Acceptance- and mindfulness-based interventions for the treatment of chronic pain: a meta-analytic review

Abstract

The number of acceptance- and mindfulness-based interventions for chronic pain, such as acceptance and commitment therapy (ACT), mindfulness-based stress reduction (MBSR), and mindfulness-based cognitive therapy (MBCT), increased in recent years. Therefore an update is warranted of our former systematic review and meta-analysis of studies that reported effects on the mental and physical health of chronic pain patients. Pubmed, EMBASE, PsycInfo and Cochrane were searched for eligible studies. Current meta-analysis only included randomized controlled trials (RCTs). Studies were rated for quality. Mean quality did not improve in recent years. Pooled standardized mean differences using the random-effect model were calculated to represent the average intervention effect and, to perform subgroup analyses. Outcome measures were pain intensity, depression, anxiety, pain interference, disability and quality of life. Included were twenty-five RCTs totaling 1285 patients with chronic pain, in which we compared acceptance- and mindfulness-based interventions to the waitlist, (medical) treatment-as-usual, and education or support control groups. Effect sizes ranged from small (on all outcome measures except anxiety and pain interference) to moderate (on anxiety and pain interference) at post-treatment and from small (on pain intensity and disability) to large (on pain interference) at follow-up. ACT showed significantly higher effects on depression and anxiety than MBSR and MBCT. Studies’ quality, attrition rate, type of pain and control group, did not moderate the effects of acceptance- and mindfulness-based interventions. Current acceptance- and mindfulness-based interventions, while not superior to traditional cognitive behavioral treatments, can be good alternatives.

Keywords:

• Mindfulness
• acceptance and commitment therapy
• meta-analysis
• chronic pain
• acceptance

Introduction

Chronic pain is a major health problem with a large impact on the emotional, physical and social functioning of patients as well as society (Breivik, Eisenberg, O’Brien, & Openminds, 2013). Unfortunately, none of the most commonly used pharmacological, medical or surgical treatments are, by themselves, sufficiently able to remove pain, or to substantially enhance physical and emotional functioning (Turk, Wilson, & Cahana, 2011). This calls for the inclusion of psychosocial factors in addressing chronic pain and pain-related disabilities. Among psychological treatments, cognitive behavioral treatments (CBTs) have accumulated the most evidence in the research literature. However, as is the case for biomedical treatment solutions, effect sizes of CBTs are modest (Williams, Eccleston, & Morley, 2012).

The modest effects of both biomedical interventions and CBTs on pain removal and control are not surprising given the complex nature of chronic pain. They only shallowly reflect the large advances that have been made over the last decades in our understanding and treatment of chronic pain. They do, however, also point out a present necessity to negotiate with patients’ realistic expectations from treatments. In general total pain relief is an unrealistic goal and treatments should focus on improving function (Turk et al., 2011). Acceptance and Commitment Therapy (ACT) emphasizes the necessity of pain acceptance in order to improve function. ACT is based on the relational frame theory and focusses on psychological flexibility as the ultimate goal of treatment. Psychological flexibility refers to the capacity to change or maintain one’s behavior in open contact with thoughts and feelings and, with attention to the opportunities of the current situation in order to realize valued life goals (Hayes, Luoma, Bond, Masuda, & Lillis, 2006). In the context of chronic pain, psychological flexibility means that painful sensations, feelings, and thoughts are accepted, that attention is focused on the opportunities of the current situations rather than on ruminating about the lost past or catastrophizing about the future, and that behavior is focused on realizing valued goals instead of pain control (McCracken & Vowles, 2014). In ACT, psychological flexibility is attained by acquiring six psychological skills: acceptance, cognitive defusion, present moment awareness, contact with self-as context, values formulation and committed actions (Hayes, Strosahl, & Wilson, 2012). These skills can be divided in two overlapping processes, i.e. mindfulness and acceptance on the one hand and commitment and behavior change on the other hand (Hayes, Luoma, Bond, Masuda, & Lillis, 2006). In ACT mindfulness exercises in the form of meditation are advisable but not a precept.

Formal meditation is the central technique in mindfulness based stress reduction (MBSR) (Kabat-Zinn, 1990) and mindfulness-based cognitive therapy (MBCT) (Segal, Williams, & Teasdale, 2002). Mindfulness is defined as intentional and non-judgmental awareness (Kabat-Zinn, 1990). It can be conceptualized as a multifaceted construct, consisting of the facets: observe, describe, act with awareness, nonjudge and nonreact. These last two facets are strongly related to acceptance (Baer, Smith, Hopkins, Krietemeyer, & Toney, 2006). Although ACT, MBSR and MBCT can be differentiated theoretically, they share an underlying focus on the concepts of acceptance and mindfulness. In the systematic review and meta-analysis with studies published before January 2009, we concluded that acceptance and mindfulness-based therapies in chronic pain produced small effects that are similar to CBT (Veehof, Oskam, Schreurs, & Bohlmeijer, 2011). Since our last analysis, the number of studies on acceptance- and mindfulness-based studies has expanded considerably. Systematic reviews that qualitatively assess the efficacy of ACT, MBSR and MBCT generally find beneficial effects in chronic pain conditions (Lakhan & Schofield, 2013; McCracken & Vowles, 2014). The rapid increase of acceptance- and mindfulness-based studies in chronic pain allowed for a more rigorous assessment of their effectiveness. The aim of the present study, therefore, was to conduct a meta-analysis to assess the effectiveness of acceptance- and mindfulness-based treatments for chronic pain patients across randomized controlled studies.

Methods

A protocol based on the Cochrane Manual for Systematic Reviews of Interventions was used.

Search strategy

We included studies identified in our previous meta-analysis (Veehof et al., 2011) (up to December 2008), and updated this list of studies by conducting a new search (from January 2009 to December 2013) in the four electronic databases: PubMed, EMBASE, PsycInfo, and the Cochrane Central Register of Controlled Trials. The databases were searched for English language studies using the same terms as in our previous search: ‘mindfulness’ or ‘vipassana’ or ‘meditation’ or ‘mindfulness-based stress reduction’ or ‘MBSR’ or ‘mindfulness-based cognitive therapy’ or ‘MBCT’ or ‘acceptance-based’ or ‘acceptance based’ or ‘acceptance and commitment’, in combination with ‘chronic pain’ or specific chronic pain conditions including ‘fibromyalgia’ or ‘chronic low back pain’ or ‘whiplash associated disorder’ or ‘WAD’ or ‘repetitive strain injury’ or ‘RSI’ or ‘dystrophy’. Given recent developments in diagnosing specific chronic pain conditions, we added the following terms: ‘complaints arm neck shoulder’ or ‘CANS’ ‘complex regional pain syndrome’ or ‘CRPS’. Furthermore, the reference lists of newly included studies as well as systematic reviews were examined for additional eligible studies.

Selection procedure and eligibility criteria

Inclusion criteria were the same as the ones in our previous review, except that all studies had to have a randomised controlled design (Veehof et al., 2011). The new search initially yielded 1393 titles (PubMed: 927, EMBASE: 222, PsycInfo: 151, Cochrane: 93). After removal of duplicates, two reviewers (MV and KS) independently selected potentially eligible studies on the basis of title and abstract. Systematic reviews were initially included in this phase. The inter-rater reliability was satisfactory (κ = .79) (Everitt, 1992), and disagreements were resolved by consensus.

A total of 163 studies (including 76 systematic reviews) were identified as being potentially eligible for inclusion in the study, and we subsequently requested the full-text articles. The final selection was then made by two reviewers (MV and KS). The inter-rater reliability was very good with a κ value of .93 (Everitt, 1992). Studies were included if they reported on the effectiveness of a standardized acceptance- (ACT) or mindfulness (MBSR/MBCT)-based treatment program in patients with chronic pain or chronic pain related conditions. Given the increasing number of (randomized) controlled studies published in this field in recent years, only randomized controlled studies were included. Both published and unpublished (e.g. dissertations) studies were included. Studies were excluded if (1) mindfulness or acceptance was just one of several modalities provided simultaneously to the treatment group, (2) the intervention consisted of a single treatment session, (3) insufficient data was reported (and could not be obtained) to calculate standardized mean differences (SMDs), or (4) absence of appropriate outcome measures (pain, depression, anxiety, pain interference, disability or quality of life). Systematic reviews were only examined for additional potentially eligible studies and were not definitely included.

Data extraction

Data extraction was performed using a standardized data abstraction form created for the study. Data were extracted on participant characteristics, intervention type, control group and attrition rate. In agreement with the Clinical Importance of Treatment Outcomes in Chronic Pain Clinical Trials recommendations (Dworkin et al., 2005), our outcome measures were pain intensity, depression, anxiety, disability, pain interference, and quality of life.

Quality assessment

Methodological quality of included studies was independently assessed by two reviewers (MV and KS) using a 8-point scale, based on criteria by the Cochrane Collaboration (Higgins & Altman, 2008) and the validated Jadad scale (Jadad, 1996) tailored for the included studies (see Table 1). The inter-rater reliability was satisfactory (κ = .78) (Everitt, 1992), and disagreements were resolved by consensus. When seven or more criteria were met, the quality of the study was assessed as high. The study quality was assessed as medium when four, five or six criteria were met and low when three or less criteria were met.

Table 1. Methodological quality criteria.^a

1.	Allocation to conditions was based on randomization according to the text	1/0
2.	The randomization scheme was described and appropriate, e.g. using a computer, random number table	1/0
3.	A drop-out analysis was conducted (comparing drop-outs and completers at baseline on outcome measures), or there were no drop-outs	1
	Reasons of attrition were reported, but no analysis was conducted	0
4.	Intention to treat analysis was performed, or there were no drop-outs	1/0
5.	At least one of the trainers was experienced or trained in teaching Mindfulness or ACT	1
	Specific experience or training was not reported	0
6.	Patient’s pain was diagnosed by a physician or rheumatologist, or patients were referred from a pain clinic where diagnosis was prior to admission	1
	Recruitment through the media, or diagnosis based on a scale and self-report, or diagnosis not mentioned	0
7.	The study had a minimal level of statistical power to find significant effects of the treatment, and included 50 or more persons in the comparison analysis between treatment and control group. (This allows the study to find standardized effect sizes of .80 and larger, assuming a statistical power of .80 and α of .05.)	1
	Sample smaller than 50, or the total the sample was bigger than 50, but the results were only reported divided by different studies	0
8.	Treatment integrity was checked during the study by supervision of the therapists during treatment, or by recording the treatment sessions, or by systematic screening of protocol adherence by a standardized measurement instrument	1
	Treatment integrity was not checked, or integrity was supervised by one of the therapists, or they tried to keep the intervention sound by intensive consultation	0

^a Scores are summed, 1 point for each fulfilled criterion, range 0–8.

Data analysis

Primary goal was to compare the effects of acceptance- and mindfulness-based interventions with no treatment or an attention control condition. Therefore, for our main analyses, we only included studies with a waitlist, a (medical) treatment as usual [(M)TAU] or, an education/support group as a control condition. Secondary goal was to compare effects of acceptance- and mindfulness-based treatments with an active treatment like CBT, multidisciplinary treatment (MDT) or relaxation therapy.

We used RevMan software version 5.2 for calculating (pooled) SMDs to test heterogeneity and to perform subgroup analyses. SMDs were calculated at post-treatment and follow-up using post-treatment and follow-up scores respectively with standard deviations from the experimental group and the control group. Given the presence of statistical heterogeneity between the studies (tested with the χ² test and I²-statistic), pooled SMDs were calculated using the random-effects model representing the average intervention effect. Cohen (Cohen, 1977) has described effect sizes .2, .5, and .8 as small, moderate, and large, respectively.

In our main analysis, subgroup analyses were performed for all outcome measures by comparing pooled SMDs for significant differences between subgroups. Since these analyses were exploratory, no hypotheses were formulated beforehand. Subgroups were created based on (1) quality score (divided by low, medium and high quality as mentioned earlier), (2) control group [waitlist or (M)TAU vs. education/support], (3) intervention type (ACT vs. mindfulness-based), (4) type of pain (categorized as chronic pain without further specification, fibromyalgia, specific site pain, or rheumatoid arthritis), and (5) attrition rate (higher or lower than 25%).

To assess for publication bias, we performed a funnel plot for all outcome measures by plotting the pooled SMD at post-treatment against its standard error (as a measure of study size). When publication bias is absent, the studies are expected to be distributed symmetrically around the pooled effect size. Bias can be expected when the plot shows a higher concentration of studies on one side of the pooled SMD than on the other.

Results

A flow chart of the selection procedure is shown in Figure 1. A total of 87 studies were identified as being potentially eligible for inclusion in the study. The full-text versions of these studies were independently assessed by two reviewers (MV and KS). Of the 87 articles, 67 were excluded for the following reasons: no acceptance- and mindfulness-based intervention (Coleman, 2011; Cramer et al., 2013; Curtis, Osadchuk, & Katz, 2011; Elinoff, Lynn, Ochiai, & Hallquist, 2009; Elomaa, Williams, & Kalso, 2009; Evans et al., 2011; Friedberg et al., 2013; Hirschfeld et al., 2013; Hsu et al., 2010; Liu et al., 2012; Paoloni et al., 2013; Rasmussen, Mikkelsen, Haugen, Pripp, & Forre, 2009; Rasmussen et al., 2012; Sawynok, Hiew, & Marcon, 2013; Stinson et al., 2010; Tekur, Chametcha, Hongasandra, & Raghuram, 2010; Tekur, Nagarathna, Chametcha, Hankey, & Nagendra, 2012; Vincent, Hill, Kruk, Cha, & Bauer, 2010), mindfulness or acceptance is not the main component of the intervention (Carbonell-Baeza et al., 2011; Carson, Carson, Jones, Mist, & Bennett, 2012; Carson et al., 2010; Cassidy, Atherton, Robertson, Walsh, & Gillett, 2012; Smeeding, Bradshaw, Kumpfer, Trevithick, & Stoddard, 2011), no intervention study (Morone et al., 2012), only CFS no chronic pain (Rimes & Wingrove, 2013), no control group (Abbey & Nanke, 2013; Baer, Carmody, & Hunsinger, 2012; Duggan et al., 2015; Fox, Flynn, & Allen, 2011; Gauntlett-Gilbert, Connell, Clinch, & McCracken, 2013; Hawtin & Sullivan, 2011; Hennard, 2011; Ilgen et al., 2011; Jelin, Granum, & Eide, 2012; Kristjánsdóttir et al., 2011; Ljotsson et al., 2013; Lunde & Nordhus, 2009; Lush et al., 2009; Mathias, Parry-Jones, & Huws, 2013; McCracken & Gutiérrez-Martínez, 2011; McCracken & Jones, 2012; Oberg, Rempe, & Bradley, 2013; Rosenzweig et al., 2010; Vernon, 2011; Vowles & McCracken, 2010; Vowles, McCracken, & O’Brien, 2011; Vowles, Witkiewitz, Sowden, & Ashworth, 2014), no randomization (Cusens, Duggan, Thorne, & Burch, 2010; Vowles, Wetherell, & Sorrell, 2009), sample contained subjects other than pain patients (Branstetter-Rost, Cushing, & Douleh, 2009; Fjorback et al., 2012, 2013; Hartmann & Vlieger, 2012; Sampalli, Berlasso, Fox, & Petter, 2009), no appropriate outcome measure(s) (Cathcart, Barone, Immink, & Proeve, 2013; Cathcart, Vedova, Immink, Proeve, & Hayball, 2013; Garland & Howard, 2013; McCracken, Sato, & Taylor, 2013; Vago & Nakamura, 2011), 100% attrition in the control group (Jastrowski Mano et al., 2013), duplicate study data (Jensen et al., 2012; Wicksell, Olsson, & Hayes, 2010, 2011), insufficient quantitative data or none at all (Esmer, Blum, Rulf, & Pier, 2010; Olsson et al., 2013; S. Y. Wong, 2009), single treatment session (Martin, 2012). As a result, 20 new studies were included. Together with the eight randomized controlled studies included in our previous meta-analysis, we definitely included 28 studies. Our main analyses consists of 25 studies comparing acceptance- and mindfulness-based interventions with waitlist, (M)TAU or, education/support groups. Two of these 25 studies also used an active intervention as control (Schmidt et al., 2011; Zautra et al., 2008). Three studies only made a comparison with an active intervention (Thorsell et al., 2011; Wetherell et al., 2011; Wicksell, Melin, Lekander, & Olsson, 2009). So, five studies were included for our secondary analyses, comparing acceptance- and mindfulness-based interventions with active treatment.

Figure 1. Flow chart of the selection procedure.

Characteristics of included studies

Characteristics of the selected studies are presented in Table 2. The 25 studies included in the main analyses evaluated 1285 subjects. In general, the participants were adults with a mean age between 35 and 60 years. In two studies (Morone, Greco, & Weiner, 2008; Morone, Rollman, Moore, Li, & Weiner, 2009) the mean age of the participants was higher (> 70 years). In all studies the majority of the participants were female. Nine studies referred to patients with chronic pain without further specification and, one to patients with musculoskeletal pain Seven studies included patients with fibromyalgia, six with specific-site pain (e.g. chronic low back pain, chronic headache), and two with rheumatoid arthritis. Study sizes ranged from a small pilot study of 14 subjects to a large-scale study involving 112 subjects, with an average of 51 subjects. In ten studies the attrition rate (in one or both groups) was higher than 25% (range 26.7% to 67.3%). Eleven studies used a MBSR(-based) program, nine an ACT(-based) program, two a MBCT(-based) program, and one a combination of MBCT and MBSR. Two studies used other mindfulness-based programs: Mindfulness-based Pain Management (MBPM) and Four Step Mindfulness-based Therapy (FSMT). Most mindfulness-based programs (MBSR, MBCT, MBPM, FSMT) consisted of eight weekly group sessions, each session ranging in length from 1.5 to 2.5 h. In one MBSR study a self-help program was used. A self-help program was also used in two ACT studies. In the seven other ACT studies, four to twelve weekly sessions were held, each session ranging in length from 1.0 to 2.5 h. In two of these studies, the sessions were held with individual patients. In the other five studies, the sessions were group-based. Ten studies used a waitlist as the comparison group, eight used an education/support group, and seven used a (M)TAU as the control group. The majority (n = 14) of the studies scored ‘medium’ on the quality criteria, six scored ‘high’ and five scored ‘low.’ None of the 25 studies met all the quality criteria.¹

Table 2. Characteristics of included studies.

Author	Follow-up		Quality score	Pain	Mean age (SD or range)	Male %	Intervention	Group size	Sessions: number, duration	Control group	n	Attrition rate % (follow-up)	Outcome measures
Comparison with education/support, (M)TAU or waitlist as control group
Astin et al., 2003		16 w	5	Fibromyalgia	T: 47.7 (10.6)	I: 1.6	MBSR + Qigong	10–20	8, 2.5 h	Education/support	I: 32	T: 39.0 (49.0)	Pain: SF-36
						C: 0.0					C: 33		Depression: BDI
													QoL: FIQ
Brown & Jones, 2013		–	2	Musculo-skeletal pain	I: 48 (10)	I: 20.0	MBPM	Max 15	8, 2.5 h	TAU	I: 15	I: 25.0	Pain: MPQ (sensory pain)
					C: 45 (12)	C:30.8					C: 13	C: 35.0	Disability: SF-36 PCS
Buhrmao et al., 2013		6 m experimental group only	5	Chronic pain	I: 48.8 (9.5)	I: 44.7	ACT	Self-help (online)	7	Education/support	I: 38	I: 15.8	Pain: MPI
					C: 49.3 (11.3)	C: 36.8					C:38	C:23.7	Depression: HADS
													Anxiety: HADS
													Interference: MPI
													QoL: QOLI
Cathcart, Galatis, Immink, Proeve, & Petkov, 2014		–	2	Chronic tension-type headache	I: 45.8 (13.1)	I: 43.0	MBCT/MBSR		6, 2 h	Waitlist	I: 22	I: 20.7	Pain: 0–5 scale
					C: 45.3 (14.2)	C: 31.5					C: 19	C: 34.5
Dahl, Wilson, & Nilsson, 2004		6 m	4	Chronic pain	I: 36.7 (12.5)	I: 9.1	ACT	Individual	4, 1 h	MTAU	I: 11	0	Pain: NRS
					C: 44.4 (13.6)	C: 12.5					C: 8	0	QoL: LSQ
Davis & Zautra, 2013		–	5	Fibromyalgia	T: 46.1	T: 2.0	MBSR	Self-help	12 (6 w)	Education/support	I: 19	I: 51.3	Pain: NRS
											C: 25	C: 37.5
Garland, Manusov, Froeliger, Kelly & Williams, 2013		3 m	7	Chronic pain	I: 49.3 (13.7)	I: 30	MBCT + positive psychology	8–12	8, 2 h	Education/	I: 31	I:45.6 (59.6)	Pain: BPI
					C: 47.4 (13.6)	C: 34				support	C: 38	C: 34.4 (51.7)	Depression: C-SOSI
													Interference: BPI-SF
Johnston, Foster, Shennan, Starkey, & Johnson, 2010		–	2	Chronic pain	T: median 43	T: 37.5	ACT	Self-help	6	Waitlist	I: 6	I: 50.0	Pain: SF-MPQ
											C: 8	C: 33.3	Depression: CMDI Anxiety: BAI
													QoL: QOLI
Luciano et al., 2014		6 m	7	Fibromyalgia	I: 48.9 (5.9)	I: 3.9	ACT	10–15	8, 2.5 h	Waitlist	I: 51	I: 9.8 (11.8)	Pain: VAS
					C: 48.3 (5.7)	C: 5,7					C: 53	C: 5.7 (11.3)	Depression: HADS
													Anxiety: HADS
													QoL: FIQ
Mawani, 2009		–	2	Chronic pain	I: 46.2 (12.1)	I: 16.0	MBSR	4–8	8, 2 h	Education/	I: 14	I: 51.7	Pain: MPQ PRI
					C: 44.7 (13.2)	C: 31.0				support	C: 14	C: 56.3
McCracken et al., 2013		3 m	6	Chronic pain	T: 58.0 (12.8)	T: 31.5	ACT	12–13	4, 4 h	TAU	I: 28	I: 16.2 (24.3)	Pain: NRS
											C: 26	C: 25.0 (22.2)	Depression: PHQ-9
													Disability: RMDQ
Morone et al., 2008		–	3	Chronic low back pain	I:74.1 (6.1)	T: 43.2	MBSR		8, 1.5 h	Waitlist	I: 19	I: 31.6	Pain: MPQ-SF
					C: 75.6 (5.0)						C: 18	C: 5.6	Disability: SF-36 PCS
													QoL: SF-36 GH
Morone et al., 2009		2 m	4	Chronic low back pain	I: 78 (7.1)	I: 31.2	MBSR		8, 1.5 h	Education/	I: 16	I: 20.0 (20.0)	Pain: MPQ-SF current pain score
					C: 73 (6.2)	C: 42.1				support	C: 19	C: 5.0 (5.0)	Disability: RMDQ
Mo’tamedi et al., 2012		–	5	Primary chronic headache	I:34.2 (7.4)	T: .0	ACT	15	8, 1.5 h	MTAU	I: 15	I:26.7	Pain: MPQ sensory dimension
					C: 37.9 (8.7)						C: 15	C:0	Disability: MIDAS
Nash-Mc Feron, 2006		–	4	Chronic headache	I: 50 (14.6)	I: .0	MBSR	20	8, 1.5 h	TAU	I: 19	I: 5	Pain: SF-36
					C: 51 (12.2)	C: 35.0					C: 18	C:10	Disability: SF-36 PF
Parra-Delgado & Latorre-Postigo, 2013		3 m	4	Fibromyalgia	I: 53.1 (10.5)	T: .0	MBCT	17	8, 2.5 h	TAU	I: 15	I: 11.8 (11.8)	Depression: BDI
					C: 52.7 (10.6)						C: 16	C: .0	QoL: FIQ
Plumb Vilardaga, 2012		–	4	Chronic pain	T: 52.3 (13.0)	T: 25.0	ACT	15	6, 2 h	Waitlist	I: 11	I: 26.7	Pain: PIR
											C: 9	C: 30.8	Depression: DASS
													Disability: PDI
													QoL: QOLS
Pradhan et al., 2007		4 m	7	Rheumatoid arthritis	I: 56 (9)	I: 16.0	MBSR	13–18	8, 2.5 h	Waitlist	I: 28	I: 9.7 (9.7)	Depression: SCL-90
					C: 53 (11)	C: 9.0					C: 32	C: 0 (6.3)
Schmidt et al., 2011		8 w	6	Fibromyalgia	I: 53.4 (8.7)	T: .0	MBSR	Max 12	8, 2.5 h	Waitlist	I: 53	I:15.1 (11.3)	Pain: PPS (sensory pain)
					C: 52.3 (10.9)						C: 59	C: 11.9 (5.1)	Depression: CES-D
													QoL: FIQ
Sephton et al., 2007		2 m	6	Fibromyalgia	I: 48.4 (8.9)	T: .0	MBSR	25	8, 2.5 h	Waitlist	I: 51	28.6 (38.5)	Depression: BDI
					C:47.6 (11.5)						C: 39
Wicksell, Ahlqvist, Bring, Melin, & Olsson, 2008		4 m	6	Whiplash-associated disorders	I: 48.2 (7.8)	I: 18.2	ACT	Individual	10, 1 h	Waitlist + TAU	I: 11	I: .0 (9.1)	Pain: VAS
					C: 55.1 (11.2)	C: 30.0					C: 9	C: 10.0 (.0)	Depression: HADS
													Anxiety: HADS
													Disability: PDI
													Interference: VAS
													QoL: SWLS
Wicksell et al., 2013		3–4 m	6	Fibromyalgia	T: 45.1 (6.6)	T: 0	ACT	6	12, 1.5 h	Waitlist	I: 20	I: 13.0 (17.4)	Pain: NRS
											C:16	C: 5.9 (17.6)	Depression: BDI
													Anxiety: STAI state
													Disability: SF-36 PCS
													QoL: FIQ
C. M. Wong, 2009		–	7	Chronic pain	I: 48.3 (8.2)	I: 47.4	FSMT	10–12	8, 2 h	TAU	I: 38	I: 32.0	Pain: NRS
					C: 40.1 (12.9)	C: 41.2					C: 27	C: 44.9	Depression: BDI
													Interference: MPI
Wong et al.,2011;		3 m, 6 m	7	Chronic pain	I: 48.7 (7.8)		MBSR		8, 2.5 h	Education/support	I: 51	I: 13.7 (21.6, 19.6)	Pain: NRS
					C: 47.1 (7.8)						C: 48	C: 4.2	Depression: CES-D
												(6.3, 6.3)	Anxiety: STAI state
													Disability: SF-12 PCS
Zautra et al., 2008		–	7	Rheumatoid arthritis	I: 57.3 (15.3)	I:46.3	MBSR	6–10	8	Education/support	I: 41	I: 8.3	Pain: NRS
					C: 52.4 (13.0)	C 23.3					C: 30	C: 2.3	Depression: own scale
Comparison with CBT
Zautra et al., 2008;	–		7	Rheumatoid arthritis	I: 57.3 (15.3)	I:46.3	MBSR	6–10	8	CBT	I: 41	I: 8.3	Pain: NRS
					C: 56.1 (13.5)	C:40.0					C: 35	C: 4.0	Depression: own scale
Wetherell et al., 2011;	6 m		8	Chronic pain	T: 54.9 (12.5)	T: 49.1	ACT	6	8, 1.5 h	CBT	I:57	I: 24.6	Pain: BPI-SF
											C: 57	C: 26.3	Depression: BDI
													Disability: SF-12 PCS
													Interference: BPI-SF
Comparison with MDT/relaxation
Wicksell et al., 2009	3.5 m, 6.5 m		5	Longstanding idiopathic pain	T: 14.8 (2.4)	T: 21.9	ACT	Individual	10, 1 h	MDT	I: 16	I: 6.3 (12.5, 18.8)	Pain: VAS
											C:16	C: 12.5 (31.3, 31.3)	Depression: CES-D
													Disability: SF-36 PCS
													Interference: MPI and BPI
Schmidt et al., 2011;	8 w		6	Fibromyalgia	I: 53.4 (8.7)	T: .0	MBSR	Max 12	8, 2.5 h	Relaxation	I: 53	I:15.1 (11.3)	Pain: PPS (sensory pain)
					C: 51.9 (9.2)						C: 56	C: 8.9 (12.5)	Depression: CES-D
													QoL: FIQ
Thorsell et al., 2011	6 m,		6	Chronic pain	T: 46.0 (12.3)	T: 35.6	ACT	Individual	9	Relaxation	I: 52	I: 46.2 (48.1, 67.3)	Pain: NRS
	12 m										C: 38	C: 28.9 (31.6, 60.5)	Depression: HADS
													Anxiety : HADS
													Disability: ÖMPQ

Notes: ACT: acceptance and commitment therapy; C: control group; CBT: cognitive behavioural therapy; FSMT: four step mindfulness-based therapy; h: hours; I: intervention group; m: months; MBCT: mindfulness-based cognitive therapy; MBPM: mindfulness-based pain management; MBSR: mindfulness-based stress reduction; MDT: multidisciplinary treatment; (M)TAU: (medical) treatment as usual; SD: standard deviation; T: total group; w: weeks.

Outcome measures: BAI: Beck Anxiety Inventory; BDI: Beck Depression Inventory; BPI: Brief Pain Inventory; CES-D: Center for Epidemiologic Studies Depression Scale; CMDI: Chicago Multi-scale Depression Inventory; C-SOSI: Calgary Symptoms of Stress Inventory; DASS: Depression Anxiety and Stress Scale; FIQ: Fibromyalgia Impact Questionnaire; HADS: Hospital Anxiety and Depression Scale; LSQ: Life Satisfaction Questionnaire; MIDAS: Migraine Disability Assessment Scale; MPI: Multidimensional Pain Inventory; MPQ: McGill Pain Questionnaire; NRS: Numerical Rating Scale; ÖMPQ: Örebro Musculoskeletal Pain Questionnaire; PDI: Pain Disability Index; PHQ-9: Patient Health Questionnaire-9; PIR: Pain Intensity Rating; PPS: Pain Perception Scale; PRI: Pain Rating Index (MPQ); QoL: Quality of Life Profile for the Chronically Ill; QOLI: Quality Of Life Inventory; QOLS: Quality Of Life Scale; RMDQ: Roland and Morris Disability Questionnaire; SCL-90: 90-item Symptom Checklist; SF-12/SF-36: 12/36-item Short-Form Health Survey; SF-36 GH/PCS/PF: SF-36 General Health/Physical Component Summary/Physical Functioning; STAI: Stait-Trait Anxiety Inventory; SWLS: Satisfaction With Life Scale; VAS: Visual Analogue Scale; WSAS: Work and Social Adjustment Scale.

n = sample size of post-treatment analyses, lowest sample size in case sample sizes differed among outcome measures.

Characteristics of the three additional studies that only used an active control group are given in Table 2.

Publication bias

Some indication for publication bias was found for the outcome measure of depression. The funnel plot for depression is presented in Figure 2, and shows asymmetrically distributed studies in the bottom of the figure, which displays smaller studies. This might be due to the fact that smaller studies are more likely to be published if they have larger-than-average effects (Borenstein, 2005). The funnel plots for the other outcome measures were symmetrically distributed around the pooled SMD, which is an indication for the absence of publication bias.

Figure 2. Funnel plot for depression.

Post-treatment effects

Twenty-two studies assessed the effects at post-treatment on pain intensity, 16 on depression, six on anxiety, four on pain interference, 10 on disability, and 11 on quality of life. Pooled SMDs and the results of the tests for heterogeneity and overall effect are presented in Table 3. Small effects were found for pain intensity, depression, disability, and quality of life with mean SMDs of respectively .24 (95% CI = .06, .42), .43 (95% CI = .18, .68), .40 (95% CI = .01, .79), and .44 (95% CI = −.05, .93). A moderate effect was found for anxiety and pain interference with mean SMD of respectively .51 (95% CI = .10, .92) and .62 (95% CI = .21, 1.03). All effects were statistically significant, except for the effect on quality of life (p = .08). Statistical heterogeneity between the studies was caused by the studies of Luciano et al. (2014) (for pain, depression, anxiety and quality of life) and Mo’tamedi, Rezaiemaram, and Tavallaie (2012) (for disability). Both studies, which were respectively of high and medium quality, found large beneficial effects of acceptance- and mindfulness-based interventions.

Table 3. Effects at post-treatment and follow-up.

Outcome measures	Time	n	Studies	Heterogeneity	Pooled SMD (95% CI)	Test for overall effect
Pain	Post-treatment	22	Astin et al., 2003; Brown & Jones, 2013; Buhrman et al., 2013; Cathcart et al., 2014; Dahl et al., 2004; Davis & Zautra, 2013; Garland & Fredrickson, 2013; Johnston et al., 2010; Luciano et al., 2014; Mawani, 2009; McCracken et al., 2013; Morone et al., 2008, 2009; Mo’tamedi et al., 2012; Nash-Mc Feron, 2006; Plumb Vilardaga, 2012; Schmidt et al., 2011; Wicksell et al., 2008, 2013; C. M. Wong, 2009; Wong et al., 2011; Zautra et al., 2008	χ^2 = 41.53, df = 21 (p = .005); I^2 = 49%	.24 (.06, .42)	Z = 2.68 (p = .007)
	Follow-up	10	Astin et al., 2003; Dahl et al., 2004; Garland & Fredrickson, 2013; Luciano et al., 2014; McCracken et al., 2013; Morone et al., 2009; Schmidt et al., 2011; Wicksell et al., 2008, 2013; Wong et al., 2011	χ^2 = 20.93, df = 9 (p = .01); I^2 = 57%	.41 (.14, .68)	Z = 3.02 (p = .003)
Depression	Post-treatment	16	Astin et al., 2003; Buhrman et al., 2013; Garland & Fredrickson, 2013; Johnston et al., 2010; Luciano et al., 2014; McCracken et al., 2013; Parra-Delgado & Latorre-Postigo, 2013; Plumb Vilardaga, 2012; Pradhan et al., 2007; Schmidt et al., 2011; Sephton et al., 2007; Wicksell et al., 2008, 2013; C. M. Wong, 2009; Wong et al., 2011; Zautra et al., 2008	χ^2 = 52.30, df = 15 (p < .00001); I^2 = 71%	.43 (. 180, .68)	Z = 3.38 (p = .0007)
	Follow-up	10	Astin et al., 2003; Luciano et al., 2014; McCracken et al., 2013; Parra-Delgado & Latorre-Postigo, 2013; Pradhan et al., 2007; Schmidt et al., 2011; Sephton et al., 2007; Wicksell et al., 2008, 2013; Wong et al., 2011	χ^2 = 30.65, df = 9 (p = .0003); I^2 = 71%	.50 (.19, .80)	Z = 3.22 (p = .001)
Anxiety	Post-treatment	6	Buhrman et al., 2013, Johnston et al., 2010; Luciano et al., 2014; Wicksell et al., 2008, 2013; Wong et al., 2011	χ^2 = 15.13, df = 5 (p = .01); I^2 = 67%	.51 (.10, .92)	Z = 2.42 (p= .02)
	Follow-up	4	Luciano et al., 2014; Wicksell et al., 2008, 2013; Wong et al., 2011	χ^2 = 12.07, df = 3 (p = .007); I^2 = 75%	.57 (.00, 1.14)	Z = 1.97 (p = .05)
Pain interference	Post-treatment	4	Buhrman et al., 2013; Garland & Fredrickson, 2013; Wicksell et al., 2008; C. M. Wong, 2009	χ^2 = 6.50, df = 3 (p = .09); I^2 = 54%	.62 (.21, 1.03)	Z = 2.94 (p = .003)
	Follow-up	2	Garland & Fredrickson, 2013; Wicksell et al., 2008	χ^2 = .58, df = 1 (p = .44); I^2 = 0%	1.05 (.55, 1.56)	Z = 4.09 (p < .0001)
Disability	Post-treatment	10	Brown & Jones, 2013; McCracken et al., 2013; Morone et al., 2008, 2009; Mo’tamedi et al., 2012; Nash-Mc Feron, 2006; Plumb Vilardaga, 2012; Wicksell et al., 2008, 2013; Wong et al., 2011	χ^2 = 30.12, df = 9 (p = .0004); I^2 = 70%	.40 (.01, .79)	Z = 2.01 (p = .04)
	Follow-up	5	[68, 74, 85, 110, 111, 117] McCracken et al., 2013; Morone et al., 2009; Wicksell et al., 2008, 2013; Wong et al., 2011	χ^2 = 4.54, df = 4 (p = .34); I^2 = 12%	.39 (.11, .67)	Z = 2.71 (p = .007)
Quality of life	Post-treatment	11	Astin et al., 2003; Buhrman et al., 2013; Dahl et al., 2004; Johnston et al., 2010; Luciano et al., 2014; Morone et al., 2008; Parra-Delgado & Latorre-Postigo, 2013; Plumb Vilardaga, 2012; Schmidt et al., 2011; Wicksell et al., 2008, 2013	χ^2 = 68.44, df = 10 (p < .00001); I^2 = 85%	.44 (−.05, .93)	Z = 1.77 (p = .08)
	Follow-up	7	Astin et al., 2003; Dahl et al., 2004; Luciano et al., 2014; Parra-Delgado & Latorre-Postigo, 2013; Schmidt et al., 2011; Wicksell et al., 2008, 2013	χ^2 = 41.41, df = 6 (p < .00001); I^2 = 86%	.66 (.06, 1.26)	Z = 2.16 (p = .03)

Notes:

SMD, standardized mean difference; 95% CI, 95% confidence interval; df, degrees of freedom; n.a., not applicable.

Follow-up effects

Thirteen studies assessed long-term effects. Follow-up periods ranged from 2 months to 6 months after completing treatment. Wong et al. (2011) reported follow-up results at both 3 and 6 months. In this case, we included the longest follow-up period. Ten studies assessed the effects on pain intensity, 10 on depression, four on anxiety, two on pain interference, five on disability, and seven on quality of life. Pooled SMDs and the results of the tests for heterogeneity and overall effect are presented in Table 3. The effect on pain increased but remained small with a mean SMD of .41 (95% CI = .14, .68). The effects on depression and quality of life (slightly) increased and became moderate with mean SMDs of respectively .53 (95% CI = .24, .82) and .66 (95% CI = .06, 1.26). The effect on pain interference increased and became large with a mean SMD of 1.05 (95% CI = .55, 1.56). The effects on anxiety and disability remained moderate with mean SMDs of respectively .59 (95% CI = .14, 1.04) and .39 (95% CI = .11, .67). All effects were statistically significant.

Subgroup analyses

Results of the subgroup analyses are presented in Table 4. Significant differences in mean SMDs were found between some subgroups. ACT interventions reported a statistically significant higher mean effect on depression (SMD .82 vs. .18) and anxiety (SMD .64 vs. −.01) than mindfulness-based interventions. Furthermore, studies using a (M)TAU group as the control group (SMD .85) showed a higher mean effect on anxiety than studies using an education/support group as an active control group (SDM .07). Finally, medium quality studies (SMD .17) showed a lower mean effect on quality of life than the low (SMD 1.01) and high (2.33) quality studies.

Table 4. Subgroup analyses on outcome measures at post-treatment.

Outcome measure	Criteria	Subgroup	n	SMD [95% CI]	Test for subgroup differences
Pain	Quality	Low	5	.27(−.06, .60)	χ^2 = 1.33, df = 2 (p = .52), I^2 = 0%
		Medium	12	.13 (−.06, .32)
		High	5	.39 (−.08, .85)
	Intervention	ACT	9	.38 (.00, .76)	χ^2 = 1.23, df = 1 (p = .27), I^2 = 18.8%
		Mindfulness	13	.15 (−.01, .30)
	Control group	Education/support	8	.13 (−.08, .34)	χ^2 = 1.11, df = 1 (p = .29), I^2 = 9.8%
		(M)TAU/waitlist	14	.31 (.06, .57)
	Pain type	Chronic pain	10	.19 (.01, .38)	χ^2 = .15, df = 3 (p = .98), I^2 = 0%
		Fibromyalgia	5	.23 (−.34, .80)
		Specific-site pain	6	.27 (−.06, .59)
		Rheumatoid arthritis	1	.22 (−.26, .69)
	Attrition rate	<25%	11	.38 (.12, .63)	χ^2 = 3.16, df = 1 (p = .08), I^2 = 68.3%
		>25%	11	.08 (−.14, .29)
Depression	Quality	Low quality	1	.60 (−.49, 1.70)	χ^2 = .34, df = 2 (p = .84), I^2 = 0%
		Medium quality	9	.34 (.16, .51)
		High quality	6	.45 (−.13, 1.02)
	Intervention	ACT	7	.82 (.30, 1.33)	χ^2 = 5.36, df = 1 (p = .02), I^2 = 81.3%
		Mindfulness	9	.18 (.03, .34)
	Control group	Education/support	5	.21 (.00, .41)	χ^2 = 2.66, df = 1 (p = .10), I^2 = 62.4%
		(M)TAU/waitlist	11	.56 (.19, .93)
	Pain type	Chronic pain	7	.35 (.15, .55)	χ^2 = 6.56, df = 3 (p = .09), I^2 = 54.3%
		Fibromyalgia	6	.54 (−.02, 1.11)
		Specific-site pain	1	1.09 (.13, 2.05)
		Rheumatoid arthritis	2	.00 (−.34, .34)
	Attrition rate	<25%	10	.46 (.07, .84)	χ^2 = .20, df = 1 (p = .66), I^2 = 0%
		>25%	6	.36 (.13, .58)
Anxiety	Quality	Low quality	1	1.40 (.17, 2.62)	χ^2 = 2.74, df = 2 (p = .25), I^2 = 27.1%
		Medium quality	3	.33 (−.02, .67)
		High quality	2	.47 (−.48, 1.42)
	Intervention	ACT	5	.64 (.24, 1.05)	χ^2 = 5.09, df = 1 (p = .02), I^2 = 80.4%
		Mindfulness	1	−.01 (−.40, .38)
	Control group	Education/support	2	.07 (−.23, .37)	χ^2 = 12.42, df = 1 (p = .0004), I^2 = 91.9%
		(M)TAU/waitlist	4	.85 (.53, 1.16)
	Pain type	Chronic pain	3	.25 (−.25, .76)	χ^2 = 2.76, df = 2 (p = .25), I^2 = 27.7%
		Fibromyalgia	2	.81 (.39, 1.23)
		Specific-site pain	1	.62 (−.29, 1.53)
		Rheumatoid arthritis	0
	Attrition rate	<25%	5	.43 (.01, .84)	χ^2 = 2.16, df = 1 (p = .14), I^2 = 53.82%
		>25%	1	1.40 (.17, 2.62)
Disability	Quality	Low quality	2	.41 (.05, .77)	χ^2 = 1.81, df = 2 (p = .41), I^2 = 0%
		Medium quality	7	.57 (.00, 1.14)
		High quality	1	.35 (−.04, .75)
	Intervention	ACT	5	.75 (−.10, 1.61)	χ^2 = 1.42, df = 1 (p = .23), I^2 = 29.7%
		Mindfulness	5	.21 (−.05, .47)
	Control group	Education/support	2	.34 (−.01, .68)	χ^2 = .12, df = 1 (p = .73), I^2 = 0%
		(M)TAU/waitlist	8	.45 (−.09, .98)
	Pain type	Chronic pain	4	.22 (−.13, .56)	χ^2 = 3.65, df = 2 (p = .16), I^2 = 45.2%
		Fibromyalgia	1	−.19 (−.85, . 47)
		Specific-site pain	5	.81 (.03, 1.59)
		Rheumatoid arthritis	0
	Attrition rate	<25%	7	.36 (.13, .58)	χ^2 = .14, df = 1 (p = .70), I^2 = 0%
		>25%	4	.59 (.60, 1.78)
Quality of life	Quality	Low quality	1	1.01 (−.14, 2.16)	χ^2 = 61.50, df = 2 (p < .00001), I^2 = 96.7%
		Medium quality	9	.17 (−.02, .37)
		High quality	1	2.30 (1.80, 2.80)
	Intervention	ACT	6	.77 (−.13, 1.67)	χ^2 = 1.74, df = 1 (p = .19), I^2 = 42.4%
		Mindfulness	5	.14 (−.10, .39)
	Control group	Education/support	2	.08 (−.25, .41)	χ^2 = 1.57, df = 1 (p = .21), I^2 = 36.2%
		(M)TAU/waitlist	9	.53 (−.09, 1.16)
	Pain type	Chronic pain	4	.03 (−.46, .52)	χ^2 = 2.13, df = 2 (p = .35), I^2 = 5.9%
		Fibromyalgia	5	.67 (−.19, 1.52)
		Specific-site pain	2	.52 (−.28, 1.32)
		Rheumatoid arthritis	0
	Attrition rate	<25%	7	.60 (−.09, 1.28)	χ^2 = 1.35, df = 1 (p = .25), I^2 = 26.0%
		>25%	4	.11 (−.32, .54)
Pain interference	Quality	Low quality	0		χ^2 = .03, df = 1 (p = .86), I^2 = 0%
		Medium quality	2	.64 (.23, 1.06)
		High quality	2	.56 (.28, 1.40)
	Intervention	ACT	2	.64 (−.23, 1.06)	χ^2 = .03, df = 1 (p = .86), I^2 = 0%
		Mindfulness	2	.56 (−.28, 1.40)
	Control group	Education/support	2	.76 (−.34, 1.18)	χ^2 = .34, df = 1 (p = .56), I^2 = 0%
		(M)TAU/waitlist	2	.48 [−.37, 1.32)
	Pain type	Chronic pain	3	.56 (.09, 1.03)	χ^2 = .72, df = 1 (p = .40), I^2 = 0%
		Fibromyalgia	0
		Specific-site pain	1	1.02 (.07, 1.96)
		Rheumatoid arthritis	0
	Attrition rate	<25%	2	.64 (.23, 1.06)	χ^2 = .03, df = 1 (p = .86), I^2 = 0%
		>25%	2	.56 (−.28, 1.40)

Notes: SMD, standardized mean difference; 95% CI, 95% confidence interval; ACT, acceptance and commitment therapy; (M)TAU, (medical) treatment as usual.

Comparison with active treatments

In two of the 25 studies, a comparison was made with both a waitlist or education/support group and an active intervention such as CBT (Zautra et al., 2008) or relaxation (Schmidt et al., 2011). Three additional studies made only a comparison with an active intervention like CBT (Wetherell et al., 2011), MDT (Wicksell et al., 2009) or relaxation (Thorsell et al., 2011). The results of these studies were pooled in a meta-analysis to compare the effects of acceptance- and mindfulness-based interventions with the effects of other active treatments. Pooled SMDs and the results of the tests for heterogeneity and overall effect are presented in Table 5. Small effects were found on depression, pain interference and disability, in favour of CBT compared to acceptance- and mindfulness-based interventions. Moderate to large effects were found on pain, depression, pain interference, disability and quality of life in favour of acceptance- and mindfulness-based interventions compared to MDT/relaxation. None of these differences in pooled SMDs were significantly different.

Table 5. Comparison of acceptance-based treatments with CBT and MDT/relaxation at post-treatment.

Outcome measures	n	Studies	Heterogeneity	Pooled SMD (95% CI)	Test for overall effect
Comparison with CBT
Pain	2	Wetherell et al., 2011; Zautra et al., 2008	χ^2 = 6.04, df = 1 (p = .01); I^2 = 83%	−.02 (−.74, .70)	Z = .05 (p = .96)
Depression	2	Wetherell et al., 2011; Zautra et al., 2008	χ^2 = 1.29, df = 1 (p = .26); I^2 = 22%	−.25 (−.58, .08)	Z = 1.48 (p = .14)
Anxiety	0
Pain interference	1	Wetherell et al., 2011	n.a.	−.23 (−.60, .14)	Z = 1.24 (p = .21)
Disability	1	[109]	n.a.	−.22 (−.58, .15)	Z = 1.15 (p = .25)
Quality of life	0
Comparison with MDT/relaxation
Pain	3	Schmidt et al., 2011; Thorsell et al., 2011; Wicksell et al., 2009	χ^2 = 27.09, df = 2 (p < .00001); I^2 = 93%	.94 (−.16, 2.04)	Z = 1.68 (p = .09)
Depression	3	Schmidt et al., 2011; Thorsell et al., 2011; Wicksell et al., 2009	χ^2 = 51.94, df = 2 (p < .00001); I^2 = 96%	.83 (−.70, 2.37)	Z = 1.06 (p = .29)
Anxiety	1	Thorsell et al., 2011	n.a.	.00 (−.42, .42)	Z = .00 (p = 1.00)
Pain interference	1	Wicksell et al., 2009	n.a.	.69 (−.03, 1.41)	Z = 1.89 (p = .06)
Disability	2	Thorsell et al., 2011; Wicksell et al., 2009	χ^2 = 49.90, df = 1 (p < .00001); I^2 = 98%	2.54 (−1.20, 6.29)	Z = 1.33 (p = .18)
Quality of life	2	Schmidt et al., 2011; Thorsell et al., 2011	χ^2 = 63.08, df = 1 (p < .00001); I^2 = 98%	1.56 (−1.31, 4.43)	Z = 1.07 (p = .29)

Note: SMD, standardized mean difference; 95% CI, 95% confidence interval; df, degrees of freedom; n.a., not applicable.

Discussion

Main findings

The increased number of studies on the effectiveness of acceptance- and mindfulness-based interventions for people with chronic pain enabled a more robust assessment of their effects on the physical and mental health of patients. Twenty-five randomized controlled trials (RCTs) were included in this meta-analysis. At post treatment, small effects were found for pain intensity, depression, disability and quality of life; and moderate effects for anxiety and pain-interference. These results confirm the findings of the former meta-analysis (Veehof et al., 2011). The larger effect size for pain interference is congruent with acceptance- and mindfulness-based treatment models, mainly because pain interference, as opposed to pain intensity, is a more proximate indicator of the aim of acceptance- and mindfulness-based interventions, i.e. for patients to go along with life even in the face of pain. Disabilities caused by chronic pain as experienced by patients also depend on the degree to which the patients allow pain sensations to interfere with their daily life. Increased acceptance of pain sensations instead of fighting them, may be related to a larger decrease of pain interference. Only four studies, however, included pain interference as an outcome measure. To ensure conformity, comparability and especially, consistency with theory, we advise for future acceptance- and mindfulness-based intervention trials to apply pain interference or pain disabilities as the primary outcome.

At two to six months after treatment, the effect sizes for depression, anxiety and quality of life were moderate. The effect size for pain interference was large. These (in general) larger effects at follow-up indicate that patients, even in the presence of enduring pain, in the long term manage to apply the principles of acceptance- and mindfulness-based treatments in their daily lives.

The effect sizes for pain intensity were small, both at post-treatment and follow-up. This confirms the findings of a recent systematic review of mindfulness-based interventions and ACT in mixed samples (Reiner, Tibi, & Lipsitz, 2013). Given the chronic nature of pain, it is doubtful whether psychological treatments can achieve larger effects on pain intensity. Pain control and reduction are not the primary aims of acceptance- and mindfulness-based interventions, but some impact on pain intensity may be reached because mindfulness skills have been found to influence brain mechanisms that may alter the pain experience (Zeidan et al., 2011). In addition, acceptance- and mindfulness-based interventions can have indirect effects on pain intensity since increased acceptance may buffer the degree to which pain sensations are experienced as stressful events to be immediately avoided (Shapiro, Carlson, Astin, & Freedman, 2006).

Subgroup analyses showed that the effect sizes of ACT were higher in comparison to MBSR and MBCT for all outcome measures and significantly higher for depression and anxiety. The difference on depression remained after excluding an ACT study that found large effects on depression and anxiety (Luciano et al., 2014) . The explicit formulation of life values and the emphasis on committed action of the ACT-model may install hope and increase behavioral change and thereby explain the larger effects of ACT when compared to MBSR and MBCT. The smaller effects of MBSR and MBCT in this meta-analysis corroborate with a recent meta-analysis of the effects of mindfulness on anxiety and depression (Hofmann, Sawyer, Witt, & Oh, 2010). For fibromyalgia, none or small effects of MBSR or MBCT were found in (randomized) controlled studies (Lakhan & Schofield, 2013; Lauche, Cramer, Dobos, Langhorst, & Schmidt, 2013).

Except for quality of life, the quality of the studies was not found to moderate outcomes. For quality of life, studies of medium quality were found to have significant lower effect. However, this result needs to be interpreted with caution since the number of high- and low-quality studies were only one study each.

Overall, the effect sizes found in this study were at least comparable to those found in a recent meta-analysis of CBT for chronic pain patients (Williams et al., 2012). Williams and co-workers found small post treatment effect sizes on pain intensity, disability and mood in their meta-analysis of sixteen RCTs, in which CBT was compared to TAU or a waiting list. However, only the effect size of mood was maintained at follow-up (Williams et al., 2012). This finding is in contrast to those of the current meta-analysis which showed that the effects on all outcomes, except pain intensity, increased at follow-up.

The subgroup analysis on control group showed that (M)TAU and waitlist did not differ from the active control conditions of support groups or education, apart from a significant difference in favour of (M)TAU on anxiety. Pain education as developed by Moseley and his colleagues has been shown to have beneficial effects as a treatment on its own (Moseley & Butler, 2015). This pain education program, however, differs fundamentally form the pain management education used in most of the studies in this meta-analysis, by the important role for explanation of the biological basis of pain. It thereby offers alternative and more helpful conceptualisations of pain and of inadequate (catastrophic) cognitions patients may be fused with. Five studies compared acceptance- and mindfulness-based interventions to active treatments such as CBT (Wetherell et al., 2011; Zautra et al., 2008), relaxation (Schmidt et al., 2011; Thorsell et al., 2011) or MDT (Wicksell et al., 2009). Small effects were found in favour of CBT, and moderate to large effects were found in favour of acceptance- and mindfulness-based treatments compared to relaxation or MDT. However, the number of studies is small and none of differences were found to be significant in the pooled analyses.

Study limitations and strengths

The number of studies on acceptance- and mindfulness-based interventions has increased considerably since our former systematic review and meta-analysis. Therefore, we were able to restrict ourselves to the more rigorous test of RCTs. However, this study also contains a number of limitations. First, the average quality of studies has not improved over the past six years. Although the number of high quality studies has increased compared to our previous search up to 2009, a number of studies of low quality have still been recently published. The rating of quality may be an underestimation since we rated a criterion as not-fulfilled if it was not reported in the article. It is also plausible that mean quality would be enhanced if all researchers would routinely report on their randomization procedures and the training of therapists. Second, we combined studies of ACT with mindfulness-based studies. This allows for a comparison of the recent results with the results of the former systematic review and meta-analysis. However, the ACT and mindfulness based treatments can only be compared at post-treatment, since the number of studies were too small to perform subgroup analyses at follow-up. As studies continue to increase, future meta-analyses may differentiate between acceptance- and mindfulness-based interventions to understand their unique effects. Third, a small number of studies were represented in some of the subgroup analyses.

The data extraction of this study revealed a great number of qualitative, narrative or systematic reviews. By now, the field is saturated with reviews. The meta-analytic approach of this study and the reported effect sizes not only offer insight into the strength of effects, but can also be used to benchmark pre- to post-treatment as well as follow-up measurements in daily practice (Morley, 2011).

Conclusions and recommendations

To conclude, acceptance- and mindfulness-based interventions for chronic pain are, on the whole, moderately effective on a number of beneficial outcomes, especially in the long term. Nevertheless, there is room for progress and treatment improvement.

It is necessary to study whether interventions work in accordance to their underlying theoretical basis. In CBT, the main hypothesized working mechanism is the changing of the content of thoughts such as catastrophizing cognitions and of emotions such as kinesiophobia (Crombez, Eccleston, Van Damme, Vlaeyen, & Karoly, 2012). In acceptance- and mindfulness-based treatments, mindfulness and psychological flexibility are hypothesized to be the main process of change (McCracken & Morley, 2014). In a MBSR study among adults with problematic stress resulting from chronic pain, chronic disease or other life circumstances, increases in mindfulness preceded decreases in perceived stress (Baer et al., 2012). A meta-analysis of samples with a mix of mental and physical health problems, showed that ACT seemed to work through its proposed mechanism of change while CBT did not (Ruiz, 2012). In a mediation analysis within a sample of patients with chronic whiplash-associated disorders, psychological inflexibility was identified as a mediator on pain-related disability and life satisfaction, but kinesiophobia was not (Wicksell et al., 2010). The mediating role of psychological inflexibility was also found in individuals with pain that finished an online ACT program, but there were indications that catastrophizing, a putative mediating variable in a CBT model, was also an independent mediator on pain interference (Trompetter, Bohlmeijer, Fox, & Schreurs, 2015). Therefore, it is recommended to include variables from the CBT, mindfulness, as well as from the psychological flexibility model in future studies.

Knowledge on mechanisms of change can help to enhance or integrate our existing theoretical models of chronic pain, thereby progressing opportunities for more effective treatment. More than one measurement moment must be scheduled during the intervention phase to discern how processes unfold over time. But the scheduling of more intensive measurement is advisable to really grasp time-varying processes in relation to specific intervention characteristics and subsequent changes in outcome. We suggest that the application of intensive measurements once per day, or even more times per day by the use of mobile technology gives real insight into unfolding processes over time within individuals (Richardson & Reid, 2013). When scheduling the intensity and timing of measurements during interventions, however, we must not merely measure mechanisms just for the sake of measuring. Timing and spacing of measurements needs to be well thought out and be related to theory and intervention characteristics. It must be scheduled in such a way that one is most certain to catch unfolding of processes one is interested in (Selig & Preacher, 2009).

Furthermore, outcomes could also be improved if treatments are matched to the characteristics of patients. In general, sample sizes of current RCTs are too small to answer questions on what works for whom; therefore, studies should be designed to investigate specific hypotheses.

Finally, clinicians and researchers should invest in treatment integrity. In this meta-analysis, only seven studies performed a treatment integrity check. Clearly, this is a challenging criterion that is still too liberally applied in this and other meta-analyses since the mere report of an integrity check already receives credit. If we want to improve the effects of interventions, it is absolutely necessary to ensure that interventions are provided as intended.

Overall, we can conclude that individuals with pain, in general, respond rather well to acceptance- and mindfulness-based interventions and that beneficial effects are retained after treatment. ACT seems to be more effective in treating depression and anxiety in individuals with pain than mindfulness-based treatments.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Information on the quality assessment per study is obtainable from the corresponding author.

Appendix 1. Search strategy Pubmed (OVID)

(“attention”[MeSH Terms] OR “attention”[All Fields]) OR (“mindfulness”[MeSH Terms] OR “mindfulness”[All Fields]) OR vipassana[All Fields] OR (“meditation”[MeSH Terms] OR “meditation”[All Fields]) OR MBSR[All Fields] OR MBCT[All Fields] OR “mindfulness-based stress reduction”[All Fields] OR “mindfulness-based cognitive therapy”[All Fields] OR “acceptance based”[All Fields] OR “acceptance-based”[All Fields] OR (acceptance[All Fields] AND (“Commitment”[Journal] OR “commitment”[All Fields]))

AND

((“chronic pain”[MeSH Terms] OR (“chronic”[All Fields] AND “pain”[All Fields]) OR “chronic pain”[All Fields]) OR (“fibromyalgia”[MeSH Terms] OR “fibromyalgia”[All Fields]) OR “chronic fatigue syndrome”[All Fields] OR “whiplash associated disorder”[All Fields] OR WAD[All Fields] OR “repetitive strain injury”[All Fields] OR RSI[All Fields] OR dystrophy[All Fields] OR (complaints[All Fields] AND (“arm”[MeSH Terms] OR “arm”[All Fields]) AND (“neck”[MeSH Terms] OR “neck”[All Fields]) AND (“shoulder”[MeSH Terms] OR “shoulder”[All Fields])) OR CANS[All Fields] OR “complex regional pain syndrome”[All Fields] OR CRPS[All Fields])

AND

(hasabstract[text]

AND

(“2009/01/01”[PDAT] : “2013/12/31”[PDAT])

AND

English[lang])