Of the 2463 titles identified, 11 studies (12 articles) were eligible for inclusion (Figure 2).17–19, 22–30
Included studies
Of the included studies, eight were conducted in the US,17,18,22–24,26–28 one in Germany,29 one in Hong Kong,30 and one in the UK.25 The majority of participants were female (with the exception of one study26) and white (with the exception of one study30). The mean age of participants ranged from 47–52 years, except the two studies with older adults27,28 (mean age 75 years). The majority of participants had completed secondary education and/or had a college qualification. Where employment status was reported, the majority were either employed or retired.19,22,23,27,28,30 Chronic pain populations studied were fibromyalgia (three studies),19,22–24 rheumatoid arthritis (two studies),17,29 chronic musculoskeletal pain (two studies),18,27 failed back surgery syndrome (one study),26 and mixed aetiology (three studies).27,28,30 Control groups included active comparator (six studies),17–19,24,27,30 waiting list control group (five studies),19,22,23,26,28,29 and standard care (two studies).18,25 Active comparators were educational control group,17,19,24,27 massage,18 multidimensional pain intervention (mainly educational),30 and cognitive behavioural therapy for pain.17 There were three 3-armed studies.17–19 Ten studies were randomised by patient18,19,22–30 and one was cluster randomised.17 Only four studies17,19,22,30 had involved sample size calculations and were adequately powered to detect an effect. Of the remaining seven studies, four were feasibility or pilot studies with small sample sizes,18,26–28 one failed to recruit,29 and two to retain24,25 target numbers. Included studies and their characteristics are displayed in Table 1.
Table 1. Methodological details of individual randomised controlled trials included (outcome data shown in Figure 3).
Four studies looked at economic outcomes: prescribed and over-the-counter medication use (three studies);19,24,26 use of complementary/alternative therapies (one study);24 doctors’ appointments (two studies);19,24 and sickness absence from work (one study).30
All eleven studies looked at clinical outcomes: pain intensity (eight studies);17,18,24–28,30 and depressive symptoms (six studies).17,19,22,24,29,30 Only three studies27,28,29 specifically collected data on adverse events but in practice none were reported.
Health-related quality of life (six studies)18,19,25,27,28,30 and physical functioning (five studies)19,24,26–28 were the most commonly reported humanistic outcomes. Three studies measured patient-centred outcomes (personal treatment goals,19 personal value of components of intervention,29 global impression of change,27 and perceived success at helping their back problem27). All outcomes measured are shown in Figure 3.
Figure 3. Outcome measures with primary outcomes identified where these are the basis for power calculation. EEG = electroencephalography. RA = rheumatoid arthritis.
Study quality and risk of bias
The YQRS scores of included studies varied from 10–31 points out of 35. The quality criteria commonly not met by studies included: adherence to manual (10 studies18,19,22–30), 6-month follow-up (nine studies18,19,22–29), attrition bias (differential rates of attrition between groups) (eight studies20–22,24–27,29,30), performance bias (differential equivalence in treatment expectations between groups) (eight studies17,18,22–26,28,29), and selection bias (independent allocation bias) (seven studies18,19,22–26,28). The quality rating for each of the studies is illustrated in Figure 4.
Figure 4: Yates Quality Rating of Studies. Items with two cells were allocated 2 points on the Yates Scale. Items with only one cell were allocated 1 point.
There is also the wider issue of publication bias that may have arisen due to dissemination of research findings being influenced by the nature and direction of results.20 Due to the small number of included studies, a funnel plot was not appropriate.
Recruitment and attrition
Recruitment to studies was generally by advertisements (posters, flyers, and at health fairs). An average of 47% of participants assessed and invited were randomised (range 4%25 to 87%26). Rates of attrition from studies are shown in Table 1. Drop-out from the mindfulness intervention ranged from 2%17 to 50%18 (median 20%), and drop-out from the mindfulness 8-week programme was higher than from active control group programmes, with two exceptions.17,24 Loss to follow-up ranged from 8–50% of intervention groups (median 20%) and from 0–52% of control groups (median 11%). Individual studies reported that drop-outs had a lower level of education (P = 0.05),28 were older (P = 0.03),27 and had greater baseline symptom severity (P = 0.005)22,23 and poorer physical functioning (P = 0.05).22,23
Humanistic outcomes
Health-related quality of life (HRQoL)
Four studies18,24,28,30 reported both a physical health component of HRQoL (combined effect size: 0.16 (95% CI = −0.15 to 0.47; I2 = 8%) and a mental health component (combined effect size of 0.37 (95% CI = −0.07 to 0.82; I2 = 46%).
Physical functioning
Five studies19,24,26–28 reported physical functioning (combined effect size: 0.22 (95% CI = 0.00 to 0.45; I2 = 0%).
Pain-related humanistic outcomes
Pain-related humanistic outcomes were looked at in six studies.17,24–28 There was a combined effect size of 1.58 (95% CI = −0.57 to 3.74) for pain acceptance (two studies26,28) (I2 = 91%) and of 0.58 (95% CI = 0.23 to 0.93) for perceived pain control (two studies17,25) (I2 = 0%).
Mindfulness
There was a combined effect size of 0.03 (95% CI = −0.66 to 0.72) for mindfulness (four studies19,25,27,29) with some evidence of heterogeneity (I2 = 85%).
Other outcomes
For patient-centred outcomes,19,23,27,29 results from mindfulness groups were generally positive but validated scales were often not used.