Spsc_a_442893_o.pdf

Pre-treatment patient variables as predictors of drop-out and treatment outcome in cognitive behavioural therapy for social phobia: A systematic review ANITA ESKILDSEN, ESBEN HOUGAARD, NICOLE K. ROSENBERG Eskildsen A, Hougaard E, Rosenberg NK. Pre-treatment patient variables as predictors of drop-out and treatment outcome in cognitive behavioural therapy for social phobia: A systematic review. Nord J Psychiatry 2009;63:000.
Background: Although cognitive behavioural therapy (CBT) has been shown to be an efficacious treatment for social phobia (SP), many patients drop out or achieve little or no benefit from treatment. This fact is generally considered an argument for the importance of studies of predic-tor variables. Aims: This paper systematically reviews pre-treatment patient variables as predictors of drop-out from and outcome of CBT for SP. Method: A structured literature search was con-ducted in PsycINFO, Embase and PubMed. Results: 28 published studies with nr60 were located. No pre-treatment patient variables were found to predict drop-out. Consistently across studies, higher levels of pre-treatment symptomatic severity predicted higher levels of end-state symptomatic severity, but not degree of improvement. There was some evidence that comorbid depression and avoidant personality disorder before treatment negatively influenced post-treatment end-state functioning, but not consistently improvement. No other patient variables consistently predicted outcome across studies. Conclusions: Generally, the results are in line with the conclu-sion that more disturbed patients with SP both begin and end treatment at a higher symptomatic level but with a similar degree of improvement. There is, however, little clinically or theoretically relevant knowledge to be gained from existing studies of pre-treatment patient variables as pre-dictors of drop-out and treatment outcome in CBT for patients with SP. The field is in need of conceptual and methodological improvements if more solid findings should be hoped for.
u Cognitive behavioural therapy, Drop-out, Patient variables, Predictive variables, Prognostic variables, Social phobia Anita Eskildsen, Cand.psych., Institute of Psychology, Clinic of OCD, Aarhus University Hospital, Risskov Nobelparken, Jens Chr. Skous Vej 4, 8000 Aarhus C, Denmark, E-mail: eskildsen.anita@gmail.com; Accepted 7 October 2009.
Nord J Psychiatry Downloaded from informahealthcare.com by Statsbiblioteket Tidsskriftafdeling Social phobia (SP) or social anxiety disorder is char- If untreated, the disorder tends to run a chronic acterized by a marked and persistent fear of social or course and to be associated with lower work productivity, performance situations in which embarrassment may impaired functioning in social and romantic relationships, occur (1). Epidemiological studies suggest lifetime prev- financial dependency, comorbid psychiatric disorders alence rates between 7% and 13% in most western coun- and poor quality of life (4–6). Mental disorders comor- tries based on DSM-III-R or DSM-IV (2). However, the bid with SP especially include major depression, gener-figures vary considerably in different studies; probably alized anxiety disorder, personality disorders and alcohol mainly because of different cut-off lines for clinical abuse.
cases, since the required degree of distress or functional Cognitive behavioural therapy (CBT) is the most impairment is not specified in the diagnostic systems. A widely used psychological treatment for SP (7, 8). CBT new, very large epidemiological study in six different for SP is often administered in a group format and usu-European countries found a lifetime prevalence estimate ally combines exposure exercises with cognitive restruc-for SP of only 2.4% (3).
turing, although other methods such as applied relaxation, 2009 Informa UK Ltd. (Informa Healthcare, Taylor & Francis As) rational emotive therapy and social skills training are treatment outcomes). The databases descriptors were used also included under the CBT heading.
when possible. In addition, the articles located were CBT has been shown to be an efficacious treatment inspected for further relevant references.
for SP. Effect sizes of outcome for CBT derived from
meta-analyses are high (9, 10), e.g. Cohen’s (11) d1.27 Inclusion criteria
(pre- to post-treatment) in a meta-analysis of randomized Studies included in the review examined pre-treatment studies (9), although the effect sizes reported in meta- patient variables as predictors of drop-out rate and/or analyses including non-randomized studies comparing treatment outcome in CBT. The inclusion criteria were: with placebo control conditions or using intention- (a) published studies (b) in English, German or French to-treat analyses have been considerably lower (5, 12). language (c) that examined pre-treatment patient variables The outcome of CBT has been found to be comparable as predictors of outcome and/or drop-out in (d) CBT, e.g. with pharmacological interventions, but with more dura- cognitive therapy, social skills training, exposure therapy, ble effects of CBT (4, 13). Many patients with SP, or applied relaxation with (e) adult patients (r18 years) however, do not respond to CBT, and the proportion of with (f) a formal diagnosis of SP (DSM-III or later ver- patients who drop out during treatment is high (10–20%) sion) and (g) a sample size of at least 60. With a sample (9, 10, 14). When drop-out rates are included, clinical size of 60 and a moderate effect size (r0.30), the power trials suggest that 40–50% of patients with SP show of a study conducting correlation analysis is 0.65 little or no improvement after CBT (8, 9, 15). Because (A0.05, two-tailed) (11).
of this, there has been an upsurge of interest in prog- Excluded were dissertation abstracts, and studies nostic patient variables in CBT for SP. Empirical studies where different treatment conditions other than CBTs on patient variables relevant for prediction of risk of drop- (e.g. medication) were aggregated in the analyses, unless out or low benefit from treatment seem to be both theo- controlling for differential effects of the predictors in the retically and clinically informative. Knowledge about respective conditions.
which clients are likely to fail in therapy, and why they
would do so, might help in modifying treatment strate-
gies and in delineating critical variables for matching Data analysis
clients to the most suitable treatment programme (16). The varied statistical procedures used in the studies do
In their declaration on evidence-based practice in psy-
not allow for the calculation of a pooled effect size for chology, the American Psychological Association (17) the degree of association between variables. Therefore, suggests that studies of patient characteristics as moder- the box-score method with counts of statistical signifi- ators of treatment response should be among future cant results is used. Most of the studies predicting out-research priorities. Predictive patient variables associated come used several dependent variables and/or performed with outcome in a specific treatment might imply that several statistical analyses on the same variable without the variables are functioning as moderators even though statistical correction. In these cases, it was decided to no formal test of patient×treatment interaction (e.g. 18) count a result as substantial supportive evidence if at has been performed.
least half of the investigated associations were signifi- Zaider & Heimberg (19), Rodebaugh et al. (10) and cant. The results are also expressed as the means of the Lincoln et al. (20) have reviewed the literature on pre- percentages of the examined associations within each dictors of response to CBT for SP, although neither of study that were statistical significant. Associations in the these are systematic reviews with comprehensive litera- opposite direction of the general trend in the studies Nord J Psychiatry Downloaded from informahealthcare.com by Statsbiblioteket Tidsskriftafdeling ture search reported in the articles.
were subtracted in these calculations. Data are analysed The aim of this study is to conduct a systematic liter- separately for prediction of drop-out and for prediction ature review of pre-treatment patient variables associated of treatment outcome. With regard to treatment outcome, with the outcome of and drop-out from CBT in the treat- a distinction is made between prediction of 1) end-state functioning or responder status on outcome measures after treatment or at follow-up (if the authors only reported a group effect in an ANOVA analysis without a post hoc analysis regarding post-treatment differences, Search strategies
the result was not incorporated in the analysis of end- Journal articles were located in PsycINFO, Embase and state functioning), and 2) degree of change or improve-PubMed from the beginning of the databases to November ment from pre- to post-treatment or from pre-treatment 2008. Meta-analyses and literature reviews were excluded to follow-up. Decline in symptom level measured over in the search in PsychINFO. The keywords were (SP or the course of treatment by statistical growth trajectories social anxiety disorder) and (cognitive behaviour therapy (e.g. hierarchical linear modelling) is categorized as or patient drop-outs or predictor variables or prognosis or degree of improvement.
Altogether, the studies included 2927 patients. In most The three literature searches in PsycINFO, Embase and of the studies reviewed, at least one of the treatment for-PubMed yielded 268, 715 (627 further hits not located in mats administered was group CBT using the protocol of the PsycINFO search) and 580 (355 further hits not Heimberg and colleagues (e.g. 21) or a similar protocol located in the PsychINFO or Embase searches) hits, (e.g. comprehensive CBT which includes social skills respectively. After inspection of abstracts, 46 articles were training; 22, 23). Almost half of the studies used more retrieved for more detailed evaluation from PsycINFO, than one treatment condition, e.g. including both individ-30 additional articles from Embase, 16 additional articles ual and group therapy. All studies but one (24) incorpo-from PubMed and three articles were located from refer- rated some form of exposure exercises in the treatment. ence lists. Of these 95 references, 67 were excluded and All studies used DSM-III-R or a later edition of DSM.
28 included in the review, with 16 studies examining pre-
dictors of drop-out (Table 1) and 25 predictors of treat-
ment outcome (Table 2). Tables 1 and 2 are only published Comorbidity
in the online version of the journal at URL: http://www.
informahealthcare.com/10.3109/08039480903426929. A list Drop-out. Three studies in Table 1 examined if a diagnosis of excluded studies is also available online.
of major depressive disorder (MDD) predicted drop-out Table 1. Studies examining pre-treatment patient variables as predictors of treatment drop-out.
Severity (PAS, SIB, SCL-90, FQ, Clinician rated avoidance); Rational cognition (RBI and self-statements test); Behavioral test (SSIT) Treatment expectancy after the first session (RTC) ); Additional anxiety or mood disorders; Treatment expectations (the scale is not reported); Demographic variables (Age, Gender, Marital status, Education, Employment Status, Duration of complaints, Previous treatment received); Psychotropic medication use Pattern of anxious arousal in anticipation of and during exposure (BAT) Baseline severity (outcome measures); Comorbid mood disorder (DSM-III-R/DSM-IV); NSF Comorbid anxiety disorder (DSM-III-R/DSM-IV) Anger experience and expression (STAXI): State anger, Trait anger, Angry temperament, Angry reaction, Anger-In, Anger-Out, Anger control.; Demographic variables (gender, marital status, education, race and age); Baseline severity (SIAS, SPS, BFNE); Trust (RAAS-DEPEND); Quality of life (QOLI); Depressive symp. (BDI) Age; Neuroticism (EPQ-N); Depressive symp. (DASS-D); Severity (SPS, SIAS) Demographic characteristics (Marital status, Race, Education, Gender, Age) Nord J Psychiatry Downloaded from informahealthcare.com by Statsbiblioteket Tidsskriftafdeling Depressive symptoms (BDI, HRSD); Severity (BSPS) Demographic characteristics (Age,, Age of onset, Prior treatment experience, Gender, Marital status, Educational level); Severity (patient rated impairment, CSR, SCL-GSI, SCL-IS, BSQ, BAI, Number of feared situations, Total anxiety on the ADIS-R); Comorbid disorders (ADIS-R, BDI, HZI, ACQ, MI-A, WI); Chronic health problems; Alcohol use (MALT-S); Use of benzodiazepines Demographic characteristics (Age, Sex, Years of education, Family income, Ethnicity and Marital status); Severity (SPAI-SP and CSR); Subtype (GSP vs. NGSP); Depressive symptoms (BDI ); Depression (DSM-IV criteria); Number of additional Axis I diagnoses; Personality disorder traits (PDQ-4); Attitudes towards treatment (three Likert-type ratings) Severity (SPS, SIAS), depressive symptoms (BDI), medication use, education, employment status, marital status, age, sex, Note. See Appendix for the denotations of the abbreviations of measures. For the denotations of the other abbreviations, see note for Table 2. If treatment duration is not reported, it is estimated on the basis of number of sessions. Drop-outs are associated with a higher score on the predictor variable if the opposite is not indicated by a minus. *Used Bonferroni adjustment (P v0.002).
(25–27) and seven studies (20, 22, 25–29) examined pre- OTHER PERSONALITY DISORDERS AND MALADAPTIVE treatment level of depressive symptoms as a predictor of PERSONALITY TRAITSdrop-out. Only one of the seven studies (22), a large (n279) Drop-out. One study in Table 1 (27) examined multiple and well-conducted study, found a significant difference (more than one) personality disorder (PD) diagnoses, and with drop-outs reporting more depressive symptoms.
one (26) PD traits as predictors of drop-out. None of the two studies found any association between PD/PD traits and Outcome. Three studies in Table 2 examined MDD as a drop-out.
predictor of outcome (29–31) and nine examined level of depressive symptoms (15, 20, 22, 28, 29, 32–35). Only one Outcome. Four studies in Table 2 investigated PD or PD (30) of two studies (+31) specifically focusing on MDD traits, other than APD, as predictive variables (15, 27, 32, found a substantial, negative association with end-state 39), three studies investigated perfectionist personality traits functioning at post-treatment, and no such associations were (28, 34, 40), one (34) dependent personality traits, and one found at follow-up (45% and 0% significant findings, (28) neuroticism; in all cases with no substantial findings.
respectively). No associations were found with regard to improvement at post-treatment or at follow-up in the three COMORBID ANXIETY DISORDERS studies (12% significant findings for the two data points combined—NB, in the positive direction, i.e. more Drop-out. Four studies in Table 1 examined comorbid improvement among patients with MDD).
anxiety disorders as predictors of drop-out (26, 27, 30, 41) For degree of depressive symptoms four (15, 22, 33, with no statistical significant results.
35) of six (+28, 32) studies found an association with Outcome. Two studies in Table 2 (29, 30) examined the lower end-state functioning at post-treatment (60% sig- influence of comorbid anxiety disorders on outcome, one nificant findings), and none of two (32, 33) at follow-up study (41) the influence of generalized anxiety disorder, and (10% significant findings). With regard to degree of one study (27) the influence of both anxiety and mood improvement, depressive symptomatology predicted less disorder diagnoses. None of the studies found any substantial improvement in four (15, 20, 22, 33) of seven studies associations.
(+29, 32, 34) at post-treatment, and in two (20, 33) of four (+15, 32) at follow-up (34% and 48% significant findings, respectively). In summary, there is some evi- dence that an MDD diagnosis or the degree of depres- Drop-out. Two studies in Table 1 (20, 26) examined the sive symptoms is negatively influencing post-treatment relationship between the number of comorbid Axis I diag- end-state functioning after therapy. The association noses and drop-out with no statistical significant results.
between depression and improvement is less consistent Outcome. Two studies in Table 2 examined alcohol with some studies (29–31) even finding MDD or depres- problems (20, 29) and one substance abuse (42) with no sive symptomatology associated with more improvement.
significant findings. One study (20) examined the influence of miscellaneous comorbid diagnoses (measured on different scales) on degree of improvement with only four of 28 Drop-out. Three studies in Table 1 examined if avoidant significant findings (including depressive symptomatology personality disorder (APD) predicted drop-out from CBT [covered in the analysis above]; agoraphobic cognition; (27, 36, 37), and one (26) examined APD traits. None of the four studies found that APD or APD traits had any influence Nord J Psychiatry Downloaded from informahealthcare.com by Statsbiblioteket Tidsskriftafdeling Symptomatic severity
Outcome. Six studies in Table 2 examined APD (24, 27, GENERAL SYMPTOMATIC SEVERITY33, 36–38) and two APD traits (15, 32) as predictors of Symptomatic severity covers somewhat different concepts treatment outcome. APD/APD traits predicted lower end- and scales in the different studies, e.g. level of social state functioning at post-treatment in two (24, 33) of six anxiety and avoidance; clinician rated global severity; or studies (+(15, 27, 32, 36), and in two (24, 33) of five studies scores on general symptom scales like the General Symp- (+15, 27, 32) at follow-up (37% and 45% significant findings, respectively). APD/APD traits predicted less improvement in two (24, 33) of the eight studies at post-treatment, and in Drop-out. Nine studies in Table 1 (20, 22, 25–30, 44) none of six studies (15, 24, 27, 32, 33, 38) at follow-up (12% examined if baseline symptomatic severity predicted drop-out. and 0% significant findings, respectively). Thus, although Seven studies did not find a significant difference in severity some studies suggest that APD/APD traits predict lower between drop-outs and completers, and two found opposing end-state functioning at post-treatment and follow-up, there results, i.e. in one study (44) drop-outs were more impaired at is no evidence that it has any influence on improvement.
baseline, while in the other (28) they were less impaired. Thus, Results (Quotients of significant associations)/Comments Nord J Psychiatry Downloaded from informahealthcare.com by Statsbiblioteket Tidsskriftafdeling . Studies examining pre-treatment patient variables as predictors of treatment outcome.
Results (Quotients of significant associations)/Comments Nord J Psychiatry Downloaded from informahealthcare.com by Statsbiblioteket Tidsskriftafdeling ive behavioural group therapy; CCBT, comprehensive Results (Quotients of significant associations)/Comments g, and a “+” denotes an association in the opposite direction; der; OCD, obsessive–compulsive disorder; SP, social phobia.
groups; TCT, task concentrations training.
Nord J Psychiatry Downloaded from informahealthcare.com by Statsbiblioteket Tidsskriftafdeling : A “−” denotes that a higher score on the predictor variable is associated with less improvement or lower end-state functionin : APD, avoidant personality disorder; PD, personality disorders; GAD, generalized anxiety disorder; MDD, major depressive disor : AR, applied relaxation; BT, behaviour therapy; BGT, behaviour group therapy; CBT, cognitive behavioural therapy; CBGT, cognit : Abbreviations of measures are explained in Appendix A.
NR denotes that no data are reported; BAT, Behavioral Activation Test; CRC, Clinician rated change; NSF, No significant findings cognitive–behavioural therapy; CT, cognitive therapy (without behavioural methods); EXP, exposure; EXPG, exposure treatment in the existing studies seem to imply that baseline symptomatic Expectancy
severity does not influence drop-out rate.
Drop-out. Three studies in Table 1 (26, 27, 35) examined treatment expectancy or patients’ attitudes toward treatment Outcome. Seven studies in Table 2 examined the role of as predictors of drop-out, but only one (26) found statistical symptomatic severity in predicting treatment outcome (15, significant results for one of three variables.
20, 22, 28, 34, 35, 40). All four studies (15, 28, 35, 40) examining severity’s role in predicting end-state functioning Outcome. Two studies in Table 2 (32, 35) investigated the after treatment found that higher severity before treatment role of expectancy. One study (32) found that expectancy predicted higher severity after treatment on all outcome positively predicted improvement (pre/post-treatment and measures; and the same result was found for one study (15) pre-treatment/follow-up) in one of five outcome measures, with regard to follow-up (only one study (15) controlled for and the other (35) found that it positively predicted end-state baseline scores on the dependent variables). Four studies functioning in six of eight outcome measures (altogether, (15, 20, 22, 34) investigated severity’s role in predicting less i.e. covering end-state and improvement, at post-treatment improvement with two studies finding significant associations and follow-up, 38% significant findings).
in opposite directions, one positive (34) and one negative
(15), at post-treatment, and no association in two studies Medication use before and during therapy
(15, 20) at follow-up (7% and 21% significant findings, Drop-out. Three studies in Table 1 (20, 27, 46) examined respectively). In summary, there is consistent evidence that if medication use before and during therapy predicted drop- pre-treatment symptomatic severity predicts symptomatic out with no significant findings.
severity after therapy, but no evidence that it predicts degree of improvement.
Outcome. Three studies in Table 2 (20, 32, 46) examined the impact of medication use on treatment outcome with no substantial associations.
SP SUBTYPES
Drop-out. Only one study in Table 1 (36) compared
generalized SP with non-generalized SP and found no Social anxiety-related cognition
significant differences in drop-out rate between the groups.
Drop-out. One study in Table 1 (44) found that social One study (20) categorized clinician rated degree of anxiety anxiety-related cognition in relation to a behavioural activa- and number of feared situations under the heading of tion test (BAT) predicted drop-out on one of two measures.
“Subtypes”, but in this review, the two variables are Outcome. One study in Table 2 (15) found no associations considered measures of symptomatic severity.
between the frequency of positive and negative cognition in Outcome. The same study was also the only one to examine social situations at pre-treatment and end-state functioning SP subtypes as predictors of treatment outcome. Patients or improvement from pre- to post-treatment or from pre-with generalized SP were more severely disturbed on all treatment to follow-up.
outcome measures before and after therapy and more patients with non-generalized SP were classified as treatment Anger
responders (clinician rated) compared with the group of Drop-out. One study in Table 1 (25) found that drop-outs generalized SP. No differences were found with regard to endorsed a greater disposition to experience and express improvement on outcome measures, however.
anger on three of seven scales of the State–Trait Anger Expression Inventory (47).
Nord J Psychiatry Downloaded from informahealthcare.com by Statsbiblioteket Tidsskriftafdeling Demographic variables
Outcome. The same study found that three of seven Drop-out. Six studies in Table 1 (20, 26, 27, 29, 30, 45) measures of anger predicted three of four measures of end- examined the role of demographic variables (age, age of state functioning, while one additional measure of anger onset, prior treatment experience, gender, marital status, predicted one measure at post-treatment.
educational level, family income, ethnicity, and employment
status) in the prediction of drop-out. Two studies (29, 45)
found significant associations in that younger and female Patterns of anxious arousal
patients were more likely to drop out.
Drop-out. One study in Table 1 (48) examined patterns of Outcome. The one study in Table 2 (20) examining demo- anxious arousal during a BAT and classified patients into graphic variables as predictors of outcome found no signi- four groups: “high anxiety”, “increasing/high anxiety”, ficant associations between such variables and improvement “moderate anxiety” and “mild anxiety”, according to their from pre- to post-treatment. At 1-year follow-up, females, responses to eight SUDS probes during the BAT. No married patients and patients with higher education were association was found between patterns of anxious arousal more improved on one of two outcome measures.
Outcome. The same study also examined the predictive that even severely disturbed SP patients could profit from value of patterns of anxious arousal on treatment outcome. short time CBT, although severely disturbed patients A significant association was found between lower anxious might not achieve end-state functioning close to normality. arousal before treatment and higher responder rate after The finding of few prognostic factors should, however, treatment on two of four measures and between higher be interpreted on the background of the restricted range anxious arousal before treatment and more improvement of severity/comorbidity in most studies. Few clinicians during therapy on two of nine measures.
doubt that e.g. severe depression, severe substance abuse or severe personality disorders, would highly impede Miscellaneous variables
Studies have examined several other pre-treatment patient Research on predictor variables in psychotherapy is a variables as predictors of drop-out or treatment outcome, challenging field, conceptually and methodological. Con-but failed to find significant results: Social impairment ceptually, many of the relevant variables of predictors, as (Table 2; 32); trust and quality of life (Table 1; 25); well as of outcome, could be supposed to load highly on chronic health problems (Tables 1 and 2; 20); negative negative affect(ivity). Thus, the dependent variables might affect (Table 2; 40) and locus of control (Table 2; 49) be contaminated by the independent. It should come as (see Tables for information on the studies).
little surprise that symptomatic severity before treatment is associated with severity after treatment. The association Discussion
between pre-treatment severity and improvement or degree More than 15 years after Steketee & Chambless (16, of change might be influenced by regression towards the p. 390) declared prediction research “a stepchild to mean. This is taken into account in most studies by using behavioural outcome research”, the field has witnessed an residual change scores or firstly entering pre-treatment increased attention with more than 60% of the studies in scores in the regression analysis but in case of predictor this sample being from 2000 or later. There are, however, variables loading on negative affect, residual change still rather few studies, and their divergent variables and scores partly correct for the predictors of interest. Besides, methods limit any conclusions to be drawn from them.
if there is no association between pre-treatment variables Almost no pre-treatment patient variables predicted loading on negative affect and improvement, this might be drop-out from CBT for SP in the reviewed studies. This related to a combination of two forces drawing in differ-result is in line with conclusions from other reviews that ent directions, namely regression towards the mean and the psychotherapy termination literature in general is less responsiveness among severely disturbed patients. The considered inconclusive (50).
problem is a common one in psychological research with Only a few variables predicted outcome. MDD or variables characterized by loose boundaries and conceptual depressive symptoms prior to therapy were associated and/or measurement overlap (cf. 53).
with lower end-state functioning after therapy, but only Optimally, the study should try to disentangle the spe- inconsistently with less improvement during therapy. cific contribution of the predictor variable after taking nega-Correspondingly, APD or APD symptomatology was in tive affect into account, which was only attempted in a few some studies associated with lower end-state functioning, of the studies. Thus, in two studies on perfectionism with but not with less improvement. Comorbidity with other a similar design (28, 40), pre-treatment scores on the out-disorders, e.g. other anxiety disorders, substance abuse, come variable was firstly entered in the regression analysis, PD or PD traits was not found to be associated with out- secondly measures of negative affect, and then, thirdly, the come in the studies. Treatment expectancy and anger perfectionism dimensions of interest. Perfectionism was not Nord J Psychiatry Downloaded from informahealthcare.com by Statsbiblioteket Tidsskriftafdeling expression showed promising results in a few studies.
found to be a specific predictor in these studies.
The only consistent finding in this review was that prior As in all psychotherapy research, the power problem is symptomatic severity predicted lower end-state functioning a serious one. The correlations between predictors and after therapy and at follow-up, although it was not associ- outcome are mostly small to moderate demanding large ated with degree of improvement. This finding is in line samples to achieve acceptable power (nr200 if r0.2, and with Rodebaugh et al.’s (10) conclusion, that more dis- power0.8 in a correlation analysis). Only three studies turbed patients with SP (including patients with generalized in the sample had n200. Exclusion of severe cases in SP and APD) both begin and end treatment at a higher controlled studies will narrow the range in many relevant symptomatic level, but with a similar degree of improve- predictor variables (e.g. severity and comorbidity) thus ment as less disturbed patients; as well as with conclusions lowering the chance of significant findings. Comparing on predictor variables in CBT for other anxiety disorders groups of unequal size further reduces power, which is a (e.g. 51, 52). This conclusion might also cover the findings highly relevant concern with studies of drop-outs, and regarding depression and APD in this review.
with dichotomous predictors, e.g. comorbid diagnoses. If From a positive point of view, the few prognostic fac- e.g. the expected drop-out rate is 20% (a high percentage tors in CBT for SP might be interpreted as indicating according to prior studies), a sample of at least 200 is needed for an 80% chance of finding a difference between The most important conclusion from the review is that drop-outs and completers if the population difference is there is little clinically or theoretically relevant knowl-moderate (d0.50; A0.05, two-tailed) (11). Conse- edge to be gained from existing studies of pre-treatment quently, it is not surprising that a high proportion of the patient variables as predictors of drop-out and treatment studies had statistically insignificant findings.
outcome in CBT for patients with SP. It seems unlikely Further research would profit from large, naturalistic that more studies of the same kind would result in a studies with a high degree of variability in the relevant markedly different conclusion. The field is in need of
predictor variables. Study samples could be enlarged by conceptual and methodological improvements if more
cooperation between research centres. Some strategies, solid findings should be hoped for.
now widely recommended for outcome research (54, 55),
might also help to promote the field, namely to use one Declaration of interest: The authors report no con-
primary measure of outcome, and always report effect flicts of interest. The authors alone are responsible for sizes for an association (e.g. correlations with end-state the content and writing of the paper.
and residual gain score, or d-values). Reporting effect sizes (or relevant variables for calculating them) would highly facilitate research syntheses.
References
1. American Psychiatric Association (APA). Diagnostic and statistical Preferably, future studies should include more theoreti- manual of mental disorders (DSM-IV), 4th edition. Washington, cally derived variables, e.g. related to the proposed sub- types of SP by Hofmann et al. (56). Ultimately, the most 2. Furmark T. Social phobia: Overview of community surveys. Acta interesting question is, of course, about the relationship 3. Alonso J, Angermeyer MC, Bernert S, Bruffaerts R, Brugha TS, between differentially moderating variables in different Bryson H, et al. Prevalence of mental disorders in Europe: forms of treatments, so-called aptitude–treatment interac- Results from the European Study of the Epidemiology of Mental Disorders (ESEMeD) project. Acta Psychiatr Scand 2004;109 tions (57). The problems inherent in the much less com- plex study of predictive variables within one form of 4. Coles ME, Horng B. Social anxiety disorder. In: Andrasik F, editor. treatment, as exemplified in the present review, might Comprehensive handbook of personality and psychopathology: Vol. 2: Adult psychopathology. Hoboken, NJ: John Wiley & Sons help in understanding why little progress has been made 5. Fedoroff IC, Taylor S. Psychological and pharmacological treat- Most probably, prognostic patient variables interact in ments of social phobia: A meta-analysis. J Clin Psychopharmacol 2001;21:311–24.
complex patterns of causal relationships. With large sam- 6. Kessler RC. The impairments caused by social phobia in the gen- ples (nc. 300) structural equation modelling (58) could eral population: Implications for intervention. Acta Psychiatr be used to test specific models of the associations between 7. Rowa K, Antony MM. Psychological treatments for social phobia. different pre-treatment variables and outcome. From quite another corner of the research landscape, multiple inten- 8. Hofmann SG, Bögels SM. Recent advances in the treatment of sive case studies have been suggested as a method suited social phobia: Introduction to the special issue. J Cogn Psychother. Special Issue: Recent advances in the treatment of social phobia for investigation of the complex, idiographic patterns of patient prognostic variables (59, 60). At present there are, 9. Norton PJ, Price EC. A meta-analytic review of adult cognitive– however, no studies in the field using structural equation behavioral treatment outcome across the anxiety disorders. J Nerv Ment Dis 2007;195:521–31.
modelling, and few systematic case studies.
10. Rodebaugh TL, Holaway RM, Heimberg RG. The treatment of The review has some limitations. Thus, only published social anxiety disorder. Clin Psychol Rev 2004;24:883–908.
studies were included, but most well performed large stud- 11. Cohen J. Statistical power analysis for the behavioral sciences, 2nd Nord J Psychiatry Downloaded from informahealthcare.com by Statsbiblioteket Tidsskriftafdeling edition. Mahwah, NJ: Lawrence Erlbaum Associates; 1988.
ies will be published. Because of characteristics of the 12. Hofmann SG, Smits JA. Cognitive–behavioral therapy for adult studies, it was not possible to use meta-analytic data-syn- anxiety disorders: A meta-analysis of randomized placebo- thesis, which is better suited than the box-score method to controlled trials. J Clin Psychiatry 2008;69:621–32.
13. Gould RA, Buckminster S, Pollack MH, Otto MW, Yap L. Cognitive– take small associations in underpowered studies into behavioral and pharmacological treatment for social phobia: A meta- account. By excluding studies with n60, and by using analysis. Clin Psychol: Sci Pract 1997;4:291–306.
the criteria for counting results mentioned in the section on 14. Ruipérez MA, García-Palacios A, Botella C. Clinical features and treatment response in social phobia: Axis II comorbidity and social data analysis the review might compensate for one of the phobia subtypes. Psicothema 2002;14:426–33.
main problems of box score reviews, namely to give undue 15. Scholing A, Emmelkamp PMG. Prediction of treatment outcome in weight to findings from small studies. The nature of the social phobia: A cross-validation. Behav Res Ther 1999;37:659–670.
16. Steketee G, Chambless DL. Methodological issues in prediction of results with very few significant findings, furthermore, treatment outcome. Clin Psychol Rev 1992;12:387–400.
makes it implausible, that more sophisticated data synthesis 17. American Psychological Association. Evidence-based practice in methods (if they had been possible) would have resulted psychology: APA Task Force on Evidence-based Practice. Am Psychol 2006;61:271–85.
in substantially different conclusions. Anyhow, the review 18. Baron RM, Kenny DA. The moderator-mediator variable distinction should be considered an improvement over former, more in social psychological research: Conceptual, strategic, and statistical impressionistic box score or narrative reviews (10, 20).
considerations. J Pers Soc Psychol 1986;51:1173–82.
19. Zaider TI, Heimberg RG. Non-pharmacologic treatments for social 40. Ashbaugh A, Antony MM, Liss A, Summerfeldt LJ, McCabe RE, anxiety disorder. Acta Psychiatr Scand Suppl 2003;108:72–84.
Swinson RP. Changes in perfectionism following cognitive–behavioral 20. Lincoln TM, Rief W, Hahlweg K, Frank M, Von Witzleben I, treatment for social phobia. Depress Anxiety 2007;24:169–77.
Schroeder B, et al. Who comes, who stays, who profits? Predicting 41. Mennin DS, Heimberg RG, Jack MS. Comorbid generalized anxiety refusal, dropout, success, and relapse in a short intervention for disorder in primary social phobia: Symptom severity, functional social phobia. Psychother Res 2005;15:210–25.
impairment, and treatment response. J Anxiety Disord 21. Heimberg RG, Juster HR, Hope DA, Mattia JI. Cognitive–behavioral group treatment: Description, case presentation, and empirical support. 42. McEvoy PM, Shand F. The effect of comorbid substance use disor- Washington, DC: American Psychiatric Association; 1995.
ders on treatment outcome for anxiety disorders. J Anxiety Disord 22. Ledley DR, Huppert JD, Foa EB, Davidson JR, Keefe FJ, Potts NL. Impact of depressive symptoms on the treatment of generalized 43. Derogatis LR. SCL-90-R: Administration, scoring, and procedures social anxiety disorder. Depress Anxiety 2005;22:161–7.
manual I. Baltimore, MD: Clinical Psychometric Research; 1977.
23. Huppert JD, Strunk DR, Ledley DR, Davidson JR, Foa EB. Gener- 44. Mersch PP, Emmelkamp PM, Bögels SM, Van der Sleen J. Social alized social anxiety disorder and avoidant personality disorder: phobia: Individual response patterns and the effects of behavioral Structural analysis and treatment outcome. Depress Anxiety and cognitive interventions. Behav Res Ther 1989;27:421–34.
45. Herbert JD, Gaudiano BA, Rheingold AA, Myers VH, Dalrymple 24. Oosterbaan DB, van Balkom AJLM, Spinhoven P, de Meij TGJ, K, Nolan EM. Social skills training augments the effectiveness of van Dyck R. The influence on treatment gain of comorbid avoidant cognitive behavioral group therapy for social anxiety disorder. personality disorder in patients with social phobia. J Nerv Ment Dis 46. Rosser S, Erskine A, Crino R. Pre-existing antidepressants and the 25. Erwin BA, Heimberg RG, Schneier FR, Liebowitz MR. Anger outcome of group cognitive behaviour therapy for social phobia. experience and expression in social anxiety disorder: Pretreatment Aust NZ J Psychiatry 2004;38:233–9.
profile and predictors of attrition and response to cognitive– 47. Spielberger DD. Manual for the State–Trait Anger Expression behavioral treatment. Behav Ther 2003;34:331–50.
Inventory. Odessa, FL: Psychological Assessment Resources; 1988.
26. Hofmann SG, Suvak M. Treatment attrition during group therapy 48. Coles ME, Heimberg RG. Patterns of anxious arousal during expo- for social phobia. J Anxiety Disord 2006;20:961–72.
sure to feared situations in individuals with social phobia. Behav 27. Van Velzen CJM, Emmelkamp PMG, Scholing A. The impact of personality disorders on behavioral treatment outcome for social 49. Leung AW, Heimberg RG. Homework compliance, perceptions of phobia. Behav Res Ther 1997;35:889–900.
control, and outcome of cognitive–behavioral treatment of social 28. Rosser S, Issakidis C, Peters L. Perfectionism and social phobia: phobia. Behav Res Ther 1996;34:423–32.
Relationship between the constructs and impact on cognitive behav- 50. Corning AF, Malofeeva EV. The application of survival analysis to ior therapy. Cogn Ther Res 2003;27:143–51.
the study of psychotherapy termination. J Counsel Psychol 29. McEvoy PM. Effectiveness of cognitive behavioural group therapy for social phobia in a community clinic: A benchmarking study. 51. Taylor S. Understanding and treating panic disorder: Cognitive behavioural approaches. New York: Wiley & Sons; 2000.
30. Erwin BA, Heimberg RG, Juster H, Mindlin M. Comorbid anxiety 52. National Institute of Clinical Excellence. Post-traumatic stress disor- and mood disorders among persons with social anxiety disorder. der: The management of PTSD in adults and children in primary and secondary care. National Clinical Practice Guideline Number 31. Joormann J, Kosfelder J, Schulte D. The impact of comorbidity of 26. London: National Institute for Clinical Excellence; 2005.
depression on the course of anxiety treatments. Cogn Ther Res 53. Suls D, Bunde D. Anger, anxiety, and depression as risk factors for cardiovascular disease: The problems and implications of overlap- 32. Chambless DL, Tran GQ, Glass CR. Predictors of response to cog- ping affective disposition. Psychol Bull 2005;131:260–300.
nitive–behavioral group therapy for social phobia. J Anxiety Disord 54. CONSORT Group. The revised CONSORT statement for reporting randomized trials: Explanation and elaboration. Ann Intern Med 33. Feske U, Perry KJ, Chambless DL, Renneberg B, Goldstein AJ. Avoidant personality disorder as a predictor for treatment outcome 55. Boutron I, Moher F, Altman D, Schulz KF, Ravaud P. Extending among generalized social phobics. J Personal Disord the Consort statement to randomized trials of nonpharmacological treatment: Explanation and elaboration. Ann Intern Med 34. Cox BJ, Walker JR, Enns MW, Karpinski DC. Self-criticism in gen- eralized social phobia and response to cognitive–behavioral treat- 56. Hofmann SG, Heinrichs N, Moscovitch DA. The nature and expres- sion of social phobia: Toward a new classification. Clin Psychol 35. Safren SA, Heimberg RG, Juster HR. Clients’ expectancies and their relationship to pretreatment symptomatology and outcome of 57. Dance KA, Neufield RWJ. Aptitude–treatment interaction research cognitive–behavioral group treatment for social phobia. J Consult in the clinical setting: A review of attempts to dispel the “patient Nord J Psychiatry Downloaded from informahealthcare.com by Statsbiblioteket Tidsskriftafdeling uniformity” myth. Psychol Bull 1988;104:192–213.
36. Brown EJ, Heimberg RG, Juster HR. Social phobia subtype and 58. Maruyama GM. Basics of structural equation modeling. Thousand avoidant personality disorder: Effect on severity of social phobia, impairment, and outcome of cognitive behavioral treatment. Behav 59. Edwards DJA, Dattilio FM, Bromley DB. Developing evidence- based practice: The role of case-based research. Prof Psychol: Res 37. Huppert JD, Strunk DR, Ledley DR, Davidson JR, Foa EB. Generalized social anxiety disorder and avoidant personality 60. Fishman DB. Editor’s introduction to PCSP—From single case to disorder: Structural analysis and treatment outcome. Depress database: A new method for enhancing psychotherapy practice. Pragm Case Stud Psychother 2005;1:1–50.
38. Stangier U, Trenk-Hinterberger I, Erfert N, Heidenreich T. Soziale 61. Hofmann SG. Cognitive mediation of treatment change in social Phobie und Vermeidend-Selbstunsichere Persönlichkeitsstörung: phobia. J Consult Clin Psychol 2004;72:392–9.
Komorbidität und Behandlungsprognose in kognitiv-behavioraler Therapie [Social phobia and avoidant personality disorder: Comor- Anita Eskildsen, Cand.psych., Institute of Psychology, Aarhus University, bidity and treatment prognosis in cognitive–behavioral therapy]. Nobelparken, Aarhus C, and Clinic of OCD, Aarhus University Hospital, PTT: Persönlichkeitsstörungen Theorie und Therapie 2001;5:248–62.
39. Bogels SM. Task concentration training versus applied relaxation, in Esben Hougaard, Professor at Institute of Psychology, Aarhus University, combination with cognitive therapy, for social phobia patients with fear of blushing, trembling, and sweating. Behav Res Ther Nicole K. Rosenberg, Chief psychologist of Clinic of Anxiety, Aarhus University Hospital, Risskov, Denmark.
Appendix
Abbreviations of scales
ACQ, the Agoraphobia Cognition Questionnaire; ADIS-R, Anxiety Disorders Interview Schedule-Revised; BAI, Beck Anxiety Inventory; BAT, Behavioral Assessment Test; BDI, Beck Depression Inventory; BFNE, Brief Fear of Negative Evaluation; BSPS, the Brief Social Phobia Scale; BSQ, Body Sensation Questionnaire; CGI, the Clinical Global Impression Scale; CSR, Clinician Severity Rating (ADIS-R or ADIS-IV); CSAQ, the Cognitive–Somatic Anxiety Questionnaire; DASS-D, the Depression subscale of the Depression Anxiety and Stress Scale; DEQ, the Depressive Experiences Questionnaire; EPQ-N, Eysenck Personality Questionnaire—Neuroticism Scale; FNE, the Fear of Negative Evaluation Scale; FQ-SP, the Social Phobia subscale of the Fear Questionnaire; GISDS, the Global Impairment of Social Domains; GRAI, the Gambrill and Richey Assertion Inventory; HRSD, Hamilton Rating Scale for Depression; HZI, the Hamburg Obsessive-Compulsive Inventory; LOCS, The Levenson Locus of Control Scale; LSAS, the Liebowitz Social Anxiety Scale; LSAS-AX, Anxiety subscale of the LSAS; LSAS-AV, Avoidance sub-scale of the LSAS; LSAS-I, LSAS social interaction subscale; LSAS-P, LSAS performance subscale; LWASQ-B, the Behavior subscale of the Lehrer-Woolfolk Anxiety Symptom Questionnaire; MALT-S, the Self-evaluation scale of the Munich Alcoholism Test; MCMI, the Millon Clinical Multiaxial Inventory; MI, the Mobility Inventory; MI-A, the Alone subscale of the MI; MPS-F, the Multidimensional Perfectionism Scale; PAS, The phobic anxiety scale; PDE, The Personality Disorders Examination; PDQ-4, the Personality Diagnostic Questionnaire for DSM-IV; PERI-D, the Demoralization scale of the Psychiatric Epidemiology Research Interview; PSS, the Personal Self Scale of the Tennessee Self-Concept Scale; QOLI, the Quality of Life Inventory; RAAS-DEPEND, The DEPEND subscale of the Revised Adult Attachment Scale; RBI, the Rational Behavior Inventory; RGI, The rating of global improvement; RTQ, Reaction to Treatment Questionnaire; SADS, the Social Avoidance and Distress Scale; SAS-SR, the Social Adjustment Scale, Self-Report version; SASSI, the Social Anxiety Self-Statements Inventory; SCL-90, The Symptom Checklist; SCL-D, the Depression subscale of the SCL-90; SCL-GSI, The global severity index of the SCL-90; SCL-IS, the Interpersonal sensitivity subscale of the SCL-90; SIAS, the Social Interaction Anxiety Scale; SIB, Scale for Interpersonal Behavior; SPAI-SP, the Social Phobia scale of the Social Phobia and Anxiety Inventory; SPS, the Social Phobia Scale; SSIT, This test consists of 8 brief social interactions and was used to assess behavioral skills; STAI-S, the State section of the State–Trait Anxiety Inventory; STAI-T, the Trait section of the State–Trait Anxiety Inventory; STAXI, the State–Trait Anger Expression Inventory; SUDS, patients provided 0 to 100 anxiety ratings during a BAT; TES, the 4-item Treatment Expectancy Scale; WI, The Whiteley Index (Hypochondriasis).
Nord J Psychiatry Downloaded from informahealthcare.com by Statsbiblioteket Tidsskriftafdeling

Source: http://www.amkherning.dk/userfiles/file/Eskildsen%20Hougaard%20og%20Rosenberg%20(2010).pdf