D1: Landmark Papers & Trials in Psychotherapy — Deep Study
Table of Contents
- The Dodo Bird Verdict and Common Factors
- CBT Landmark Trials
- DBT Landmark Trials
- Psychodynamic Landmark Trials
- Schema Therapy Landmark Trials
- Exposure Therapy and EMDR
- Family and Couples Therapy
- ACT and MBCT
- Comparative and Integration Studies
- Process Research
1. THE DODO BIRD VERDICT AND COMMON FACTORS
1.1 Luborsky, Singer & Luborsky (1975) — "Everybody Has Won and All Must Have Prizes"
This paper is the origin of the "Dodo bird verdict" in psychotherapy research. Luborsky and colleagues reviewed comparative outcome studies and concluded that different forms of psychotherapy produce roughly equivalent outcomes. The title references the Dodo bird from Alice in Wonderland, who declares after a race that "everybody has won and all must have prizes."
Design: Narrative review of comparative outcome studies
Key finding: No consistent differences in efficacy between bona fide psychotherapies
Clinical implication: The factors common to all therapies (alliance, empathy, expectation) may matter more than technique-specific ingredients
The Dodo bird verdict does NOT claim that therapy is ineffective or that all therapies are equally good for all conditions. It claims that when two bona fide treatments are compared head-to-head for the same condition, differences in outcome are typically small or absent. This is the most misunderstood finding in psychotherapy research.
1.2 Lambert (1992) — The Pie Chart of Outcome Variance
Michael Lambert proposed a model of what accounts for therapeutic change, based on his review of outcome research. His framework became one of the most cited in psychotherapy training.
The Four Factors:
| Factor | Variance Explained | Description |
|---|---|---|
| Extratherapeutic change | 40% | Client factors, spontaneous remission, life events |
| Therapeutic relationship | 30% | Alliance, empathy, warmth, positive regard |
| Expectancy (placebo) | 15% | Hope, credibility of treatment rationale |
| Technique | 15% | Specific model-driven interventions |
Clinical implication: The relationship accounts for roughly twice the variance of technique. Training that focuses exclusively on technique while neglecting relational skills is misallocated.
Lambert's 40/30/15/15 split is the single most frequently examined breakdown of therapeutic outcome factors. The exact percentages have been debated, but the rank order is consistent across re-analyses: client/extratherapeutic factors > relationship > expectancy >= technique.
When asked "what is the most important factor in psychotherapy outcome," the answer is extratherapeutic/client factors (40%). When asked "what is the most important therapist-delivered factor," the answer is the therapeutic relationship (30%).
1.3 Wampold (2001, 2015) — The Great Psychotherapy Debate
Bruce Wampold's book and subsequent meta-analytic work provided the most rigorous quantitative test of the Dodo bird verdict. He introduced the distinction between the medical model (specific ingredients cure specific disorders) and the contextual model (common factors account for outcomes).
Design: Meta-analyses of comparative outcome studies
Key findings:
- The overall effect size of psychotherapy vs no treatment: d = 0.80 (large)
- The effect size of differences between bona fide therapies: d = 0.00-0.20 (negligible to small)
- Therapist effects (variability between therapists within the same treatment) account for 5-9% of outcome variance — larger than the difference between treatments
- Alliance accounts for approximately r = 0.28 of outcome variance (medium effect)
Wampold's work means that who delivers the therapy matters more than which therapy is delivered. A skilled CBT therapist and a skilled psychodynamic therapist will get comparable outcomes. A poor therapist delivering an "evidence-based" protocol will get poor outcomes. This has direct implications for training, supervision, and quality assurance.
Wampold's effect size for the overall benefit of psychotherapy (d = 0.80) is comparable to the effect of many medical interventions. Psychotherapy is not a "soft" intervention — its effect size exceeds that of many medications for common conditions.
2. CBT LANDMARK TRIALS
2.1 Rush, Beck, Kovacs & Hollon (1977) — The First CBT vs Pharmacotherapy Trial
This is the trial that put CBT on the map. Aaron Beck's cognitive therapy was tested against imipramine (then the gold-standard antidepressant) for outpatient depression.
Design: RCT
Sample: N = 41 outpatients with major depression
Intervention: Cognitive therapy (max 20 sessions over 12 weeks) vs imipramine
Key findings:
- Cognitive therapy showed significantly greater improvement on self-report (BDI) and clinician-rated (HRSD) measures
- 79% of CT group showed marked improvement vs 23% of imipramine group
- CT group had lower dropout rate (5% vs 32%)
Clinical implication: A psychological treatment could match or exceed pharmacotherapy for depression. This was revolutionary in an era when depression was considered a purely biological illness requiring medication.
The Rush et al. (1977) trial has been criticised for its small sample, non-blind design, and possible inadequate pharmacotherapy dosing. But its historical significance is enormous — it launched the entire CBT for depression research programme.
2.2 Elkin et al. (1989) — NIMH Treatment of Depression Collaborative Research Program (TDCRP)
The TDCRP was the first large-scale, multi-site, federally funded trial comparing psychotherapies head-to-head with pharmacotherapy and placebo. It remains one of the most discussed trials in psychotherapy research.
Design: Multi-site RCT, 4 arms
Sample: N = 250 outpatients with major depression
Arms: CBT vs IPT vs imipramine + clinical management vs placebo + clinical management
Duration: 16 weeks
Key findings:
- For the full sample, all active treatments outperformed placebo, but differences between active treatments were not significant
- For more severely depressed patients (HRSD >= 20): imipramine and IPT outperformed placebo; CBT did not significantly differ from placebo
- General pattern: imipramine >= IPT >= CBT >= placebo (but most comparisons non-significant)
- Recovery rates: imipramine 57%, IPT 55%, CBT 51%, placebo 29%
The TDCRP is frequently cited as evidence that CBT is less effective for severe depression. This interpretation is contested: CBT performance varied dramatically across sites, suggesting therapist competence issues rather than a limitation of the model. Site 2 (University of Oklahoma), where CBT therapist competence was rated lowest, dragged down the CBT results.
The TDCRP taught the field three lasting lessons: (1) multi-site trials require rigorous therapist training and monitoring; (2) severity moderates treatment response; and (3) the placebo response rate in depression trials is substantial (~29%).
2.3 DeRubeis et al. (2005) — CBT vs Medications for Moderate-to-Severe Depression
This trial directly addressed the TDCRP criticism that CBT does not work for severe depression.
Design: RCT, 2 sites (University of Pennsylvania, Vanderbilt)
Sample: N = 240 outpatients with moderate-to-severe MDD (HRSD >= 20)
Arms: Cognitive therapy vs paroxetine vs placebo (8 weeks, then active arms continued to 16 weeks)
Key findings:
- At 8 weeks: both CT and paroxetine significantly outperformed placebo
- At 16 weeks: CT and paroxetine showed equivalent response rates (58% vs 58%)
- Site moderated results — Penn (Beck's institution) had better CT outcomes
Clinical implication: CBT is effective for moderate-to-severe depression, directly refuting the TDCRP conclusion.
DeRubeis et al. (2005) is the definitive rebuttal to "CBT doesn't work for severe depression." The response rates were identical to medication at 58% each. Know this trial and the TDCRP as a pair — examiners love this comparison.
2.4 Hollon et al. (2005) — The Enduring Effects of Cognitive Therapy
This is the companion paper to DeRubeis et al. (2005), following patients after treatment discontinuation.
Design: Naturalistic follow-up of the DeRubeis (2005) RCT
Sample: Responders from CT (n = 52) vs responders from paroxetine randomised to continuation medication or placebo withdrawal
Duration: 12-month follow-up after 16-week acute treatment
Key findings:
- Prior CT had significantly lower relapse rate (31%) compared to medication withdrawal (76%, p = 0.004)
- Prior CT was statistically equivalent to continuing medication (31% vs 47%, ns)
- CT's enduring effect was not due to continued therapy — patients were not receiving any treatment during follow-up
Hollon et al. (2005) established that CBT has an enduring effect that persists after treatment ends. Medication works while you take it; CBT continues to protect after you stop. This is the single most important argument for CBT in depression guidelines.
The relapse rate after medication discontinuation (76%) is strikingly high. This is why most guidelines now recommend either long-term maintenance medication OR a course of CBT for recurrent depression — not just a single course of antidepressants.
2.5 Clark et al. (2003, 2006) — CBT for Social Anxiety and PTSD
David M. Clark's Oxford group produced two landmark trials that established disorder-specific cognitive therapy as a front-runner for anxiety disorders.
Clark et al. (2003) — Social Anxiety Disorder
Design: RCT
Sample: N = 60 with social phobia
Arms: Individual CT (Clark & Wells model) vs fluoxetine + self-exposure vs placebo
Key findings: CT was significantly superior to both fluoxetine and placebo at post-treatment and follow-up. Effect sizes were very large (d > 1.0 vs placebo).
Ehlers, Clark et al. (2003, 2005) — PTSD
Design: RCT
Sample: N = 85 with chronic PTSD (Ehlers et al., 2005 trial)
Arms: Cognitive therapy (Ehlers & Clark model) vs self-help booklet vs waitlist
Key findings: CT produced large effects (d = 1.27 vs waitlist). 71% achieved reliable and clinically significant change.
Clark's cognitive therapy models for social anxiety and PTSD are now considered among the most effective psychological treatments for any condition. The effect sizes (d > 1.0) exceed typical psychotherapy benchmarks.
3. DBT LANDMARK TRIALS
3.1 Linehan et al. (1991) — The First RCT of DBT for BPD
Marsha Linehan's first RCT was a watershed moment for BPD treatment. Before this trial, the prevailing view was that BPD was untreatable and that patients should be avoided.
Design: RCT
Sample: N = 44 women with BPD and recent parasuicide
Arms: DBT (1 year) vs treatment as usual (TAU)
Key findings:
- DBT significantly reduced frequency of parasuicidal acts (including self-harm and suicide attempts)
- DBT significantly reduced number of inpatient psychiatric days
- DBT had significantly better treatment retention (83% vs 42%)
- No significant difference in depression, hopelessness, or suicidal ideation scores
Linehan (1991) showed DBT reduces self-harm and hospitalisation but NOT depression or suicidal ideation. This distinction matters: DBT targets behavioural dysregulation, not mood per se. When asked about DBT efficacy, specify the outcomes — it reduces the behaviour, not necessarily the distress.
3.2 Linehan et al. (2006) — DBT vs Community Expert Treatment
The major criticism of early DBT trials was that the comparison condition (TAU) was a weak control. This trial addressed that directly.
Design: RCT
Sample: N = 101 women with BPD, recent suicidal or self-injurious behaviour
Arms: DBT (1 year) vs community treatment by experts (CTBE) — therapists nominated as expert BPD treaters
Key findings:
- DBT was significantly superior to CTBE on: suicide attempts (hazard ratio 2.66, fewer in DBT), medical severity of self-harm, emergency department visits, and psychiatric hospitalisations
- DBT had half the rate of suicide attempts compared to CTBE (23% vs 46%)
- Treatment retention was equivalent
Linehan (2006) is the strongest evidence for DBT superiority over genuine expert treatment (not just TAU). The comparison was deliberately rigorous — community experts who treated BPD regularly, not randomly assigned novices.
A key finding was that even expert clinicians without a structured BPD protocol achieved significantly worse outcomes than DBT. Structure, protocol adherence, and skills training matter — clinical experience alone is insufficient for this population.
3.3 Rathus & Miller (2002) — Adolescent DBT
Design: Quasi-experimental (non-randomised comparison)
Sample: N = 111 suicidal adolescents with BPD features
Arms: DBT-A (12 weeks, adapted for adolescents) vs TAU (12-week supportive-psychodynamic therapy)
Key findings: DBT-A had significantly fewer psychiatric hospitalisations and higher treatment completion. No suicide attempts in the DBT-A group vs 3 in TAU.
Rathus & Miller (2002) is the foundational study for adolescent DBT (DBT-A). Although not a true RCT, it demonstrated feasibility and safety, leading to the later McCauley et al. (2018) definitive trial showing DBT-A reduces self-harm in adolescents.
4. PSYCHODYNAMIC LANDMARK TRIALS
4.1 Shedler (2010) — The Efficacy of Psychodynamic Psychotherapy
Jonathan Shedler's American Psychologist paper was a landmark review that challenged the narrative that psychodynamic therapy lacked an evidence base.
Design: Meta-analytic review
Key findings:
- Effect sizes for psychodynamic therapy (d = 0.97 for targeted outcomes, d = 0.73 for general symptom improvement) are as large as those reported for other evidence-based therapies
- Uniquely, patients who receive psychodynamic therapy continue to improve after treatment ends (the "sleeper effect")
- Non-psychodynamic therapies may work, in part, because they contain psychodynamic processes (e.g., working alliance, exploration of affect, identification of recurring patterns)
Shedler (2010) showed that the effect size for psychodynamic therapy (d = 0.97) matches CBT benchmarks. The "sleeper effect" — continuing improvement after termination — was unique to psychodynamic therapy and was not seen with CBT or medication.
If asked "is psychodynamic therapy evidence-based?", cite Shedler (2010) and Leichsenring & Rabung (2008, 2011). The evidence base is smaller than CBT's, but the effect sizes are comparable.
4.2 Leichsenring & Rabung (2008, 2011) — Long-Term Psychodynamic Psychotherapy Meta-Analyses
Design: Meta-analyses of RCTs and controlled trials of long-term psychodynamic psychotherapy (LTPP, defined as > 1 year or > 50 sessions)
Samples: 2008 meta-analysis: 23 studies (N = 1,053). 2011 update: 10 RCTs
Key findings:
- LTPP showed large effect sizes for overall effectiveness (d = 0.96, 2008), complex mental disorders (d = 1.03, personality disorders: d = 0.97)
- LTPP was significantly more effective than shorter forms of therapy for complex/comorbid presentations
- The 2011 update confirmed findings with stricter RCT-only inclusion: LTPP superior to less intensive interventions (d = 0.44)
"Complex mental disorders" in these meta-analyses means personality disorders, chronic depression, multiple comorbidities, and treatment-resistant presentations. The finding that LTPP is superior for complexity, not for simple single-diagnosis cases, is clinically important — it guides appropriate treatment matching.
4.3 Bateman & Fonagy (1999, 2001, 2008) — Mentalisation-Based Treatment for BPD
Anthony Bateman and Peter Fonagy developed MBT specifically for BPD and tested it in a landmark partial hospitalisation trial.
Bateman & Fonagy (1999) — Original Trial
Design: RCT
Sample: N = 38 with BPD, partial hospitalisation setting
Arms: MBT-based partial hospitalisation (18 months) vs standard psychiatric care
Key findings: MBT significantly reduced self-harm, suicide attempts, inpatient days, depression, anxiety, and interpersonal distress. Effects appeared after 6 months and continued to increase.
Bateman & Fonagy (2001, 2008) — Follow-Up Studies
18-month follow-up (2001): MBT gains were maintained and continued to improve, while the control group showed no change.
8-year follow-up (2008): MBT group maintained significantly better functioning. Suicidality and service use remained lower. 13% of MBT patients still met BPD criteria vs 87% of control.
Bateman & Fonagy's 8-year follow-up (2008) is the longest follow-up in any BPD treatment trial. Only 13% of MBT patients still met BPD criteria at 8 years vs 87% of controls. This is the strongest longitudinal evidence for any BPD treatment.
4.4 Clarkin et al. (2007) — TFP vs DBT vs Supportive Therapy for BPD
Design: RCT
Sample: N = 90 women with BPD
Arms: Transference-Focused Psychotherapy (TFP, 1 year) vs DBT (1 year) vs psychodynamic supportive therapy (1 year)
Key findings:
- All three treatments produced significant improvement across multiple domains
- TFP and DBT significantly reduced suicidality; supportive therapy did not
- Only TFP significantly improved reflective functioning (mentalisation) and attachment organisation
- DBT uniquely improved impulsivity-related behaviours
Clarkin (2007) is the only trial comparing TFP, DBT, and supportive therapy head-to-head. TFP was the only treatment that improved reflective functioning and attachment. DBT was the only one that improved impulsivity. Each modality has a distinct mechanism signature.
5. SCHEMA THERAPY LANDMARK TRIALS
5.1 Giesen-Bloo et al. (2006) — Schema Therapy vs TFP for BPD
This is the largest and most methodologically rigorous trial of any psychotherapy for BPD. It was conducted across four sites in the Netherlands and directly compared two structured, long-term treatments.
Design: Multi-centre RCT
Sample: N = 86 with BPD (44 ST, 42 TFP), 3-year treatment
Arms: Schema therapy (2 sessions/week, 3 years) vs Transference-Focused Psychotherapy (2 sessions/week, 3 years)
Key findings:
- ST showed significantly greater improvement on all primary outcomes (BPD severity, quality of life, psychopathology)
- Recovery rate: ST 45.5% vs TFP 24.4%
- Clinical improvement: ST 65.9% vs TFP 42.9%
- ST had significantly lower dropout: 26.7% vs 50.6% for TFP
- Effect sizes favoured ST from year 1 onwards, with advantages increasing over time
Giesen-Bloo et al. (2006) produced the highest recovery rate ever reported for BPD in an RCT (45.5%). Know this number. The dropout rate (26.7%) was also notably low for a BPD trial, reflecting ST's emphasis on limited reparenting and alliance.
The TFP dropout rate (50.6%) was comparable to dropout from unstructured treatment. The ST dropout rate (26.7%) was half that of TFP. For a population notorious for treatment attrition, this is clinically significant. Limited reparenting and the warm therapeutic stance in ST appear to be protective against dropout.
If asked to compare ST, TFP, DBT, and MBT for BPD, know the trial evidence for each: Giesen-Bloo (2006) for ST, Clarkin (2007) for TFP, Linehan (2006) for DBT, Bateman & Fonagy (1999/2008) for MBT. All show efficacy; ST has the highest recovery rate, MBT has the longest follow-up, DBT has the strongest evidence for reducing self-harm behaviour.
5.2 Nadort et al. (2009) — Schema Therapy With and Without Phone Crisis Support
Design: Multi-centre RCT
Sample: N = 62 with BPD
Arms: ST with therapist phone availability between sessions vs ST without phone availability
Key findings:
- Both groups showed large improvements in BPD symptoms, with recovery rates comparable to Giesen-Bloo et al.
- No significant differences between groups — phone crisis support did not add benefit beyond ST alone
- This demonstrated ST's effectiveness without requiring 24/7 therapist availability (a criticism often levelled at DBT)
Nadort et al. (2009) showed that ST's efficacy does not depend on between-session phone contact, making it more practical and less burdensome than DBT for therapists. ST works through the therapeutic relationship in-session, not through crisis coaching.
5.3 Bamelis et al. (2014) — Schema Therapy for Cluster C Personality Disorders
Design: Multi-centre RCT
Sample: N = 323 with Cluster C PDs (avoidant, dependent, obsessive-compulsive), PD NOS, or paranoid PD
Arms: ST (50 sessions) vs Clarification-Oriented Psychotherapy (COP) vs TAU
Key findings:
- ST showed significantly higher recovery rates than COP and TAU (81% vs 60% vs 52% at the 1-year follow-up)
- ST was more cost-effective per quality-adjusted life year (QALY) gained
- ST was effective across all Cluster C subtypes
Bamelis (2014) is the largest trial of any psychotherapy for personality disorders (N = 323). It established ST's efficacy beyond BPD — specifically for Cluster C PDs. The 81% recovery rate for ST is remarkable for personality disorders.
6. EXPOSURE THERAPY AND EMDR
6.1 Foa, Dancu, Hembree et al. (1999, 2005) — Prolonged Exposure for PTSD
Edna Foa's work on Prolonged Exposure (PE) is the foundation of trauma-focused treatment.
Foa et al. (1999)
Design: RCT
Sample: N = 96 female sexual assault survivors with chronic PTSD
Arms: PE vs stress inoculation training (SIT) vs combined PE + SIT vs waitlist
Key findings:
- All active treatments significantly outperformed waitlist
- PE alone showed the best overall outcome at follow-up
- The combined treatment was NOT superior to PE alone — a counterintuitive finding
Foa et al. (2005)
Design: RCT
Sample: N = 179 women with PTSD
Arms: PE vs PE + cognitive restructuring vs waitlist
Key findings: PE alone and PE + CR were equally effective. Adding cognitive restructuring did not improve outcomes beyond exposure alone.
Foa's work consistently shows that exposure is the critical ingredient in PTSD treatment. Adding cognitive restructuring, stress inoculation, or other components does not significantly improve outcomes beyond PE alone. Exposure is necessary and may be sufficient.
6.2 Resick et al. (2002) — Cognitive Processing Therapy vs Prolonged Exposure
Design: RCT
Sample: N = 171 female sexual assault survivors with PTSD
Arms: CPT (12 sessions) vs PE (9 sessions) vs minimal attention control
Key findings:
- Both CPT and PE were significantly superior to control
- CPT and PE showed equivalent outcomes on PTSD severity
- Both maintained gains at 9-month follow-up
- CPT showed a slight (non-significant) advantage on guilt cognitions
CPT and PE are the two front-line trauma treatments recommended by the VA/DoD guidelines. For clinical matching: PE is preferred when avoidance of trauma memories is the primary maintaining factor. CPT is preferred when "stuck points" (maladaptive beliefs about the trauma) are prominent.
6.3 Shapiro (1989) — The Original EMDR Study
Design: RCT
Sample: N = 22 with traumatic memories (rape, molestation, Vietnam veterans)
Arms: EMDR (single session) vs modified flooding with saccades
Key findings: EMDR produced significant reductions in subjective distress (SUD scores) and increases in positive cognition validity (VOC scores) after a single session.
Shapiro (1989) has been heavily criticised for: small sample, therapist allegiance (Shapiro administered all treatments), non-blind assessment, and inadequate comparison condition. However, it generated a research programme that has since accumulated strong evidence. Know the criticisms but also know the subsequent positive trials.
6.4 Bisson et al. (2007) — Cochrane Review of Psychological Treatments for PTSD
Design: Cochrane systematic review and meta-analysis
Sample: 38 RCTs
Key findings:
- Trauma-focused CBT (including PE and CPT) and EMDR were the only treatments with sufficient evidence for recommendation
- Both TF-CBT and EMDR significantly outperformed waitlist and non-trauma-focused therapies
- No significant difference between TF-CBT and EMDR
- Non-trauma-focused therapies (supportive counselling, relaxation) showed limited evidence
- Group CBT was less effective than individual CBT
The Cochrane review (Bisson, 2007) gave NICE and WHO their evidence base for recommending TF-CBT and EMDR as first-line PTSD treatments. The specific finding that trauma-focused approaches outperform non-trauma-focused ones is critical — supportive counselling is not adequate for PTSD.
7. FAMILY AND COUPLES THERAPY
7.1 Gottman (1999) — The Mathematics of Divorce Prediction
John Gottman's research programme, culminating in his book "The Seven Principles for Making Marriage Work," used observational coding of marital interactions to predict divorce with remarkable accuracy.
Design: Prospective longitudinal studies with observational coding (Specific Affect Coding System, SPAFF)
Samples: Multiple cohorts totalling over 3,000 couples
Key findings:
- The "Four Horsemen" — criticism, contempt, defensiveness, and stonewalling — predicted divorce with 93.6% accuracy in a 6-year follow-up
- The ratio of positive to negative interactions during conflict was the strongest predictor. Stable couples showed a 5:1 ratio; divorce-bound couples showed 0.8:1
- Contempt was the single most toxic behaviour, predicting not only divorce but also physical illness in the recipient
- Physiological arousal (heart rate, skin conductance) during conflict predicted relationship deterioration
Gottman's Four Horsemen of the Apocalypse: Criticism, Contempt, Defensiveness, Stonewalling. Contempt is the most damaging. The 5:1 positive-to-negative ratio during conflict is the stability threshold. These are among the most asked questions in couples therapy examinations.
7.2 Johnson, Hunsley, Greenberg & Schindler (1999) — EFT for Couples
Design: Meta-analysis
Sample: 7 studies of Emotionally Focused Couple Therapy (EFT-C)
Key findings:
- Large effect size: d = 1.30 (post-treatment change)
- 70-73% of couples moved to recovery; 90% showed significant improvement
- Gains were maintained at follow-up (2 years in some studies)
- Moderators: greater initial distress predicted larger gains; EFT was effective across different presenting problems
EFT-C's effect size of d = 1.30 is the largest reported for any couples therapy approach. Know that EFT-C (Johnson) works through accessing attachment emotions and restructuring interactional cycles — it is NOT the same as Greenberg's EFT for individuals, though they share theoretical roots.
7.3 Leff et al. (2000) — Family Intervention for Schizophrenia
Design: Meta-analysis and critical review of family intervention trials
Key findings:
- Family interventions (psychoeducation, communication training, problem-solving) reduced relapse rates in schizophrenia by approximately 50% compared to standard care
- 9-month relapse: ~25% with family intervention vs ~50% without
- Family interventions reduced high expressed emotion (EE), the strongest psychosocial predictor of relapse
- Most effective when: lasting > 6 months, including the patient, and providing concrete coping skills
High expressed emotion (criticism, hostility, emotional over-involvement) in family members is one of the most replicated predictors of relapse in schizophrenia. Family interventions targeting EE are now recommended in every major schizophrenia guideline (NICE, APA, PORT). The evidence is strong enough that NOT offering family intervention represents a gap in care.
8. ACT AND MBCT
8.1 Teasdale et al. (2000) — MBCT Prevents Depression Relapse
Design: Multi-site RCT
Sample: N = 145 with recurrent MDD (>= 3 prior episodes), currently in remission
Arms: MBCT (8-week group programme) + TAU vs TAU alone
Key findings:
- For patients with >= 3 prior episodes (77% of sample): MBCT significantly reduced relapse — 37% relapsed vs 66% in TAU (p = 0.004)
- For patients with only 2 prior episodes: MBCT did NOT reduce relapse
- MBCT's mechanism: decentering from depressive cognitions, breaking the reactivation of ruminative processing patterns
Teasdale (2000) showed MBCT prevents relapse in patients with >= 3 prior episodes of depression, NOT in those with only 2 episodes. This specificity is crucial: MBCT works by disrupting cognitive reactivity, which is more established in highly recurrent depression. The 3+ episodes threshold is the standard clinical indication.
8.2 Kuyken et al. (2015) — MBCT vs Maintenance Antidepressants
Design: Multi-centre RCT (the PREVENT trial)
Sample: N = 424 with recurrent MDD (>= 3 episodes), on maintenance antidepressants
Arms: MBCT with support to taper antidepressants vs maintenance antidepressants
Duration: 24-month follow-up
Key findings:
- MBCT with antidepressant tapering was non-inferior to maintenance antidepressants for preventing depressive relapse over 24 months
- Relapse: MBCT 44% vs maintenance medication 47% (hazard ratio 0.89, ns)
- MBCT was more effective than medication for patients with higher severity of childhood abuse
Kuyken (2015) is the definitive MBCT trial. It established MBCT as a viable alternative to long-term antidepressant maintenance. The finding that childhood abuse survivors benefited more from MBCT than medication suggests MBCT may particularly help patients whose depression is maintained by ruminative processing of early adversity.
Know the triad of MBCT evidence: Teasdale (2000) = prevention, Ma & Teasdale (2004) = replication, Kuyken (2015) = non-inferiority to medication.
8.3 A-Tjak et al. (2015) — ACT Meta-Analysis
Design: Meta-analysis of RCTs
Sample: 39 RCTs across multiple conditions
Key findings:
- ACT outperformed control conditions (waitlist, placebo, TAU) with a small-to-medium effect: g = 0.57 (95% CI 0.44-0.70)
- ACT was NOT significantly different from established treatments (CBT): g = 0.06 (ns)
- ACT was effective for chronic pain (g = 0.62), anxiety disorders (g = 0.51), depression (g = 0.56), and substance use (g = 0.45)
- Process measures: psychological flexibility mediated outcomes across studies
A-Tjak (2015) is the benchmark ACT meta-analysis. ACT works — it outperforms control conditions with medium effect sizes. But it does NOT outperform traditional CBT. The claim that ACT is "better" than CBT is not supported; the claim that ACT is an effective, evidence-based alternative is.
9. COMPARATIVE AND INTEGRATION STUDIES
9.1 Cuijpers et al. (2008, 2019) — Meta-Analyses of Psychotherapy for Depression
Pim Cuijpers and his group at VU Amsterdam have produced the most comprehensive meta-analytic programme in psychotherapy research.
Cuijpers et al. (2008)
Design: Meta-analysis
Sample: 53 RCTs of psychotherapy for adult depression
Key findings:
- Overall effect size: d = 0.67 (moderate to large)
- CBT, behavioural activation, IPT, psychodynamic therapy, and problem-solving therapy all showed significant effects
- No significant differences between bona fide therapies (supporting the Dodo bird verdict)
- When corrected for publication bias, the effect size dropped to d = 0.42
Cuijpers (2008) estimated that publication bias inflates psychotherapy effect sizes by approximately 25%. The corrected effect size (d = 0.42) is still clinically meaningful but more modest than the uncorrected estimate. Always consider publication bias when interpreting psychotherapy meta-analyses.
Cuijpers et al. (2019) — Updated Network Meta-Analysis
Design: Network meta-analysis
Sample: 331 RCTs (N > 34,000)
Key findings:
- All seven types of psychotherapy were more effective than controls for adult depression
- Behavioural activation, CBT, IPT, problem-solving therapy, psychodynamic therapy, life review, and third-wave approaches all showed significant effects
- The effects did not differ significantly between types (again supporting equivalence)
- Combined psychotherapy + pharmacotherapy was more effective than either alone
The consistent finding across Cuijpers' meta-analyses is that combination treatment (therapy + medication) outperforms either alone. For moderate-to-severe depression, the evidence strongly supports combination. This is reflected in NICE, APA, and CANMAT guidelines.
9.2 Steinert et al. (2017) — Psychodynamic Therapy vs CBT
Design: Systematic review and meta-analysis
Sample: 23 RCTs directly comparing psychodynamic therapy with CBT
Key findings:
- No significant difference in primary outcomes between psychodynamic therapy and CBT: g = -0.07 (95% CI -0.21 to 0.08)
- Results were consistent across depression, anxiety, and personality disorders
- Neither approach was consistently superior to the other
Steinert (2017) is the most cited direct comparison meta-analysis of psychodynamic vs CBT. The effect difference was essentially zero (g = -0.07). Use this to counter the narrative that CBT is "more evidence-based" — it has more trials, but when directly compared, outcomes are equivalent.
9.3 Barlow's Unified Protocol (UP)
The Unified Protocol for Transdiagnostic Treatment of Emotional Disorders (Barlow et al., 2011) represents the integration movement in CBT — treating the shared mechanisms underlying all emotional disorders rather than using disorder-specific protocols.
Core modules: Mindful emotion awareness, cognitive flexibility, countering emotional behaviours (exposure), interoceptive awareness
Key trials:
- Farchione et al. (2012): N = 37, open trial, significant improvement across anxiety disorders (d = 0.52-1.14)
- Barlow et al. (2017): N = 223 RCT, UP vs single-disorder protocols vs waitlist. UP was non-inferior to single-disorder protocols and superior to waitlist across all anxiety disorders
- Cassiello-Robbins et al. (2020): UP showed equivalent outcomes to disorder-specific CBT with similar treatment gains
Barlow's Unified Protocol is the most researched transdiagnostic CBT. It treats neuroticism (the shared vulnerability across emotional disorders) rather than disorder-specific symptoms. The non-inferiority finding vs disorder-specific protocols is significant — one protocol can treat comorbid anxiety and mood disorders simultaneously.
10. PROCESS RESEARCH
10.1 Bordin (1979) — The Working Alliance
Edward Bordin proposed a pantheoretical model of the therapeutic alliance with three components:
- Tasks — agreement on what happens in therapy
- Goals — agreement on desired outcomes
- Bond — mutual trust, liking, and attachment
Impact: The Working Alliance Inventory (WAI), based on Bordin's model, became the most widely used alliance measure. Subsequent meta-analyses (Horvath & Symonds, 1991; Horvath et al., 2011) confirmed a consistent relationship between alliance and outcome: r = 0.275 (medium effect), stable across therapies, raters, and time of measurement.
Bordin's three alliance components: Tasks, Goals, Bond. The alliance-outcome correlation (r = 0.275) is one of the most replicated findings in psychotherapy research. It accounts for roughly 7.5% of outcome variance — modest but consistent and significant.
The alliance is not just "rapport." It includes agreement on what you are doing (tasks) and why (goals). A therapist who has excellent rapport but uses interventions the patient does not understand or agree with has a poor task alliance — and this predicts poorer outcomes.
10.2 Safran & Muran (2000) — Alliance Rupture and Repair
Jeremy Safran and J. Christopher Muran's programme of research on rupture-repair sequences revolutionised how clinicians understand the therapeutic relationship.
Key concepts:
- Rupture: A deterioration or strain in the alliance — can be a confrontation (patient expresses anger, dissatisfaction) or a withdrawal (patient disengages, becomes compliant, changes topic)
- Repair: The process of recognising, addressing, and resolving the rupture within the therapy
- Resolution: Successful repair predicts better outcome than never having a rupture at all
Research findings:
- Rupture-repair sequences predict positive outcome more strongly than consistently high alliance ratings
- The ability to repair ruptures distinguishes effective from ineffective therapists
- Withdrawal ruptures are harder to detect than confrontation ruptures and are more common
Safran & Muran's key finding: repaired ruptures predict BETTER outcomes than no rupture at all. A therapy that goes smoothly may mean the patient is being compliant rather than engaged. The rupture-repair cycle is itself therapeutic — it provides a corrective emotional experience.
There are two types of rupture: confrontation (the patient pushes back) and withdrawal (the patient shuts down). Withdrawal ruptures are more dangerous because they are harder to detect. If a patient suddenly becomes agreeable and compliant after previously being engaged, suspect a withdrawal rupture.
10.3 Kazdin (2007) — Mediators and Mechanisms of Change
Alan Kazdin's landmark paper challenged the field to move beyond "does therapy work?" to "how and why does therapy work?"
Key arguments:
- Establishing that a treatment works (efficacy) is not the same as understanding why it works (mechanism)
- A mediator is a statistical concept (a variable that accounts for the treatment-outcome relationship). A mechanism is a causal process (the actual reason change occurs)
- Most claimed mechanisms in psychotherapy (cognitive change in CBT, insight in psychodynamic therapy, mentalisation in MBT) have not met the criteria for true mechanism demonstration
- Requirements for mechanism evidence: strong association with outcome, specificity to that treatment, temporal precedence (the mechanism changes before the outcome changes), gradient (more mechanism change = more outcome change), and a plausible theoretical account
Kazdin (2007) distinguished mediators from mechanisms. A mediator is a statistical finding (X accounts for the Y-Z relationship). A mechanism requires temporal precedence, specificity, gradient, and plausibility. Very few psychotherapy "mechanisms" have been truly demonstrated — most remain at the mediator level.
This paper has practical implications. If we do not know how therapy works, we cannot efficiently train therapists, improve protocols, or match patients to treatments. Kazdin's call to study mechanisms has shaped the next generation of psychotherapy research, including the dismantling studies and process-outcome studies that now dominate the field.
MASTER TABLE: LANDMARK TRIALS IN PSYCHOTHERAPY
| Trial | Year | Design | N | Condition | Comparison | Key Finding | Effect Size |
|---|---|---|---|---|---|---|---|
| Luborsky et al. | 1975 | Review | Multiple | Various | Various psychotherapies | "Dodo bird verdict" — equivalent outcomes | — |
| Rush et al. | 1977 | RCT | 41 | Depression | CT vs imipramine | CT superior on all measures | — |
| Bordin | 1979 | Theory | — | — | — | Working alliance: tasks, goals, bond | r = 0.275 |
| Shapiro | 1989 | RCT | 22 | Trauma | EMDR vs flooding variant | EMDR reduced distress rapidly | — |
| Elkin et al. (TDCRP) | 1989 | Multi-site RCT | 250 | Depression | CBT vs IPT vs imipramine vs placebo | All active > placebo; imipramine best for severe | d = 0.37 (vs placebo) |
| Linehan et al. | 1991 | RCT | 44 | BPD | DBT vs TAU | DBT reduced parasuicide + hospitalisations | — |
| Lambert | 1992 | Review | — | Various | — | 40/30/15/15 outcome variance model | — |
| Bateman & Fonagy | 1999 | RCT | 38 | BPD | MBT partial hospital vs standard care | MBT reduced self-harm, depression, hospitalisations | — |
| Foa et al. | 1999 | RCT | 96 | PTSD | PE vs SIT vs combined vs waitlist | PE alone best at follow-up | — |
| Gottman | 1999 | Longitudinal | >3000 | Couples | Observational coding | Four Horsemen predict divorce (93.6%) | — |
| Johnson et al. | 1999 | Meta-analysis | 7 studies | Couples | EFT-C vs controls | EFT-C d = 1.30; 70-73% recovery | d = 1.30 |
| Leff et al. | 2000 | Meta-analysis | Multiple | Schizophrenia | Family intervention vs standard care | 50% relapse reduction | — |
| Safran & Muran | 2000 | Process research | Multiple | Various | — | Rupture-repair predicts better outcomes | — |
| Teasdale et al. | 2000 | RCT | 145 | Recurrent MDD | MBCT + TAU vs TAU | MBCT halved relapse for 3+ episodes | — |
| Wampold | 2001 | Meta-analysis | Multiple | Various | Between-therapy comparisons | Therapy differences d = 0.00-0.20 | d = 0.80 (vs no tx) |
| Rathus & Miller | 2002 | Quasi-exp. | 111 | Adolescent BPD | DBT-A vs TAU | DBT-A reduced hospitalisations, 0 suicide attempts | — |
| Resick et al. | 2002 | RCT | 171 | PTSD | CPT vs PE vs control | CPT = PE; both superior to control | — |
| Clark et al. | 2003 | RCT | 60 | Social anxiety | CT vs fluoxetine vs placebo | CT superior to fluoxetine and placebo | d > 1.0 |
| DeRubeis et al. | 2005 | RCT | 240 | Moderate-severe MDD | CT vs paroxetine vs placebo | CT = paroxetine (58% response each) | — |
| Foa et al. | 2005 | RCT | 179 | PTSD | PE vs PE + CR vs waitlist | PE alone = PE + CR | — |
| Hollon et al. | 2005 | Follow-up | 104 | Depression | Prior CT vs medication continuation vs withdrawal | Prior CT relapse 31% vs withdrawal 76% | — |
| Giesen-Bloo et al. | 2006 | Multi-site RCT | 86 | BPD | ST vs TFP (3 years) | ST recovery 45.5% vs TFP 24.4% | — |
| Linehan et al. | 2006 | RCT | 101 | BPD | DBT vs community experts | DBT halved suicide attempts (23% vs 46%) | — |
| Bisson et al. | 2007 | Cochrane review | 38 RCTs | PTSD | Multiple comparisons | TF-CBT and EMDR recommended; equivalent to each other | — |
| Clarkin et al. | 2007 | RCT | 90 | BPD | TFP vs DBT vs supportive | All improved; TFP improved mentalisation; DBT improved impulsivity | — |
| Kazdin | 2007 | Conceptual | — | — | — | Mediators are not mechanisms; criteria for mechanism evidence | — |
| Cuijpers et al. | 2008 | Meta-analysis | 53 RCTs | Depression | Multiple psychotherapies | d = 0.67 (0.42 corrected for pub bias); no between-therapy differences | d = 0.67 |
| Leichsenring & Rabung | 2008 | Meta-analysis | 23 studies | Complex disorders | LTPP vs shorter therapies | LTPP d = 0.96 for complex presentations | d = 0.96 |
| Bateman & Fonagy | 2008 | 8-year follow-up | 38 | BPD | MBT vs standard care | 13% still met BPD criteria vs 87% control | — |
| Nadort et al. | 2009 | RCT | 62 | BPD | ST + phone vs ST alone | No added benefit of phone support | — |
| Shedler | 2010 | Meta-review | Multiple | Various | Psychodynamic vs benchmarks | d = 0.97; "sleeper effect" unique to psychodynamic | d = 0.97 |
| Bamelis et al. | 2014 | Multi-site RCT | 323 | Cluster C PD | ST vs COP vs TAU | ST 81% recovery; largest PD trial | — |
| A-Tjak et al. | 2015 | Meta-analysis | 39 RCTs | Various | ACT vs controls and CBT | ACT g = 0.57 vs controls; g = 0.06 vs CBT (ns) | g = 0.57 |
| Kuyken et al. | 2015 | RCT | 424 | Recurrent MDD | MBCT + taper vs maintenance meds | MBCT non-inferior to medication over 24 months | — |
| Steinert et al. | 2017 | Meta-analysis | 23 RCTs | Various | Psychodynamic vs CBT | No significant difference: g = -0.07 | g = -0.07 |
| Cuijpers et al. | 2019 | Network meta-analysis | 331 RCTs | Depression | 7 psychotherapy types | All effective; no between-therapy differences; combo > monotherapy | — |
KEY NUMBERS TO MEMORISE
| Number | What It Represents | Source |
|---|---|---|
| 40/30/15/15 | Outcome variance: client/relationship/expectancy/technique | Lambert (1992) |
| d = 0.80 | Overall effect of psychotherapy vs no treatment | Wampold (2001) |
| r = 0.275 | Alliance-outcome correlation | Horvath et al. (2011) |
| 5:1 | Positive-to-negative interaction ratio for stable couples | Gottman (1999) |
| 45.5% | ST recovery rate for BPD | Giesen-Bloo (2006) |
| 31% vs 76% | Relapse after prior CT vs medication withdrawal | Hollon (2005) |
| d = 1.30 | EFT-C effect size | Johnson et al. (1999) |
| 3+ episodes | MBCT relapse prevention threshold | Teasdale (2000) |
| 93.6% | Gottman's divorce prediction accuracy | Gottman (1999) |
| 13% vs 87% | BPD criteria at 8-year follow-up (MBT vs control) | Bateman & Fonagy (2008) |
Weave Psychotherapy Vol. 12 — Landmark Papers & Trials
Dr. Wilfred Dsouza