تجزیه و تحلیل مواد افزودنی از زیرگروه های تشخیصی اختلال اضطراب اجتماعی: پیامدها برای DSM-V

Abstract Background and objectives Much controversy exists regarding diagnostic subtypes of social anxiety disorder (SAD). The present study used admixture analysis to examine whether individuals with generalized and nongeneralized SAD belong to the same or different populations of origin. This can inform diagnostic subtyping of SAD in the forthcoming DSM-V. Methods Treatment-seeking individuals with generalized SAD (n = 154) and nongeneralized SAD (n = 48) completed a battery of questionnaires. Based on participants’ responses to the Liebowitz Social Anxiety Scale (LSAS), we estimated log likelihood and chi-square goodness-of-fit for models with 1, 2, 3, or 4 populations of origin, and compared models using forward stepwise estimation and maximum likelihood ratio tests. Results Admixture analyses suggested that the two diagnostic subtypes of SAD belong to the same underlying population of origin. In addition, observable differences in depression, general anxiety, and comorbidity were no longer significant when controlling for social anxiety severity. Limitations Our sample was recruited in the U.S. and was a treatment-seeking sample. Future studies should examine whether our results generalize to different cultures, and community samples. Conclusions Support for qualitative differences between SAD subtypes was not found. Rather, our findings support the notion that the diagnostic subtypes of SAD differ quantitatively, and that SAD exists on a continuum of severity. This finding informs diagnostic subtyping of SAD in the forthcoming DSM-V.

1. Introduction Individuals with social anxiety disorder (SAD) fear and avoid diverse social situations. The DSM-IV instructs clinicians to assign a “generalized” subtype of SAD if “fears include most social situations” (APA, 1994, p. 417). Although the DSM does not define the residual category of SAD, many researchers and clinicians have referred to individuals who do not fulfill the criterion above as belonging to the “nongeneralized” or “specific” subtype (e.g., Heimberg et al., 1993, Hofmann et al., 1995 and Hook and Valentiner, 2002). As the DSM definition does not set clear quantitative criteria, several operational definitions for SAD subtypes have been suggested (for a review see Hofmann, Heinrichs, & Moscovitch, 2004). Consistent with previous reviews (Blöte et al., 2009, Hofmann et al., 2004 and Hook and Valentiner, 2002) as well as the DSM definition, we will distinguish between individuals who have fears pertaining to most social situations (generalized subtype) and the residual subgroup of individuals who fear fewer situations (nongeneralized subtype). Much controversy exists regarding the nature of the difference between the two diagnostic subtypes of SAD. Some researchers view these subtypes as representing different points on a continuum of social anxiety severity (Hofmann, 2000, Rapee, 1995, Rapee and Spence, 2004, Stein et al., 2000 and Vriends et al., 2007). In contrast, others have argued for qualitative differences between the subtypes and view them as distinct disorders that are associated with differences in diagnostic comorbidity, etiology, and the type of feared situations (Carter and Wu, 2010, Hook and Valentiner, 2002 and Knappe et al., 2011). The empirical evidence regarding the quantitative and qualitative views of SAD subtypes is equivocal. Some studies reported differences in the demographic profile of individuals with generalized and nongeneralized SAD (Brown et al., 1995, Heimberg et al., 1990 and Levin et al., 1993), whereas other studies found no such differences (Herbert et al., 1992, Hofmann and Roth, 1996, Holt et al., 1992, Mannuzza et al., 1995 and McNeil et al., 1995). Similarly, some authors found that individuals with generalized SAD have more comorbid disorders compared to individuals with nongeneralized SAD (Holt et al., 1992, Mannuzza et al., 1995 and Turner et al., 1992), whereas another study reported that these differences disappeared after controlling for the number of feared situations (El-Gabalawy, Cox, Clara, & Mackenzie, 2010). It has further been reported that the generalized subtype is associated more with fear of interpersonal situations, whereas the nongeneralized subtype is associated more with fear of performance situations (Cox et al., 2008 and Stemberger et al., 1995). However, other studies did not find the type of social fears to differentiate between the two subtypes (e.g., Stein et al., 2000), supporting a dimensional structure (Furmark et al., 2000 and Kollman et al., 2006). Finally, some studies found the generalized subtype to be less responsive to treatment compared to the nongeneralized subtype (Brown et al., 1995), whereas others found it to be more responsive (Marom, Gilboa-Schechtman, Aderka, Weizman, & Hermesh, 2009) and still others have found no differences (Stein, Stein, Goodwin, Kumar, & Hunter, 2001; for a comprehensive review, see Hofmann et al., 2004). A complicating issue in the interpretation of the results from the SAD subtype literature is that finding differences between subtypes can be consistent with either a qualitative or a quantitative view. For instance, individuals with the generalized subtype have been found to report fewer positive thoughts and more negative thoughts during an interpersonal task compared to individuals with the nongeneralized subtype (Beazley, Glass, Chambless, & Arnkoff, 2001). This may be due to either qualitatively distinct cognitive processes or quantitatively greater social anxiety severity. In order to effectively juxtapose the two views, it is imperative to control for social anxiety severity and examine whether differences remain (e.g., El-Gabalawy et al., 2010). A qualitative distinction is likely if differences remain after accounting for severity of social anxiety. Alternatively, differences that are in the reverse direction assumed by the quantitative view (e.g., individuals with nongeneralized SAD endorsing a certain social fear to a greater extent compared to individuals with generalized SAD) would also provide support for a qualitative view (Hook & Valentiner, 2002). In the present study, we used a novel statistical technique to examine SAD subtypes in a large treatment-seeking population. Specifically, we used admixture analysis, which examines whether a sample distribution is likely derived from one or more normally distributed populations of origin (Delorme et al., 2005 and Kolenikov, 2001). Admixture analysis has advantages compared to both cluster analysis and taxonomic analysis. Cluster analysis is a non-inferential technique that examines groups of individuals on the sample level. In contrast, admixture is an inferential technique that examines groups on the population level. Taxonomic analysis examines whether latent structures are more likely taxonic or dimensional by generating simulated sample data that is compared to real data (Meehl, 1999). In contrast, admixture provides the means and standard deviations of the population distributions that are most likely to generate the data (Kolenikov, 2001). In the context of SAD, admixture can determine if social anxiety symptoms among individuals with the generalized and nongeneralized subtypes are the product of a single population of origin, thus supporting the quantitative view, or the product of two or more populations of origin, thus supporting the qualitative view. A single population of origin is inconsistent with the qualitative view as it posits that the subtypes are two distinct disorders. To our knowledge, this is the first study to examine SAD subtypes using admixture analyses. We also examined whether the observed differences between the subtypes can be accounted for by levels of social anxiety (supporting the quantitative view) or if differences remain after controlling for social anxiety (supporting the qualitative view). Finally, we examined whether individuals with the nongeneralized subtype would endorse certain social fears to a greater degree than individuals with the generalized subtype as some studies suggest (Cox et al., 2008, Hook and Valentiner, 2002 and Stemberger et al., 1995)

3. Results 3.1. Admixture analysis We examined LSAS scores for all study participants, and computed estimated log likelihood and χ2 goodness-of-fit for models with 1, 2, 3 or 4 normally distributed populations of origin. The estimated log likelihoods were −934.8, −933.1, −932.1, and −929.6, respectively. Maximum likelihood ratio tests indicated that the difference between 1 and 2 populations (1.7) was lower than the threshold value (7.8) and non-significant, View the MathML sourceχ(3)2 = 1.7, p > 0.05. Thus, assuming a second population of origin did not increase model fit indicating that the best-fitting model is the one including only a single normally distributed population. The single population of origin had a mean of 70.34 (95% Confidence Interval = 66.93–73.76), and a standard deviation of 24.75 (95% Confidence Interval = 24.07–25.44). We also examined χ2 goodness-of-fit for all models (View the MathML sourceχ(2)2 = 3.21, p = 0.67; View the MathML sourceχ(2)2 = 2.41, p = 0.30; View the MathML sourceχ(2)2 = 5.16, p = 0.08; View the MathML sourceχ(2)2 = 7.25, p = 0.03, for models with 1, 2, 3, and 4 populations respectively). According to Kolenikov (2001), the highest χ2 goodness-of-fit probability value indicates the best-fitting model. This is because non-significant χ2 indicates good model fit, whereas a significant result indicates the model significantly differs from the data. Thus, the model with a single population (p = 0.67) is also the best-fitting model according to this criterion. To sum, admixture analysis indicated that the data were most likely derived from a single population. Fig. 1 presents LSAS frequencies in the sample and Fig. 2 presents the estimated population of origin. Frequencies of LSAS in the sample. Fig. 1. Frequencies of LSAS in the sample. Figure options Estimated population of origin. Fig. 2. Estimated population of origin. Figure options 3.2. Differences in demographic and clinical profile We compared individuals with generalized and nongeneralized SAD on all demographic and clinical measures. Results are presented in Table 1. No significant differences between the subtypes emerged on any demographic measure. Individuals with the generalized subtype reported higher LSAS, BDI and STAI-T scores, and had more comorbid disorders (see Table 1). However, when controlling for levels of social anxiety (LSAS scores), no differences were found on the BDI (F(1, 103) = 2.2, p = 0.14, n.s.), STAI-T (F(1, 184) = 2.8, p = 0.10, n.s.), or the presence of a comorbid disorder (Wald statistic = 2.2, df = 1, p = 0.14, n.s.). Thus, the differences found between the subtypes are best accounted for by the degree of social anxiety. 3.3. Differences in feared situations We compared individuals with generalized and nongeneralized SAD on LSAS factors. We examined LSAS factors from 2 models found among clinical populations: a 4-factor model (Safren et al., 1999) and a 5-factor model (Perugi et al., 2001). Compared to individuals with the nongeneralized subtype, individuals with the generalized subtype had significantly elevated scores on all factors except the formal speaking anxiety factor, for which no significant differences were found (Table 2). In all comparisons, individuals with the generalized subtype reported more symptoms compared to individuals with the nongeneralized subtype. Table 2. Comparison between social anxiety disorder subtypes on LSAS factors. Generalized subtype Nongeneralized subtype Statistic p 4-Factor model (Safren et al., 1999) Social interaction 30.9 (9.3) 14.8 (10.2) F(1, 196) = 100.2 <0.001 Observation fear 5.5 (4.5) 2.3 (2.6) F(1, 199) = 21.1 <0.001 Public speaking 24.0 (5.6) 22.1 (5.4) F(1, 200) = 4.3 <0.05 Eating and drinking in public 3.5 (0.3) 0.8 (0.1) F(1, 198) = 30.8 <0.001 5-Factor model (Perugi et al., 2001) Interpersonal anxiety 31.1 (10.0) 15.0 (10.0) F(1, 197) = 91.9 <0.001 Formal speaking anxiety 20.0 (4.4) 19.0 (4.3) F(1, 200) = 1.9 0.17 Stranger-authority anxiety 11.2 (4.2) 5.7 (4.0) F(1, 199) = 63.3 <0.001 Eating and drinking while observed 3.1 (3.5) 0.3 (0.8) F(1, 198) = 30.8 <0.001 Doing something while observed 6.9 (4.1) 2.6 (2.7) F(1, 200) = 45.8 <0.001