A Double-Blind Randomized Controlled Trial of Augmentation and Switch Strategies for Refractory Social Anxiety Disorder

Social anxiety disorder (also referred to as social phobia) is among the most common psychiatric conditions, with a lifetime prevalence of 12% (1), and is associated with significant distress and dysfunction in affected individuals. The emotional and social costs of social anxiety disorder are considerable, with greater associated physical and emotional health difficulties, alcohol and drug abuse, suicide attempts, financial dependence, use of general medical resources, poorer marital functioning, and lower educational and vocational attainment and productivity (1, 2). The onset of generalized social anxiety disorder is typically in early adolescence; thus, overall functioning and productivity are affected early and persist over much of the lifecycle, increasing the risk for complications such as depression and suicidality (2).

Despite the availability of evidence-based treatments for generalized social anxiety disorder, many patients remain symptomatic after initial intervention. For instance, in a two-site study comparing group cognitive-behavioral therapy (CBT) to treatment with the monoamine oxidase inhibitor (MAOI) phenelzine for patients with generalized social anxiety disorder, 42% of patients receiving group CBT and 35% of those taking phenelzine were nonresponders at 12 weeks in the intent-to-treat analysis (3). Similarly, controlled studies of the selective serotonin reuptake inhibitors (SSRIs) and serotonin-norepinephrine reuptake inhibitors (SNRIs) in generalized social anxiety disorder reported rates of nonresponse in the range of 40%−60% over 3–5 months of treatment (4–8). In addition, in most studies, responder criteria include patients who, although improved, were still symptomatic. For instance, in a study of the SSRI paroxetine for generalized social anxiety disorder (7), less than one-half of the total responders were “very much improved” at week 12 (a potential surrogate of remission); 78%−81% were still symptomatic. Thus, outcome data from treatment studies in social anxiety disorder suggest that roughly three-quarters or more of patients fail to achieve remission after initial intervention.

Therefore, although currently available treatments for social anxiety disorder are efficacious, most patients remain symptomatic after initial intervention. Clinicians regularly face the question of what to do next for these patients, yet despite the prevalence and attendant morbidity of this condition, no empirically derived data to guide clinical practice regarding the relative benefits of “next step” strategies to improve outcomes are available. Herein, we present what is to our knowledge the first study to provide systematic, prospectively derived data on the relative benefits of commonly available and applied pharmacotherapies to improve outcomes for persistently symptomatic patients with this common, distressing, and disabling disorder.

Our study comprised a 12-week, randomized, double-blind placebo-controlled trial at three sites comparing the relative benefits of three potential strategies for patients with generalized social anxiety disorder who remained symptomatic after a 10-week trial of the SSRI sertraline alone. These strategies were 1) augmentation of the SSRI with the benzodiazepine clonazepam (sertraline plus clonazepam), which was selected as the augmentation agent because of previous evidence of its efficacy as monotherapy (9) and because of its relative ease of use; 2) switch to the SNRI venlafaxine (extended release), which was selected as an antidepressant agent with a different mechanism of action (i.e., dual rather than serotonin selective); or 3) continued sertraline with the addition of a placebo pill (sertraline plus placebo), which was used to control for the additional time on the SSRI and the potential change in intervention.

Method

Participants

Prospective participants were recruited between August 2006 and July 2011 through advertisements and clinical referrals to Massachusetts General Hospital, McMaster University, and the University of California at San Diego. Inclusion criteria were 1) outpatients who were at least 18 years old with a current principal diagnosis of generalized social anxiety disorder by DSM-IV criteria (including designation by the patient as the most important source of distress or impairment), 2) a total score on the Liebowitz Social Anxiety Scale (LSAS) (10) of 60 or greater, and 3) no clinically significant medical abnormalities based on physical and laboratory examination.

Exclusion criteria were 1) a history of more than two unsuccessful adequate pharmacological treatment trials, operationalized as lack of response to at least 10 weeks of any of the following: SSRIs at adequate dosage (e.g., paroxetine, 40 mg/day or its equivalent); benzodiazepines (e.g., clonazepam, >2.5 mg/day) plus an antidepressant (at adequate dosages as above); MAOIs (e.g., phenelzine, 60 mg/day or its equivalent); or a single failed trial of at least 10 weeks of venlafaxine (≥150 mg/day); 2) pregnant women, lactating women, and women of childbearing potential not using medically accepted forms of contraception; 3) psychotic disorders, mental retardation, organic medical disorders, bipolar disorder, obsessive-compulsive disorder (OCD) with a Yale-Brown Obsessive Compulsive Scale (YBOCS) score ≥25, or a recent history of eating disorders or alcohol or substance abuse or dependence; 4) concurrent use of other psychotropic medications with discontinuation of regular benzodiazepine or antidepressant therapy at least 2 weeks before the baseline evaluation (5 weeks for fluoxetine); patients with comorbid adult attention deficit hyperactivity disorder (ADHD) could remain on stimulant medication if the dosage had been stable for at least 1 month; 5) current or recent significant suicidality; and 6) any concurrent psychotherapy initiated within 3 months or ongoing psychotherapy of any duration directed specifically toward the treatment of the generalized social anxiety disorder (e.g., CBT). Inclusion of patients with major depression, panic disorder, generalized anxiety disorder, adult ADHD, or posttraumatic stress disorder (PTSD) was permitted if the generalized social anxiety disorder was judged to be the predominant disorder.

Participants underwent an initial telephone screening followed by on-site psychiatric and medical assessment. Table 1 summarizes the demographic, comorbidity, and symptom severity information for those patients who were randomly assigned to a treatment group. Demographic characteristics and symptom severity did not differ across the three treatment conditions at the baseline evaluation.

Study Design and Procedures

Eligible participants were enrolled into the two-phase protocol. Phase 1 comprised 10 weeks of open prospective treatment with sertraline initiated at 25 mg/day, with all symptomatic patients titrated up to their maximally tolerated dosage (≤200 mg/day) by week 8. Participants who were unable to reach at least 50 mg/day by the end of phase 1 were discontinued from the study and referred for clinical treatment. Participants were seen weekly for the first 2 weeks of phase 1 and then biweekly thereafter. In addition, phase 1 participants were randomly assigned to receive either brief exposure instructions or no specific instructions regarding activity. Responders to open treatment (phase 1) at week 10 were continued on the same dosage of sertraline and seen at week 14 and week 22 for efficacy and safety assessments.

In phase 2, participants who remained symptomatic at the end of phase 1 (operationalized as an LSAS score >50 to identify individuals with at least a moderate level of severity and for whom additional intervention was warranted) were randomly assigned to one of three treatment groups: sertraline plus placebo; sertraline plus clonazepam, up to 3.0 mg/day; or switch to venlafaxine with flexible titration up to 225 mg/day. Randomization was stratified by site (Massachusetts General Hospital, University of California at San Diego, and McMaster University) and by LSAS severity at week 10 (LSAS total score ≤70 compared with >70). The sertraline dosage in the sertraline plus placebo and sertraline plus clonazepam arms was held at the same level as at entry into phase 2. For patients randomly selected for the venlafaxine group, the first 2 weeks consisted of a double-blind cross-titration with venlafaxine dosages set at 37.5 mg/day for the first 5 days and then at 75 mg/day. The dosage was increased to 150 mg/day at week 12, with flexible titration to a maximum of 225 mg/day thereafter. Sertraline was tapered off in weeks 10 and 11 by 50 mg every 5 days. The goal of the dose titration schedule was for patients to receive augmentation with either a maximum of 3.0 mg/day of clonazepam or placebo or switch to a maximum of 225 mg/day venlafaxine by week 20. Dosing was flexible, permitting clinicians to slow or suspend the titration of the medication because of side effects or response, but patients had to receive no less than 0.5 mg of clonazepam, 75 mg venlafaxine, or one capsule of placebo per day in order to remain in the study. A double-dummy design was employed in which patients received an identical number of study capsules at each visit (a mixture of study medication and placebo depending on treatment assignment and dose) to enable maintenance of the double blind for all three treatment arms. Patients were asked at each visit about the presence of adverse effects associated with the study medication. The protocol was approved by the institutional review boards of Massachusetts General Hospital, McMaster University, and University of California at San Diego, and all patients provided written informed consent.

Measures

Clinician-rated instruments.

The initial diagnoses were made by trained study clinicians using the Structured Clinical Interview for DSM-IV Axis I Disorders, Patient Edition (SCID-P) (11), a widely used instrument for the reliable diagnosis of axis I disorders in clinical populations. All patients met the criteria for the generalized subtype of social anxiety disorder. The LSAS is a 24-item scale assessing fear and avoidance in social and performance situations; it is widely used in studies of pharmacological treatment of generalized social anxiety disorder (10). The LSAS was administered at every visit and used in determining the primary categorical outcome of remission (i.e., LSAS score ≤30) (12), in representing the primary continuous outcome of change in LSAS over time, and to classify participants as responders (LSAS score ≤50) or nonresponders. Training in LSAS administration was provided to clinicians at the start of the study and periodically thereafter (approximately twice a year). The Clinical Global Impressions improvement scale (CGI-I) (13) is a clinician-rated instrument used to assess change and overall evaluation of global severity of symptoms and was used as a secondary outcome measure. The Montgomery-Åsberg Depression Rating Scale (MADRS) was designed to measure the overall severity of depressive symptoms and has demonstrated good reliability and specificity for depressive compared with anxiety symptoms as well as sensitivity to change with treatment (14). The MADRS was used at baseline, week 4, week 10 (randomization), week 14, week 18, and week 22 (endpoint) as a secondary measure to evaluate the effect of treatment on mood. The Hamilton Anxiety Rating Scale (HAM-A) (15) was developed to assess general anxiety symptoms in a clinical population and has proven sensitive to change with treatment. It was administered at baseline, week 4, week 10 (randomization), week 14, week 18, and week 22 (endpoint) as a secondary outcome measure.

Self-report questionnaires.

The Sheehan Disability Scale (16) is a brief scale that assesses overall function and in the domains of work, social life, and family life and has been demonstrated responsive to changes in treatment. It was administered at every visit of phases 1 and 2.

The Quality of Life Enjoyment and Satisfaction Questionnaire (17) rates 16 aspects of quality of life, including physical health, mood, activities of daily living, and overall life satisfaction. It was administered at baseline, week 4, week 10 (randomization), week 14, and week 22 (endpoint) and used as a secondary measure to examine changes in overall quality of life with treatment.

Statistical Methods

The primary endpoint of remission was defined as a clinician-rated LSAS total score of ≤30 at the last phase 2 visit. The sample size was determined to provide sufficient power to examine the primary study question: the comparison of categorical remission (defined as LSAS score ≤30) for each of the enhanced treatment groups (venlafaxine switch and clonazepam adjunct) compared separately with placebo adjunct in phase 2 of the study. With 90 individuals per treatment arm (270 randomly assigned participants total), the study had 80% power to detect a difference of at least a doubling of remission rates (i.e., 20% compared with 40%) between the venlafaxine switch or clonazepam augmentation arms and the placebo augmentation arm, assuming a two-sided alpha of 0.05. However, enrollment was terminated after the Data and Safety Monitoring Board reviewed the data and determined that additional recruitment was unlikely to materially affect the results of the study, and thus additional recruitment was unwarranted. Post hoc comparisons of outcomes between the venlafaxine and sertraline plus clonazepam arms were also conducted assuming a two-sided alpha of 0.05.

Logistic regression modeling was used to examine the relationship between phase 2 random treatment assignment and remission. We also examined whether treatment predicted improvement in LSAS severity over time. Per protocol, we planned to analyze improvement by calculating the overall change in LSAS score (last phase 2 LSAS score minus week 10 LSAS score) and by examining change over time using mixed modeling with random effects for slope and intercept. Normality of the LSAS total score was verified, and likelihood ratio tests and graphical methods were used to examine the assumption of linear change in LSAS score by week.

To assess if randomization was successful in balancing demographic characteristics across the treatment groups, we compared age at study entry, sex, race, ethnicity, and phase 2 completion status by treatment assignment using Fisher’s exact tests (categorical) and Student’s t tests (continuous). We also compared these covariates by phase 2 remission status to determine the potential for confounding; any covariates with a two-sided p<0.05 were adjusted for multivariate regression modeling. Logistic regression models were stratified by site and week 10 LSAS score (>70 compared with ≤70), while mixed models accounted for site and week 10 LSAS score through covariate adjustment. Models of tolerability and dropout accounted for site but not for week 10 LSAS score. We did not adjust for the phase 1 exposure instructions received in any of our modeling, as the instructions did not significantly affect phase 1 response rates (reported separately). Site-by-treatment interactions were examined in all models (except those of secondary outcome measures) to determine if the directionality of treatment effects differed between sites. All analyses were conducting using Statistical Analysis System, version 9.3 (SAS Institute, Cary, NC).

Results

Overview

A Consolidated Standards of Reporting Trials (CONSORT) diagram of the participant flow is presented in Figure 1. A total of 397 individuals received at least one dose of sertraline; this phase 1 intention-to-treat population comprised 139 women and 258 men with a mean LSAS score of 91 (SD=18) at baseline. A total of 312 individuals (79%) completed the initial 10 weeks of treatment. Among the intention-to-treat population (N=397), 129 patients (32%) achieved response (LSAS score ≤50) and 53 (13%) achieved remission (LSAS score ≤30). Among the participants who completed phase 1 (N=312), 117 (38%) achieved response, including 46 (15%) who achieved remission.

Thirteen nonresponders who completed phase 1 withdrew from the study before phase 2, and one phase 1 nonresponder who was randomly assigned in phase 2 dropped out before taking the study medication and was excluded from the phase 2 analyses. A total of 181 (61%) of the remaining 295 patients were nonresponders at week 10 (LSAS score >50), and therefore comprised the phase 2 intention-to-treat sample that was randomly assigned to receive 12 weeks of treatment with either sertraline plus placebo (N=59), sertraline plus clonazepam (N=63), or venlafaxine (N=59).

Phase 2 visits for nonresponders occurred weekly from weeks 11 to 14 and every 2 weeks thereafter through week 22. A total of 27 patients (15%) dropped out or were withdrawn before week 22 (nine in the sertraline plus placebo group, six in the sertraline plus clonazepam group, and 12 in the venlafaxine group). Analyses were conducted in the phase 2 intention-to-treat population (N=181), with remission based on either LSAS score at week 22 for participants who completed the study (N=154) or the last-observed phase 2 LSAS score for the 27 patients who left the study early.

Baseline characteristics of the intention-to-treat population are presented by phase 2 treatment assignment in Table 1. No significant differences by treatment group were observed.

Remission

A total of 38 intention-to-treat patients (21%) achieved phase 2 remission (LSAS score ≤30). The remission rates differed significantly by LSAS severity at week 10 (a blocking factor for randomization) and also by study completion status, which we therefore included in all models of outcome (but not in models of tolerability or dropout). In all, 27% of patients in the sertraline plus clonazepam group achieved remission compared with 17% in the sertraline plus placebo group and 19% in the venlafaxine group, but the pairwise differences did not reach significance. Conclusions were unchanged when using the Cochran Mantel-Haenszel statistic (i.e., when not controlling for completion status).

Change in LSAS Score

Phase 2 overall change in LSAS score (last phase 2 LSAS score minus week 10 LSAS score) as well as change in LSAS score by week were analyzed. The pattern of change in social anxiety disorder severity over time was nonlinear (Figure 2), with improvement tapering over time. Rather than use a nonlinear model, splines, or higher polynomial terms to model this nonlinear change in LSAS score by week, which would be difficult to interpret, we compared the change in LSAS score from week 10 to each phase 2 visit, since our interest was in the greatest magnitude of change (rather than the rate of change), and we conservatively applied the Bonferroni correction to these comparisons (alpha=0.006).

Overall and within each site, patients in the sertraline plus clonazepam group had a significantly greater reduction in social anxiety severity from week 10 through the last visit (mean=27 point drop, SD=24.4 overall) compared with patients in the sertraline plus placebo group (mean=16 point drop, SD=23.0 overall). Patients who were randomly assigned to venlafaxine also had a greater reduction in LSAS total score from week 10 through the last visit (mean=18 point drop, SD=19.0) compared with patients in the sertraline plus placebo group, but the separation never reached significance. Table 2 summarizes the comparison of change in LSAS score from week 10 to last phase 2 visit (i.e., at week 22 or early departure) by treatment.

Response

To supplement protocol-specified analyses focusing on remission and change in LSAS score, we conducted analyses examining response rates by treatment assignment. A total of 83 intention-to-treat patients (46%) achieved phase 2 response (LSAS score ≤50). After controlling for week 10 social anxiety disorder severity (a blocking factor for randomization) and study completion status, response rates differed significantly by treatment, with a greater proportion of patients in the sertraline plus clonazepam group (56%) achieving response compared with patients in the sertraline plus placebo group (36%; p=0.029), as summarized in Table 2. In all, 46% of patients in the venlafaxine group achieved response, a nonsignificantly different rate from patients in the sertraline plus placebo group (p=0.10). Conclusions were unchanged when using the Cochran Mantel-Haenszel statistic (i.e., when not controlling for completion status).

Dosing and Treatment Tolerability

The mean dose of sertraline at week 10 (end of phase 1) was 181 mg (SD=37) among the 181 patients in the phase 2 intention-to-treat population overall, including 181 mg (SD=39) for patients who were subsequently randomly assigned to the sertraline plus placebo group and 180 mg (SD=38) for patients subsequently assigned to the sertraline plus clonazepam group. The sertraline dosage was held constant during phase 2 in both arms. Over the course of phase 2, the mean doses administered by treatment arm were: sertraline plus clonazepam, 1.5 mg (SD=0.9); sertraline plus placebo, 180 mg (SD=39); and venlafaxine, 168 mg (SD=65). At the last phase 2 visit, the mean doses administered were: sertraline plus clonazepam, 2.3 mg (SD=0.8); sertraline plus placebo, 179 mg (SD=39); and venlafaxine, 197 mg (SD=56).

The analyses of treatment tolerability included a total of 155 patients (86%) in the intention-to-treat population who had at least one nonserious adverse event during phase 2. Of these, 139 patients (90%) had at least one nonserious adverse event that was at least possibly a result of treatment, and the reports were generally consistent with the known side-effect profiles of the medications (see Table S1 in the data supplement that accompanies the online edition of this article for the incidence of nonserious adverse events by type that occurred in ≥5% of patients in any given treatment arm). Of note, no significant differences were observed in the occurrence of nausea and dyspepsia (analyzed separately or together) or anxiety and nervousness (analyzed separately or together). Only the occurrence of somnolence differed significantly between the treatment arms, with significantly more patients in the sertraline plus clonazepam group experiencing somnolence (32%) compared with patients in the venlafaxine group (15%); 23% of patients in the sertraline plus placebo group experienced somnolence, but pairwise comparisons did not reach significance.

Comparing serious adverse events by treatment arm was not statistically feasible given that only six serious adverse events were observed during the study (and only three events during phase 2 among three patients who received sertraline plus clonazepam). Only one of these serious adverse events, fatigue and shortness of breath, was felt to be potentially related to the study medication (sertraline plus clonazepam), and it resolved after discontinuation of the study drug. Early discontinuation rates were highest among patients randomly assigned to venlafaxine (20%), followed by patients assigned to sertraline plus placebo (15%) and sertraline plus clonazepam (10%), although the differences were not significant.

Secondary Outcomes

We compared change in secondary measures (from week 10 to last phase 2 assessment), including the MADRS, the HAM-A, the Sheehan Disability Scale, and the Quality of Life Enjoyment and Satisfaction Questionnaire, by treatment group. As summarized in Table 3, only the decrease in Sheehan Disability total score differed significantly between patients in the sertraline plus clonazepam group and patients in the sertraline plus placebo group. Patients randomly assigned to venlafaxine also had a greater decrease in Sheehan Disability Scale total score compared with those in the sertraline plus placebo group, but the difference did not reach significance.

Discussion

This is, to our knowledge, the first large-scale, randomized, double-blind placebo-controlled clinical trial to examine “next-step” strategies for patients with an anxiety disorder (in this case generalized social anxiety disorder) who remain symptomatic despite treatment with a standard first-line pharmacotherapy. Rates of remission and response in 397 patients receiving initial pharmacotherapy with sertraline were 13% and 32%, respectively. The response rate in this treatment phase was lower than has been reported in other large acute pharmacotherapy trials in social anxiety disorder (6–8), which may be attributable in part to our use of a specific cutoff score on the LSAS rather than a clinical global improvement assessment, as was used in these other trials, as well as our inclusion of individuals with a broad range of clinically relevant comorbidities. Of the 181 nonresponders to initial sertraline who were randomly assigned to phase 2 interventions, an additional 21% of patients overall achieved remission and 46% achieved response. While there was a numerical advantage for the addition of clonazepam relative to a switch to venlafaxine or continuation on sertraline (plus placebo), as assessed by the prespecified primary outcome measure of remission, the difference between clonazepam and placebo was not statistically significant (nor was the difference between venlafaxine and sertraline plus placebo).

The relative treatment effects on social phobia symptoms, as assessed by the LSAS continuous measure or on categorically defined response (a lower hurdle than remission), more strongly indicated benefit for the clonazepam augmentation strategy, with significant and sustained effects emerging early in treatment relative to placebo. The 10-point difference in magnitude of change between clonazepam and placebo augmentation on LSAS score is in the range of what has been reported for effective pharmacotherapies for social anxiety disorder, as evinced by 12-week trials of SSRIs and SNRIs (6–8), suggesting that this response represents a clinically meaningful effect. Although the use of a specific cutoff score to determine response (LSAS score ≤50) raises the possibility that some patients may have achieved responder status with relatively small score changes, this method is unlikely to be a significant driver of the differential findings for responders, given the lack of difference between the groups in mean baseline severity (phase 2) and the significantly greater drop in LSAS scores for participants in the sertraline plus clonazepam group. Post hoc analyses demonstrate a number needed to treat of five for an additional patient to achieve response with clonazepam augmentation compared with placebo, and a number needed to treat of 10 to achieve response with clonazepam augmentation compared with a switch to venlafaxine, suggesting benefit for this strategy on improving outcomes, despite some residual symptoms at the endpoint for many patients. It is unknown whether participants would have further gains over time, although available data do suggest continued gains for up to 1 year with generalized social anxiety disorder treatment (18). The clonazepam augmentation strategy also indicated superiority on anxiety-related disability, as assessed by the Sheehan Disability Scale.

Rates of dropout from randomized treatment were in the range of 10%−20%, suggesting that all treatments were generally well tolerated. As in most pharmacotherapy clinical trials, adverse events and outcomes were assessed by the same research clinician. It is possible that this concurrent information could have biased assessors in one way or another, but the reliance on the LSAS—essentially a clinician-elicited patient self-report of symptoms (19)—would mitigate this possibility.

The dosages of clonazepam used in this study were somewhat lower than the highest dosages achieved in an earlier monotherapy study of clonazepam for social anxiety disorder (18). Given the good tolerability observed in our study of dosages up to 3 mg/day, it is possible that even higher dosages could be beneficial for some patients and may yield even better responses. Despite some theoretical concerns about the use of benzodiazepines in avoidance-based disorders such as generalized social anxiety disorder (20), these data support the idea that clonazepam augmentation appears to be an efficacious option for those with generalized social anxiety disorder refractory to an initial antidepressant trial, and suggest it may have greater therapeutic impact than additional time on the initial agent alone or switch to an alternative antidepressant.

For patients in this study, the overall remission rate by week 22 after treatment with the SSRI sertraline for 10 weeks and subsequent randomization to one of three alternatives for nonresponders was on the order of 21%. This result may be contrasted with results from the STAR-D study (21) for patients with major depression, where remission rates in each of the first two phases of treatment were about 30%. Although there are some important differences in the STAR-D study, including a more flexible equipoise stratified randomization for nonresponders compared with the blinded randomization in this trial, the marked difference in remission rates suggests that generalized social anxiety disorder had an overall less robust response to initial pharmacotherapy than major depression. It is likely that absolute response and remission rates may have been even lower had we included patients with more than two failed adequate pharmacotherapy trials. Additional pharmacological and psychosocial therapeutic strategies, as well as a longer course of treatment, may be necessary to move patients with generalized social anxiety disorder into remission; it is of note that patients remaining on a constant dosage of sertraline during phase 2 continued to improve from weeks 10 to 22, to a degree comparable to patients who were switched to a different antidepressant. Our findings of relatively modest remission rates, however, underscore the importance of efforts to develop novel strategies to improve outcomes for individuals with generalized social anxiety disorder.