تظاهرات بالینی و واکنش به دارو درمانی در اختلال اضطراب اجتماعی:بررسی اثر باورهای اتیولوژیک

Abstract Therapies for social anxiety disorder (SAD) leave many patients symptomatic at the end of treatment and little is known about predictors of treatment response. This study investigated the predictive relationship of patients׳ etiological attributions to initial clinical features and response to pharmacotherapy. One hundred thirty-seven individuals seeking treatment for SAD received 12 weeks of open treatment with paroxetine. Participants completed the Attributions for the Etiology of Social Anxiety Scale at baseline in addition to measures of social anxiety and depression at baseline and over the course of treatment. A latent class analysis suggested four profiles of etiological beliefs about one׳s SAD that may be characterized as: Familial Factors, Need to be Liked, Bad Social Experiences, and Diffuse Beliefs. Patients in the more psychosocially-driven classes, Need to be Liked and Bad Social Experiences, had the most severe social anxiety and depression at baseline. Patients in the Familial Factors class, who attributed their SAD to genetic, biological, and early life experiences, had the most rapid response to paroxetine.These results highlight the effect of biological and genetically-oriented etiological beliefs on pharmacological intervention, have implications for person-specific treatment selection, and identify potential points of intervention to augment treatment response.

. Introduction Social anxiety disorder (SAD) is highly prevalent, with up to 13% of the U.S. population experiencing SAD at some point during their lives (Ruscio et al., 2008). SAD is characterized by a marked and/or persistent fear of one or more social situations and is associated with significant functional impairment (Aderka et al., 2012, American Psychiatric Association, 2013 and Schneier et al., 1994). Individuals with SAD have higher rates of alcohol and drug dependence, depression, suicide, and use of medical resources, as well as diminished vocational and educational attainment (Acarturk et al., 2009, Katzelnick et al., 2001 and Van Ameringen et al., 2003). Although there are well-validated treatments for SAD (Heimberg and Magee, 2014, Schneier et al., 2014 and Wong et al., 2012), response rates for even the best empirically supported treatments suggest that many treated patients remain symptomatic. For instance, in one study, 35% of patients receiving the monoamine oxidase inhibitor phenelzine and 42% of patients receiving group cognitive behavioral therapy (CBT) were classified as non-responders (Heimberg et al., 1998). Moreover, trials of serotonin-norepinephrine reuptake inhibitors and selective serotonin reuptake inhibitors (SSRIs) suggest similar rates of non-response. For example, non-response rates ranged from 41% with venlafaxine (Liebowitz et al., 2005) to 45% with paroxetine (Stein et al., 1998) to 47% with sertraline (Van Ameringen et al., 2001). In an effort to augment treatment response for mental disorders, the National Institute of Mental Health (NIMH) called for the study of elements of personalized mental health care in its Strategic Plan (NIMH, 2008). Personalized medicine seeks to identify variables related to both patient and treatment modality that optimize treatment outcomes. The scope of possible avenues of research is wide and include pharmacogenetics, pharmacotherapy dosing schedules, and predictors of treatment outcome that inform patient-treatment matching (Arch and Ayers, 2013). Only a handful of studies have examined variables that impact treatment outcome in SAD. Having an expectation of benefiting from group CBT predicts enhanced treatment response (Chambless et al., 1997 and Safren et al., 1997). Among patients with SAD, certain cognitive characteristics (e.g., cognitive reappraisal self-efficacy, negative cognitive appraisal) may mediate response to CBT (Goldin et al., 2012 and Hofmann, 2000). However, only a few published studies have examined predictors of response to pharmacotherapy for SAD (Bruce et al., 2012). To our knowledge, there are no published studies investigating whether and to what extent patients׳ cognitive characteristics impact response to pharmacotherapy for SAD, information potentially relevant to tailoring to personalizing therapeutic intervention. 1.1. Attributions as a potential predictor of pharmacotherapy response in SAD Etiological attributions (i.e., causal explanations of the etiology of one׳s disorder) are one type of cognitive characteristic that may impact treatment response. Causal attributions have been associated with both etiology and maintenance of SAD (Hope et al., 1989). For example, individuals with SAD tend to attribute positive outcomes to external factors and negative outcomes to internal factors. This attributional bias strengthens as social anxiety intensifies (Coles et al., 2001). Though this literature is not specifically focused on attributions about etiology, attributions about causality play an important role in SAD. In the context of psychopathology more generally, the attributions that individuals make about their disorder may influence the steps they take in the pursuit of treatment (Roth and Eng, 2002). Although there is currently no research that investigates whether etiological beliefs directly influence treatment response, research suggests that etiological beliefs influence perceived efficacy of treatment ( Furnham, 1995). Thus, matching treatments to patients׳ etiological beliefs may lead to better treatment response based on expectancy effects, which as noted above, have been associated with response to CBT for SAD ( Chambless et al., 1997 and Safren et al., 1997). Thus, given the relevance of causal attributions to SAD, investigating the clinical effect of etiological attributions is warranted. 1.2. Present study Aligned with the NIMH׳s call for the personalization of mental health care, we sought to identify baseline patient characteristics differentially related to clinical presentation and response to pharmacotherapy, information which may inform ways to augment treatment response. This work has the potential to add to an emerging body of literature on personalization of treatment for SAD (e.g., Craske et al., 2014) and to extend this research by investigating personalization within the context of etiological attributions and pharmacotherapy, two previously unexplored domains. Thus, we examined whether individuals׳ attributions about the etiology of their SAD are related to their the initial severity of their symptoms and predictive of their response to pharmacotherapy with paroxetine, an SSRI demonstrated to be efficacious in the treatment of SAD (Allgulander, 1999, Baldwin et al., 1999 and Stein et al., 1998). We hypothesized that individuals would differ both in the types of attributions that they endorse (e.g., genetics, family environment, stressful social experiences), as well as frequency of attributions (e.g., moderate vs. high levels of genetic attributions). We used latent class analysis (LCA) to identify distinct profiles of SAD-related etiological attributions. We expected that classes of individuals whose profiles are characterized by genetic/biological attributions would be associated with a more severe clinical presentation at baseline. Additionally, we expected classes of individuals with profiles emphasizing genetic/biological attributions to exhibit better response to paroxetine, as these types of etiological attributions would best match the treatment modality and, in line with previous research (Furnham, 1995), lead to higher expectancy of treatment efficacy.

. Results 3.1. Characteristics of sample The sample had a mean age of 32.82 (S.D.=11.34), ranged from 18 to 62 years, and was 37.9% female. The mean years of education were 15.29 years. Sixty-five participants (47.4%) self-identified as Caucasian, 18 (13.1%) as Asian or Pacific Islander, 29 (21.2%) as Black, and 25 (18.2%) as “Other.” At baseline, the sample was mildly depressed (BDI-II; M =16.91; S.D. =12.42) and displayed high levels of social anxiety (LSAS; M =75.98; S.D. =21.04). 3.2. Latent class analysis LCA models were first conducted with all 17 AESAS items as indicators of class membership. However, only the one-class model fit the data; the two-class model could not be replicated because of extreme scores or lack of variability in responses. Specifically, one class was composed of individuals who rated the items “highly perfectionistic parents” and “unstable or abusive family life” as high in importance, suggesting that lack of variance for these items was leading to difficulties replicating the model solution. Thus, we removed these items as predictors of class membership. The final set of LCA models was conducted using the remaining 15 AESAS items as predictors of class membership. As indicated in Table 1, the lowest BIC was associated with the three-class model, and the lowest AIC and ABIC were associated with the four-class model. However, the BLRT indicated that the four-class model better fit the data than the three-class model, suggesting that the four-class model had the best overall fit. The five-class model produced two small classes, one of which (n=7) was only 5% of the sample. This small class size suggests that the five-class model was unlikely to be replicated; therefore, it was not considered further. Table 1. Fit indices for latent class analysis models with 1–4 classes. Number of classes Number of free parameters Log likelihood AIC BIC ABIC BLRT 1 30 −3464.40 6988.81 7076.41 6981.50 N/A 2 46 −3346.00 6783.99 6918.31 6772.79 <0.001 3 62 −3308.78 6741.57 6922.60a 6726.46 <0.001 4 78 −3272.47 6700.94a 6928.70 6681.94a <0.001 AIC=Akaike Information Criterion; BIC=Bayesian Information Criterion; ABIC=sample-size adjusted BIC; BLRT=Bootstrap likelihood ratio test. A five-class model produced a class composed of only 5% of the sample and is therefore not reported. a Indicates best fitting model according to each index. Table options Mean AESAS item scores across the four classes are displayed in Fig. 1. Consistent with LCA conventions, we named the classes based on the frequency and quality of the profiles of responses regarding individuals׳ beliefs about the etiology of their SAD. The Familial Factors class (n=12) exhibited the greatest endorsement of family-related etiological factors, both biological (e.g., “something inherited from my parents/family genetics”) and early environmental (e.g., “very critical parents”). The Need to be Liked class (n=53) exhibited lower endorsement of familial factors and relatively stronger endorsement of feelings of low self-esteem and a strong desire to make a good impression or to be liked by others. The Bad Social Experiences class (n=48) exhibited greater endorsement than did all other classes of the belief that “having been teased as a child or rejected by peers” and having “one or more bad experiences in social situations that resulted in embarrassment or humiliation” played an important role in the development of their SAD. They also endorsed feelings of low self-esteem, sensitivity to criticism, and a tendency to think pessimistically. The Diffuse Beliefs class (n=24) rated all of the AESAS items as having just below slight to moderate importance regarding the development of their SAD. 3.3. Comparisons of external validators among classes 3.3.1. Demographic characteristics Omnibus χ2 analyses indicated that the classes did not differ on age, sex, race, ethnicity, employment status, marital status, or religion (χ2s<2.02, ps≥0.383). 3.3.2. Baseline symptom severity Omnibus χ2 analyses revealed significant between-class differences for baseline social anxiety and depressive symptoms (see Table 2). Pairwise comparisons indicated that the Need to be Liked and Bad Social Experiences classes exhibited greater social anxiety and depressive symptoms at baseline than the other two classes (χ2s> 4.33, ps≤ 0.037). Table 2. Comparisons of external validators in the four-class model. Class 1: Diffuse Beliefs (n=24) Class 2: Familial Factors (n=12) Class 3: Need to be Liked (n=53) Class 4: Bad Social Experiences (n=48) Omnibus χ2 test Pairwise comparisonsa M S.D. M S.D. M S.D. M S.D. χ2 p Social Anxiety Symptoms: Week 0 (baseline) LSAS – total score 63.28 4.71 67.34 4.04 77.46 2.67 83.13 3.25 15.43 0.001 Classes 3,4>1,2 Depressive symptoms: Week 0 (baseline) BDI-II – total score 6.98 1.71 11.21 2.08 17.53 1.79 21.68 1.83 30.11 <0.001 Classes 3,4>1,2 Social Anxiety Symptoms: Week 4 LSAS – total score 47.82 6.66 38.52 4.28 60.80 3.68 67.12 3.93 8.25 b 0.041 Class 3,4>2 LSAS=Liebowitz Social Anxiety Scale; BDI-II=Beck Depression Inventory – II. Comparisons on change in Social Anxiety Symptoms by Weeks 8 and 12 and change in depressive symptoms by Week 12 are not reported here; however, they were conducted and no significant between-class differences were found. a All pairwise comparisons are significant at the p<0.05 level. b Between class comparisons for differences in LSAS – total score by Week 4 were conducted using LSAS residualized change scores for each class (which were computed by regressing LSAS – total at week 4 on LSAS – total at week 0). Table options 3.3.3. Treatment response Omnibus χ2 analyses revealed significant between-class differences for residualized social anxiety change scores 2 by Week 4 of treatment (see Table 2). Pairwise comparisons indicated that the Familial Factors class exhibited greater reductions in social anxiety by Week 4 (M [S.D.] =−0.61 [0.26]) than did the Need to be Liked (M [S.D.] =0.10 [0.15]) and Bad Social Experiences classes (M [S.D.] =0.20 [0.16]; see Fig. 2). The Diffuse Beliefs class did not differ from any of the other classes. There were no between-class differences in social anxiety change at Weeks 8 or 12. Although depressive symptoms were not assessed at Weeks 4 or 8, there were no observed differences in depressive symptom change at end of treatment (Week 12). Results suggest that there are four distinct profiles of beliefs about the causes of SAD and these profiles were differentially associated with baseline severity of symptoms and initial response to pharmacotherapy. Mean Liebowitz Social Anxiety Scale (LSAS) scores for each class in the ... Fig. 2. Mean Liebowitz Social Anxiety Scale (LSAS) scores for each class in the four-class model over the course of treatment with paroxetine.