Randomized, Placebo-Controlled Trial of Fluoxetine for Acute Treatment of Minor Depressive Disorder

Subjects were recruited from three sites (University of California, San Diego; University of Pittsburgh; and University of Texas, Southwestern) after raters underwent an intensive 3-day training session in which they attained good interrater reliability on the diagnostic criteria and rating instruments used in the investigation. After a telephone screening, subjects who gave written informed consent underwent a physical examination and diagnostic evaluation with the Structured Clinical Interview for DSM-IV (SCID) (25). The initial evaluation also included the depression module of the National Institute of Mental Health Diagnostic Interview Schedule (DIS) (26); the GAF; the MOS-36; the 28-item Hamilton Depression Rating Scale, which was also scored for the 17-item and 21-item versions; the 14-item Hamilton Anxiety Rating Scale (27); the 30-item Inventory of Depressive Symptomatology, clinician-rated version; the 21-item Beck Depression Inventory; the 90-item Hopkins Symptom Checklist (SCL-90) (28); the CGI severity scale; a demographic assessment; and Research Diagnostic Criteria for family history diagnoses (29).

Subjects were reevaluated weekly during a 4-week placebo lead-in period and while in the 12-week treatment phase with the DIS, Inventory of Depressive Symptomatology, 28-item Hamilton Depression Rating Scale, MOS-36, GAF, CGI severity scale, and CGI improvement scale. Clinician- and patient-rated CGI improvement ratings were also made at each weekly visit. The 21-item Beck Depression Inventory was administered at the initial and final study visit of the 12-week treatment period.

As presented more fully in our prior paper (9), we adopted the DIS as our primary diagnostic tool for defining minor depressive disorder and required a level of functional disability based on the GAF and at least one of two MOS-36 subscales (Appendix 1 shows the criteria for minor depressive disorder). To qualify as having “confirmed minor depressive disorder” for the treatment phase of the study, patients needed to meet these criteria at their initial diagnostic visit and at least 3 of the 4 subsequent weeks, including the last 2 weeks of the placebo lead-in period. If, at any time, a subject had developed five or more symptoms of a major depressive episode, as assessed with the DIS depression section, the depression module of the SCID was administered to determine whether the patient had progressed to major depressive disorder—in which case the patient was released from this study and treated outside the protocol.

Subjects were eligible for the study if they were at least age 18 years, healthy, conversant in English, and willing to participate in the full 40-week study consisting of a 4-week single-blind placebo lead-in period, the 12-week double-blind acute treatment period reported on here, and a 24-week crossover treatment period. Subjects also had to have normal results on a physical examination and on laboratory tests, including CBC, urine toxicology screen, urine analysis, and serum chemistries for hepatic and renal function.

Patients were excluded from the study if they had major depressive disorder or dysthymic disorder currently or within the past 2 years; major depressive disorder in partial remission; a loss of a loved one or significant other within the past year that could account for the mood disorder; serious suicidal risk; substance or alcohol abuse or dependence within the past year; a current diagnosis of any axis I disorder; a lifetime diagnosis of bipolar disorder (type I), borderline personality disorder, antisocial personality disorder, psychotic disorder, organic mood disorder, organic psychotic disorder, or schizophrenia; use of psychotropic drugs except chloral hydrate within 7 days of study entry or use of a monoamine oxidase inhibitor within 14 days of starting active treatment; the presence of a serious medical condition not presently stabilized; seizure disorder within the past year; a history of severe allergies or multiple adverse drug reactions; previous nonresponse or adverse reaction to fluoxetine; or previous participation in a fluoxetine study. Patients who met all of the inclusion criteria were randomly assigned to receive fluoxetine or placebo treatment for this 12-week, double-blind investigation.

Patients were instructed to take the study medication each morning, beginning with one 10-mg capsule of fluoxetine or a look-alike placebo. After 2 weeks, patients without intolerable side effects and without significant improvement were instructed to increase the dose to two capsules (20 mg/day of fluoxetine or placebo). For patients who experienced adverse effects, the dose could be titrated down once to 10 mg/day and then back up if symptoms persisted. Chloral hydrate was the only other psychotropic drug allowed during the treatment period. Nonpharmacologic antidepressant treatments, such as light therapy and psychotherapy, were also prohibited.

Demographic characteristics, baseline clinical characteristics, time in study, and adverse events of the fluoxetine and placebo groups were compared by using analysis of variance (ANOVA) for continuous measures and chi-square tests for categorical variables. Ordinal variables (e.g., CGI scale scores) and extremely nonnormally distributed variables (e.g., MOS-36 scores) were analyzed with the Wilcoxon rank sum test.

A much longer than usual placebo lead-in period was used in this study to identify placebo responders and to highlight true drug response. For patients who participated in the acute treatment phase, depression and psychosocial functioning scores from the start and end of the placebo lead-in period were analyzed by repeated-measures ANOVA.

The main hypothesis, that fluoxetine would be superior to placebo in ameliorating minor depressive symptoms, was tested with a last-observation-carried-forward repeated-measures ANOVA that used the SAS GLM procedure (30) to analyze changes in Inventory of Depressive Symptomatology scores from baseline to the last visit in the acute treatment phase. Similar analyses were conducted to test secondary hypotheses concerning differences between treatment groups on all other measures of depressive symptom severity and functioning. Complementary and confirmatory information about the treatment effects of fluoxetine versus placebo was obtained by comparing rates (slopes) of improvement from mixed (random-effects) regression analysis conducted on all weekly measures of depressive symptoms severity and functioning over the 12-week treatment period. Mixed regression analysis of continuous measures was performed with MIXREG (31) software, and analysis of ordinal outcomes (MOS-36 and CGI severity scale scores) was performed with the MIXOR (32) program. In the mixed regression analyses, each dependent variable was modeled on the basis of a random intercept term, a random effect representing time (weeks) in treatment, and fixed effects representing treatment (fluoxetine versus placebo), as well as a treatment-by-week interaction term for group differences in outcome over the course of the study. A first-order autoregressive process (a stationary AR[1] process) was used to characterize the autocorrelation structure. Initial analyses were conducted to test for treatment-by-time-by-site and treatment-by-time-by-gender interactions. Since these interactions were nonsignificant, site and gender were dropped from the ANOVA and mixed regression models that are reported here. A two-tailed alpha level of p=0.05 was used to determine statistical significance. It was not appropriate to make an adjustment for the multiple dependent variables used in this study because of the high correlations among the different measures of depression or psychosocial functioning.

Of the 226 patients who qualified for the study, 64 were not randomly assigned to treatment groups after the 4-week placebo lead-in period for the following reasons: onset of a major depressive episode (N=5), failure to meet minor depressive disorder criteria (N=14), experience of adverse effects (N=4), inability to be contacted (N=6), patient’s decision to withdraw (N=22), contraindicated medical condition (N=10), or other protocol violations (N=3). Of the 162 patients who were randomly assigned to treatment groups (N=81 for the fluoxetine group; N=81 for the placebo group), 59.3% were female, and 90.1% were Caucasian; the average age was 43.5 years (SD=11.7, range=18–72).

As shown in Table 1, the subjects with minor depressive disorder were mildly to moderately depressed at intake and had a mild to moderate level of psychosocial impairment. During the single-blind placebo lead-in period, patients improved significantly (p<0.0001, repeated-measures ANOVA) on all measures of depression severity (range of effect sizes=–0.35 to –0.62) as well as on all measures of psychosocial functioning (range of effect sizes=0.28 to 0.41) (Table 1). At the start of double-blind treatment, the fluoxetine and placebo groups were not significantly different on any demographic characteristics, clinical measures, or psychosocial measures.

Seventy-three percent (59 of 81) of each group completed 12 weeks of treatment. The mean time in treatment was 9.8 weeks for both groups (median for both groups=12 weeks). The 22 dropouts in each treatment group were due to: lack of efficacy (fluoxetine group: N=6; placebo group: N=9), adverse events (fluoxetine: N=3; placebo: N=4), loss to contact (N=3 in each group), moving away (fluoxetine: N=2; placebo: N=1), scheduling problems (fluoxetine: N=7; placebo: N=3), development of a major depressive episode (fluoxetine: N=0; placebo: N=1), and protocol violations (N=1 in each group).

Repeated-measures ANOVA showed that from baseline until the last treatment visit, the fluoxetine group improved significantly more than the placebo group on four of five measures of depressive symptom severity (Table 2). For the clinician-rated Inventory of Depressive Symptomatology score, which was defined a priori as the primary outcome measure for this study, the treatment effect size for fluoxetine was approximately double that for placebo (–1.19 versus –0.61). Compared with the placebo group, the subjects treated with fluoxetine also had significantly greater improvement as measured by the self-rated Beck Depression Inventory and the 17-item and 21-item versions of the Hamilton Depression Rating Scale but not the 28-item version. Secondary analysis with mixed (random-effects) regression analysis showed a statistically greater rate of improvement in the fluoxetine group than in the placebo group in Inventory of Depressive Symptomatology scores (z=2.40, p<0.02), 17-item Hamilton Depression Rating Scale scores (z=2.06, p=0.04), and 21-item Hamilton Depression Rating Scale scores (z=2.19, p<0.03). The fluoxetine group’s greater rate of improvement in 28-item Hamilton Depression Rating Scale scores approached significance (z=1.93, p=0.053). Change in Beck Depression Inventory scores was not analyzed by mixed regression analysis because this instrument was administered only at baseline and at the last visit, rather than on a weekly basis.

Figure 1 is a plot of the mean weekly Inventory of Depressive Symptomatology scores for each treatment group. Both groups showed improvement during the placebo lead-in period. Week-by-week repeated-measures ANOVAs showed that the difference in improvement between the fluoxetine and placebo groups became statistically significant at 7 weeks of treatment. Patterns of response similar to that seen in Figure 1 were found for other measures of depressive symptom severity.

Repeated-measures ANOVA showed that the fluoxetine group did not improve more than the placebo group on any of three measures of psychosocial functioning (Table 2). The secondary analysis with mixed regression, however, showed that the rate of improvement in the GAF score in the fluoxetine group was significantly greater than that in the placebo group (z=2.10, p<0.04).

At baseline, the fluoxetine and placebo groups had similar CGI severity ratings, ranging from “mildly ill” to “markedly ill.” At exit, 40.5% of the fluoxetine patients, compared to 24.1% of the placebo group, were rated as “normal, not at all depressed” (χ²=6.63, df=1, p=0.01). Repeated-measures ANOVA showed a significantly greater improvement on the CGI severity rating for the fluoxetine group than for the placebo group (Table 2). Ordinal mixed regression analysis showed a statistically greater rate of improvement for the fluoxetine group than for the placebo group on this measure (z=2.56, p=0.01). The results of a Wilcoxon rank sum test showed that clinicians’ CGI improvement ratings at the last assessment were significantly better for the fluoxetine group than for the placebo group (p<0.04); the CGI improvement ratings made by the patients were better in the fluoxetine group than in the placebo group, but the difference only approached significance (p=0.057).

Repeated-measures ANOVA showed that the fluoxetine and placebo groups did not differ significantly in improvement as measured by the SCL-90 total score (F=2.30, df=1, 151, p=0.13) or the Hamilton Anxiety Rating Scale score (F=0.56, df=1, 155, p=0.46).

After receiving 10 mg/day of fluoxetine or placebo for 2 weeks, 74 patients (91.4%) in each group were treated with 20 mg/day. By their last visit, each group spent about 70% of their study weeks at this target dose. Adverse events were recorded on the basis of spontaneous reports from study participants. During the placebo lead-in period, 92.6% of the patients in each group reported one or more adverse events. The mean number of adverse events during the 12-week course of treatment was 5.2 (SD=3.6) in the fluoxetine group and 4.6 (SD=3.6) in the placebo group (z=1.45, p=0.15, by Wilcoxon rank sum test). There were no significant differences between the fluoxetine and placebo patients on the worst reported level of severity of adverse events (z=0.46, p=0.65). The only significant difference between the treatment groups was that insomnia was experienced by a greater percentage of the fluoxetine patients than of the placebo patients (24.7% versus 12.4%; χ²=4.09, df=1, p<0.05). No differences were found in sexual side effects, although these effects tend to be underestimated by the spontaneous report method. In summary, apart from the between-group difference in the rate of insomnia, the two groups were similar in the number and severity of adverse events as spontaneously reported during treatment.