Source: DeRubeis et al., Archives of General Psychiatry, Vol 62, Apr 2005
Tags: cognitive therapy, CT, antidepressant medications, ADM, moderate to severe depression, major depressive disorder, MDD, placebo-controlled trial, randomised controlled trial, RCT, paroxetine, Hamilton Depression Rating Scale, HDRS, TDCRP, cognitive behavioural therapy, CBT, treatment efficacy, psychotherapy vs pharmacotherapy
This study is the first large, placebo-controlled trial to directly compare cognitive therapy (CT) with antidepressant medications (ADM) specifically in patients with moderate to severe major depression. Two hundred and forty outpatients across two university sites were randomised to 16 weeks of medication, 16 weeks of CT, or 8 weeks of pill placebo. It was designed to address a gap left by the influential TDCRP trial, which had concluded that medications should be the default for this severity range.
Cognitive therapy (CT)
A structured, time-limited form of cognitive behavioural therapy (CBT) developed by Aaron Beck and colleagues. Focuses on identifying and modifying distorted cognitions and maladaptive beliefs that maintain depression. In this trial, delivered via individual sessions following standard Beck protocols.
Antidepressant medication (ADM)
Pharmacological treatment for depression. In this trial, the primary medication was paroxetine (an SSRI), with augmentation by lithium carbonate or desipramine hydrochloride for non-responders after 8 weeks.
Paroxetine
A selective serotonin reuptake inhibitor (SSRI) used as the primary ADM in this trial. Dosage ranged from 10 to 50 mg daily. Notable for its anxiolytic (anxiety-reducing) properties, which became relevant in the results.
Hamilton Depression Rating Scale (HDRS)
A clinician-rated instrument used to measure the severity of depressive symptoms. The modified 17-item version was the primary outcome measure. A score of 20 or higher was required for study entry, placing all participants in the moderate to severe range.
Pill placebo
An inert substance given to participants under double-blind conditions. Used here as a control for the medication arm only (not for CT, since you cannot blind a talking therapy). Placebo condition lasted 8 weeks for ethical reasons.
Response (treatment response)
Meeting a predefined threshold of symptom reduction. At 8 weeks: HDRS score of 12 or lower, with last observation carried forward for dropouts. At 16 weeks: a more complex set of criteria designed to prevent a single bad week from disqualifying a patient who was otherwise improving.
Remission
A stricter outcome than response. Requires meeting all response criteria plus a final HDRS score of 7 or lower, consistent with the MacArthur recommendation for defining remission.
TDCRP (Treatment of Depression Collaborative Research Program)
A landmark NIMH-funded study from the late 1980s that compared CT, ADM, pill placebo, and interpersonal psychotherapy. Its secondary analysis found ADM superior to both CT and placebo in more severely depressed patients, which became the basis for the APA guideline recommending medications over CT for moderate to severe depression. The DeRubeis et al. trial was explicitly designed to re-examine that finding.
Hierarchical linear model (HLM)
A statistical technique (also called multilevel or random coefficient models) used to analyse repeated-measures data where observations are nested within subjects. Each patient's trajectory over time is modelled with person-specific intercepts and slopes, then treatment effects are estimated at the group level. This approach uses all available data, making it a full intent-to-treat analysis even when patients drop out.
Effect size
A standardised measure of the magnitude of a difference between groups. In this study, derived from subject-specific slopes of the HLM model. By Cohen's conventions: 0.2 to 0.5 is small, 0.5 to 0.8 is medium, above 0.8 is large.
The study emerged from a specific clinical policy question. At the time, the APA Practice Guidelines recommended medications, not CT, as first-line treatment for moderate to severe MDD. That recommendation rested almost entirely on one study: the TDCRP.
The TDCRP's main analysis found no overall difference between CT and ADM
A secondary analysis of more severely depressed patients (HDRS 20+) found ADM superior to both CT and placebo, while CT was not significantly better than placebo
However, concerns had been raised about the quality of CT delivered in the TDCRP
A mega-analysis of four prior studies that focused on more severely depressed patients did not find ADM superior to CT
No previous trial had combined a large sample of moderate-to-severe patients with a placebo control
The authors set out to fill that gap with a properly powered, placebo-controlled comparison, making special efforts to ensure both treatments reflected "best practices."
Multisite: University of Pennsylvania (Philadelphia) and Vanderbilt University (Nashville)
240 outpatients randomised to three arms:
ADM: n = 120 (16 weeks)
CT: n = 60 (16 weeks)
Pill placebo: n = 60 (8 weeks only, for ethical reasons)
The ADM cell was twice the size of the others because medication responders at week 16 were to be re-randomised into a companion relapse-prevention study
Randomisation was stratified by sex and number of prior depressive episodes
Evaluations were conducted blind to treatment condition
Inclusion required a DSM-IV diagnosis of MDD, age 18 to 70, English-speaking, and HDRS scores of 20 or higher at both screening and baseline visits (separated by at least 7 days).
Of 437 patients evaluated, 96 did not meet diagnostic or severity criteria, and 101 were excluded for reasons including:
History of bipolar I disorder
Substance abuse or dependence requiring treatment
Current or past psychosis
Another Axis I disorder requiring priority treatment
Certain Axis II disorders (antisocial, borderline, schizotypal)
Acute suicide risk requiring hospitalisation
Medical contraindications to study medications
Prior non-response to an adequate trial of paroxetine
Primary drug: paroxetine, starting at 10 to 20 mg/day, raised in 10 to 20 mg increments to a maximum of 50 mg/day by week 6
All five pharmacotherapists were male, board-certified psychiatrists with 9 to 23 years' experience
Sessions were weekly for the first 4 weeks, then every other week
Patients and pharmacotherapists were blind to drug vs placebo for the first 8 weeks
After 8 weeks, the blind was broken; ADM patients continued, placebo patients were offered free treatment
Non-responders at week 8 could receive augmentation with lithium carbonate or desipramine hydrochloride (47 of 101 continuing ADM patients received augmentation)
Sessions followed the TDCRP clinical management manual, with CT-specific techniques explicitly prohibited
Six therapists across the two sites (3 per site), following standard Beck CT manuals for depression and comorbid personality disorders
Sessions: twice weekly for the first 4 weeks, once or twice weekly for weeks 5 to 12, once weekly for weeks 13 to 16
Weekly 90-minute case consultation meetings at each site
A critical detail for the results: all three Pennsylvania therapists and one Vanderbilt therapist had extensive CT experience (7 to 21 years). The other two Vanderbilt therapists each had only 2 years of CT experience at the start of the trial. They received supplementary training through the Beck Institute during the study.
The typical patient was middle-aged (mean 40 years), white (82%), with some college education and modest income. About 59% were female. One third were married or cohabiting.
Key clinical features:
Mean HDRS score at baseline: 23.4 (SD 2.9), with no difference between conditions or sites
90% had chronic or recurrent MDD
Mean age of onset: 22 years
Mean duration of current episode: 46 months (nearly 4 years)
72% had at least one Axis I comorbidity (anxiety disorders were the most common)
48% had at least one Axis II (personality) disorder
12% had a history of psychiatric hospitalisation
60% had a history of prior ADM treatment
The Vanderbilt sample was notably more complex: higher rates of dysthymia (41% vs 9%), Axis I comorbidity (82% vs 63%), comorbid anxiety disorders (65% vs 41%), comorbid PTSD (28% vs 5%), and Cluster C personality disorders (38% vs 19%).
Dropout rates were moderate and broadly similar across conditions:
First 8 weeks: 11% ADM, 15% CT, 13% placebo
Over the full 16 weeks: 16% ADM, 15% CT
No patient in the CT cell dropped out during the second 8 weeks
One patient in the ADM cell died by suicide during the second week of treatment
⚠️ The unequal cell sizes (ADM = 120, CT = 60, placebo = 60) were deliberate, not a flaw. The ADM group was doubled to power a companion relapse-prevention study.
⚠️ The placebo arm ran for only 8 weeks, so the 16-week comparison is ADM vs CT only, with no placebo benchmark. This is an important design limitation to note.
⚠️ The difference in therapist experience between sites (Pennsylvania therapists were far more experienced in CT than Vanderbilt therapists) is not a minor footnote. It becomes central to interpreting the results.
⚠️ This trial used the 17-item HDRS, not the 24-item version. The cutoff of 20 for "moderate to severe" matches the TDCRP definition, making the results directly comparable to that earlier, influential study.
⚠️ The study population was highly chronic (mean episode duration of nearly 4 years) and highly comorbid. These are not mild, first-episode patients.
Q: Why was this study conducted, given that earlier trials had already compared CT and ADM?
A: No prior study had combined a large sample of patients with moderate to severe depression (HDRS 20+) with a placebo control. The influential TDCRP had found medications superior to CT in this severity range, but concerns about the quality of CT in that trial left the question open. This study was designed to provide a more definitive test.
Q: Why was the pill placebo condition limited to 8 weeks?
A: For ethical reasons. Eight weeks was considered sufficient to establish whether medications outperformed placebo (as shown in prior trials), and it would have been unethical to leave severely depressed patients on an inert substance for longer.
Q: What statistical method was used for continuous outcome data, and why?
A: Hierarchical linear models (HLM), because the data had a nested structure (repeated measures within subjects). HLM models each patient's individual trajectory over time and accounts for within-subject correlation. It also allows a full intent-to-treat analysis, using all available data from each patient even if they dropped out.
Q: How did the two study sites differ in their patient populations?
A: Vanderbilt patients were more chronically ill on several dimensions: higher rates of dysthymia, Axis I comorbidity (especially anxiety disorders and PTSD), Axis II comorbidity (especially Cluster C personality disorders), and earlier age of onset. This became relevant in explaining why ADM performed better at Vanderbilt.
Q: What was the key procedural difference between the two sites?
A: Therapist experience in CT. All three Pennsylvania CT therapists had 7 to 21 years of CT experience. Two of Vanderbilt's three CT therapists had only 2 years of experience at the start of the trial. This imbalance contributed to a significant site-by-treatment interaction in the results.
cognitive therapy, cognitive behavioural therapy, CBT, antidepressant medication, pharmacotherapy, paroxetine, SSRI, lithium augmentation, desipramine, moderate to severe depression, major depressive disorder, MDD, Hamilton Depression Rating Scale, HDRS, HDRS-17, placebo-controlled trial, randomised controlled trial, RCT, TDCRP, NIMH, treatment response, remission, hierarchical linear model, HLM, multilevel model, random coefficient model, intent-to-treat, effect size, Cohen's d, Beck, DeRubeis, Hollon, site-by-treatment interaction, therapist experience, therapist competence, comorbidity, Axis I, Axis II, anxiety disorders, treatment guidelines, APA Practice Guidelines