عوامل پیش بینی کننده در رفتاردرمانی شناختی با ارائه اینترنتی و مدیریت استرس اضطراب شدید سلامت

Severe health anxiety can be effectively treated with exposure-based Internet-delivered cognitive behavior therapy (ICBT), but information about which factors that predict outcome is scarce. Using data from a recently conducted RCT comparing ICBT (n = 79) with Internet-delivered behavioral stress management (IBSM) (n = 79) the presented study investigated predictors of treatment outcome. Analyses were conducted using a two-step linear regression approach and the dependent variable was operationalized both as end state health anxiety at post-treatment and as baseline-to post-treatment improvement. A hypothesis driven approach was used where predictors expected to influence outcome were based on a previous predictor study by our research group. As hypothesized, the results showed that baseline health anxiety and treatment adherence predicted both end state health anxiety and improvement. In addition, anxiety sensitivity, treatment credibility, and working alliance were significant predictors of health anxiety improvement. Demographic variables, i.e. age, gender, marital status, computer skills, educational level, and having children, had no significant predictive value. We conclude that it is possible to predict a substantial proportion of the outcome variance in ICBT and IBSM for severe health anxiety. The findings of the present study can be of high clinical value as they provide information about factors of importance for outcome in the treatment of severe health anxiety.

Severe health anxiety is common in medical settings, chronic for the majority of affected individuals if untreated, and associated with functional impairment and substantial suffering (Abramowitz et al., 2007, Barsky et al., 1998 and Tyrer et al., 2011). In the present paper the term severe health anxiety is used synonymously with DSM-IV hypochondriasis (American Psychiatric Association, 2000). In two recently conducted randomized controlled trials (RCTs) our research group has shown that Internet-delivered exposure-based cognitive behavior therapy (ICBT) is effective in the treatment of severe health anxiety (Hedman et al., 2011 and Hedman et al., 2014). ICBT can be described as therapist-guided online bibliotherapy and requires less therapist time than face-to-face CBT and has the important advantage that each therapist can treat as many as 80 patients simultaneously (Andersson, 2009 and Andrews et al., 2010). For anxiety disorders and depression this type of treatment can yield large effect sizes and seems to be similarly effective as the most well-established psychological treatments delivered in a face-to-face format (Hedman, Ljótsson, & Lindefors, 2012). Although ICBT for severe health anxiety is generally effective about a third of patients are insufficiently improved. For the clinician it is therefore important to know how likely it is that a given patient will respond to treatment. Gaining more knowledge about these likelihoods can be achieved by investigating prognostic factors, i.e. predictors of treatment response. Using this knowledge in clinical contexts can lead to a larger proportion of successfully treated patients (Kraemer, Wilson, Fairburn, & Agras, 2002). This can be achieved by using empirical data regarding predictors when making treatment recommendations, i.e. to offer ICBT only to those who are likely to respond it. It can also be used to generate hypotheses about how the treatment might be improved or how it could be individually tailored. For example, if it would be found that persons with reading difficulties do not respond to treatment it may be of interest to develop a treatment relying more on video clips and audio files than large amounts of text. Based on the first published RCT of ICBT for severe health anxiety (Hedman et al., 2011), we analyzed clinical, demographic and therapy process-related predictors and found that more health anxiety at baseline predicted more improvement whereas more depressive symptoms were related to less improvement (Hedman et al., 2013). In that study, treatment adherence in terms of number of completed treatment modules was also positively associated with improvement, while demographic factors were largely unrelated to outcome (Hedman et al., 2013). However, a limitation of much predictor research is that findings tend to be fairly inconsistent across studies and it has been suggested that predictors found in one sample should be validated in a second sample to avoid type I errors (e.g. Hellstrom & Ost, 1996). We have found no replication studies on predictors of psychological treatment of severe health anxiety, i.e. where the same treatment is tested and the same methods for investigating predictors are used. The firm structure of ICBT and the limited between-therapist effects makes it highly suitable for predictor replication research. This means that if a predictor is identified in one study it should also be present in another, given adequate power and that participants are recruited from the same population. In a recently conducted RCT we compared ICBT with Internet-delivered behavioral stress management (IBSM) for severe health anxiety (Hedman et al., 2014). The results showed that both treatments yielded large reductions of health anxiety but that exposure-based ICBT was more effective (Hedman et al., 2014). This second trial provides a very good framework for hypothesis-driven testing of predictors. That is, the predictors found in the first trial of ICBT for severe health anxiety should be associated with outcome also in the second trial if they are of true predictive value. The design of this RCT also allows for testing of moderators, i.e. treatment specific predictors. As ICBT is based on exposure to health anxiety-provoking stimuli and IBSM is based on symptom control through applied relaxation and stress management it could be that predictors differ between treatments. In other words, it could be that one treatment is more suitable than the other depending on patient characteristics. Although not previously investigated in ICBT for severe health anxiety, five additional potential predictors were considered of interest in the present study due to previous research. These were somatosensory amplification, perceived competence, mindfulness, working alliance and reading skills. Somatosensory amplification can be described as “the tendency to experience somatic and visceral sensation as unusually intensive, obnoxious or disturbing” and has been suggested to be involved in the pathogenesis of health anxiety and found to be elevated in persons with severe health anxiety (Barsky, Wyshak, & Klerman, 1990). Perceived competence, according to self-determination theory, is related to anxiety and some evidence suggests that perceived competence can predict symptoms of anxiety and depression (Uhrlass, Schofield, Coles, & Gibb, 2009). Mindfulness, which has been defined as bringing awareness to the present moment in an accepting way (Baer, Smith, Hopkins, Krietemeyer, & Toney, 2006), was regarded potentially relevant in the present study as mindfulness training is part of ICBT for severe health anxiety throughout the treatment. We viewed the mindfulness facet non-reactivity to internal events to be of specific interest as that is something that could facilitate successful exposure and response prevention. Working alliance can be described as the degree to which the patient and the therapist agree on goals and tasks and how strong their relational bond is (Horvath & Greenberg, 1989). A previous study has shown that working alliance can predict improvement in ICBT for post-traumatic stress disorder (Knaevelsrud & Maercker, 2007). Finally, as ICBT could be viewed as a form of bibliotherapy and patients read about 100 pages of text during the treatment it could be that reading ability is a factor that is related to outcome. To our knowledge, no prior study has investigated whether these five factors predict outcome in ICBT for severe health anxiety. The aim of this study was to investigate clinical (e.g. symptom levels), demographic (e.g. age) and therapy process-related (e.g. adherence) predictors in ICBT and IBSM for severe health anxiety using data from an RCT. Employing a replication design, we hypothesized that higher baseline health anxiety would predict larger improvements but higher end state health anxiety. Depressive symptoms were expected to be associated with less improvement as well as with higher end state health anxiety. We also hypothesized that more completed modules would predict larger improvements and less end state health anxiety. Based on the results from the previous predictor study of ICBT (Hedman et al., 2013), the remainder of tested factors was hypothesized to have little predictive value. Somatosensory amplification, self-efficacy, mindfulness, working alliance and reading skills, were analyzed within an exploratory framework.

Results Attrition According to Little's MCAR test data were missing completely at random (χ2 = 321.6; df = 289; p = .09). Of the 158 participants, 158 (100%) provided data at baseline while 151 (96%) completed assessments at post-treatment. Treatment efficacy summary Participants in both ICBT and IBSM made large and significant improvements on the primary outcome measure HAI. The baseline to post-treatment effect size was d = 1.8 in the ICBT group and d = 1.2 in the IBSM group. Mixed models analysis showed a significant interaction of group and time indicating larger improvements in the ICBT group compared to IBSM (F = 3.9; df = 2, 121; p = .022). Predictors of end state health anxiety Significant predictors (HAI) that were included in the final multiple linear regression model explained 47% of the variance of end state health anxiety (R2 = 0.47; F = 20.8; p < .001). Table 2 presents the regression coefficients of predictors that remained significant in the final multiple and quantile regression models. Below, results from each predictor domain are presented. Table 2. Multiple linear and quantile regression presenting the final significant predictors using HAI scores at post-treatment as dependent variable. Linear regression Quantile regression Predictors B SE β t P-value Coef. t P-value Health anxiety (HAI) 0.65 0.09 0.48 7.58 <0.001 0.58 3.34 <0.01 Completed modules −1.47 0.53 −0.18 −2.71 <0.01 −2.05 −2.42 <0.02 Treatment credibility (C-scale) −0.66 0.22 −0.23 −3.00 <0.01 −0.53 −1.62 0.11 Abbreviations: HAI, Health Anxiety Inventory; Coef., Coefficient. Table options Clinical characteristics Univariate linear regression analyses showed that more baseline health anxiety (HAI), more depressive symptoms (MADRS-S), and having a comorbid diagnosis of depression significantly predicted more health anxiety at post-treatment. General anxiety (BAI), anxiety sensitivity (ASI), psychotropic medication with SSRI/SNRI and having a comorbid anxiety disorder were not significantly associated with outcome and there were no significant moderators. In the final multiple and quantile regression models, baseline health anxiety remained significant (Table 2). Demographic characteristics Among the demographic variables, only age was a significant predictor from the initial univariate linear regression. Being older predicted significantly less health anxiety at post-treatment. Age did not remain significant in the final multiple regression model. None of the other investigated demographic variables, i.e. gender, marital status, computer skills, educational level, and having children or not, significantly predicted health anxiety at post-treatment and there were no significant moderators. Therapy process related variables Univariate regression analysis showed that more completed modules, higher levels of treatment credibility, mindful non-reactivity, and perceived competence, and better working alliance predicted less health anxiety at post-treatment. Subjective reading ability and time spent reading each week did not predict outcome and there were no significant moderators. In the final multiple regression analysis, number of completed modules and treatment credibility remained significant predictors of end state health anxiety. In the quantile regression analysis number of completed modules was a significant predictor but treatment credibility was not. Predictors of health anxiety improvement Table 3 presents significant predictors in the final multiple and quantile regression models using baseline to post-treatment improvement as dependent variable. The final multiple regression model explained 31% of the variance in health anxiety improvement (R2 = 0.31; F = 13.0; p < .001). Table 3. Multiple linear and quantile regression presenting the final significant predictors of baseline to post-treatment change scores of the HAI as dependent variable. Linear regression Quantile regression Predictors B SE β t P-value Coef. t P-value Health anxiety (HAI) 0.24 0.10 0.20 2.52 <0.02 0.22 2.01 0.046 Anxiety sensitivity (ASI) 0.44 0.17 0.21 2.59 <0.02 0.42 2.40 0.018 Completed modules 1.48 0.50 0.21 2.93 <0.01 1.88 3.49 0.001 Treatment credibility (C-scale) 0.46 0.23 0.18 2.03 <0.05 0.44 2.39 0.018 Working alliance (WAI) 0.47 0.18 0.21 2.59 <0.02 0.36 2.04 0.043 Abbreviations: HAI, Health Anxiety Inventory; ASI, Anxiety Sensititivity Index; WAI, revised short version of Working Alliance Inventory; Coef., Coefficient. Table options Clinical characteristics Univariate regression analyses showed that more baseline health anxiety (HAI), general anxiety (BAI), and anxiety sensitivity (ASI) predicted more improvement in health anxiety. Depressive symptoms, years with severe health anxiety, concurrent stable psychotropic medication with SSRI/SNRI, comorbid depression, comorbid anxiety disorder did not significantly predict improvement. In the final multiple and quantile regression models, baseline health anxiety and anxiety sensitivity remained significant predictors. Demographic variables No demographic variables, i.e. age, gender, having children, computer skills, educational level, or marital status significantly predicted improvement in health anxiety. Therapy process related factors More completed modules, higher levels of treatment credibility, somatosensory amplification, and mindful non-reactivity, and better working alliance were significant predictors in the univariate regression analyses. Reading ability and time spent reading were not significantly related to outcome and there were no moderators. In the final multiple regression and quantile models number of completed modules, treatment credibility and working alliance remained significant predictors. Decision tree based on signal detection analysis Fig. 1 displays the results of the signal detection analysis based on recursive partitioning. As shown in Fig. 1 treatment credibility and baseline health anxiety were significant predictors of clinically significant improvement. Full-size image (46 K) Fig. 1. Decision tree of predictors from signal detection analysis. Receiver operator characteristics (ROC) for predictor C-scale at baseline: a) Positive predictive value, 62%, b) Negative predictive value, 76%; ROC for predictor Health anxiety at baseline (HAI cut-point 110): a) Positive predictive value, 81%, b) Negative predictive value, 66%; ROC for predictor Health anxiety at baseline (HAI cut-point 105): a) Positive predictive value, 42%, b) Negative predictive value, 88%. Abbreviation: HAI, Health Anxiety Inventory.