A Multicenter, Randomized, Double-blind, Placebo-Controlled Trial ofParoxetine in Children and Adolescents With Social Anxiety Disorder

Background  Social anxiety disorder is a debilitating, highly prevalent disorder in children and adolescents. If left untreated, it can interfere with emotional, social, and school functioning.

Objective  To evaluate the efficacy and tolerability of paroxetine in children and adolescents with social anxiety disorder.

Design and Setting  Multicenter, 16-week, randomized, double-blind, placebo-controlled, flexible-dose, parallel-group, outpatient study.

Patients  A total of 322 children (8-11 years of age) and adolescents (12-17 years of age) with social anxiety disorder as their predominant psychiatric illness.

Intervention  Eligible patients were randomized (1:1) to receive paroxetine (10-50 mg/d) or placebo.

Results  Four hundred twenty-five patients were screened, and 322 were randomized to treatment. Of these, 319 were included in the intention-to-treat population (paroxetine, n = 163; placebo, n = 156). At the week 16 last observation carried forward end point, the odds of responding (Clinical Global Impression-Improvement score of 1 or 2) were statistically significantly greater for paroxetine (77.6% response [125/161]) than for placebo (38.3% response [59/154]) (adjusted odds ratio, 7.02; 95% confidence interval, 4.07 to 12.11; P<.001). The proportion of patients who were “very much” improved (Clinical Global Impression-Improvement score of 1) was 47.8% (77/161) for paroxetine compared with 14.9% (23/154) for placebo. Adverse events occurring at an incidence of 5% or greater for paroxetine and twice that for placebo were insomnia (14.1% vs 5.8%), decreased appetite (8.0% vs 3.2%), and vomiting (6.7% vs 1.9%). Withdrawals due to adverse events were infrequent (5.5% [9/163] for paroxetine and 1.3% [2/156] for placebo).

Conclusion  Paroxetine is an effective, generally well-tolerated treatment for pediatric social anxiety disorder.

Social anxiety disorder is a common condition in children and adolescents,with prevalence rates ranging from 0.9% to 7% in youths.^1,2 Childrenwith social anxiety disorder, as defined by the DSM-IV,³ have a fear of social and performance situations inwhich embarrassment or humiliation may occur. When exposed to these situations,they experience intense anxiety that substantially interferes with normalchildhood activities. Ordinary social interactions, such as starting or joininga conversation, and performances, such as playing sports or participatingin dance recitals, cause significant distress for these children. Their fearsand avoidance result in loneliness, dysphoria, and inadequate social skills.^4,5 The detrimental effects of social anxietydisorder are not limited to childhood. Adolescents with social anxiety disordermay be at increased risk for depression,⁶ substanceabuse, nicotine use, suicidal behavior, and educational underachievement inyoung adulthood.⁷ Moreover, social anxietydisorder in adolescents can persist to adulthood.⁸ Asadults, these individuals frequently have substantial impairment in work andsocial functioning⁹ and reduced quality oflife.¹⁰

Given the significant morbidity and chronicity of childhood social anxietydisorder, it is essential to identify efficacious treatments. Unfortunately,little attention has been directed at the pharmacological treatment of thisdisorder in the pediatric population. Most pharmacological studies have beenaimed at the treatment of adults. The selective serotonin reuptake inhibitors(SSRIs) paroxetine and sertraline and the serotonin norepinephrine reuptakeinhibitor venlafaxine have been shown in multicenter controlled trials tobe effective for the treatment of social anxiety disorder in adults.^11-14

Open-label studies with SSRIs for the treatment of childhood socialanxiety disorder have shown some promising results. Sertraline treatment forsocial anxiety disorder,¹⁵ citalopram combinedwith psychoeducation for social anxiety disorder,¹⁶ andfluoxetine for mixed anxiety disorders^17,18 demonstratedsome improvement in anxiety symptoms. A small, controlled, randomized trialshowed efficacy of fluoxetine treatment of child anxiety disorders.¹⁹ In a multicenter, double-blind, placebo-controlledtrial of fluvoxamine treatment for children and adolescents with anxiety disorders,fluvoxamine was superior to placebo in the reduction of anxiety.²⁰

This study is the first multicenter, double-blind, randomized pharmacologicaltrial directed specifically at the treatment of social anxiety disorder inchildren and adolescents. Given the established efficacy of paroxetine inadults with this disorder, this study was designed to examine the efficacyand safety of paroxetine for the treatment of social anxiety disorder in childrenand adolescents.

Male and female children (8-11 years of age) and adolescents (12-17years of age) who met DSM-IV diagnostic criteriafor social anxiety disorder, according to the Anxiety Disorders InterviewSchedule for DSM-IV (child and parent versions, presentdisorders),²¹ were eligible. Patients wereenrolled from 38 centers (22 in the United States, 10 in South Africa, 4 inCanada, and 2 in Belgium) from November 30, 1999, to October 19, 2001. Centersin the United States and South Africa enrolled the majority of patients, 182(57.1%) and 100 (31.3%), respectively.

The study was conducted in accordance with good clinical practice guidelinesand the Declaration of Helsinki (amended in Somerset West, Republic of SouthAfrica, October 1996), with the protocol and statement of informed consentapproved by the institutional review boards/ethics committees prior to eachcenter’s initiation. Written informed consent/assent was obtained fromall patients and their parents/guardians prior to study entry.

Patients were evaluated at the screening and baseline visits and wereexcluded if they had a clinically predominant Axis I disorder (based on theinvestigator’s judgment) other than social anxiety disorder (eg, dysthymia,simple phobia, obsessive-compulsive disorder, panic disorder, body dysmorphicdisorder, attention-deficit/hyperactivity disorder, generalized anxiety disorder,or separation anxiety disorder) or if they had a concurrent major depressiveepisode. Patients were also excluded if they had any history of a psychoticepisode (including schizophrenia), bipolar disorder, or a pervasive developmentaldisorder. Other grounds for exclusion were concurrent psychotherapy, concurrentpsychoactive medication use, substance abuse/dependence, hypersensitivity(defined as previous intolerance) to SSRIs, pregnancy/lactation, recent electroconvulsivetherapy, serious suicidal/homicidal risk, clinically significant abnormallaboratory or electrocardiogram findings, and a positive test result for illicitdrug use. In addition, patients were excluded if they had a serious medicalcondition that would preclude the administration of paroxetine.

This was a prospective, multicenter, randomized, double-blind, placebo-controlled,flexible-dose, parallel-group, outpatient study. A 1-week screening phasewas used to determine eligibility and to conduct baseline efficacy and safetyassessments. Patients meeting eligibility criteria were randomized (1:1) toreceive paroxetine (10-50 mg/d) or placebo for 16 weeks. Study assessmentswere scheduled at the end of weeks 1, 2, 3, 4, 6, 8, 10, 12, and 16 or onearly withdrawal. Disorder-specific and global assessment scales were administeredat each visit, and adverse events (AEs) and vital signs were monitored.

In light of the recognized importance of patient and family educationin the management of any chronic disease, particularly one that affects childrenand adolescents in the context of their families, this study incorporatedage-appropriate psychoeducational pamphlets (1 each for children, adolescents,and parents/guardians). These pamphlets were provided at baseline and includedinformation on the nature and course of the disorder and suggestions for self-help.^22-24 The pamphlet forparents/guardians also included reasons for children/adolescents’ developingsocial anxiety disorder, symptoms, types of treatment, and what parents cando to help. Extended discussions or more specific cognitive or behavioralinterventions were not permitted.

A computer-generated randomization list was used to assign patientsto each treatment group. The randomization code was not stratified by ageor sex. Placebo and paroxetine capsules were identical in appearance, so thatall study personnel and patients were blinded to treatment.

During the first week of the double-blind treatment phase, patientsreceived 10 mg/d of paroxetine or matching placebo. The dose could then beup-titrated (10 mg/d) no more frequently than every 7 days to a maximum doseof 50 mg/d. A dose reduction to the next lower dose consequent to an AE waspermitted after week 2. At the conclusion of the treatment phase or upon earlywithdrawal, patients ending treatment at 20 mg/d or higher were required togradually reduce study medication by 10 mg/d each week up to a maximum of4 weeks prior to stopping therapy. Following completion of this taper phase,patients returned to the clinic for a taper end visit. A follow-up visit wasrequired (14 ± 3 days) after the last dose including tapermedication.

A psychiatrist, clinical psychologist, or psychometrician with at least2 to 3 years’ experience with pediatric patients conducted all diagnosticand efficacy assessments. For consistency, detailed instruction on the useof the Anxiety Disorders Interview Schedule instrument and on the efficacyrating scales (Liebowitz Social Anxiety Scale for Children and Adolescents[LSAS-CA],²⁵ Social Phobia and Anxiety Inventory/SocialPhobia and Anxiety Inventory for Children,²⁶ andKutcher Generalized Social Anxiety Disorder Scale for Adolescents²⁷) was provided at the prestudy investigator’smeeting. Additionally, the same rater performed assessments on individualsthroughout the study when possible.

Social anxiety disorder in children and adolescents has only recentlyreceived recognition. Consequently, at the time this study was conducted,there were no validated, clinician-rated pediatric social anxiety scales availablefor use. As a result, the Clinical Global Impression-Improvement (CGI-I) scale²⁸ was used as the primary efficacy end point (proportionof responders based on a score of 1 [“very much improved”] or2 [“much improved”] at the week 16 last observation carried forward[LOCF] end point). In addition, several disorder-specific and global scaleswere included as secondary efficacy end points and summarized as the changefrom baseline at the week 16 LOCF end point. Secondary efficacy measures wereLSAS-CA, Clinical Global Impression-Severity of Illness (CGI-S),²⁸ KutcherGeneralized Social Anxiety Disorder Scale for Adolescents (≥11 years ofage), Social Phobia and Anxiety Inventory for Children (<14 years of age),Social Phobia and Anxiety Inventory for Adolescents (≥14 years of age),²⁹ and Global Assessment of Functioning scale. Age subgroupanalyses were conducted post hoc for CGI-I and LSAS-CA.

Safety was assessed at every visit through AE monitoring and vital signdetermination (eg, blood pressure and pulse). A serious AE was defined asany event that was fatal, life threatening, disabling/incapacitating or resultedin hospitalization, prolonged a hospital stay, or was associated with congenitalabnormality, cancer, or overdose (either accidental or intentional). In addition,any experience that the investigator regarded as serious or that suggestedany significant hazard, contraindication, adverse effect, or precaution thatmay have been associated with the use of the drug was documented as a seriousevent. Clinical laboratory evaluations (eg, hematology and serum chemistry)and physical examinations (including weight) were performed at baseline andweek 16 or upon early withdrawal.

It was estimated that 130 patients per treatment group who could beevaluated would be sufficient to detect a 20% difference between paroxetineand placebo in the proportion of patients with a CGI-I score of 1 or 2 atthe week 16 LOCF end point. The number of patients who could be evaluatedwas based on a placebo response rate of 40%. This difference is detectablewith a power of 90%, given a significance level of 5% and using a 2-sidedsignificance test.

All patients who were randomized into the treatment phase, who receivedat least 1 dose of study medication, and who had at least 1 postbaseline safetyor efficacy assessment were included in the intention-to-treat (ITT) population.

Statistical conclusions concerning the efficacy of paroxetine were madeusing the week 16 LOCF and the week 16 observed cases data sets, based onthe ITT population. Primary inference was based on week 16 LOCF, with week16 observed cases used to assess the robustness of the conclusions. All hypothesistests were 2-sided, and the effect of interactions was assessed at the 10%level of significance (primary end point at week 16 LOCF only). All otherstatistical tests were performed at the 5% level of significance. Binary data(proportion of responders based on CGI-I) were analyzed using logistic regressionwith results presented as adjusted odds ratios, 95% confidence intervals (CI)around the odds ratios, and associated P values.Continuous efficacy variables were analyzed by analysis of variance techniqueswith results presented as the difference in adjusted means for change frombaseline, 95% CI for the differences, and associated P values.Estimates of treatment difference for binary and continuous efficacy variableswere adjusted for age group, sex, applicable baseline score, and country group(United States, Canada, or South Africa/Belgium combined). Baseline CGI-Swas used as baseline score for CGI-I analysis. The change from baseline inCGI-S was analyzed using the Wilcoxon rank sum test with results presentedas the median difference and P value for the Wilcoxonrank sum test. The incidence of AEs of interest was compared between treatmentgroups post hoc using the continuity adjusted χ² test or Fisherexact test where there were small expected frequencies (<5).

The rigorous criterion for remission of a 70% or greater reduction inthe LSAS has been recommended for social anxiety disorder in adults.³⁰ Post hoc remission analysis was performed followingthe current study to provide a thorough assessment of remission in this groupof patients. Here, a 70% or greater reduction from baseline in LSAS-CA (reductionin symptomatology) was utilized in retrospective analyses. The other remissioncriterion that was analyzed was the CGI-I, where a CGI-I score of 1 (“verymuch improved”) indicated improvement in well-being/overall CGI diseaseseverity. Estimates of treatment difference were adjusted for age group (agesubgroups, combined analysis only), sex, baseline score (LSAS-CA or CGI-S),and country group. Additionally, the proportion of patients meeting both remissioncriteria was summarized.

A total of 425 patients were screened, and 322 were randomized to double-blindtreatment. The ITT population consisted of 319 patients (163 paroxetine and156 placebo). Of the 319 patients in the ITT population, 91 (28.5%) were children,and 228 (71.5%) were adolescents; 160 (50.2%) were male, and 159 (49.8%) werefemale.

There were no marked imbalances between treatment groups in demographiccharacteristics except for sex (Table 1).The percentage of males in the paroxetine group overall (43.6% [71/163]) waslower than in the placebo group (57.1% [89/156]). This was also the case inthe adolescent subgroup.

At baseline, the vast majority of patients (95.6%) were given a CGI-Srating of “moderately ill” (45.1%), “markedly ill”(38.5%), or “severely ill” (12.0%). The 2 treatment groups weresimilar with respect to their mean baseline scores on all symptom-severityrating scales, indicating comparable levels of social anxiety disorder severity(Table 2, Table 3).

The percentage of patients with a current comorbid psychiatric illnesswas slightly greater in the paroxetine group (56.4% [92/163]) than in theplacebo group (48.7% [76/156]). The most common (>10%) comorbid psychiatricconditions in the overall population were specific phobia (24.8% [79/319]),generalized anxiety disorder (23.5% [75/319]), and separation anxiety disorder(16.3% [52/319]).

The progress of patients through the study and details of reasons forwithdrawal are shown in Figure 1. Atotal of 228 (71.5%) of 319 patients completed the 16-week double-blind phase(ITT population). Overall, the percentage of patients who withdrew was higherin the placebo group (33.3% [52/156]) than in the paroxetine group (23.9%[39/163]). Rates of withdrawal due to AEs were low in both groups (5.5% paroxetinevs 1.3% placebo). Withdrawals due to a lack of efficacy were higher in theplacebo group (14.1% placebo vs 3.7% paroxetine).

A total of 72.2% (117/162) of patients in the paroxetine group receiveda dose higher than 20 mg/d. Slightly more than half the children (56.5% [26/46])received a dose higher than 20 mg/d, compared with 78.4% (91/116) of the adolescents.The number of patients in the paroxetine group who received a maximum doseof 50 mg/d (26.5% [43/162]) was approximately half that for placebo (49.4%[77/156]). At end point (week 16 LOCF), the mean dose for paroxetine was 32.6mg/d for all patients (26.5 mg/d for children and 35.0 mg/d for adolescents),whereas the overall mean dose for paroxetine-treated patients was slightlylower (24.8 mg/d for all patients, 21.7 mg/d for children, and 26.1 mg/d foradolescents).

Overall, 77.6% (125/161) of patients randomized to receive paroxetinewere defined as CGI-I responders at week 16 LOCF end point compared with 38.3%(59/154) for placebo (adjusted odds ratio, 7.02; 95% CI, 4.07 to 12.11; P<.001) (Figure 2).This result was supported by the week 16 observed cases analysis. The benefitof paroxetine was apparent within the first 4 weeks of treatment (Figure 3). The CGI-I responder data demonstrateda statistically significant benefit of paroxetine over placebo in both agesubgroups, again, for both the week 16 LOCF results and the week 16 observedcases data (post hoc analysis). There was no evidence of any statisticallysignificant treatment by covariate interactions, indicating that the treatmenteffect was consistent across age group, sex, baseline score, and country grouping.

Results for the treatment effect of paroxetine compared with placeboat the week 16 LOCF end point for the secondary efficacy parameters are shownin Table 4 and demonstrate a statisticallysignificant benefit of paroxetine over placebo for all 5 parameters.

Patients in the paroxetine group had greater reductions in LSAS-CA totalscore at all time points (Figure 4).The adjusted mean difference in LSAS-CA total score between paroxetine andplacebo at week 16 LOCF for the ITT population was –23.75 points infavor of paroxetine (95% CI, –29.77 to –17.74; P<.001), indicating a statistically significant benefit of paroxetineover placebo (Table 4, Figure 5). The treatment effect shown for LSAS-CA was consistentacross age subgroups (post hoc analysis). The adjusted mean difference inchange from baseline in LSAS-CA total score between paroxetine and placeboat week 16 LOCF for the ITT population in the child subgroup was –28.82points in favor of paroxetine (95% CI, –38.71 to –18.92; P<.001), and in the adolescent subgroup it was –20.62points in favor of paroxetine (95% CI, –28.10 to –13.14; P<.001).

The analysis for CGI-S was performed separately for each age group.For both children and adolescents, the median difference between paroxetineand placebo at week 16 LOCF end point in change from baseline in CGI-S scorewas –1.0 (P<.001), indicating a statisticallysignificant benefit in terms of the severity of illness of paroxetine overplacebo (Table 4). Similar results wereobserved for the week 16 observed cases analysis. At week 16 LOCF based onthe CGI-S, 47.5% (77/162) of paroxetine-treated patients vs 20.7% (32/154)of placebo patients were rated normal (“not at all ill”) or borderlinementally ill (Figure 6).

At the week 16 LOCF end point, the odds of being in remission accordingto the criterion of a 70% or greater reduction (from baseline) in LSAS-CAtotal score were statistically significantly greater for patients receivingparoxetine (remission proportion, 47.2% [75/159]) than for those receivingplacebo (remission proportion, 13.3% [20/150]) (adjusted odds ratio, 6.05;95% CI, 3.38 to 10.82; P<.001) (Table 5).

At the week 16 LOCF end point, the odds of being in remission accordingto the criterion of a CGI-I score of 1 were statistically significantly greaterfor patients receiving paroxetine (remission proportion, 47.8% [77/161]) thanfor those receiving placebo (remission proportion, 14.9% [23/154]) (adjustedodds ratio, 5.44; 95% CI, 3.09 to 9.57; P<.001)(Table 5). The remission analyses inboth age subgroups were consistent with the results in the total population(Table 5).

More than 4 times as many patients randomized to paroxetine met bothremission criteria (≥70% reduction in LSAS-CA and CGI-I of 1) at week 16LOCF end point compared with those receiving placebo (34.6% [55/159] vs 8.0%[12/150], respectively). This was also the case in both age subgroups at week16 LOCF end point: children, 38.6% (17/44) vs 8.9% (4/45), respectively; adolescents,33.0% (38/115) vs 7.6% (8/105), respectively.

Adverse events considered possibly treatment emergent (incidence ≥5%for paroxetine and at least twice that for placebo in both age groups combined)were insomnia (14.1% [23/163] vs 5.8% [9/156]; χ²₁= 5.26; P = .02), decreased appetite (8.0% [13/163] vs3.2% [5/156]; χ²₁= 2.57; P = .11), and vomiting (6.7% [11/163] vs 1.9% [3/156]; χ²₁ = 3.35; P = .07)(Table 6). Most AEs were mild to moderatein intensity.

The behavior-related AEs with an incidence of 5% or greater for paroxetineand at least twice that for placebo were somewhat different between childrenand adolescents. Nervousness, hyperkinesia, asthenia, and hostility met thesecriteria in children who had received paroxetine, and somnolence and insomniamet these criteria in adolescents who had received paroxetine. Other behavioralAEs of interest not included in Table 6 wereagitation (1.8% [ 3/163] vs 1.3% [2/156] for paroxetine and placebo, respectively),manic reaction (1.8% [3/163] vs 0% for paroxetine and placebo, respectively),and emotional lability (2.5% [4/163] vs 1.3% [2/156] for paroxetine and placebo,respectively).

The 4 paroxetine-treated patients who had an AE of emotional labilityhad expressed suicidal ideation or had threatened suicide (vs no patientsin the placebo group with a similar AE; P = .12using Fisher exact test). One of these patients, in addition to experiencingsuicidal thoughts, also exhibited self-harm behavior. For 3 of these patientsthe event occurred while receiving treatment, whereas for the fourth patientthe event occurred during the follow-up phase. A second instance of self harm(superficial cuts to the arms to get parents’ attention) was reportedin a fifth patient in the paroxetine group. This event was categorized asan AE of neurosis because the investigator indicated that the event was dueto impulsive behavior. If this event is combined with the 4 cases mentionedabove, the P value becomes. 06 for paroxetine vsplacebo. None of these 5 cases were considered serious, none involved clearevidence of a suicide attempt, and none of the events were attributed to thestudy medication by the investigator.

Serious AEs were reported in 3 patients in the paroxetine group (anemia,fear and depression related to social anxiety disorder, and broken arm) and1 in the placebo group (unintentional overdose). All except the unintentionaloverdose were considered unrelated to study medication.

The proportion of patients withdrawing from the study because of anAE was low in both groups (5.5% and 1.3% for paroxetine and placebo, respectively).Two patients in the paroxetine group were withdrawn because of manic reaction;this was the only AE that led to the withdrawal of more than 1 paroxetine-treatedpatient from the study. Dose reduction consequent to an AE was permitted oncea patient had reached 20 mg/d of paroxetine or matching placebo. In the paroxetinegroup, the proportion of patients who had dose reductions because of an AEwas statistically significantly greater than for placebo (17.2% [28/163] comparedwith 3.8% [6/156], respectively; χ²₁ = 13.51; P<.001).

The proportion of patients with an AE emerging upon treatment discontinuation(ie, with an onset either during taper dosing or during the follow-up phase)was significantly greater in the paroxetine group (47.2% [68/144]) than inthe placebo group (32.6% [42/129]; χ²₁ = 5.49; P = .02). Discontinuation-emergent AEs, whichoccurred at an incidence of 5% or greater in the paroxetine group and alsoat twice that incidence for the placebo group, were nausea (11.1% [16/144]vs 2.3% [3/129]; χ²₁ = 6.81; P<.01), dizziness (11.1% [16/144] vs 1.6% [2/129]; χ²₁ = 8.61; P<.01), andabdominal pain (6.9% [10/144] vs 1.6% [2/129]; χ²₁ = 3.52; P = .06).

No clinically significant changes in blood pressure, pulse, weight,or laboratory values were found.

Paroxetine was significantly superior to placebo on the primary efficacymeasure and all 5 secondary efficacy measures. Seventy-seven percent of theparoxetine-treated group were “much improved” or “very muchimproved” at week 16 LOCF end point compared with 38% of the placebogroup. Response rates were comparable for children and adolescents.

Beyond response, a major goal of clinical treatment is remission.³⁰ Significantly more paroxetine-treated patients (almost50%) achieved remission, defined as either a 70% or greater reduction in symptomsof anxiety (LSAS-CA) or a CGI-I score of “very much” improved.The proportion of patients meeting both remission criteria was 34.6% for theparoxetine group and 8% for the placebo group. This finding is particularlystriking because more than 50% of the study population at baseline was ratedas “markedly ill” to “among the most extremely ill”based on CGI-S.

It is interesting to note that the response rate to paroxetine in thisgroup of children and adolescents (77.6%) was higher than that reported foradults treated with paroxetine for social anxiety disorder (55%).¹¹ The mean dose of paroxetine treatment in this studywas 24.8 mg/d, whereas for adults the mean dose was 36.6 mg/d. Because mostadults report having symptoms of social anxiety disorder in childhood,³¹ it may be the case that this disorder is less treatmentresponsive in adulthood.

The majority of AEs ranged from mild to moderate in intensity. Adverseeffects considered possibly treatment emergent (incidence ≥5% for the paroxetinegroup and at least twice that for the placebo group) were insomnia, decreasedappetite, and vomiting. When comparing AEs between age subgroups, there wassome indication that the AE profile in children may differ slightly from thatin adolescents. Withdrawal due to AEs was low in the paroxetine group. Therewere no clinically significant changes in vital signs, weight, or laboratoryvalues in the paroxetine-treated group.

The use of antidepressants, including paroxetine, in patients youngerthan 18 years of age has recently come under scrutiny by regulatory authorities(and the media) because of concerns of a possible increased risk of suicidalthinking, suicide attempts, or self-harm. The Food and Drug Administrationhas initiated analyses of the pertinent safety data from all placebo-controlledtrials of antidepressants in pediatric patients. In addition, the Food andDrug Administration has requested that all antidepressant manufacturers includea warning statement in product labeling that recommends close observationof patients with major depressive disorder or other indications for the emergenceof suicidality when treated with antidepressants.

There are a number of limitations to this study. At the time it wasconducted, no validated instruments were available specifically to measuresymptoms of childhood social anxiety disorder; therefore, this study reliedon CGI-I as the primary outcome measure. However, the LSAS-CA and the KutcherGeneralized Social Anxiety Disorder Scale for Adolescents, which producedresults consistent with the CGI-I in this study, have since been validated.^25,27 Some commonly occurring comorbidconditions in adolescents with social anxiety, such as major depression andsubstance abuse,^32,33 were exclusioncriteria for this study. Furthermore, the diagnostic instrument (Anxiety DisordersInterview Schedule) was designed primarily to assess current anxiety disordersand therefore by design did not assess the presence of some psychiatric disorders(eg, bipolar disorder) or any past disorders. The duration of the treatmentperiod was relatively short, given the chronicity of childhood social anxiety.Although further improvement may have been possible with a longer treatmenttrial, the long-term maintenance of effect is best addressed via a relapse-preventiondesign trial. No other concomitant psychotherapy was permissible, other thanthe psychoeducation of patients about their illness. Lastly, some cliniciansmay have conducted both the efficacy and the safety assessments, which couldhave led to unblinding and potential bias. These limitations may impact thegeneralizability of the findings (efficacy and safety) to a broader population.

Psychotropic prescription for children has almost reached adult-userates, although the evidence base for their use in children is not nearlyas strong as in adults.^34,35 Thisstudy is therefore a welcome addition to the growing evidence base for theutility of SSRIs in the treatment of anxiety disorders in children.^15,18,19 While specializedindividual and group behavioral therapies have also shown benefit in the treatmentof childhood social anxiety disorder,^36-40 apaucity of skilled therapists limits accessibility to such treatments. Itis important that clinicians determine the need for pharmacological interventionin each case of social anxiety disorder, particularly when parental educationand supportive counseling prove ineffective. Paroxetine is not currently approvedfor use in pediatric patients.

For the purpose of demonstrating a pharmacological effect, our studydesign was constrained in the use of other nonpharmacological techniques thatmight enhance effectiveness outside of research settings. Yet there is reasonto expect that the provision of simple instructions about exposure in thecontext of the provision of pharmacotherapy (eg, “As the medicationbegins to work, encourage your child to do new things; reward your child forspeaking to peers”) might enhance response.¹⁶ Ifnothing else, such instructions are likely to enhance compliance by providingparents (and children) with a model for how they can help themselves whilethe children take the medication. These instructions should be shared (eitherby the health care provider or by the parent) with teachers and school counselors,all of whom will play an integral role in promoting socialization, increasingconfidence, and helping the child to build and maintain enriching peer networks.Future research should be conducted to demonstrate the merits of combinedpharmacotherapy and basic psychoeducation and/or behavioral therapy for childhoodanxiety disorders.

In conclusion, this is the first multicenter, well-controlled trialto demonstrate the efficacy of paroxetine in the treatment of social anxietydisorder in children and adolescents. This finding is important and clinicallymeaningful given the chronicity and substantial functional and social impairmentof untreated childhood social anxiety disorder.

Submitted for Publication: November 19, 2003;final revision received May 11, 2004; accepted June 9, 2004.

Correspondence: Karen Dineen Wagner, MD,PhD, Division of Child and Adolescent Psychiatry, Department of Psychiatryand Behavioral Sciences, University of Texas Medical Branch, 301 UniversityBlvd, Galveston, TX 77555-0188 (kwagner@utmb.edu).

Funding/Support: This study was funded by GlaxoSmithKline,King of Prussia, Pa. Drs Wagner, Berard, and Stein were participating investigatorsand have been paid consultants for GlaxoSmithKline. Drs Wagner and Berardhave received research support and speaking honoraria from GlaxoSmithKline.Drs Perera and Gee, Ms Wetherhold, Mr Carpenter, Ms Davy, and Ms Machin areemployees of GlaxoSmithKline and may own stock in the company.

Acknowledgment:We thank the following investigatorsfor their contribution to this study: Deborah Beidal, MD; Jerome Kaufman,MD; Samuel Turner, PhD; Joshua Calhoun, MD; David Duesenberg, MD; James Connor,MD; Graham Emslie, MD; Miguel Mandoki, MD; Wayne Goodman, MD; Chris Hayward,MD, MPH; Rakesh Jain, MD, MPH; James Kyser, MD; Jeffrey Kelsey, MD, PhD; ThomasCummings, MD; Michael Liebowitz, MD; Dale Mortimer, MD; Michael Duran, MD;Elizabeth Reeve, MD; Ronald Landbloom, MD; Robert Reichler, MD; Randall Ricardi,DO; Floyd Sallee, MD, PhD; David Sheehan, MD, MBA; Elizabeth Weller, MD; OwenHagino, MD; Craig Wronski, DO; Michael Labellarte, MD; Golda Ginsburg, PhD;Joseph Biederman, MD; Matthew Brams, MD; Kevin Kjernisted, MD; Margaret Weiss,MD, PhD; Michael Van Amerigen, MD; Normand Carrey, MD; Christine Reynaert,MD; Emile Willems, ADEJ; Karien Botha, MMed; Karin Fuls, MMed; Wally Landsberg,MMed; Catherine Maud, MMed; Solomon Rataemane, MMed; Desiree Rossouw, MMed;Christopher Szabo, MMed; and Gerrit Verster, MMed.