BGPT: Paper Review: Efficacy of Mindfulness-Based Cognitive Therapy in Prevention of Depressive Relapse

Explore by Goal

Quick Explanation Copied

Core finding: In an individual patient data (IPD) meta-analysis of 9 randomized trials (N=1329), mindfulness-based cognitive therapy (MBCT) was associated with a lower hazard of depressive relapse within 60 weeks vs non-MBCT controls (random-effects HR ≈ 0.69, 95% CI 0.58–0.82).

Skeptical caveat: Evidence of time-varying effects and potential publication bias signals (funnel-plot asymmetry; Egger test limited by only 9 trials), plus missing moderator data (e.g., race/ethnicity and employment), mean effect magnitude for specific subgroups remains uncertain.

Long Explanation

BGPT Paper Review (IPD meta-analysis)

Efficacy of Mindfulness-Based Cognitive Therapy in Prevention of Depressive Relapse

Individual patient data (IPD) meta-analysis of randomized trials; primary outcome: depressive relapse within 60 weeks.

DOI: 10.1001/jamapsychiatry.2016.0076 Design: IPD meta-analysis N=1329; 9 trials

What the paper set out to answer

Whether MBCT reduces depressive relapse compared with usual care and other active treatments (including antidepressants), using an IPD approach.
Whether treatment effects vary by patient-level moderators (demographic and depression-related variables).

Figure 1 — Study-level hazard ratios (MBCT vs non-MBCT)

Source of HR/CI values: Table/Figure content from the paper’s random-effects meta-analysis (2-stage).

Figure 2 — Pooled effects across control types

Comparisons taken from paper-reported 2-stage random-effects pooled results (HRs and 95% CIs).

Figure 3 — Time-varying effect signal (conceptual, paper-reported)

The authors report evidence that the effect of MBCT on relapse risk decreases as follow-up time progresses when using 1-stage fixed-effects with a time-varying specification: MBCT status HR ~0.34 (95% CI 0.19–0.60) plus MBCT×log(time) interaction HR ~1.28 (95% CI 1.06–1.55). Important skeptic note: this is not a full survival curve reconstruction because the paper does not provide the necessary baseline hazard function or time-grid outputs here.

Method quality checkpoints (IPD-focused)

Strength: IPD enables patient-level moderator analysis versus aggregate meta-analysis.

Strength: Follow-up harmonized to 60 weeks and relapse defined by diagnostic clinical interview; censoring at 60 weeks.

Bias risk: Not all trials blinded assessors; 2/9 did not blind assessors and 1 had incomplete outcome data.

Moderator results: what seems to generalize?

No support for differential effects by sex, age, education, or relationship status, suggesting generalizability across these demographics within available data.
Baseline depression severity showed a significant interaction with MBCT status: patients with higher baseline depression z-scores received greater benefit in relapse prevention.
However, the paper also acknowledges that moderator analyses were exploratory and not formally powered, increasing type I error risk.

Adverse events: what can we trust?

Adverse events were formally recorded in 6/9 studies, and none were attributed to MBCT.
Skeptical caveat: incomplete and non-systematic AE capture across trials limits the ability to rule out rare harms.

Bias & uncertainty audit (what could mislead us?)

Known unknowns are explicitly listed below; BGPT does not assume missingness is random.

Publication bias / small-study effects: funnel plot asymmetry reported; Egger test non-significant but limited power with 9 studies.
Missing moderator data: some important moderators (e.g., race/ethnicity, employment) were not available consistently for moderator analyses.
Allegiance effects: constituent trials were largely conducted by proponents of MBCT; authors used risk-of-bias analysis, but residual influence cannot be ruled out.
Data availability limitation: the paper reports unavailability of at least one trial’s IPD (Meadows et al.), impeding transparency and external scrutiny.

Conclusion (with confidence framing)

Known (from the paper): MBCT is associated with a reduced hazard of depressive relapse within 60 weeks vs non-MBCT controls (HR≈0.69, 95% CI 0.58–0.82), with comparably favorable pooled effects vs active treatments and vs antidepressant medication.
Uncertain (and why): moderator-specific claims (who benefits most) are based on exploratory moderation with some missing moderator data and assessor blinding limitations in a subset of trials; funnel asymmetry suggests small-study effects may be present.

Author-review links (skeptical cross-check)

Feedback:

Updated: April 12, 2026