Generalized anxiety disorder is characterized by excessive anxiety and worry about a number of everyday and routine events or activities such as work or academic performance. Afflicted persons find it difficult to control the worry and are restless, easily fatigued and irritable, and frequently complain of muscle tension and difficulty concentrating and sleeping. As defined by DSM-IV, the anxiety, worry, and physical symptoms are associated with significant subjective distress and impairment in social and occupational functioning.

Generalized anxiety disorder is a common anxiety disorder in the United States, with a lifetime prevalence of 5.1% (1). According to an international multicenter study sponsored by the World Health Organization, generalized anxiety disorder is, after major depression, the second most frequent psychiatric disorder in the primary care setting (2), with approximately 8% of patients suffering from this illness (3). The course of generalized anxiety disorder tends to be chronic and recurrent (4), with less than one-half of cases remitting without treatment (5). Epidemiological research has also shown that the impairment caused by generalized anxiety disorder is equivalent in magnitude to that of major depression, with sufferers having a 2.5-times greater risk of social role impairment than healthy persons as measured by scales of social support and negative interaction with family and friends (6). Individuals with generalized anxiety disorder are high utilizers of medical care. At some point during their illness, 48% of generalized anxiety disorder patients seek professional help, and 25% of these patients require medication (7, 8).

Benzodiazepines have traditionally been the mainstay in the pharmacotherapy of anxiety symptoms, and treatment trials demonstrate that this medication class is of benefit in treating generalized anxiety disorder (9, 10). Although effective in the short term, the benzodiazepines are associated with physical dependence, particularly with prolonged use (11). Current practice is therefore to restrict the use of these agents to less than 4 weeks (12), a period that may be insufficient for treating generalized anxiety disorder. Buspirone, a partial agonist at the 5-HT_1A receptor, has also been studied as a treatment for generalized anxiety disorder. In studies with patients having symptoms corresponding to DSM-III generalized anxiety disorder, buspirone was shown to be effective (13). However, in a clinical trial assessing treatment of patients meeting DSM-IV criteria for generalized anxiety disorder, this medication failed to separate from placebo on several outcome measures (14).

Antidepressants have also been investigated for their antianxiety properties. Rickels et al. (15) compared the efficacy of imipramine and trazodone to diazepam in 230 patients with generalized anxiety disorder. Diazepam was more effective in the first 2 weeks of therapy, notably in alleviating somatic symptoms, but throughout the remainder of the 8-week treatment the antidepressants demonstrated comparable anxiolytic effects. Both antidepressants were more effective in alleviating psychic symptoms of anxiety than in alleviating somatic symptoms. However, possibly because of the relatively high frequency of unpleasant side effects, older-generation antidepressants have not been further investigated in this area. Among the newer antidepressants, the extended-release formulation of the serotonin-noradrenaline reuptake inhibitor venlafaxine has been evaluated in three placebo-controlled trials (14, 16, 17) as a treatment for DSM-IV generalized anxiety disorder. In two 8-week studies, fixed doses of extended-release venlafaxine (75 and 150 mg/day [14] and 75, 150, and 225 mg/day [16]) demonstrated greater effects than placebo on several measures, with the greatest effects observed at the highest dose. A 6-month, flexible-dose trial also demonstrated efficacy for extended-release venlafaxine, 75–225 mg/day, with the greatest portion of the effect occurring by week 8 (17).

Despite the widespread use of selective serotonin reuptake inhibitors for the treatment of depression and anxiety disorders, paroxetine is the only agent of this class to have been investigated for the treatment of generalized anxiety disorder. In a randomized trial involving 81 outpatients, paroxetine was compared to imipramine and 2′-chlordesmethyl-diazepam (not available in the United States) (18). As in the study by Rickels et al. (15), the benzodiazepine was more effective during the first 2 weeks but was surpassed thereafter by the antidepressants. Improvement of generalized anxiety disorder symptoms with paroxetine continued to the 8-week study endpoint, and both antidepressants were more effective than the benzodiazepine in alleviating psychic symptoms of anxiety rather than the somatic symptoms. These findings, together with consideration of paroxetine’s anxiolytic effect in depressed patients (19) and efficacy in the treatment of other anxiety disorders (20, 21), led to the development of the present study.

This report presents the results of a placebo-controlled, fixed-dose study of paroxetine for the treatment of generalized anxiety disorder. Since the DSM-IV definition is explicit regarding the presence of significant social and occupational impairment, the design of this study also included a measure of treatment effect on patient disability.

Patients

The subjects were male and female outpatients, 18 years and older, who fulfilled DSM-IV criteria for generalized anxiety disorder as determined by psychiatric evaluation, which included the Mini International Neuropsychiatric Interview (22) as a screening instrument. A score of at least 20 on the Hamilton Rating Scale for Anxiety was required, including a score of 2 or more on items 1 (anxious mood) and 2 (tension). Patients with another psychiatric disorder, including major depression, were excluded, as were patients with a score >17 on the Montgomery-Åsberg Depression Rating Scale (23). Patients were not eligible for the study if they were taking other psychoactive medications, had received electroconvulsive therapy or formal psychotherapy within the 3 months before the initial assessment, or had a clinically significant medical condition. Also excluded were women of childbearing potential who did not practice a reliable method of contraception.

This study was conducted at 50 sites in the United States and Canada. The protocol was approved by institutional review boards at each of the centers or by a regional review board. Following complete description of the study, written informed consent was obtained from each subject before any procedures were carried out.

Study Design

Patients who met the inclusion criteria began a 1-week placebo run-in phase to evaluate compliance with study procedures and to eliminate early placebo responders. At the conclusion of the run-in phase, eligible patients were randomly assigned to a double-blind study medication if the initial Hamilton anxiety scale total score did not decrease by ≥20% and all other entry criteria were fulfilled. Subjects who were noncompliant or had unresolved clinically significant abnormalities in hematology, blood chemistry, ECG, or physical examination findings were not assigned to a condition. All patients receiving paroxetine started at 10 mg/day for the first week of treatment. Thereafter, the paroxetine dose was increased by 10 mg/week until the fixed dose was reached (at the beginning of week 2 and week 4 for the 20-mg and 40-mg arms, respectively). The total treatment period was 8 weeks; assessments were at weeks 1, 2, 3, 4, 6, and 8.

Efficacy Assessments

The mean change from baseline in total score on the Hamilton anxiety scale was defined as the primary variable. Because previous research indicated differential treatment effects on the psychic and somatic components of generalized anxiety disorder, the mean change from baseline in the psychic and somatic subscales of the Hamilton anxiety scale were selected as relevant secondary efficacy parameters. Changes in items 1 and 2 of the Hamilton anxiety scale, which encompass the cardinal symptoms of generalized anxiety disorder (anxiety, excessive worrying, and tension), were analyzed separately. Other secondary measures were the mean change from baseline in score on the anxiety subscale of the patient-rated Hospital Anxiety and Depression Scale (24) and the Clinical Global Impression (CGI) severity of illness scale. Response was defined as a score of 1 ("very much improved") or 2 ("much improved") on the global improvement rating of the CGI. Remission, which represents complete or near complete symptom resolution including resolution of functional impairment, was defined as a Hamilton anxiety scale total score of ≤7 (25). Functional impairment was assessed at baseline and at weeks 4 and 8 with the Sheehan Disability Scale (26), which quantifies the patient’s perception of disability within the three domains of work, social life, and family life.

Data Analysis

The study was designed to provide 90% power for the analysis of the primary variable at an overall significance level of 5%. Continuous efficacy variables were analyzed by analysis of variance (ANOVA) model with treatment and site effects. Dichotomous data (response, remission) were analyzed by using logistic analysis with treatment in the model. The effect of site was not included in this model because of the small number of patients per treatment group for some sites. The results for these analyses were generated by using the general linear model and categorical modeling procedures, respectively, in SAS (SAS Institute, Cary, N.C.). All hypothesis tests were two-sided. For the continuous variables, the effect of the treatment-by-site interaction was assessed during the model building process at the 10% level of significance. The treatment-by-site interaction was not significant (p>0.10) for any variable with the exception of the somatic subscale of the Hamilton anxiety scale. For this variable, the treatment effect, however, remained nonsignificant regardless of whether the interaction was included in the model or not. As a result, further analysis in the form of pairwise comparisons has not been presented for the Hamilton anxiety scale somatic subscale. All other statistical tests were performed at the 5% significance level. Pairwise comparisons (paroxetine, 20 mg/day, versus placebo and paroxetine, 40 mg/day, versus placebo) were carried out only if the overall F test was significant. When comparing the two paroxetine groups with placebo, Dunnett’s multiple comparison procedure was used to maintain an overall alpha level of 0.05. This led to an adjusted significance level of 0.027 for the two comparisons. The confidence interval was constructed by using the same adjusted alpha level used for treatment group comparisons.

All analyses were carried out with the intent-to-treat population, i.e., all patients who were randomly assigned to a double-blind study medication condition and had at least one postbaseline efficacy assessment. Two data sets were used: 1) the last-observation-carried-forward data set, in which missing data at subsequent timepoints were estimated by using the last recorded data, and 2) the observed cases data set, in which the available data at each study visit are used.

Patient Characteristics

Six hundred sixty-one patients entered the placebo run-in phase, of whom 566 were randomly assigned to a double-blind medication condition (intent-to-treat population). The reasons for dropout during the placebo run-in phase and the patient progression during the 8-week treatment period are presented in F1. The demographic and baseline clinical characteristics of the randomly assigned patients are summarized in t1. There were no significant differences among the treatment groups in terms of severity of illness or impairment. Likewise, there were no meaningful differences in age at study entry, gender, race, age at onset of generalized anxiety disorder, or duration of generalized anxiety disorder symptoms.

Four hundred twenty-six patients (75.3%) completed the 8-week treatment. Completion rates did not differ substantially among patients given placebo (77.8%), 20 mg of paroxetine (76.1%), or 40 mg of paroxetine (72.6%) (χ²=1.42, df=2, p<0.50), and there were no significant differences in the demographic or clinical characteristics between the patients who dropped out and those who completed the study. The most frequent reason for study discontinuation was the occurrence of adverse events (placebo group: N=12 [6.7%]; 20-mg paroxetine group: N=19 [10.1%]; 40-mg paroxetine group: N=24 [12.2%]), which did not significantly differ by treatment group (χ²=3.30, df=2, p<0.20). The number of patient withdrawals from the study because of lack of efficacy, protocol violations, or being lost to follow-up was similar across the three treatments (F1).

Efficacy

F2 illustrates the effects of the study treatments on the total score and the psychic anxiety subscale score of the Hamilton anxiety scale over the 8-week treatment period. For the last-observation-carried-forward data set, patients treated with either 20 or 40 mg/day of paroxetine showed a significantly greater mean change from baseline in total score on the Hamilton anxiety scale compared with patients taking placebo (t2 presents a summary of ANOVA results). The mean changes in the psychic anxiety subscale score were also significantly greater in the paroxetine groups than in the placebo group.

The effects of paroxetine observed on these parameters of the Hamilton anxiety scale were corroborated by the findings for the other efficacy measures. The results are summarized in t2 and demonstrate that, with the exception of the somatic anxiety subscale, there were significant group differences in the mean changes from baseline for all secondary efficacy variables, including the patient-completed anxiety subscale from the Hospital Anxiety and Depression Scale.

For the last-observation-carried-forward data set, 61.7% and 68.0% of the patients in the 20-mg and 40-mg paroxetine groups, respectively, fulfilled the defined response criterion (CGI improvement rating of 1 or 2) at endpoint compared with 45.6% of the placebo patients (χ²=20.2, df=2, p<0.001). Of the paroxetine patients who completed 8 weeks of treatment (observed cases data set), 68% and 80% achieved response in the 20- and 40-mg regimens, respectively, compared with 52% of patients given placebo (χ²=24.3, df=2, p<0.001).

For the last-observation-carried-forward data set, significantly greater rates of remission were achieved with both paroxetine regimens relative to placebo (placebo=20%; paroxetine, 20 mg=30%; paroxetine, 40 mg=36%) (χ²=11.20, df=2, p=0.004). The proportion of patients who completed 8 weeks of treatment and met the criteria for remission was 36% and 42% in the 20- and 40-mg paroxetine groups, respectively, compared with 24% in the placebo group (χ²=14.26, df=2, p<0.001).

The results of the last-observation-carried-forward analysis also demonstrated significantly greater improvement for the paroxetine groups compared with the placebo group in social and occupational functioning as measured by the Sheehan Disability Scale (t2). This improvement reflects the statistically significant changes seen in paroxetine-treated patients on each of the three domains of the Sheehan Disability Scale: work, social life, and family life (F3).

Tolerability

Fewer patients in the placebo group (74%) reported at least one adverse event during the treatment period, compared with 88% and 86% in the paroxetine 20- and 40-mg groups, respectively (χ²=13.21, df=2, p<0.002). The most common adverse events occurring with paroxetine treatment are shown in t3. Most adverse events were mild to moderate in severity and were more likely to occur at the onset of treatment and to diminish over time. For the most common adverse event (somnolence), discontinuation of treatment occurred in 0.6% of the placebo group (N=1 of 180), 3.7% of the 20-mg paroxetine group (N=7 of 189), and 1.5% of the 40-mg paroxetine group (N=3 of 197) (χ²=5.10, df=2, p<0.08).

The incidence of laboratory abnormalities was low and similar in the placebo and paroxetine groups (3.9% and 4.6%, respectively). Clinically meaningful decreases in blood pressure and changes in pulse were infrequent and were observed more with placebo (2.2%) than with paroxetine (1.3%). Relative to baseline, the mean body weight in the placebo group (78.9 kg) rose by 0.4% (0.3 kg), and the mean body weight in the paroxetine groups (77.2 kg) decreased by 0.3% (0.2 kg).

This 8-week, placebo-controlled study examined the efficacy and safety of fixed doses of paroxetine (20 and 40 mg/day) in patients with generalized anxiety disorder. Considering that the mean duration of symptoms was 10 years, it is apparent that the patients in this study were experiencing long-term illness. Both paroxetine doses were effective in alleviating generalized anxiety disorder symptoms, as demonstrated by mean decreases from baseline in total score on the Hamilton anxiety scale that were significantly greater than those seen with placebo. These results were confirmed by significantly greater improvements seen in all but one of the secondary outcome measures for paroxetine over placebo (t2). Improvements in score on the psychic anxiety subscale of the Hamilton anxiety scale, which encompasses most of the DSM-IV criteria for generalized anxiety disorder, improved in the paroxetine groups as early as week 3 (F2).

The mean improvement in total score on the Hamilton anxiety scale in the last-observation-carried-forward analysis was 12.5 points and 12.2 points in the 20- and 40-mg paroxetine groups, respectively, compared with 9.3 points for those given placebo. For the patients taking paroxetine, these results represent a 50% reduction in the severity of generalized anxiety disorder symptoms. This reduction is considered clinically significant, since the symptoms that were moderate to severe at the outset of treatment were, on average, of mild intensity after 8 weeks of treatment with paroxetine. The clinical significance of the observed change in Hamilton anxiety scale scores is corroborated by high rates of response and remission. For paroxetine patients who completed the study, the response rates were as high as 80%. Full remission, in which anxiety patients are essentially indistinguishable from healthy counterparts (26), was achieved by 30%–40% of the paroxetine patients who completed the full course of treatment, compared with 20% of patients given placebo.

In this study, the effects of treatment on functional impairment were measured with a validated and user-friendly instrument. Across the treatment groups, the mean baseline score on the Sheehan Disability Scale was 13.9, indicating that patients were moderately to severely impaired. Similar levels of disability have been observed for untreated major depression and panic disorder (26). At week 8, both paroxetine treatment groups demonstrated improvement that was significantly greater than that of placebo in each of the three Sheehan Disability Scale domains. The improvement in the self-rated Sheehan Disability Scale clearly reflects the improvement that was documented by the changes in the clinician-rated Hamilton anxiety scale and CGI.

Paroxetine was generally well tolerated. The dropout rate due to adverse events was 11% for both active-treatment groups. This rate is consistent with the results of studies investigating paroxetine treatment of depression and other anxiety disorders (21, 27, 28) and lower than the rates of 15% and 26% reported in generalized anxiety disorder studies of buspirone (14) and extended-release venlafaxine (14, 16, 17). A clear dose-response relationship was not shown, indicating that 20 mg/day of paroxetine may be effective for the majority of generalized anxiety disorder patients. However, although this study was not designed to test dose response, the numerically greater response and remission rates of the 40-mg treatment group at week 8 suggest that higher doses of paroxetine may be useful in individual cases.

The results of this study and of a flexible-dose study with a similar design (29) offer strong evidence for the efficacy of paroxetine in the treatment of generalized anxiety disorder. However, long-term studies are required to demonstrate that patients with generalized anxiety disorder can benefit from maintenance treatment. Preliminary evidence for the need of maintenance treatment in generalized anxiety disorder has been obtained in a 6-month maintenance study involving 652 patients (30).

To demonstrate generalized anxiety disorder-specific effects of paroxetine, patients with major depression or another anxiety disorder were excluded from the study. Thus, the population consisted of patients with relatively "pure" generalized anxiety disorder. Given the evidence that in primary practice up to 75% of generalized anxiety disorder cases coexist with depression or another anxiety disorder (3), we are aware of the fact that many physicians will infrequently see cases of pure generalized anxiety disorder. Because of this situation, the development of rational approaches to treating pure and comorbid generalized anxiety disorder will require further epidemiological investigations concerning the influence of prior or comorbid psychiatric and somatic illness on treatment outcomes. Nonetheless, over the last 10 years, paroxetine has been established as a treatment for depression and for three anxiety disorders that are frequently comorbid with generalized anxiety disorder (21, 27, 28). The versatility of this drug may therefore represent an advantage in managing an illness that in its own right causes chronic suffering and impairment and that compounded by other psychiatric disorders often poses a difficult medical challenge.

The authors want to acknowledge the following clinical investigators who participated in the collaboration phase of this study: Mohammed Bari, M.D., Chula Vista, Calif.; David Beck, M.D., Columbia, Mo.; Robert Birnbaum, M.D., Boston; Jacques Bradwejn, M.D., Ottawa; David Brown, M.D., Austin, Tex.; William Burke, M.D., Omaha; Bruce Cohen, M.D., Charlottesville, Va.; Pedro Delgado, M.D., Tucson, Ariz.; Eugene Du Boff, M.D., Denver; James Macham Ferguson, M.D., Salt Lake City; William Gilmer, M.D., Chicago; Wayne Goodman, M.D., Gainesville, Fla.; Francis Haines, M.D., East Providence, R.I.; James Hartford, M.D., Cincinnati; Jon F. Heiser, M.D., Irvine, Calif.; Rudolf Hoehn-Saric, M.D., Baltimore; Barbara Kennedy, M.D., Louisville, Ky.; Arifulla Khan, M.D., Bellevue, Wash.; Lorrin Koran, M.D., Stanford, Calif.; Stanley Kutcher, M.D., Halifax, N.S.; Ronald Landbloom, M.D., St. Paul; Anthony Levitt, M.D., Toronto; Michael Liebowitz, M.D., New York; Sidney Lorfald, M.D., Charleston, W.Va.; Denis Mee-Lee, M.D., Honolulu; Matthew Menza, M.D., Piscataway, N.J.; Charles Merideth, M.D., San Diego, Calif.; Kevin Miller, M.D., St. Louis; Charles Nemeroff, M.D., Atlanta; Julie Oldroyd, M.D., Irvine, Calif.; Laszlo Papp, M.D., New York; Teresa Pigott, M.D., Galveston, Tex.; Francisco Jose Pinero-Medina, M.D., Sherbrooke, Que.; Lucy Puryear, M.D., Houston; Charles Ravaris, M.D., Lebanon, N.H.; Karl Rickels, M.D., Philadelphia; Robert Risenberg, M.D., Decatur, Ga.; Pierre Savard, M.D., Montreal; Howard Schwartz, M.D., Miami; Leslie Seiden, M.D., New York; Hope Selarnick, M.D., Philadelphia; Anantha Shekhar, M.D., Indianapolis; Jeffrey Simon, M.D., Brown Deer, Wisc.; John Swerneman, M.D., South Bend, Ind.; Richard Swinson, M.D., Hamilton, Ont.; Karen Weihs, M.D., Washington, D.C.; Richard Weisler, M.D., Raleigh, N.C.; Kenneth Weiss, M.D., King of Prussia, Pa.; Andrew Winokur, M.D., Lebanon, N.H.; Dan Zimbroff, M.D., Upland, Calif.

Presented in part at the 153rd annual meeting of the American Psychiatric Association, Chicago, May 15–20, 2000. Received May 2, 2001; revisions received Oct. 15, 2001, and June 4 and Sept. 24, 2002; accepted Oct. 8, 2002. From the Department of Psychiatry, University of Pennsylvania Medical Center; GlaxoSmithKline, Research and Development, Collegeville, Pa.; and the Institute for Research in Psychiatry, University of South Florida, Tampa. Address reprint requests to Dr. Rickels, Department of Psychiatry, University of Pennsylvania Medical Center, 3535 Market St., Suite 670, Philadelphia, PA 19104-3309. Supported by GlaxoSmithKline. Dr. Rickels assumes full responsibility for the scientific content of this article. He was involved to a significant degree in the protocol development, data collection, data analysis, and manuscript preparation phases of this project. Dr. Rickels had full access to all data collected and major input into the data analysis plan.