Differential Response to Placebo Among Patients With Social Phobia, Panic Disorder, and Obsessive-Compulsive Disorder

Data were obtained from 1) 26 patients with OCD who received placebo in a randomized clinical trial that compared 12 weeks of cognitive behavior therapy, clomipramine, their combination, and placebo (8); 2) 60 patients with social phobia who received placebo in a randomized clinical trial that compared 14 weeks of group cognitive behavior therapy, fluoxetine, their combination, placebo, and cognitive behavior therapy plus placebo (9); and 3) 24 patients with panic disorder who received 12 weeks of placebo in a randomized clinical trial that compared 12 weeks of cognitive behavior therapy, imipramine, their combination, placebo, and cognitive behavior therapy plus placebo (10). In all three studies, written informed consent was obtained after full explanation of the procedures. All studies had similar inclusion criteria (e.g., adults with primary OCD, social phobia, or panic disorder) and exclusion criteria (e.g., no psychosis, mania, or substance abuse). However, unlike the other studies, the panic disorder study (10) did not exclude patients with major depression but did exclude moderate to severe agoraphobia.

All studies included independent raters blind to treatment assignment and used the Clinical Global Impression (CGI) (11) severity and improvement scales, the 17-item Hamilton Depression Rating Scale (12), and a symptom-specific measure: the Yale-Brown Obsessive Compulsive Scale (13) for OCD, the Duke Brief Social Phobia Scale (14) for social phobia, or the Panic Disorder Severity Scale (15) for panic disorder. Psychopharmacologists saw patients for similar amounts of time in each study, and exposure to anxiety-provoking situations was not systematically encouraged during these visits. Effect sizes were calculated for the symptom measures. CGI severity and Hamilton depression scale scores were compared between and within groups by using analyses of variance for pretreatment scores, analyses of covariance for posttreatment scores (covarying pretreatment severity), and paired-sample t tests for within-subject change scores. CGI improvement scores were compared by means of chi-square tests of ordinal data. Analyses were conducted separately for the completers and for all randomly assigned patients with any available data by using the last observation carried forward (intent-to-treat group).

The OCD patients receiving placebo had less symptom improvement than the patients with either social phobia or panic disorder in both the completer (Table 1) and intent-to-treat groups. When the most severe cases were excluded (CGI severity score=6) to eliminate significant pretreatment CGI severity differences between the patient groups (F=2.2, df=2, 56, p=0.12), the patients with OCD continued to show less improvement at posttreatment than those with social phobia or panic disorder (p<0.01, ANCOVA). Ratios of the effect sizes for active medication and exposure therapy to the effect size for placebo were calculated to determine the treatment responsiveness of the groups. All groups showed at least a 40% larger effect size for the active treatments than for placebo, with the OCD patients showing the largest treatment effects, owing to the smallest placebo effects.

Consistent with our hypothesis and earlier literature (6, 16), patients with social phobia and panic disorder who received placebo evidenced a greater reduction in core symptoms than patients with OCD. This difference cannot be explained by patient expectancy (because those ratings did not differ among groups), by pretreatment severity (because differences in outcome persisted when severity was similar), by differential dropout rates (because the rates were similar and the results for the intent-to-treat group were similar to the results for the completers), or by treatment responsiveness (all groups responded to active treatments). One possibility is that placebos may help patients to confront feared situations (i.e., exposure or reduced avoidance) by increasing motivation or tolerance of anxiety. In OCD, compulsions may interfere with the benefit of such exposures (17). More research is needed to determine how, or whether, the subtle avoidance behaviors engaged in by patients with panic disorder or social phobia differ from compulsions and whether these differences account for placebo effects. Alternatively, OCD may be less likely to spontaneously remit than panic disorder or social phobia, although the chronic course of most anxiety disorders makes this less likely (18). The mechanism interfering with placebo response in OCD may also account for the lower rate of response to benzodiazepines than in patients with other anxiety disorders (19). The mechanisms of action of placebo response among the anxiety disorders should be further examined and should be compared with the mechanisms of action of active treatments in studies using larger subject groups, multiple contexts (e.g., medication-only trials), and other disorders to understand better the psychological and neurobiological mechanisms involved in the maintenance and treatment of anxiety disorders.

Received May 8, 2003; revision received Oct. 31, 2003; accepted Feb. 10, 2004. From the Center for Treatment and Study of Anxiety, University of Pennsylvania School of Medicine; the Center for Anxiety and Related Disorders, Boston University; the Anxiety and Traumatic Stress Program, Duke University, Durham, N.C.; the Department of Psychiatry, Mt. Sinai School of Medicine, New York; the Panic, Anxiety, and Traumatic Grief Program, Western Psychiatric Institute, Pittsburgh; the Anxiety Disorders Clinic, New York State Psychiatric Institute, New York; and the Department of Psychiatry, Yale University School of Medicine, New Haven, Conn. Address reprint requests to Dr. Huppert, Center for Treatment and Study of Anxiety, University of Pennsylvania School of Medicine, 6th Floor, 3535 Market St., Philadelphia, PA 19104; [email protected] (e-mail).