Consort diagram. CBGT indicates cognitive behavioral group therapy.
Response and remission rates by treatment group at week 12. CBGT indicates cognitive behavioral group therapy; CGI-I, Clinical Global Impression Improvement Scale; and LSAS, Liebowitz Social Anxiety Scale.
Blanco C, Heimberg RG, Schneier FR, Fresco DM, Chen H, Turk CL, Vermes D, Erwin BA, Schmidt AB, Juster HR, Campeas R, Liebowitz MR. A Placebo-Controlled Trial of Phenelzine, Cognitive Behavioral Group Therapy, and Their Combination for Social Anxiety Disorder. Arch Gen Psychiatry. 2010;67(3):286-295. doi:10.1001/archgenpsychiatry.2010.11
Medication and cognitive behavioral treatment are the best-established treatments for social anxiety disorder, yet many individuals remain symptomatic after treatment.
To determine whether combined medication and cognitive behavioral treatment is superior to either monotherapy or pill placebo.
Randomized, double-blind, placebo-controlled trial.
Research clinics at Columbia University and Temple University.
One hundred twenty-eight individuals with a primary DSM-IV diagnosis of social anxiety disorder.
Cognitive behavioral group therapy (CBGT), phenelzine sulfate, pill placebo, and combined CBGT plus phenelzine.
Main Outcome Measures
Liebowitz Social Anxiety Scale and Clinical Global Impression (CGI) scale scores at weeks 12 and 24.
Linear mixed-effects models showed a specific order of effects, with steepest reductions in Liebowitz Social Anxiety Scale scores for the combined group, followed by the monotherapies, and the least reduction in the placebo group (Williams test = 4.97, P < .01). The CGI response rates in the intention-to-treat sample at week 12 were 9 of 27 (33.3%) (placebo), 16 of 34 (47.1%) (CBGT), 19 of 35 (54.3%) (phenelzine), and 23 of 32 (71.9%) (combined treatment) (χ21 = 8.76, P < .01). Corresponding remission rates (CGI = 1) were 2 of 27 (7.4%), 3 of 34 (8.8%), 8 of 35 (22.9%), and 15 of 32 (46.9%) (χ21 = 15.92, P < .01). At week 24, response rates were 9 of 27 (33.3%), 18 of 34 (52.9%), 17 of 35 (48.6%), and 25 of 32 (78.1%) (χ21 = 12.02, P = .001). Remission rates were 4 of 27 (14.8%), 8 of 34 (23.5%), 9 of 35 (25.7%), and 17 of 32 (53.1%) (χ21 = 10.72, P = .001).
Combined phenelzine and CBGT treatment is superior to either treatment alone and to placebo on dimensional measures and on rates of response and remission.
Social anxiety disorder (SAD) is a highly prevalent1- 3 chronic and disabling anxiety disorder associated with substantial impairment, decreased quality of life,4- 7 and psychiatric comorbidity.8,9 Although cognitive behavioral therapy (CBT) and pharmacotherapy are the most efficacious treatments for SAD,10- 15 only two-thirds of patients who receive these treatments are considered responders, of which only half are considered remitters.16 Most patients remain symptomatic after initial treatment.
Six controlled trials have examined the efficacy of combining medication and psychosocial treatments for SAD. The first study17 compared social skills training plus propranolol hydrochloride with social skills training plus placebo. There were no significant differences in efficacy between the groups. The second study18 compared buspirone hydrochloride, placebo, CBT plus buspirone, and CBT plus placebo; CBT resulted in improvement in SAD symptoms, but buspirone alone was not superior to placebo and did not augment the efficacy of CBT.
In the third study,19 patients were randomized to receive sertraline hydrochloride or placebo and separately to receive exposure therapy or general medical care. Sertraline was associated with greater efficacy than was placebo, whereas exposure alone was not. The fourth study15 examined the efficacy of fluoxetine hydrochloride, pill placebo, group CBT, CBT plus fluoxetine, and CBT plus pill placebo. All active treatments had greater efficacy than did pill placebo, but there were no differences among the active treatments.
Two recent studies20,21 examined the effect of administration of DQ-cycloserine before exposures in the context of CBT based on clinical and preclinical data on the effects of D-cycloserine on learning. Both studies found that D-cycloserine was superior to placebo as an adjunct to CBT. In a previous study,10 our group compared the monoamine oxidase inhibitor phenelzine sulfate, cognitive behavioral group therapy (CBGT), pill placebo, and a psychosocial control treatment. We found that phenelzine and CBGT were superior to pill placebo and the psychosocial control on a variety of measures. At the end of the study, many patients in both active treatments were still symptomatic, suggesting the need for more efficacious treatments. The goal of the present study was to examine whether a combination of 2 partially efficacious treatments with different mechanisms of action—pharmacotherapy and CBGT—would be superior to each monotherapy in the treatment of SAD. We selected phenelzine as the medication because it was the best-established medication for the treatment of SAD at the time this study was initiated.
This study was conducted at 2 academic centers with outpatient anxiety disorder programs and complementary expertise: the New York State Psychiatric Institute/Columbia University (pharmacotherapy expertise) (hereinafter New York) and the Adult Anxiety Clinic of Temple University (expertise in CBT) in Philadelphia (hereinafter Philadelphia). Enrollment began June 1, 1995, and continued through October 31, 2001. Biweekly conference calls were held to ensure homogeneity of procedures. The institutional review board at each site approved the protocol, and all the patients provided written informed consent.
The sample consisted of 128 patients referred by local mental health or medical practitioners or responding to advertisements in local media. Eighty-four patients were treated in New York and 44 in Philadelphia. The inclusion criteria were (1) a primary DSM-IV diagnosis of SAD and (2) age 18 to 65 years. To increase comparability with other treatment studies of SAD and to eliminate the possibility that improvements in SAD could be attributed to the antidepressant effects of phenelzine, the exclusion criteria were (1) a comorbid anxiety disorder more clinically salient for the patient; (2) a lifetime history of schizophrenia, bipolar disorder, or mental disorder due to a general medical condition; (3) major depressive disorder or substance use disorder in the past 6 months; (4) previous failure of treatment with phenelzine or CBT, defined as nonresponse to 60 mg or more of phenelzine (or the equivalent dose of another monoamine oxidase inhibitor) for at least 4 weeks or to 6 sessions of CBT for SAD; (5) concurrent psychiatric or psychological treatment; and (6) pregnancy, lactation, or inability or unwillingness to use contraceptive measures for the duration of the study.
At each site, patients were randomly assigned, in groups of 4 to 6, to 1 of 4 conditions: (1) phenelzine, (2) CBGT, (3) combined treatment (CBGT plus phenelzine), or (4) pill placebo. Patients were randomized according to a table of pseudorandom numbers by the New York site data manager (A.B.S.), who had no patient contact. Patient allocation was concealed from all other research personnel at both sites before randomization and from independent evaluators providing the clinician-administered assessments throughout the study. Medication or pill placebo was administered and monitored by a psychiatrist. All CBGT sessions were conducted by masters- or doctoral-level therapists. All CBGT sessions were audiotaped and evaluated by one of us (R.G.H.), who supervised therapists at both sites weekly.
The study had 4 phases. The first phase (acute treatment) lasted 12 weeks. Medication visits occurred weekly for 4 weeks, then every 2 weeks during this phase. The CBGT sessions took place weekly. Patients with at least minimal improvement on a modified version of the Clinical Global Impression Improvement Scale (CGI-I) that included anchors for each level of improvement22 entered the second phase. In this 12-week intensive continuation phase, patients received the same treatment, with CBGT sessions taking place weekly and the frequency of medication visits reduced to once per month. At the end of the continuation phase, patients who were at least much improved on the CGI-I entered the third phase, a 28-week maintenance phase during which they received the same treatment modality but with monthly visits for both modalities. Patients who were at least much improved on the CGI-I at the end of the third phase entered a 12-month naturalistic follow-up. In this article, we present the results of the acute treatment phase and the main findings of the continuation phase.
Two therapists administered CBGT in twelve 2.5-hour sessions to groups of 4 to 6 participants. In the first 2 sessions, patients were taught to identify negative cognitions (automatic thoughts), to observe the covariation between anxiety and automatic thoughts, to challenge logical errors in automatic thoughts, and to formulate rational alternatives. Thereafter, they confronted increasingly difficult feared situations, first through role-playing in the session and then in real life, while applying cognitive skills. Patients worked on their personal target situations following a standard sequence: (1) identification of automatic thoughts, (2) identification of logical errors in automatic thoughts, (3) disputation of automatic thoughts and formulation of rational responses, and (4) establishment of observable behavioral goals. Patients practiced cognitive skills while completing behavioral tasks (eg, conversing with another group member). Goal attainment and use of cognitive skills were reviewed. Patients were given assignments for exposures between sessions and completed self-administered cognitive restructuring exercises before and after these assignments.
Pharmacotherapy patients began with phenelzine sulfate, 15 mg/d, or matching placebo for 3 days, then 30 mg/d for 4 days, 45 mg/d for week 2, and 60 mg/d for weeks 3 and 4. Depending on clinical progress and adverse effects, the dosage could be raised to 75 mg for week 5 and to 90 mg for weeks 6 to 12. Patients were instructed to expose themselves to anxiety-provoking situations and were told that the role of medication was to make such exposure easier. However, no systematic exposure instructions or programmed practice was offered. No other psychotropic medication was permitted except chloral hydrate, 500 to 1000 mg, and zolpidem, 5 to 10 mg, as needed for sleep. Patients were instructed about the dietary restrictions appropriate to phenelzine, symptoms that could occur if the restrictions were violated, and procedures to follow in that event.
Patients assigned to combined treatment received CBGT and phenelzine as described in the preceding paragraph beginning in the same week. To remove potential bias in the performance of treatments, neither pharmacotherapists nor CBGT therapists were informed as to whether a specific patient was also receiving the other treatment, and they could not consult each other or attempt to integrate their treatment efforts. Patients were also coached to withhold information that would indicate whether they were receiving combined treatment. Although all patients undergoing combined treatment actually received phenelzine, they were told, with the approval of the institutional review board at each site, that they might receive either active medication or placebo.
Assessments were conducted at baseline (week 0) and at weeks 6, 12, and 24. Information was collected using clinician- and self-administered instruments.
Measures administered by independent evaluators blinded to treatment condition included (1) the Liebowitz Social Anxiety Scale (LSAS), a 24-item scale that assesses fear and avoidance of a range of social interaction and performance situations12,23- 25; (2) the Anxiety Disorders Interview Schedule for DSM-IV Clinician's Severity Rating (ADIS), a rating from 0 to 8 of the severity of symptoms and impairment associated with SAD26; (3) the modified CGI with anchor points defined specifically in reference to SAD22; and (4) the 29-item version of the Hamilton Rating Scale for Depression.27
Patient-rated symptom measures included (1) the Fear Questionnaire Social Phobia Subscale,28 a measure of the assessment of avoidance due to SAD; (2) the Social Interaction Anxiety Scale,29- 32 a measure of anxiety in dyads and groups; (3) the Social Phobia Scale, a measure of anxiety when being observed by others28- 30; and (4) the Sheehan Disability Scale,33 a 4-item scale to assess impairment in work, social life and leisure activity, family life and home responsibilities, and overall functioning as a result of a psychiatric disorder.
The pharmacotherapist used a checklist (available on request) to inquire about the presence of 28 potential monoamine oxidase inhibitor adverse effects at each visit and rated the severity of each on a scale from 0 to 3 (none, mild, moderate, or severe). Emergent adverse effects were identified by an increase of at least 2 points from baseline to any of the assessment points.15 Adverse effects were assessed in patients randomized to the placebo, phenelzine, and combined treatment groups but not in those randomized to the CBGT alone group.
The following hypotheses were tested: (1) phenelzine, CBGT, and combined treatment are superior to pill placebo in ameliorating the symptoms and disability of SAD and (2) there is a gradient of efficacy across the treatments, with combined treatment being superior to each monotherapy, which, in turn, is each superior to pill placebo.
To examine hypothesis 1 using continuous measures, outcomes at baseline and at weeks 6, 12, and 24 were modeled as a function of time, treatment, and the treatment × time interaction using linear mixed-effects models (LMMs),34 which take into account baseline differences across groups and the intercorrelations of repeated observations, and a statistical software program (SAS Proc MIXED; SAS Institute Inc, Cary, North Carolina). Site effects were assessed by including site in the models and by examining the interactions of site with treatment, time, and treatment × time. Treatment group differences were assessed by the significance of the interaction term and the comparison of LMM estimates at the end point (week 12 for the acute treatment phase and week 24 for the maintenance phase). Response and remission rates were compared between groups using χ2 tests of independence, using the last observation carried forward for individuals who dropped out before the end point. Odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated to assess the magnitude of the differences in categorical outcomes for each treatment arm compared with placebo. Responder status was defined as a score of much improved or very much improved on the CGI-I at week 12 (ie, a score of 1 or 2). In accord with previous work, 2 definitions of remission were used: (1) an LSAS score of 30 or less, previously found to be the optimal value to discriminate between individuals in or outside of the clinical range,24 and (2) a score of 1 on the CGI-I.35
To examine hypothesis 2 using categorical measures, we used the linear-by-linear association χ2 test. In contrast to χ2 tests of independence, this test assumes a specific gradation in the magnitude of responses.36 To examine hypothesis 2 using continuous measures, we used the Williams test.37 The linear-by-linear association test and the Williams test are part of a larger family of tests of constrained statistical inference.38 The use of constrained statistical inference is indicated when the hypothesis to be tested implies an ordering of effects, as in the study of dose-response relationships or augmentation strategies, and has the advantage of providing more powerful tests in those situations.38,39
For both hypotheses, the primary outcome measures were change over time in the total LSAS score and responder classification using the CGI-I. Secondary measures were changes in scores on the ADIS, CGI Severity Scale, Fear Questionnaire Social Phobia Subscale, Social Interaction Anxiety Scale, Social Phobia Scale, and Sheehan Disability Scale. The Hamilton Rating Scale for Depression was included to determine whether effects of treatment were due to reductions in depression rather than to test differences across treatments in reducing depressive symptoms in individuals with SAD.
We calculated the slope of outcomes on each continuous measure for each participant using LMMs as implemented in SAS Proc MIXED.40,41 We entered these individual slopes in the program ORIOGEN42 to calculate the Williams test statistic and the P value. The critical value and the P value of the Williams test statistic are determined using nonparametric bootstrapping procedures. For each bootstrap sample, the Williams test statistic was computed and was compared with the Williams test statistic for the actual sample. This process was repeated 100 000 times. Then, the bootstrap P value was defined to be the proportion of times that the Williams test statistic for the bootstrap sample exceeded that for the actual sample. Note that unlike the original Williams test, which assumes that the data are normally distributed, ORIOGEN is distribution free.42
All tests were considered significant at α = .05, 2-tailed. All analyses were based on the intention-to-treat sample, defined as those who received at least 1 dose of medication (or placebo) or attended at least 1 CBGT session. To examine the sensitivity of the results, we repeated the analyses with comorbidity as a predictor. Because comorbidity did not predict outcome, it was excluded from the final models. Power calculations for linear trends43 and for continuous variables44 under order restrictions indicated that the sample size provided 90% power to detect a linear trend in response rates and a linear trend in the change of LSAS score and 80% power to detect a linear trend in remission rates.
Of 726 patients who were screened, 166 were randomized to 1 of the 4 treatment groups (Figure 1). The most common reasons for screening failure were not meeting the inclusion criteria (n = 141) and lack of interest in participating in the research study due to time commitments or unwillingness to be randomized to treatment (n = 242).
Of the 166 individuals randomized, 12 from the placebo group, 10 from the phenelzine group, 6 from the CBGT group, and 10 from the combined group withdrew from the study before receiving any treatment (χ23 = 3.0, P = .40) and were excluded from the analyses. The remaining 128 participants composed the intention-to-treat sample as follows: phenelzine (n = 35), CBGT (n = 34), CBGT and phenelzine (n = 32), and pill placebo (n = 27). Groups did not differ significantly in demographic characteristics (Table 1). Differences between groups existed, however, in the baseline severity of SAD. Individuals randomized to the combined treatment group had significantly lower baseline values on the Fear Questionnaire Social Phobia Subscale, Social Interaction Anxiety Scale, and Social Phobia Scale than did those randomized to the other treatment conditions. Differences in baseline scores on the LSAS approached significance (Table 2). Some between-site differences were also observed. There were fewer married and Hispanic patients in Philadelphia than in New York. Mean age was lower in Philadelphia than in New York. Patients in Philadelphia had lower ADIS ratings than did those in New York. However, no site × treatment condition interactions were observed (data available on request). Of the 128 participants, 44 (34.4%) had at least 1 comorbid disorder, generally a comorbid anxiety disorder (26 patients [20.3%]) or dysthymia (19 [14.8%]).
Rates of discontinuation were 37.1% (13 of 35) in the phenelzine group, 35.3% (12 of 34) in the CBGT group, 28.1% (9 of 32) in the combined treatment group, and 18.5% (5 of 27) in the placebo group. Those rates were not significantly different when examining all groups jointly (χ23 = 3.0, P = .40) or in pairwise treatment comparisons (P > .10 for all). There were no differences in demographic or baseline measures between patients who dropped out and those who completed the acute treatment phase (data available on request).
Mean scores and standard deviations for all primary and secondary continuous measures at the primary end point (week 12) are given in Table 2. Using LMM analyses, we found significant differences in the outcomes of the 4 treatment groups for most measures, as indicated by the F tests for the treatment × time interaction effects. Pairwise comparisons of each treatment group vs placebo showed that the slope of the LSAS score change was significantly greater in the combined treatment (t = 4.3) and phenelzine (t = 3.3) groups than in the placebo group (P = .001 for both), whereas there was no significant difference between the slopes of the CBGT and placebo groups (t = 0.68, P = .50). The slope of change for the ADIS was significantly larger in the combined treatment group than in the placebo group (t = 3.1, P = .002), whereas the slope of the Social Phobia Scale score change was larger in the phenelzine group than in the placebo group (t = 2.1, P = .04). No other significant differences in slopes were seen between any of the treatment groups and placebo (data available on request).
Table 3 provides pairwise comparisons and corresponding effect sizes (Cohen d) of all the primary and secondary outcome measures at the acute phase end point for each treatment group vs the placebo group adjusting for baseline scores. Results were similar to the LMM findings. Phenelzine and combined treatment were superior to placebo on the LSAS, Fear Questionnaire Social Phobia Subscale, and Social Interaction Anxiety Scale. Combined treatment was also superior to placebo on the ADIS and the Social Phobia Scale. No significant differences were seen between the CBGT and placebo groups on any of the outcome measures. Effect sizes were generally small for CBGT, medium for phenelzine, and large for combined treatment.
Categorical measures yielded similar results. Patients randomized to receive combined treatment were significantly more likely than those randomized to receive placebo to be classified as responders (OR, 5.11; 95% CI, 1.68-15.52). There were no significant differences in the probability of response in patients randomized to receive phenelzine (OR, 2.38; 95% CI, 0.84-6.72), CBGT (1.78, 0.63-5.06), or placebo. Rates of remission were also significantly higher for patients randomized to combined treatment than for those randomized to placebo (Figure 2). Using the definition of CGI-I = 1, 46.9% of patients who received combined treatment were classified as remitters compared with 7.4% taking placebo (OR, 11.03; 95% CI, 2.23-54.57). In contrast, the percentage of CGI-I remitters was 22.1% in the phenelzine group (OR, 3.70; 95% CI, 0.82-19.14) and 8.8% in the CBGT group (1.21, 0.19-7.81), neither significantly different from the rate in the pill placebo group. When remission was defined by an LSAS score of 30 or less, 59.4% of patients in the combined treatment group and 11.1% in the placebo group were classified as remitters (OR, 11.69; 95% CI, 2.91-47.05). The percentage of remitters was 20.0% in the phenelzine group (OR, 2.00; 95% CI, 0.47-8.60) and 20.6% in the CBGT group (2.07, 0.48-8.93). Here again, the monotherapies did not distinguish themselves from placebo.
Table 4 provides mean slopes of change for all continuous measures and rates of response and remission for all the treatment groups. Across all measures, the results of the Williams test were highly significant. Examination of categorical measures produced similar results.
Because in a previous study10 phenelzine was superior to CBGT on several continuous measures after acute treatment, in the exploratory analyses of the present study, we reexamined the models hypothesizing the following order: combined treatment, phenelzine, CBGT, and placebo. The results also support this ordering of treatment effects (Table 4). Additional analyses restricted the sample to responders to examine whether responders to each treatment differed in magnitude of improvement. The mean slope of the LSAS score change was significantly larger for combined treatment than for the monotherapies considered separately or pooled (Table 5).
The mean (SD) daily dose of phenelzine sulfate at the end of week 12 in the medication group (65.9 [22.5] mg/d) and in the combined group (62.0 [24.6] mg/d) did not differ significantly (t = 0.7, P = .11). However, significant differences in rates of treatment-emergent events were noted for 7 symptoms: insomnia, lightheadedness, dry mouth, weight gain, constipation, anorgasmia, and nervousness (Table 6). For 3 symptoms, incidence was highest in the combined group; the incidence of 3 other symptoms was highest in the phenelzine group; and the incidence of 1 symptom was highest in the placebo group.
At week 24, mean (SD) LSAS scores were 59.3 (23.5) for the placebo group, 51.0 (22.9) for the CBGT group, 52.6 (24.0) for the phenelzine group, and 32.0 (19.6) for the combined group, resulting in effect sizes (Cohen d) of 0.36, 0.35, and 0.91 respectively. The Williams test statistic was 4.7 (P < .001). The proportion of responders at week 24 was 9 of 27 (33.3%) in the placebo group, 35 of 69 (50.7%) in those receiving monotherapy (18 of 34 [52.9%] for CBGT and 17 of 35 [48.6%] for phenelzine), and 25 of 32 (78.1%) in those randomized to combined treatment, yielding a linear-by-linear χ21 = 12.02 (P = .001). Rates of remission were 4 of 27 (14.8%), 17 of 69 (24.6%; 8 of 34 [23.5%] and 9 of 35 [25.7%]), and 17 of 32 (53.1%), respectively, resulting in a linear-by-linear χ21 = 10.72 (P = .001), when remission was defined as an LSAS score of 30 or less. Rates of remission were 2 of 27 (7.4%), 14 of 69 (20.3%; 5 of 34 [14.7%] and 9 of 35 [25.7%]), and 15 of 32 (46.9%), with a linear-by-linear χ21 = 12.78 (P < .001), when remission was defined as CGI-I = 1.
To our knowledge, this is the first study to show the superiority of a combined treatment over medication, psychotherapy, and placebo in the acute treatment of SAD. In addition, we found that phenelzine, but not CBGT alone, was superior to placebo. These results were consistent across several outcome measures and analytic strategies and were maintained throughout the 12-week continuation phase.
Supporting the main hypothesis, combined treatment was superior to both monotherapies and to placebo. Two mechanisms could explain the higher efficacy of combined treatment: (1) distinct groups of patients with SAD could respond to only phenelzine or CBGT (by receiving both, patients in the combined treatment group would have increased their chances of receiving at least 1 treatment that was efficacious for them) and (2) combined treatment may exert a truly additive or synergistic effect in the treatment of SAD beyond the effects of either monotherapy alone.
If only the first mechanism was at work, responders in the combined group would not have had larger average improvements than responders in the monotherapy groups. However, individuals receiving combined treatment had larger average improvements than did those receiving phenelzine or CBGT alone. This finding suggests an additive or synergistic effect of these treatment modalities, possibly due to their different mechanisms of action or by mutually facilitating the other's effect. For example, phenelzine may reduce anxiety and increase the chances of successful exposures to feared situations, whereas the skills learned through CBGT may help those taking phenelzine profit more from their exposures.
The present findings of the superiority of combined treatment are at variance with those of previous studies15,17,18 of combination treatment for SAD but in accord with some other studies and meta-analyses45- 48 that have shown the superiority of combined treatment over monotherapies in other mood and anxiety disorders. Discrepancies in the SAD results may be due in some cases to the use of medications with a mixed15 or poor record of efficacy in the treatment of SAD.17,18
The findings of Blomhoff and colleagues19 are more difficult to interpret. Although the study did not find an additional benefit of combined treatment over sertraline monotherapy, this result may have been due to the use of pairwise comparisons rather than tests for ordered responses implied in the design. We reanalyzed the rates of response from that study assuming a gradation of response from placebo to monotherapies to combined treatment using a linear-by-linear test, which yielded χ21 = 8.0 (P = .005). An even more significant result was obtained when the gradation was assumed to be placebo, exposure therapy, sertraline, and combined treatment (χ21 = 9.9, P = .002). More recent work by the same group,49 although not formally tested for ordered responses, also suggests a gradient of efficacy in the acute treatment of SAD, with placebo having the lowest degree of response, followed by monotherapies (exposure therapy and sertraline), and combined treatment having the highest efficacy at week 24. Taken together, the available evidence seems to support the superiority of combined treatment over medication or exposure and CBT alone for the treatment of SAD.
Consistent with previous studies,10,50- 52 we found that phenelzine was superior to placebo on most measures, providing additional documentation of its efficacy. Phenelzine and CBGT, however, were less efficacious than in previous studies,10 and CBGT was generally not superior to placebo in pairwise comparisons, although it was superior to placebo in the analyses using tests of constrained statistical inference, was not different from phenelzine monotherapy in pairwise comparisons at week 12, and achieved the same efficacy as phenelzine at week 24. The lower efficacy of CBGT in this study is surprising to us and may be due to sample differences (R.G. H. moved from Albany to Philadelphia between the time of the previous trial10 and the present one). However, there were no site × treatment interactions in either study, making this explanation less likely. Furthermore, in another 2-site trial53 conducted since the time of the study reported herein, CBT proved highly efficacious. Recent meta-analyses54- 56 and qualitative reviews57 continue to support the efficacy of CBT for SAD.
The present study has the limitations common to most efficacy trials. First, treatments were provided by experts and may show lower efficacy in less specialized settings. Second, participants had to be willing to be randomized to any of the 4 treatment conditions. Individuals who dropped out of the study after randomization but before receiving any treatment were not included in the analyses. These results may not generalize to them. Similarly, because individuals were recruited from advertisements and word of mouth, the results may not be generalizable to all patients with SAD. Third, because the study lacked a CBGT plus pill placebo group, the nonspecific effects of phenelzine in the combined treatment cannot be ruled out. Fourth, because self-exposure was neither assessed nor explicitly discouraged in the pill-only groups, it is possible that more spontaneous exposure occurred in the phenelzine group, which may have contributed to their improvement. Fifth, the study examined only 1 medication, 1 psychotherapy, and their combination rather than a broader array of treatments. Thus, these findings may not extend to individuals treated with selective serotonin reuptake inhibitors or selective noradrenergic reuptake inhibitors. However, the reanalysis of the study by Blomhoff et al19 and the findings of Haug et al49 suggest that a gradation of response from placebo to monotherapy to combined treatment may extend to other medications and empirically supported psychotherapies. Further research is needed to confirm those findings. Sixth, there were some baseline differences across treatment groups and sites. However, the results remained significant after appropriate statistical adjustments, suggesting the robustness of the findings.
In summary, this study is the first, to our knowledge, to provide an empirical rationale for the use of combined treatment for SAD. Future studies should prospectively examine whether the combination of a selective serotonin reuptake inhibitor plus behavioral therapy or CBT is superior to either treatment alone and the acceptability, efficacy, and cost-effectiveness of combined vs sequentially administered or augmented treatments.
Correspondence: Carlos Blanco, MD, PhD, New York State Psychiatric Institute, 1051 Riverside Dr, Box 69, New York, NY 10032 (firstname.lastname@example.org).
Submitted for Publication: October 15, 2008; final revision received June 22, 2009; accepted June 22, 2009.
Financial Disclosure: Dr Blanco reports support from Pfizer and GlaxoSmithKline. Dr Heimberg reports support from research funding from GlaxoSmithKline. Dr Schneier reports support from research funding from Forest Laboratories and Pfizer and is on the scientific advisory board of Jazz Pharmaceuticals. Dr Liebowitz reports equity ownership in ChiMatrix LLC, electronic data capture, Liebowitz Social Anxiety Scale; consulting for Avera, AstraZeneca, Pherin, Tikvah, and Wyeth; licensing software of LSAS with GlaxoSmithKline, Pfizer, Avera, Tikvah, Lilly, Indevus, Servier Speaking, and Wyeth; and has clinical trial contracts with Pfizer, GlaxoSmithKline, Forest, Lilly, Pherin, Novartis, Sepracor, Takeda, Horizon, and Johnson & Johnson. Dr Liebowitz is also on the speaking bureau of Bristol Myers Squibb.
Funding/Support: This study was supported in part by grants DA023200 (Dr Blanco), MH44119 (Dr Heimberg), and MH57148 (Dr Liebowitz) from the National Institutes of Health; by the New York State Psychiatric Institute (Drs Blanco, Schneier, Campeas, and Liebowitz and Ms Vermes); and in part by General Clinical Research Center grant RR00349 from the National Center for Research Resources, National Institutes of Health, to Temple University.