Impact of depression on treatment effectiveness and gains maintenance in social phobia: a naturalistic study of cognitive behavior group therapy

EFFECTIVENESS AND GAINS MAINTENANCE IN SOCIAL PHOBIA: A NATURALISTIC STUDY OF COGNITIVE Sofi Marom, Ph.D.,1Ã Eva Gilboa-Schechtman, Ph.D.,2 Idan M. Aderka, M.A.,2 Abraham Weizman, M.D.,1 Background: The impact of depression on cognitive behavioral group therapy(CBGT) for social phobia (SP) in a naturalistic outpatient setting was examinedafter treatment termination and at 1-year follow-up. Methods: Consecutive SPoutpatients (N 5 219) were diagnosed using a structured interview. CBGT wasprovided in 18 1.5-hr weekly sessions. At pretreatment and posttreatmentquestionnaires and clinician ratings were administered. Self-report measureswere obtained at 1-year follow-up. The main outcome measure was the LiebowitzSocial Anxiety Scale. Results: CBGT was found to be effective in reducing bothsocial anxiety (effect size 5 1.23) as well as depression (effect size 5 0.94).
Individuals with generalized social phobia (GSP) and individuals with specificsocial phobia (SSP) differed in their presenting psychopathology and in theirresponse to CBGT. Among treatment completers, 44% GSPs and 37% SSPsachieved at least 50% improvement, and 44% GSPs and 87% SSPs reporteddistress and functioning within the normal range at the end of treatment. AmongSPs diagnosed with major depressive disorder (MDD) at the onset of treatment,SP symptoms aggravated during the follow-up period, whereas SPs not diagnosedwith MDD experienced a further alleviation of SP symptoms during follow-up.
CBGT provided in a public clinic to non-selected, mostly unmedicated andcomorbid patients, is an effective treatment for the majority of SP sufferers.
Conclusions: MDD at the onset of CBGT was not associated with poorertreatment response, but predicted exacerbation of SP symptoms followingtreatment termination. Depressed SPs may need additional intervention tomaintain CBGT gains. SSPs may benefit from less intensive CBGT than GSPs.
Depression and Anxiety 0:1–12, 2009.
Key words: social phobia; cognitive behavioral therapy; depression;co-morbidity; treatment outcome ÃCorrespondence to: Sofi Marom Ph.D., Anxiety Disorders andBehavior Therapy Unit, Outpatient Department, Geha MentalHealth Center, P.O. Box 102, Petah-Tikva 49100, Israel.
E-mail: ehud@barak.net.il Received for publication 4 February 2007; Revised 23 July 2007; Anxiety Disorders and Behavior Therapy Unit, Outpatient Department, Geha Mental Health Center, Tel-Aviv University,Tel-Aviv, Israel 2Psychology Department, Bar-Ilan University, Ramat-Gan, Published online in Wiley InterScience (www.interscience.wiley.
treatment outcome for the former group.[17,18] Exam- ining the differential response of GSPs and SSPs to ocial phobia (SP) is the third most common CBGT is the second goal of this study.
psychiatric disorder, with an estimated lifetime pre- Using pretreatment level of depression to predict valence rate of 7–13% in western countries.[1] SP CBGT outcome resulted in inconclusive results. For usually starts in childhood or adolescence and is instance,[19] found that individuals with SP and typically comorbid with other axis-I disorders, parti- comorbid depression had more severe SP symptoms cularly anxiety and mood disorders.[2,3] In the absence than individuals with SP alone. However, the degrees of treatment, SP is a chronic lifelong disorder with of improvement and rates of improvement were similar little spontaneous remission course.[4] The DSM-IV- for both groups. This suggests that despite being a TR[5] distinguishes between generalized social phobia graver condition, comorbidity with depression does not (GSP) and non-generalized or specific social phobia affect the course of treatment or the response to (SSP) subtypes. GSP is characterized by fear of treatment. In a similar vain,[20] concluded that depres- multiple social situations, whereas SSPs fears are sion has no detrimental effect on treatment or circumscribed to a small number of situations.
improvement during treatment of SP. In contrast, Randomized controlled trials have repeatedly de- other studies found that depression was related to monstrated that SP can be successfully treated with dropping out during treatment and to diminished either pharmachotherapy or cognitive behavioral ther- improvement during treatment.[15, 21] and [22] also apy [CBT].[6,7] One of the most widely studied variants found higher levels of depression to be predictive of of CBT—group CBT (CBGT)—has received robust diminished gains at outcome. Importantly, no study has empirical support.[8–10] In spite of the well-established yet examined that patient characteristics predict the therapeutic effects of CBGT in SP, no single study maintenance of treatment gains. The examination of examined these effects in a naturalistic setting, with a the effects of depression at the onset of CBGT on both large sample size and an extended follow-up.[11] treatment outcome and treatment gains maintenance is conducted a treatment outcome study in SP using a large (n 5 387) sample, and so did[12] (n 5 295).
However, both studies were randomized double-blindplacebo-controlled trials, by which the virtue of their design limit the generalization of their findings tonaturalistic settings. In addition, the lack of follow-up assessments in the Davidson et al.’s[12] study precludes Upon contacting the outpatient clinic (T0), patients were any conclusions regarding gains maintenance. Other interviewed using the Mini International Neuropsychiatric Inter- studies investigated the treatment of SP in more view.[23] The data concerning all participants were gathered as a part naturalistic settings. For instance,[13] compared treat- of routine clinical practice and were used for the purpose of this study ment of SP in a research unit with the same treatment (the procedure was approved by the Institutional Review Board of theGeha Mental Health Center).
in an independent private practice and found no The interval between the intake (T0) and the first treatment differences in outcome between the settings. However, session (T1) averaged 4.8 months (SD 5 1.2). CBGT was conducted this study included a moderate sample size (50–60 per by a Ph.D.-level clinical psychologist and a co-therapist (a psychiatry/ condition) and follow-up was conducted only 3 months psychology resident). There were 18 weekly sessions of 1.5-hr after termination of treatment. Another important duration. The treatment protocol was based on the protocols of [24] naturalistic study was conducted by,[14] who examined and the theoretical writings of [25] The protocol included (a) 217 participants undergoing treatment in naturalistic psychoeducation about SP; (b) exposure to feared social situations, settings and reported results comparable to controlled (c) reduction of safety behaviors, (d) cognitive restructuring, (e) trials. However, patients were assessed only up to 6 instruction on external focus of attention, and (f) social skills training.
weeks after treatment termination. To sum, many A typical session was divided into three parts: theoretical introduc-tion, in vivo exposure to feared social situations (e.g. reading in front recent studies examined various aspects of the general- of the group), and cognitive work. Homework was assigned weekly izability and maintenance of CBGT, but none com- and included in vivo exposure and self-administered cognitive bined naturalistic setting, large sample size, and restructuring. Each member of the treatment group took part in extended follow-up assessment. This combination is three to six in-session exposure exercises during the course of the entire treatment, involving situations that are commonly stressful to The existing data are also limited with respect to most individuals with SP. Weekly homework assignments typically CBGT’s differential treatment effects for the two included exposure corresponding to the participants’ unique fears.
subtypes of SP. GSPs begin treatment more impaired Participants comprised 21 CBT groups. Number of participants per than SSPs and some studies suggest that patients with group ranged between six and 20 with a mean of 10.8 (SD 5 4.3).
SSP are more responsive to CBGT—than are GSPs,[15] During the last session (T2) patients were assessed using the samemeasures as at intake.
whereas other studies found a similar degree of Drop-out was defined as (a) a premature cessation of participating improvement in the two groups.[16,17] Clearly, if GSPs in CBGT, that is, nonparticipation in the last session or (b) absence begin treatment more impaired than SSPs, similar from six or more CBGT sessions. No additional sessions (i.e.
degree of improvement lead to a worse end-of- ‘‘booster sessions’’) were conducted after termination of CBGT.
Approximately, 1 year after termination (T3) all participants who Both patient and clinician versions of the Clinical Global completed CBGTreceived questionnaires by mail and were requested Impressions-Improvement Scale [CGI-I][33] were administered at T2 to assess participants’ improvement. The CGI-I is a single itemassessing the degree of change in a person’s distress, on a 7-pointscale, such that 1 represents the greatest improvement and 7 represents a significant worsening. The Clinical Global Impres- Two hundred ninety-three consecutive outpatients seeking treat- sions-Severity scale [CGI-S][33] was used at T1 and T2 to assess ment for SP at an anxiety disorders clinic of a regional mental health participants’ global clinical condition. The CGI-S is a single center in the greater Tel-Aviv area were assessed between 1998 and item assessing severity of ‘‘illness’’ using a 7-point Likert scale 2002. The outpatient clinic provides treatment services free of charge.
ranging from ‘‘normal’’ to ‘‘very severely ill’’, and has good reliability Participants were either self-referred or referred by a medical doctor or mental health professional. Sixty-four individuals (22%) opted not Depression was assessed by the Montgomery and Asberg Depres- to participate in treatment. They cited interference with work/study, sion Rating Scale [MADRS][35] at T1 and T2. The MADRS is a 10- and/or participation in an active pharmacological or psychological item scale in which each item is rated on a scale of 0 (no evidence of treatment as their primary reasons for nonparticipation.
symptom) to 6 (pervasive evidence). The MADRS has good reliability Inclusion criteria for study participation were: (a) a current scores[36] and is sensitive to change during treatment.[35] diagnosis of SP according to the DSM-IV; (b) a minimum ofa 1-year duration of SP; (c) primary diagnosis of SP, that is, in cases with comorbidity, SP was deemed as the most distressing andclinically significant condition among the comorbid disorders; (d) Descriptive analyses. Measures of a continuous nature stable pharmacotherapy, that is, participants receiving a pharmaco- were compared using independent t tests. Measures of categorical or logical treatment who were taking a stable medication for at least 3 proportionate nature were compared between groups using w2 tests.
months before the beginning of CBGT; and (e) age between 18 and Bonferroni correction was used when computing multiple compar- 60 years. Exclusion criteria were: (a) past or present diagnosis of psychotic state and schizophrenia; (b) another psychotherapeutic Efficacy analyses. We used multiple analysis of variance treatment during CBGT, and (c) change in medication status during (MANOVA) to check treatment efficacy. Dependant variables were the CBGT. Ten individuals were excluded due to the above criteria.
the self-report measures used, and independent variables were Time Thus, at the onset of treatment the sample included 219 individuals.
(i.e. before treatment, after treatment, and follow-up) and Subtype It is important to note that none of the participants were excluded (GSP versus SSP, or depressed versus nondepressed participants).
due to severity of SP. Moreover, participants were aware of group size Multiple regression was used when predictors were continuous.
Diagnoses were determined by the Mini International Neuropsy- chiatric Interview and were conducted by Ph.D.-level clinical psychologists or experienced graduated psychology students trained in their administration. The sub-diagnosis of GSP was based onDSM-IV criteria. Participants were diagnosed as suffering from GSP Table 1 shows demographic, diagnostic, and if they demonstrated fears and avoidance in most social situations. If symptom severity levels for the complete sample. Out participants demonstrated anxiety or avoidance pertaining to one of the 219 participants at T1, 177 (80.8%) met or two specific social situations (e.g. public speaking, eating in diagnostic criteria for GSP and 42 (19.2%) met criteria public), they were diagnosed as suffering from SSP.
for SSP. The majority of the sample (56.2%) received A pilot study (n 5 20), conducted before the beginning of this one or more additional DSM-IV diagnoses: 32.4% study, examined the interviewers’ diagnostic validity and reliability bycomparing their diagnosis to those of a senior psychologist and a suffered from exactly one additional anxiety disorder, senior psychiatrist. The k coefficient was .92.
8.7% suffered from two or more additional anxietydisorders, 12.3% were diagnosed as having majordepressive disorder (MDD), and 11% were diagnosed The main outcome measure used in this study was the Liebowitz Pre-CBGT severity of SP, as measured by the LSAS, Social Anxiety Scale [LSAS].[26] This study used the self-report significantly correlated with pre-CBGT severity of version of the LSAS, which has been shown to have high internal depression as measured by the MADRS (r 5 0.40, consistency, strong convergent, and discriminant validity, and high Po0.001) and with the number of comorbid disorders test–retest reliability.[27,28] The Hebrew version of the self-report (r 5 0.18, Po0.01). Participants using psychotropic LSAS has been validated in previous research.[29] Outcome was drugs pre-CBGT and during CBGT (n 5 51) scored measured by changes in LSAS scores between pretreatment (T1) and significantly higher on the pre-CBGT LSAS compared posttreatment (T2), and maintenance of gains was measured by with non-medicated participants [81.1(22.6) versus changes in LSAS scores between posttreatment (T2) and 1-year 68.6(23.6), respectively, t(203) 5 À3.32, Po0.01].
Changes in the degree of impairment in work, social, and family areas were assessed using the Sheehan Disabilities Scale [SDS][30] at T1, T2, and T3. The SDS evaluates the degree of impairment inwork, social, and family areas on a 10-point Likert scale. Reliability Participants completed the LSAS twice before the and construct validity of the SDS are high[31] and it has strong beginning of treatment: at their initial interview (T0) correlations with symptoms of social anxiety, depressive symptoms, and at their first session (T1). The average waiting and quality of life among individuals with SP.[32] period for treatment was 4.8 months (SD 5 1.2). The TABLE 1. Demographic, diagnostic, and illness severity characteristics of GSPs and SSPs pre-CBGT: percentageor mean (standard deviation) an 5 165.
bn 5 38.
cRemains significant following Bonferroni correction.
GSPs, generalized social phobics; SSPs, specific social phobics; SP, social phobia; CBGT, cognitive behavioral group treatment.
LSAS scores at the two time points were almost elevated levels of depression compared with comple- identical [70.5 (SD 5 24.1) and 70.4 (SD 5 20.0) for T0 ters. No significant differences were found on any and T1, respectively] and the correlation between these other demographic or clinical characteristics.
two pre-CBGT LSAS scores was highly significant(r 5 0.82, Po0.001). These results mirror earlierfindings implying that SP symptoms no not subsidein the absence of treatment:[29] found test–retest reliability of the LSAS to be 0.87–0.91 over 7 weeks Table 2 shows pre-CBGT, post-CBGT, and follow- and[27] found it to be r 5 83, Po0.01 over a period of 3 up scores for GSPs and SSPs. To examine treatment effectiveness in GSP and SSP, we conducted a 2(Time:pre-CBGT versus post-CBGT) Â 2(Subtype: GSP versus SSP) MANOVAs with Time as a within-subjectfactor,and Subtype as a between subject factor.
The clinical characteristics of SSPs were significantly Dependent variables were LSAS, SDS, MADRS, and milder compared with GSPs (see Table 1). SSPs had a later age of onset [t(201) 5 À5.60, Po0.001], and fewer The main effect of Time was significant (Wilks’ comorbid diagnoses [w2(1) 5 5.20, Po0.05]. In addi- l 5 0.36, Po0.001) indicating that post-CBGT scores tion, a significantly higher percentage of SSP partici- were significantly lower than pre-CBGT scores on all pants were married [w2(1) 5 4.26, Po0.05]. SSPs scored significantly lower than GSPs on all severity Po0.001; SDS: F(1, 149) 5 171.6, Po0.001; MADRS; measures: LSAS [t(148) 5 À19.55, Po0.01], SDS [t(216) 5 À5.98, Po0.01], MADRS [t(216) 5 À5.02, 149) 5 203.8, Po0.001]. The effect size (ES; Cohen’s Po0.01], and CGI-S [t(214) 5 À7.28, Po0.01].
d) of CBGT (on LSAS scores) was 1.23. A main effectfor Subtype was also found (Wilks’ l 5 0.64, Po0.001) indicating that GSPs had higher scores than SSPs on Fifty-six participants (25.6%) dropped out of treat- all dependant variables [LSAS: F(1, 149) 5 82.7, ment. More men dropped out of treatment compared Po0.001; SDS: F(1, 149) 5 30.3, Po0.001; MADRS: with women [w2(1) 5 4.57, Po0.05] and more medica- F(1, 149) 5 13.3, Po0.001; CGI-S; F(1, 149) 5 33.8, tion users dropped out of treatment compared with Po0.001]. We also examined the possible effect of nonmedication users [w2(1) 5 4.3, Po0.05]. It is medication on treatment outcome. Medication use did important to note that drop-outs did not differ from not affect LSAS scores at T2 [t(164) 5 À1.48, P 5 0.14, completers on MADRS scores [P40.05, ns (not ns], nor did it affect improvement during treatment significant)] indicating that drop-outs did not have TABLE 2. Clinical measures among GSP and SSP completers of CBGT [presented as either means and (standarddeviations)/or percentages] aNumber of cases varies due to missing data.
GSPs, generalized social phobics; SSPs, specific social phobics; LSAS, Liebowitz Social Anxiety Scale; SDS, Sheehan Disabilities Scale; MADRS,Montgomery and Asberg Depression Rating Scale; CGI, Clinical Global Impression Severity of Illness; RCI, reliable change index. Good end-state, LSASo50 and SDSo3, and MADRSo8.
See results for statistical analysis.
A significant Time  Subtype interaction was found functioning compared to GSPs (79.3 versus 42.7%, (Wilks’ l 5 0.86, Po0.001), but when further explored w2(1) 5 11, Po0.001). As SSPs began treatment less was found to be significant only on LSAS and MADRS impaired than did GSPs, their absolute gains, although scores [F(1, 217) 5 13.8, Po0.001; F(1, 217) 5 4.9, smaller than those of GSPs, led to a better final good Po0.05; for LSAS and MADRS, respectively]. On the end-state. Importantly, no GSPs and 17.1% of SSPs fit SDS and CGI-S no interaction was identified (both the definition of good end-state functioning before Fo1). The interaction on LSAS and MADRS scores treatment (see Table 2). Even though 17.1% of SSPs was such that GSPs exhibited greater reductions in SP met requirements for good end-state before treatment, and depressive symptoms compared to SSPs. The ESs significantly more SSPs (79.3%) reached good end- of CBGT (on LSAS scores) were 1.54 for GSPs and 1.16 for SSPs. The ESs of CBGT (on MADRS scores), were 1.00 for GSPs, 0.64 for SSPs, and 0.94 for the We used[41] reliable change index to assess recovery and improvement in treatment based on means and The majority of participants who completed treat- standard deviations reported by.[28] According to ment (63.2% of GSPs and 60.0% of SSPs) improved by recommendations by,[41] we set the criterion for at least 30% on the LSAS, and 43.6% of GSPs and recovery to be two standard deviations above the mean 36.7% of SSPs improved at least by 50% on the LSAS.
of the functional population [LSAS scores lower than Severity of SP as assessed by the LSAS at termination 38.9 indicated recovery, see].[28] Percentages of indivi- of CBGT was aggravated in 6.4%, and in 6.6% of the duals achieving improvement and recovery are re- participants the improvement was o10%.
ported in Table 2 (for completer analysis) and in Good end-state functioning was conceptualized as Table 3 [for intent-to-treat (ITT) analysis].
return to nonpathological levels of distress[37] andoperationalized as LSAS o50, SDS o3 and MADRAS o12. These three cutoff points were chosen because they were found to be 1 SD above the average scores of Of 163 CBGT completers, 95 (58.3%) completed nonanxious and nondepressed individuals and therefore 1-year follow-up assessment. Table 2 presents data on well within the normal range [for LSAS scores in Israel, follow-up assessment for GSPs and SSPs.
see[38] for SDS scores,[39] for MADRS scores].[40] To examine changes between termination (T2) and According to this definition, at the end of CBGT, follow-up (T3), we conducted a 2(Time: post- versus significantly more SSPs reached good end-state follow-up) Â 2(Subtype: GSP versus SSP) MANOVA TABLE 3. Clinical measures of intent-to-treat sample using last-observation-carried-forward [presented as eithermeans and (standard deviations)/or percentages] GSPs, generalized social phobics; SSPs, specific social phobics; LSAS, Liebowitz Social Anxiety Scale; SDS, Sheehan Disabilities Scale; MADRS,Montgomery and Asberg Depression Rating Scale; CGI, Clinical Global Impression Severity of Illness; RCI, reliable change index. Good end-state 5 LSASo50 and SDSo3, and MADRSo8.
See results for statistical analysis.
As expected, GSPs had higher scores than SSPs onboth LSAS [F(1, 93) 5 15.6, Po0.001], and SDS [F(1,93) 5 8.2, Po0.01].
EFFECTS OF DEPRESSION ON TREATMENTOUTCOME AND GAINS MAINTENANCE In our sample, 23 participants were diagnosed with past MDD; 27 with current MDD; 24 with dysthymia;two were diagnosed as having both MDD (past orcurrent) and dysthymia. Both past MDD and dysthy-mia did not have significant effects on treatmentoutcome, on improvement during treatment, and ongains maintenance (all P 40.05).
Current MDD did not have a significant effect on most treatment outcome measures. Specifically,SPs with current MDD did not differ from SPs with- Figure 1. LSAS means for GSPs and SSPs. LSAS, Liebowitz out current MDD on LSAS at T2 [t(164) 5 À1.59, Social Anxiety Scale; GSP, generalized social phobia; SSP, P 5 0.11, ns], or CGI-S at T2 [t(159) 5 À1.37, specific social phobia; T1, pretreatment; T2, posttreatment; P 5 0.17, ns]. However, SPs with current MDD reported higher levels of impairment on the SDScompared with Time as a within-subject factor, and Subtype as a [t(163) 5 À2.36, Po0.05]. To further explore this between-subject factor. Dependent variables were difference, we examined the effect of current MDD LSAS and SDS scores. Main effects for both Time on each area of impairment included in the SDS. SPs (Wilks’ l 5 0.93, Po0.05), and Subtype (Wilks’ with current MDD had greater impairments at work l 5 0.86, Po0.01) were found suggesting a further [F(1, 162) 5 7.17, Po0.01] compared with SPs without decrease of SP severity. No significant Time  Subtype current MDD, but no differences were found in the interaction was found (Wilks’ l 5 0.98, P 5 0.461, ns).
social or family areas (both P40.05). In addition, At T3, both LSAS and SDS scores were significantly current MDD did not have an adverse effect on lower than at T2 [F(1, 93) 5 5.0, Po0.05, F(1, the improvement during treatment (pre-scores minus 93) 5 5.8, Po0.05, for LSAS and SDS, respectively], post-scores) as measured by LSAS [t(164) 5 0.06, indicating additional, albeit minor, improvement of P 5 0.95, ns], SDS [t(162) 5 0.16, P 5 0.87, ns], or symptoms following treatment termination (see Fig. 1).
CGI-S [t(159) 5 À0.52, P 5 0.61, ns]. As expected, mentioned effects of current MDD are due to anycomorbidity, or due to its depression component per se,we examined whether comorbidity with anxiety dis-orders would generate the same pattern of results. Weused a 2(Time: post-tx versus follow-up)  2(AnxietyComorbidity: none versus comorbid anxiety) MANO-VA with Time as a within-subject factor, and AnxietyComorbidity as a between-subject factor. In addition,current MDD was a covariate. Dependent variableswere LSAS and SDS scores. Results indicated no maineffect for Anxiety Comorbidity and no Time  AnxietyComorbidity interaction (all P40.05). Anxiety Comor-bidity did not significantly affect maintenance of gains,as opposed to Current MDD.
When analyzed as a continuous variable using Figure 2. LSAS means for participants with MDD and without regression analysis, depression (but not SP symptoms) MDD. LSAS, Liebowitz Social Anxiety Scale; History of at the onset of treatment predicted exacerbation of SP Depression, past or current MDD, major depressive disorder at follow-up. Specifically, baseline MADRS scores or dysthymia; T1, pretreatment; T2, posttreatment; T3 5 1-year predicted SP symptom aggravation between posttreat- ment and follow-up, whereas baseline LSAS scores didnot (MADRAS: T1-b 5 À0.23, Po0.05, but LSAS:T1-b 5 À0.15, P 5 0.17, ns).
current MDD had a positive effect on improvement ofdepressive [t(151) 5 À3.34, Po0.01] (i.e. SPs with current MDD Table 3 contains clinical measures for the intent-to- showed more improvement in depressive symptoms treat (ITT) sample using the last-observation-carried- compared with SPs without current MDD).
forward procedure. We repeated the effectiveness Next, we examined the effect of current MDD on analyses and follow-up analyses using the ITT sample.
gains maintenance. A repeated measures MANOVA Results were identical to those reported for the with Time (post-tx vs. follow-up) as a within-subject completer sample. However, in the ITT sample only variable and Current MDD as between-subject variable 33.3% of GSPs and 26.2% of SSPs had more than 50% revealed a significant Time  Current MDD interac- decrease in LSAS scores at termination of treatment. At tion (Wilks’ l 5 0.85, Po0.01), and a main effect for follow-up 35.6% of GSPs and 47.6% of SSPs had more Current MDD (Wilks’ l 5 0.88, Po0.01). The main effect of Current MDD was found to be significant forboth LSAS [F(1, 93) 5 7.8, Po0.01] and SDS [F(1,93) 5 12.6, Po0.01] measures. As expected, SPs with current MDD reported higher social anxiety and This study investigated the impact of depression on impairment compared with SPs without current MDD.
effectiveness of CBGT for SP in a large naturalistic The Time  Current MDD interaction was found to sample. Completers were also assessed 1 year after be significant on both LSAS scores [F(1, 93) 5 14.9, treatment termination and potential predictors of gains Po0.001]; and on SDS scores [F(1, 93) 5 9.5, Po0.01] separately. The interaction was such that on bothmeasures, the scores of SPs with current MDDsignificantly increased from termination to follow-up [t(12) 5 À2.2, Po0.05], whereas SPs without current MDD experienced a further alleviation of their In a recent paper,[42] proposed and tested seven symptoms [t(82) 5 3.4, Po0.01; Fig. 2].
criteria for distinguishing between efficacy (explana- To check whether this effect of current MDD on tory) studies and effectiveness (pragmatic) studies.
maintenance of gains was not due to current MDD Efficacy studies are conducted in ideal conditions with being correlated with comorbidity with other anxiety high internal validity, whereas effectiveness studies are disorders, we repeated the above analysis with anxiety conducted under ‘‘real world’’ conditions emphasizing comorbidity as a covariate. Results were identical in external validity. The first criterion for an effectiveness nature: a significant main effect for current MDD was (naturalistic) study is treating or examining populations found (Wilks’ l 5 0.87, Po0.01), and a significant in primary clinical settings. As our clinic is located in a Time  Current MDD interaction was found (Wilks’ primary clinical setting and offers treatment free of l 5 0.85, Po0.01). The significant current MDD effect charge, we believe the first criterion is satisfied. The remained even after controlling for comorbidity of second criterion is less stringent eligibility. In this other anxiety disorders. To examine if the above- study, only 10 individuals were excluded as a result of our exclusion criteria. The inclusion/exclusion ratio also mirror those reported by,[20] who found no was 21.9, meaning that one person was excluded for detrimental effects of depression on treatment outcome every 22 that were included. This ratio is much higher and on improvement during treatment. Again, con- than in other studies (e.g. [43] with an inclusion/ sistent with our results,[20] also found that among exclusion ratio of 2) and therefore fulfills the criterion depressed individuals, depression declined during of less stringent eligibility. Our study also fulfills the treatment but remained elevated compared with non- criterion of measuring health outcomes as the SDS and depressed individuals. These results can be understood CGI are used, and the criteria of long study duration, when considering the interplay between anxiety and adequate sample size, and ITT analysis. The only depression during treatment of SP.[47] reported that criterion that our study does not fulfill is the during CBGT, changes in anxiety account for 91% of measurement of adverse events during treatment. To changes in depression, but changes in depression sum, our study fulfils six out of seven criteria of account for only 6% of changes in anxiety. These data effectiveness studies and is much closer to the suggest that depression reduction is primarily an naturalistic prototype than to the randomized con- outcome of social anxiety reduction among individuals with SP, and that focally treating SP results in a The mean LSAS score of our sample (70.5) was high, or higher than the mean scores in clinical controlled However, our results are at odds with several other trials in the literature. For instance,[44] aggregated studies investigating the interplay between depression results from four clinical SP samples and found an and social anxiety during treatment. For instance,[15] average LSAS score of 67.2 before treatment. In found that depression was a predictor of diminished another study by[27], the average LSAS score among improvement during treatment. A possible reason for 175 diagnosed but untreated participants was 69.1.
this discrepancy may be that the outcome measures These data suggest that the present sample, which is used by[15] were mainly self-reports of performance in naturalistic in nature, is comparable to samples reported behavioral tasks compared to questionnaires assessing in clinical randomized studies in terms of SP severity.
both anxiety and avoidance used in this study. In arecent study,[21] found lower improvement rates fordepressed individuals. However, a diagnosis of MDD was an exclusion criterion in the[21] study and may have restricted the range of depressive symptoms examined.
The majority of SPs completing CBGT experienced In our sample, the ES for anxiety was 1.54 for GSPs a clinically significant improvement at the end of and 1.16 for SSPs, and ESs for depression were 1.00 for treatment, and these effects were maintained at 1-year GSP and 0.64 for SSP. These ESs for SP symptoms follow-up. In our naturalistic sample, participants were somewhat larger than those reported in previous improved irrespective of baseline severity of SP and meta-analyses [0.80–0.90 in [48–50]]. Our ESs were also irrespective of comorbidity of SP with other anxiety or higher than those reported in naturalistic settings by[14] mood disorders. Thus, even SPs with serious comor- (0.71–0.88 for SP measures) and by[13] (1.0–1.2 for SP bidity of anxiety and depression can markedly benefit measures). When compared with other group treat- from CBGT. It is important to note that our study ments our ESs were also high [e.g. [51]—0.53–0.60 for contained a naturalistic control condition. The waiting SP measures, [52]—0.56 for SP measures]. However, period for treatment averaged 18 weeks, a duration our ESs were lower than those reported in recent similar to the active treatment length. During this controlled treatment outcome studies using individual period, participants’ SP symptoms remained un- format [e.g. 2.14 in[43] 2.63 in[45]].
changed. Thus, the improvement in treatment cannot Despite these relatively large ESs, it is somewhat sobering to realize that only about 50% GSPs Although CBGT focally targeted social anxiety, completing the treatment reached good end-state participants exhibited a significant simultaneous de- functioning. If we take into account the ITT sample, crease in depressive symptoms at treatment termina- this figure goes down to approximately 35%. This tion. Several studies found a similar therapeutic effect general pattern of considerable improvement during e.g.[45,46] Namely, CBGT for SP resulted in improve- treatment on one hand, and marked residual sympto- ment in a wide range of distress symptoms. In addition, matology on the other, has been found in previous depressed individuals were identical to nondepressed research[12]. When using reliable change index criter- individuals at treatment completion on all measures but ia[41] results were similar: 56% of completers achieved one (higher impairment at work, as measured by a significant improvement and 52% were categorized as subscale of the SDS). In other words, depression did recovered. When considering the ITT sample only not have a detrimental effect on treatment outcome, 43% achieved significant improvement and 41% were nor did it affect the degree of improvement during categorized as recovered. These rates are very similar treatment. These results are in line with,[19] who found to the ones reported by[14] in a naturalistic setting: 56% that comorbid depression does not result in differential reliably improving and 57% categorized as recovered.
rates of improvement during CBGT of SP. Our results Our rates of improvement are also slightly higher than the ones reported by[51]: 36–44% of patients achieving groups included 50–100% more participants than reliable improvement on SP measures after group groups in recent studies [e.g. [58]—6–8 participants; treatment. Therefore, our results are comparable with [51]—4–7 participants]. Perhaps, this may have caused other naturalistic and group treatments.
participants to withdraw prematurely from CBGT.
Third, our higher attrition rate may be due to a stricter definition of dropping out (e.g. lack of participation inthe final session counted as a drop-out due to Consistent with the literature, our data show that procedural reasons). Fourth, ours was a naturalistic GSP and SSP subgroups differed on most demographic setting with a minimal exclusion rate and hence it and on most psychopathology measures: GSPs had an probably reflects a more realistic public outpatient–- earlier age of onset; they presented graver indices of SP and depression at pretreatment and posttreatment; and Participants who dropped out of CBGT did not had a higher prevalence of both anxiety and mood differ from completers of CBGT on any demographic disorders [for similar findings, see].[3,53] Taken to- or clinical measure (including symptoms of SP and gether, these findings are consistent with the notion depression) except for a higher percentage of men and that GSP and SSP are distinct nosological cate- medication users among drop-outs. This indicates that our sample of completers is likely to be representative Although both GSP and SSP participants improved of treatment seeking SPs. A similar conclusion was significantly, a higher percentage of SSP participants, reached by,[59] who compared drop-outs to treatment compared with GSP participants, reported symptoms completers in CBGT for SP and found only minor in the nonclinical range at termination of CBGT. Put differences between the two groups. They conclude differently, although there was a substantial improve- that it is unlikely that drop-outs pose a serious threat to ment from pretreatment for the majority of partici- external validity in CBGT for SP. Regrettably, the pants, a large portion of GSPs retained residual dropout participants in our sample were not assessed at symptomatology that did not constitute a return to the time of their early termination, so it is not clear normal mental health and functioning. It is important whether their dropping out was associated with an to note that more GSPs (61.8%) improved compared adequate improvement or aggravation of symptoms.
with SSPs (30.0%) but that less GSPs (44.1%) were Obviously, more frequent assessments along the course considered recovered compared to SSPs (86.7%). This of CBGT would partially resolve this issue.
pattern of results is due to the fact that GSPs weremore impaired than SSPs at the onset of treatment.
Many GSPs were able to achieve greater improvement during treatment on the one hand but not sufficient Sixty-three percent of completers of CBGT re- improvement to be categorized as recovered on the sponded to the follow-up questionnaires. No differ- other. SSPs improved less but were closer to the ences were detected between follow-up responders and definition of recovery to begin with, and so achieved nonresponders on any measure, rendering the respon- recovery more easily. These findings suggest that GPSs ders representative of CBGT completers. Many studies may require more intensive, individual, and prolonged have found that most participants extend treatment interventions (either psychological and psychopharma- gains over follow-up periods [e.g.].[45,51,57,60] In con- cological) than do SSPs. However, SSPs might benefit gruence with these studies, completers of our CBGT from a concise group format, in which a more targeted, maintained or improved the gains measured at less extended, and hence less costly course of therapy is termination of CBGT at a 1-year follow-up. The maintenance and even augmentation of the treatmentresponse following CBGT represents an advantage over pharmachotherapy to SP, which is associated with The attrition rate in our study was somewhat higher a significant likelihood of relapse when the drug is than the average attrition reported in the literature [25.6 versus 18.6% in a meta-analysis of 21 CBT Our study is the first to examine the influence of studies].[48] However, many nonnaturalistic studies depression at intake on the maintenance of gains report similar attrition rates [e.g. [55]—reporting a during extended follow-up periods. Indeed, we found 22% attrition rate for an exposure condition;[56]— that although depression did not affect the degree of reporting a 31% attrition rate;[57]—reporting a 25% improvement at CBGT termination, it significantly attrition rate]. In this study, several factors might be interfered with the maintenance of treatment gains. In responsible for raising the attrition rate above the rate fact, MDD was a better predictor of gains maintenance common in controlled studies [yet maintaining it well than comorbid anxiety or SP severity at intake. Several within the attrition rates of naturalistic studies].[14] mechanisms may be responsible for this pattern. First, First, our relatively long CBGT course (18 sessions) it is possible that the increase in social anxiety might have increased the difficulty to complete symptoms is due to the reoccurrence of another treatment and thus increased attrition. Second, our depressive episode in depression-prone participants.
The occurrence of a depressive episode is typically Acknowledgments. The authors thank I. Gelern- associated with an increase in social isolation, which in ter M.A. and Prof. Y. Benjamini, from the Statistical turn may increase symptoms of social anxiety. Second, Laboratory, Sackler School of Mathematics, Tel Aviv it may be that comorbid depression and SP represents a University, for the statistical consultation, and Gadi more severe type of mental disorder,[21] and therefore Presburger, Ph.D., and Yael Vaknin, Ph.D., and Dan individuals afflicted by it are more susceptible to full or Raveh, Ph.D., for their help in data collection.
partial relapse. Indeed, individuals with comorbidMDD present a more severe clinical picture at the onset of CBGT than do nondepressed individuals.
Although depressed SPs may be capable of modifying 1. Furmark T. Social phobia: overview of community surveys. Acta their social behavior and regulatory strategies while in active CBGT, the presumed rigidity of their cognitive 2. Chartier MJ, Walker JR, Stein MB. Considering comorbidity in schemas might interfere with the implementation of social phobia. Soc Psychiatry Psychiatr Epidemiol 2003; newly learned skills after termination.
3. Kessler RC, Stein MB, Berglund P. Social phobia subtypes in the National Comorbidity Survey. Am J Psychiatry 1998;155:613–619.
4. deWit DJ, Ogborne A, Offord DR, MacDonald K. Antecedents Our study suffered from several limitations. First, of the risk of recovery from DSM-III-R social phobia. Psychol treatment gains were not assessed blindly and were American Psychiatric Association. Diagnostic and statistical comprised of self-report measures and clinician ratings.
manual of mental disorders, Fourth Edition, Text Revision.
Self-report instruments may be susceptible to partici- Washington, DC: American Psychiatric Association; 2000.
pants’ and clinicians’ expectations regarding improve- 6. Blanco C. Evidence-based pharmacotherapy of social anxiety ment. Future studies should make use not only of self- disorder. CNS Spectr 2005;10:10–11.
report and clinician ratings, but also of behavioral and 7. Radomsky AS, Otto MW. Cognitive behavioral therapy for social implicit measures. Second, only data pertaining to anxiety disorder. The Psychiatr Clin N Am 2001;24:805–815.
social anxiety and depression were collected. Our 8. Heimberg RG, Becker RE. Treatment of social fears and knowledge of treatment outcome would be enhanced phobias. New York: Guilford Press; 2002.
if additional parameters of psychological distress (e.g.
9. Hofmann SG, Bo¨gels SM. Recent advances in the treatment of measures of anxiety) as well as broader measures of social phobia: introduction to the special issue. J CognPsychother 2006;20:3–5.
interpersonal functioning (e.g. satisfaction with inter- 10. Turk CL, Heimberg RG, Hope DA. Social anxiety disorder. In: personal relationships, cognitions, and attitudes per- Barlow DH, editor. Clinical handbook of psychological dis- taining to one’s social functioning) were assessed.
orders, 3rd ed. New York: The Guilford Press; 2001.
Third, in this study we lack the knowledge regarding 11. Blomhoff S, Haug TT, Helstro¨m K, Holme I, Humble M, the main life events and (possibly) additional therapeu- Madsbu HP, Wold JE. Randomized controlled general practice tic interventions that may have occurred between trial of sertraline, exposure therapy and combined treatment in treatment termination and follow-up. Such data would generalized social phobia. Br J Psychiatry 2001;179:23–30.
be important in furthering our understanding regard- 12. Davidson JR, Foa EB, Huppert JD, Keefe FJ, Franklin ME, ing the factors contributing to changes in emotional Compton JS, Zhao N, Connor KM, Lynch TR, Gadde KM.
distress following treatment termination.
Fluoxetine, comprehensive cognitive behavioral therapy, andplacebo in generalized social phobia. Arch Gen Psychiatry2004;61:1005–1013.
13. Gaston JE, Abbott MJ, Rapee RM, Neary SA. Do empirically supported treatments generalize to private practice? A bench- CBGTwas an effective treatment for chronic SP, and mark study of a cognitive-behavioural group treatment pro-gramme for social phobia. Br J Clin Psychol 2006;45:33–48.
its therapeutic gains were maintained and even slightly 14. Lincoln TM, Rief W, Hahlweg K, Frank M, von Witzleben I, improved at a 1-year follow-up. CBGTwas found to be Schroeder B, Fiegenbaum W. Effectiveness of an empirically effective in individuals with comorbid disorders. SSPs supported treatment for social phobia in the field. Behav Res present a more benign clinical picture than do GSPs on most measures of psychopathology. More SSPs reached 15. Chambless DL, Tran GQ, Glass CR. Predictors of response to good-end state functioning than GSPs. It appears that cognitive-behavioral group therapy for social phobia. J Anxiety SSPs could benefit from a shorter and more focused CBGT. GSPs, in contrast, may need a longer course of 16. Hope DA, Herbert JD, White C. Diagnostic subtype, CBGT, and smaller groups than ours, which could avoidant personality disorder, and efficacy of cognitive behavioral include additional components such as an individual group therapy for social phobia. Cogn Ther Res 1995;19:399–417.
treatment or booster sessions. In addition, individuals 17. Turner SM, Beidel DC, Wolff PL, Spaulding S, Jacob RG.
suffering from both SP and depression experience an Clinical features affecting treatment outcome in social phobia.
aggravation in SP symptoms after treatment termina- tion, and may need follow-up treatment to prevent 18. Brown EJ, Heimberg RG, Juster HR. Social phobia subtype and exacerbation of their symptoms at long term.
avoidant personality disorder: effect on severity of social phobia, impairment, and outcome of cognitive-behavioral treatment.
37. Feeny NC, Zoellner LA, Foa EB. Treatment outcome for chronic PTSD among female assault victims with borderline 19. Erwin BA, Heimberg RG, Juster H, Mindlin M. Comorbid personality characteristics: a preliminary examination. J Personal anxiety and mood disorders among persons with social anxiety disorder. Behav Res Ther 2002;40:19–35.
38. Iancu I, Levin J, Hermesh H, Dannon P, Poreh A, Ben-Yehuda Y, 20. Joormann J, Kosfelder J, Schulte D. The impact of comorbidity Kaplan Z, Marom S, Kotler M. Social phobia symptoms: of depression on the course of anxiety treatments. Cogn Ther prevalence, sociodemographic correlates, and overlap with specific phobia symptoms. Compr Psychiatry 2006;47:399–405.
21. Ledley DR, Huppert JD, Foa EB, Davidson JRT, Keefe FJ, Potts NLS. Impact of depressive symptoms on the treatment of depressive disorders and their comorbidity: effect on central nervous system noradrenergic function. Biol Psychichiatry 2004; 22. Scholing A, Emmelkamp PMG. Prediction of treatment out- 40. Von-Knorring AL, Olsson GI, Thomsen PH, Lemming OM, come in social phobia: a replication. Behav Res Ther 1999; Hulten A. A randomized, double-blind, placebo-controlled study of citalopram in adolescents with major depressive disorder. J 23. Sheehan DV, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Clin Psychopharmacol 2006;26:311–315.
Weiller E, Hergueta T, Baker R, Dunbar GC. The Mini 41. Jacobson NS, Truax P. Clinical significance: a statistical approach International Neuropsychiatric Interview (M.I.N.I.): the devel- to defining meaningful change in psychotherapy research. J opment and validation of a structured diagnostic psychiatric Consult Clin Psychol 1991;59:12–19.
interview for DSM-IV and ICD-10. J Clin Psychiatry 1998; 42. Gartlehner G, Hansen RA, Nissman D, Lohr KN, Carey KS. A simple and valid tool distinguished efficacy from effectiveness 24. Heimberg RG, Juster HR, Hope DA, Mattia JI. Cognitive- studies. J Clin Epidemiol 2006;59:1040–1048.
behavioral group treatment: description, case presentation, and 43. Clark DM, Ehlers A, McManus F, Hackmann A, Fennell M, empirical support. In: Stein MB, editor. Social phobia: clinical Campbell H, Flower T, Davenport C, Louis B. Cognitive and research perspectives. Washington, DC: American Psychia- therapy versus fluoxetine in generalized social phobia: a randomized placebo-controlled trial. J Consult Clin Psychol 25. Clark DM, Wells A. A cognitive model of social phobia. In: Heimberg RG, Liebowitz MR, Hope DA, Schneier FR, editors.
44. Heimberg RG, Horner HR, Juster SA, Safran SA, Brown FR, Social phobia: diagnosis, assessment, and treatment. New York: Liebowitz MR. Psychometric properties of the Liebowitz Social Anxiety Scale. Psychol Med 1999;29:199–212.
26. Liebowitz MR. Social phobia. Mod Probl Pharmacopsychiatry 45. Clark DM, Ehlers A, Hackmann A, McManus F, Fennell M, Grey N, Waddington L, Wild J. Cognitive therapy versus 27. Baker SL, Heinrichs N, Kim HJ, Hofmann SG. The Liebowitz exposure and applied relaxation in social phobia: a randomized Social Anxiety Scale as a self-report instrument: a preliminary controlled trial. J Consult Clin Psychol 2006;74:568–578.
psychometric analysis. Behav Res Ther 2002;40:701–715.
46. Haug TT, Blomhoff S, Hellstrom K, Holme I, Humble M, 28. Fresco DM, Coles ME, Heimberg RG, Liebowitz MR, Hami S, Madsbu HP, Wold JE. Exposure therapy and sertraline in social Stein MB, Goetz D. The Liebowitz Social Anxiety Scale: a phobia: 1-year follow-up of a randomized controlled trial. Br J comparison of the psychometric properties of self-report and 47. Moscovitch DA, Hofmann SG, Suvak MK, In-Albon T.
Mediation of changes in anxiety and depression during treatment 29. Levin JB, Marom S, Gur S, Wechter D, Hermesh H.
Psychometric properties and three proposed subscales of a self- of social phobia. J Consult Clin Psychol 2005;73:945–952.
report version of the Liebowitz Social Anxiety Scale translated 48. Federoff IC, Taylor S. Psychological and pharmacological into Hebrew. Depress Anxiety 2002;16:143–151.
treatments for social phobia: a meta-analysis. J Clin Psycho- 30. Sheehan D. The anxiety disease. New York: Scribner; 1983.
31. Leon AC, Olfson M, Portera L, Farber L, Sheehan DV.
49. Feske U, Chambless DL. Cognitive behavioral versus exposure Assessing psychiatric impairment in primary care with the only treatment for social phobia: a meta-analysis. Behav Ther Sheehan Disability Scale. Int J Psychiatry Med 1997;27:93–105.
32. Hambrick JP, Turk CL, Heimberg RG, Schneier FR, Liebowitz 50. Gould RA, Buckminster S, Pollack MH. Cognitive-behavioral MR. Psychometric properties of disability measures among and pharmacological treatment for social phobia: a meta-analysis.
patients with social anxiety disorder. J Anxiety Disord 2004; Clin Psychol Sci Pract 1997;4:291–306.
51. Stangier U, Heidenreich T, Peitz M, Lauterbach W, Clark DM.
33. Guy W. ECDEU Assessment Manual for Psychopharmacology, Cognitive therapy for social phobia: individual versus group Revised. Rockville: US Dept of Health, Education and Welfare therapy. Behav Res Ther 2003;41:991–1007.
52. Mo¨rtberg E, Karlsson A, Fyring C, Sundin O 34. Dahlke F, Lohaus A, Gutzmann H. Reliability and clinical cognitive-behavioral group treatment (CBGT) of social phobia: concepts underlying global judgments in dementia: implications a randomized controlled study. J Anxiety Disord 2006; for clinical research. Psychopharmacol Bull 1992;28:425–432.
35. Montgomery SA, Asberg MA. A new depression scale 53. Brown TA, Cambell LA, Lehman CL, Grisham GR, Mancill RB.
designed to be sensitive to change. Br J Psychiatry 1979; Current and lifetime comorbidity of the DSM-IV anxiety and mood disorders in a large clinical sample. J Abnorm Psychol 36. Davidson J, Turnbull CD, Strickland R, Miller R, Graves K. The Montgomery-Asberg Depression Scale: reliability and validity.
54. Hook J, Valentiner D. Are specific and generalized social phobias Acta Psychiatr Scand 1986;74:544–548.
qualitatively distinct? Clin Psychol Sci Pract 2002;9:379–395.
55. Smits JA, Powers MB, Buxkamper R, Telch MJ. The efficacy of 59. Hofmann SG, Suvak M. Treatment attrition during group videotape feedback for enhancing the effects of exposure-based therapy for social phobia. J Anxiety Disord 2006;20:961–972.
treatment for social anxiety disorder: a controlled investigation.
60. Liebowitz MR, Heimberg RG, Schneier FR, Hope DA, Davis S, Holt CS, Goetz D, Juster HR, Lin SH, Bruch MA, Marshall RD, 56. Herbert JD, Gaudiano BA, Rheingold AA, Myers VH, Dalrym- Klein DF. Cognitive-behavioral group therapy versus phenelzine ple K, Nolan EM. Social skills training augments the effective- in social phobia: long term outcome. Depress Anxiety ness of cognitive behavioral group therapy for social anxiety disorder. Behav Ther 2005;36:125–138.
61. Blanco C, Raza MS, Schneier FR, Liebowitz MR. The evidence- 57. Otto MW, Smiths JA, Reese HE. Cognitive-behavioral therapy based pharmacological treatment of social anxiety disorder. Int J for the treatment of anxiety disorders. J Clin Psychiatry Clin Neuropsychopharmacol 2003;6:427–442.
62. Montgomery SA, Nil R, Du¨rr-Pal N, Loft H, Boulenger JP. A 58. Rosser S, Erskine A, Crino R. Pre-existing antidepressants 24-Week randomized, double-blind, placebo-controlled study of and the outcome of group cognitive behaviour therapy for social escitalopram for the prevention of generalized social anxiety phobia. Aust N Z J Psychiatry 2004;38:233–239.
disorder. J Clin Psychiatry 2005;66:1270–1278.

Source: http://www.ruppin.ac.il/download/files/%20on%20treatment%20effectiveness%20and%20gains%20maintenance%20in%20social%20phobia-%20a%20naturalistic%20study%20of%20cognitive%20behavior%20group%20therapy.pdf