PATIENTS AND METHODS
The protocol was approved by the institutional review board. Subjects (recruited through media, printed announcements, and self or practitioner referrals) underwent triage by telephone. Outpatients (N = 366) with the complaint of depression participated in the Structured Clinical Interview for DSM-III-R
which uses the DSM-III-R23
criteria for MDD and other disorders. To assess MDD with atypical features, the Atypical Depression Diagnostic Scale (Jonathan W. Stewart, MD, written communication, October 20, 1988, and March 20, 1990) was administered for the initial episode. If the diagnosis was absent at the nadir, symptoms were reassessed at “any other time during the episode.” Criteria (according to the Atypical Depression Diagnostic Scale) for depression with atypical features included (1) maintains reactive mood and (2) shows 2 or more of the following: (a
) increased appetite or weight gain; (b
) over-sleeping; (c
) sensation of leaden paralysis or extreme heaviness of arms or legs, while depressed; and (d
) lifetime sensitivity to interpersonal rejection.1,2
Diagnoses were confirmed by a faculty-level diagnostician at a follow-up interview. Entry criteria were (1) DSM-III-R
MDD, (2) definite atypical depression, and (3) score of 14 or more on the 21-item Hamilton Rating Scale for Depression (HRSD-21)24
at the initial or follow-up interview. All patients provided a medical history and a physician reviewed laboratory screening.
Patients were excluded if they (1) had a concurrent medical disorder or treatment that might cause depressive symptoms or required medication incompatible with MAOIs; (2) refused to be randomized or to maintain a tyramine-free diet; (3) had other current primary comorbid psychiatric disorders (eg, organic mental disorders, psychotic disorders, schizophrenia, schizoaffective disorders, alcoholism, or drug abuse or dependency in the last 6 months); (4) scored less than 14 on the HRSD-21 at diagnostic evaluation and follow-up, or before randomization (see description of nonspecific treatment below); (5) could not complete questionnaires; (6) represented an imminent suicide risk; or (7) had previously had an adequate trial of MAOIs or cognitive therapy that failed.
Of the 366 patients studied, 287 (78.4%) were diagnosed as having MDD; of the patients with MDD, 242 (66.1% of those studied) were also diagnosed as having atypical depression. One hundred eighty-one (49.5%) were eligible. Thirty-nine (21.5%) refused consent (generally because of scheduling problems or the desire to receive or avoid a study treatment). The nonspecific treatment baseline (designed to identify and exclude patients who respond to early, nonspecific effects) was initiated when patients signed consent (n = 142). Subjects participated in 2 sessions (during 14 days) of nonspecific treatment by watching a videotape on mood disorders and reporting on symptoms from the HRSD-21. At the end of nonspecific treatment, 13 patients had responded (ie, HRSD-21 <14 or no MDD) and were referred.
Final eligibility for randomization was evaluated by confirming that the inclusion criteria were met. One hundred eight (76.0% of the consenting and eligible) patients were randomized () under the supervision of the statistician (D.M.), who kept research personnel blind to assignment (phenelzine or placebo) during the study.
Demographic and Severity Characteristics of Randomized Depressed Outpatients With Atypical Features
Cognitive therapy was conducted as described by Beck et al29
for 10 weeks, in 20 individual sessions held twice weekly. Three experienced male therapists provided cognitive therapy. Two were doctoral-level clinical psychologists and 1 was a psychiatrist. An offsite consultant (see acknowledgments) used the Cognitive Therapy Scale30,31
to evaluate competence and provide feedback to therapists and investigators.
Therapists participated in weekly group supervision. Of 64 total Cognitive Therapy Scale ratings (ie, 2 planned per patient), 9 (14%) received scores of less than 40. The grand mean Cognitive Therapy Scale score across all therapists was 46.1 ± 4.1. The analyses of variance showed no statistically significant differences in the mean Cognitive Therapy Scale scores among therapists (F2= 2.50; P = .09) or across years (F4= 1.40; P = .24).
Phenelzine and Placebo
A treatment manual32
modeled after the National Institute of Mental Health Treatment of Depression Collaborative Research Program33
guided the psychiatrist’s (M.S.) clinical management of phenelzine or placebo. Each patient was introduced to precautions necessary for using phenelzine safely. The 11 sessions spread over 10 weeks involved adjusting medication and recording symptoms, side effects, weight, and blood pressure.
When symptom reduction and monoamine oxidase inhibition of 80% or more were achieved, the patient continued to receive that dose. Compliance was assessed by pill counts and patient diaries. Dosage levels and schedules were changed to optimize response while keeping side effects within the tolerable range and reducing dropouts.
Phenelzine and placebo were identical in appearance. Patients treated in both conditions followed a lowtyramine MAOI diet. The dose (taken once or twice daily) was increased gradually during the 10 weeks to achieve a therapeutic response to a phenelzine sulfate dose of approximately 0.85 mg/kg (<50 kg, 2 tablets; 50-65 kg, 3 tablets; 66-80 kg, 4 tablets; and ≥81 kg, 5 or 6 tablets) or 1 mg/kg in all patients not responding to a lower dose.34
Patients receiving phenelzine sulfate reported taking the following average amounts per day: week 4, 60.1 ± 2.3 mg; week 7, 64.4 ± 2.9 mg; and week 10, 64.0 ± 2.4 mg.
Platelet monoamine oxidase activity was determined twice before beginning medication and inhibition was determined (as described by Orsulak et al35
). To protect the double blind, blood was drawn from both patients receiving phenelzine and those receiving placebo, and the laboratory provided plausible fictitious results for the placebo group. Actual levels were received for patients taking phenelzine. A target of greater than 80% inhibition was set. Thirty (83%) of 36 patients treated with phenelzine achieved at least 80% of platelet inhibition for 2 consecutive weeks. The average monoamine oxidase percentage inhibition levels for patients treated with phenelzine were as follows: week 4, 87.6% ± 2.9%; week 8, 89.4% ± 1.7%; and week 10,90.5% ± 0.9%. The percentage of patients with monoamine oxidase inhibition of 80% or more was 91% at week 4, 93% at week 8, and 96% at week 10.
The 5 domains assessed were psychiatric diagnoses and symptom severity (reported herein), and cognitive, interpersonal, and personality functioning (to be reported separately). The blind evaluators collected the following symptom severity measures and scored DSM-III-R criteria for MDD at treatment weeks 4, 7, and 10 or at patient exit.
Symptom severity was measured as follows to compare these data with existing studies. The HRSD-2124
was collected at initial and follow-up evaluations during nonspecific treatment, randomization, each blind evaluation, weekly, and at exit. The 21-item Beck Depression Inventory36
was collected at initial evaluation and weekly. The Clinical Global Impression Scale,37
a 7-point Likert scale assessing overall clinical status, was collected at initial evaluation, randomization, each blind evaluation, and weekly.
The hypothesized response rates were 33% for placebo and 64% for both phenelzine and cognitive therapy. The primary dependent variable (identified a priori) was the HRSD-21 score collected at week 10 by an evaluator blind to treatment assignment. Randomization was restricted by stratification. Strata included length of current episode (≤2 years vs >2 years) and marital status at the time of the initial diagnostic evaluation (cohabiting or married vs living without a partner [single, divorced, separated, or widowed]).
Analysis of covariance (ANCOVA) was selected a priori as the primary statistical analysis. A repeated-measures ANCOVA with 3 treatments (cognitive therapy, phenelzine, and placebo) by 3 times (blind evaluation at weeks 4, 7, and 10) was conducted, where the covariates were age at onset and HRSD-21 at randomization. When attrition occurred, the end point was carried forward.
Within secondary analyses, a random coefficients regression (RCR) analysis was used to complement the ANCOVA, because it accommodates attrition of subjects over time by using available data without carrying over end points. The primary parameters estimated by RCR are overall slopes for each treatment supplemented by tests of differences among the slopes. Inclusion of covariates in the RCR models was evaluated by means of backward elimination with a P<.10 criterion for retention.
All analyses were of an intention-to-treat strategy (ie, using the 108 patients randomized). Discrete variables were reported as percentage of frequency (eg, number and percentage), while continuous variables were reported with a mean ± SE. An α level of P≤.05 was used to define significance. Fisher exact tests were 2 tailed, while χ2 tests, analyses of variance, ANCOVAs, and RCR analyses were 1 tailed.