The prevalence of pain in bipolar disorder: a systematic review and large-scale meta-analysis

Authors


Abstract

Objective

To conduct a meta-analysis investigating the prevalence of pain in people with bipolar disorder (BD).

Method

A systematic review and random effects meta-analysis searching major electronic databases from inception till 01/2014 in accordance with the PRISMA statement. We included articles reporting quantitative data on the prevalence of pain in people with BD with or without a healthy control group. Two independent authors conducted searches, extracted data, and completed methodological quality assessment.

Results

Twenty two cross-sectional studies were included, representing 12 375 644 individuals (BD n = 171 352, n controls = 12 204 292). The prevalence of pain in people with BD was 28.9% (95% CI = 16.4–43.4%, BD n = 171 352). The relative risk (RR) of pain in BD compared to controls was 2.14 (95% CI = 1.67–2.75%, n = 12 342 577). The prevalence of migraine was 14.2% (95% CI = 10.6–18.3%, BD n = 127 905), and the RR was 3.30 (95% CI = 2.27–4.80%, n = 6 732 220).About 23.7% (95% CI = 13.1–36.3%, n = 106 214) of people with BD experienced chronic pain. Age, percentage of males, methodological quality, and method of BD classification did not explain the observed heterogeneity.

Conclusion

People with BD experience significantly increased levels of pain (particularly chronic pain and migraine). The assessment and treatment of pain should form an integral part of the management of BD.

Summations

  • The pooled prevalence of clinical pain in people with bipolar disorder is approximately 28.9%, while 23.7% and 14.2% are affected by chronic pain and migraines respectively.
  • Compared with the general population, people with bipolar disorder are at significantly increased risk of reported clinically relevant pain [Relative Risk (RR) = 2.14] and migraine (RR = 3.30).
  • Because pain has a range of deleterious impacts on an individual's health and quality of life and may worsen psychiatric symptoms, we recommend that pain assessment and treatment should form part of the routine care of people with bipolar disorder.

Considerations

  • There was inconsistency in the assessment methods used to measure pain across the studies.
  • There was considerable heterogeneity in each of the pooled analysis that could not be explained by mean age, percentage of males, method of diagnosing bipolar disorder, and methodological quality of the included studies.
  • There was insufficient information to determine the influence of the phase of illness and polarity as well as acuity of bipolar symptoms on the observed results.

Introduction

Pain has a deleterious impact on an individual's health and wellbeing [1], and common painful conditions, such as chronic musculoskeletal disorders, contribute to a significant number of years lived with disability across the globe [2]. Chronic pain in particular is associated with greatly reduced quality of life and difficulties with activities of daily living and often has a negative impact on an individual's emotional and mental health [3]. A substantial body of the literature suggests that those with chronic pain have higher rates of depressive and anxiety symptoms than those without chronic pain [4-6].

Despite this, the prevalence of chronic pain in persons with severe mental illness (SMI) has received little attention [7, 8]. This is surprising as persons with SMI such as schizophrenia and bipolar disorder (BD) have a highly increased risk for a plethora of painful physical illnesses including cardiopulmonary diseases, metabolic diseases, bone disorders, viral infections, and cancer [9-12]. In addition, pain in people with SMI is also associated with a worsening of psychiatric symptoms [7]. Despite this increased risk of severe comorbid physical illnesses, most persons with SMI do not receive adequate physical healthcare provision and treatment [13-15]. Mental health specialists report barriers limiting their ability to treat physical comorbidity and people with SMI are less likely to recognize or monitor co-occurring medical conditions than the general population [16, 17]. Additionally, many healthcare professionals fail to take people with SMI seriously when they report physical health problems [18]. When compared with those without SMI, persons with SMI appear to have an increased likelihood of experiencing conditions that cause pain while at the same time having a lower likelihood of receiving adequate care to manage it [9, 10].

A recent systematic review established that people with schizophrenia, who have been known to have a higher pain threshold for pain than the general population, have a lower prevalence of pain than people with other psychiatric disorders, particularly compared to those with BD [19]. However, to date, no systematic review or meta-analysis of pain in individuals with BD exists, despite the fact this group appears to be particularly more likely to experience chronic pain and less likely to seek medical help [8]. In fact, people with BD reported almost four pain complaints at any one time [20]. Moreover, people with BD who are treatment adherent report statistically lower levels of pain than their non-treatment adherent counterparts [21]. Clearly, a better understanding of the risk and burden of pain is an important step toward improving clinical outcomes for individuals with BD.

Aims of the study

In recognition of the potential for pain to be problematic for people with bipolar disorder (BD), the paper had the following two aims: (i) to establish the prevalence of pain and its moderators in people with BD and (ii) to compare the prevalence of pain in BD with general population controls.

Material and methods

This systematic review was conducted according to the PRISMA statement [22] following a predetermined, but unpublished protocol.

Inclusion and exclusion criteria

Studies were eligible that fulfilled the following criteria: (i) inclusion of participants with BD, diagnosed according to diagnostic criteria [e.g. DSM-IV [23] or ICD 10 [24]], a valid screening measure (e.g. Alcohol Use Disorder and Associated Disabilities Interview Schedule—DSM-IV Version) or through medical record review. When we encountered studies containing groups of mixed participants (e.g. with major depressive disorder), we contacted the authors up to two times over a month period to ascertain the variables of interest in BD subjects. If these data were not available, we excluded the study. (ii) Reporting of the prevalence of pain (of any type) or assessment of pain with a continuous measure with or without comparison to a control group that did not have a mental illness. When a study measured pain with a continuous measure, but did not specify prevalence rates with a cut-off point, we contacted the authors up to two times to obtain this information.

We did not place a language restriction upon our searches. If we came across studies that reported data from the same sample at different time points, we used the most recent data and/or the largest data set. We excluded studies that (i) reported pain as an adverse event of a drug trial (e.g. for headache), (ii) reported the prevalence of BD in a sample of patients who all had pain (no other comorbidities were excluded), or (iii) in which the pain was experimentally induced. When we encountered studies without a control group that assessed pain in a sample with a continuous measure [e.g. SF 36 bodily pain scale [25]], but did not have a cut-off to determine the prevalence of pain, we excluded the study if the authors did not respond to requests for additional data.

Information sources

Two reviewers (BS, DV) independently conducted searches on Academic Search Premier, MEDLINE, EMBASE, Psychology and Behavioral Sciences Collection, PsycINFO, SPORTDiscus, CINAHL Plus, and Pubmed. In addition, the reference lists of all eligible articles and recent systematic reviews of the literature were scanned to assess eligibility of additional studies.

Searches

Two independent reviewers (BS, DV) employed the predetermined search strategy using the key words ‘bipolar disorder’ and ‘pain’ or ‘pain perception’ or ‘pain management’ or ‘pain measurement’ or ‘musculoskeletal pain’ or ‘pain intensity’ or ‘chronic pain’ or ‘neuropathic pain’ or ‘pain*’.

Study selection

After the removal of duplicates, two independent reviewers (BS, DV) screened the titles and abstracts of all potentially eligible articles. Both authors applied the eligibility criteria, and a list of full text articles was developed through consensus. Two reviewers (BS, DV) then considered the full texts of these articles, and the final list of included articles was reached through consensus.

Data extraction

Two authors (BS, DV) independently conducted data extraction using a predetermined form. The data collected from each article included the following: study design, geographical location, bipolar sample and control sample characteristics (number, % male, mean age), bipolar diagnosis method, method of pain assessment (including site, severity, and interference of pain where available), and the prevalence of pain in people with BD and controls as defined by the authors.

Methodological quality assessment

Two independent authors (BS, DV) completed methodological quality assessment of included articles using the Newcastle Ottawa Scale (NOS) [26]. Due to the anticipated paucity of data, we also included studies without a control group. These studies were considered as case–control studies for the purposes of methodological assessment in accordance with a previous review [27]. The NOS is utilized to assess the methodological quality of non-randomized trials and has acceptable validity and reliability [26]. The assessment tool focuses on three main methodological features: (i) the selection of the groups, (ii) the comparability of the groups, and (iii) the ascertainment of the outcome of interest. The NOS can be modified, and we adapted the NOS to take into account age and gender as comparability measures and considered pain assessment in the exposure category. Studies are given a score from 0 to 9, with a score of 5 or greater being indicative of satisfactory methodological quality. We anticipated studies without a control group would score below this and present their results with due consideration.

Meta-analysis

We pooled individual study data using DerSimonian–Laird proportion method [28]. Our predetermined protocol stipulated that heterogeneity would be assessed with the Cochran Q statistic [29]. As we found significant heterogeneity [Cochran Q = 66988.29 (df = 24), P < 0.0001], a random effects meta-analysis was employed using StatsDirect. We calculated the RR to investigate the differences in pain between those with BD and members of the general population when there were three or more studies (Aim ii). When possible, we conducted subgroup analyses to investigate the prevalence of migraine and chronic pain because the literature has suggested that these are prevalent in people with BD [8]. In order to investigate sources of heterogeneity, we conducted moderator analysis with mean age, percentage of males, NOS score, and the method of BD classification (comparing DSIM, ICD, or any other screening measure). We assessed publication bias with a visual inspection of funnel plots, yet gave priority to quantitative testing through the Begg–Mazumdar Kendall's tau [30] and Egger bias tests [31].

Results

Study selection

The original search yielded 2713 potential hits which were reduced to 2319 after the removal of duplicates. At the eligibility screening stage, a total of 72 articles were deemed potentially eligible and full texts were obtained and reviewed by two authors. In total, 50 articles were excluded with reasons and 22 articles met the eligibility criteria and were included in the review [8, 32-52]. The full search strategy including reasons for exclusion is presented in Fig. 1.

Figure 1.

PRISMA 2009 flow diagram for search strategy.

Study characteristics

In total, 171 352 people with BD and 12 204 292 general population controls (total sample size = 12 375 644) were included in the 22 meta-analyzed studies. Details of the included studies are presented in Table 1. All of the studies adopted a cross-sectional measurement of pain, and seven of these (n with BD = 138 285) [8, 33, 38, 41, 44, 49, 52] had a control group without a mental illness. The sample size of persons with BD across the studies ranged from 10 [40] to 96 186 [8], and the control populations ranged from 32 333 [44] to 4 247 684 [8]. The mean age of participants with BD ranged from 39 [34] to over 65 years [51].

Table 1. Included study characteristics and methodological quality
Study NoLocation and designBipolar diagnosisParticipant with BD characteristicsControl participant characteristicsNOS score
  1. BD = bipolar disease; GP = general practitioner; ICD-9-CM = International Classification of Diseases, 9th Revision, Clinical Modification; DSM-IV-TR = Diagnostic and Statistical Manual of Mental Disorders Fourth Edition Text Revision; CIDI 3.0 = World Mental Health Composite International Diagnostic Interview version 3.0; AUDADIS-IV = NIAAA Alcohol Use Disorder and Associated Disabilities Interview Schedule-DSM-IV Version; NOS = Newcastle Ottawa Scale; BDRS = Bipolar Depression Rating Scale; YMRS = Young Mania Rating Scale.

8

USA

Cross-sectional design collecting data over 1 year

ICD-9-CM

Derived from patient electronic medical records

N = 96 186

Age <35 to >80 years

Males 81 757 (85.0%)

No data on BD severity or medication

N = 4 247 684

Age <35 to >80 years

Males 3 882 806 (91.4%)

7
32

Brazil

Cross-sectional

DSM IV

N = 339 split in two groups

Migraine (n = 115)

41.6 ± 11.20 years

Males 16 (17.4%)

None migraine (n = 224)

41.5 ± 12.32 years

Male 60 (26.7%)

No control group3
33

USA

Cross-sectional study collecting data over 5 year period

ICD-9

N = 3557

39.3 ± 11.8 years

Males 1395 (39.2%)

BD more likely have substance use disorder OR 2.92; (95% CI, 2.59–3.29%) & alcohol use disorder AOR 19.63; (95% CI, 17.59–21.90%)

N = 726 262

37.7 ± 12.8 years

Male 345 146 (47.5%)

7
34

USA

Cross-sectional

Composite International Diagnostic Interview Version 3.0

N = 740

39 (±10.6) years

Males 414 (56%)

No data on BD severity or medication

No control group3
35

Australia

Cross-sectional

DSM IV

N = 67

Males 35.8% (n = 24)

40.4 (±13.5) years

BDRS = 11.5 ± 9.3

No control group3
36

Spain

Cross-sectional

DSM-IV-TR

N = 121

50.7 years (±12.3)

Males 45 (37.8%)

50.7% had suicidal ideation

No control group3
37

Italy

Cross-sectional

DSM-IV-TR

N = 248

Demographic information not available

No control group3
38

USA

Cross-sectional

AUDADIS-IV

N = 883

36.9 ± 0.3 years

Males 380 (43%)

N = 42 210

45.4 ± 0.1 years

Males 20 261 (48%)

7
39

UK

Cross-sectional retrospective

Not stated

N = 169

Demographic information not available

No control group3
40

Taiwan

Cross-sectional

DSM-IV-TR

N = 10

Demographic data not available

No control group3
41

USA

Cross-sectional retrospective analysis of data over 1 year

ICD-9 medical records

N = 4310

Males 3879 (90%)

53 ± 13 years

BD more likely have SUD (P < 0.0001)

N = 3 408 760

Males 3 067 884 (90%)

58 years

8
42

USA

Cross-sectional

DSM IV

N = 111

44.8 ± 13.2 years

Males 35 (32.4%)

No control group3
43

Italy

Cross-sectional

DSM III

N = 30

Demographic data not available

No control group3
44

Canada

Cross-sectional

CIDI

N = 938

Age 25–64 years

Males 436 (46.4%)

N = 32 333

Demographic information not available

7
45

USA

Cross-sectional (baseline from RCT)

ICD-9 criteria

N = 384

42.07 ± 11.3 years

Males 128 (33.3%)

No control group3
46

Australia

Cross-sectional

ICD 9

N = 27

10 males (37.0%)

No control group3
47

Canada

Cross-sectional

DSM IV

N = 296 with BD 1 and BD 2

49.8 ± 12.7 years

% Males not available

No control group3
48

South Korea

Cross-sectional

DSM-IV

N = 190

Demographic data not available

No control group3
49

Scotland

Cross-sectional retrospective analysis

GP databases

N = 2582

54.5 years

Males 1021 (39.5%)

N = 1 421 796

47.9 years (P < 0.001)

Males 698 408 (49.1%)

8
50

Singapore

Cross-sectional

CIDI 3.0

N = 93

Age 18–65> years

Males 47 (50.5%)

66% BD I had severe or moderate manic/hypomanic & 100% respondents with BP-II reported mild clinical severity on the YMRS

Not reported3
51

United States

Cross-sectional

ICD 9

N = 24206

All >65 years nursing home residents

No specific data on demographics

No control group3
52

United States

Cross-sectional retrospective analysis

ICD 9

N = 27 054

Demographics not available

N = 2 325 247

Demographics not available

8

Methodological quality

The NOS summary score for each article is presented in Table 1. All seven studies that had a control group scored high (mean NOS score 7.2 ± 0.48) and were considered good quality. The 15 studies that did not have a control group all scored lower than 5 on the NOS, which was attributable to the absence of a control group; these studies scored zero (out of a possible 5 points) in the areas that compare the bipolar and control groups on selection, comparability, and exposure.

Measurement and location of pain in the bipolar populations

A range of different types of pain were considered. The most commonly investigated pain was headache/migraine [8, 32, 33, 35, 39, 40, 42-44, 47, 52] while six studies investigated chronic pain [8, 34, 45, 48-50]. A wide range of methods were employed to ascertain pain in people with BD and are presented in Table 2.

Table 2. Results of pain in included studies
StudyType of painMethod of pain assessment/ascertainmentPain results BDPain results in controlOther results
  1. BD = bipolar disorder; VAS = visual analogue scale; ADL = activities of daily living; ICHD-2 = International Classification of Headache Disorders, 2nd edition; LBP = low back pain; SF 36 = short form 36; nr = number; SUD = substance use disorder; BDRS = Bipolar Depression Rating Scale; OCD = obsessive compulsive disorder.*Statistically significant.

8

Arthritis Back pain

Chronic Pain

Migraine Other headache

Psychogenic

Neuropathic

ICD-9-CM based on electronic patient records

Total n = 96 186 in BD sample

Any pain 61.3% (n = 58 983)

Arthritis 45.3% (n = 43 595)

Back pain 33.5% (n = 32 264)

Chronic pain 3.4% (n = 3316)

Migraine 4.9% (n = 4677)

Other Headache 6.7% (n = 6419)

Psychogenic pain 0.9% (n = 833)

Neuropathic pain 5.4% (n = 5180)

Total n = 4 247 684 in control sample

Any pain 42.3% (n = 1 795 600)

Arthritis 32.2% (n = 1 365 901)

Back pain 17.0% (n = 721 372)

Chronic pain 0.7% (n = 27 758)

Migraine 1.1% (n = 46 015)

Other headache 2.0% (n = 86 126)

Psychogenic pain 0.1% (n = 3646)

Neuropathic pain 3.7% (n = 156 393)

OR comparing BD and controls:

Any pain OR 2.17 (CI 2.14–2.19%)*

Arthritis OR 1.75 (CI 1.73–1.77%)*

Back pain OR 2.47 (CI 2.43–2.50%)*

Chronic pain OR 5.43 (CI 5.23–5.63%)*

Migraine OR 4.67 (CI 4.53–4.82%)*

Other headache OR 3.46 (CI 3.37–3.55%)*

Psychogenic pain OR 10.17 (CI 9.43–10.96%)*

Neuropathic pain OR 1.49 (CI 1.45–1.53%)*

32MigrainePhysician diagnosis33.9% (n = 115) had migrainesNo Control Group

Migraine group higher nr of psychiatric comorbidity (72.6%) vs. non migraine group (47.4%), P < 0.001

Migraine group more likely to have anxiety disorder (P < 0.001) and depressive polarity

33LBP Headache

Elixhauser

Comorbidity

Index

25.8% (n = 919/3557) had LBP

19.3% (n = 685/3557) headaches

13.3% (n = 96 201/726 262) had LBP

5.7% (n = 41 234/726 262) headaches

BD more likely to have LBP (P < 0.0001) and headaches (P < 0.0001)
34Chronic pain interfering with ADLSingle item question46% (n = 338/641) had chronic pain interfering with ADLNo control group 
35(i) migraines and (ii) body achesSelf-report 4 point Likert scale questions rating pain during depressive episode

68.5% sample (46/67) had headaches; 29.9% slight, 20.9% moderate and 17.9% major problem

62.6% sample had body aches; 22.4% slight, 19.4% moderate, and 20.9 major problems

No control groupCurrent BDRS score predicted headaches (P = 0.012)* BDRS severity not related to body aches P = 0.3
36General pain

VAS interviewer administered to assess pain over last 6 weeks (>40 on VAS)

Noted severity, duration, and interference with ADL

51.2% (n = 62/121) had pain

Duration of pain 62.5 (±90.9) months Severity 67.5 (±14.9)

Interference with ADL 67.7 (±21.2)

Location of pain: Head 66.1%; Neck 66.1%; Back 74.2%; Limbs 67.7%; Joints 64.5%

Nr. of pain locations: 3.44 (±1.46) 75% of pain musculoskeletal pathology

No control group

Older age associated with pain (OR 1.03 (CI 1.00–1.07%)

Sex, education, marital status, diagnostic group, depressed mood, sleep disorders, and depression not related to pain

37Painful somatic symptomsMedical records22.6% (n = 56/248)No control group 
38General pain interfering with activitiesSingle item question about pain interfering with ADL over past 4 weeks (0 = not at all to 5 = extremely)24.8% (n = 219/883) had moderate or worse pain interfering with ADL11.9% (n = 5023/42 210) had moderate or worse pain interfering with ADL P < 0.001Comorbid anxiety (OR 1.72, 95% CI 1.41–2.10%), being married (OR 1.33, 95% CI 1.08–1.64%), and SUD (OR 1.91, 95% CI 1.56–2.34%) associated with interfering pain. Age, lower income associated with pain (P < 0.001)*
39MigraineUnclear, searched patient records4.7% (n = 8/169)No control group 
40Migraine/headacheICHD-270% (n = 7/10) had migraineNo control group 
41LBPLBP classified as present yes/no from national patient electronic records database15.4% (n = 663/4310) had LBP10.6% (n = 361 868/3 408 760), P < 0.0001BD more likely to have LBP (P < 0.0001)*
42MigraineQuestion on lifetime prevalence of migraine39.8% (n = 43/108) had lifetime prevalence of migraineNo control groupNumber of psychiatric admission higher in BD without migraine (P = 0.046)*, no difference in suicide attempts or SUD
43Migraine/headachePhysician diagnosed and classified

20% (n = 6/30) migraine

33.3% (n = 10/30) muscle tension headache

No control group 
44MigraineSurvey on previous diagnosis by physician14.9% for males and 34.7% for females had migraines5.8% for men and 14.7% females had migrainesBD males with migraine more likely to report earlier BD onset (P < 0.05)*, and anxiety (P < 0.05)*
45Arthritis/chronic painSingle item self-report question48.9% (n = 188/384) had chronic painNo control groupChronic pain associated with worse physical HRQOL (P < 0.001)* but better mental HRQOL scores (P = 0.01)*
46Joint painCheck list18.5% (n = 5/27) had joint painNo control group 
47MigraineID migraine questionnaire according to International Headache Society23.9% (n = 71/296) had migraineNo control groupMigraine associated with BD diagnostic subtype (P < 0.001)*, History suicidal behaviour (P = 0.03)*, social phobia 7 panic disorder (P < 0.001)*, OCD, and anxiety (P < 0.001)*
48Medically unexplained pain (MUS)Asked if has pain lasting >6 months in past year that was severe/interfered with ADL and could not be explained0.3% had severe chronic pain interfering with ADL (n = 190/6328) OR 5.93 (1.71–20.60)No control group data 
49Chronic painGP database ≥4 analgesic prescriptions in last year OR ≥4 antiepileptics in the absence of an epilepsy17.5% (n = 451/2582) had chronic pain8.8% (n = 125 680/1 421 796) had chronic painOR 1.88 P < 0.001 Chronic pain in BD vs. control*
50Chronic painModified CIDI checklist for medical disorders40.4% (n = 38/93) had chronic painNot reportedBD associated with chronic pain OR 3.0 (CI 1.5–5.8%), P < 0.001*
51General painClassified from medical records present = yes/no18.1% had pain (n = 4381/24 206)No control group 
52MigraineDefined from medical records2.0% migraine (n = 530/27 054)0.7% (n = 16 383/2 325 247) 

Prevalence of pain in persons with bipolar disease

In total, 25 types of pain were investigated, and the pooled prevalence of pain was 28.9% [95% CI: 16.4–43.4%, n = 171 352, Cochran Q = 66988.29 (df = 24), P < 0.0001, Fig. 2a]. The funnel plot was asymmetrical (Fig. 2b); however, both the Begg–Mazumdar (Kendall's τ = −0.013; P = 0.908) and Egger bias (Kendall's τ = 11.51; P = 0.4897) tests did not demonstrate any evidence of publication bias. Next, we pooled the prevalence of pain using only one pain measurement from each of the 22 studies, thus including only the highest prevalence of pain from three studies that contained data on pain at two sites [33, 35, 43]. The prevalence of clinical pain across 22 studies was 28.4% [95% CI = 15.0–44.1%, Cochran Q = 66477.17 (df = 21), P < 0.0001]. Within this analysis, there was also no evidence of publication bias (Egger: bias = 12.44, P = 0.5176, Begg–Mazumdar: Kendall's τ = 0.021, P = 0.9113).

Figure 2.

Random effects pooled prevalence of pain in bipolar samples (n = 22, n = 171 352). Pooled proportion = 28.9% (95% CI = 16.4–43.4%). Cochran Q = 66988.29 (df = 24), P < 0.0001. (b) Funnel plot. Begg–Mazumdar: Kendall's τ = −0.013, P = 0.90; Egger: bias = 11.510, P = 0.48.

Moderators of the prevalence of pain in people with BD

Ten studies [32-36, 38, 41, 42, 45, 49] had sufficient data on mean age, percentage of males, and bipolar diagnosis method to enable moderator analyses. The moderator analyses demonstrated that mean age (b1 = −0.038, z = −0.311, P = 0.75), % male (b2 = −0.074, z = −1.013, P = 0.311), and method of diagnosing BD (b3 = −0.0935, z = −0.092, P = 0.92) did not explain the heterogeneity in the prevalence of pain. We investigated the effect of methodological quality (NOS score) on the prevalence of pain across the 22 studies, and this suggested that a low NOS score was associated with a high prevalence of pain but this did not reach statistical significance (b1 = 0.532, z = 1.875, P = 0.06). Lastly, we investigated the influence of the method of BD diagnosis on the prevalence across all studies, and this demonstrated that the classification used to diagnose BD had no significant effect on the prevalence of pain (b1 = 0.310, z = 0.524, P = 0.59).

Comparing the prevalence of pain in people with BD versus control groups

In each of the seven studies with a control group, persons with bipolar disease consistently reported a higher prevalence of pain than the comparison group. One study [33] provided pain data for two different types of pain and was corrected for multiple comparisons in the pooled analysis. In total, data from 12 342 577 unique individuals (n with BD = 138 285 and control n = 12 204 292) indicated that the RR of pain in people with BD was 2.14 [95% CI = 1.67–2.75%, χ2 = 36.623 (df = 1), P < 0.0001; Cochran Q = 1078.49 (df = 7), P < 0.0001]. The results from the meta-analysis are presented in Fig. 3. The funnel plot of the seven included studies was not symmetrical indicating possible publication bias. However, the Eggers test (10.931, P = 0.013), but not the Begg–Mazumdar: test (Kendall's τ = 0.14; P = 0.7195), showed evidence of publication bias.

Figure 3.

Relative risk (RR) of pain in people with bipolar disorder compared t controls (n = 7, n = 12 342 577). Pooled RR = 2.14 (95% CI = 1.676–2.75%), χ2 = 36.623 (df = 1), P < 0.0001; Cochran Q = 1078.49 (df = 7), P < 0.0001.

Pooled prevalence of migraine in people with BD

We also calculated the pooled prevalence of migraine in 127 905 individuals across nine studies [8, 32, 39, 40, 42-44, 47, 52], and this yielded a prevalence of 14.2% [95% CI = 10.6–18.3%; Cochran Q = 1080.29 (df = 8), P < 0.0001].

Comparing the prevalence of migraine in people with BD versus control groups

It was possible to pool the data from three comparative studies [8, 44, 52] involving 6 732 220 unique individuals (n with BD = 126 956, n controls = 6 605 264). The RR was 3.30 (95% CI = 2.27–4.80%, χ2 = 39.408 (df = 1), P < 0.0001).

Pooled prevalence of chronic pain in people with BD

It was possible to calculate the pooled prevalence of chronic pain in 106 214 individuals with BD across six studies [8, 34, 45, 48-50]. The pooled prevalence of chronic pain was 23.7% [95% CI = 13.1–36.3%, Cochran Q = 2200.77 (df = 5), P < 0.0001]. Only two comparative studies [8, 49] contained data on chronic pain, and it was therefore not possible to meta-analyze these data.

Discussion

To our knowledge, this is the first systematic review and meta-analysis investigating the prevalence of pain in people with BD. In this large review involving 171 352 persons with BD and 12 204 292 controls, we found that a substantial proportion of patients with BD reported clinically relevant levels of pain. The overall pooled analysis of pain in people with BD was 28.9%, and the RR was over double for people with BD compared to members of the general population. In terms of specific types of pain, the pooled prevalence of chronic pain was high with almost one in four (23.7%) being affected. In addition, migraine affected one in seven (14.2%) persons with BD, and the comparative analysis demonstrated that people with BD are over three times more likely to experience migraines than members of the general population.

Increased levels of pain in persons with BD may be explained by several mechanisms. For instance, BD and migraine appear to share some specific polymorphisms, with the KIAA0564 gene being particularly implicated, thus suggesting a close association [53, 54]. Also, people with BD have an increased prevalence of depression [8, 36], and depression has been associated with increased physical complaints, and possibly, greater pain sensitivity [55], opposite to findings in schizophrenia [19]. For example, neuroimaging studies in major depressive disorder indicate that heightened amygdala activity, in part, explains the high comorbidity of pain and depression when these conditions become chronic [56]. However, due to limitations in the available data, we could not investigate the influence of depressive symptoms on the observed results. Other studies have suggested serotonergic and noradrenergic pathway involvement [7, 57]. In addition, specific neuroinflammatory mechanisms responsible for an elevated risk of painful physical comorbidity in people with BD may contribute to the higher levels of observed pain [58]. Previous research [59] has found that migraine and BD symptoms are closely related, and the presence of migraine can influence pain perception. As we found that 14.2% of people with BD experienced migraine, this could have influenced the variance in the prevalence of pain. Lastly, recent findings [60] also suggest that limited cognitive flexibility and memory capacities may be linked to the mechanisms of pain chronicity and probably also to its neuropathic quality. This may imply that people with BD who are known to have deficits in executive functioning or memory have a greater risk of pain chronicity after a painful event. This seems particularly pertinent given the fact that we found across 106 214 individuals with BD that almost one in four is affected by chronic pain.

Clinical implications

The results of this review are concerning because pain and in particular chronic pain in people with BD is associated with impaired recovery [45], greater functional incapacitation [44, 61], lower quality of life [8], and increased risk of suicide compared to people without pain [62]. As BD is already associated with a greatly increased risk of suicide [63], it is imperative that this population receives adequate pain assessment and management [36]. A central component to this is the training and education of psychiatrists who are in a critical place to oversee the pharmacological management of pain [7]. We advocate that systematic assessment of pain should be undertaken as part of the management of BDs and that pain should be monitored during the course of treatment. Equally, healthcare professionals dealing with pain should consider mental health complications. Previous work suggests that clinicians are more likely to attend to pain than mental distress [64]. The potential benefits of early identification and treatment of pain may not only include a reduction in pain and of its impact on the individual, but may also extend to a reduction of healthcare costs and improvement of mental health outcomes.

Of great concern are the high levels of chronic pain experienced by people with BD. A better understanding of the association of BD and chronic pain could help limit harmful/adverse pharmacological side-effects. For instance, in the general population, chronic pain is often managed with tri-cyclical antidepressants [65], yet prescription of such medication to a person with BD may inadvertently trigger a manic phase of illness if prescribed in the absence of a mood stabilizer [66]. Commonly used analgesic medications also need careful consideration. For instance, there is sound evidence that non-steroidal anti-inflammatory medications can increase serum lithium levels, impairing renal lithium excretion and possibly eliciting lithium toxicity [67]. Similarly, some stronger analgesic medications such as opioids may have mood altering qualities increasing the risk of eliciting a manic episode [68].

Limitations of the review

Several limitations, especially of the included literature need to be considered when interpreting the results of our review. First, BD is a complex and heterogeneous disorder, and reporting of pain likely varies according to different phases, polarity and acuity of the disease. The paucity of information regarding these illness characteristics made it impossible to systematically evaluate their effects on pain prevalence in patients with BD. In addition, the perception and therefore prevalence of pain is known to vary according to the type of BD (I or II) [59], but due to limitations in the data, we were not able to disentangle this relationship. In addition, gender may also cause some variance, but our moderator analysis did not elucidate any evidence of a gender effect. Second, all of the included studies utilized a cross-sectional measurement of pain and did not correlate pain with mood state or severity of symptoms. Therefore, prospective longitudinal studies that assess pain prevalence and severity over time and in relationship to mood symptoms and treatments are essential. Third, our results may have been suspect to Berkson's bias, which states that clinical samples are more impaired and experience more pain than non-clinical samples due to self-referral to a clinical setting. Berkson's bias has been observed in the mood dimensions of BD [69] and may account for an underreporting within the pooling of epidemiological data. Fourth, none of the included studies used a validated pain assessment scale, and subsequent information about the severity, location, variability, and interference of pain during activities is lacking. Fifth, all of the meta-analytic results were heterogeneous. and some demonstrated a degree of publication bias. In our moderator analysis, we were not able to explain the heterogeneity with mean age, % males, or the methodological quality of method of classification of BD. This finding demonstrates that unknown/unmeasured factors contribute to the observed heterogeneity. Regarding publication bias, the funnel plot for the main analysis (Fig. 2b) appeared asymmetrical, yet the quantitative investigation of bias did not demonstrate any evidence to support this. This discrepancy may be due to the fact that there is a trend for publication bias, but its magnitude is insufficient to reach statistical significance according to the Eggers test or Begg–Mazumdar test. In addition, the comparative analysis (Fig. 3) demonstrated some publication bias with the Eggers test, but this finding should be interpreted with caution due to the low number of studies (<10) [70]. Sixth, there was insufficient information about psychotropic and analgesic medication within the BD cohorts to enable statistical investigation of these variables on the observed results. Future research should seek to investigate the influence of psychotropic and analgesic medications on pain, and particular attention should be paid to the prevalence of pain in people with BD who are drug naïve. In the same way, future research should investigate the role of psychiatric comorbidities including anxiety and substance use disorders on the prevalence of clinical pain in these patients. Finally, we included 15 studies that received low methodological quality ratings. However, the low methodological quality ratings were due to the absence of a control group, and the moderator analysis demonstrated that these studies had no significant effect on the observed results. Despite the aforementioned, higher levels of pain were reported consistently among people with BD than in the comparison groups.

Future research

It is essential that future research seeks to clearly assess pain characteristics including noting the site, severity, variability, and chronicity. There were insufficient data to analyze these pain characteristics in our meta-analysis. In addition, only one study [8] measured psychogenic pain, and it would be important to investigate if this differs from physiological pain in people with BD. An important question, also unaddressed, is what is the impact of comorbid pain, particularly chronic pain, on daily activities? It is likely that chronic pain amplifies the effect of BD on disability and reduced quality of life. Future prospective studies should be conducted in order to truly capture the prevalence of pain and disentangle its impact and contributing factors. Such research should establish how pain impacts on a person's mental health and wellbeing, with longitudinal studies being most important. Future research should also explore the extent to which those with BD are more or less responsive to behavioural, pharmacological, and non-pharmacological treatments for pain. For example, studies have not yet examined the impact on pain of antiepileptic medications such as lamotrigine, valproate, and topiramate among persons with BD. In addition, in the general population the promotion of physical activity is a key factor preventing the onset of chronic pain but is also encouraged to treat it [71], and many people with chronic pain are inactive [72]. However, research [73] has established that most people with BD are sedentary. Therefore, strategies to encourage people with BD to become active that do not exacerbate their pain are likely to be key in the prevention and management of pain and physical therapists can lead this process [74]. In addition, the barriers and facilitators to pain management should be explored in people with BD with emphasis on the perspective of the patient and the treating multidisciplinary team. Lastly, within our review, there were limited studies assessing pain in patients with BD and those with other psychiatric conditions, making it impossible to directly compare the prevalence of clinical pain in people with BD and other psychiatric diagnosis. More research is required to directly compare clinical pain across different psychiatric disorders.

To conclude, almost 30% of persons with BD experience clinically relevant pain, which was twice as common compared to general population controls. Chronic pain was prevalent affecting almost one in four people, and migraine was over three times as common than in the general population. Pain has a range of adverse and deleterious impact upon the individual and may impede recovery, reduce quality of life and have adverse effects on psychiatric symptoms. Therefore, it is essential that treating psychiatrists and the wider multidisciplinary team seek to provide adequate assessment and treatment of pain in people with BDs.

Declaration of interest

BS, AJM, AS, SR, and LE have no conflicts of interest to declare. DV is supported by the Research Foundation – Flanders (FWO – Vlaanderen). MDH declares that he has been a consultant for, received grant and/or research support and honoraria from, and been on the speakers' bureaus and/or advisory boards of the following companies: Astra Zeneca, Bristol-Myers Squibb, Eli Lilly, Janssen-Cilag, Lundbeck, Pfizer, Sanofi Aventis, and Takeda. CUC has been a consultant and/or advisor to or has received honoraria from: Actelion, Alexza; Bristol-Myers Squibb, Cephalon, Eli Lilly, Genentech, Gerson Lehrman Group, IntraCellular Therapies, Janssen/J&J, Lundbeck, Medavante, Medscape, Merck, Otsuka, Pfizer, ProPhase, Roche, Sunovion, Takeda, Teva, and Vanda. He has received grant support from BMS, Janssen/J&J, Novo Nordisk A/S, and Otsuka.

Ancillary