Prevalence and incidence of musculoskeletal extremity complaints in children and adolescents. A systematic review

Different traditions and interest in different levels of complaint-severity are reasons for various terms used in research about musculoskeletal extremity complaints: injury, disorder, discomfort, complaint, pain, ache etc. In this review we analysed musculoskeletal extremity complaints in the general population, and therefore found a broad term of musculoskeletal extremity complaint (MEC) to be most comprehensive and this will be used throughout the rest of this article, regardless of the term used in the referenced articles. When possible, the MEC were divided according to causation, and were categorised as either traumatic or non-traumatic. A traumatic complaint was defined as an injury resulting from a specific identifiable event, whereas a non-traumatic episode was not related to an identifiable event [10]. For example, a traumatic complaint could be pain due to fall from a horse, and a non-traumatic complaint could be pain of unspecific origin developed over a longer period of time.

The search was made in collaboration with a research librarian. Two electronic databases, MEDLINE and EMBASE, were searched for articles published before September 2015. The following search terms were used both as MeSH terms and free text in MEDLINE, and as Subject heading and abstract term in EMBASE: “prevalence”, “incidence”, “musculoskeletal disorder”, “musculoskeletal injury”, “musculoskeletal pain”, “extremity”, “limb”, “children”, “adolescents”, “paediatric”. As free text/abstract term “toddlers” and “teenager” were searched as well. Different forms of spelling and synonyms were used. The full search strategy is seen in Additional file 1. In addition, the reference lists of the relevant obtained articles were screened for additional relevant articles.

The article had to report the prevalence or incidence of musculoskeletal disorders, complaints, injuries, pain or other description of complaints in the upper and/or lower extremities in general or clinical populations of children and adolescents. All levels of prevalence and incidence rates were included, and could be both parental reported or self-reported values. All study designs were included, but they had to be published in English, Swedish, Norwegian or Danish.

Special settings or groups, e.g. children with other diseases such as diabetes or other chronic diseases, or children from a specific sport setting (e.g. football players) were excluded, because their pattern of injuries might not be comparable to rest of the population.

The first and third author (SF and LH) reviewed the titles and abstracts and identified relevant articles to be read in full text. Inclusion of articles based on full text was decided by agreement between the first and third author (SF and LH).

We were not aware of quality assessment tools specifically designed for studies of prevalence and/or incidence, since most quality assessment tools are designed for studies of associations or comparative effectiveness in either observational studies [20‐23] or randomised clinical trials [24]. Therefore the quality of the included articles was assessed by a modified version of the Quality In Prognosis Studies (QUIPS) tool (Additional file 2), which was originally developed to assess bias in studies of prognostic factors [25]. This tool identifies six domains to consider when evaluating risk of bias in studies of prognostic factors: ‘Study Participation’, ‘Study Attrition’, ‘Prognostic Factor Measurement’, ‘Outcome Measurement’, ‘Study Confounding’ and ‘Statistically Analysis and Reporting’ [25]. Three of these domains: ‘Study Participation’, ‘Outcome Measurement’ and ‘Statistical Analysis and Reporting’ were considered to be important, relevant and adequate for studies investigating prevalence, whereas the attrition domain was also included for studies investigating incidence, due to the main interest in relation to representativeness of the population of interest and the validity of the outcome measurement. In the ‘Statistical Analysis and Reporting’ domain, the items “strategy for model building is appropriate and is based on a conceptual framework or model” and “the selected statistical model is adequate for the design of the study” were not relevant and therefore ignored in the assessment. ‘Study Attrition’ is irrelevant for cross-sectional studies of prevalence and was therefore not included in the assessment of these studies, and ‘Prognostic Factors Measurement’ and ‘Study Confounding’ were not relevant either, since our results did not include analyses of prognosis or associations. The domains are described in detail by Hayden et al. [25].

Each of the three (four) domains was categorized as “yes”, “partly” or “no” according to whether the level of quality was adequately fullfilled or not. In evaluation of risk of bias in studies of prevalence and incidence, the most important parameter to consider is whether the study sample is representative of the source population, thus leaving ‘Study Participation’ and ‘Study Attrition‘ the most important domains in this review. This is supported by ‘Study Participation’ being the most frequently used parameter across different tools of evaluation of quality and susceptibility to bias in observational studies [26]. Therefore, ‘Study Participation’ and ‘Study Attrition‘ (if included) needed to be considered as satisfactory (“yes”) for the study to be classified in the low risk of bias category. Furthermore, the other two domains had to be at least partly fulfilled. If ‘Study Participation’ and ‘Study Attrition’ (if included) was not considered satisfactory (“no”) the study was categorized as high risk of bias, regardless the judgements in the other domains. If ‘Study Participation’ and ‘Study Attrition’ (if included) was only partly fulfilled and both the other domains were not fulfilled, the study was classified as high risk as well. All other combinations were considered to be medium risk of bias.

The studies were divided into general population studies and clinical population studies. Data were extracted to a descriptive table, a modified version of the STROBE statement [27]. The following relevant items from the STROBE Statement were included in the table: study design, setting, age, study size including response rates, data sources/types of measurement, area of complaint and main results. In addition, type of prevalence or incidence, mode of data acquisition and the level of bias were added to the table.

Data extraction and assessment of quality were individually performed by two authors (SF and KBD) and results compared. In case of disagreement, the third author (LH) was consulted and consensus reached. The review was conducted according to the PRISMA guidelines, and the PRISMA checklist can be seen in Additional file 3.

Obviously, clinical populations cannot be used to assess prevalence of disease. However, they can be used to identify the type of complaints, for which care is sought. Therefore both general population studies and clinical population studies were included in the review, but data were reported separately.

Some of the general population studies had a longitudinal design. Due to the focus on prevalence and incidence rates, these studies were considered as series of cross-sectional studies and each time point reported separately. If the study included any type of intervention, only baseline estimates were reported.

It is known that the prevalence of MEC change with age [28]. Therefore, to give the most relevant description, results were reported by age groups. Cut points between age groups were based on reporting in the included articles.

Where possible, the difference between rates of non-traumatic and traumatic complaints was reported.

Some of the studies reported combined prevalence or incidence rates of the complete upper or lower extremity regions. To identify a possible pattern in distribution of MEC, ratios between the prevalence or incidence rates of the upper and lower extremities were calculated.

The quality assessment categorized fourteen articles to have low risk of bias [14, 15, 35‐46], two to have medium risk of bias [47, 48] and six as having high risk of bias [49‐54] (Table 1). The clinical population studies were all in the low risk category. There was no disagreement in the independent quality assessments by the two authors (SF and KBD) regarding most of the articles, with the exception of two instances [47, 50] within the ‘study participation’ domain which were solved through discussion without need of mediation by the third author.

The search resulted in 19 general population studies and three clinical population studies (Table 2). All the included studies covered children and adolescents of both sexes. Most of the studies were conducted in the northern part of Europe, but also other parts of the world were represented with three studies from North America [35, 43, 55], one from Australia [45] and one from India [49]. Of the 19 general population studies, seven were prospective, four with one follow up evaluation [37, 43, 50, 51], and three studies reporting incidence over time [14, 39, 54]. In most of the studies, data were collected via questionnaires, either self-reported or filled in by parents, but telephone interviews, mobile phone text messages and diagnoses from general practice were also used. The 22 studies used 14 different outcome measurements such as point prevalence, twelve month prevalence and incidence per 1000 exposure hours. Estimates are simply presented in the tables as they appear in the original articles because prevalence and incidence rates were reported in very different ways. For the same reason, meta-analysis was not possible, and even a reliable range of an estimate of frequency, i.e. incidence or prevalence, could not be presented either. However, the findings relating to differences in frequency between different age groups, between different anatomical sites and between different populations, are independent of the absolute prevalence or incidence rates and are therefore presented in detail. We attempted to report findings of younger children and older children separately, but due to different cut points in age between the studies, it was not possible to do this. Throughout most of the analyses, a division between younger children aged 0–12, and older children aged 10–19 was used, but to accommodate all studies, some results were reported for the following age groups: 0–9, 9–12, 10–19, 6–17 and 2–19. A summary of findings is presented in Tables 3, 4, 5 and 6.

In both the younger (aged 0–12) and the older children (aged 10–19), lower extremity complaints were more common than upper extremity complaints, and ankle/ft and knee were the most frequent sites of MEC. Among the younger children, two studies reported ankle/ft complaints to be about twice as frequent as knee complaints [39, 54], whereas the last study reported almost similar prevalence rates for the two sites [15] (Table 3). Among the older children, five of the six included studies reported 0.2 to 2.8 times more knee complaints than ankle/ft complaints [40, 42, 43, 50, 52] (Table 4). In the upper extremities, wrist/hand/fingers was the most common site of complaint in younger children [39, 54] (Table 3), whereas shoulder complaints were more common among older children [42, 50, 52] (Table 4). The least frequent anatomical site of complaint reported in both younger and older children was the elbow (Tables 3 and 4).

Also in the clinical population studies, lower extremity complaints were more frequent than complaints from the upper extremities [44‐46]. Two of the three clinical population studies were based on data from the same cohort of children, but one reported on upper extremities [44] and the other on lower [46]. In the younger children the incidence rate of ankle/ft complaints was about three times higher than for the knee, whereas the incidence rates were almost equal for the two pain sites in the older children [46]. In the upper extremity, wrist/hand/fingers was the most common site in both age groups [44]. The least frequent site of complaint was the shoulder in the younger group, and shoulder and hip/groin in the older age group (Table 5).

In the younger children, three of the general population studies classified the complaints of the lower extremities into traumatic or non-traumatic mode of onset [14, 15, 39], but two of these were based on the same cohort of children [14, 39]. All three reported about two times more non-traumatic complaints compared to traumatic complaints (1:0.5 [15], 1:0.3 and 1:0.4 [14], respectively). One of the studies also reported mode of onset in upper extremities and found the opposite relationship, with non-traumatic complaints less frequently reported than traumatic complaints (1:1.5 and 1:2.5, for prevalence and incidence respectively) [14] (Table 6). There were no reports of mode of onset in other studies, neither in the older age group nor in the clinical population studies.

Comparing lower and upper extremities in general, MEC were reported up to ten times more often from the lower than from the upper extremities (mean ratio 1:4.0; range 1:1.1 to 1:10.3) in the general population studies (Tables 3 and 4), whereas the difference was much smaller in the clinical population studies with ratios from 1:1.2 to 1:1.6 (mean 1:1.4) (Table 5). In the general population studies, the difference was more significant in the younger children (mean ratio 1:5.1; range 1:2.3 to 1:10.3) than in the older children (mean ratio 1:2.5; range 1:1.1 to 1:4.6).

No pattern of differences in reporting could be detected between studies of high and low risk of bias.