A Randomized Placebo-Controlled Trial of Fluoxetine in Body Dysmorphic Disorder FREE

BODY DYSMORPHIC disorder (BDD), also known as dysmorphophobia, consists of a distressing or impairing preoccupation with an imagined or slight defect in appearance. Although BDD was first described more than a century ago,¹ research on its pharmacologic treatment remains limited and no placebo-controlled pharmacotherapy studies have been done to our knowledge. Such research is needed since BDD causes severe distress and marked impairment in functioning.^{1 - 6} A high percentage of patients require hospitalization, become housebound, and attempt suicide.^{3 ,7} Completed suicide has been reported in both psychiatric¹ and dermatologic^{8 - 9} settings, and quality of life is notably poor.¹⁰

Body dysmorphic disorder seems to be relatively common in community,^{11 - 13} psychiatric,^{14 - 16} cosmetic surgery,^{17 - 19} and dermatologic²⁰ settings. As many as 50% of patients with BDD receive surgery or dermatologic treatment, often with poor outcomes.^{2 - 3 ,7} In all of these settings, BDD has been reported to be extremely difficult to treat.^{21 - 23}

Early case reports noted mixed but largely negative outcomes with a variety of psychotropic agents and electroconvulsive therapy.¹ However, subsequent data from case series and 2 open-label fluvoxamine maleate trials suggest that BDD may respond to serotonin reuptake inhibitors (SRIs).^{3 ,7 ,24 - 30} The only published controlled pharmacotherapy trial on BDD to our knowledge was a double-blind cross-over study, which found that the SRI clomipramine hydrochloride was more effective than the non-SRI antidepressant desipramine hydrochloride,³¹ supporting earlier retrospective findings^{3 ,7 ,25} that SRIs may be selectively effective for BDD and that the treatment response of BDD differs from that of depression.

Most patients with BDD have poor insight or are delusional regarding their appearance flaws,⁷ which has the potential to complicate treatment. Available data suggest that patients with delusional BDD respond to SRIs as well as^{3 ,7 ,28 ,32} or even better than³¹ nondelusional patients, although most studies did not assess delusionality (insight) with a reliable and valid scale. In addition, several studies found that delusionality improves with SRI treatment.^{31 - 32} Although data are very limited, antipsychotics alone seem ineffective for delusional BDD.^{3 ,7}

We report the first placebo-controlled treatment study of BDD and its delusional variant. We hypothesized that (1) fluoxetine hydrochloride would be more effective than placebo (the primary hypothesis); (2) delusional BDD would respond as well as nondelusional BDD to fluoxetine; and (3) illness severity and the presence of major depression, obsessive-compulsive disorder(OCD), or a personality disorder would not predict outcome.

The study was done in outpatients at a single academic site. Patients were entered into the study from August 1995 through February 2000. All patients met DSM-IV criteria for BDD³³ :(1) preoccupation with an imagined defect in appearance; if a slight physical anomaly is present, the person's concern is markedly excessive; (2) the preoccupation causes clinically significant distress or impairment in social, occupational, or other important areas of functioning; and (3) the preoccupation is not better accounted for by another mental disorder (eg, dissatisfaction with body shape and size in anorexia nervosa). Because the Structured Clinical Interview for DSM-III-R (SCID-P)^{34 - 35} did not include BDD, BDD was diagnosed with the Body Dysmorphic Disorder Diagnostic Module, a reliable semistructured SCID-like diagnostic instrument based on DSM-IV criteria.³⁶ Patients with delusional beliefs about their appearance (delusional disorder, somatic type) were included because the delusional and nondelusional forms of BDD seem to constitute the same disorder,⁷ and patients with delusional BDD may be diagnosed with both BDD and delusional disorder according to the DSM-IV³³ (patients with other types of somatic delusions but no appearance-related delusions were excluded). Body dysmorphic disorder was diagnosed by the consensus of the first 2 authors. A family member or other informant was interviewed(for the 36 patients willing and able to do this); in all cases, the BDD diagnosis was confirmed. Comorbid disorders were diagnosed by the first author with the SCID-P and the Structured Clinical Interview for DSM-III-R Personality Disorders.^{37 - 38} Data on the clinical characteristics of BDD were obtained with a semistructured instrument (K.A.P., unpublished data, 1992).

Inclusion criteria were (1) presence of DSM-IV BDD or its delusional variant currently and for at least 6 months; (2) age18 to 65 years; (3) score of 24 or higher on the Yale-Brown Obsessive Compulsive Scale Modified for Body Dysmorphic Disorder (BDD-YBOCS)³⁹ ;(4) score of at least moderate on the Clinical Global Impression Scale for BDD (BDD-CGI)⁴⁰ ; (5) ability to communicate and give written informed consent.

Exclusion criteria were (1) schizophrenia, schizoaffective disorder, or another current or lifetime psychotic disorder not attributable to delusional BDD; (2) current or lifetime bipolar disorder; (3) alcohol or substance dependence or abuse in the past 6 months; (4) body image concerns better accounted for by an eating disorder, including eating disorder not otherwise specified;(5) primary body image concern with weight and BDD criteria not met if weight concerns were excluded from consideration; (6) recent suicide attempt or clinically significant suicidal ideation; (7) use of psychoactive medication within 2 weeks of placebo lead-in (6 weeks for fluoxetine); (8) use of investigational medication within 3 months or a depot neuroleptic within 6 months of placebo lead-in; (9) past treatment with fluoxetine for 8 weeks or longer or with a dose of 40 mg/d or more; (10) initiation of psychotherapy (including cognitive-behavioral therapy) within 4 months of lead-in; (11) significant or unstable medical illness; (12) history of seizures; (13) pregnancy, lactation, or lack of contraception in women of childbearing potential; (14) clinically relevant abnormal laboratory tests; (15) requirement for psychotropic medication; (16) inability to cooperate with the protocol.

Potential participants (n = 296) were screened by telephone, 158 of whom seemed to qualify and were evaluated in person by the first author. Seventy-four patients were enrolled in the study; 54% (n = 40) were self-referred, 35%(n = 26) were referred by a professional, and 11% (n = 8) responded to an advertisement. The protocol and informed consent documents were approved by the institutional review board. After a thorough description of the study to patients, including its rationale, procedures, and potential risks and benefits, voluntary written informed consent was obtained.

After completing the screening evaluations during 2 to 3 weeks, patients received single-blind pill placebos for 1 week. Patients whose BDD-YBOCS score decreased by 30% or more during this week were terminated from the study; those who still qualified were randomly assigned in a 1:1 ratio to 12 weeks of double-blind treatment with fluoxetine or placebo. Randomization was performed by a technician with no clinical contact who kept the code during the trial. A computer-generated urn randomization procedure⁴¹ balanced the 2 study groups for current major depression, current OCD, and whether the appearance-related beliefs were currently delusional.

Patients received fluoxetine, 20 mg/d, or pill placebo–equivalent(identical-appearing capsules) for 2 weeks. The dose was increased by 20 mg/d every 10 days to a maximum of 80 mg/d, as tolerated, and could be decreased by 10 mg/d or 20 mg/d at any time as clinically indicated. Patients who could not tolerate at least 20 mg of fluoxetine or placebo per day were terminated from the study. No other psychotropic medications were taken except chloral hydrate, 0.5 to 2.0 g/d, no more than 3 times per week if needed for insomnia. Psychotherapy of any type was not initiated during the study.

One of us (K.A.P.), who was blind to adverse events, did all of the ratings. Adverse events were assessed by the second author, who also adjusted the medication dose. Returned medication was counted to verify compliance. After completing the double-blind phase, placebo-treated patients were offered12 weeks of open-label fluoxetine treatment.

The BDD-YBOCS was selected a priori as the primary outcome measure. This 12-item semistructured clinician-rated scale,³⁹ adapted from the Y-BOCS,^{42 - 43} assesses BDD severity during the past week. Each item is scored on a 5-point scale from 0 (least symptomatic) to 4 (most symptomatic), with a total score of 0 to 48. Items assess preoccupation with the perceived defect (time occupied, interference with functioning due to the preoccupation, distress, resistance, and control), associated repetitive behaviors, such as mirror-checking (time spent, interference with functioning, distress if the behaviors are prevented, resistance, and control), insight, and avoidance. The scale is reliable, valid, and sensitive to change.³⁹ Response was defined as a 30% or greater decrease in total score, an empirically derived cut point that corresponds to clinically significant improvement in BDD symptoms.³⁹ Secondary measures of BDD outcome were the CGI improvement scale⁴⁰ and the National Institute of Mental Health Global Obsessive Compulsive Scale,⁴⁴ a 15-point global rating of BDD severity (BDD-NIMH). Clinical Global Impressions Scale ratings were done for BDD symptoms and for global outcome; patients and the clinician provided separate ratings. A CGI score of much or very much improved (score of 1 or 2) was defined as improvement. The CGI severity scale⁴⁰ assessed illness severity at baseline.

The delusionality (insight) of beliefs about appearance (patient's conviction that his or her appearance was abnormal) was assessed with the Brown Assessment of Beliefs Scale. This 7-item semistructured clinician-administered scale assesses delusionality during the past week both dimensionally and categorically.⁴⁵ Patients were categorized at baseline as delusional(n = 27) or nondelusional (n = 37) using an a priori empirically derived cut point.⁴⁵ The scale is reliable, valid, and sensitive to change.⁴⁵ Items are conviction, perception of others' views, explanation of differing views, fixity, attempts to disprove beliefs, insight (recognition that the belief has a psychiatric etiology), and ideas/delusions of reference. Scores range from 0 to 24. Three patients could not be assessed because they had clearly noticeable skin lesions due to skin picking.

The 17-item Hamilton Rating Scale for Depression (HAM-D)⁴⁶ assessed current severity of depressive symptoms, and the Brief Psychiatric Rating Scale (BPRS)⁴⁷ assessed severity of psychopathology. The Social and Occupational Functioning Scale (SOFAS)³³ evaluated psychosocial functioning, and the Global Assessment of Functioning (GAF)⁴⁸ rated symptom severity and functioning.

All patients received an electrocardiogram, physical examination, and standard laboratory tests, including a drug screen; another physical examination was done at study end point. The BDD-YBOCS, BDD-NIMH, Brown Assessment of Beliefs Scale, and HAM-D were administered at each visit, the CGI severity scale was rated at baseline, and the CGI improvement scale was rated at all visits subsequent to randomization. The BPRS, GAF, and SOFAS were completed at baseline and end point. At each visit, patients were asked whether they had any adverse physical symptoms since the last visit, which were rated for severity, action taken, outcome, and seriousness. At termination, patients and the first author judged which treatment had been received.

All data were double entered to ensure accuracy. Data were analyzed with the Statistical Package for the Social Sciences, version 6.1 for the Macintosh (SPSS Inc, Chicago, Ill). All tests of group differences on outcome variables used an intent-to-treat analysis plan that included all randomized patients, with last observation carried forward for dropouts. Except for CGI improvement ratings, the dependent variable was the change in outcome from baseline. All tests were 2-tailed. An α level of .05 was used to determine statistical significance.

Efficacy analyses for continuous variables were performed using analysis of covariance (ANCOVA), with baseline measures as the covariate. A repeated-measures ANCOVA further tested for a time effect, a treatment effect, and a time-by-treatment interaction on the dependent measure between groups. The ANCOVA, using the baseline score as a covariate, determined when significant drug-placebo differences occurred. When tests of sphericity were violated, the Huynh-Feldt statistic was used. Independent sample t tests were also used to analyze continuous variables (when tests of homogeneity of variance were violated, the nonparametric alternative, the Mann-Whitney U test, was used). The χ² analysis and Fisher exact test were used to analyze dichotomous variables. The effect size (f) and 95% confidence intervals were calculated to measure treatment strength. An effect size (f) of 0.25 is considered a medium effect and 0.40 is considered large.⁴⁹ The Pearson product moment correlation coefficient was used to examine correlations between variables, and stepwise multiple regression assessed predictors of treatment outcome. The incidence of adverse events was based on the number of patients who reported a given treatment-emergent event (ie, an event that first appeared or worsened during double-blind therapy).

Of the 74 enrolled patients, 6 were discontinued from the study during the screening period and 1 was discontinued after the single-blind placebo lead-in week because her BDD-YBOCS score decreased to less than 24 (Figure 1). No patients responded to the single-blind placebo. Sixty-seven patients were randomized to receive double-blind treatment with fluoxetine (n = 34) or placebo (n = 33). Three patients (9%) randomized to receive fluoxetine and 5 (15%) randomized to receive placebo discontinued study participation (χ²₁ = 0.64; P = .42).

There were no significant differences between the fluoxetine (n = 34) and placebo (n = 33) groups on baseline demographic and clinical characteristics(Table 1). For example, skin (eg, acne) and hair (eg, hair loss) were the most common appearance concerns (skin:25 [74%] in the fluoxetine group and 26 [79%] in the placebo group; hair:14 [41%] in the fluoxetine group and 19 [58%] in the placebo group). Twelve patients (37.5%) in the fluoxetine group and 15 patients (46.9%) in the placebo group were delusional at baseline. Ongoing psychotherapy (begun before study entry) was received during the study by 3 patients in the fluoxetine group and 3 patients in the placebo group; none of these patients were receiving cognitive-behavioral therapy.

Fluoxetine was superior to placebo for BDD symptoms as measured by the primary and secondary BDD outcome measures and both clinician and patient ratings (Table 2). Controlling for baseline group differences in BDD severity, on the BDD-YBOCS (the primary outcome measure), fluoxetine was more effective than placebo (F_1,64= 16.5; P<.001) beginning at week 8 (F_1,64 = 4.63; P = .04) and continuing at weeks 10 and 12 (Figure 2). The mean change from baseline in the BDD-YBOCS total score was more than twice as large with fluoxetine as with placebo treatment (35% vs 14% decrease; t₆₅ = 3.54; P = .001), with a similar decrease for BDD preoccupations and repetitive behaviors. The response rate on the BDD-YBOCS to fluoxetine was 53% (18/34) vs 18% (6/33) to placebo (χ²₁ = 8.8; P = .003). The treatment effect size was medium to large (f = 0.35; 95% confidence interval, 0.22-0.48; d equivalent = 0.70).

On the clinician-rated BDD-CGI, 14 patients (41%) treated with fluoxetine were much improved and 5 (15%) were very much improved. Functional impairment as assessed by the GAF and SOFAS improved more with fluoxetine than placebo(Table 2).

The mean ± SD time to fluoxetine response (as assessed by a 30% decrease in BDD-YBOCS score) was 7.7 ± 3.5 (range, 2-12) weeks and to placebo was 5.3 ± 3.2 (range, 1-8) weeks (t₂₉ = 1.97; P = .6). The mean ± SD fluoxetine dose at end point was 77.7 ± 8.0 (range, 40-80) mg/d; the fluoxetine equivalent in the placebo group was 76.0 ± 13.1 (range, 20-80) mg/d.

Of the 21 patients treated with open-label fluoxetine after placebo treatment during the double-blind phase (mean ± SD fluoxetine dose at end point, 61.1 ± 21.4 mg/d), 5 (24%) responded to the BDD-YBOCS. Scores decreased from a mean ± SD score of 29.3 ± 7.4 to 22.3 ± 7.2 (t₂₀ = 5.14; P<.001). On the clinician-rated BDD-CGI, 9 (43%) responded, with7 (33%) much improved and 2 (10%) very much improved.

In 34 cases (69%), the clinician correctly judged whether the patient had received fluoxetine or placebo; this was the case for 25 (63%) patients. The clinician's judgment was incorrect in 6 (12%) cases and the patient's in 10 (25%). The clinician was unsure of group assignment in 9 (18%) of cases and the patient in 5 (13%).

We tested for a difference in the amount of improvement in BDD symptoms from baseline to end point for delusional (n = 27) vs nondelusional (n = 37) patients, covarying for baseline BDD severity. We found no interaction between delusionality and improvement in BDD symptoms over time using a repeated-measures analysis (F_2.7,162 = 1.06; P = .36). The BDD symptoms of patients categorized as delusional at baseline were as likely as those of patients who were nondelusional at baseline to respond to fluoxetine(50% [6/12] vs 55% [11/20]; χ²₁ = 0.08; P = .78). However, BDD symptoms of delusional patients were significantly less likely than those of nondelusional patients to respond to placebo (0%[0/15] vs 35% [6/17]; χ²₁ = 6.51;P = .01). Although delusional patients had more severe BDD symptoms at baseline (delusional patients' mean ± SD BDD-YBOCS scores, 33.0 ± 5.1; nondelusional patients, 29.4 ± 5.6; t₆₂ = 2.65; P = .01), an ANCOVA that controlled for BDD severity at baseline indicated that this did not account for their lower placebo response rates (F_1,61 = 1.91; P = .17). In delusional patients, the response rate of BDD symptoms to fluoxetine was significantly higher than to placebo (50% vs 0%; χ² = 9.6₁; P = .002). This was not the case for nondelusional patients (55% vs 35%; χ²₁= 1.44; P = .23), although power was limited (1−β= .27).

We also examined a second question: with treatment, did patients' conviction that their appearance was abnormal (delusionality) change with fluoxetine compared with placebo treatment? While Brown Assessment of Beliefs Scale scores decreased between baseline and end point for both the fluoxetine and placebo groups, the difference between them was not significant (Table 2). However, scores decreased significantly more in treatment responders than in treatment nonresponders (for both treatment groups combined(F_3.8,233.6 = 9.5; P<.001).

Hamilton Rating Scale for Depression scores improved significantly more with fluoxetine than with placebo (Table2). Change in BDD-YBOCS and HAM-D scores was correlated r = 0.65 (P<.001) for the fluoxetine group and r = 0.58 (P<.001) for the placebo group. Two patients (6%) treated with fluoxetine and 10 (30%) treated with placebo had an increase (ie, worsening) on the HAM-D suicidal ideation item between baseline and end point (P =.001).

In the entire sample, BDD duration, BDD severity, and the presence of a personality disorder, current OCD, or current major depression did not predict response of BDD in a stepwise regression analysis. Furthermore, with regard to depression, there was no main effect of the diagnosis of major depression at baseline on BDD outcome (F_1,63 = 1.0; P= .32) and no interaction between major depression at baseline and treatment group on BDD outcome (F_1,63 = 0.70; P= .41). There was a significant effect of fluoxetine on BDD symptoms even after covarying for main and interactive effects of baseline depressive symptoms(HAM-D score; F_1,64 = 5.4; P = .02). As previously noted, delusional patients were less likely than nondelusional patients to respond to placebo. Treatment outcome did not differ by sex or minority status, although power to assess the latter was limited.

Treatment-emergent adverse events (irrespective of relationship to study drug) were reported by 82% of patients (n = 28) treated with fluoxetine and64% (n = 21) treated with placebo (χ²₁ = 3.0; P = .08). The only adverse events that were significantly more frequent with fluoxetine treatment were drowsiness and stomach/abdominal discomfort (Table 3). Five patients(7%) took chloral hydrate for insomnia. Adverse events were often transient, and nearly all were of mild to moderate severity and well tolerated. No patients discontinued the study because of adverse events. No serious adverse events(eg, suicide attempts or hospitalizations) occurred.

This study, the first placebo-controlled trial on BDD, indicates that fluoxetine is safe and more effective than placebo for BDD, including delusional patients. Fluoxetine was more effective for BDD on the primary and secondary BDD outcome measures and as assessed by both clinician and patient ratings, with a medium to large effect size. Depressive symptoms, global symptomatology, and functioning also improved significantly more with fluoxetine than with placebo.

Consistent with previous studies, fluoxetine was as effective for BDD symptoms in delusional as in nondelusional patients.^{3 ,7 ,28 ,31 - 32} Unlike one previous study³¹ in which an SRI was even more effective for delusional than nondelusional patients, we found a similar fluoxetine response rate in both groups, although in our study, the incremental effect of fluoxetine over placebo for BDD symptoms was greater for delusional than nondelusional patients. While patients with delusional symptoms are generally treated with antipsychotics, this finding suggests that SRIs may be effective for patients with certain types of delusional symptoms.³² Whether antipsychotics alone are effective for delusional BDD has received virtually no investigation and needs to be studied. It is striking that no delusional patients responded to placebo. Given that delusional and nondelusional patients had significantly different placebo response rates, delusionality should be carefully assessed in future BDD treatment studies.

Our finding that delusionality (patient's conviction that his or her appearance was abnormal) did not improve significantly more with fluoxetine than with placebo contrasts with a previous study in which delusionality (insight) improved more with clomipramine than with desipramine (although we did find that delusionality improved more in treatment responders than in nonresponders).³¹ The reason for these somewhat discrepant findings is unclear, although the 2 studies used different measures of delusionality. Further investigation of change in insight with SRI treatment is needed.

Although the dose was increased relatively rapidly, mean ± SD time to fluoxetine response (7.7 ± 3.5 weeks) was lengthy, consistent with previous studies indicating time to response of 6 to 9 weeks.^{25 ,28 - 29} The mean ± SD fluoxetine dose at study end point was relatively high (77.7 ± 8.0 mg/d), although we attempted to reach 80 mg/d to avoid undertreatment. The efficacy of lower fluoxetine doses is unknown, and dose-finding studies are needed to ascertain the optimal dose of fluoxetine and other SRIs. Despite the relatively rapid titration and the high mean dose attained, the medication was generally well tolerated.

The only predictor of treatment outcome was delusionality, with delusional patients less likely than nondelusional patients to respond to placebo. Consistent with previous studies,^{28 - 29} BDD symptoms improved regardless of whether patients had major depression or OCD at baseline. Although BDD severity did not predict treatment outcome, all subjects were required to have at least moderately severe BDD symptoms. Including patients with milder BDD might have yielded an association between BDD severity and treatment response and produced a higher placebo response rate, as has been found in other disorders.^{50 - 52} This issue, however, requires investigation.

While the response rate on the BDD-YBOCS to open-label fluoxetine (subsequent to the double-blind phase) was relatively low, several patients were not classified as responders but their BDD-YBOCS scores decreased by nearly 30%. The mean magnitude of change on the BDD-YBOCS with open-label fluoxetine treatment was nearly as great as with double-blind treatment. In addition, given that86% of patients treated with open-label fluoxetine had been placebo nonresponders, a somewhat lower response rate to subsequent open-label fluoxetine (compared with double-blind fluoxetine) might be expected.

This study has several limitations characteristic of efficacy trials. It was conducted in a university-affiliated private psychiatric hospital, and the sample was selected to meet strict inclusion and exclusion criteria. Patients with milder BDD symptoms were excluded as were patients who were highly suicidal or who needed inpatient treatment. Future studies are required to determine how generalizable the results are to other populations of patients with BDD. Another limitation is that a longer treatment trial might have yielded a slightly higher fluoxetine response rate since an open-label fluvoxamine trial found that 5.3% of responders required more than 12 weeks to respond.²⁹

These results, while promising, require replication. Placebo-controlled studies of other SRIs and parallel-group studies comparing SRIs with other medications (eg, antipsychotics) are needed, as are longer-term treatment studies (eg, continuation and maintenance studies), especially because BDD seems to be a chronic illness.⁷ While it is our clinical impression that the response of BDD to SRIs is usually maintained or further increases over time with continued treatment, this impression requires empirical validation. It is worth underscoring that only slightly more than half of patients responded to fluoxetine, and even though their response was clinically significant, it was usually partial. It is therefore critically important to determine whether adding other pharmacologic agents or psychotherapy(eg, cognitive-behavioral therapy) to fluoxetine or other SRIs might enhance treatment outcome. In the meantime, this study suggests that fluoxetine is a safe and effective treatment for BDD—a distressing, relatively common, and severe mental disorder.

Submitted for publication February 28, 2001; final revision received August 3, 2001; accepted August 13, 2001.

This study was supported by a 1995-2000 FIRST Award (R-29; MH54841) from the National Institutes of Mental Health, Bethesda, Md (Dr Phillips). The study medication and matching placebo were supplied by Eli Lilly & Co, Indianapolis, Ind.

We thank Jason Siniscalchi, MS, for assistance with data analysis and graphics.

Corresponding author and reprints: Katharine A. Phillips, MD, Butler Hospital, 345 Blackstone Blvd, Providence, RI 02906 (e-mail: katharine_phillips@brown.edu).