Background

Chronic pain can contribute to disability, depression and anxiety, sleep disturbances, poor quality of life, and increased health care costs, with close to 20 % of the adult population in Europe reporting chronic pain [1].

Chronic pain is recognised as a long-term condition in its own right or a major comorbidity of other long-term conditions. An aim of the management of all long-term conditions, embodied in the idea of self-management, is that the patient should be an active participant in the management of their own condition. In chronic pain management, such thinking is based on the assertion that “self-care and management underpins all activities in the care pathway and should be considered [at all stages]… pain management is most effective when it engages the patient in self-management” [2]. To empower the person to self-manage, it is advocated that education and training is provided about the nature of pain and its effects and how to live with pain. Responding to recommendations to establish what educational interventions in pain management work best and for whom [3] is complex not least because of the many different methods and combinations of methods of education and training that are available and in use.

The intervention being investigated in this study is education of the patient to facilitate their knowledge of chronic pain that does not include behaviour modification or training in skills of pain management.

Whilst the use of other non-pharmacological interventions can generally be informed by good quality systematic reviews of the literature (e.g. cognitive behavioural and behavioural therapy [4–6], TENS [7] and low-intensity movement therapy [6, 8]), that level of evidence is less readily available to inform the use of education to facilitate knowledge about chronic pain in adults.

Therefore, a systematic review of available studies that have evaluated the effectiveness of education to facilitate knowledge about chronic pain in adults was conducted to assess the current situation and inform future research and guidelines.

Review objectives

The primary objective was to determine the current level of evidence of the effect of education to facilitate knowledge about chronic pain for adults on pain and disability. A secondary objective was to determine (from the selected papers) the evidence of effect of the educational interventions on psychosocial outcomes.

Methods

Inclusion/exclusion criteria

Study type

Suitable for inclusion were randomised controlled trials (RCTs) and cluster-RCTs published and pre-published (electronically) in peer-reviewed journals; studies were accepted as randomised when described by the author as such. Studies were included if they were published in English. Studies were excluded if they were only available in abstract form.

Type of participants

Studies of adults (aged 18 years or older) reporting pain for at least 3 months (12 weeks) at any body-site(s) were included. Mixed age samples were included if data could be separated (adult/child). We excluded studies that were focused on specific diagnoses (e.g. osteoarthritis) and conditions where chronic pain is not necessarily the main symptom (e.g. irritable bowel syndrome), as the advice or guidance was likely to be based on the specific impact and management of the condition (e.g. joint stiffness, bowel dysfunction) at least as much as on the management of pain. We also excluded cancer-related pain.

We only included studies that focused on chronic pain, which is consistent with a strategy used in previous national guidelines [6].

Interventions

Studies were included where the effectiveness of education to facilitate knowledge about chronic pain could be assessed as a stand-alone intervention compared with usual care, or where different types of education could be compared with each other. Therefore, we excluded investigations of multi-disciplinary programmes (such as pain management programmes) which did not allow assessment of the effectiveness of the education component.

Outcome measures

Studies were included if they reported outcomes of pain severity and/or physical function.

Data collection and analysis

Electronic searches

We searched EBSCOhost [MEDLINE, CINAHL Plus, OmniFile (Full text), eBook Collection] and CENTRAL [Cochrane Central Register of Controlled Trials; MEDLINE, EMBASE, Cochrane Review Groups Specialist Registers] databases from inception to 31 December 2013. No language restrictions were imposed whilst searching; English language criterion was applied later in the process. The search strategy was developed for use in MEDLINE and adapted for each database (Additional file 1).

Other resources

Reference lists of reviews and the 22 articles reviewed in detail by all of the authors (see below) were checked for additional studies, and citation searches were performed on key articles to minimise publication bias.

Ongoing trials were searched using the World Health Organization International Clinical Trials Registry Platform (ICTRP) (http://apps.who.int/trialsearch/) and the metaRegister of Controlled Trials (mRCT) (http://www.controlledtrials.com/), which includes the ISRCTN Register (international), Action Medical Research (UK), NIH ClinicalTrials.gov Register (international), the Wellcome Trust (UK) and UK trials (UK).

Selection of studies

Following initial scanning of titles, abstracts and full papers by one author, two authors read the remaining full papers and assessed them separately. Authors independently selected studies that met the inclusion criteria using a purpose-designed checklist, limiting inclusion to studies that were randomised (or cluster-randomised) as a minimum. Further discussion was required when the authors could not reach a consensus on the studies to be included. Authors with psychological expertise reviewed the studies to remove those deemed to have formally implemented an underlying psychological therapy as part of, or including, education to facilitate knowledge about chronic pain. In this manner, 22 studies were included after screening and were then assessed by all authors. Finally, those studies that fulfilled all of the inclusion criteria were selected for analysis, a total of nine.

Data extraction and management

Data were extracted using a standardised form which included information regarding study design, participants, trial characteristics, intervention, comparison (control) and outcomes. Data were collected manually on paper extraction forms and entered into intermediate software (Microsoft Excel for Windows) before being entered in to RevMan 5.3 [9]. This intermediary stage allowed for any necessary statistical conversions. Only one pain measure was selected per study. When there were multiple measures of pain in a study, we used only the measure of average pain intensity. When there was more than one report of average pain intensity, then the mean of these was calculated. Results from visual analogues scores (VAS) were prioritised over measures such as the McGill Pain Questionnaire (MPQ) if both were reported.

Risk of bias assessment

Two authors independently assessed risk of bias for each study. Arbitration by a third author was not necessary as inconsistencies were resolved through discussion. The domain-based evaluation presented in the Cochrane Handbook for Systematic Reviews of Interventions (Chapter 8, version 5.1.0 [10]) was used to assess risk of bias. We assessed the following for each study: random sequence generation (checking for possible selection bias), allocation concealment (selection bias), blinding of patients and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias due to the amount, nature and handling of incomplete outcome data) and selective reporting (reporting bias). For “other” sources of bias, we included criteria to evaluate study sample size, where fewer than 50 participants per treatment arm was considered an increased risk of bias.

Risk of bias across all included studies was categorised according to the following:

  • Low risk of bias (✓) – plausible bias unlikely to seriously alter the results if most information was obtained from studies at low risk of bias

  • Unclear risk of bias (?) – plausible bias that raised some doubt about the results if most information was obtained from studies at low or unclear risk of bias

  • High risk of bias (X) – plausible bias that seriously weakens confidence in the results if the proportion of information was obtained from studies at high risk of bias sufficient to affect interpretation of results

Measurement of treatment effect

Data from included studies were reviewed separately and then, where possible, combined quantitatively by population, intervention, comparison and outcomes. Continuous data were expressed as mean difference (MD) or standardised mean difference (SMD) with 95 % confidence intervals (95 % CI); dichotomous data were reported by just one study [11] for a single outcome measure (session attendance), and this finding has been reported in the results of this review as text only.

Assessment of heterogeneity

We assessed heterogeneity according to the standard method using the Chi [2] test and the I [2] statistic, calculated for each comparison on each outcome. I [2] values above 50 % suggest high heterogeneity, 25–50 % medium heterogeneity and below 25 % low heterogeneity, though this is only used as a guide.

A standard random effects analysis was used to avoid over-weighting large studies and potentially losing small study effects.

Assessment of reporting bias

We intended to use funnel plots to assess small-study effects, following the guidance of the Cochrane Handbook of Systematic Reviews for Interventions (Chapter 10) [10], but studies were insufficient in number (n = 9) to undertake this effectively.

Data synthesis

Data were entered into RevMan 5.3 [9] by one author and checked by a second. Data were largely presented as mean and standard deviation (SD); though one study [11] reported results as mean and standard error (SE), these data were converted to mean and standard deviation using the RevMan calculator, and results were checked by hand. One study reported only average (mean) results with no measure of variation [12], and we were unable to extract plausible data for inclusion in the meta-analyses or other form of data presentation within this review.

Analysis

Studies with a comparator group of usual care were analysed separately from the studies that compared different types of education. The issues with aging and pain are more complex than simply pain being a direct correlate of biological age, though there does appear to be justification for considering pain in older people as a distinct issue [13–16]. Therefore, where data on older people (>65 years) could be extracted, these were analysed separately. Sensitivity analysis was also planned to determine if the effect size was affected by the methodological quality of the study (risk of bias). However, these analyses were not possible due to the small number of included studies.

Using the same approach as above, we also carried out post-hoc analyses of the following psychosocial variables which were reported in these studies: catastrophising, mood, knowledge of chronic pain, self-efficacy, global health and social function.

Results

Search results

Results of the search are shown in Fig. 1. Summary information of the studies included in the review is shown in Table 1.

Fig. 1
figure 1

PRISMA [43] flow chart demonstrating database searches, identification, screening and selection of included studies

Full size image
Table 1 Characteristics of Included studies
Full size table

Searches for ongoing trials revealed 148 trials, of which twelve titles fulfilled the inclusion criteria. From these, six were excluded after further examination of the abstract/trial registration detail or due to duplication with previously located studies. The remaining six studies have potential to be included in future updates but are currently unavailable to this review (Additional file 2).

Description of studies

Nine studies were included (Ferrell et al. 1997 [17], Gallagher et al. 2013 [18]; Linton et al. 1997 [11], Morrison et al. 1988 [12], Moseley et al. 2004 [19], Ruehlman et al. 2012 [20], Soares et al. 2002 [21], Sparkes et al. 2012 [22] and van Oosterwijck et al. 2013 [23]). In each study, all of the participants reported pain for at least 3 months. Two studies were conducted on people with fibromyalgia [21, 23], three were on people with chronic pain or chronic musculoskeletal pain [17–19], and four studies specifically examined back pain [11, 12, 19, 22]. Individual study sample size ranged from 20 [17] to 305 [20]. Participants were over the age of 18 years in all studies, though one did not specify this in the criterion [19]. There was no upper age limit in five studies [12, 17, 19, 20, 22], one of which used a sample that was exclusively over 65 years old [17]. Others excluded individuals over the age of 60 [11], 65 [21, 23] and 75 years [18].

All studies except one [23] reported the process of recruiting, which was largely through a general practitioner (GP) or specialist referrals and pain programme waiting lists. Trials were conducted in Canada [12], USA [17, 20], Sweden [11, 21], Australia [18, 19], UK [22] and in both Belgium and the UK [23].

Education to facilitate knowledge about chronic pain was in the form of lectures [12, 17, 19, 23], individual or group discussions [11, 21], written text [18, 22] or website interaction [20]. These interventions took place during a single session [17, 19, 23], numerous interactions (multiple sessions of the same format [11, 12], individual and then group sessions [21]) or with no direct contact [18, 20, 22]. Education varied in focus from understanding the neurophysiology and biology of pain [18, 19, 23] to management of symptoms through accessing physical help, such as medication, hot/cold packs, ergonomics [17, 21] and a combination of these and other topics (e.g. anatomy, physiology, body mechanics, posture, pain relief and first aid techniques [11, 12, 20, 22]).

The comparator group was usual care in five studies [11, 12, 20–22], and the other four studies compared different methods of education [17–19, 23].

The nine studies included in this review reported at least one of the primary outcome measures (Table 4). All studies except one [19] reported an assessment of pain, although there was a variety of measures used for the assessment. Physical function or disability was measured objectively in two studies using validated performance tests [17] and standardised tests for the study [12]. It was also measured subjectively (patient-reported) in a further four studies using a validated disability questionnaire (Roland Morris Disability Questionnaire) [19, 22] and validated assessment of interference or impact on daily life [11, 20].

Excluded studies

Thirteen studies were excluded (Table 2). Three of these were excluded due to their multi-disciplinary intervention, where the effect of education alone could not be assessed [24–26]; five were excluded after assessment as having a psychological, rather than an educational content [27–31]; and three more were excluded due to inconsistency with the study design criteria [32–34].

Table 2 Excluded studies
Full size table

Risk of bias in included studies

Low or unclear/uncertain risk was identified across the majority of the six domains. Risk of bias for each included study is shown in Table 3.

Table 3 Risk of bias summary showing the review authors’ judgements about each methodological quality item for each included study
Full size table

Selection bias (random sequence generation and allocation concealment)

Four studies fulfilled both criteria for low risk of bias [28, 32, 33, 35], and one fulfilled one of the two criteria [36]. Three studies mentioned that the participants were randomised and allocation-concealed but did not specify the method constituting an unclear risk of bias [29, 31, 34]. One study described itself as randomised, but patients were “consecutively allocated” to each group and so held a high risk of bias [37].

Performance bias (blinding of participants and personnel)

All of the most recent studies [32–37] showed low risk of bias and reported blinding of participants and personnel where necessary. Earlier publications [28, 29, 31] did not mention blinding.

Detection bias (blinding of outcome assessment)

In five studies, there was no blinding of outcome assessments [31–34, 36], but the review authors judged that the outcome measure was unlikely to be influenced by this knowledge as questionnaires were completed alone by the participant. Two studies reported blinding for outcome measures [35, 37], only one of which reported assessing the success of blinding of both the participants and personnel [35]. The study by Morrison et al. [29] was labelled high risk of bias as each group was only assessed once (the control group at pre-intervention, and treatment group post-intervention only).

Attrition bias (incomplete outcome data)

Dropouts and withdrawals were noted and explained in all studies (low risk of bias) except one where there was no mention of incomplete data [29].

Reporting bias (selective reporting)

No published protocols were found, and so we cannot say with absolute certainty that all outcome measures were reported, and all included studies were therefore awarded an unclear/uncertain risk of bias.

Other potential sources of bias

Study size was assessed as an additional risk of bias, as a small study size could bias the results. The methods specified that fewer than 50 participants per treatment arm would be an increased risk of bias as seen in all but two studies (low risk of bias n = 79 after crossover from control [32], n = 162 [34]). The remaining seven studies were further separated into those with fewer than 20 participants in the treatment group as high risk of bias in three studies (n = 10 [31], n = 18 [37], n = 15 [35]), uncertain risk for those where n ~ 30 in the treatment group [28, 33, 36] and unclear risk for one study [29] that reported no separate treatment/control sample size.

Intervention effect

  1. 1.

    Average pain intensity (Table 4 post-intervention and Table 5 follow-up).

    Table 4 Pain and disability outcome measures - post-intervention
    Full size table
    Table 5 Pain and disability outcome measures - follow-up
    Full size table

Education versus usual care

Adults >18 years: All four studies measured pain at the post-intervention measurement point [11, 20–22], and two of these studies also reported follow-up assessments [20, 21].

None of the studies showed significant effects post-intervention. Pooling the data of all four studies showed low heterogeneity (I [2] = 0 %), and the effect size was small and statistically non-significant (Fig. 2). In neither of the two studies reporting results around 3 months after the end of the intervention was there a significant effect. Again, pooling of the data showed a small effect size that was not statistically significant (Fig. 3).

Fig. 2
figure 2

Forest plot showing pain intensity (education versus usual care)—post-intervention

Full size image
Fig. 3
figure 3

Forest plot showing pain intensity (education versus usual care)—follow-up (3 months)

Full size image

Adults >65 years: No specific data were available.

Comparison of different types of education

Adults >18 years: Measures of average pain could only be extracted from one study [23]. In this study, which used a very small sample size to compare pain neurophysiology education (PNE) with another form of information provision, there were no statistically significant differences in average pain intensity between the two forms of information provision 2 weeks after the intervention period or 3 months after the intervention ended.

Adults >65 years: One study exclusively investigated adults aged >65 years [17]. The study, which had a very small sample size, found no significant differences between information provision about pain and information provision about physical methods to reduce pain immediately after the intervention period or 6 weeks after completion.

  1. 2.

    Disability (Table 4 post-intervention and Table 5 follow-up).

Education versus usual care

Adults >18 years: Disability was assessed in three out of the four studies post-intervention [11, 20, 22]. Only one of the studies had a follow-up assessment, reporting results at 3 months from the end of the intervention [20].

There were no significant effects on disability in any of the studies immediately after the end of the intervention. When data were pooled, heterogeneity was high amongst these studies (I [2] = 49 %), and the overall effect size was low (Z = 0.11) and statistically non-significant (Fig. 4). The single study that assessed disability at 3 months following the end of the intervention showed no significant differences in disability between groups (Fig. 5).

Fig. 4
figure 4

Forest plot showing disability (education versus usual care)—post-intervention

Full size image
Fig. 5
figure 5

Forest plot showing disability (education versus usual care)—follow-up (3 months)

Full size image

Adults >65 years: No specific data were available.

Comparison of different types of education

Adults >18 years: Disability was assessed in two studies, both of which compared pain neurophysiological education (PNE) with other information provision types [19, 23]. Only one study contained a follow-up assessment at 3 months following the end of the intervention; this study used a very small sample size [23].

PNE showed a significantly better effect than its comparator on the Roland Morris Disability Questionnaire (RMDQ) scores immediately after the intervention in one study [19]. In the other study, which had a very small sample size [23], the mean difference in favour of PNE (5.3 points on the SF36 physical function subscale 2 weeks after the intervention had ended) was not statistically significant. Pooling the data from both studies showed low heterogeneity (I [2] = 0 %), and there was a statistically significant difference in favour of PNE (Fig. 6).

Fig. 6
figure 6

Forest plot showing disability and physical function (comparison of different types of education)—post-intervention

Full size image

In the study [23] that contained a follow-up assessment (3 months after the intervention had ended), there was a mean difference of 8.4 points on the SF36 physical function subscale (range 0–100) in favour of PNE, which was not statistically significant (Fig. 7).

Fig. 7
figure 7

Forest plot showing disability (comparison of different types of education)—follow-up (3 months)

Full size image

Adults >65 years: The one study exclusively investigating older adults (>65 years) used a very small sample size [17]. There were no significant differences between education about pain and education about physical methods to reduce pain either immediately after the intervention period or 6 weeks from its end.

  1. 3.

    Psychosocial outcomes (Table 6 post-intervention and Table 7 follow-up).

    Table 6 Psychosocial outcome measures - post-intervention
    Full size table
    Table 7 Psychosocial outcome measures - follow-up
    Full size table

Education versus usual care

Adults >18 years: Catastrophising was assessed in three of the four studies post-intervention [11, 20, 21]. Two had a follow-up assessment at 3 months from the end of the intervention [20, 21]. There was no effect, post-intervention, in any of the studies individually or when the data were pooled (Fig. 8), nor was there an effect at follow-up (Fig. 9). One study measured self-efficacy, only at post-intervention, and showed no effect [21]. Depression was measured in one study, and there were no changes post-intervention or at follow-up [20].

Fig. 8
figure 8

Forest plot showing catastrophising (education versus usual care)—post-intervention

Full size image
Fig. 9
figure 9

Forest plot showing catastrophising (education versus usual care)—follow-up (3 months)

Full size image
Fig. 10
figure 10

Forest plot showing catastrophising (comparison of different types of education)—post-intervention

Full size image
Fig. 11
figure 11

Forest plot showing catastrophising (comparison of different types of education)—follow-up (3 months)

Full size image
Fig. 12
figure 12

Forest plot showing knowledge of pain (comparison of different types of education)—post-intervention

Full size image
Fig. 13
figure 13

Forest plot showing knowledge of pain (comparison of different types of education)—follow-up (3 months)

Full size image

Adults >65 years: One study carried out a bespoke measure of participants’ knowledge about pain and reported a significant improvement, post-intervention and at follow-up, in favour of the intervention [17].

Comparison of different types of education

Adults >18 years: Catastrophising was assessed in each of the three studies post-intervention [18, 19, 23], with two providing follow-up data [18, 23]. In each study and in the pooled data, there was a positive effect in favour of PNE at both assessment times (Fig. 6a, b). Two of the three studies assessed knowledge of pain post-intervention and at follow-up [18, 23]. A positive effect in pooled data in favour of PNE reflected the positive effects in both studies post-intervention and at follow-up (Fig. 7a, b). Only one of the studies assessed other relevant outcomes—mood, global health and social function [23]. For mood, there were positive effects post-intervention in favour of PNE, which did not reach statistical significance (p > 0.05) but did so at follow-up. There were no significant effects in global health or social function at either measurement point.

Adults >65 years: No specific data were available.

Discussion

We systematically reviewed RCTs that investigated the effects of education to facilitate knowledge of chronic pain in adults on pain intensity and disability. Our analysis of the nine studies that fit the inclusion criteria found no evidence of an effect on pain intensity. However, for disability, there was evidence of a significant improvement immediately following a course of a particular type of education—pain neurophysiology education (PNE). Such an effect was not seen for the other types of education investigated in the studies.

Only one study specifically looked at people over 65 years old, also showing no significant effect on pain or disability.

Other reviews have been published in the past 5 years examining education for cancer pain [37], PNE for chronic musculoskeletal pain [36], PNE for chronic low back pain [35], education for neck pain [38, 39], educational interventions by pharmacists for chronic pain [40] and knowledge translation for chronic non-cancer pain management [41]. This last review included interventions aimed at health professionals, patients and a combination of target groups. Of these reviews, only three were able to combine studies to perform some meta-analyses within their reviews [35, 38, 40], whilst others reported results in the narrative.

Pain severity/intensity

As in the current review, educational interventions had no significant impact on pain severity or intensity in whiplash-associated disorders (neck pain [38]) and no clinical significance in chronic low back pain [35], though it was shown to be significantly effective in reviews of education in cancer pain [37] and chronic musculoskeletal pain [38]. The variation in results with regards to the change (or lack thereof) in pain intensity may largely be due to the nature of the patient population (cancer patients [37]) or the intervention itself (multi-disciplinary approach combining education with physiotherapy or cognition-targeted motor control training [38]).

Disability/physical function

Other reviews [35, 38] have not revealed significant change in levels of disability and function, consistent with the present review.

Conversely, Louw et al. [38] described a significant effect from education in those with musculoskeletal pain. Included in that review [38] was the one study in the present review that showed a significant improvement in disability as a result of the intervention [19], and others by the same research team (five out of eight trials), potentially skewing the results of the review to reflect this one intervention. The review of knowledge translations targeting patients showed short-term improvements in patient function with chronic low back pain, but no change in migraine-related complaints [41], suggesting as we have in the inclusion and exclusion criteria of the present review that underlying conditions (such as migraine) should be treated and analysed separately to other manifestations of chronic pain.

Psychosocial outcomes

The most interesting findings were the significant improvements in catastrophising and knowledge of pain. The improvements in catastrophising were only found in those studies that utilised pain neurophysiological education (PNE) in the intervention. This fits with one of the primary aims of PNE, to reconceptualise thinking about pain, away from the belief that “hurt” always equates to “physical harm”. The change in knowledge about pain, which were also seen with PNE, also point towards achievement of this primary aim. However, the design of the studies and reliance on questionnaire findings do not allow the depth of investigation needed to fully explore this suggestion, and appropriate qualitative investigation is called for. Interestingly, knowledge of pain was found to increase in the single study on older people exclusively [17] (not using PNE). This suggests that such an aim (to reconceptualise thinking) is not limited by older age. However, the measurement used in the study was very superficial, and again, the most appropriate action would be to explore this in more depth.

This review was conducted using the most robust techniques available. Electronic searches included full access to four databases (MEDLINE, CINAHL Plus, EMBASE and CENTRAL [Cochrane Central Register of Controlled Trials]), alongside international trial registries and author personal libraries. In total, 8519 titles were assessed for inclusion (8371 of published papers, 148 from ongoing trial registry). The review summarises the highest quality evidence available using RCTs of reasonable quality. The use of meta-analytical methods to pool data from different studies, which had relatively small individual sample sizes, maximised the strength of the findings.

The search was undertaken from database inception until 31 December 2013, and all data were extracted and analysed within 6 months of this date. The resources available prevent us from updating and re-analysing the search, and we are unable to examine the effect that any subsequent studies may have on our findings. This is an area for ongoing research, which will be supported by our included list of ongoing trials noted at the time of our analysis.

The review was limited by the small number of studies suitable for analysis, reflecting the availability of relevant published studies. This meant that findings were based on relatively low sample sizes, although this was overcome to an extent by pooling of data where appropriate. Because of the small numbers available, we were not able to carry out subgroup analyses to assess the influence of study quality on outcomes nor were we able to comprehensively assess the influence of older age.

The scope of the review was deliberately restricted to investigate education in isolation from other interventions, and care should therefore be taken when extrapolating the findings to the use of education delivered in combination with other interventions. However, there is room for future study of the additive effects of education in combination with other approaches to pain management.

We also used outcomes of pain intensity and disability as inclusion criteria. Therefore, in our post-hoc analysis of psychosocial outcomes, it is likely that we have excluded evidence from studies that used such measures but did not measure pain or disability; other sources of evidence should be used to make judgement on the effect of education on other outcomes such as mood, coping strategies and pain beliefs, all of which can be important in pain management.

We did not contact authors for further information and excluded papers that were not available in English at the full-text stage only. Both of these decisions were largely made as a result of the resources available to us. However, the need to contact authors only occurred in the case of a single paper [12], where details were lacking regarding group sample sizes, and no variation around the mean was reported. We decided we were unlikely to receive a response due to the considerable time period since publication (1988), and as a result, the paper was excluded from the meta-analyses. Only one paper [42] was excluded due to the language (Fig. 1), and we are unable to judge the effect of this exclusion on our results.

As highlighted by the small number of studies and the diverse range of educational methods that have been used in the current literature base, there is a general need for more high quality trials in this area. The specific findings for PNE, in this review and in others, should stimulate research to see how its promise can be optimised to further improve its effects, perhaps by comparing different methods of delivery and tailoring its content to specific populations including older people.

One study [28] compared two different modes of delivery (website versus written material), though it was excluded due to the large cognitive behavioural therapy (CBT) component delivered through the website. Equivalent research examining mode of delivery would be of interest to further examine whether online delivery of an educational intervention, for example, is equally effective across all age groups or whether it is the group element compared to individual learning that has the greatest influence on effect size.

The wide variety of assessments available for this review meant that a great deal of analysis was not possible due to the variability of the focus of the assessments or when subscales were not reported. Future meta-analyses could be improved through the standardisation of outcome measures.

Further, the effect of education on psychosocial variables as mediators of pain and disability remains to be elucidated. Research to investigate the effect of education upon knowledge and its relationship to psychosocial mediators is warranted.

Finally, a long-term follow-up should be implemented for all studies, as short-term results whilst promising, do not necessarily suggest long-term effect.

Conclusions

Of the different forms of education reported in RCTs, only PNE appears to be effective (by reducing disability) as a sole intervention for adults with chronic pain and only immediately after the intervention. However, the evidence is too limited to conclusively rule out other options.

Practical implications

Education to facilitate knowledge of chronic pain in adults remains a potentially important part of patient activation. Certainly, for people with established chronic pain (as represented by the studies reviewed here), it would be sensible to include education along with other interventions as there is little evidence to support education as a stand-alone intervention. Whilst research findings continue to emerge, clinicians should consider incorporating PNE, though it would be premature to discard other options. However, we cannot confidently conclude that education alone is effective in reducing pain intensity or related disability in chronic pain in adults.