized trials but many patients do not respond to the treatment, we therefore need to find predictors and moderators of treatment response. In this study, we analyzed predictors of ICBT response using both post-treatment as well as 24-month outcome data. As half of the participants were randomized to receive an Internet-based booster program as an adjunct to ICBT, we also investigated moderators of ICBT with or without booster. Results showed that more severe baseline OCD symptoms predicted worse end state outcome but also higher degree of change. Furthermore, high degree of working alliance predicted better outcome but patients with primary disgust emotions had worse treatment effects. The moderator analysis also indicated that scoring high on the obsessing subscale on the Obsessive–Compulsive Inventory-Revised predicted worse treatment outcome in the booster group. In conclusion, there are some possible predictors and moderators of ICBT for OCD but more research is needed with larger and clinically representative samples.

Obsessive–compulsive disorder (OCD) is a common and debilitating condition (Kessler et al., 2005 and Weissman et al., 1994) with low spontaneous remission rate (Mataix-Cols et al., 2002 and Skoog and Skoog, 1999). The most well-established psychological treatment is cognitive behavior therapy (CBT), with responder rates averaging 50–60% (Fisher & Wells, 2005) and sustained long-term effects (Whittal, Robichaud, Thordarson, & McLean, 2008). CBT can be delivered both in individual- and group format (Fisher & Wells, 2005). A recent innovation is Internet-based CBT with therapist support (ICBT), where all therapist contact is provided using interactive online platforms (Andersson, 2009). Our research group has previously tested ICBT for OCD in an open pilot study (Andersson et al., 2011) and in a randomized controlled trial (Andersson et al., 2012a), with responder rates similar to conventional CBT (60–61%). These findings correspond well with the effects found in two Australian trials of ICBT for OCD (Wootton et al., 2011 and Wootton et al., 2013). In a recent study, we reused the sample from the Andersson et al. (2012a) trial and randomized half the participants to also receive an Internet-based booster program as an adjunct to ICBT. Results showed that both groups had sustained long-term effects up to two years after receiving ICBT, but participants randomized to ICBT with booster had a slower relapse rate compared to ICBT without booster (Andersson et al., 2014). Although research has shown that both individual-, group- and Internet-based CBT are effective treatments for OCD, a significant proportion of the patients do not get an adequate treatment response and the full recovery rate of CBT has been estimated to be only about 25% (Fisher & Wells, 2005). One possible way to reduce the non-responder rates is to investigate predictors and moderators of treatment response. A predictor in a randomized controlled trial (RCT) is a pre- or post-treatment variable that has a main effect on outcome but shows no interactive effect with the treatment condition. A moderator has the same characteristic as a predictor but with the difference that it shows an interactive between-group effect on outcome. For instance, if pre-treatment depressive symptoms were found to be associated with worse outcome for both CBT and psychodynamic therapy, it would be a predictor. But if it were to be shown that pre-treatment depression only affected the outcome for patients going through the CBT, it would be a moderator. Thus, research on predictors and moderators are important because they can provide the clinician with information for whom the chosen treatment works for, thereby reduce treatment failures and also help to develop individually tailored treatments (Kazdin, 2007 and Kraemer et al., 2002). There are, to our knowledge, no published studies that have investigated predictors or moderators of ICBT for OCD. This study therefore set out to investigate this issue. We used outcome data from a previously conducted randomized trial of ICBT with a long-term follow-up data, where 101 patients received ICBT for 10 weeks and were then assessed at post-treatment and at 24-month follow-up (Andersson et al., 2012a and Andersson et al., 2014). As patients presenting with hoarding- and sexual/religious obsessions have been shown to have worse treatment response in previous studies (Jaurrieta et al., 2008, Keeley et al., 2008 and Knopp et al., 2013), our first hypothesis was that this would also be the case in ICBT. Another clinical variable that has been shown to be important in CBT for OCD is pre-treatment symptom levels (Keeley et al., 2008). Our second hypothesis was therefore that higher pre-treatment OCD symptoms would be indicative of worse treatment response. Furthermore, an interesting process variable that has been shown to be important in the treatment of OCD is the therapeutic alliance (Keeley et al., 2008 and Vogel et al., 2006), and this variable has also been shown to be associated with adherence which, in turn, is associated with better treatment response (Simpson et al., 2011). Our third hypothesis was therefore that higher degree of working alliance would predict better treatment response also in ICBT for OCD. Another process variable that may interact with treatment outcome is the role of emotions. An example of this is McKay (2006) who showed that OCD patients with self-reported disgust in relation to specific obsessions do not respond as quickly to ERP as patients with other self-reported primary emotions (e.g. fear). We therefore wanted to investigate if this process variable mattered also in a large-scale sample, and our fourth and last hypothesis was that patients with primary disgust emotions would show worse treatment response.

3.1. Predictors at post-treatment Table 2 shows the predictive values of each predictor variable in the first step. In the stepwise regression model, Y-BOCS pre-treatment score, presence of disgust emotions and low degree of working alliance predicted more symptoms at post-treatment. When using Y-BOCS change score as dependent variable, both Y-BOCS pre-treatment score and working alliance predicted higher degree of improvement, whereas presence of disgust emotions predicted worse treatment response. Detailed information about the final model at post-treatment is shown in Table 3. Signal detection analysis also yielded an interaction model with pre-treatment Y-BOCS score cutoff of >20 interaction with <69 on the WAI as best predictor of being a non-responder (Fig. 2). Table 2. Separate regression analyses on each outcome variable. Y-BOCS post-treatment score as dependent variable Predictor B SE t p-Value OCI-R hoarding 0.31 0.23 1.34 0.18 OCI-R obsessing 0.17 0.20 0.84 0.41 Y-BOCS pre 0.74 0.11 6.34 <0.001 Disgust emotions 3.70 1.42 2.61 <0.05 WAI −0.09 0.05 −1.83 0.07 Y-BOCS pre- to post-treatment change score as dependent variable Predictor B SE t p-Value OCI-R hoarding 0.02 0.20 0.09 0.93 OCI-R obsessing 0.15 0.17 0.91 0.37 Y-BOCS pre 0.26 0.12 2.18 <0.05 Disgust emotions −2.63 1.23 −2.13 <0.05 WAI −0.09 0.04 2.20 <0.05 Y-BOCS 24-month score as dependent variable Predictor B SE t p-Value OCI-R hoarding 0.49 0.24 2.05 <0.05 OCI-R obsessing 0.32 0.20 1.57 0.12 Y-BOCS pre 0.49 0.14 3.54 <0.001 Disgust emotions 2.71 1.57 1.73 0.09 WAI −0.01 0.05 −0.27 0.79 Y-BOCS pre- to 24-month change score as dependent variable Predictor B SE t p-Value OCI-R hoarding −0.19 0.24 −0.79 0.44 OCI-R obsessing 0.11 0.21 0.52 0.60 Y-BOCS pre 0.51 0.14 3.62 <0.001 Disgust emotions −1.33 1.58 −0.84 0.40 WAI 0.02 0.05 0.31 0.76 Abbreviations: Y-BOCS, Yale-Brown Obsessive–Compulsive Scale; WAI, Working Alliance Inventory; OCI-R, Obsessive–Compulsive Inventory-Revised. Table options Table 3. Multivariate linear regression presenting the final model using Y-BOCS scores at post-treatment. Y-BOCS post-treatment score as dependent variable (R2=0.38,p<0.001) Predictor B SE t p-Value Y-BOCS pre 0.72 0.11 6.31 <0.001 Disgust emotions 2.43 1.21 2.01 <0.05 WAI −0.90 0.04 −2.15 <0.05 Y-BOCS pre- to post-treatment change score as dependent variable (R2=0.24,p<0.001) Predictor B SE t p-Value Y-BOCS pre 0.29 0.11 2.52 <0.05 Disgust emotions −2.43 1.21 2.01 <0.05 WAI 0.090 0.04 2.15 <0.05 Abbreviations: Y-BOCS, Yale-Brown Obsessive–Compulsive Scale; WAI, Working Alliance Inventory. Table options Full-size image (37 K) Fig. 2. Decision tree on post-treatment outcome based on signal detection analysis. Abbreviations: Y-BOCS, Yale-Brown Obsessive–Compulsive Scale; WAI, Working Alliance Inventory. Figure options 3.2. Predictors at 24-month follow-up Table 2 shows the statistics on the separate regressions. The only predictor that was retained in the stepwise regression model was Y-BOCS pre-treatment score (B=0.50, t=3.55, p <0.001). When repeating the analysis, but instead using the Y-BOCS change score (from pre-treatment to 24-month follow-up), the only predictor that was entered in the final model was a positive relationship with Y-BOCS pre-treatment score i.e. higher baseline symptoms predicted higher degree of reduction (B=0.51, t=3.62, p <0.001). 3.3. Moderators at 24-month follow-up As the sample was divided in two groups at the 24-month follow-up (ICBT+booster vs. ICBT without booster), we repeated the analyses but also created interaction terms to see if any of the predictors were moderated by the booster treatment. Patients who scored high on the obsessing subscale had worse treatment response if they were randomized to the booster group (B=0.63, t=2.33, p <0.05).