Distribution and prevalence of musculoskeletal pain co-occurring with persistent low back pain: a systematic review

We considered observational studies (i.e., longitudinal and cross-sectional cohort studies), from clinical settings or based on cohorts of the general or working populations involving adults 18 years or older. We included studies investigating persistent LBP (e.g., pain within the anatomical region below the twelfth thoracic vertebra and the inferior gluteal fold), with or without radiation to the legs, with a duration of at least 4 wks. Furthermore, eligible studies had to assess co-occurring musculoskeletal pain (i.e., number of co-occurring pain sites and distribution across body sites) in individuals with persistent LBP. Peer-reviewed studies published in the English, Dutch, Danish, Norwegian, Portuguese, Spanish, or Swedish languages, understood by the authors of this paper, were screened for eligibility. Studies including individuals with LBP of specific pathological origin (e.g., fracture, tumour, inflammatory diseases, systemic diseases, infection, structural deformity) were excluded, as were studies including pregnant women and studies dealing with post-surgical persistent LBP. Studies with other study designs (e.g., randomised controlled trials) as well as studies with a study sample of < 100 individuals with persistent LBP were post-protocol decided to be excluded due to their limited possibility to assess prevalence.

The literature search was performed with no restrictions on date, publication type, or language within the following bibliographic databases: MEDLINE and Embase (via Ovid), CINAHL, and Scopus (for forward citation tracking), from the earliest records published to August 2nd, 2019. We updated the search on October 26th, 2020 in Medline (via Ovid) only, as there were no unique hits in the other databases and since majority of relevant studies in systematic reviews are reported to be found within a limited number of databases without introducing bias or changing results [25]. Search terms covered the following domains: LBP, co-occurring musculoskeletal pain, and number of pain sites/pain patterns, combined with study design. Pilot searches were performed on the search terminology to ensure its all-inclusiveness. The design and execution of the searches were supervised by a research librarian (see Additional file 2 for the search strategy). The reference lists of included articles and related reviews within the topic were scrutinised, and forward citation tracking was performed on key articles in order to identify any further studies. PROSPERO was inspected for ongoing or recently completed systematic reviews to identify additional articles not identified in the bibliographic databases. We did not search additional grey literature as originally planned as we do not expect to find any relevant epidemiological studies here as opposed to literature on interventions where very few relevant studies or questionable vested interests may have an impact on the results [26]. The identified articles were downloaded to and managed in EndNote X9 [27].

Relevant records were selected through a two-stage screening process by three independent reviewers (CKØ, MSJ and TFC), where one reviewer (CKØ) screened all and the other two (MSJ and TFC) shared the screening of the retrieved records. In the first stage, titles and abstracts were screened with the reviewers blinded to each other’s selections. Disagreements were discussed and resolved by a fourth independent reviewer (BN) if necessary. The studies considered not to be relevant, or that clearly did not meet the inclusion criteria, were excluded, and full-text articles of the remaining studies were obtained. Studies relevant for the topic, but with uncertain relevance for the current review were taken to the second stage for further consideration. In the second stage, the same reviewers made the final selection based on screening of the full text articles against the eligibility criteria. If necessary, study authors were contacted for additional information to resolve questions about eligibility. Consensus meetings were used to resolve any disagreement by consulting a fourth reviewer (JH). Reasons for exclusion were recorded.

Data from the included articles were extracted into pre-tested forms by two independent author pairs (CKØ + MLF/JH and TFC + MLF/JH), each pair including an experienced reviewer. Disagreements were resolved first by discussion or if necessary by a third independent group of reviewers (MSJ, BN and PJM). Data extraction included: main characteristics of the included articles, definition of LBP, prevalence of co-occurring musculoskeletal pain by anatomical location, number of co-occurring pain sites and/or pattern of distribution of co-occurring musculoskeletal pain, association between the pain pattern and/or number of pain sites and LBP-related disability, and other information relevant for the critical appraisal. We contacted two study authors by e-mail as additional information was required regarding missing data to calculate prevalence (see Table 3 for details).

We used a modified version of the Risk of Bias Tool for Prevalence Studies to assess the risk of bias [47] (see Additional file 3 for details). For the purpose of this systematic review, the tool was slightly modified; in item 1, the wording was changed from “Was the study’s target population a close representation of the national population in relation to relevant variables?” to “Was the study population representative of the target population?” whereby randomly selected or consecutively selected samples were appraised as low risk of bias and convenience samples as high risk of bias. Furthermore, item 6 was defined for LBP only and we left the example in item 7 open with regards to which study instrument that was used apart from that it must have been validated. This modified risk of bias tool was piloted to ensure that reviewers were consistent in their appraisal. Two reviewer pairs independently performed the risk of bias assessment (CKØ + JH and TFC + MLF). The overall risk of bias (i.e., low, moderate, or high) was determined for each included article based on the reviewers’ consensus, given the responses to the items in this tool. With low overall risk of bias further research is very unlikely to change the confidence in the estimate, moderate overall risk of bias indicate that further research is likely to have an important impact on our confidence in the estimate and may change the estimate, and high overall risk of bias imply that further research is very likely to have an important impact on our confidence in the estimate and is likely to change the estimate. All authors were involved in the final assessment of risk of bias. The GRADE approach was not used in this review for overall appraisal of the quality of the evidence due to lack of guidance for systematic reviews on prevalence data using this methodology [48].

The results of the literature search, risk of bias assessment and data extraction are summarized in tables and figures. The proportion of participants with persistent LBP and co-occurring musculoskeletal pain are described as prevalences and subgrouped according to study population, distribution and location(s) of co-occurring pain. The precision of the prevalences was assessed with 95% confidence intervals (CI) calculated using the exact method in Stata (StataCorp, College Station, Texas, USA). Differences in age- and sex-specific prevalences and distribution pattern of co-occurring musculoskeletal pain are described but not pooled as few studies stratified on these factors. Furthermore, we were not able to assess LBP-related disability or possible differences between working versus general populations as intended, due to the limited reporting in the included articles.

The study selection processes are illustrated in Fig. 1. A total of 7481 articles were identified through the literature searches and through other sources. After removal of duplicates, 5744 articles were screened at title/abstract level and a total of 658 articles were screened at carried forward for full-text screening. Finally, 19 articles were considered eligible for this systematic review [28‐46].

The main characteristics of the included articles are presented in Table 1. Ten articles were based on data from the general population, two articles reported on working populations and seven articles reported on clinical populations. Study designs were cross-sectional (n=11), prospective (n=5), and retrospective (n=3); for prospective and retrospective studies, baseline data were considered. The articles were published between 1998 and 2019, while the data were collected between 1985 and 2017 (not reported in two articles). All the included articles were published in English, apart from one in Norwegian [37]. The articles originated from the Scandinavian countries (n=7), other European countries (n=5), USA (n=4) and Japan (n=3). A total of 34,492 individuals with persistent LBP and co-occurring musculoskeletal pain were included in our evidence synthesis. Participation rate varied from 19.4 to 100% in different studies while number of people analysed with persistent LBP varied from 100 to 7523. Age was reported both as mean, frequencies within age categories and quartiles, preventing us from reporting an overall mean age; however, most studies included middle-aged and older people (40 to ≥75 years of age). The proportion of females ranged from 48 to 68%, apart from the study with a working population that only included females. Ten of the 19 articles collected prevalence data with questionnaires (five with own questionnaire), and nine used a combination of questionnaires, interviews, and physical examination.

The definition of LBP was typically based on location combined with duration, and in some instances with pain intensity. All the included studies relied on information from questionnaires and interviews that in six studies included pain drawings [30, 31, 35, 38, 43, 44]. Five studies used current pain duration (ranging from ≥4 to 8 weeks) as an inclusion criteria [35, 38‐40, 42, 43], while 11 studies used accumulated pain duration within the past year (ranging from ≥3 to 6 months) or variations of accumulated ‘daily pain’ during the past year as inclusion criteria [28‐34, 36, 37, 41, 45]. One study used ≥90 days of sickness absence due to LBP during a 2-year period as inclusion criteria [44] and another followed IASP Classification of Chronic Pain [46].

The risk of bias assessment is shown in Table 2. In the overall rating, 14 articles were judged to have a low risk of bias [29‐35, 38‐43, 46], three articles a moderate risk of bias [28, 36, 37], and two articles a high risk of bias [44, 45]. Overall, the internal validity was judged to be slightly better than the external validity, also, when the articles with an overall high risk of bias were excluded. Notably, seven of the 10 articles with general population samples were rated to have a high risk of non-response bias (item 4). The two articles on working populations had a low risk of bias on all items. For the studies on clinical populations with an overall low risk of bias (two high risk of bias studies disregarded), the risk of bias was mainly related to external validity and non-response bias.

Since the included studies were not considered sufficiently homogenous, we chose not to conduct a statistical meta-analysis and the results were therefore synthesised narratively. Table 3 provides a summary of the non-weighted prevalence of co-occurring musculoskeletal pain. Figure 2 gives an overall summary of the results in a forest plot and shows the results for the articles with low or moderate risk of bias. Studies with an overall high risk of bias were not included in the evidence synthesis. Based on the results from the individual articles, co-occurring pain could be grouped into three main categories: i) co-occurring axial pain, ii) co-occurring extremity pain, and iii) other co-occurring multi-site pain (from ≥1 pain site to pain in several body sites).

For co-occurring axial pain, the prevalence of neck and neck/shoulder pain ranged from 32 to 57% in the general population [28, 36], 58% in the working population [38], and from 18 to 54% in clinical populations [40, 41, 46].

For co-occurring extremity pain in the hip, knee, and foot the prevalence in the general population ranged from 20 to 48% [29, 30, 32‐34], while in the working population the prevalence of co-occurring knee pain was 27% [38]. In clinical populations, the prevalence co-occurring buttock, leg, or foot pain was 50% [46], the prevalence of pelvic and pelvic-groin pain ranged from 6 to 28% [40, 46], and the prevalence of co-occurring shoulder-arm-hand pain was 17% in the only study that considered upper extremity pain [46].

In the general population the prevalence of any co-occurring pain in arms, legs, or joints were 43% [29] and other musculoskeletal pain 79% [37], while the prevalence of co-occurring pain in arms, legs, or joints was reported to be 89% and the prevalence of widespread pain (defined as pain in most of your body) was 30% in a clinical population [40]. Another clinical population study reported the prevalence of co-occurring chest-abdomen-groin pain (‘anterior trunk pain’) to be 20% [43]; yet another clinical population study reported the prevalence of co-occurring thorax pain to be 10%, and pain in other body sites 65% [46]. Headache was not part of our search strategy, but we have included the prevalences of headache in studies that otherwise was eligible, and it was reported to be 22 to 32% in the general population [28, 29] and 29% in a clinical population [40]. Additionally, headache was included in two studies that counted number of pain sites [35, 42].

Our results clearly indicate that additional pain sites are common among people with persistent LBP, but reported pain sites are dependent on which sites were asked for in the individual studies. Given a higher number of possible options of co-occurring pain sites, persistent LBP only was reported to be 10–35% across populations (i.e., 10.4% [31]; 18.8% [42]; 21.3% [37]; 22.6% [35]; 29.4% [39]; 35.3% [46]). However, when fewer pain sites were considered the prevalence of persistent LBP only were markedly higher (i.e., 73.6% [30], and 74.9% [32], when the only option was co-occurring knee pain; 45.9% [41] with only co-occurring neck pain as option).

The prevalence of co-occurring neck pain was more than twice as high among females than among males in a general population study that reported on sex differences [28], while for co-occurring neck pain in a clinical population study, about 60% were females [46]. About 30% of those around 75 years reported co-occurring hip pain. Among those reporting co-occurring knee pain, the prevalence tended to increase with age (from 21% at age 45 up to 48% at age 74). Furthermore, the sex distribution among those reporting fewer additional pain sites was similar, while a higher proportion of those reporting a higher number of pain sites were females [31, 35, 39].

Four articles reported on co-occurring pain in relation to LBP-related disability. In the population-based study by Nordstoga et al. [31], persons who reported 3–8 additional pain sites had 16–27% lower probability of recovery from LBP over a 10–11 years period compared to persons with 1–3 additional pain sites. Similarly, in a clinical population of elderly, the Roland Morris Disability Questionnaire (RMDQ) score increased by 0.65 points (95% CI 0.43 to 0.86) for every additional pain site [40]. Additionally, co-occurring pelvic/groin pain, pain in arms, legs, or joints, and widespread pain were all associated with increased long-term LBP disability [40]. Furthermore, patients with persistent LBP who presented with co-occurring anterior trunk pain had significantly higher disability levels measured by RMDQ than those with localised LBP (adjusted group difference at baseline 2.41 [0.34 to 4.49], at 3 months 3.78 [1.37 to 6.18], at 12 months 2.89 [0.67 to 5.11]), but the presence of this co-occurring pain did not affect the rate of recovery of LBP [43]. Socioeconomic status was sparsely considered in the included articles. Though, MacLellan and co-workers [42] found patients with obesity and two or more additional pain sites (in a group where 67.5% had persistent LBP) to have a higher unemployment rate, being retired or unable to work because of disability, having two or more children, or being female compared with patients without any pain.

Strengths of this review include a published protocol with registration in PROSPERO, and adherence to PRISMA recommendations and other guidelines for systematic reviews of prevalence studies [69]. Our search was comprehensive and supervised by an experienced research librarian. Based on information given in the abstracts, it was difficult to assess eligibility. We therefore carried a total of 658 articles forward for full-text screening. This reflects the lack of studies considering persistent LBP in the context of co-occurring musculoskeletal pain and of suitable Medical Subject Headings terms and keywords to cover it. For critical appraisal we used ‘Risk of Bias Tool for Prevalence Studies’ which has been reported to have a high inter-rater agreement [47]. This tool uses overall summary risk of bias based on the rater’s judgment, which conforms with Grades of Recommendation, Assessment, Development and Evaluation (GRADE) and Cochrane approaches, rather than using cut-points from summary scores from numerical rating scales that has been discouraged [70]. We retrospectively restricted the study sample size to improve precision as also recommended by Munn et al. [71]. It is arguably inappropriate to conduct meta-analysis for prevalence studies due to the large heterogeneity and differences within characteristics of study populations, and a qualitative description of these variations across study populations has been encouraged instead [72, 73]. This heterogeneity negatively impact the overall certainty of the evidence.

The generalizability of our results is limited given relatively few studies within each co-occurring musculoskeletal area and population type, particularly working populations. Importantly, all articles originated from high-income countries. We also excluded studies with people reporting LBP of < 4 weeks duration. However, this was a pragmatic decision to, on the one hand, avoid inclusion of very few studies and capture studies including people with fluctuating persistent LBP [74], and on the other hand, avoid inclusion of shorter acute episodes of LBP. Definition of LBP varied among reviewed articles, with commonly data on localisation, duration, and sometime intensity, but none fulfilled the optimal definition of LBP for prevalence studies [75], which indicate that this definition is not well known and that awareness of this could be encouraged among researchers providing prevalence data.

For co-occurring musculoskeletal pain, number of pain sites in addition to LBP ranged from 1 to 9 with limited information on the nature of this co-occurring pain. This may have an impact on the observed patterns because the primary studies did not attempt to map underlying patterns of co-occurring pain sites. The problem with inconsistency in the number of pain sites asked for is also seen in other similar studies with variations from six to 26 pain sites [76‐79] and a need for a set number of pain sites has been proposed [51].

Persistent LBP is very often accompanied by pain in other body regions and thus part of other pain or disease clusters. The available literature on this is, however, quite varied and therefore there is a need for more uniform means of measuring co-occurring musculoskeletal pain to allow for more robust data on the actual existing patterns and to provide better context and interpretability. Understanding how pain occur in patterns and clusters can help inform prognosis as well as clinical research, because certain pain patterns among those with persistent LBP may result in more consequences and require different prevention and treatment strategies [23]. Clinically, patients with LBP and a more complex clinical picture are known to get inadequate care [80‐82]. This suggests a need for better guidance in clinical care and a move from single-site guidelines [83] to recommendations that cover multi-care pathways for people with persistent LBP including their concomitant conditions [84]. However, this requires prioritising and investment by health authorities, not only to acquire the knowledge base needed to inform healthcare personnel, but also to empower patients to better self-manage if the goal is to decrease disability and costs [85].