Of 179 participants randomized to treatment (), 2 withdrew consent before receiving study medication and were excluded from all analyses. Additionally, 7 withdrew consent after receiving study medication but without providing any follow-up data and were included in the ITT analysis but excluded from the modified ITT analysis. Of the 177 randomized patients who received study medication, 85 were randomized to escitalopram and 92 to placebo.
Flow of Patients Through the Trial
Randomized patients did not significantly vary from the entire eligible sample in any demographic characteristic (). Proportions of participants recruited from primary/specialty medical (n=88), mental health (n=17), or self-referral (n=72) did not differ between the escitalopram and placebo arms. Of the participants who received escitalopram, 16 (18.5%) dropped out of the study before week 12 (3 dropouts [3%] were due to adverse effects and 1 [1%] was due to hospitalization for a medical condition). Of the participants who received placebo, 17 (18.4%) dropped out before week 12 (4 dropouts [4%] were due to adverse effects and 1 [1%] was due to worsening GAD). All other dropouts were due to refusing to continue the study (ie, did not want to continue to participate in the trial).
Baseline Characteristics by Treatment Group
There were no statistically significant differences between the escitalopram and placebo groups in overall dropouts, timing of dropouts, or dropouts due to adverse effects. Participants who dropped out before week 12 were significantly different from completers in terms of baseline severity on Hamilton Anxiety Rating Scale (25.1; 95% confidence interval [CI], 22.9-27.2; vs 22.6; 95% CI, 21.9-23.2; P=.003) and Hamilton Depression Rating Scale (14.0; 95% CI, 12.3-15.7; vs 11.6; 95% CI, 11.0-12.2; P=.02), co-prescription of benzodiazepine (dropout rate for benzodiazepine users, 33.3% [n=9/27]; and for nonusers, 16.0% [n=24/150]; P=.04), and race (dropout rate for white patients, 15.9% [n=23/145]; and for black patients, 31.2% [n=10/32]; P=.04); however, none of these variables differed in terms of proportions of dropouts in the escitalopram vs placebo groups.
For study completers, a patient-reported treatment guess was obtained to determine the success of blinding. Treatment guesses were obtained in 140 patients; of those randomized to escitalopram, 36 of 66 (55%) guessed correctly; of those randomized to placebo, 43 of 74 (58%) guessed correctly (Fisher exact, P=.18). These results suggest adequate success of blinding. Of patients randomized to escitalopram, 34 of 36 (94%) correct guessers were responders, while in the placebo group, 14 of 43 (33%) were responders (Fisher exact, P<.001), suggesting that correct guesses were associated with response in the active treatment group only.
Estimates of the cumulative incidence of response and time to response are shown in . The cumulative incidence of response was higher in the escitalopram group than in the placebo group (mean response rate, 69%; 95% CI, 58%-80%; vs 51%; 95% CI, 40%-62%; P=.03). In the ITT analysis, the response was not different between groups (57%; 95% CI, 46%-67% for escitalopram; vs 45%; 95% CI, 35%-55% for placebo; P=.11); however, in a modified ITT analysis, response was higher in the escitalopram group than in the placebo group (60%; 95% CI, 50%-71%; vs 45%; 95% CI, 36%-56%; P=.048). In an exploratory analysis, 37 of 49 escitalopram responders (76%) and 30 of 42 placebo responders (71%) had at least 1 subsequent Clinical Global Impressions-Improvement scale score in the response range, suggesting that response typically persisted.
Kaplan-Meier Survival Curve With Escitalopram or Placebo
In contrast with the cumulative incident response results, time to response (6 weeks median in the escitalopram group vs 10 weeks in placebo) did not significantly differ (P
=.19). Examination of time-to-response showed a crossover in hazard functions (crossover of escitalopram and placebo curves between weeks 4 and 6). Therefore, as an exploratory analysis, we separately examined time-to-response analyses for the 2 intervals (baseline to week 4 and week 4 to week 12) using the Gail and Simon method.40
The initial 4-week interval showed no difference between escitalopram and placebo (P
=.43), while the final 8-week interval showed a significant difference favoring escitalopram (P
=.01) (). Week 4 was the point of the titration of study medication to 2 pills, which was done in 55 of 71 patients (77%) in the escitalopram group and 66 of 83 patients (80%) in the placebo group (Fisher exact, P
shows results from mixed-effects models of change in anxiety symptoms and role functioning. Escitalopram was significantly better than placebo for Clinical Global Impressions-Improvement scale, Penn State Worry Questionnaire, activity limitations subscale, and role-emotional impairment and social function subscales. also displays effect size (Cohen d) with 95% CIs, showing that the largest effect size was observed with Clinical Global Impressions-Improvement scale (0.93; 95% CI, 0.50-1.36; P<.001) and role-emotional impairment subscale (0.96; 95% CI, 0.03-1.90; P=.04); the rest of the scales showed effect size in the small-to-moderate range (Penn State Worry Questionnaire: 0.30; 95% CI, 0.23-0.48; P=.01; activity limitations: 0.32; 95% CI, 0.01-0.63; P=.04).
Improvements in Anxiety Symptoms and Disability and Quality of Life Measures
Baseline variables significantly associated with poorer overall time to response were presence of comorbid major depressive disorder (χ2=3.97; hazard ratio [HR], 0.58; 95% CI, 0.34-0.99; P=.046) and higher medical comorbidity scores in the Cumulative Illness Rating Scale for Geriatrics (χ2=4.16; HR, 0.94; 95% CI, 0.89-1.00; P=.04). However, a multivariate model including these variables as well as age, sex, referral source, comorbid benzodiazepine use, age of GAD onset, and MMSE score did not affect the significant association of treatment group with response (ie, no significant covariate×treatment group interactions, P values ranged from .22-.99, and no change in the significance of the treatment group main effect). Therefore, no baseline variable appeared to moderate medication efficacy.
Tolerability and Safety
Adverse effects data are shown in . Most participants reported at least 1 adverse effect during the study. Of individual adverse effects, the most common in individuals taking escitalopram was fatigue or somnolence. shows the course of adverse effect endorsement over time. The only single time in which a significantly higher proportion of participants receiving escitalopram had adverse effects (P=.007) was at week 6, which was the appointment after most participants had a dosage increase from 10 to 20 mg/d.
Proportion of Participants Who Experienced at Least 1 Adverse Effect by Treatment Group
Rates of Overall Self-reported Adverse Effects Over Time in the Escitalopram and Placebo Groups
No participants in either group had a serious or unexpected adverse event. One patient (randomized to receive escitalopram) was hospitalized for bacterial endocarditis and was removed from the study after week 3. With respect to vital sign changes, we found a significant effect by treatment group on pulse, systolic blood pressure, and diastolic blood pressure, with a greater decrease in participants receiving escitalopram than those receiving placebo.
In an exploratory, post hoc evaluation of changes in pulse and blood pressure as a function of baseline values, the greater decrease in the escitalopram group in blood pressure was isolated to those participants with high baseline values (). Although the escitalopram group had a greater decrease in pulse, prevalence of bradycardia (pulse < 60/min) did not change over time in either the escitalopram or placebo groups (19/69 [27%] at baseline vs 22/62 [35%]; t test, P=.33; vs 14/82 [17%] at baseline vs 16/69 [23%]; t test, P=.35). Weight changes over time were not significant in the overall sample or between groups by mixed-effect model (for time: standard estimate = 0.07, F1632=0.36, P=.55; and for group × time interaction: standard estimate=−0.12, F1632=0.41, P=.52).
Changes in Blood Pressure and Pulse by Treatment Group and Baseline Valuesa