Meta-Analyses of Mindfulness Trials: Effect Sizes and Active Controls

A meta-analysis pools many trials into one estimate. Because mindfulness trials are small and varied, meta-analyses are how researchers judge the whole field. This page walks through the best-known ones, explains effect sizes in plain terms, and shows why the comparison group matters. It is information only.

Reading an effect size

An effect size such as "d = 0.5" or "SMD = -0.5" expresses the difference between groups in standard-deviation units, so studies using different questionnaires can be combined. Bigger numbers mean a bigger average difference, but an average hides wide variation between people and between trials. A "prediction interval" shows the range where a future trial's result might fall. When it crosses zero, a new trial could show no benefit.

The comparison group is the key

• No treatment or waitlist. The easiest bar. Participants know they are waiting.
• Non-specific active control. Something like health education or relaxation that has attention and structure but no mindfulness.
• Specific active control or established treatment. Cognitive behavioral therapy, exercise or medicine.

The main meta-analyses

1. Goyal 2014

   Moderate evidence of better anxiety (0.38 at 8 weeks), depression (0.30) and pain (0.33); no evidence of beating active treatments [1]

   What it pooled: 47 randomized trials, 3,515 people, all with active controls [1]Notes: Searched to 2012

2. Khoury 2013

   Versus waitlist g = 0.53; versus other active treatments g = 0.33; no difference from CBT or behavioral therapies [2]

   What it pooled: 209 studies, 12,145 people [2]Notes: Includes uncontrolled studies

3. Goldberg 2018

   Versus no treatment d = 0.55; non-specific active 0.35; specific active 0.23; no difference from evidence-based treatments [3]

   What it pooled: 142 samples, 12,005 people with clinical problems [3]Notes: Most consistent for depression, pain, smoking, addictions [3]

4. Galante 2021

   Better than no intervention for anxiety, depression, distress and well-being; against active controls only depression clearly improved [4]

   What it pooled: 136 trials, 11,605 adults in non-clinical settings [4]Notes: Moderate to very low confidence [4]

5. Kuyken 2016

   Lower relapse risk over 60 weeks than usual care (hazard ratio 0.69) and than active treatments (0.79) [5]

   What it pooled: 9 trials of MBCT, 1,258 people with recurrent depression [5]Notes: Specific to MBCT

What the pattern shows

First, effects look larger against no treatment and smaller against active comparison groups. [3] In Goldberg's review the average dropped from 0.55 against no treatment to 0.23 against specific active controls. [3] Second, mindfulness programs did not differ from established treatments such as CBT, which suggests "about as good as", not "better than". [3][2]

Third, "most but not all scenarios": in Galante's review the ranges of likely true effects for anxiety, depression and distress extended to about zero or slightly beyond. [4] Fourth, USA-based trials reported smaller effects, and only the distress effect remained when higher-risk-of-bias trials were excluded. [4] Finally, a review of the reporting of results found that 87% of 124 trials reported a positive outcome in the abstract, more than expected from the average effect size. [6]

What these reviews cannot tell you

Effects are short-term snapshots: Goyal's anxiety effect fell from 0.38 at 8 weeks to 0.22 at 3 to 6 months. [1] The reviews say little about brief or app-based practice, and they do not show that mindfulness works for everything. For individual conditions see stress, sleep and anxiety, MBCT for depression and pain, attention and depression relapse. For the quality problems behind these numbers, see quality and bias in meditation research. The overall verdict is on does meditation work?.