Background

While chronic pain may be a primary complaint or secondary to an underlying disease, it is now recognised as a health condition in its own right [1] requiring specific therapy and rehabilitation [2]. In the UK, chronic pain affects between a third and a half of the population and about 10–14% of people report moderate to severely disabling chronic pain [3]. For people affected and their families, chronic pain is associated with a reduced quality of life and impacts on work and social life [4,5,6]. About two thirds of people with chronic pain report problems sleeping [7], including difficulties falling asleep, staying asleep, or waking early [6, 8,9,10], and this is evident across a range of conditions associated with chronic pain. In a large US population, 89% of people with chronic pain caused by fibromyalgia reported one, and 63% reported two or more symptoms of sleep disturbance [11]. Sleep disorders are common in people with multiple sclerosis [12, 13], rheumatoid arthritis [14], and osteoarthritis [15,16,17], with about 60–75% of people affected. Sleep disturbance is greater in people with more severe osteoarthritis symptoms [15, 18]. Other pain conditions with associated sleep disturbance include migraine and frequent headache [19,20,21], and low back [8, 22] and neck pain [23].

The relationship between sleep and pain is bidirectional [24,25,26,27]. Reduced sleep leads to greater pain, and greater pain has a negative impact on sleep. Poor sleep is also associated with the development of chronic pain [24]. In a large Norwegian cohort, women with three symptoms of insomnia (problems falling asleep, waking early and work disruption) were nearly three times more likely to develop fibromyalgia compared with those with no symptoms [28]. In addition, chronic sleeping difficulties are a predictor of acute post-operative pain in patients undergoing total knee replacement [29]. From the other direction, studies have demonstrated that reduced sleep is causally linked to greater pain [26, 30, 31], increasing both the neurotransmitters related to pain sensitivity and the inflammatory markers associated with pain [32, 33]. Restricted total sleep time and frequent waking, similar to the sleep patterns experienced by those with chronic pain, results in high spontaneous pain reports and reduced pain modulation. Improving sleep for people with chronic pain therefore has the potential to reduce pain levels and improve quality of life. The aim of this study was to use systematic review methods and meta-analysis to evaluate the effectiveness of non-pharmacological sleep interventions in improving sleep in people with chronic pain.

Methods

The protocol was registered prospectively with PROSPERO (CRD 42019093799) [34], and the research question formulated according to the PICO principle [35]. Methods were based on those described in the Cochrane Handbook [36], and reporting was in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [37]. (Supplement Table 1).

Patient and public involvement

All of our studies of sleep problems in people with chronic pain are fully supported by patient involvement. This includes regular discussions during development, conduct and reporting of research.

Eligibility criteria

Eligible studies reflected PICOS criteria:

  • Population: People aged ≥18 years with chronic non-cancer pain

  • Intervention: Non-pharmacological and non-invasive intervention to improve sleep quality or duration

  • Comparison: Comparator of standard care, no treatment, attentional, or wait list control

  • Outcomes: Primary outcomes of sleep quality and duration, secondary outcomes of other sleep outcomes, pain, health-related quality of life, and psychological wellbeing, and a primary harm outcome of adverse events. Follow up post-treatment and at 3 and 6 months after end of treatment if reported

  • Study: Evaluation in a randomised controlled trial

Information sources and searches

We searched MEDLINE, EMBASE, PsycINFO, Cochrane Library, and CINAHL from inception up to 8th April 2020. The search strategy as applied in MEDLINE is included in Supplement Table 2. Citations of key reviews and studies were tracked in Web of Science, reference lists checked and clinical trial records reported in the Cochrane Library followed up. No language restrictions were applied, and relevant non-English articles were translated. Studies reported only as abstracts or that we are unable to acquire using inter-library loans or email contact with authors were excluded.

Study screening and data extraction

Results of searches were imported into Endnote and duplicates removed. After an initial screen by one reviewer to remove clearly off-topic studies, all titles and abstracts were screened independently by two reviewers. Potentially relevant articles were acquired and independently assessed by two reviewers for eligibility with disagreements resolved in discussion with a third reviewer.

One reviewer extracted data from eligible studies into Excel and a second reviewer checked this. Extracted data comprised: country; dates of recruitment; setting; inclusion and exclusion criteria; participant characteristics (chronic pain condition, age, sex); intervention and comparator content, timing, duration and intensity; assessment times; outcome measures; and information on intervention fidelity. We contacted study authors for clarification relating to review eligibility and for missing data.

Risk of bias assessment

Risk of bias was assessed independently by two reviewers using the Cochrane tool [36], specifically relating to: randomisation process; deviations from intended interventions; missing outcome data (> 20% considered high risk), measurement of the outcome; and selection of the reported result. Studies with serious concerns relating to risk of bias were considered high risk and those with limited reporting unclear risk. Studies with wait list controls were considered to be at unclear risk of bias due to inherent lack of blinding. Studies with high or unclear risk of bias were excluded from meta-analysis in sensitivity analysis.

Data synthesis

We conducted meta-analyses with Review Manager 5.4 software to compare outcomes across studies with similar interventions and outcome measures. For continuous data, if outcomes were measured identically across studies, an overall mean difference (MD) with 95% confidence intervals (CIs) was calculated. If continuous outcomes were measured differently across studies, overall standardised mean differences (SMDs) and 95% CIs were calculated and presented alongside measures of heterogeneity (I2). Forest plots were generated. Risk of bias as a potential source of heterogeneity was considered in sensitivity analyses. If pooling of outcome data was not appropriate, a narrative synthesis was reported. In interpreting the outcomes from this review we consider effect sizes as described by Cohen: small, SMD = 0.2; medium, SMD = 0.5; large, SMD = 0.8 [38].

Results

Review progress is summarised in Fig. 1. Searches identified 4314 articles of which 305 were considered potentially eligible. After detailed screening we included 42 randomised trials which included 3346 participants. In 33 trials, 2 randomised groups were compared, in 8 there were 3 groups and in 1 there were 4. Overall, 94 groups were compared.

Fig. 1
figure 1

Study flow diagram

Full size image

Details of studies and methodological quality scores are summarised in Supplement Table 3 and risk of bias assessments in Supplement Table 4. Methodological quality was assessed as good in 32 studies and fair in ten. Overall risk of bias was judged to be low in 11, high in ten and unclear in 21 studies. The mean ages of participants in studies ranged from 31 to 73 years, with an average age across studies of 51 years. Overall, 56 groups received an intervention targeting improvements to sleep. Country of study, chronic pain condition, intervention characteristics, and outcome measures are summarised in Table 1.

Table 1 Study and intervention characteristics
Full size table

The primary areas addressed by interventions were psychological, physical exercise, physical therapy, and other. Within these groups, comparisons were with untreated controls or alternate active interventions. Sleep outcomes were questionnaires focusing on specific aspects of sleep experience, sleep diaries including aspects of time in bed, nocturnal sleep time, sleep latency, sleep efficiency, wake after sleep onset, total sleep time, and number of awakenings, or measurements with sensors such as actigraphy or polysomnography. Pain outcomes were reported in 38 studies, health-related quality of life outcomes in 24 studies and measures of psychological health in 26 studies. Adverse events were infrequently recorded. All studies reported follow up at the end of intervention or within 2 weeks of completion. Fifteen studies reported follow up at 3 months or longer.

For all studies, effect estimates comparing intervention with control or alternative intervention are summarised in Supplement Table 5 with outcomes reported in multiple studies shown in meta-analysis summaries in Table 2.

Table 2 Effect estimates for interventions with multiple studies reporting outcomes
Full size table

Psychological interventions

Cognitive behavioural therapy for insomnia (CBT-I) versus control

In 10 studies including 482 participants, CBT-I was compared with no treatment, attentional control or wait list control [39,40,41,42,43,44,45,46,47,48]. Causes of chronic pain were fibromyalgia (2 studies) [41, 43], osteoarthritis (2 studies) [45, 48], spinal pain (1 study) [42], multiple sclerosis (1 study) [39], migraine (1 study) [46], and diverse chronic pain (3 studies) [40, 44, 47]. Three studies were judged to be at low risk of bias [42, 45, 46], but no studies were considered to be at high risk.

At the end of intervention, overall questionnaire assessed sleep quality in 9 studies with 385 participants was improved in people receiving CBT-I compared with untreated controls, SMD -1.23 (95%CI -1.76, − 0.70), p < 0.00001 (Fig. 2). Heterogeneity was high, (I2 80%). The improvement was sustained but reduced at 6 months in 3 studies with data. In 3 studies at low risk of bias, the benefit for CBT-I over control was slightly reduced, SMD -1.01 (95%CI -1.79, − 0.22), p = 0.01 and heterogeneity remained high (I2 74%). Exploration of effectiveness in relation to a specific condition was only possible for 2 studies at unclear risk of bias including 79 people with fibromyalgia suggesting no benefit for CBT-I, SMD -0.57 (95%CI -1.44, 0.30), p = 0.2, but heterogeneity was high (I2 65%).

Fig. 2
figure 2

CBT-I versus control. Sleep quality

Full size image

Waking after sleep onset, as measured by actigraphy, improved in people receiving CBT-I post-treatment compared with controls and this was also apparent up to at least 6 months in those studies using sleep diaries and polysomnography. Sleep onset latency was improved up to 3 months after CBT-I when assessed by questionnaire, but not with actigraphy or polysomnography. Sleep efficiency was improved up to 6 months after CBT-I compared with control when measured in sleep diaries or by polysomnography, but not by actigraphy. Total sleep time measured by diary, actigraphy or polysomnography was not improved in those receiving CBT-I compared with controls. There were no improvements in diary recorded sleep awakenings or the Epworth Sleepiness Scale in people receiving CBT-I compared with control. Adverse events were assessed in 5 studies [42, 43, 45,46,47]. No adverse events were reported in 4 studies. In 1 study 3 adverse event cases were deemed to be study related [45], this included rash from wearing actigraph and tenderness at site of testing.

In 9 studies with 370 people randomised, pain measured by questionnaire was reduced post-treatment in people receiving CBT-I compared with controls, SMD -0.24 (95%CI -0.45, − 0.03; p = 0.02), but this was not apparent in 3 studies at low risk of bias, SMD -0.19 (95%CI -0.59, 0.220), p = 0.36, or sustained at 3 months or longer (Fig. 3). There was no evidence of heterogeneity. In 2 studies at unclear risk of bias including 79 people specifically with fibromyalgia, there was no benefit for CBT-I, SMD -0.31 (95%CI -0.75, 0.140), p = 0.18. There was no suggestion of heterogeneity.

Fig. 3
figure 3

CBT-I versus control. Pain

Full size image

No studies reported health-related quality of life outcomes. For psychological health, information from studies was mainly limited to post-treatment with benefit suggested for anxiety SMD -0.54 (95% CI -1.01, − 0.06), p = 0.03 with slight heterogeneity (I2 28%), and depression SMD -0.57 (95%CI -1.05, − 0.08), p = 0.08 with high heterogeneity (I2 65%). IN 1 study exclusively including people with fibromyalgia, anxiety and depression were reduced in the Group receiving CBT-I compared with controls.

In 2 studies, a further comparison was made between CBT-I and CBT solely for pain (CBT-P). One was a small pilot study with 11 people with chronic pain randomised to the 2 interventions [44]. In a larger study with unclear risk of bias due to high losses to follow up, there was no suggestion of benefit for any outcome for CBT-I compared with CBT-P in people with fibromyalgia [43].

Brief education with CBT component versus no treatment

In 1 study with 132 people with chronic non-cancer pain randomised, a brief educational intervention incorporating sleep hygiene and cognitive behavioural strategies was compared with wait list controls [49]. The study was at high risk of bias due to large losses to follow up. Only sleep outcomes were reported and there was no difference between groups in sleep quality or sleep diary measures excepting diary recorded sleep latency which favoured the intervention.

Cognitive behavioural therapy for insomnia (CBT-I) versus sleep hygiene

CBT-I was compared with a sleep hygiene intervention in 4 randomised trials [41, 50,51,52]. Studies included 270 participants, all with chronic pain from fibromyalgia. Risk of bias was low in 3 studies [50,51,52] and unclear in 1 due to limited reporting of methods [41].

Data on overall sleep quality was available for 3 studies [41, 50, 51]. There was no difference between randomised groups post-treatment, SMD -0.25 (95%CI -0.82, 0.33), p = 0.40 but heterogeneity was high (I2 61%). Excluding the study at unclear risk of bias removed heterogeneity and there was an improvement in overall sleep quality after CBT-I compared with sleep hygiene, SMD -0.53 (95%CI -0.94, − 0.12), p = 0.01 [50, 51]. Evidence relating to longer term outcomes was limited but with no clear suggestion of benefit for CBT-I over sleep hygiene.

In 3 studies, results for pain outcome were similar in direction to sleep quality [41, 50, 51] but a difference favouring CBT-I post-treatment was only apparent in the 2 studies at low risk of bias, SMD -0.85 (95%CI -1.26, − 0.43), p < 0.0001 [50, 51] with no evidence of heterogeneity. Health-related quality of life was improved in people receiving CBT-I compared with sleep hygiene in 2 studies with 97 people randomised, both at low risk of bias and with no heterogeneity, SMD -0.79 (95%CI -1.20, − 0.37), p = 0.0002. Improvements in pain and health-related quality of life were not evident at longer follow up. Evidence relating to psychological health was limited to 2 studies at low risk of bias with 97 patients randomised. There was no benefit for CBT-I compared with sleep hygiene for anxiety, SMD -0.32 (95%CI -0.72, 0.08), p = 0.12 with no heterogeneity, but depression was reduced, SMD -0.61 (95%CI-1.05, − 0.18), p = 0.006 with slight heterogeneity (I2 11%).

Cognitive behavioural therapy for insomnia and pain (CBT-IP) versus control

In 4 studies with 432 participants randomised, cognitive behavioural therapy focusing on insomnia and pain (CBT-IP) was compared with no treatment, wait list or attentional control [44, 53,54,55]. In 2 studies, the cause of pain was fibromyalgia [53, 54], and in 1 each, osteoarthritis [55], or diverse causes [44]. Risk of bias was high in 1 study due to large losses to follow up at the end of treatment [54], and unclear in 1 due to lack of methodological detail [53]. A third was a small pilot study [44]. For sleep quality, data for meta-analysis was available from all studies post-treatment (Fig. 4). Compared with controls, people receiving CBT-IP had marginally improved sleep quality and improved diary recorded total sleep time, SMD -0.79 (95%CI -1.58, 0.00), p = 0.05, and MD − 61.58 min (95%CI -105.25, − 17.91), p = 0.006, respectively. In the 1 study at low risk of bias, the difference in sleep quality was smaller and in 2 studies with data, the benefit relating to sleep quality was not sustained at 3 and 6 months. In 2 studies with 299 people exclusively with fibromyalgia, there was no evidence for a difference in sleep quality, SMD -0.88 (95%CI -2.16, 0.41), p = 0.18 but heterogeneity was high (I2 94%).

Fig. 4
figure 4

CBT-IP versus control. Sleep quality

Full size image

All 4 studies reported a pain outcome post-treatment (Fig. 5) [44, 53,54,55]. There was no benefit for CBT-IP compared with control except in 2 studies at 3 months [53, 54]. No benefit was seen in the 2 studies of people with fibromyalgia or the study at low risk of bias [55]. Information was limited relating to health-related quality of life and psychological health but in 2 studies there was no difference post-treatment in quality of life or depression between groups.

Fig. 5
figure 5

CBT-IP versus control. Pain

Full size image

In 4 studies with 415 people randomised, CBT-IP was compared with a control condition that included cognitive behavioural therapy for pain (CBT-P) with no specific focus on sleep [43, 44, 54, 55]. In studies with data suitable for meta-analysis there was no difference in sleep quality, pain or health-related quality of life post-treatment, or anxiety or depression post-treatment or at 3 months, and this was not changed if restricted to studies at low or unclear risk of bias or in a study exclusively including people with fibromyalgia. Heterogeneity was high (I2 75%).

In 1 study at unclear risk of bias due to limited reporting of methods with 95 people with fibromyalgia randomised, hypnosis additional to CBT-IP was evaluated [53]. Compared with CBT-IP alone, there were no differences in sleep quality, diary assessed total sleep time, or pain post-treatment and at 3 and 6 months. Health-related quality of life was improved after CBT-IP with hypnosis compared with CBT-IP but only post-treatment and at 3 months. At 6 months, CBT-IP without hypnosis showed a favourable outcome. General psychological health was improved in people receiving CBT-IP and hypnosis compared with CBT-IP post-treatment and at 3 and 6 months.

Acceptance and commitment therapy versus attentional control or exercise

In 1 study, 299 people with chronic pain were randomised to a 7-week course of acceptance and commitment therapy based stress management, or to a control discussion group of similar intensity and duration, or to group-based exercise [56]. The study was at high risk of bias due to high losses to follow up. There were no differences between randomised groups in insomnia severity, pain, anxiety or depression at the end of treatment or at 6 month follow up.

Mindfulness versus control

In 2 studies including 239 people with fibromyalgia pain, a mindfulness-based intervention was compared with waiting list controls [57], or CBT with no specific focus on sleep [58]. Risk of bias was unclear in the former mainly because of wait list controls [57], while in the latter, risk of bias was high mainly due to large losses to follow up [58]. In meta-analysis, there was benefit for improved sleep quality post-treatment after mindfulness intervention compared with wait list controls or CBT, SMD -0.41 (95%CI -0.72, − 0.11), p = 0.008, and this was consistent at 3 [57], and 6 months [58]. For pain, and health-related quality of life, there was no consistent evidence for benefit of mindfulness intervention over CBT or wait list control.

Relaxation versus control

In 1 study with 53 people with fibromyalgia pain, group and individualised applied relaxation was compared with a sleep hygiene-based educational intervention, and a wait list control [59]. Risk of bias was low for the comparison of interventions, but unclear in relation to the wait list control. There were no differences post-treatment between sleep quality, pain, or health-related quality of life in relaxation and wait list control groups. This was also the case for relaxation compared with education, with the exception of improved pain in the relaxation group at 6 month follow up. Risk of bias was high at this follow up time due to high and uneven losses to follow up.

Sleep hygiene versus control

Further to the studies comparing sleep hygiene intervention with CBT-I [41, 50,51,52], in 2 studies with 54 people with fibromyalgia pain, sleep hygiene was compared with untreated controls [41, 59]. Risk of bias in 1 study with a wait list control was unclear [59], and high in the other due to uneven randomisation and large loss to follow up [41]. In meta-analysis, there was no benefit post-treatment for sleep hygiene compared with controls for sleep quality or pain (Table 2). A difference in general psychological health favouring sleep hygiene over control was limited to a single study at high risk of bias.

Physical exercise

Group-based exercise versus control

In 5 studies with 697 people randomised, exercise programmes were compared with usual care or an attentional control [56, 60,61,62,63]. Pain conditions were, arthritis, rheumatoid arthritis, fibromyalgia, low back pain and general chronic pain. In 4, the programme was delivered at a clinic or in a group [56, 60, 62, 63], and in 1 at home [61]. Studies were at low [61], unclear [60, 62, 63], or high risk [56] of bias. One was a feasibility study [62]. For the 3 studies with data [56, 61, 63], questionnaire assessed sleep quality was not improved in the exercise groups compared with controls, SMD -0.10 (95%CI -0.31, 0.12), p = 0.39 and this was consistent in the study at low risk of bias. There was no difference in 1 study with sleep measures at 6 months [56]. One study including 321 people with arthritis comparing a low to moderate intensity physical activity programme with wait list controls and at unclear risk of bias, presented sleep data dichotomised into groups of people with no problems and those with moderate to severe problems [63]. In intention to treat analyses, the authors reported post-treatment benefit for the intervention compared with controls for the outcome waking up tired, p < 0.001, but not trouble falling asleep or staying asleep, waking up at night, or trouble staying asleep. Benefit was not maintained at 3 or 6 months. In one study including 53 people with fibromyalgia, the authors reported that sleep quality was improved in the group receiving group-based exercise compared with controls, p = 0.051 [56, 60].

Pain in 3 studies was reduced in the exercise group compared with controls, SMD -0.52 (95%CI -0.76, − 0.27), p < 0.00001 [56, 60, 61], and this was consistent in the study at low risk of bias [61] and as reported by the authors in the study specifically in people with fibromyalgia, p = 0.039 [56, 60]. In 2 studies reporting anxiety and depression, there was no benefit for exercise compared with controls [56, 60]. The study with dichotomised outcomes showed no benefit for physical activity intervention in relation to health-related quality of life or mental health [63]. Adverse events were assessed in 1 study [62] with 2 events reported, increase in low back pain (n = 1) and increase in knee pain (n = 1).

Home-based walking programme versus exercise or control

In a feasibility study with 60 people with chronic low back pain randomised, a walking intervention was compared with supervised exercise and controls [62]. Risk of bias was unclear due to lack of blinding. While no differences were apparent in sleep quality, pain, general psychological health, anxiety or depression, the authors concluded that their screening methods and new intervention could be evaluated in a fully powered trial.

Comparison of exercise interventions

In 2 studies, different exercise modalities were compared [64, 65].

In a pilot study, 40 people with multiple sclerosis pain were randomised to a clinic-based moderate-intensity aerobic exercise programme or low-intensity home-exercise programme [64]. Moderate-intensity aerobic exercise improved sleep quality and Actigraph sleep measures post-treatment compared with the low-intensity exercise. However, the study was at high risk of bias due to high and uneven loss to follow up.

A course of aquatic biodance was compared with a course of stretching exercises in 1 study with 76 people with fibromyalgia pain [65]. The randomised trial was at unclear risk of bias due to high losses to follow up at end of treatment, but the analysis reported was intention to treat. Immediately after treatment, sleep quality, pain, health-related quality of life, anxiety and depression were all improved in the group who participated in aquatic biodance compared with those who did stretching exercises.

Tai Ji Quan versus education

In 2 studies with 90 people randomised, Tai Ji Quan was compared with an education intervention [66, 67]. Chronic pain conditions included were knee osteoarthritis [66] and fibromyalgia [67]. One study was at low risk of bias [66] and 1 was at unclear risk due to limited reporting [67]. In meta-analysis there was little difference in sleep quality post-treatment, SMD -0.78 (95%CI -2.31, 0.76), p = 0.32 and this was consistent in the study at low risk of bias. No heterogeneity was evident. Also in the study at low risk of bias, sleep efficiency measured by questionnaire and diary was improved in people receiving Tai Ji Quan [66], but there were no differences in other sleep, pain or psychological measures.

Physical therapy

Physical therapy modalities were evaluated in 6 studies.

Hydrotherapy

Two studies described treatment comparisons including hydrotherapy [68, 69]. In 1 study including 81 people with fibromyalgia pain, a course of hydrotherapy with stretching was compared against hydrotherapy with Tai Chi [68]. Risk of bias was high due to large losses to follow up. Results were presented as graphs and interpreted by the authors as showing no differences between randomised groups.

A course of hydrotherapy was compared with conventional physiotherapy in 1 study at low risk of bias with 50 people with fibromyalgia randomised [69]. Post-treatment, total sleep time was marginally higher and total nap time lower in people who received hydrotherapy compared with conventional physiotherapy. Pain and general psychological health did not differ between groups post-treatment. Adverse events were assessed in 1 study [68], 3 patients dropped out of the intervention group due to pain exacerbation (n = 2) and chlorine hypersensitivity (n = 1).

Massage or manual therapy

A course of massage therapy was compared with relaxation therapy in 1 study with 30 people with low back pain [70]. Reporting was limited and risk of bias unclear. A large difference in sleep disturbance between groups was apparent at baseline. Post-treatment, there were no differences between groups. A possible favourable sleep disturbance outcome for massage therapy at follow up may have been masked by the difference at baseline.

In 1 study with 89 people with fibromyalgia pain randomised, a course of manual therapy was compared with no treatment [71]. The study was at unclear risk of bias due to lack of blinding of the intervention. Results reported separately for men and women suggested improved sleep quality, pain, health-related quality of life and depression in people who received manual therapy compared with controls.

Physical therapy programme

In 1 study including 60 people with fibromyalgia, a physical therapy programme with hot pack, ultrasound, transcutaneous electrical nerve stimulation and low power laser was compared with an untreated control group [72]. Risk of bias was high mainly through lack of blinding. The authors reported improvements in sleep quality, pain and health-related quality of life in people receiving the physical therapy programme compared with controls.

Pompage

A course of pompage was compared with controls in 1 study with 23 people with fibromyalgia pain randomised [81]. All participants received stretching and aerobic exercises. Risk of bias was high to large losses to follow up. There was no improvement in sleep quality or pain post-treatment in people receiving pompage compared with controls.

Other interventions

Acupressure

Auricular point acupressure was compared with sham auricular point acupressure in 1 study including 61 people with chronic low back pain [74]. Risk of bias was high due to large and uneven losses to follow up. People receiving active intervention reported improved sleep quality post-treatment compared with the sham group. Differences in other sleep measures were marginal. In another study, self administered relaxing or stimulating acupressure was compared with usual care in 67 people with low back pain [75]. Risk of bias was unclear due to limited reporting of methods. With results shown as graphs, the authors reported no improvement in sleep quality between acupressure groups, and compared with controls. Pain was reduced in acupressure groups but not in controls. Adverse events were assessed 1 study [75] with four mild events related to acupressure treatments involving too much pressure to acupoints.

Bright light treatment

In 1 study, visor delivered bright light treatment was compared with sham therapy in 19 people with fibromyalgia pain randomised in a crossover trial [76]. The study was at high risk of bias due to large losses to follow up. There were no differences between groups in sleep quality, hours slept per night, awakenings per night, pain, anxiety or depression.

Foot reflexology versus control

In 1 study, 68 people with pain from rheumatoid arthritis were randomised to a course of foot reflexology or control [77]. Risk of bias was unclear due to lack of blinding. Actigraphy was only completed for 25% of participants. Sleep quality and pain were improved in people receiving reflexology compared with controls.

Transcranial stimulation versus sham

One study compared transcranial stimulation with sham in 16 patients with musculoskeletal pain [78]. This was a small feasibility study with unclear risk of bias due to limited reporting. People receiving transcranial stimulation had reduced pain after the intervention compared with controls, but there was no difference in sleep quality. The authors described aspects of study conduct to advise future evaluations of transcranial stimulation.

Mattress interventions versus control

In 1 study at low risk of bias with 30 people with fibromyalgia pain randomised, a period of sleep on a magnetic mattress pad was compared with sham [79]. Sleep, pain and ADL focused quality of life were improved in the intervention group compared with sham control. Adverse events were assessed with none reported. A period of sleep on a supportive mattress overlay was compared with untreated controls in 1 study with 38 people with low back pain randomised [80]. The study was at high risk of bias due to concerns about the randomisation procedure and blinding. Post-treatment, sleep quality was similar between groups. Presented as graphs, the authors reported improved pain in the mattress overlay group compared with controls but no difference in sleep quality.

Discussion

We identified 42 randomised trials evaluating sleep interventions for people with chronic pain. CBT interventions provided the largest evidence base with CBT-I interventions demonstrating benefit post-treatment compared with controls for improved sleep quality, however evidence for a longer-term sustained benefit was lacking. Evidence in people with specific conditions was limited due to primary study limitations and statistical heterogeneity. Numerous interventions were evaluated in small numbers of studies, but evidence was insufficient to draw conclusions on effectiveness.

Findings from meta-analysis found that CBT-I (9 studies) and CBT-IP (4 studies) demonstrated a medium to large (− 0.79) effect compared with control for sleep quality at post-treatment. Differences between groups at 6 months was slightly reduced with a medium effect size for CBT-I and was not sustained for CBT-IP. CBT-I showed a small improvement in pain outcome post-treatment but this was not sustained. However, there was high heterogeneity in studies which should be considered when interpreting results. In 4 studies comparing CBT-IP with CBT-P only, no differences in pain outcome were found. Due to the active control it is therefore only possible to infer that adding insomnia specific content to CBT-P does not have an additional impact on pain outcomes. Our findings regarding the impact of CBT-I interventions on sleep quality reflect the existing literature. A recent systematic review of CBT-I therapies in patients with chronic non-malignant pain showed significant treatment effects immediately post-treatment for global measures of sleep [73]. Condition specific reviews show similar results with CBT therapies improving sleep outcomes in the short-term for patients with lower back pain, fibromyalgia, and osteoarthritis [82,83,84].

Our results demonstrate that improving sleep for people with chronic pain is possible, and that CBT approaches have the strongest evidence base. Poor sleep has a negative impact on optimism, sociability, and psychosocial functioning [85]. Poor sleep also has clear links with depression and pain catastrophising, both of which can affect pain management and coping. Pain catastrophising is linked with maladaptive coping techniques and depression is linked to lack of engagement in treatment [86, 87]. CBT approaches are already widely used in pain management with a focus on coping strategies and behavioural rehearsal [88]. This systematic review demonstrates that additional focus on sleep improvement could be of benefit. As CBT is an established treatment approach, future work should focus on how best to implement CBT sleep interventions for people with chronic pain and foster more equitable access to support, particularly for underserved populations. Results from the National Pain audit highlighted that service provision for the management of chronic pain in the UK is inadequate [89]. Only 40% of pain clinics in the UK are multidisciplinary which presents a challenge for implementation of psychological interventions such as CBT.

Evaluation of the longer-term effectiveness of sleep interventions for people with chronic pain is lacking. Although evidence suggests that CBT interventions improve sleep immediately post treatment, effects reduce over time. In addition, most studies had follow-up data collected at 3 months or less post-intervention, with 9 studies reporting 6-month outcomes, and 2 studies reporting 12-month outcomes. Understanding longer-term effectiveness of these interventions is crucial for people with chronic pain. Due to the nature of the condition, individuals with chronic pain may experience disturbed sleep for many months and years, therefore effective interventions need to have sustained effects.

Assessment of outcomes within the trials included in this review varied considerably, and this was particularly notable for the secondary outcomes. This heterogeneity limits comparison between studies, particularly for health-related quality of life and psychological wellbeing, because outcome measures assess different aspects of these constructs, for example general mood assessment versus specific anxiety or depression measures. The issue of heterogeneity across randomised trials is well established and initiatives such as COMET have been addressing this through the development of core outcome sets [90, 91]. Core outcome sets provide a minimum set out of outcomes to be used in all trials of a certain focus, ensuring comparability across multiple studies. The challenge of a review of this scope is that the interventions included are varied and include both generic and condition specific measures. As the evidence base builds, focused reviews on areas of promise, such as CBT and third wave therapies, could offer benefit. The feasibility of developing a sleep core outcome set could also be explored as this would be provide opportunity for greater comparison across interventions.

Nine of the 42 studies in this review provided data on adverse events with no serious adverse events reported [42, 43, 45,46,47, 62, 68, 75, 79]. Assessment of adverse events is vital for patient safety, however unlike in trials of pharmacological treatments where monitoring and reporting of harm outcomes is mandatory, behavioural and psychological interventions are not held to the same account [92]. In 2004 the CONSORT group provided ten recommendations for reporting harm outcomes in trials [93]. All except 5 of the studies included in our review were published after the recommendations. This demonstrates a need for evaluations of psychological and behavioural interventions to improve reporting of harm outcomes.

Strengths and limitations

This review has strengths and limitations that should be considered when interpreting the findings. The method used in this review was robust and systematic, following Cochrane guidance [36]. The review provides a comprehensive overview of the existing literature on sleep interventions for patients with chronic pain, but as included studies addressed a wide variety of interventions with small numbers of studies for each intervention, this limited opportunity for meta-analysis. The population of people with chronic pain that we considered is heterogenous with a range of underlying medical conditions. However, chronicity reflects pain that persists for 3 months or longer, and chronic pain is usually the affected person’s main clinical problem. Secondary outcome measure tools used were highly heterogenous, which limited comparison between studies.

Conclusion

CBT approaches have the potential to be an effective treatment to improve sleep for people with chronic pain, but further high-quality primary research is required to explore refinements that will ensure parallel improvements to pain, quality of life, psychological health and maintain all benefits in the long term. Individuals who experience depression and pain catastrophising may particularly benefit from sleep interventions. As CBT is an established treatment approach, future work should focus on how best to deliver these interventions, for instance by exploring any difference between online or face-to-face delivery or differences between delivery professionals. Importantly, future research could focus on how best to facilitate equal access and outcomes in underserved populations.

Primary research is also needed to evaluate the effectiveness of interventions including mindfulness, aquatic exercise and hydrotherapy, Tai Ji Quan, manual therapy, physical therapy programmes, acupressure, foot reflexology and magnetic mattress pads. Individuals who experience chronic pain could benefit from interventions that address sleep, and research is needed to assess any impact of sleep interventions on pain.