Key research papers on mindfulness

How the literature is organised

The oldest listed work is Kabat-Zinn (1982), on an outpatient behavioural medicine programme for chronic pain patients based on the practice of mindfulness meditation. Twenty years later a cluster of papers turned that clinical practice into a research construct. Baer (2003) reviewed mindfulness training as a clinical intervention, concluding that the literature contained many methodological flaws but that the interventions might help in several disorders; Kabat-Zinn (2003) set the interventions in context. Bishop et al. (2004) proposed a two-component operational definition arrived at by consensus meetings. Brown and Ryan (2003) and Baer et al. (2006) supplied the questionnaires that made survey research possible, and Shapiro et al. (2006) proposed a model of how mindfulness might produce change, a question the clinical trials had left aside.

The literature since then has two halves. Trait studies correlate questionnaire scores with wellbeing, emotion regulation and performance; Brown, Ryan and Creswell (2007) is the theoretical statement of that programme. Intervention studies test programmes such as MBSR and Mindfulness-Based Cognitive Therapy against controls: Hofmann et al. (2010) pooled 39 studies with 1,140 participants and found moderate pre-to-post effects on anxiety and mood (Hedges's g around 0.6, and around 0.95 in patients with anxiety and mood disorders), and Grossman et al. (2004) is the other meta-analysis most often cited for effectiveness.

Main debates

The argument is about what the comparison group should be, and how good the trials are. Goldberg et al. (2022) reviewed 44 meta-analyses reporting 160 effect sizes across 336 randomised trials and 30,483 participants: mindfulness-based interventions beat passive controls almost everywhere (d = 0.10 to 0.89), but effects against active controls were typically smaller and less often statistically significant, they found important sources of bias, and reporting of adverse effects was inconsistent. Galante et al. (2021), pooling 136 trials in non-clinical settings, found improvements against no intervention but no statistically significant superiority over specific active controls, and rated confidence in the evidence as moderate to very low because of inconsistency and risk of bias. Zhang et al. (2021) make the same point about study quality across the wider field.

Against that background, Hoge et al. (2023) is the trial to know: MBSR was non-inferior to escitalopram for adults with anxiety disorders over eight weeks, with study-related adverse events in 15.4% of the MBSR group against 78.6% on the drug. The useful framing is comparable to a standard treatment, not better than everything.

Where recent work is heading

Large pragmatic trials now set the tone, and they are not uniformly encouraging. MYRIAD randomised 85 UK schools and just over 8,000 students and found no evidence that school-based mindfulness training was superior to normal social-emotional teaching at one year (Kuyken et al., 2022); the companion analysis (Montero-Marin et al., 2022) reports slightly worse scores for students already at risk of mental health problems, small enough not to be clinically relevant, and concludes that universal training in this format is not indicated in early adolescence. Lenze et al. (2022), with 585 older adults over 18 months, found no effect of mindfulness training or exercise on episodic memory or executive function; Whitfield et al. (2022), meta-analysing 45 studies, put the pooled cognitive effect at g = 0.15, with working memory and executive function the only subdomains reaching significance.

Targeted trials do better than universal ones. Garland et al. (2022) found Mindfulness-Oriented Recovery Enhancement reduced opioid misuse and pain compared with supportive group therapy among 250 primary care patients, and Sun et al. (2021) found smartphone-delivered training improved depression scores among pregnant women at risk of perinatal depression. Attention to adverse effects (Britton et al., 2021) has become part of the mainstream, and Baminiwatta and Solangaarachchi (2021) map the field's growth bibliometrically.